LARGE-SCALE SPATIOTEMPORAL CORTICAL DYNAMICS IN VISUAL SHORT- TERM MEMORY: FROM SPIKING ACTIVITY TO OSCILLATIONS by Steven Joseph Hoffman A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Neuroscience MONTANA STATE UNIVERSITY Bozeman, Montana November 2020 ©COPYRIGHT by Steven Joseph Hoffman 2020 All Rights Reserved ii TABLE OF CONTENTS 1. GENERAL INTRODUCTION ....................................................................................... 1 Introduction ..................................................................................................................... 1 A Brief History of Oscillatory Electrical Brain Activity ................................................. 5 Proposed Roles of Field Potentials ................................................................................ 12 Are Oscillations Vital, or Simply and Epiphenomena .......................................... 12 Delta Oscillations (0.5-4 Hz) ................................................................................. 14 Theta Oscillations (4-8 Hz) ................................................................................... 15 Alpha Oscillations (8-12 Hz) ................................................................................. 17 Beta Oscillations (14-30 Hz) ................................................................................. 20 Gamma Oscillations (30-90 Hz) ............................................................................ 21 Methodological and Experimental Considerations In Large-Scale Neurophysiology ................................................................................... 23 Conclusions ................................................................................................................... 25 Specific Aims ................................................................................................................ 26 Specific Aim 1: Large-Scale Neural Recording .................................................... 27 Specific Aim 2: Widespread Dynamics During Visual Short-Term Memory ...... 27 Specific Aim 3: Spatial Distribution of Oscillatory Activity ................................ 27 Figures ........................................................................................................................... 28 References ..................................................................................................................... 32 2. A LARGE-SCALE SEMI-CHRONIC MICRODRIVE RECORDING SYSTEM FOR NON-HUMAN PRIMATES ............................................................................................. 47 Contribution of Authors and Co-Authors ...................................................................... 47 Manuscript Information Page ........................................................................................ 48 Summary ........................................................................................................................ 49 Introduction ................................................................................................................... 49 Results ........................................................................................................................... 52 Large-Scale Semi-Chronic Microdrive System ..................................................... 52 Implanting the Chamber and Microdrive .............................................................. 55 Large-Scale Recording in Behaving Non-Human Primates .................................. 58 Segregated Functional Networks Revealed by Relative Phase Relationships ...... 62 Discussion ...................................................................................................................... 65 Methodological Considerations ............................................................................. 66 Overall Performance .............................................................................................. 68 Future Directions ................................................................................................... 69 Figures .......................................................................................................................... 71 Supplemental Material .................................................................................................. 83 Experimental Model and Subject Details .............................................................. 83 iii TABLE OF CONTENTS -- CONTINUED Subjects ............................................................................................................. 83 Method Details ......................................................................................................... 83 Behavioral Task ................................................................................................ 83 Chamber System ............................................................................................... 84 Electrophysiological Recordings ...................................................................... 85 Anatomical Reconstruction Technique for Monkey 2 ..................................... 85 Quantification and Statistical Analysis .................................................................... 86 Network Analysis ............................................................................................. 86 References .................................................................................................................... 89 3. FEATURE-BASED VISUAL SHORT-TERM MEMORY IS WIDELY DISTRIBUTED AND HIERARCHICALLY ORGANIZED ........................................... 96 Contribution of Authors and Co-Authors ...................................................................... 96 Manuscript Information Page ........................................................................................ 97 Summary ........................................................................................................................ 98 Introduction ................................................................................................................... 98 Results ......................................................................................................................... 100 Task-Dependent Activity is Widely Distributed ................................................. 102 Embedded Hierarchy for Mnemonic Representations ......................................... 105 Microsaccades Encode Visual Memories ............................................................ 109 Discussion .................................................................................................................... 110 Methodological Caveats ...................................................................................... 112 Memory Maintenance in V1 and V2 ................................................................... 113 Decoding Items in Memory with Microsaccades ................................................ 114 The Balanced Activity State and Short-Term Memory ....................................... 115 Short-Term Memories are Maintained in a Hierarchy of Cortical Areas ............ 115 Tables ......................................................................................................................... 117 Figures ........................................................................................................................ 124 Supplemental Material ................................................................................................ 136 Experimental Model and Subject Details .............................................................. 136 Subjects ............................................................................................................. 136 Method Details ...................................................................................................... 136 Behavioral Task ................................................................................................ 136 Recording Techniques ...................................................................................... 137 Histology .......................................................................................................... 139 Anatomical Hierarchy ...................................................................................... 140 Spike Sorting .................................................................................................... 141 Selection of Units ............................................................................................. 141 Quantification and Statistical Analysis ................................................................. 142 Firing Rate Analyses ........................................................................................ 142 Microsaccade Detection Procedure .................................................................. 144 Microsaccade Modulation Analysis ................................................................. 144 iv TABLE OF CONTENTS -- CONTINUED Microsaccade Pattern Analysis ...................................................................... 145 References ................................................................................................................... 147 4. THE CORTICAL LOCAL FIELD POTENTIAL EXHIBITS DISTINCT SPATIAL GRADIENTS THAT VARY WITH FREQUENCY AND TIME DURING VISUAL SHORT-TERM MEMORY ............................................................................................. 152 Contribution of Authors and Co-Authors .................................................................... 152 Manuscript Information Page ...................................................................................... 153 Summary ..................................................................................................................... 154 Introduction ................................................................................................................. 154 Results ......................................................................................................................... 156 Spatial Organization of the LFP .......................................................................... 157 Decoding Cortical Areas ..................................................................................... 160 Feature Importance .............................................................................................. 163 Areal Misclassification ........................................................................................ 164 Discussion .................................................................................................................... 166 Methodological Considerations ........................................................................... 167 Frequency Differences ......................................................................................... 169 Anatomical Correlations ...................................................................................... 170 Detailed Methods ......................................................................................................... 171 Behavioral Task ................................................................................................... 171 Electrophysiological Recording .......................................................................... 172 Areal Grouping .................................................................................................... 173 Identification of Spectral Bands .......................................................................... 173 Classifier Features ............................................................................................... 174 Machine Learning Classifier ............................................................................... 174 Tables .......................................................................................................................... 176 Figures ......................................................................................................................... 178 References ................................................................................................................... 201 5. GENERAL DISCUSSION .......................................................................................... 205 Summary of Results .................................................................................................... 205 Future Directions ......................................................................................................... 206 Concluding Remarks ................................................................................................... 208 References ................................................................................................................... 210 CUMULATIVE REFERENCES ................................................................................. 212 v LIST OF TABLES Table Page 1. List of Areas Contained in Each Group .......................................................... 117 2. Sampling Distribution of Unit Activity ........................................................... 118 3. Recording Counts for All Areas and Areal Grouping In Monkeys E and L ........................................................................................ 176 4. Correlation Between Features Variability and Validation Accuracy .............. 177 4. Correlation Between Neuron Density and Spectral Content ........................... 177 vi LIST OF FIGURES Figure Page 1. First Reported Photograph of Electrical Brain Activity .................................... 28 2. Early Recording Device .................................................................................... 29 3. Early Recordings ............................................................................................... 30 4. Popularity of Oscillations and Brain ................................................................. 31 5. Design Drawings for Designs 1 and 2 of the Large-Scale Chamber and Microdrive System ............................................... 71 6. Steps in the Assembly and Sealing of the Microdrive System for Design 2 Only .............................................................. 72 7. Craniotomy and Microdrive Implantation Procedure for Stages II and III .......................................................................... 73 8. A Brief 300 ms Segment of Broadband Data Simultaneously Sampled from 88 Microelectrodes In Monkey 2 Just Prior to the Onset of the Match Stimulus during the dMTS Task ........................................................................ 74 9. Anatomical Reconstruction Process .................................................................. 75 10. Summary of Recording Performance .............................................................. 76 11. Results of the Functional Connectivity Analysis ............................................ 77 12. Design of the Skull Model and Definition of the Chamber Boundaries ....................................................................................... 78 13. Exploded Views of the Design Drawings of the Large-Scale Chamber and Microdrive Systems For Each of the Three Stages of Design 1 and Design 2 ................................. 79 14. Actuator Mechanism ....................................................................................... 80 vii LIST OF FIGURES CONTINUED Figure Page 15. Broadband Data Simultaneously Sampled from 88 Microelectrodes in Monkey 2 for the Full Duration of The Trial Shown in Figure 4 ........................................................................... 81 16. Recording Quality Over Time for a Single Electrode And the Changes in Electrode Impedance Across the Experiments ................................................................................... 82 17. Behavioral Task, Distribution of Recording Sites, and Raw Data Collected on a Sintle Trial ..................................................... 124 18. Task-Dependent Changes in Firing Rates ..................................................... 125 19. Task-Dependent Unit Activity is Widely Distributed ................................... 126 20. Stimulus Selective Changes in Firing Rates During The dMTS Task ............................................................................................. 127 21. Distribution of Stimulus Selectivity Across the Sampled Cortical Areas in Both Monkeys .................................................... 128 22. Embedded Hierarchy for Mnemonic Representations ................................... 129 23. Microsaccades Encode the Sample Stimulus During the Delay Period ................................................................................ 130 24. Microsaccade Modulation Analysis .............................................................. 131 25. Microsaccade Modulation Analysis Results ................................................. 132 26. Task-Dependence Analysis for All 24 Areas/Groups Over the Course of the Task ................................................... 133 27. Stimulus-Selectivity Analysis for all 24 Areas/Groups Over the Course of the Task ................................................... 134 28. Incidence of Stimulus-Selectivie Activity (Solid Lines) And the Incidence of Both Stimulus-Selective and Task- Dependent Activity (Dashed Lines) for the 9 Areas/Groups In the Hierarchy ............................................................................................. 135 viii LIST OF FIGURES CONTINUED Figure Page 29. Spatial and Task-Dependent Variation in the Spectral Organization of the LFP Across the Cortex During A Single Recording Session .............................................................................. 178 30. Distributions of Peak Frequencies ................................................................. 179 31. Spectral Variables and Their Task Dependence ............................................ 180 32. Cortical Flatmaps of Spectral Content .......................................................... 181 33. Variation of Sample Entropy (SE) Across The Cortex for Both Monkeys ...................................................................... 182 34. Baseline Epoch Areal Classification ............................................................ 183 35. Combined Epochs Areal Classification ........................................................ 184 36. Validation Accuracy Improvement with Inclusion of All Epochs ............................................................................... 185 37. Feature Importance ....................................................................................... 186 38. Correlations in Validation Accuracy and Variation ..................................... 187 39. Multidimensional Scaling of the Pairwise Confusability ............................. 187 40. State Dependent Oscillations ........................................................................ 188 40. Correlations in Cell Density and Spectral Content ....................................... 189 41. Flatmaps of the Recording Count in Cortical Area/Groups In Both Monkeys .......................................................................................... 190 42. Peak Detection Results in Monkey E .......................................................... 191 43. Peak Detection Results in Monkey L .......................................................... 192 44. Distributions of Spectral Peaks Obtained from All Channels and Sessions .................................................................................. 193 ix LIST OF FIGURES CONTINUED Figure Page 45. Boxplots of the Spectral Content for Monkey E .......................................... 194 46. Boxplots of the Peak Amplitude for Monkey E ........................................... 195 47. Boxplots of the Spectral Content for Monkey L ......................................... 196 48. Boxplots of the Peak Amplitude for Monkey L ........................................... 197 49. Sample Entropy Example ............................................................................. 198 50. Distribution of Validation Accuracies as a Function of Task Epoch for Each Area/Group in Monkey E ............................................. 199 51. Distribution of Validation Accuracies as a Function of Task Epoch for Each Area/Group in Monkey L ............................................. 200 x ABSTRACT Cognitive processes occur through coordinated activity via disparate cortical and subcortical brain structures. Although these structures may be widely separated, evolutionary pressures dictate that cognition must occur rapidly and efficiently. In order to capture these brain-wide activity patterns the tools for measuring them need to be similarly capable of measurements of both high spatial coverage, and high temporal resolution. Additionally, the measurements would ideally be of the activity of the fundamental units involved in cognition, that is the neurons, rather than an extrapolation of their activity via a different signal source. However, outside of the work presented here, current technologies are rare that allow both the requisite coverage and spatiotemporal resolution to achieve these measurements. The results of the studies presented in Chapters 2-4 provide both the tools for making such measurements, and the initial analyses of the neuronal dynamics during short-term memory. In Chapter 2 we present the technological and methodological process for recording neural activity (both action potentials and local field potentials) from across roughly a hemisphere of cortex in the macaque monkey performing a visual short- term memory task. In visual short-term memory a visual percept must be maintained then recalled when it is no longer present. This cognitive process is one we use nearly incessantly in every-day life. In Chapter 3 we found task dependent spiking activity during short-term memory is wide-spread, and that most areas display a balanced state of both increases and decreases in firing rate. Within these areas we found a hierarchically organized subset of cortical areas that also showed stimulus specific activity during the memory period of the task. In Chapter 4 we used spectral analysis to investigate the oscillatory make-up of neural activity across the recorded areas. We found within specific frequency bands there are different gradients of amplitude of spectral power across cortex. Additionally, we found that we could use a small number of spectrally derived variables in order to decode the brain area origin of the signal. This shows that areas have a characteristic spectral composition, that varies systematically across the cortical mantle. xi FORWARD “I love deadlines. I love the whooshing noise they make as they go by.” -Douglas Adams (Adams, 2002) I just wish I had more time! Working on this project has made me feel as that of a sculptor. It has taken years to shape this block down into a discernable figure, it’s getting there, but with a few more months, years or decades we can carve out a real Venus de Milo. It turns out that simultaneously recording electrical activity from hundreds of microelectrodes across the breadth and depth of a cerebral hemisphere is quite the endeavor (who would a thunk it?). Yet, when I showed up at my first Society for Neuroscience annual conference years ago, I assumed it was the norm in the field. The interpretation of these data is another story. This sentiment was predicted in the writing of a pioneer of the field of electrophysiology at the end of his nearly 7 decade career, Ted Bullock: “Even hundred-channel recording[s] of units, even if each channel were, as never before, wideband to allow both spikes and slow signals to be seen, is unlikely to make interpretation easier” (Bullock, 2002). My hope here is to contradict this statement, and to relay a semblance of understanding from what we (both as a field and as a lab) have done and a glimpse at what we have coming down the pike. 1 CHAPTER ONE GENERAL INTRODUCTION Introduction Over a century of microelectrode research has provided a strong base for our understanding of cognitive processes at the level of the single neuron (Hubel and Wiesel 1959; Fuster and Alexander, 1971; Moran and Desimone, 1985; Funahashi et al., 1989). However, it has proven difficult to record from more than a few microelectrodes at a time, and subsequently most of these studies lack the coverage to address how cognitive processes are carried out at the level of the brain as a whole. Many studies have well established that cognitive processes involve the interactions of dispersed neural populations (Mountcastle, 1978; Mesulam, 1990; Gray, 1994; Bressler, 1995; Friston, 1997; Bressler and Kelso, 2001; Fuster and Bressler, 2012; Siegel et al., 2012). There exist several modalities for the recording of neural activity, each having their own set of strengths and weaknesses. Many Non-invasive technologies such as functional imaging (functional magnetic resonance imaging; fMRI), electroencephalography (EEG), and magnetoencephalography (MEG) allow the recording of activity from widespread brain areas and corroborate that cognitive processes play out across disparate brain regions (Munk et al., 2002; Harrison and Tong 2009; da Silva 2013). These studies indicate that even simple cognitive tasks involve the coordinated actions of widespread cortical and subcortical areas. Although these non-invasive technologies allow great spatial breadth, spatial smoothing due to the scalp as in EEG deteriorate their spatial resolution, making it 2 a somewhat foggy correlate of the cortical neural activity. In the case of fMRI they further suffer in poor temporal resolution. Due to these limitations, we developed a large-scale recording device allowing for high spatiotemporal resolution of recordings with the microelectrode, and with high coverage across cortical and even subcortical areas (Dotson et al., 2015; 2017). The broadband activity recorded extracellularly from a microelectrode simultaneously provides two types of data, unit activity and the local field potential (LFP). As I will be presenting research involving both of these data types I will give a brief summary of them here. The first data type, unit activity, refers to action potentials from one (single unit) or a group of neurons (multi-unit) within a radius of about 200 microns of the microelectrode tip (although factors that affect the electrode impedance such as electrode diameter, or shape affect this; Freeman, 2007; Lewicki, 1998). The spiking events are extracted by high-pass filtering the broadband signal, and recording the time events that cross a threshold level. Often times, in a process known as “spike sorting” researchers can use the various parameters of the waveform (e.g. spike amplitude, spike width) and timing of the spikes (i.e. inter-spike-interval) in order to isolate out the activity of a single neuron or a small number of neurons (termed single unit activity; SUA; Lewicki 1998). Spiking activity that cannot be deemed attributable to a single neuron is also recorded, with this activity arising from the surrounding population of neurons (termed multi-unit activity; MUA; Lewicki 1998). The local field potential is the result of the summed effect of all the various transmembrane ionic currents occurring within a volume of cortical tissue, although the main source is that of synaptic activity (Buzsaki and Draguhn, 2004; Wang, 3 2010; Buzsaki et al., 2012). It is thought to reflect the activity of the neural population (on the order of 105 neurons/mm3; Freeman, 2007; although see Beul et al., 2017 on the heterogeneity of neuron density across cortex), with a diameter on the order of 0.25-1mm of cortical tissue (Katzner et al., 2009). The dendritic currents that give rise to the LFP spread via volume conduction (Nunez et al., 1997) to the most internal of the meninges (the pia), where the electrocorticogram (ECoG) is measured, as well as the overlying scalp which is where the electroencephalogram is measured (EEG). At each one of these steps (from LFP to ECoG to EEG) the recording electrode becomes more distal to the cortical tissue giving rise to it, which results in a loss of spatial resolution. Generally, investigation of the LFP begins by transforming the recorded voltage time-series into the frequency domain (i.e. Fourier or wavelet transform), which quantifies the oscillatory “ingredients” of the signal. Despite decades of study there is only a piecemeal understanding of how the predominant frequencies of the LFP power spectra vary across the cortex (something that is substantially better characterized in the related extra-cortical EEG signal; Niedermeyer, 2005; Maurer and Dierks, 2012). Albeit certain regions of the brain are well characterized to the extent that they are associated with their own specific frequencies of observed oscillations and their task or state dependence. These include and are not limited to the occipital alpha oscillation observed when the subject closes their eyes (Berger, 1929). The somato-motor beta oscillation in the absence of peripheral limb stimulation or movement (Jasper and Penfield, 1949). The visual gamma oscillation in the presence of stimulation of the receptive field with an optimally oriented drifting light bar (Gray and Singer, 1989). Additionally, there are state dependent oscillations such as the delta rhythm during slow 4 wave sleep (Amzica and Steriade, 1998). However, while there exists a wealth of information about many individual brain areas, the overall spectral pattern of the LFP across the cortex remains poorly characterized. In order to investigate these questions, we used a large-scale recording device allowing us to simultaneously record from many cortical areas within a cerebral hemisphere during a visual short-term memory task (Dotson et al., 2017). A primary question was to characterize the frequencies at which oscillatory activity occurred, and the spatial organization of these signals. By answering these questions, we can shed light on the spectral organization of the LFP across the cortex and the widespread spatio-temporal dynamics that occur during visual short-term memory. Additionally, the results of these analyses will allow us to focus on specific regions and frequencies for further analysis of synchronous activity and directed interactions. The remainder of the General Introduction will provide the theoretical, technological and experimental framework that motivated the studies presented in Chapters 2-4. Chapter 2 will describe the technological design that allowed the widespread recording of cortical activity. This design allowed for an unparalleled simultaneous sampling of neural data across a cortical hemisphere. Additionally, due to the actuator mechanism of the electrodes, recordings were able to be achieved from deep, sub-cortical structures. Chapter 3 describes the initial findings of widespread neural activity during visual short- term memory, with a focus on the single and multi-unit activity. In this chapter we show how essentially all cortical areas recorded from show task dependent activity1, and that 1 To quote a key individual involved in the project “The whole damn brain is involved!” 5 within a cortical area the nature of this activity resolves to an essentially balanced level of both increases and decreases in firing rate relative to the baseline level. Additionally, we found a subset of hierarchically organized areas that displayed stimulus specific modulations in firing rate during the memory period of the task. We also found that microsaccades (miniscule eye movements) during the memory period of the task encoded the stimuli held in memory, suggesting that eye movements play a role in visual short-term memory. Chapter 4 will demonstrate the widespread variations in the local field potential (LFP) and its spectral power across the recorded areas during the visual short-term memory task. In order to accomplish this analysis, we developed a peak detection process to define the frequency bands of interest. We then show that spectral power in the different frequency bands display gradients of amplitude across cortex. Suggesting that cortical areas have unique spectral profiles. Next we show that the calculated sample entropy (a proxy of the stochasticity of the signal), the relative power, and peak amplitude in each frequency band can be used as features to decode the brain area origin of the signal. Interestingly we found that the decoding accuracy increased with the addition of information from the other task epochs, suggesting that the power in the different epochs adds further information for decoding cortical area. Chapter 5, the General Discussion, will speculate on possible reasons for what is reported in the previous chapters, as well as provide potential caveats and future directions of research in this field. A Brief History of Oscillatory Electrical Brain Activity 6 Many credit Hans Berger with the discovery of oscillatory electrical activity arising from the brain, with his work measuring activity on the human scalp (Berger 1929). However this claim is in fact misguided as there are several reports of electrical brain activity measured from a variety of animals dating back more the 50 years previous to when Berger first published his oft-cited work (although Berger can still lay claim to the first one to observe this in humans). Although the phenomena of electrical activity in nervous tissue was earlier speculated on, it wasn’t until 1791 that this “intrinsic form of electricity” was demonstrated. Albeit this claim was met with some debate by the contemporaries of the time, and was largely limited to stimulations, not recordings of peripheral nerves (Galvani, 1791; Brazier, 1961; Piccolino, 1998). Later, with improvements made to the galvanometer, Du Bois-Reymond and von Helmholtz contributed more stable evidence to this claim, also measuring activity and even calculating conduction velocity in a peripheral nerve (Du Bois-Raymond, 1848; Helmholtz, 1850). Likely driven by the lesion techniques employed by Ferrier, who was attempting to attribute various parts of the brain to specific functions, we find the first (rather brief) reported recordings from cortical tissue made in 1875 by Richard Caton, whom subsequently gave a slightly more detailed report. Caton’s drive appeared to be to elucidate the localization in the brain where a peripheral stimulus was represented (Ferrier, 1874; Caton, 1875; Caton, 1877). Despite largely failing in this regard Caton did report on the oscillatory brain activity in recordings made from a number of different animals, including monkeys (Caton, 1875). He noted “the current is usually in constant fluctuation…”, and even hinted at a link to cognition noting fluctuations that may 7 coincide with “the animals’ mental condition” (Caton, 1875). However, due to the lack of recording technologies and analytical tools, these earliest observations of oscillatory brain activity were just brief written descriptions. Unfortunately it appears that the diffusion of these reports of Caton’s failed to occur or that it had it fallen on deaf ears, as in the subsequent years several others working separately and unaware of each other attempted to lay claim to having first observed electrical brain activity (Danilevsky, 1877; Beck, 1891; and Fleischl von Marxow, 18832). The debate for primacy amongst these early electrophysiologists wasn’t ultimately cleared up until Caton chimed in notifying them all of his observations published prior to all their claims (Brazier, 1961). Of note among these pioneers was Adolf Beck, who in 1891 appears to have been one of the first to emphasize the oscillatory nature of the phenomenon they were recording, observing: “…there was a continuous waxing and waning variation taking place which neither was related to the respiratory rhythm nor was it synchronous with the pulse, nor finally was it in any way dependent on movement of the animal, since it was present in curuarized dogs.” He later went on to posit a relation to behavior saying: “I therefore believe it justifiable to consider that these fluctuations are the expression of a fundamental variation which is the state of activity taking place in the cortical centres.” (Beck, 1891; Brazier, 1961; pp 30). 2 This was a sealed letter detailing a brief unpublished observation Fleischl von Marxow had originally written in 1883. 8 Others continued investigations, with the first publication containing a visual reference to oscillations through the use of a galvanometer equipped with a string allowing the trace to be recorded on photographic media (Figure 1; Pravdich-Neminski, 1913; Brazier, 1984). Pravdich-Neminski was not only the first to provide photographic visualizations of his recordings but was also the first to characterize the frequency of the observed oscillations, even going as far as to categorize them in the five “classes of rhythm”. Others continued the development of photographic visualization techniques, including a notable innovation called the Kymograph, employed by Beck’s advisor, Cybulski as well as Macieszyna (Brazier, 1961). The modern digital era experimenter may find the techniques of these early explorers somewhat amusing-- Cybulski’s technique, for example, was that the experimenter would observe the fluctuations on the galvanometer needle through a telescope, meanwhile attempting to mirror the observed fluctuations by adjusting a signal pen on a rotating smoke drum. Thus, allowing them to record oscillations over time as etchings in the ash coated drum (Brazier, 1961). Early analyses were crude, with brief descriptions being limited mentions of “cycles per second”. It would take half a century for analysis techniques from various fields (such as Fouier decomposition) to be applied to electrophysiological data. One may speculate that the field of studying the oscillations of the brain was somewhat dissipated with the growth of Santiago Ramon y Cajal’s neuron doctrine (Ramón 1899). Where many in the field turned to the investigation of the anatomical make-up of the brain and single neurons. We will see that this was really just the beginning of a century long series of pendulum swings of focus in neuroscience between single cells and larger 9 volumes of tissue. Indeed, to this day this debate between the importance of single cell recordings and the oscillations of the cortical milieu continues. Although before arriving at the present day, I would like to review a couple more periods of this oscillating field. In the late 1920’s focus began to swing back to oscillations with the work of Hans Berger, who first measured the scalp EEG in humans, recording the data in 1924, but not publishing the data until 19293. Indeed, it was he who began associating the observed oscillations with behavioral correlates, such as the extinguishing of the occipital alpha oscillation when the subject (often his son) opened their eyes (Berger 1929). Berger further characterized the oscillations in terms of their frequency, observing the slower occipital “alpha” rhythm, compared to the faster somato-motor “beta” rhythm. It appears though that these early findings of Berger did not come into the mainstream until they were subsequently confirmed by Adrian, as Berger (a devotee of various psychic phenomenon, and “…having rather the reputation of being a crank”; Walter, 1953) lacked the knowledge, or at least ability in explaining the methodological and technical aspects of exactly what he was recording (Karakas and Barry, 2017). However, this initial concentration on the recording of the “mass action potentials” somewhat shifted with the pioneering development by Adrian of a more adequate electrode (i.e. high impedance) and recording technologies allowing the recording of single units (Adrian and Moruzzi, 1939). Although Adrian was not solely focused on the recording of single units, and made important contributions to the subject of field potentials, such as the observation of the 40 Hz “sniff” rhythm in the hedgehog olfactory bulb (Adrian and Mathews, 1942). The difficulty in 3 I know the feeling… 10 interpretation of field potentials led to a more or less plateau in focus in them, where as other fields of neuroscience such as the study of actions of single neurons continued to be developed for much of the middle part of the 20th century. It should be mentioned that there were a handful of researchers who continued to work on field potentials, with a focus on their interpretation and genesis (Bishop 1936; Marshal et al., 1937; Bremer 1949). Among these early pioneers was Ted Bullock, who by the estimate of this author must hold the record for number of different species recorded from (for a review of the field at the time see Bullock, 1945). During this era we see the field swing its focus to the biophysics of the single unit activity with groundbreaking work of Hodgkin, Huxley and Katz on the mathematical description of the dynamics of ion channels of the isolated squid giant axon which give rise to the action potential (Hodgkin and Huxley, 1945; 1953; Hodgkin et al., 1951). These works were contemporaneous with those of Hartline, Hubel and Wiesel’s series of outstanding experiments describing the concept of receptive fields (Hartline 1940) and tuning properties of single unit recordings in anesthetized cat visual cortex (Hubel and Wiesel, 1959). These works swayed the field towards the direction of focus on single units, where it adapted an almost dismissive posture toward the study of field potentials. The areas where there continued to be a focus on oscillatory activity were largely where the invasive techniques necessary to obtain single unit recordings were not possible (i.e. human EEG recordings). However, contemporaneous with these works we find the ambitious investigations of Herbert Jasper and Wilder Penfield. They proposed the idea that different parts of cortex responsible for different functions would exhibit different patterns of electrical activity by making electrode recordings from the cortical surface in epileptic 11 patients (Figure 2-3; Jasper and Andrews, 1938; Jasper and Penfield, 1949). It is in this era where the study of “evoked potentials” began via the development of an analysis technique that relied on the aggregate signal over many presentations of the sensory stimulus (Dawson, 1951). This led later to the discovery of the P300 (oddball) event-related potential (ERP), that is a positive deflection 300ms after presentation of a deviant stimulus within a sequence of expected stimuli (Sutton et al., 1965). An interesting sidebar to these studies is the work of John C. Lilly who while working at NIH at the time was developing techniques for the visualization of brain activity from multiple electrodes (the 25 channel BAVATRON; Lilly 1950), as well as discovering an electrical stimulus pulse that would not damage the cortical tissue (Lilly et al., 1955), and developing a cannulae insertion procedure for electrode recording that negated the need of performing a craniotomy (Lilly, 1958). In this later study he reports to have “…explored about 500 zones in two monkeys…” with one monkey being implanted simultaneously with 20 cannulae, although what exactly these “zones” were remains ambiguous. From these studies Lilly hypothesized that the brain was simply a stimulus response structure, which led him (while still at NIMH) to develop the sensory deprivation tank (Lilly, 1997). Ultimately studies of field potentials were brought further into the mainstream by the works of those such as Walter Freeman, who with early theoretical work and complementary experimental backing showed how they are implicated in a range of cognitive behaviors (Freeman, 19754). In the 1980’s the subject of field potentials experienced a resurgence of interest with such highlights as: increased use of array 4 This viewpoint may be biased due to my personal academic genealogy. 12 recordings (Bressler and Freeman, 1980); links to spiking activity, stimulus specificity, and their relation to the “binding problem” (Gray and Singer, 1987; Eckhorn et al., 1988; Gray and Singer, 1989). With these studies the field was brought into the modern era, and as Buzsáki put it out of the “depression” it was experiencing in the early 1980’s (Buzsáki, 2006). Today both field potentials and unit activity continue to be explored in their relation to a wide range of cognitive, and behavioral processes. Proposed Roles of Field Potentials “Field potentials, although easy to record, are difficult to interpret” Ulla Mitzdorf (Mitzdorf, 1985) Are Oscillations Vital, or Simply an Epiphenomena? Any neurophysiologist recording broadband activity with a goal to investigate the dynamics of spiking activity will jointly be recording the field potential activity surrounding the microelectrode. However, due to the difficulties in interpretation of these LFP data, they are often disregarded. It is curious to think that many investigators deem the LFP an “epiphenomenon” and in essence wholeheartedly ignore half of the data they collect! In the late 90’s the pioneering neurophysiologist Ted Bullock shared his frustrations on the subject writing: “The dichotomy between two groups of workers on neuroelectrical activity is retarding progress” (Bullock 1997). In their comprehensive text “Electronic Fields of the Brain”, authors Nunez and Srinivasan express these sentiments, attributing those whose dismiss field potentials as an “epiphenomenon” as guilty of exercising “chauvinism of a spatial scale, the assumption that data and theory at one’s favored scale is the most useful, or in extreme cases an inability to recognize scientific merit outside of 13 one’s own subfield” (Nunez and Srinivasan, 2006). Hopefully in the subsequent sections I will be able to lay these claims to rest by showing the variety of utilities of field potentials. Commonly field potential activity is transformed into the frequency domain which allows the analysis of the spectral dynamics of the signal. The spectra are often divided into different canonical frequency ranges with each given a Greek letter (which corresponds not to its order in terms of frequency range but to its order of chronological discovery/acceptance into the literature). These ranges of “rhythms” have been somewhat arbitrarily drawn on the basis of the number of cycles per second (Hz), with the most common designations being: delta (0.5-4 Hz), theta (4-8 Hz), alpha (8-14 Hz), beta (14-30 Hz), and gamma (30-90 Hz). Although this list can vary in frequency ranges and may also include other types of oscillations such as: ultra slow (< 1Hz; Steriade et al., 1993a), the Rolandic mu (wicket) (7.5-12.5 Hz; often associated with voluntary movement, similar to a motor alpha rhythm), beta 1 (12.5-16 Hz), beta 2 (15-20 Hz), high beta (20-30 Hz), and “high gamma”, “ultrafast rhythms”, very high frequency oscillations (VHFO), or epsilon rhythms (90 Hz +) (Buzsaki, 2006; Gonzalez et al., 2006). The generation, importance, and interpretation of field potentials continues to be an active area of both theoretical and experimental research (Buzsaki et al., 2012; Einvoll et al., 2013; Herreras 2016; Pesaran et al., 2018). Indeed, the amount of research in the area of brain oscillations has steadily increased since the 1940’s (Figure 4). In some areas, such as clinical epilepsy diagnosis and localization of treatment, the utility of field potentials has been evident for decades (Binnie and Stefan, 1999; Le Van Quyen et al., 2001). However, in other areas, such as 14 mental health research, the clinical applicability (e.g. the use as a biomarker for psychiatric disorder) and interpretation of field potentials remains more ambiguous (Mathalon et al., 2015). Here I will attempt to summarize some of the more general proposed roles and interpretations of field potentials recorded both from the scalp (electroencephalogram; EEG) and intracortically (LFP)), as a function of the various specific frequencies of them5. I stress that my intention here is not to construct a comprehensive review of each of these rhythms in their entirety (which would be a gargantuan task), but to briefly summarize some of the highlights. Delta Oscillations (0.5-4 Hz) Delta rhythms were the fourth described type of oscillation (following Berger’s Alpha and Beta oscillations, and Jasper and Adrian’s Gamma oscillations) and were first described in the early 30’s by William Grey Walter (Grey Walter, 1963). Delta rhythms are of high amplitude comparatively and are typically associated with subject being in the unconscious state, whether it be due to sleep, coma, or anesthesia. Indeed, the majority of publications relating to them are in their relation to sleep, and they are a key characteristic of slow wave sleep (Steriade and McCarley, 2005). Clinically they are used in the determination of the stage of sleep the subject is in (Steriade et al., 1993; Berry et al., 2012), and disruptions in delta activity are associated with a variety of parasomniatic sleep disorders (Pilon et al., 2006). Decreases in the spectral power of the delta rhythm have been associated with alcoholism (Colarin et al., 2009), and aging 5 A note on terminology: I will be using rhythms, oscillations, and field potentials rather interchangeably to refer to activity in that band. Often times the term oscillation is used for describing activity in a particular band even if there is not a bona-fide oscillation present. 15 (Colrain et al., 2010). Delta oscillations have also been implicated in their roles in a number of cognitive tasks (Harmony, 2013), as well as in in homeostatic motivational drives such as hunger (Knyazev, 2007; Dubbelink et al., 2008). Increases in delta power have also been observed in prefrontal and frontal regions upon success in working memory tasks (Fernández et al., 2002) in memory tasks of greater relative difficulty (Harmony et al., 1996) and in attentional processing (Shroeder and Lakotos, 2009). The literature is relatively sparse when it come to research on intracortical delta oscillations (LFPs) during cognitive tasks, with a few exceptions; In a numerosity matching task delay period increases in delta/theta activity in prefrontal cortex (PFC) have been observed (along with a simultaneous decreases in ventral interparietal area (VIP) beta activity (Jacob et al., 2018); and interestingly in a freeviewing task the phase of delta/theta LFP activity was found to be locked to the saccade endpoint (fixation onset) as opposed with the higher frequency activity being locked to the saccade onset (Ito et al., 2013). Theta Oscillations (4-8 Hz) Of all the oscillation “flavors” associated with particular brain structures, theta rhythms are certainly a contender with their affiliation with the hippocampus. They additionally are likely a contender of one of the most intensively researched topics. The first reports of theta activity arise in the 1930’s in subcortical (including the hippocampus) recordings from the rabbit (Jung and Kornmüller, 1938). The authors report near sinusoidal activity in the 5-6 Hz range, using a multiple electrode recording device. These reports of “very regular waves of 5-7/sec were subsequently corroborated (and this time in English, which facilitates 16 comprehension for this author) in the hippocampi of both rabbits and cats under various forms of anesthesia, and stimulation (presentation of various olfactory, auditory and visual stimuli; Green and Arduini, 1953; Grastyán et al., 1959). Today there is strong evidence for the relationship of not only hippocampal theta, but scalp recorded theta and its relationship with learning and memory across a range of species (Winston, 1978; Mizumori et al., 1990; Klimesch et al., 1996; Klimesch 1999; Fell et al., 2001; Benchenane et al., 2010). Intracortical LFP recordings support this where it has been shown in monkeys performing a delayed match-to sample task elevated delay period theta activity in visual area V4, which was correlated to increased phase locking of recorded spikes (Lee et al., 2005), as well as increased coherence within V4 in the theta band (Hoerzer et al., 2010), and an increase in theta synchronization between lateral PFC and V4 (Liebe et al., 2012). Suggesting that theta oscillations may play a role in facilitating long range communication between brain areas (Von Stein and Sarnthein, 2000; Sauseng et al., 2005). Another aspect of theta oscillations that has been well studies is their role in spatial navigation. Following their discovery of place cells in the rat hippocampus (O’Keefe and Dostrovsky, 1971; O’Keefe, 1976) O’Keefe and colleagues found an interesting relationship between the spike timing of the place cells with the ongoing theta oscillation. It appeared that as the rat entered and moved through the place field, the firing of the place cell shifted along the phase of the theta oscillation (O’Keefe and Recce, 1993). These findings tied together two aspects of hippocampal theta rhythms, their roles in both memory processes (specifically episodic memory), and navigation. We find here an 17 argument for this relationship in that both aspects involve a specific procession of events. In episodic memory it is the specific procession of events occurring in time, where as in navigation is the specific procession of spatial locations (Buzsaki, 2006). Alpha Oscillations (8-12 Hz) Alpha oscillations were the first described rhythms, which is reflected in their name, something that their discoverer would likely be displeased with in that the term “Berger waves” has failed to catch on (Berger 1929). After suffering a somewhat lull in focus through most of the middle part of the 20th century there appeared to be a resurgence in interest in the mid-1990s. This lull was likely due to interest in alpha oscillations being leap-frogged by interest in the subsequently discovered beta and gamma oscillations, at one point during this era alpha oscillations were regarded as “noise”, “smoke” or “idling of the brain” (recounted in Adey, 1988). Their origin is likely linked with the thalamus (Roux et al., 2012) and are implicated in a wide range of functions (Basar, 2012). One proposed role of alpha oscillations that they are a mechanism for active inhibition of brain areas not involved in the task at hand (Jensen and Mazaheri, 2010). This theory stems from the long thought idea that alpha activity is correlated to “cortical idling”, such as is exemplified by the canonical observation of higher EEG alpha activity over visual cortex when an individual’s eyes are closed (Berger, 1929; Adrian and Mathews, 1934; Klimesch et al., 1996; Pfurtscheller et al., 1996). However, this idea of alpha activity reflecting cortical idling is contradicted by other studies that show an increase in alpha activity in a memory task during the delay period dependent on memory load (Jensen et al., 2002; Tuladhar et al., 2007), and in frontal, parietal and occipital regions during conscious visual perception (Babiloni et al., 2006). If one thinks of alpha activity of 18 reflecting idling, and the idle state as being a binary “on” or “off”, we would not suspect to see varying degrees of “offness”. Further, these latter studies found that the magnitude of alpha activity was predictive of performance in the task, with higher alpha activity in the irrelevant brain areas corresponding to higher performance. Thus, Jensen and colleagues suggest that alpha activity is rather a form of active inhibition, where it acts in gating resource allocation, and that increases in its amplitude should be observed in task irrelevant brain areas (Jensen and Mazaheri, 2010). It was further shown using trans-cranial magnetic stimulation (TMS) at the alpha frequency that working memory performance could be depleted upon stimulation over parietal cortex (Riddle et al., 2020). Directly contrary to these views as alpha activity reflecting an active inhibition process, we find the views of Palva and Palva, whom argue that increases in alpha activity during short-term memory tasks is actually a reflection of active memory maintenance (Palva and Palva, 2007). The details of this argument stem from a general disagreement on exactly what brain areas are involved in in the behavioral task. Where it is likely that a broad network of brain loci are involved in tasks such as attention or short-term memory (Courtney et al., 1997; Munk et al., 2002; Constantinidis and Procyk, 2004; Postle, 2006), thus it is difficult to perfunctorily attribute a brain area as task irrelevant. Thus, arises a dichotomy on whether alpha power increases reflect an active or an inhibitory process. Citing this ambiguity in alpha amplitude, Palva and Palva instead contend that it is within the phase relationship of the alpha oscillation where the true mechanism of alpha activity lies. They note that although the variation in alpha power remains ambiguous, reflecting both increases and decreases during several attention, or memory related tasks, there was 19 across the board an increase in alpha synchrony (coherence) in brain areas deemed task relevant (Halgren et al., 2002; Gail et al., 2004; Hanslmayr et al., 2005). Others support this idea, with evidence of inhibition/excitation in the firing rate of recorded neurons during specific phases of the alpha oscillation (See Van Diepen et al., 2019 for a review). A consensus on these two hypotheses has yet to be determined, however having access to a large-scale recordings during a visual short-term memory may help further evidence in one direction or the other. Finally as with theta, alpha oscillations have also been implicated in facilitating long range communication between brain regions (Von Stein and Sarnthein, 2000; Sauseng et al., 2005) either through synchronization of spiking activity (Chapeton et al., 2019) or through multiplexing type mechanism with higher frequency oscillations being nested within alpha oscillations (Bonnefond et al., 2017). One potential confound is that although there is a rich literature of the study of alpha oscillations measured from the surface of the scalp (EEG-alpha), less is known if this phenomena is analogous to the intracortically measured alpha oscillation (LFP-alpha). Thus, proposed roles of the EEG-alpha may not be applicable to LFP-alpha, and the two phenomena may only be similar in that they share a similar range of frequencies. Clinically abnormalities (mainly reduction) in alpha oscillations have also been implicated in a number of pathologies including a reduction in alpha coherence in Alzheimer’s Disease (Hogan et al., 2003), reduction in alpha power in Bipolar Disorder (Basar et al., 2012), and reduced phase locking in Schizophrenia (Brockhaus-Dumke et al., 2008). 20 Beta Oscillations (14-30 Hz) Beta oscillations were initially described alongside alpha oscillations by Berger (Berger, 1929). Early on their widespread nature and relation to sensorimotor cortices was shown in early ECoG experiments in human patients, and beta activity was even suggested as a biomarker for these regions of cortex (Jasper and Penfield 1949). The decrease in beta power, or even the cessation of beta activity, often termed movement-related beta desynchronization (MRBD) is a well-known phenomenon that is observed just prior to and during motor movements, followed by a “rebound” of power upon movement cessation (Pfurtscheller, 1977; Pfurtscheller and Aranibar, 1979; Pfurtscheller and Da Silva, 1999). It is from these studies that one notion of the role of beta activity is that of cortical (in this case sensorimotor cortex) “idling”. Similar to this is that beta activity is not actually that of idling, but is a reflection of an active, ongoing process (Engel and Fries, 2010). Beta oscillations have been implicated to dynamically work with alpha and gamma oscillations, in the framework of the “communication through coherence” hypothesis, where the beta oscillations provide top-down control or a feedback function to complement the feedforward function of gamma (Fries, 2005; Buschman and Miller, 2007; Siegel et al., 2012; Bastos et al., 2015; Fries 2015). Contrary to the MRBD, which is a decrease in beta activity, it has been shown in humans performing a visual working memory task that there is an increases in occipital beta during memory maintenance (Tallon-Baudry et al., 1998). In Monkeys elevated delay period activity has been observed in ventral PFC, and lateral PFC (Pipa et al., 2009; Siegel et al., 2009). Intra areal beta LFP synchrony in the temporal lobe has been shown to be implicated in during memory maintenance (Tallon-Baudry et al., 2004), and in 21 sensorimotor areas beta has been implicated in linking and directing information flow between these areas (Brovelli et al., 2004). Beta LFP power has also been shown to provide stimulus motion direction better than that of spiking activity the motion selective middle temporal (MT) area (Mendoza-Halliday et al., 2014). This finding of a difference in motion discriminability between the LFP and spikes has been implicated into the notion of the role of beta activity providing a top-down signal from higher order areas (namely PFC). Interestingly a number of studies have shown that beta oscillatory synchronization appear to be not only task dependent but content specific, in that they show synchronization that is specific to stimulus (Salazar et al., 2012), or reflects the rules of the cognitive task (Buschman et al., 2012). Further, it has been shown that this correlation structure (the relative phase) is highly dynamic and can rapidly transition from in phase to antiphase (Dotson et al., 2014). These somewhat dichotomous findings paint the picture for perhaps two roles of beta activity, that which is associated with sensorimotor tasks, and that which is associated with cognitive processes. Gamma Oscillations (30-90 Hz) Although first very briefly described in humans in the late 1930’s, (Jasper and Andrews, 1938) then subsequently shown in response to olfactory stimuli in hedgehogs (Adrian, 1942) research on gamma oscillations largely plateaued until a resurgence (and quite the resurgence it was) in the 1980s. This resurgence was the result of the works of Walter Freeman, Steven Bressler, Wolf Singer and some other guy who’s name escapes me at the moment6…(Bressler and Freeman, 1980; Gray and Singer, 1989). First described in the olfactory system, gamma activity was found to 6 Ah, It’s just come to me: Charles Gray. 22 display odor specific spatial patterns of activity (Bressler and Freeman, 1980; Skarda and Freeman, 1987). This theory of olfactory perception allowing for specific odors to be represented in the amplitude patterns (Bressler, 1984) was next explored in the visual domain by recording from arrays in visual cortex of monkeys (Freeman and van Dijk, 1987). These findings were advanced by the work of Gray and Singer with the discovery of synchronization of the phase of gamma oscillations to that of the spikes in the visual cortex of cats (Gray and Singer, 1987; Gray and Singer, 1989; corroborated by Eckhorn et al., 1988). The synchronization occurred between sites separated by several millimeters with similar receptive fields when stimulated by light bars moving in the same direction (Gray et al., 1989). The effect was found to be enhanced by a single light bar encompassing both receptive fields, and was diminished if the light bars were moving in opposite directions. The relevance of these studies vaulted gamma into the spotlight (evidenced by the slew of subsequent contemporaneous studies; Engel et al., 1991a; Engel et al., 1991b; Engel et al., 1991c; Murthy and Fetz, 1992) by not only demonstrating the synchronization of spikes in the visual cortex to that of the phase of the gamma oscillation, but when taken with respect to their implications to theories of cell assemblies (Hebb, 1949; Milner 1974; Edelman and Mountcastle, 1978; von der Malsburg, 1985). This finding allowed for a mechanistic answer to the “Binding Problem” of coordinated activity between cell assemblies, and was followed by the “Temporal correlation hypothesis”, and the “binding by synchrony” proposal (Singer and Gray, 1995; Gray, 1999; Singer, 1999). Importantly these works provided a link between neuronal unit activity and field potentials, giving credence to the EEG and MEG communities. Although focused on visual perception, in 23 essence this would allow disparate cell assemblies, with different functions, to integrate their information. The temporal synchronization will increase the efficacy of the outputs from these cell assemblies on their downstream targets (Konig et al., 1996). These findings have also been interpreted as a mechanism of information transfer (Bressler, 1990), which today resides as the “communication through coherence” (CTC) hypothesis (Womelsdorf 2007; Fries, 2005; Fries, 2015). The CTC posits that the phase of the LFP determines the efficacy of the pre-synaptic spikes, where higher receptivity post-synaptically occurs during the receptive phase of the LFP (Haider and McCormick 2009). Aside from these studies gamma activity continues to be the topic of research, with the focus generally being on synchronization (coherence or phase locking). Elevated delay period gamma power and synchrony have been observed in the lateral interparietal area (LIP) for preferred saccade location (Pesaran et al., 2002). Similarly, but in the beta and gamma frequencies elevated delay period activity has been observed in vPFC and lPFC (Pipa et al., 2009; Siegel et al., 2009), and in transient form in both encoding and retrieval in a color matching task (Lundqvist et al., 2016). In humans gamma activity has been shown to be stimulus specific (Tallon-Baudry et al., 1996), associated with working memory load (Howard et al., 2003), and in visual motion selective areas have been shown to be modulated with motion stimuli strength (Siegel et al., 2007). Methodological and Experimental Considerations in Large-Scale Neurophysiology Here I will give a general overview of the work that goes into conducting a study of large-scale neural dynamics during a cognitive task, specifically in an animal model. The details of the development and implementation of the large-scale recording device 24 will be omitted here as they are the focus of Chapter 2. Although we use working memory constantly while awake and conscious (as you are using it right now reading this sentence), studying it in the laboratory requires the elimination of as many confounding variables as possible. In this way researchers are trying to limit any observed physiological phenomena to those having a relation to the memory task. The elimination of all confounds is ultimately futile, and often their discovery only comes in the analysis phase of the experiment. An example of this is arises in Chapter 3, where we realized that neuronal firing due to very small eye movements (microsaccades) needed to be accounted for. Due to technical constraints, often animal research in this field utilizes stimuli that are presented visually, where aspects such as their precise site, shape, and luminance can be accounted for. Although depending on the focus other stimuli may utilized (e.g. auditory, olfactory, tactile; as is often done in human studies). In NHP research the animal is often head fixated in order to eliminate movement related confounds, as well as to position the animals gaze optimally. Generally, the animal is presented (via computer monitor) one or a few items to be remembered, followed by a delay period typically ranging from half a second to a few seconds. After this delay period comes a test or match period where the animal must indicate the corresponding item previously presented via directing their gaze towards it. While asking a human to perform this task would be trivial, achieving this in animals can be quite the feat. Training can take years, where complexity is slowly added in an iterative form. The human experimenter may have a theory on what strategy the animal is implementing to perform the task, but there is rarely any way of elucidating this. While 25 obtaining proficiency is the goal of the training, this itself can be a negative, as often times the animals become so proficient in the trained task, that there is little availability of behavioral analysis for aspects such as errors. While all these efforts to reduce confounds allow a focus on the neurophysiological variable of choice, they ultimately result in a set-up that is completely unnatural and far removed from what occurs in the real world. As far as I know the observation of a macaque monkey playing a computer- based memory game is not part of the normal behavioral repertoire in the wild... When studying cognitive processes, such as visual short-term memory, many aspects need to be addressed. These include which brain areas are believed to be involved, what is the modality being studied, is it spiking activity or field potentials, or maybe both, and the dynamics of these brain areas. This could be modulations in firing rate compared to a baseline, modulations in spectral power, or even the coordinated dynamics between brain areas. It is widely believed that many brain regions are involved, even in relatively simple cognitive tasks (Mountcastle, 1978; Mesulam, 1990; Bressler 1995; Fuster and Bressler, 2012; Varela 2001; Siegel et al., 2012) such as visual short- term memory. It is therefore why we utilized a large-scale device allowing for great coverage across many cortical areas, but with the resolution to allow the recording of both spiking activity and field potentials. Conclusions As the previous sections have shown, oscillatory activity in all frequency bands has been implicated in a variety of attentional, sensorimotor, and cognitive processes (something which was proposed nearly 30 years ago if not earlier; Lopes da Silva, 1991) 26 as well as with homeostatic drives and sleep. Indeed, entire books have been written on the subject (Walter 1953; 1963; Koch 2004; Buzsaki 2006). It still stands that the chemical synapse is the essential mechanism for interaction between neurons, yet it has been shown that oscillations play an important role in neuronal function as well. A more comprehensive role of oscillations may be that they fall into a hierarchical organization. Slower oscillations having a longer period length and thus a larger timescale of interaction relating to long range interactions and general processes. While faster oscillations being involved in coordination more locally with faster timescale of interaction (Engel et al., 2010). While these myriad of functions have been attributed to oscillations, the field still (especially in the case of intracortical; LFP oscillations) lacks a basic understanding of how they are spatially distributed across cortex. If for example oscillations at a particular frequency are responsible for a communicative role in the brain, one would expect oscillations to be at least present in the interacting areas prior to investigating the nature of the interaction. Specific Aims In order to elucidate the broad-spanning coordinated neural dynamics involved in even simple tasks we must be able similarly record activity in a broad-spanning synchronous way. Large-scale microelectrode recordings provide both high spatiotemporal resolution and high coverage in order to answer these questions. The specific aims of this study are to elucidate the technological leaps we have made in addressing this issue, as 27 well as offer initial analyses during a visual short-term memory task from the resultant data provided in large-scale recordings. Specific Aim 1: Large-Scale Neural Recording To develop, construct, and implement a Large-Scale Device allowing the recording of neural activity. Sub-aims: 1) To record intracortical activity spanning the breadth and depth of a cortical hemisphere. 2) To accurately determine anatomical origin of recordings. Specific Aim 2: Widespread Dynamics During Visual Short-Term Memory To test the hypothesis that visual short-term memory is reflected in wide-spread dynamics of neuronal firing. Sub-aims: 1) Describe the overall activity profile of firing rate modulation resolves into a balanced state. 2) Describe that stimulus specific changes in firing rate are reflected in a subset of cortical areas that pertain to a cortical hierarchy. Specific Aim 3: Spatial Distribution of Oscillatory Activity To characterize the spatial patterns of oscillatory activity across cortex. Sub-aims: 1) To determine the predominant frequencies at which oscillatory activity is observed. 2) To determine the spatial distribution of oscillatory activity. 3) To utilize these defined spectral patterns to show that cortical areas display specific oscillatory profiles both spatially and in a task dependent manner. 28 Figures Figure 1. First reported photograph of electrical brain activity. First reported photograph of electrical activity recorded from the motor cortex of a dog. Timeline on the top in 1/5 sec. Adopted from Pravdich-Neminsky 1913a. 29 Figure 2. Early Recording Device. Original caption: “Electrodes in adjustable holder attached to bone clamp as used for electrocoricograms in man. The entire electrode assembly with its cable may be sterilized by the autoclave.” From Jasper and Penfield 1949. The device allowed them to measure from 8 disparate sites across the surface of a hemisphere. 30 Figure 3. Early Recordings. Original caption: “Sample electrocorticograms showing characteristic spontaneous electrical activity from specific normal cortical areas in the unanaesthetized human subject.” From Jasper and Penfield 1949. Recordings were made from the device shown in Figure 2. Note the variability in oscillations from frontal to occipital cortex. 31 Figure 4. Popularity of Oscillations and Brain. PubMED search result for the number of manuscripts containing the words “Oscillations” and “Brain” From 1945 until present. 32 References Adey, W.R., 1988. Do EEG-like processes influence brain function at a physiological level?. In Dynamics of Sensory and Cognitive Processing by the Brain (pp. 362- 367). Springer, Berlin, Heidelberg. Adrian, E.D. and Moruzzi, G., 1939. Impulses in the pyramidal tract. The Journal of physiology, 97(2), pp. 153-199. Adrian, E.D., 1942. Olfactory reactions in the brain of the hedgehog. The Journal of physiology, 100(4), p.459. Amzica, F. and Steriade, M., 1998. Electrophysiological correlates of sleep delta waves. Electroencephalography and clinical neurophysiology, 107(2), pp.69-83. Babiloni, C., Vecchio, F., Bultrini, A., Luca Romani, G. and Rossini, P.M., 2006. Pre-and poststimulus alpha rhythms are related to conscious visual perception: a high- resolution EEG study. Cerebral cortex, 16(12), pp.1690-1700. Baddeley, A., 1992. Working memory. Science, 255(5044), pp.556-559. Başar, E., Güntekin, B., Atagün, I., Gölbaşı, B.T., Tülay, E. and Özerdem, A., 2012. Brain’s alpha activity is highly reduced in euthymic bipolar disorder patients. Cognitive neurodynamics, 6(1), pp.11-20. Başar, E., 2012. A review of alpha activity in integrative brain function: fundamental physiology, sensory coding, cognition and pathology. International Journal of Psychophysiology, 86(1), pp.1-24. Bastos, A.M., Vezoli, J., Bosman, C.A., Schoffelen, J.M., Oostenveld, R., Dowdall, J.R., De Weerd, P., Kennedy, H. and Fries, P., 2015. Visual areas exert feedforward and feedback influences through distinct frequency channels. Neuron, 85(2), pp.390-401. Beck, A., 1891. Oznaczenie lokalizacyi w mózgu i rdzeniu za pomoczjawisk elektrycznych [The determination of localizations in the brain and spinal cord with the aid of electrical phenomena]. Rozprawy Akademii Umiejętności, wydział matematyczno-przyrodniczy II(1): 187–232. Benchenane, K., Peyrache, A., Khamassi, M., Tierney, P.L., Gioanni, Y., Battaglia, F.P. and Wiener, S.I., 2010. Coherent theta oscillations and reorganization of spike timing in the hippocampal-prefrontal network upon learning. Neuron, 66(6), pp.921-936. 33 Berger, H., 1929. Über das elektroenkephalogramm des menschen. Archiv für psychiatrie und nervenkrankheiten, 87(1), pp.527-570. Berry, R.B., Brooks, R., Gamaldo, C.E., Harding, S.M., Marcus, C. and Vaughn, B.V., 2012. The AASM manual for the scoring of sleep and associated events. Rules, Terminology and Technical Specifications, Darien, Illinois, American Academy of Sleep Medicine, 176, p.2012. Beul, S.F., Barbas, H. and Hilgetag, C.C., 2017. A predictive structural model of the primate connectome. Scientific reports, 7(1), pp.1-12. Binnie, C.D. and Stefan, H., 1999. Modern electroencephalography: its role in epilepsy management. Clinical Neurophysiology, 110(10), pp.1671-1697. Bishop, G.H., 1936, January. The interpretation of cortical potentials. In Cold Spring Harbor Symposia on Quantitative Biology (Vol. 4, pp. 305-319). Cold Spring Harbor Laboratory Press. Bonnefond, M., Kastner, S. and Jensen, O., 2017. Communication between brain areas based on nested oscillations. eneuro, 4(2). Brazier, M.A., 1961. A history of the electrical activity of the brain: The first half- century. Brazier, M.A., 1984. Pioneers in the discovery of evoked potentials. Electroencephalography and Clinical Neurophysiology/Evoked Potentials Section, 59(1), pp.2-8. Bremer, F., 1949. Considérations sur l'origine et la nature des⪡ ondes⪢ cérébrales. Electroencephalography and clinical neurophysiology, 1(1-4), pp.177- 193. Bressler, S.L., 1984. Spatial organization of EEGs from olfactory bulb and cortex. Electroencephalography and clinical neurophysiology, 57(3), pp.270-276. Bressler, S.L., 1995. Large-scale cortical networks and cognition. Brain Research Reviews, 20(3), pp.288-304. Bressler, S.L. and Freeman, W.J., 1980. Frequency analysis of olfactory system EEG in cat, rabbit, and rat. Electroencephalography and clinical neurophysiology, 50(1- 2), pp.19-24. Bressler, S.L. and Kelso, J.S., 2001. Cortical coordination dynamics and cognition. Trends in cognitive sciences, 5(1), pp.26-36. 34 Brockhaus-Dumke, A., Mueller, R., Faigle, U. and Klosterkoetter, J., 2008. Sensory gating revisited: relation between brain oscillations and auditory evoked potentials in schizophrenia. Schizophrenia research, 99(1-3), pp.238-249. Brovelli, A., Ding, M., Ledberg, A., Chen, Y., Nakamura, R. and Bressler, S.L., 2004. Beta oscillations in a large-scale sensorimotor cortical network: directional influences revealed by Granger causality. Proceedings of the National Academy of Sciences, 101(26), pp.9849-9854. Brown, E.N., Lydic, R. and Schiff, N.D., 2010. General anesthesia, sleep, and coma. New England Journal of Medicine, 363(27), pp.2638-2650. Bullock, T.H., 1945. Problems in the comparative study of brain waves. The Yale journal of biology and medicine, 17(5), p.657. Bullock, T.H., 1997. Signals and signs in the nervous system: the dynamic anatomy of electrical activity is probably information-rich. Proceedings of the National Academy of Sciences, 94(1), pp.1-6. Bullock, T .H., 2002. Biology of Brain waves: natural history and evolution of an information-rich sign of activity. In Advances in Electrophysiolgy in Clinical Practice and Research. Kjellberg, Inc.. Buschman, T.J., Denovellis, E.L., Diogo, C., Bullock, D. and Miller, E.K., 2012. Synchronous oscillatory neural ensembles for rules in the prefrontal cortex. Neuron, 76(4), pp.838-846. Buzsáki, G. and Draguhn, A., 2004. Neuronal oscillations in cortical networks. science, 304(5679), pp.1926-1929. Buzsaki, G., 2006. Rhythms of the Brain. Oxford University Press. Buzsáki, G., Anastassiou, C.A. and Koch, C., 2012. The origin of extracellular fields and currents—EEG, ECoG, LFP and spikes. Nature reviews neuroscience, 13(6), pp.407-420. Caton, R., 1875. The electric currents of the brain. Br Med J, 2, p.278. Caton, R., 1877. Interim report on investigations of the electric currents of the brain. Br Med J, Suppl. L: 62-65. 35 Chapeton, J.I., Haque, R., Wittig Jr, J.H., Inati, S.K. and Zaghloul, K.A., 2019. Large- Scale communication in the human brain Is rhythmically modulated through alpha coherence. Current Biology, 29(17), pp.2801-2811. Colrain, I.M., Crowley, K.E., Nicholas, C.L., Afifi, L., Baker, F.C., Padilla, M., Turlington, S.R. and Trinder, J., 2010. Sleep evoked delta frequency responses show a linear decline in amplitude across the adult lifespan. Neurobiology of aging, 31(5), pp.874-883. Colrain, I.M., Turlington, S. and Baker, F.C., 2009. Impact of alcoholism on sleep architecture and EEG power spectra in men and women. Sleep, 32(10), pp.1341- 1352. Constantinidis, C. and Procyk, E., 2004. The primate working memory networks. Cognitive, Affective, & Behavioral Neuroscience, 4(4), pp.444-465. Courtney, S.M., Ungerleider, L.G., Keil, K. and Haxby, J.V., 1997. Transient and sustained activity in a distributed neural system for human working memory. Nature, 386(6625), pp.608-611. da Silva, F.L., 1991. Neural mechanisms underlying brain waves: from neural membranes to networks. Electroencephalography and clinical neurophysiology, 79(2), pp.81-93. da Silva, F.L., 2013. EEG and MEG: relevance to neuroscience. Neuron, 80(5), pp.1112- 1128. Danilevsky, V.Y., 1877. Investigations into the physiology of the brain. Doctoral Thesis, University of Kharkov. Dawson, G.D., 1951. A summation technique for detecting small signals in a large irregular background. J physiol, 115, p.2. Dotson, N.M., Goodell, B., Salazar, R.F., Hoffman, S.J. and Gray, C.M., 2015. Methods, caveats and the future of large-scale microelectrode recordings in the non-human primate. Frontiers in systems neuroscience, 9, p.149. Dotson, N.M., Hoffman, S.J., Goodell, B. and Gray, C.M., 2017. A large-scale semi- chronic microdrive recording system for non-human primates. Neuron, 96(4), pp.769-782. Dotson, N.M., Hoffman, S.J., Goodell, B. and Gray, C.M., 2018. Feature-based visual short-term memory is widely distributed and hierarchically organized. Neuron, 99(1), pp.215-226. 36 Dotson, N.M., Salazar, R.F. and Gray, C.M., 2014. Frontoparietal correlation dynamics reveal interplay between integration and segregation during visual working memory. Journal of Neuroscience, 34(41), pp.13600-13613. Dubbelink, K.T.O., Felius, A., Verbunt, J.P., Van Dijk, B.W., Berendse, H.W., Stam, C.J. and Delemarre-van de Waal, H.A., 2008. Increased resting-state functional connectivity in obese adolescents; a magnetoencephalographic pilot study. PLoS One, 3(7), p.e2827. Du Bois-Reymond, E., 1848. Untersuchungen über thierische Elektricität, Berlin, Reimer, Vol. I. Eckhorn, R., Bauer, R., Jordan, W., Brosch, M., Kruse, W., Munk, M. and Reitboeck, H.J., 1988. Coherent oscillations: A mechanism of feature linking in the visual cortex?. Biological cybernetics, 60(2), pp.121-130. Edelman, G.M. and Mountcastle, V.B., 1978. The mindful brain: cortical organization and the group-selective theory of higher brain function. Massachusetts Inst of Technology Pr. Einevoll, G.T., Kayser, C., Logothetis, N.K. and Panzeri, S., 2013. Modelling and analysis of local field potentials for studying the function of cortical circuits. Nature Reviews Neuroscience, 14(11), pp.770-785. Engel, A.K. and Fries, P., 2010. Beta-band oscillations—signalling the status quo?. Current opinion in neurobiology, 20(2), pp.156-165. Engel, A.K., Friston, K., Kelso, J.A., König, P., Kovács, I., MacDonald III, A., Miller, E.K., Phillips, W.A., Silverstein, S.M., Tallon-Baudry, C. and Triesch, J., 2010. Coordination in behavior and cognition. Engel, A.K., König, P., Kreiter, A.K. and Singer, W., 1991a. Interhemispheric synchronization of oscillatory neuronal responses in cat visual cortex. Science, pp.1177-1179. Engel, A.K., König, P., Kreiter, A.K. and Singer, W., 1991b. Interhemispheric synchronization of oscillatory neuronal responses in cat visual cortex. Science, pp.1177-1179. Engel, A.K., Kreiter, A.K., König, P. and Singer, W., 1991. Synchronization of oscillatory neuronal responses between striate and extrastriate visual cortical areas of the cat. Proceedings of the National Academy of Sciences, 88(14), pp.6048- 6052. 37 Fell, J., Klaver, P., Lehnertz, K., Grunwald, T., Schaller, C., Elger, C.E. and Fernández, G., 2001. Human memory formation is accompanied by rhinal–hippocampal coupling and decoupling. Nature neuroscience, 4(12), pp.1259-1264. Fernandez, T., Harmony, T., Gersenowies, J., Silva-Pereyra, J., Fernández-Bouzas, A., Galán, L. and Díaz-Comas, L., 2002. Sources of EEG activity during a verbal working memory task in adults and children. In Supplements to Clinical Neurophysiology (Vol. 54, pp. 269-283). Elsevier. Ferrier, D., 1874. The localization of function in the brain. Proceedings of the Royal Society of London, 22(148-155), pp.228-232. Fleischl von Marxow, E., 1890. Mittheilung, betreffend die Physiologie der Hirnrinde. Centralblatt fur Physiologie, 4, pp.537-540. Freeman, W.J., 1975. Mass action in the nervous system (Vol. 2004). Academic Press, New York. Freeman, W.J., 2007. Definitions of state variables and state space for brain-computer interface. Cognitive Neurodynamics, 1(1), pp.3-14. Freeman, W.J. and van Dijk, B.W., 1987. Spatial patterns of visual cortical fast EEG during conditioned reflex in a rhesus monkey. Brain research, 422(2), pp.267- 276. Fries, P., 2005. A mechanism for cognitive dynamics: neuronal communication through neuronal coherence. Trends in cognitive sciences, 9(10), pp.474-480. Fries, P., 2015. Rhythms for cognition: communication through coherence. Neuron, 88(1), pp.220-235. Friston, K.J., 1997. Another neural code?. Neuroimage, 5(3), pp.213-220. Fuster, J.M. and Alexander, G.E., 1971. Neuron activity related to short-term memory. Science, 173(3997), pp.652-654. Fuster, J.M. and Bressler, S.L., 2012. Cognit activation: a mechanism enabling temporal integration in working memory. Trends in cognitive sciences, 16(4), pp.207-218. Galvani, L., 1791. De viribus electricitatis in motu musculari: Commentarius. Bologna: Tip. Istituto delle Scienze, 1791; 58 p.: 4 tavv. ft; in 4.; DCC. f. 70. 38 Gonzalez, S.L., de Peralta, R.G., Thut, G., Millán, J.D.R., Morier, P. and Landis, T., 2006. Very high frequency oscillations (VHFO) as a predictor of movement intentions. Neuroimage, 32(1), pp.170-179. Grastyan, E., Lissak, K., Madarasz, I. and Donhoffer, H., 1959. Hippocampal electrical activity during the development of conditioned reflexes. Electroencephalography and clinical neurophysiology, 11(3), pp.409-430. Gray, C. M. & Singer, W. (1987) Soc. Neurosci. Abstr. 404.3. Gray, C.M. and Singer, W., 1989. Stimulus-specific neuronal oscillations in orientation columns of cat visual cortex. Proceedings of the National Academy of Sciences, 86(5), pp.1698-1702. Gray, C.M., 1994. Synchronous oscillations in neuronal systems: mechanisms and functions. Journal of computational neuroscience, 1(1-2), pp.11-38. Gray, C.M., 1999. The temporal correlation hypothesis of visual feature integration: still alive and well. Neuron, 24(1), pp.31-47. Grey Walter, W., 1963. The living brain. Haider, B. and McCormick, D.A., 2009. Rapid neocortical dynamics: cellular and network mechanisms. Neuron, 62(2), pp.171-189. Harmony, T., 2013. The functional significance of delta oscillations in cognitive processing. Frontiers in integrative neuroscience, 7, p.83. Harmony, T., Fernández, T., Silva, J., Bernal, J., Díaz-Comas, L., Reyes, A., Marosi, E., Rodríguez, M. and Rodríguez, M., 1996. EEG delta activity: an indicator of attention to internal processing during performance of mental tasks. International journal of psychophysiology, 24(1-2), pp.161-171. Harrison, S.A. and Tong, F., 2009. Decoding reveals the contents of visual working memory in early visual areas. Nature, 458(7238), pp.632-635. Hartline, H.K., 1940. The receptive fields of optic nerve fibers. American Journal of Physiology-Legacy Content, 130(4), pp.690-699. Hebb, D.O., 1949. John Wiley & Sons; 1949. The organization of behavior: A neuropsychological approach. Herreras, O., 2016. Local field potentials: myths and misunderstandings. Frontiers in neural circuits, 10, p.101. 39 Helmholtz, H., 1850. Über die Fortpflanzungsgeschwindigkeit der Nervenreizung. Bericht über die Bekanntmachung geeigneten Verhandlungen der Königlichen Preussischen Akademie der Wissenschaften zu Berlin, 14, pp.14-15. Hodgkin, A.L. and Huxley, A.F., 1945. Resting and action potentials in single nerve fibres. The Journal of physiology, 104(2), p.176. Hodgkin, A.L. and Huxley, A.F., 1953. Movement of radioactive potassium and membrane current in a giant axon. The Journal of physiology, 121(2), p.403. Hodgkin, A.L., Huxley, A.F. and Katz, B., 1952. Measurement of current‐voltage relations in the membrane of the giant axon of Loligo. The Journal of physiology, 116(4), pp.424-448. Hogan, M.J., Swanwick, G.R., Kaiser, J., Rowan, M. and Lawlor, B., 2003. Memory- related EEG power and coherence reductions in mild Alzheimer's disease. International Journal of Psychophysiology, 49(2), pp.147-163. Hoerzer, G.M., Liebe, S., Schloegl, A., Logothetis, N.K. and Rainer, G., 2010. Directed coupling in local field potentials of macaque v4 during visual short-term memory revealed by multivariate autoregressive models. Frontiers in computational neuroscience, 4, p.14. Howard, M.W., Rizzuto, D.S., Caplan, J.B., Madsen, J.R., Lisman, J., Aschenbrenner- Scheibe, R., Schulze-Bonhage, A. and Kahana, M.J., 2003. Gamma oscillations correlate with working memory load in humans. Cerebral cortex, 13(12), pp.1369-1374. Hubel, D.H. and Wiesel, T.N., 1959. Receptive fields of single neurones in the cat's striate cortex. The Journal of physiology, 148(3), pp.574-591. Ito, J., Maldonado, P.E. and Grün, S., 2013. Cross-frequency interaction of the eye- movement related LFP signals in V1 of freely viewing monkeys. Frontiers in systems neuroscience, 7, p.1. Jacob, S.N., Hähnke, D. and Nieder, A., 2018. Structuring of abstract working memory content by fronto-parietal synchrony in primate cortex. Neuron, 99(3), pp. 588- 597. Jasper, H.H. and Andrews, H.L., 1938. Electro-encephalography: III. Normal differentiation of occipital and precentral regions in man. Archives of Neurology & Psychiatry, 39(1), pp.96-115. 40 Jasper, H. and Penfield, W., 1949. Electrocorticograms in man: effect of voluntary movement upon the electrical activity of the precentral gyrus. Archiv für Psychiatrie und Nervenkrankheiten, 183(1-2), pp.163-174. Jensen, O., Gelfand, J., Kounios, J. and Lisman, J.E., 2002. Oscillations in the alpha band (9–12 Hz) increase with memory load during retention in a short-term memory task. Cerebral cortex, 12(8), pp.877-882. Jensen, O. and Mazaheri, A., 2010. Shaping functional architecture by oscillatory alpha activity: gating by inhibition. Frontiers in human neuroscience, 4, p.186. Jensen, O. and Tesche, C.D., 2002. Frontal theta activity in humans increases with memory load in a working memory task. European journal of Neuroscience, 15(8), pp.1395-1399. Jung, R. and Kornmüller, A.E., 1938. Eine methodik der ableitung iokalisierter potentialschwankungen aus subcorticalen hirngebieten. Archiv für Psychiatrie und Nervenkrankheiten, 109(1), pp.1-30. Karakaş, S. and Barry, R.J., 2017. A brief historical perspective on the advent of brain oscillations in the biological and psychological disciplines. Neuroscience & Biobehavioral Reviews, 75, pp.335-347. Katzner, S., Nauhaus, I., Benucci, A., Bonin, V., Ringach, D.L. and Carandini, M., 2009. Local origin of field potentials in visual cortex. Neuron, 61(1), pp.35-41. Klimesch, W., 1999. EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis. Brain research reviews, 29(2-3), pp.169-195. Klimesch, W., Doppelmayr, M., Russegger, H., and Pachinger, T., 1996. Theta band power in the human scalp EEG and the encoding of new information. Neuroreport, 7, pp.1235-1240. Knyazev, G.G., 2007. Motivation, emotion, and their inhibitory control mirrored in brain oscillations. Neuroscience & Biobehavioral Reviews, 31(3), pp.377-395. Koch, C., 2004. The quest for consciousness. Engineering and Science, 67(2), pp.28-34. König, P., 1994. A method for the quantification of synchrony and oscillatory properties of neuronal activity. Journal of neuroscience methods, 54(1), pp.31-37 König, P., Engel, A.K. and Singer, W., 1996. Integrator or coincidence detector? The role of the cortical neuron revisited. Trends in neurosciences, 19(4), pp.130-137. 41 Lachaux, J.P., Rodriguez, E., Martinerie, J. and Varela, F.J., 1999. Measuring phase synchrony in brain signals. Human brain mapping, 8(4), pp.194-208. Lee, H., Simpson, G.V., Logothetis, N.K. and Rainer, G., 2005. Phase locking of single neuron activity to theta oscillations during working memory in monkey extrastriate visual cortex. Neuron, 45(1), pp.147-156. Le Van Quyen, M., Martinerie, J., Navarro, V., Boon, P., D'Havé, M., Adam, C., Renault, B., Varela, F. and Baulac, M., 2001. Anticipation of epileptic seizures from standard EEG recordings. The Lancet, 357(9251), pp.183-188. Lewicki MS. A review of methods for spike sorting: the detection and classification of neural action potentials. Network: Computation in Neural Systems. 1998 Jan 1;9(4):R53-78. Liebe, S., Hoerzer, G.M., Logothetis, N.K. and Rainer, G., 2012. Theta coupling between V4 and prefrontal cortex predicts visual short-term memory performance. Nature neuroscience, 15(3), p.456. Lilly, J.C., 1950. A method of recording the moving electrical potential gradients in the brain: the 25 channel bavatron and electroiconograms. In Conference on Electronic Instrumentation in Nucleonics and Medicine. New York: American Institute of Electronic Engineering (pp. 37-3). Lilly, J.C., Hughes, J.R., Alvord Jr, E.C. and Galkin, T.W., 1955. Brief, noninjurious electric waveform for stimulation of the brain. Science. Lilly, J.C., 1958. Electrode and cannulae implantation in the brain by a simple percutaneous method. Science. Lilly, J.C., 1997. The Scientist: A metaphysical autobiography. Ronin Publishing. Lundqvist, M., Rose, J., Herman, P., Brincat, S.L., Buschman, T.J. and Miller, E.K., 2016. Gamma and beta bursts underlie working memory. Neuron, 90(1), pp.152- 164. Marshall, W.H., Woolsey, C.N. and Bard, P., 1937. Cortical representation of tactile sensibility as indicated by cortical potentials. Science, 85(2207), pp.388-390. Mathalon, D.H. and Sohal, V.S., 2015. Neural oscillations and synchrony in brain dysfunction and neuropsychiatric disorders: it’s about time. JAMA psychiatry, 72(8), pp.840-844. Maurer, K. and Dierks, T., 2012. Atlas of brain mapping: topographic mapping of EEG and evoked potentials. Springer Science & Business Media. 42 Mendoza-Halliday, D., Torres, S. and Martinez-Trujillo, J.C., 2014. Sharp emergence of feature-selective sustained activity along the dorsal visual pathway. Nature neuroscience, 17(9), p. 1255. Mesulam, M.M., 1990. Large‐scale neurocognitive networks and distributed processing for attention, language, and memory. Annals of Neurology: Official Journal of the American Neurological Association and the Child Neurology Society, 28(5), pp.597-613. Milner, P.M., 1974. A model for visual shape recognition. Psychological review, 81(6), p.521. Mitzdorf, U., 1985. Current source-density method and application in cat cerebral cortex: investigation of evoked potentials and EEG phenomena. Physiological reviews, 65(1), pp.37-100. Mizumori, S.J.Y., Perez, G.M., Alvarado, M.C., Barnes, C.A. and McNaughton, B.L., 1990. Reversible inactivation of the medial septum differentially affects two forms of learning in rats. Brain research, 528(1), pp.12-20. Moran, R.J., Campo, P., Maestu, F., Reilly, R.B., Dolan, R.J. and Strange, B.A., 2010. Peak frequency in the theta and alpha bands correlates with human working memory capacity. Frontiers in human neuroscience, 4, p.200. Mountcastle, V., 1978. An organizing principle for cerebral function: the unit module and the distributed system. The mindful brain. Munk, M.H., Linden, D.E., Muckli, L., Lanfermann, H., Zanella, F.E., Singer, W. and Goebel, R., 2002. Distributed cortical systems in visual short-term memory revealed by event-related functional magnetic resonance imaging. Cerebral Cortex, 12(8), pp.866-876. Murthy, V.N. and Fetz, E.E., 1992. Coherent 25-to 35-Hz oscillations in the sensorimotor cortex of awake behaving monkeys. Proceedings of the National Academy of Sciences, 89(12), pp.5670-5674. Niedermeyer, E., 2005. The normal EEG of the waking adult. Electroencephalography: Basic principles, clinical applications, and related fields, 167, pp.155-164. Nunez, P.L., Srinivasan, R., Westdorp, A.F., Wijesinghe, R.S., Tucker, D.M., Silberstein, R.B. and Cadusch, P.J., 1997. EEG coherency: I: statistics, reference electrode, volume conduction, Laplacians, cortical imaging, and interpretation at multiple 43 scales. Electroencephalography and clinical neurophysiology, 103(5), pp.499- 515. Nunez, P.L. and Srinivasan, R., 2006. Electric fields of the brain: the neurophysics of EEG. Oxford University Press, USA. O'Keefe, J. and Dostrovsky, J., 1971. The hippocampus as a spatial map: Preliminary evidence from unit activity in the freely-moving rat. Brain research. O'Keefe, J., 1976. Place units in the hippocampus of the freely moving rat. Experimental neurology, 51(1), pp.78-109. O'Keefe, J. and Recce, M.L., 1993. Phase relationship between hippocampal place units and the EEG theta rhythm. Hippocampus, 3(3), pp.317-330. Pesaran, B., Pezaris, J.S., Sahani, M., Mitra, P.P. and Andersen, R.A., 2002. Temporal structure in neuronal activity during working memory in macaque parietal cortex. Nature neuroscience, 5(8), pp.805-811. Pesaran, B., Vinck, M., Einevoll, G.T., Sirota, A., Fries, P., Siegel, M., Truccolo, W., Schroeder, C.E. and Srinivasan, R., 2018. Investigating large-scale brain dynamics using field potential recordings: analysis and interpretation. Nature neuroscience, 21(7), pp.903-919. Pfurtscheller, G., 1977. Graphical display and statistical evaluation of event-related desynchronization (ERD). Electroencephalography and clinical neurophysiology, 43(5), pp.757-760. Pfurtscheller G, Aranibar A. Evaluation of event-related desynchronization (ERD) preceding and following voluntary self-paced movement. Electroencephalography and clinical neurophysiology. 1979 Feb 1;46(2):138-46. Pfurtscheller, G. and Da Silva, F.L., 1999. Event-related EEG/MEG synchronization and desynchronization: basic principles. Clinical neurophysiology, 110(11), pp.1842- 1857. Pfurtscheller, G., Stancak Jr, A. and Neuper, C., 1996. Event-related synchronization (ERS) in the alpha band—an electrophysiological correlate of cortical idling: a review. International journal of psychophysiology, 24(1-2), pp.39-46. Piccolino, M., 1998. Animal electricity and the birth of electrophysiology: the legacy of Luigi Galvani. Brain research bulletin, 46(5), pp.381-407. 44 Pilon, M., Zadra, A., Joncas, S. and Montplaisir, J., 2006. Hypersynchronous delta waves and somnambulism: brain topography and effect of sleep deprivation. Sleep, 29(1), pp.77-84. Pipa, G., Städtler, E.S., Rodriguez, E.F., Waltz, J.A., Muckli, L., Singer, W., Goebel, R. and Munk, M.H., 2009. Performance-and stimulus-dependent oscillations in monkey prefrontal cortex during short-term memory. Frontiers in integrative neuroscience, 3, p.25. Postle, B.R., 2006. Working memory as an emergent property of the mind and brain. Neuroscience, 139(1), pp.23-38. Pravdich-Neminsky, V.V., 1913. Experiments on the registration of the electrical phenomena of the mammalian brain. Zbl. Physiol, 27, pp.951-960. Pravdich-Neminsky, V.V., 1913. Experiments on the registration of the electrical phenomena of the mammalian brain. Zbl. Physiol, 27, pp.951-960. Raghavachari, S., Kahana, M.J., Rizzuto, D.S., Caplan, J.B., Kirschen, M.P., Bourgeois, B., Madsen, J.R. and Lisman, J.E., 2001. Gating of human theta oscillations by a working memory task. Journal of Neuroscience, 21(9), pp.3175-3183. Ramón, S. and Cajal, Y., 1904. Textura del Sistema Nervioso del Hombre y de los Vertebrados. Imprenta y Librería de Nicolás Moya, Madrid (1899–1904). Riddle, J., Scimeca, J.M., Cellier, D., Dhanani, S. and D’Esposito, M., 2020. Causal evidence for a role of theta and alpha oscillations in the control of working memory. Current Biology. Roux, F. and Uhlhaas, P.J., 2014. Working memory and neural oscillations: alpha– gamma versus theta–gamma codes for distinct WM information?. Trends in cognitive sciences, 18(1), pp.16-25. Roux, F., Wibral, M., Singer, W., Aru, J. and Uhlhaas, P.J., 2013. The phase of thalamic alpha activity modulates cortical gamma-band activity: evidence from resting- state MEG recordings. Journal of Neuroscience, 33(45), pp.17827-17835. Salazar, R.F., Dotson, N.M., Bressler, S.L. and Gray, C.M., 2012. Content-specific fronto- parietal synchronization during visual working memory. Science, 338(6110), pp.1097-1100. Sauseng, P., Klimesch, W., Schabus, M. and Doppelmayr, M., 2005. Fronto-parietal EEG coherence in theta and upper alpha reflect central executive functions of working memory. International journal of Psychophysiology, 57(2), pp.97-103. 45 Schroeder, C.E. and Lakatos, P., 2009. Low-frequency neuronal oscillations as instruments of sensory selection. Trends in neurosciences, 32(1), pp.9-18. Siegel, M., Donner, T.H. and Engel, A.K., 2012. Spectral fingerprints of large-scale neuronal interactions. Nature Reviews Neuroscience, 13(2), pp.121-134. Siegel, M., Donner, T.H., Oostenveld, R., Fries, P. and Engel, A.K., 2007. High- frequency activity in human visual cortex is modulated by visual motion strength. Cerebral Cortex, 17(3), pp.732-741. Siegel, M., Warden, M.R. and Miller, E.K., 2009. Phase-dependent neuronal coding of objects in short-term memory. Proceedings of the National Academy of Sciences, 106(50), pp.21341-21346. Singer, W. and Gray, C.M., 1995. Visual feature integration and the temporal correlation hypothesis. Annual review of neuroscience, 18(1), pp.555-586. Skarda, C.A. and Freeman, W.J., 1987. How brains make chaos in order to make sense of the world. Behavioral and brain sciences, 10(2), pp.161-173. Steriade, M. and McCarley, R.W., 2005. Synchronized brain oscillations leading to neuronal plasticity during waking and sleep states. Brain control of wakefulness and sleep, pp.255-344. Steriade, M., McCormick, D.A. and Sejnowski, T.J., 1993b. Thalamocortical oscillations in the sleeping and aroused brain. Science, 262(5134), pp.679-685. Steriade, M., Nuñez, A. and Amzica, F., 1993a. Intracellular analysis of relations between the slow (< 1 Hz) neocortical oscillation and other sleep rhythms of the electroencephalogram. Journal of Neuroscience, 13(8), pp.3266-3283. Sutton, S., Braren, M., Zubin, J. and John, E.R., 1965. Evoked-potential correlates of stimulus uncertainty. Science, 150(3700), pp.1187-1188. Tallon-Baudry, C., Bertrand, O., Delpuech, C. and Pernier, J., 1996. Stimulus specificity of phase-locked and non-phase-locked 40 Hz visual responses in human. Journal of Neuroscience, 16(13), pp.4240-4249. Tallon-Baudry, C., Bertrand, O., Peronnet, F. and Pernier, J., 1998. Induced γ-band activity during the delay of a visual short-term memory task in humans. Journal of Neuroscience, 18(11), pp.4244-4254. 46 Tallon-Baudry, C., Mandon, S., Freiwald, W.A. and Kreiter, A.K., 2004. Oscillatory synchrony in the monkey temporal lobe correlates with performance in a visual short-term memory task. Cerebral cortex, 14(7), pp.713-720. Tuladhar, A.M., Huurne, N.T., Schoffelen, J.M., Maris, E., Oostenveld, R. and Jensen, O., 2007. Parieto‐occipital sources account for the increase in alpha activity with working memory load. Human brain mapping, 28(8), pp.785-792. Von Der Malsburg, C., 1985. Nervous structures with dynamical links. Berichte der Bunsengesellschaft für physikalische Chemie, 89(6), pp.703-710. Van Diepen, R., Foxe, J.J. and Mazaheri, A., 2019. The functional role of alpha-band activity in attentional processing: the current zeitgeist and future outlook. Current opinion in psychology. Von Stein, A. and Sarnthein, J., 2000. Different frequencies for different scales of cortical integration: from local gamma to long range alpha/theta synchronization. International journal of psychophysiology, 38(3), pp.301-313. Wang, X.J., 2010. Neurophysiological and computational principles of cortical rhythms in cognition. Physiological reviews, 90(3), pp.1195-1268. Walter, W.G., 1953. The living brain. Winson, J., 1978. Loss of hippocampal theta rhythm results in spatial memory deficit in the rat. Science, 201(4351), pp.160-163. Womelsdorf, T., Schoffelen, J.M., Oostenveld, R., Singer, W., Desimone, R., Engel, A.K. and Fries, P., 2007. Modulation of neuronal interactions through neuronal synchronization. science, 316(5831), pp.1609-1612. 47 CHAPTER TWO A LARGE-SCALE SEMI-CHRONIC MICRODRIVE RECORDING SYSTEM FOR NON-HUMAN PRIMATES Contribution of Authors and Co-Authors Manuscript in Chapter 2 Author: Nicholas M. Dotson Contributions: study design, data collection, data analysis, and preparing manuscript. Author: Steven J. Hoffman Contributions: data collection, data analysis, and editing manuscript. Co-Author: Baldwin Goodell Contributions: technology design and fabrication, and editing manuscript Co-Author: Charles M. Gray Contributions: study design, data collection, and preparing manuscript 48 Manuscript Information Page Nicholas M. Dotson, Steven J. Hoffman, Baldwin Goodell, Charles M. Gray Neuron Status of Manuscript: ____ Prepared for submission to a peer-reviewed journal ____ Officially submitted to a peer-review journal ____ Accepted by a peer-reviewed journal __X__ Published in a peer-reviewed journal CellPress November 2017 Issue 96 49 Summary Multi-electrode recordings in the non-human primate provide a critical method for measuring the widely distributed activity patterns that underlie brain function. However, common techniques rely on small, often immovable arrays, or microdrives that are only capable of manipulating a small number of closely spaced probes. These techniques restrict the number of cortical areas that can be simultaneously sampled and are typically not capable of reaching subcortical targets. To overcome these limitations, we developed a large-scale, semi-chronic microdrive recording system with up to 256 independently movable microelectrodes spanning an entire cerebral hemisphere. The microdrive system is hermetically sealed, free of internal connecting wires, and has been used to simultaneously record from up to 37 cortical and subcortical areas in awake behaving monkeys for up to 9 months. As a proof-of-principle, we demonstrate the capability of this technique to address network level questions using a graph theoretic analysis of functional connectivity data. Introduction Understanding how the concerted actions of large-scale neural circuits mediate cognitive functions is one of the major challenges in neuroscience. Conventional methods for monitoring neuronal activity in non-human primates usually involve acute recordings from one to several 10s of electrodes that are advanced into and then removed from the brain each day. The electrodes are typically advanced through the dural membrane or within guide tubes that are inserted at the same time (Mountcastle et al., 1991; Prut et al., 50 1998; Asaad et al., 2000; Purushothaman et al., 2006; Gray et al., 2007; Buschman and Miller, 2007; Miller and Wilson, 2008; Hernández et al., 2008). While this technique has been very effective, it requires bulky, often expensive, micromanipulators that involve long setup times and can increase the vulnerability of the animals to infection. Moreover, this approach precludes the possibility of recording from the same units over longer periods of time. These constraints have led to the development of chronically implanted arrays of microelectrodes. In these methods, groups of individual micro-wires (Fries et al., 1997; Roelfsema et al., 1997; Donoghue et al., 1998), arrays of rigidly linked micro-wires (Nicolelis et al., 1997, 2003; Musallam et al., 2007), or fabricated silicon electrode arrays (Nordhausen et al., 1996; Rousche et al., 2001; Maynard et al., 1999; Kipke et al., 2003; Vetter et al., 2004; Kim et al., 2006; Dickey et al., 2009) are chronically implanted. The implants are designed either to float with the pulsations of the brain, or be rigidly mounted to the skull. While useful for making long-term recordings, these methods also suffer from several limitations. Once the electrodes are in place, they can no longer be repositioned to isolate new activity. Thus, the experimenter is confined to studying only one set of neurons, and if neuronal activity on a particular electrode is not well isolated, or the signal is lost, that electrode can no longer yield useful data. It is also very difficult to access deep structures within the brain if large numbers of rigidly linked electrodes are being implanted. A second approach has utilized miniature microdrives that allow the implantation and positional control of large numbers of microelectrodes. These methods improve upon the static, irreversible implants, by retaining varying degrees of control over the positioning of 51 the probes. Some methods enable position control for each individual probe (Wilson and McNaughton, 1993; deCharms et al., 1999; Venkatachalam et al., 1999; Erickson and Desimone, 1999; Vos et al., 1999; Fee and Leonardo, 2001; Szabo et al., 2001; Swadlow et al., 2004; Tolias et al., 2007; Eliades and Wang, 2008; Battaglia et al., 2009; Kloosterman et al., 2009; Nguyen et al., 2009), while others employ single or multiple devices to position arrays of electrodes together (Csicsvari et al., 2003; Krupa et al., 2004; Schwarz et al., 2014; Mendoza et al., 2016). Both approaches have been effective, but in general, most of the microdrives have been designed to manipulate a relatively small number of closely spaced probes. This limits the number of neurons and brain regions that can be simultaneously recorded (see Hoffman and McNaughton, 2002, Feingold et al., 2012, and Schwarz et al., 2014 for notable exceptions). There are several obstacles to overcome in developing microdrive systems that can routinely record from large numbers of sites spanning widely distributed cortical and subcortical circuits. First, it becomes increasingly difficult to make large numbers of reliable electrical connections. In small systems, electrical connections are established by bonding or soldering flexible wires to each probe (e.g. Nguyen et al., 2009). The wires must have sufficient length to provide slack as the probe is moved. Increasing the number of probes leads to problems with packaging the many wires and the need for connectors of increasing size. Second, increasing the number of probes requires further miniaturization of the actuators. Without such improvements the microdrives become impractically large. A third obstacle is that there is no generally accepted method for introducing large numbers of independently moveable microelectrodes over widespread regions of the primate brain. 52 The hardware employed for rodents can be bulky and easily damaged (but see Tolias et al., 2007), and surgical placement, as well as care and maintenance, of the implant becomes difficult (Hoffman and McNaughton, 2002; Feingold et al., 2012). To overcome these obstacles, we developed a large-scale, semi-chronic microdrive recording system with miniature actuators and no internal connecting wires that spans an entire cerebral hemisphere (Gray and Goodell, 2007; Goodell and Gray 2008; Dotson et al., 2015). This system dramatically expands the number of cortical and subcortical areas that can be recorded simultaneously. We developed and tested two different versions of the system in two non-human primates trained to perform an object based delayed match-to- sample task. The first animal was implanted with a 256-channel system with 20 mm of travel for each electrode. Several revisions were made to the design of the recording system, and the second animal was implanted with a 252-channel system with actuators that permit 33 mm or 41 mm of electrode travel. Results Large-Scale Semi-Chronic Microdrive System Developing and implementing the large-scale recording system posed several significant challenges. A microdrive system was needed to house the requisite number of actuators, each capable of independent control of an electrode at high resolution, while keeping the overall size to a minimum. A compact and reliable method was needed for connecting to the electrodes while avoiding the use of loose wires. It was necessary to hermetically seal the microdrive to eliminate leakage of fluid into the device. An enclosure was required that could be firmly anchored to the animal’s skull while providing robust 53 protection against impacts and minimizing the risk of infection. And a method was needed to gain access to the brain over an area spanning a cerebral hemisphere. To meet these requirements, we designed two versions of a form-fitting chamber and microdrive system that was customized to each animal and that could be implanted and evaluated in stages (Fig. 1; Fig. S2). We first constructed an MRI-based 3D model of each monkey’s skull (Fig. 1A; Fig. S1). We used the resulting model along with the coronal sections of the brain to determine the boundaries of the chamber wall that would provide the greatest possible access to the left cerebral hemisphere while keeping the overall footprint to a minimum. The bottom contour of the chamber wall was matched to the cranial surface along these boundaries. This design facilitated surgical placement of the chamber and the sealing of the chamber-cranial interface with bone cement. The microdrive system was based on an earlier design (Gray and Goodell, 2007, 2011; Goodell and Gray, 2008; Markowitz et al., 2011; Salazar et al., 2012; Dotson et al., 2014, 2015), scaled up to match the dimensions of the chamber (Fig. 1B, C; Fig. S2). It consists of a guide array, an actuator block (top and bottom components), a printed circuit board (PCB), and a screw guide. The actuator block houses a set of linear actuators (256 in design 1, 252 in design 2), each consisting of a miniature stainless steel leadscrew, a threaded brass shuttle, and a compression spring (Fig. 1B, C; Fig. S3). In the first design (used for monkey 1), each actuator provided 20 mm of electrode travel at a resolution of 8 turns/mm. The second design (used for monkey 2), incorporated actuators with 33 mm or 41 mm of electrode travel at a resolution of 5 turns/mm. In both designs the actuators were spaced at 2.34 mm intervals. 54 The microdrives were assembled and loaded in multiple steps using sterile procedures (Fig. 2). First, each microelectrode (Glass insulated tungsten; 250 micron outer diameter; ~1MΩ impedance; Alpha Omega, Inc.) was cut to a specified length and bonded to a shuttle using a heat-cured conductive epoxy (Loctite 3880) (Fig. S3C). The microdrive components were assembled using machine screws after each opposing component surface was coated with a thin layer of sterile silicone grease. For design 1, the guide tubes were press fit into a counter bored hole on the bottom of the actuator block (Fig. 1B inset). Then for both designs, each electrode/shuttle assembly was front-loaded into the actuator block using polyimide tubing to guide the electrodes into place without damaging the tips (Fig. 2A). Once all electrodes were loaded, the polyimide guide tubing was removed and the PCB was fixed to the top of the actuator block. A leadscrew was passed through each hole in the PCB, rotated to capture the threads of the shuttle, and adjusted until the end of the leadscrew passed through the shuttle and became seated in a blind hole in the bottom portion of the actuator block (Fig. 2B). A compression spring was placed over the head of each leadscrew (Fig. 2C) and the screw guide was fixed on top of the PCB (Fig. 2D). When all the parts are assembled, the screw guide compresses the springs and forces contact between each leadscrew and the corresponding contact on the PCB. In this manner, a signal path is established between the electrode, shuttle, leadscrew and PCB that does not require any loose wires (Fig. 1B,C). The PCB routes the signals to eight 36-pin connectors (Omnetics Inc.). In the final step, the exposed electrodes (Fig. 2E) were retracted so the tips lie 0.5 mm inside the distal opening of the guide tubes (design 1) or 1.5 mm inside the 55 bottom of the guide array (design 2). The entire assembly was then gas sterilized before the sealing procedure. The microdrives were sealed in several steps. In design 1, each guide tube was injected with sterile silicone oil (M1000, Thomas Recording, Inc.) to act as a barrier to fluid flow (not shown). In design 2, we took a different approach and injected each guide hole with a calibrated volume of sterile silicone grease (Fig. 2F) (High-Vacuum Grease, Dow Corning, Inc.) and the bottom surface of the guide array was painted with several uniform layers of sterile silicone sealant (Fig. 2G) (734 Flowable Sealant, Dow Corning, Inc.). The combination of techniques used in design 2 insured that each guide hole was watertight, allowing for the downward movement of the electrode without any inward flow of fluid. Separate tests demonstrated that electrodes would easily pass through the silicone sealant without any changes in tip impedance. Implanting the Chamber and Microdrive We implanted the recording system in three stages – chamber implantation, craniotomy, and microdrive implantation – each followed by a period of recovery and testing to insure the health of the animals before proceeding to the next stage. All surgical and daily care, behavioral training and animal handling procedures were performed in accordance with NIH guidelines and the Institutional Animal Care and Use Committee of Montana State University. To implant the chamber, the skin, fascia, muscle and periosteum were successively retracted from the cranial bone over an area slightly larger than the chamber. The chamber was placed in position and sealed around its perimeter with a thin bead of C&B-Metabond 56 (Parkell, Inc.) cement. It was then anchored to the skull using Titanium bone screws (Gray Matter Research, LLC) and acrylic bone cement. Design 1 incorporated legs that were welded to the bottom surface of the chamber at ~1 cm intervals and included beveled holes for flathead bone screws (Fig. 1B, D). The second design included a series of horizontal grooves around the perimeter of the chamber to allow bone cement to firmly anchor the chamber to a set of bone screws mounted around the perimeter (Fig. 1C, E). After implanting the chamber, the cranial bone was left intact and a silicone gasket and short plug were mounted on to the chamber and covered with a short protective cap (Fig. S2A). The animals were given a minimum of 6 weeks to recover from this procedure before proceeding to the next stage. This allowed us to ensure that they were in good health and that the chamber was well anchored and free of both external and internal infection. The second and third stages of implantation differed between the two designs. In the first design, we used stainless steel guide tubes (24 gauge), press fit into the bottom of the actuator block, to protect the tip of each electrode (Fig. 1B). The guide array was mounted within the chamber as a separate part and served to insure that the array of guide tubes maintained accurate alignment (Fig. 1B). The bottom surface of the guide array was offset from the top surface of the cranial bone by 1 mm to avoid mechanical contact. The guide tube lengths were adjusted so that their distal ends extended into each cranial hole and came to rest just above the dura when the microdrive was mounted within the chamber (Fig. 1B). A separate drill guide was constructed with the same shape as the guide array so that a hole could be accurately drilled into the cranial bone at the correct location for each electrode. Cranial holes were drilled (stage 2) and the microdrive was implanted (stage 3) 57 in one sterile surgical procedure (Fig. 3A-E; Fig. S2). After removing the cap, plug and gasket, the chamber was rinsed and the cranial surface debrided of any connective tissue (Fig. 3A). The drill guide was mounted onto the chamber and 256 holes were drilled through the cranial bone using a hand-held motorized drill (Fig. 3B). After removing the guide, the bone was rinsed until all bleeding had ceased and the gasket, guide array and microdrive were installed and fixed to the chamber with machine screws (Fig. 3C-E). In the second design, the guide array was constructed to conform to the inner surface of the skull and was anchored to the actuator block with machine screws (Fig. 1C). Each hole in the guide array served to support and guide each electrode as it passed through the exposed dura and into the brain. For the second design, we separated the craniotomy procedure (stage 2) from the microdrive implantation (stage 3) (Fig. S2). Approximately 6 weeks after the chamber implant, we performed a craniotomy within the interior of the chamber while leaving the dural membrane intact (Fig. 3F-H). A Kerrison Rongeur was used to avoid damage to the dura and insure an accurate opening in the bone. The chamber was sealed with a gasket and a form-fitting plug that was contoured to the internal cranial surface (Fig. S2). This re-sealed the cranial cavity and prevented herniation. The animal was given 12 weeks to recover from the procedure. During this time we continually assessed the health of the animal and cultured the fluid inside the chamber at 3-week intervals using one of the drainage ports to gain access. Each of these tests was negative for infection, allowing us to proceed to the final stage with some assurance of the sterility of the chamber. In this procedure the form-fitting plug was removed, the chamber interior was cleaned of connective tissue, and the assembled microdrive mounted after replacing 58 the silicone gasket (Fig. 3I, J). The microdrive was secured with machine screws passing through the actuator block to the chamber. For both designs, the assemblies were covered with a protective Aluminum cap following the microdrive implantation (Fig. 1D, E). In both experiments, once the microdrive was mounted, it remained on the animal for the duration of the experiment (6-months in monkey 1, 9-months in monkey 2). The chamber served as the reference and ground connection to the animal using a machine screw linked to a trace on the PCB. Large-Scale Recordings in Behaving Non-Human Primates We trained both monkeys to perform an object-based, delayed match-to-sample task (dMTS), and a passive fixation task. Once the monkeys reached criterion performance on the dMTS task (>85% correct), we carried out the implantation sequence and began neural recordings when the animals were fully recovered, healthy and performing the task normally. To initiate recordings, we gradually moved all the electrodes through the dura and into the cortex over a period of 2-4 weeks. This was done in an incremental manner by advancing a subset of 10-30 electrodes each day until unit activity was detected. During this process we routinely measured electrode impedance and ceased advancing an electrode whenever its impedance was < ~50 kΩ or > ~2.5 MΩ. We considered low impedance electrodes to have damaged tips and did not move them further. We attempted to adjust the high impedance electrodes and recover the signal. If this failed, we considered the actuator or electrode to be damaged and did not move these electrodes further. Once we identified all of the functioning electrodes, we carried out daily recording sessions 3-5 days/week over a period of 6 and 9 months, in monkeys 1 and 2, respectively. 59 A total of 25 and 62 recording sessions were selected for further analysis for monkey 1 and monkey 2, respectively. Sessions were excluded for various reasons common to electrophysiological experiments, such as the animal’s behavioral performance, the running of separate tasks, or technical issues with the recording and task presentation hardware/software. Figure 4 shows an example of the broadband signals simultaneously recorded from 88 selected example electrodes in monkey 2 (Fig. S4 shows the data obtained for the entire trial). When moving the electrodes, our objective was to maximize the number of channels with detectable unit activity. However, we quickly discovered that it was too time consuming to search for unit activity on all channels each day. So at the beginning of each session, we inspected the signals and chose a subset of electrodes (ranging from 5 to 60) for further movement. We advanced these electrodes in increments of ¼ to 1 turn of the leadscrew (~30-120 μm) and typically did not advance any electrode more than 1 mm on any given day. For subsequent analysis of the data, it was important to know the anatomical location of each recording site over the course of both experiments. We therefore chose to end the recording phase of the experiments and proceed with the anatomical reconstructions. Here we describe the technique for monkey 2, the technique for monkey 1 is provided in the STAR methods. To estimate the location of each recording site, we combined several sources of information using the Computerized Anatomical Reconstruction and Editing Toolkit (Caret) (Van Essen et al., 2012). During the experiment, we kept a careful record of the depth of each recording site by counting the number of rotations of the leadscrew on each actuator. Once all recordings were completed, we made a small electrolytic lesion on 60 each functioning electrode (10μA DC for 25s), and then euthanized and perfused the animal with fixative with the microdrive in place and the electrodes extended to their final positions. We then removed the microdrive and later measured the distance that each electrode extended beyond the bottom surface of the microdrive in order to compare with the experimental depth measurements. The snug fit between the lower portion of the microdrive and the inner wall of the chamber kept off-axis movements to a minimum. We then removed and photographed the recorded hemisphere and prepared it for histology. Frozen sections were cut in the coronal plane (60μm thickness) and every 5th section was stained for Nissl substance and photographed (FD Neurotechnologies). During sectioning, we also photographed the frozen block-face of the brain to accurately record the shape of each slice. Figure 5A shows four examples of the Nissl-stained sections obtained from prefrontal cortex. The arrows mark the locations of identified electrode tracks or lesions. We imported the block-face photographs into Caret and then traced the outer cortical surface and marked the location of each electrode track or lesion from the corresponding Nissl-stained section (Fig. 5B). Using this information, the entire hemisphere was parsed into cortical areas based on the atlas reported in Markov et al. (2014) (Fig. 5C). Then, using the identified electrode tracks and lesions from the histology, we reconstructed each electrode track over the entire hemisphere and assigned a cortical area to each recording site in the entire data set (Fig. 5D). This process was facilitated by the fact that the absence of unit data often corresponded with the electrode being positioned in white matter. Following complete reconstruction of both data sets, we determined the total number of areas recorded from as well as the number of areas in which we made simultaneous 61 measurements. We excluded all measurements in which there was no detectable unit activity. In monkey 1, we sampled neuronal activity from 114 different electrodes spanning a period of 155 days. These data included recordings from 37 different cortical areas with a maximum of 21 cortical areas measured simultaneously. In monkey 2, we recorded neuronal activity from 143 different electrodes spanning a period of 260 days. Data were collected from 56 separate cortical areas and 4 subcortical nuclei (Caudate, Putamen, Thalamus, Claustrum). The maximum number of cortical areas recorded simultaneously in monkey 2 was 37. Combining the results from both experiments, we made recordings from a total of 61 cortical areas and 4 subcortical nuclei (Figure 6A). These data yielded a total of 335 well-isolated single unit recordings and 5,365 multi-unit recordings. In each of these recordings we also recovered the local field potential (LFP) signal. Next, we wanted to know how the system performed over time. Figure 6B, C shows the number of electrodes with detectable unit activity during each recording session, and figure 6D, E shows the number of cortical/sub-cortical areas in which these signals were measured simultaneously during each recording session. Over time, as we slowly advanced the electrodes through the tissue, the number of electrodes with detectable unit activity remained fairly stable (Fig. 6B, C), but the composition of recorded cortical areas changed over time – especially in monkey 2 (Fig. 6C, E). As electrodes were advanced some of them passed from gray matter into white matter, causing them to be excluded from the sample, while others followed the opposite pattern. Similarly, over time there was a reduction in measurements of motor and somatosensory areas accompanied by an increase in recordings from subcortical nuclei (especially in monkey 2). In general the recording 62 quality remained stable throughout the entire experiment and the electrodes displayed both increases and decreases in impedance over time (Fig. S5). Segregated Functional Networks Revealed by Relative Phase Relationships To demonstrate the power of this technique for understanding large-scale neuronal circuits, we used graph theoretic tools to study functional connectivity networks during the working memory task. The relative phase of the oscillatory activity between separate neural populations provides a candidate mechanism for modulating the integration/segregation of functional networks (Dotson et al., 2014; Maris et al., 2016). We have previously shown that the relative phase relationships between LFPs in widely separated cortical areas in the frontoparietal network are highly dynamic and concentrated near 0° (in-phase) or 180° (anti-phase) in the beta frequency band (12-25 Hz) (Dotson et al., 2014; Dotson and Gray, 2016). This finding suggests that anti-phase synchronization could be a mechanism for segregating functional networks. However, we know nearly nothing about the large-scale spatial distribution of phase relationships or their function. We addressed this issue by asking how the modularity of functional connectivity networks, computed using phase locking analysis, is modified when relative-phase information is taken into account. We performed a phase locking analysis of the LFP signals at multiple narrow bands to quantify the phase locking value (PLV) and its relative phase angle for each pair of signals as a function of frequency. This data was then used to create the functional connectivity networks. The phase locking analysis was performed following Tass et al. (1998), and we assessed the statistical significance of each PLV by comparing the observed PLV to surrogate distributions (see Methods section for more details). This analysis was performed 63 on a recording session from monkey 2. We used a 400ms window during the delay period, starting 400ms after the sample offset. We first wanted to know if the relative phase angles were bimodally distributed, similar to what we observed in the frontoparietal network (Dotson et al., 2014). Figure 7A shows that the relative phase distributions for 3 frequency bands (6-9Hz, 12-21Hz, and 24-30Hz) during the delay period of the dMTS task are indeed bimodal. Based on the assumption that anti-phase links serve to functionally segregate networks, we wanted to know how removing the anti-phase links affected the community structure of the functional connectivity network. To accomplish this, we used a graph theoretic measure called modularity, which determines the community structure of a network by dividing the network in a way that maximizes the within community links and minimizes the between community links (Rubinov and Sporns, 2010). Networks with a high modularity display largely non-overlapping communities. We used the phase locking data in the 12-21 Hz band to create functional connectivity networks for the full network (both in-phase and anti-phase links) and the in-phase only network. The anti-phase only network was too sparse to be considered. We chose to focus on the 12-21 Hz band, due to its relevance to working memory (Salazar et al., 2012; Dotson et al., 2014). We calculated the modularity for the full network and the in-phase only network using 9 different PLV thresholds. Multiple thresholds were used to enable the exploration of high density and low density networks, ensuring that our results are not a result of a specific threshold (Basset et al., 2008; Bullmore and Basset, 2011; Power et al., 2011). The light blue line and black line in Figure 7B shows the modularity values for the full network and in-phase only 64 network at each PLV threshold, respectively. This revealed that the modularity of the in- phase only network is well above the full network at all thresholds. Since simply removing links will result in an increase in modularity, we created a population of surrogate networks (both in-phase and anti-phase links) with the same number of links as the in-phase only network. We have previously reported that anti-phase links are predominately inter-areal (Dotson et al., 2014), so we created the surrogate networks by kicking out a set of randomly selected inter-areal links from the full network. We show the mean ± 3 std of the surrogate distributions (solid and dashed red lines) in Figure 7B. This shows that the in-phase only network is much more modular than the surrogate networks (p < .01, at all thresholds), indicating that the increase in modularity is due to exclusion of the anti-phase links. In order to visualize these networks, we imported the functional connectivity networks into the software package Gephi and used a built in graph layout algorithm (Bastian et al., 2009; Jacomy et al., 2014). Figure 7C shows the full network, with dot colors coded by modules (12-21 Hz; PLV threshold = 0.015). We see that there are 4 modules distributed across the recording sites (inset). In figure 7D we see that after removing the anti-phase links, the visual module (orange) is entirely separated from the other modules, indicating that the visual module is linked to the parietal, frontal and prefrontal areas with anti-phase links. There is also an additional module revealed. The green module covering part of the posterior parietal cortex in Figure 7C appears to be split across the intra-parietal sulcus (IPS) in Figure 7D, similar to Dotson et al. (2014) where the signals on opposite sides of the IPS were shown to be anti-phase. 65 Discussion We have developed and implemented a large-scale semi-chronic chamber and microdrive system that enables long-term recordings of neuronal activity from large numbers of independently movable microelectrodes spanning the depth and breadth of one hemisphere of the brain in behaving non-human primates. This approach provides an unprecedented advance in the ability of researchers to measure and characterize the dynamics of large-scale neural circuit activity and its relation to behavior. It greatly expands the area over which simultaneous measurements can be made, and allows the targeting of specific, but widely distributed neural circuits. By enabling the independent control of electrode position, the approach circumvents the limitations of chronically implanted fixed recording arrays and greatly expands the number of electrodes that can be implanted with acute recording techniques. Because the electrodes remain implanted for long periods, the time required to setup and take down an experiment are reduced to minutes, rather than hours. This method has the potential to open up new realms of investigation. It is becoming increasingly evident that the relative phase relationships within and between cortical areas may function to both integrate and segregate cortical networks (Dotson et al., 2014; Maris et al., 2016). We demonstrate, as a “proof-of-principle”, the ability to address network level questions that incorporate relative phase relationships in the LFP signals, which is difficult to do with other large-scale recording methods such as Magnetoencephalography (MEG) and Electrocorticography (ECoG). We are able to characterize the large-scale structure of functional networks and find that excluding the anti-phase links leads to a jump in the 66 modularity of the network. This change is larger than one would expect by simply removing the same number of randomly selected inter-areal links. And the result indicates that anti-phase links are not randomly distributed amongst cortical areas, rather they selectively segregate certain cortical areas from one another. Methodological Considerations Several aspects of the design and approach pose significant challenges. First, the method relies on a wide array of technology, including structural magnetic resonance imaging, computer-aided design software, multi-axis computer numerical control (CNC) machining, and Swiss screw machine technology capable of making the exceptionally long, fine-gage leadscrews. These technologies can be expensive and are not always routinely available. For these reasons, we contracted much of the machining and manufacture of the precision components. Second, the assembly and loading of the microdrive system requires a high degree of patience, manual dexterity, and attention to detail. All of the components must be assembled using sterile technique and great care must be taken to insure that all avenues for the entry of fluid into the system are sealed. Third, the surgical procedures for implanting the chamber, performing the craniotomy, and maintaining the good health of the animal also pose a number of challenges. The chamber system must be held rigidly to the cranial bone with a large number of bone screws linked with bone cement to the external chamber wall. And it’s necessary to establish a watertight seal at the interface between the cranial bone and the bone cement. Flaws in either of these procedures can lead to infection within the chamber or subsequent 67 mechanical failure of the implant. The skin surrounding the implant must be carefully sutured to avoid gaps that can lead to external infection, inflammation and discomfort for the animal. Both methods we employed for gaining access to the brain, the drilling of hundreds of individual holes or performing a hemispherical craniotomy, required great care and attention. We found the latter approach to be preferable because it eliminated the risk of damaging the underlying dura when using a high-speed drill without visual guidance. Fourth, the passage of large numbers of semi-chronically implanted microelectrodes through the dura can lead to leakage of cerebrospinal fluid (CSF) into the epidural space. If this fluid accumulates it can lead to an increase in pressure and potential compression of the brain. While we designed the ports in the chamber wall to enable the sampling, and subsequent culturing, of fluid from inside the chamber (design 2), it proved difficult to use the same ports to remove accumulated fluid during the recording phase of the experiment. However, this is an important concern and future versions of the system could be designed to enable the removal of fluid and to provide the ability to flush the epidural space in the event of an infection. Finally, we were unable to identify the anatomical locations of the recording sites when the experiment was in progress and had to wait for the histological reconstruction to obtain this valuable data. This is a significant issue because it requires the sacrifice of the animals, many months to complete the reconstruction, and severely constrains the timely reporting of research findings. A simple solution to this problem is to fabricate as many of the components of the system as possible out of MRI-compatible plastic, such as ULTEM, PEEK or 3D-printed materials. During various stages of the experiment the animals could 68 be re-scanned with fiducial markers mounted to the hardware and the electrode tip positions reconstructed on the basis of the scan and the recorded depths. Overall Performance Overall the animals tolerated the implants quite well. Both animals quickly learned to avoid bumping the protective cap against the walls of their home cage and the hardware remained stable for the duration of both experiments. We took great care to maintain the cleanliness of the wound margins around the perimeter of the implant. The first animal showed some bone decay within the chamber, which we attributed to compromised circulation within the bone due to the large number of cranial holes. On the second animal we repeatedly tested the fluid within the chamber prior to implantation of the microdrive and found no indication of infection. At the end of the experiment we removed the microdrive and found the residual fluid to be clear, with no sign of infection. Moreover, we found that the bone regrew a thin membrane around the inside perimeter of the chamber, which did not encroach on the recording electrodes. While the recordings were generally quite stable, and displayed excellent signal-to- noise ratio, several factors led to reduced yields in the sampling of neuronal activity. In the first animal, a substantial number of channels were lost when fluid entered a small number of guide tubes and traveled back up to the PCB and short-circuited a number of channels. We solved this problem in the second experiment by back filling each guide hole with sterile silicone grease and sealing the bottom surface with a silicone sealant. Another problem occurred when rotation of the leadscrew led to some rotation of the shuttle. This caused the electrode to bend, which affected the recording quality and occasionally led to 69 breakage of the glass insulation on the electrode and failure of that channel. We attributed this problem to variance in the internal diameter of the guide hole within the actuator block, allowing the shuttle to rotate as the leadscrew was rotated over part of the travel distance. Another problem, largely out of our control, occurred when the delicate electrode tips changed impedance or broke when passing through the dural membrane. Finally, over the course of both experiments each electrode passed through substantial sections of white matter, preventing us from recording neuronal activity. In spite of these combined factors, we were nonetheless able to sample neuronal activity from a large number of cortical sites spanning a large fraction of the cortical mantle. Future Directions Since completing this study, we have made a number of improvements in the design of the system. To prevent rotation of the shuttles when advancing the electrodes, we designed a shuttle with a teardrop cross-sectional shape and utilized 3D printing technology to fabricate the actuator block with matching teardrop shaped channels (Goodell and Gray, 2015). This greatly increased the reliability of the actuators, increased the efficiency of fabrication of the actuator block, and enabled us to modify the exit path of the guide holes so that the inter-electrode spacing could be decreased. We have improved the design of the leadscrews to provide up to 42 mm of electrode travel with a thread pitch of 8 turns/mm, and modified the design of the screw head and screwdriver to a half-moon shape. These changes have enabled the device to target structures on the ventral surface of the brain and enhanced the screwdriver-leadscrew interface. 70 Several further design changes could greatly enhance the performance of the system and the ease with which it is implanted and maintained. First, control of each actuator could be automated. This would require the design of a robotic screwdriver capable of indexing each channel in the array, delivering a specified change in electrode depth, and tracking all of the information in conjunction with data acquisition. Second, the recording yield of the system could be increased by an order of magnitude by introducing multi-contact silicon polytrodes. Such probes would likely require long flexible ribbon cables to tether the signals to a set of electrode interface boards or active on-board electronics for digitizing and multiplexing the large numbers of signals. This would require a change to the design of the actuator mechanism and the incorporation of probes with sufficient rigidity to pass through the dura unaffected. Finally, the future implementation of this large-scale approach would benefit greatly from a design that minimizes the number of surgical procedures and eliminates the need for acrylic bone cement. We think this could be accomplished by designing a prosthetic platform that conforms to each animal’s cranial surface and could be mounted in a single surgical procedure without the use of acrylic bone cement (Mulliken et al., 2015). The device would include prefabricated ports, in which a craniotomy could be made and specialized microdrives implanted for recording and manipulating neuronal activity in large-scale circuits. 71 Figures Figure 1. Design Drawings for Designs 1 and 2 of the Large-Scale Chamber and Microdrive System (A) Illustration of the MRI is shown on the left with the inner and outer surfaces of the cranial bone outlined in yellow. The middle and right images show the skull model highlighting the boundaries of the intended craniotomy. (B and C) Exploded views of the design drawings for design 1 (B) and design 2 (C) of the complete system. The insets show a cutaway view (not to scale) of the actuator mechanism. (D and E) Assembled views of the design drawings for design 1 (D) and design 2 (E) mounted on their respective cranial models. Abbreviations: PCB, printed circuit board: ABt, actuator block top; ABb, actuator block bottom; CS, compression spring; S, shuttle; LS, lead-screw, E, electrode. 72 Figure 2. Steps in the Assembly and Sealing of the Microdrive System for Design 2 Only (A) An oblique view of the partially assembled microdrive (top and bottom actuator blocks and guide array) mounted in a holder (gray) while an electrode is being front loaded into one of the polyimide tubes. (B) Top-down view of the Microdrive with the PCB mounted and a set of leadscrews inserted. (C) The same view as in (B) with all of the leadscrews and compression springs loaded. (D) Larger view of the microdrive with screw guide securely mounted. (E) Side view of the assembled Microdrive with all electrodes fully extended. (F) Image showing how the sterile silicone grease is injected into each guide hole. (G) Bottom-up view showing the layer of silicone sealant on the bottom surface of the guide array. 73 Figure 3. Craniotomy and Microdrive Implantation Procedures for Stages II and III The images in (A)-(E) Illustrate the implantation procedure for monkey 1 (design 1). (A) View of the interior of the chamber following the aspiration of fluid and connective tissue. (B) Drilling cranial holes using the drill guide (white). (C) View of the chamber interior after drilling the cranial holes. (D) Insertion of the guide array. (E) Mounting of the Microdrive. The images in (F)-(J) show similar steps for monkey 2 (design 2). (F) View of the chamber interior 6 weeks after implantation. Accumulated fluid is visible. (G) Same view after aspiration of the fluid. (G) Chamber interior immediately following the craniotomy. (I) Chamber interior 12 weeks after the craniotomy. The dural membrane is healthy and shows signs of growth around the perimeter of the form-fitting plug. (J) Mounting of the Microdrive (note: this image shows a test of the microdrive fit immediately after the craniotomy was made, not the actual implantation, which was done 12 weeks later.). 74 Figure 4. A Brief 300 ms Segment of Broadband Data Simultaneously Sampled from 88 Microelectrodes in Monkey 2 Just Prior to the Onset of the Match Stimulus during the dMTS Task. The cortical area of each recording is indicated just to the left of each trace according to the nomenclature of Markov et al., (2014). 75 Figure 5. Anatomical Reconstruction Process (A) Nissl-stained sections of prefrontal cortex. Arrows indicate lesions or small holes made by the electrodes. (B) Photographs of the face of the frozen tissue block during sectioning aligned to the histological sections in (A). The arrows correspond to thos shown in (A). The dotted lines mark the outer cortical surface, where each distinct color marks a different cortical area according to the atlas of Markov et al. (2014). The solid green lines mark the inner cortical surface. (C) A portion of prefrontal cortex reconstructed using the Caret software. Colored dots indicate cortical areas. (D) Example of the reconstructed electrode tracts using the information from (A)- (C). 76 Figure 6. Summary of Recording Performance (A) Flatmap of the cortex taken from Markov et al. (2014) showing the cortical areas in which signals were recorded in both monkeys. (B and C) Number of electrodes with detectable unit activity on each recording session for monkeys 1 (B) and 2 (C), respectively. (D and E) Number of cortical areas (and subcortical nuclei) simultaneously sampled on each recording session for monkeys 1(D) and 2 (E), respectively. In all plots, the colors indicate brain region (see legend). 77 Figure 5. Results of the Functional Connectivity Analysis (A) Histograms of the relative phase distributions for three frequency bands in the LFP during the delay period of the task. The solid and dashed lines show all the statistically significant values (see STAR Methods) and those that exceeded a PLV magnitude of 0.015, respectively. (B) Modularity as a function of the PLV magnitude threshold. The light blue line shows the modularity value for the network with all links (full network). The black line is the modularity value for the in-phase-only network. The solid and dashed red lines show the mean +/- 3 SD of the surrogate distributions. The red asterisk indicates the PLV threshold value (0.015) used for the visualization shown in (C) and (D). (C and D) Visualization of the cortical network with all links (C) and with only in-phase links (D). Colors indicate modules. Inset brain images show the corresponding locations of electrode entry into the brain. 78 Figure S1. Design of the skull model and definition of the chanber boundaries (related to Figure 1). (A) The left image shows a coronal section of the MRI with outlines of the inner and outer boundaries of the skull in yellow. A sequence of the inner and outer outlines (8 shown) are linked, converted to their respective solid models, and subtracted from one another to yield the final 3D model of the skull (right). (B) To define the boundary of the internal chanber wall the lateral and medial edges are marked in each coronal MRI section (three examples shown). The green lines mark the boundaries and the red circles mark the junctions of these lines on the skull surface. (C) Dorsal and lateral views of the 3D skull model with the boundaries of the craniotomy outlined in black (left) along with an exploded view of the intended craniotomy (right). 79 Figure S2. Exploded views of the design drawings of the large-scale chamber and Microdrive systems for each of the three stages of design 1 and design 2 (related to Figure 1). (A) Stage I designs illustrating the chamber, short plug and protective cap. (B) Stage II designs illustrating the drill guide and chamber for design 1 (left) and the chamber, form-fitting plug and protective cap for design 2. (C) Stage III design drawings showing the relationships between the Microdrive components and the chambers and protective caps. 80 Figure S3. Actuator mechanism (related to Figure 1). (A) Design drawing of the actuator showing the relationships between the leadscrew, shuttle and compression spring. (B) Photographs of the three different leadscrews and two types of shuttles used in the two designs. The 41 mm and 33 mm leadscrews used in design 2 are shown in 1 and 2, respectively. The 20 mm leadscrew and associated shuttle used in design 1 are shown in 3. (C) Photographs illustrating the bonding of the electrode to the brass shuttle using a heat-treated conductive epoxy (arrows). 81 Figure S4. Broadband data simultaneously sampled from 88 microelectrodes in monkey 2 for the full duration of the trial shown in Figure 4. The cortical area of each recording is indicated just to the left of each trace according to the nomenclature of Markov et al. (2014). The black vertical lines indicate the sample onset, sample offset, and the match onset. 82 Figure S5. Recording quality over time for a single electrode and the changes in electrode impedance across the experiments (related to Figure 6). (A) The image on the left shows the outline of a coronal section through the prefrontal cortex with areal boundaries indicated by grey lines according to Markov et al. (2014). The dashed green line shows the path of a reconstructed electrode track. The points marked on the track indicate the recording locations for the four raw data traces shown in the middle of the plot. These signals were recorded in monkey 2 on chronological days 3, 42, 140 and 202, sampled from areas 8B, 24c, 32 and 14, respectively. The plots on the right show 100 superimposed waveforms extracted from each recording (MUA (black) and SUA (red)). (B-C) Histograms of the difference in electrode impedance from the last set of impedance measurements relative to the first set, for all applicable electrodes in Monkey 1 (B) and Monkey 2 (C). Data were restricted to functioning electrodes with impedance values ranging between 0.2 to 1.5 MW. There are a wide range of changes in impedance in both experiments, and a consistent reduction in impedance in monkey 2 of ~330 kW. 83 Supplemental Material Experimental Model and Subject Details Subjects: Data was collected from two adult female macaque monkeys. All procedures were performed in accordance with NIH guidelines and the Institutional Animal Care and Use Committee of Montana State University. The monkeys were head fixed using a cranial head post (Gray Matter Research, LLC), and positioned 57 cm from a 19-inch monitor. Method Details Behavioral Task: We used MonkeyLogic software to run the experiment and record the eye position (Asaad and Eskandar, 2008a,b). Eye position was monitored using an infrared eye-tracking system (240 Hz; ISCAN, Inc.). Eye signals were recorded and converted to degrees of visual angle (dva) by MonkeyLogic. A trial begins when the monkey acquires and holds fixation on a small fixation spot (fixation window = 3 dva). At a latency of 500 ms for monkey 1 or 800 ms for monkey 2, one of five possible sample images (size: 2.4x2.4 dva) is presented for 500 ms in the center of the screen (obscuring the fixation point). During the sample period the monkey has to maintain it’s gaze in the same 3 dva window used during the fixation period. The sample stimulus is followed by a randomized delay, 800-1200 ms for monkey 1 and 1000-1500 ms for monkey 2, in which no stimulus is present. At the end of the delay period, the fixation target is extinguished and the matching image and a non-matching image (one of the four other images) appear 5 dva from the center of the screen. For monkey 1, the match and non-match were always placed across from each other on the horizontal plane. The location (left or right) of the match and non-match were randomized on each trial. For 84 monkey 2, the images were aligned either vertically, horizontally or diagonally. The location of the match and non-match and the alignment was randomly chosen on each trial. While the match image is visible, the monkey must make a saccadic eye movement to the matching image and maintain fixation for a brief period of time (200 ms for monkey 1 and 500 ms for monkey 2). Correct trials were rewarded with a drop of juice. Chamber System: All custom components were designed using the Solidworks CAD package (Dassault Systems, Inc.). The recording systems were designed to fit each animal by creating a 3D model of each animal’s skull. To achieve this, we constructed an MRI-based 3D graphical model of each monkey’s skull using Solidworks (Fig. S1). Coronal sections were taken at 2 mm intervals spanning the infra-orbital and occipital ridges. We outlined the external surface of the skull and the internal surface of the cranial vault in each section using a spline interpolation algorithm. Successive outlines were linked to form a solid 3D model. We delineated the internal dimensions of the chamber using the outer boundaries of the cerebral hemisphere to define a closed contour on the surface of the model. The chamber wall was 0.1” thick and its bottom surface was matched to the cranial surface at all locations. In some locations this caused the external boundary of the chamber to extend too far, so we reduced the dimensions to obtain a suitable fit to the skull. In design 2, we introduced sealed ports in the chamber wall to enable the sampling or release of fluids inside the chamber and the introduction of disinfectants or antibiotics. 85 Electrophysiological Recordings: Broadband signals (0.1 Hz - 9 kHz, sampled at 32 kHz) were recorded using a 256-channel Digital Lynx system (Neuralynx, Inc.). Signals were lowpass filtered (1-100 Hz) and down-sampled to 1 kHz to obtain the local field potential (LFP). Spike sorting was performed in a similar manner as in earlier studies (Salazar et al., 2012; Dotson et al., 2014). First, broadband signals were highpass filtered (monkey 1: 500 Hz – 9 kHz; monkey 2: 500 Hz – 4 kHz). Second, a threshold of 5 standard deviations of the background signal was used to identify spikes. 32 data points were saved for each spike (11 points before and 21 points after and including the minimum). Waveforms were clustered using KlustaKwik (Rossant et al., 2016). Clusters were merged and artifacts were discarded using MClust (http://redishlab.neuroscience.umn.edu/MClust/MClust.html). To be considered a single unit (SUA), waveforms in the cluster were required to be stable over time, non- overlapping with all other clusters, and have an inter-spike interval histogram with a clear refractory period. Anatomical Reconstruction Technique for Monkey 2: Following perfusion, the brain was removed, and sunk in a solution of fixative with 30% sucrose several days before being sectioned (60 µm) and stained for Nissl substance (FD Neurotechnologies, Inc.). To reconstruct the brain, the stained sections were photographed and then imported into Free-D (Andrey and Maurin, 2005). Sections were manually registered and then electrolytic lesions and microelectrode tracks were marked on the images. This information provided a 3D reconstruction of all the microelectrode tracks. We used this information and the record of microelectrode depths to estimate the microelectrode tip 86 position and identify the recording locations of each microelectrode on each recording session. Quantification and Statistical Analysis Network Analysis: The data used in the network analysis was collected during a single recording session from monkey 2 (106 total signals). Data is from areas 7A, 7B, 8B, 8L, 8M, 8r, AIP, DP, F1, F2, F6, F7, LIP, MT, TPt, V1, V2, V6A, 1, 2, 3, 5, 24c, 44, 45B, and 46D. The phase locking analysis was performed on a 400ms window during the delay period, starting 400ms after the sample offset. The phase locking analysis was implemented following Tass et al. (1998). First, each signal was bandpass filtered (zero phase forward and reverse digital IIR fourth order Butterworth filter) in 3 Hz steps with a 5 Hz band from 3 to 90 Hz, and then down sampled to 200 Hz. Next, the Hilbert transform was applied to extract time series of the instantaneous phase angles [ϕ(t)]. Then, using the instantaneous phase angles, we calculated the distribution of relative phase angles [Ψ = (𝜙( − 𝜙*) 𝑚𝑜𝑑2𝜋] between each pair of signals. The resulting data was binned from 0° to 360° in 10° bins (N = 36 bins). The phase locking value [PLV = (𝑆789 − 𝑆)/𝑆789] was then calculated using Shannon entropy (S), where 𝑆 = −∑@AB( 𝑃(𝑖) 𝑙𝑜𝑔*𝑃(𝑖), and P(i) is the probability of observing a relative phase angle in each bin. The maximum entropy (𝑆789) is calculated as [𝑆789 = 𝑙𝑜𝑔*(𝑁)]. We assessed the statistical significance of each phase locking value by comparing the observed PLV values to surrogate distributions. Surrogates were created by randomizing 87 trial labels and then computing the PLV. Each surrogate distribution (200 surrogates) was fit with a generalized extreme value function in order to estimate p-values less than 0.005 (smallest p-value using just the surrogate distribution is 1/200 = 0.005). Phase locking values between LFPs are typically much larger than shuffled surrogates (Antzoulatos and Miller, 2014). If the synchronized activity is not locked to task events, then the phase locking values estimated from trial shuffled data are very small. To account for this, we chose a very small significance threshold of p<0.000001, which is slightly smaller than what would have been chosen based on Bonferroni correction for each functional connectivity network (with 106 nodes there are 5,565 pairs of signals; with a desired significance threshold of p<.05, the Bonferroni corrected threshold is 0.05/5,5565 = 0.000009). Although, we found that the exact significance threshold made little difference, it did help to prevent noisy data points from entering the analysis. For the functional connectivity analysis, if multiple PLVs were significant within a frequency range, then the PLV with the highest magnitude and the corresponding relative phase angle was used. The modularity analysis was performed using the Brain Connectivity Toolbox (Rubinov and Sporns, 2010). Graph theoretic analyses typically explore network properties using a range of magnitude thresholds for a given metric, which enables the exploration of high density to low density networks (Basset et al., 2008; Bullmore and Basset, 2011; Power et al., 2011). Subsequently, after selecting data points for significance based on the trial shuffled data, we used multiple magnitude thresholds based on the PLVs. This allowed us to quantify how the modularity changes as a function of network density and to determine whether our results are robust to the chosen parameters. Undirected binary 88 graphs were created using the results from the phase locking analysis. Network visualization was performed in the software package Gephi using force atlas and a built in modularity analysis (Bastian et al., 2009; Jacomy et al., 2014). 89 References Andrey, P. and Maurin, Y., 2005. Free-D: an integrated environment for three-dimensional reconstruction from serial sections. Journal of neuroscience methods, 145(1), pp.233-244. Antzoulatos, E.G. and Miller, E.K., 2014. Increases in functional connectivity between prefrontal cortex and striatum during category learning. Neuron, 83(1), pp.216- 225. Asaad, W.F., Rainer, G. and Miller, E.K., 2000. Task-specific neural activity in the primate prefrontal cortex. Journal of Neurophysiology, 84(1), pp.451-459. Asaad, W.F. and Eskandar, E.N., 2008a. Achieving behavioral control with millisecond resolution in a high-level programming environment. Journal of neuroscience methods, 173(2), pp.235-240. Asaad, W.F. and Eskandar, E.N., 2008b. A flexible software tool for temporally-precise behavioral control in Matlab. Journal of neuroscience methods, 174(2), pp.245- 258. Bassett, D.S., Bullmore, E., Verchinski, B.A., Mattay, V.S., Weinberger, D.R. and Meyer-Lindenberg, A., 2008. Hierarchical organization of human cortical networks in health and schizophrenia. Journal of Neuroscience , 28(37), pp.9239- 9248. Bastian, M., Heymann, S. and Jacomy, M., 2009. Gephi: an open source software for exploring and manipulating networks. ICWSM, 8, pp.361-362. Battaglia, F.P., Kalenscher, T., Cabral, H., Winkel, J., Bos, J., Manuputy, R., van Lieshout, T., Pinkse, F., Beukers, H. and Pennartz, C., 2009. The Lantern: an ultra-light micro-drive for multi-tetrode recordings in mice and other small animals. Journal of neuroscience methods, 178(2), pp.291-300. Bullmore, E.T., & Bassett, D.S., 2011. Brain graphs: graphical models of the human brain connectome. Annual review of clinical psychology , 7 , pp.113-140. Buschman, T.J. and Miller, E.K., 2007. Top-down versus bottom-up control of attention in the prefrontal and posterior parietal cortices. science, 315(5820), pp.1860-1862. Christopher deCharms, R., Blake, D.T. and Merzenich, M.M., 1999. A multielectrode implant device for the cerebral cortex. Journal of neuroscience methods, 93(1), pp.27-35. 90 Csicsvari, J., Henze, D.A., Jamieson, B., Harris, K.D., Sirota, A., Barthó, P., Wise, K.D. and Buzsáki, G., 2003. Massively parallel recording of unit and local field potentials with silicon-based electrodes. Journal of neurophysiology, 90(2), pp.1314-1323. Dickey, A.S., Suminski, A., Amit, Y. and Hatsopoulos, N.G., 2009. Single-unit stability using chronically implanted multielectrode arrays. Journal of neurophysiology, 102(2), pp.1331-1339. Donoghue, J.P., Sanes, J.N., Hatsopoulos, N.G. and Gaál, G., 1998. Neural discharge and local field potential oscillations in primate motor cortex during voluntary movements. Journal of neurophysiology, 79(1), pp.159-173. Dotson, N.M., Salazar, R.F. and Gray, C.M., 2014. Frontoparietal correlation dynamics reveal interplay between integration and segregation during visual working memory. Journal of Neuroscience, 34(41), pp.13600-13613. Dotson, N.M., Goodell, B., Salazar, R.F., Hoffman, S.J. and Gray, C.M., 2015. Methods, caveats and the future of large-scale microelectrode recordings in the non-human primate. Frontiers in systems neuroscience, 9, pp.149-149. Dotson, N.M. and Gray, C.M., 2016. Experimental observation of phase-flip transitions in the brain. Physical Review E, 94(4), p.042420. Eliades, S.J. and Wang, X., 2008. Chronic multi-electrode neural recording in free-roaming monkeys. Journal of neuroscience methods, 172(2), pp.201-214. Erickson, C.A. and Desimone, R., 1999. Responses of macaque perirhinal neurons during and after visual stimulus association learning. Journal of Neuroscience, 19(23), pp.10404-10416. Fee, M.S. and Leonardo, A., 2001. Miniature motorized microdrive and commutator system for chronic neural recording in small animals. Journal of neuroscience methods, 112(2), pp.83-94. Feingold, J., Desrochers, T.M., Fujii, N., Harlan, R., Tierney, P.L., Shimazu, H., Amemori, K.I. and Graybiel, A.M., 2012. A system for recording neural activity chronically and simultaneously from multiple cortical and subcortical regions in nonhuman primates. Journal of neurophysiology, 107(7), pp.1979-1995. Fries, P., Roelfsema, P.R., Engel, A.K., König, P. and Singer, W., 1997. Synchronization of oscillatory responses in visual cortex correlates with perception in interocular rivalry. Proceedings of the National Academy of Sciences, 94(23), pp.12699- 12704. 91 Goodell A.B., Gray C.M., 2008. A large-scale, distributed recording system for semi- chronic monitoring of cortical and sub-cortical neuronal activity in alert monkeys. Soc. Neurosci. Abs., 101.7. Goodell A.B., Gray C.M., 2015. Next generation of large-scale, semi-chronic micromanipulator systems for measuring neuronal activity in distributed neural circuits in behaving, non-human primates. Soc. Neurosci. Abs., 267.06. Gray, C.M., Goodell, B. and Lear, A., 2007. Multichannel micromanipulator and chamber system for recording multineuronal activity in alert, non-human primates. Journal of neurophysiology, 98(1), pp.527-536. Gray, C.M., and Goodell, B., 2007. A high-density, large-scale, distributed recording system for semi-chronic monitoring of cortical and sub-cortical neuronal activity in alert monkeys. In Soc Neurosci Abstr (Vol. 33, No. 624.8). Gray C.M., and Goodell B. 2011. Spatiotemporal Dynamics of Synchronous Activity Across Multiple Areas of the Visual Cortex in the Alert Monkey. In: The Dynamic Brain: An Exploration of Neuronal Variability and Its Functional Significance. Pgs 233-254, Eds. Mingzhou Ding and Dennis Glanzman. Oxford University Press. Hernández, A., Nácher, V., Luna, R., Alvarez, M., Zainos, A., Cordero, S., Camarillo, L., Vázquez, Y., Lemus, L. and Romo, R., 2008. Procedure for recording the simultaneous activity of single neurons distributed across cortical areas during sensory discrimination. Proceedings of the National Academy of Sciences, 105(43), pp.16785-16790. Hoffman, K.L. and McNaughton, B.L., 2002. Coordinated reactivation of distributed memory traces in primate neocortex. Science, 297(5589), pp.2070-2073. Jacomy, M., Venturini, T., Heymann, S. and Bastian, M., 2014. ForceAtlas2, a continuous graph layout algorithm for handy network visualization designed for the Gephi software. PloS one, 9(6), p.e98679. Kim, T., Troyk, P.R. and Bak, M., 2006, August. Active floating micro electrode arrays (AFMA). In Engineering in Medicine and Biology Society, 2006. EMBS'06. 28th Annual International Conference of the IEEE (pp. 2807-2810). IEEE. Kipke, D.R., Vetter, R.J., Williams, J.C. and Hetke, J.F., 2003. Silicon-substrate intracortical microelectrode arrays for long-term recording of neuronal spike activity in cerebral cortex. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 11(2), pp.151-155. 92 Kloosterman, F., Davidson, T.J., Gomperts, S.N., Layton, S.P., Hale, G., Nguyen, D.P. and Wilson, M.A., 2009. Micro-drive array for chronic in vivo recording: drive fabrication. JoVE (Journal of Visualized Experiments), (26), pp.e1094-e1094. Krupa, D.J., Wiest, M.C., Shuler, M.G., Laubach, M. and Nicolelis, M.A., 2004. Layer- specific somatosensory cortical activation during active tactile discrimination. Science, 304(5679), pp.1989-1992. Maris, E., Fries, P. and van Ede, F., 2016. Diverse phase relations among neuronal rhythms and their potential function. Trends in neurosciences, 39(2), pp.86-99. Markov, N.T., Ercsey-Ravasz, M.M., Ribeiro Gomes, A.R., Lamy, C., Magrou, L., Vezoli, J., Misery, P., Falchier, A., Quilodran, R., Gariel, M.A. and Sallet, J., Gamanut, R., Huissoud, C., Clavagnier, S., Giroud, P., Sappey-Marinier, D., Barone, P., Dehay, C., Toroczkai, Z., Knoblauch, K., Van Essen, D.C., and Kennedy, H., 2014. A weighted and directed interareal connectivity matrix for macaque cerebral cortex. Cerebral cortex, 24(1), pp.17-36. Markowitz, D.A., Wong, Y.T., Gray, C.M. and Pesaran, B., 2011. Optimizing the decoding of movement goals from local field potentials in macaque cortex. Journal of Neuroscience, 31(50), pp.18412-18422. Maynard, E.M., Hatsopoulos, N.G., Ojakangas, C.L., Acuna, B.D., Sanes, J.N., Normann, R.A. and Donoghue, J.P., 1999. Neuronal interactions improve cortical population coding of movement direction. Journal of Neuroscience, 19(18), pp.8083-8093. Mendoza, G., Peyrache, A., Gámez, J., Prado, L., Buzsáki, G. and Merchant, H., 2016. Recording extracellular neural activity in the behaving monkey using a semichronic and high-density electrode system. Journal of Neurophysiology, 116(2), pp.563- 574. Miller, E.K. and Wilson, M.A., 2008. All my circuits: using multiple electrodes to understand functioning neural networks. Neuron, 60(3), pp.483-488. Mountcastle, V.B., Reitboeck, H.J., Poggio, G.F. and Steinmetz, M.A., 1991. Adaptation of the Reitboeck method of multiple microelectrode recording to the neocortex of the waking monkey. Journal of neuroscience methods, 36(1), pp.77-84. Mulliken, G.H., Bichot, N.P., Ghadooshahy, A., Sharma, J., Kornblith, S., Philcock, M. and Desimone, R., 2015. Custom-fit radiolucent cranial implants for neurophysiological recording and stimulation. Journal of neuroscience methods, 241, pp.146-154. 93 Musallam, S., Bak, M.J., Troyk, P.R. and Andersen, R.A., 2007. A floating metal microelectrode array for chronic implantation. Journal of neuroscience methods, 160(1), pp.122-127. Nguyen, D.P., Layton, S.P., Hale, G., Gomperts, S.N., Davidson, T.J., Kloosterman, F. and Wilson, M.A., 2009. Micro-drive array for chronic in vivo recording: tetrode assembly. JoVE (Journal of Visualized Experiments), (26), pp.e1098-e1098. Nicolelis, M.A., Ghazanfar, A.A., Faggin, B.M., Votaw, S. and Oliveira, L.M., 1997. Reconstructing the engram: simultaneous, multisite, many single neuron recordings. Neuron, 18(4), pp.529-537. Nicolelis, M.A., Dimitrov, D., Carmena, J.M., Crist, R., Lehew, G., Kralik, J.D. and Wise, S.P., 2003. Chronic, multisite, multielectrode recordings in macaque monkeys. Proceedings of the National Academy of Sciences, 100(19), pp.11041- 11046. Nordhausen, C.T., Maynard, E.M. and Normann, R.A., 1996. Single unit recording capabilities of a 100 microelectrode array. Brain research, 726(1), pp.129-140. Power, J.D., et al., 2011. Functional network organization of the human brain. Neuron 72(4), pp.665-678. Prut, Y., Vaadia, E., Bergman, H., Haalman, I., Slovin, H. and Abeles, M., 1998. Spatiotemporal structure of cortical activity: properties and behavioral relevance. Journal of neurophysiology, 79(6), pp.2857-2874. Purushothaman, G., Scott, B.B. and Bradley, D.C., 2006. An acute method for multielectrode recording from the interior of sulci and other deep brain areas. Journal of neuroscience methods, 153(1), pp.86-94. Roelfsema, P.R., Engel, A.K. and Singer, W., 1997. Visuomotor integration is associated with zero time-lag synchronization among cortical areas. Nature, 385(6612), p.157. Rossant, C., Kadir, S.N., Goodman, D.F., Schulman, J., Hunter, M.L., Saleem, A.B., Grosmark, A., Belluscio, M., Denfield, G.H., Ecker, A.S. and Tolias, A.S., 2016. Spike sorting for large, dense electrode arrays. Nature neuroscience, 19(4), pp.634-641. Rousche, P.J., Pellinen, D.S., Pivin, D.P., Williams, J.C., Vetter, R.J. and Kipke, D.R., 2001. Flexible polyimide-based intracortical electrode arrays with bioactive capability. IEEE Transactions on biomedical engineering, 48(3), pp.361-371. 94 Rubinov, M. and Sporns, O., 2010. Complex network measures of brain connectivity: uses and interpretations. Neuroimage, 52(3), pp.1059-1069. Salazar, R.F., Dotson, N.M., Bressler, S.L. and Gray, C.M., 2012. Content-specific fronto- parietal synchronization during visual working memory. Science, 338(6110), pp.1097-1100. Schwarz, D.A., Lebedev, M.A., Hanson, T.L., Dimitrov, D.F., Lehew, G., Meloy, J., Rajangam, S., Subramanian, V., Ifft, P.J., Li, Z. and Ramakrishnan, A., 2014. Chronic, wireless recordings of large-scale brain activity in freely moving rhesus monkeys. Nature methods, 11(6), pp.670-676. Swadlow, H.A., Bereshpolova, Y., Bezdudnaya, T., Cano, M. and Stoelzel, C.R., 2005. A multi-channel, implantable microdrive system for use with sharp, ultra-fine “Reitboeck” microelectrodes. Journal of neurophysiology, 93(5), pp.2959-2965. Szabo, I., Czurko, A., Csicsvari, J., Hirase, H., Leinekugel, X. and Buzsáki, G., 2001. The application of printed circuit board technology for fabrication of multi-channel micro-drives. Journal of neuroscience methods, 105(1), pp.105-110. Tass, P., Rosenblum, M.G., Weule, J., Kurths, J., Pikovsky, A., Volkmann, J., Schnitzler, A. and Freund, H.J., 1998. Detection of n: m phase locking from noisy data: application to magnetoencephalography. Physical review letters, 81(15), p.3291. Tolias, A.S., Ecker, A.S., Siapas, A.G., Hoenselaar, A., Keliris, G.A. and Logothetis, N.K., 2007. Recording chronically from the same neurons in awake, behaving primates. Journal of neurophysiology, 98(6), pp.3780-3790. Van Essen, D.C., 2012. Cortical cartography and Caret software. Neuroimage, 62(2), pp.757-764. Venkatachalam, S., Fee, M.S. and Kleinfeld, D., 1999. Ultra-miniature headstage with 6- channel drive and vacuum-assisted micro-wire implantation for chronic recording from the neocortex. Journal of Neuroscience Methods, 90(1), pp.37-46. Vetter, R.J., Williams, J.C., Hetke, J.F., Nunamaker, E.A. and Kipke, D.R., 2004. Chronic neural recording using silicon-substrate microelectrode arrays implanted in cerebral cortex. IEEE transactions on biomedical engineering, 51(6), pp.896-904. Vos, B.P., Wijnants, M., Taeymans, S. and De Schutter, E., 1999. Miniature carrier with six independently moveable electrodes for recording of multiple single-units in the cerebellar cortex of awake rats. Journal of Neuroscience Methods, 94(1), pp.19-26. 95 Wilson, M.A. and McNaughton, B.L. (1993) Dynamics of the hippocampal ensemble code for space. Science, 261(5124), pp.1055-1058. 96 CHAPTER THREE FEATURE-BASED VISUAL SHORT-TERM MEMORY IS WIDELY DISTRIBUTED AND HIERARCHICALLY ORGANIZED Contribution of Authors and Co-Authors Manuscript(s) in Chapter 3 Author: Nicholas M. Dotson Contributions: study design, data collection, data analysis, and preparing manuscript. Co-Author: Steven J. Hoffman Contributions: data collection, assisted in data analysis, and editing manuscript. Co-Author: Baldwin Goodell Contributions: technology design and fabrication Co-Author: Charles M. Gray Contributions: study design, data collection, data analysis and preparing manuscript 97 Manuscript Information Page Nicholas M. Dotson, Steven J. Hoffman, Baldwin Goodell, Charles M. Gray Neuron Status of Manuscript: ____ Prepared for submission to a peer-reviewed journal ____ Officially submitted to a peer-review journal ____ Accepted by a peer-reviewed journal __X__ Published in a peer-reviewed journal CellPress July 2018 Issue 99 98 Summary Feature-based visual short-term memory is known to engage both sensory and association cortices. However, the extent of the participating circuit and the neural mechanisms underlying memory maintenance are still a matter of vigorous debate. To address these questions, we recorded neuronal activity from 42 cortical areas in monkeys performing a feature-based visual short-term memory task and an interleaved fixation task. We find that task-dependent differences in firing rates are widely distributed throughout the cortex, while stimulus specific changes in firing rates are more restricted and hierarchically organized. We also show that microsaccades during the memory delay encode the stimuli held in memory, and units modulated by microsaccades are more likely to exhibit stimulus-specificity, suggesting that eye movements contribute to visual short- term memory processes. These results support a framework in which most cortical areas, within a modality, contribute to mnemonic representations at time scales that increase along the cortical hierarchy. Introduction Working memory is essential to cognition. It enables the short-term retention and utilization of behaviorally relevant information for virtually all cognitive tasks (Baddeley, 2003). The neural mechanisms that mediate feature-based visual working memory have been intensively studied for several decades (Luck and Vogel, 2013; Sreenivasan et al., 2014; Lara and Wallis, 2015; Christophel et al., 2017). In non-human primates (NHP), feature-based mnemonic representations - measured as differences in neural firing rates 99 between stimuli held in memory - have been demonstrated in prefrontal (Fuster et al., 1982; Quintana et al., 1988; Miller et al., 1996; Peng et al., 2008; Salazar et al., 2012; Mendoza- Halliday et al., 2014), posterior parietal (Sereno and Maunsell, 1998; Salazar et al., 2012; Sarma et al., 2015), inferotemporal (Fuster and Jervey, 1981; Miyashita and Chang, 1988; Miller et al., 1991, 1993), and extra-striate visual cortical areas (Bisley et al., 2004; Mendoza-Halliday et al., 2014). Functional imaging studies have corroborated and extended these findings in humans (Courtney et al., 1997; Owen et al., 1998; D’Esposito et al., 2000; Postle et al., 2003). However, because our understanding of feature-based visual working memory is derived from studies using different tasks, stimuli and recording techniques, the full extent of the participating circuit and the underlying neural mechanisms remain highly debated. In particular, fMRI-based decoding analyses of the content of visual short-term memory (Harrison and Tong, 2009; Serences et al., 2009; Ester et al., 2015) have revealed involvement of early visual cortex, prompting debate on the nature and role of sensory areas in the storage and maintenance of mnemonic representations. A salient issue is whether feature-based mnemonic representations are present in neural spiking activity in early visual areas (Serences 2016). However, the few studies that have made such measurements have reached conflicting conclusions (Bisley et al., 2004; Zaksas and Pasternak, 2006; Mendoza-Halliday et al., 2014). The role of prefrontal cortex is also heavily debated. Many argue that it is primarily involved in executive functions (Sreenivasan et al., 2014; D’Esposito and Postle, 2015; Lara and Wallis, 2015), while others emphasize a key role in both executive functions and mnemonic representations 100 (Serences, 2016; Hasson et al., 2016; Christophel et al., 2017). Finally, because studies of neural spiking activity typically focus on one cortical area at a time, the relative contribution of widely distributed cortical areas to working memory processes are largely unknown. Resolving these issues is necessary for developing a mechanistic theory of how and where working memory processes are carried out. Here we focus on two basic questions about feature-based visual short-term memory that remain largely unanswered. Which cortical areas are involved in the neural circuit that mediates visual short-term memory and how do the individual components of this circuit differ in their functional roles? To address these questions we performed large- scale microelectrode recordings of neuronal activity in NHPs performing a feature-based visual short-term memory task and determined both the task dependence and stimulus selectivity of the recorded neurons. Results We recorded broadband neuronal activity from a total of 42 cortical areas in two NHPs (Monkeys E and L) while they performed a feature-based delayed match-to-sample (dMTS) task and an interleaved visual fixation task. The dMTS task required the monkeys to remember a centrally presented sample image (1 of 5 possible images) for a minimum of 800 ms (800-1200 ms in Monkey E; 1000-1500 ms in Monkey L), before making a choice between a matching and a non-matching image (Fig. 1A). During the fixation task, occurring on ~10% of the trials, the monkeys simply had to fixate the central target for the same duration as the dMTS task. Data was collected using large-scale semi-chronic recording devices with up to 256 independently moveable microelectrodes (Dotson et al., 101 2015, 2017). Neuronal spiking activity was extracted and sorted off-line. Simultaneous recordings were made from up to 12 and 29 different cortical areas in Monkeys E and L, respectively. The recording locations and sample sizes, combined from both animals, are shown in Figure 1B (Table S1). Sparsely sampled cortical areas, and areas with similar functional properties (e.g., somatosensory areas 1, 2 and 3) were merged with adjacent areas, resulting in a total of 24 different areas/groups combined across monkeys (Table 1; Table S1). Figure 1C shows an example of eye position signals (bottom) and broadband electrophysiological data sampled simultaneously from microelectrodes located in 27 separate cortical areas in Monkey L. Prior to determining the task-dependence and stimulus selectivity of each cortical area, we first identified units that showed modulations in their firing rates locked to microsaccades (Martinez-Conde et al., 2013; see Methods; Table S1; Fig. S1-2). These microsaccade-modulated (MSM) units occurred in 22 of the 24 cortical areas/groups in our sample, but were most common (>10% incidence) in areas V1, V2, DP and 8L (Fig. S2A). Because of the possible confound introduced by eye movement related activity, we excluded all of these MSM units from our analyses unless otherwise stated. Second, because our recordings in areas V1 and V2 spanned a large portion of the retinotopic map, many of the units in these areas had receptive field locations (i.e., >2 deg eccentricity) that prevented them from responding directly to the sample stimuli. We therefore identified units in areas V1 and V2 that displayed a short latency (40-100 ms) excitatory visual response (SLVR) to at least one of the sample stimuli, and analyzed these units separately from the remaining units in V1 and V2 (see Methods; Table S1). 102 Task-Dependent Activity is Widely Distributed The first objective of our analysis was to determine the extent to which each unit differentially participated in the two tasks. We assessed this ‘task-dependence’ by comparing the firing rates during the dMTS task (combined across all 5 stimuli) to the firing rates occurring during the interleaved fixation task, using a 200 ms time bin, stepped every 50 ms (Wilcoxon rank sum test, p<.05, FDR (false discovery rate) corrected). As mentioned above, all MSM units were excluded from this analysis. Figure 2A (I-XII) shows example peri-stimulus time histograms (PSTHs) from units in 9 different cortical areas sampled from both monkeys. Each plot shows the average firing rate for the dMTS (blue) and fixation (red) tasks for an individual unit from the cortical area indicated at the top (significant bins are marked with a black square). In general, the neuronal responses during the two tasks differed substantially, but often in unexpected and heterogeneous ways. For example, the selected units from 9/46V, V2, and V1 (I, X and XII) conformed to our expectation of a relatively stable rate during the fixation trials and a clear task- dependent modulation of rate during the dMTS task. However, we also found instances of the opposite pattern, where units displayed a stable firing rate during the dMTS task and a robust rate modulation during the fixation task (V (F1), VI (3)). Moreover, we found many instances of a more complex multi-phasic relationship between the two tasks. These included cases ranging from a simple push-pull pattern, where a biphasic change in firing rate during one of the tasks was mirrored by a similar but opposite pattern during the other task (II (46v), IV (F7), VII (AIP)), to cases where the rates would diverge early in the tasks and converge back to a similar rate by the end of the tasks (VIII (AIP), IX (V2), XI (V1)). 103 Interestingly, we also found numerous instances of ascending rates during the fixation task in areas of the visual hierarchy as early as V1 (XI). These and many other examples demonstrated that task-dependent changes in rate are widely and heterogeneously distributed across the cortex, and that activity during the fixation task is itself highly dynamic, indicating that this task reflects a distinct cognitive process. To determine the task-dependence for each cortical area/group, we first calculated the incidence of significant differences in firing rates between the two tasks. We then separated the resulting distribution at each time point according to whether each unit’s response during the dMTS task was greater than (enhanced) or less than (suppressed) the response during the fixation task. The plots in Figure 2B and C show the results of these calculations for area vPFC (n=124 units). The incidence of task-dependent differences in activity increased rapidly during the sample period to a peak value near 50% and remained near 40% throughout the delay period (Fig. 2B). The colored bar at the top shows the same data plotted as a heat map. Figure 2C shows that the incidence of enhancement and suppression relative to the fixation task is split roughly equally throughout the task for vPFC. To visualize the ratio of enhancement to suppression, we calculated a rate modulation index (RMI) as a function of time (shown as a heat map in Fig. 2C). This is the percentage difference of enhanced (E) and suppressed (S) units at each time bin (RMI = [(#E-#S) / (#E+#S)]*100). To determine if the incidences of these two processes differed from one another, we tested the two counts at each time bin using a binomial test (p<.05, FDR corrected). In this cortical area (vPFC) we found no significant differences and conclude that there is an equal distribution of activity during the dMTS task where the 104 firing rates are greater or less than the rates occurring during the interleaved fixation task. We refer to this as balanced activity. We applied these calculations to all 24 cortical areas/groups (Fig. S3). The incidence of significant differences in firing rates revealed widespread differential involvement in the two tasks (Fig. 3A). Every area/group we studied demonstrated some degree of task-dependence that could broadly be separated into three types of activity profiles: visually responsive, ramping and sustained. Visually responsive areas showed clear sample related task-dependence, which either tapered off gradually (e.g., V2-SLVR), or remained elevated (e.g., vPFC and AIP/VIP) throughout the delay period. In several other areas, the incidence of task-dependence began ramping up in the middle of the sample period and continued until the end of the delay (e.g., F2 and F7). Finally, many of the areas simply showed weak but sustained task-dependent activity during all task epochs (e.g., dPFC and 7b). Interestingly, primary motor (F1) and somatosensory (1/2/3) areas were also clearly task-dependent. Analysis of the differences in firing rates between the two tasks (enhancement or suppression) revealed a more complex pattern of the task-dependence across the cortex. To visualize this pattern, we plotted the RMI as a function of time for all 24 areas/groups (Fig. 3B). We then tested the two counts at each time bin using a binomial test (p<.05, FDR corrected) and marked each significant bin with a white cross. As expected, we found that SLVR units in V1 and V2 were strongly enhanced during the sample period, but then were suppressed during the late phase of the fixed delay, and returned to a balanced state prior to the match onset (bottom of Fig. 3B). The large majority of units in V1 and V2 (i.e. those 105 having receptive field locations >2 deg eccentricity) were suppressed during both the sample and delay periods, and displayed a rebound enhancement during the match-locked period. A number of other cortical areas also displayed suppression throughout much of the task, including the primary somatosensory areas (1/2/3), and medial posterior parietal areas (7op and 5/MIP). The remaining cortical areas, including most prefrontal (e.g., vPFC and dPFC) and lateral posterior parietal areas (e.g., AIP/VIP and 7b), exhibited a balanced distribution of enhancement and suppression. To further visualize how these relationships are distributed across the cortex, we plotted the occurrence of enhanced, suppressed and balanced activity during the end of the fixed delay period (gray shading and arrows in figure 3B) on a cortical flatmap (Fig. 3C). This revealed an interesting pattern during much of the delay period, where early sensory areas are suppressed and the remaining areas are primarily balanced. Thus, while the responses in association areas (e.g., vPFC) are highly dynamic and heterogeneous between the two tasks, they tend to balance out at the population level. Early sensory areas on the other hand (i.e., V1 and V2, and somatosensory areas 1, 2 and 3) tend to be suppressed at the population level during the delay. Embedded Hierarchy for Mnemonic Representations The previous analysis revealed that task-dependent activity is widely distributed. Do the same areas encode the stimulus identity? To address this question, we determined the time course of stimulus-selectivity during the dMTS task by comparing the firing rates across stimuli in 200ms windows with a 50ms step (Kruskal-Wallis test, p<.05, FDR corrected). As with the previous analysis, all MSM units were excluded from this analysis. Figure 4A (I-XIV) shows example PSTHs with stimulus selective responses from units 106 recorded in 11 different cortical areas, revealing a striking variety of stimulus-specific responses among both association and sensory areas. Example units in vPFC (I-II) and posterior parietal cortex (VI-IX) displayed highly selective activity in response to the different stimuli that lasts throughout the sample and delay periods. Other units in prefrontal areas exhibit transient selectivity that is superimposed on a background of apparent suppression (III) or that ramps up during the delay period (IV-V). Activity in early visual areas V1, V2 and V4 (X-XIV) also displayed transient and sustained periods of selectivity. This occurred even when the delay-period firing rates were quite low (XI) or ramping up during the delay (XII). To characterize the overall behavior of the data, we calculated the incidence of selectivity as a function of time for each of the 24 cortical areas/groups. The result of this analysis for area vPFC is shown in figure 4B. And the results for all areas/groups are shown in figure 5A (also see Fig. S4 for line plots of this data). This analysis revealed a pattern of stimulus selectivity that was sparser compared to the widespread distribution of task- dependence (Fig. 3). Units in V1 and V2 that were excited by the sample stimuli (bottom 2 plots in Fig. 5A, SLVR) displayed a high incidence of stimulus selectivity throughout the sample and well into delay. This delay period selectivity occurred even though the firing rates during this period were typically low and suppressed relative to the fixation task (Fig. 3B, C). Other cortical areas displayed more sustained stimulus selectivity throughout the delay (e.g., vPFC, 8L, AIP/VIP, 7b) (Fig. 5A). And a number of areas showed little or no selectivity throughout the task (e.g., F1, 7op, 5/MIP, 1/2/3, V6A) (Fig. 5A). To further characterize these effects, we calculated the median incidence of selectivity during the late 107 delay period for all areas/groups (gray shading/arrows Fig. 5A). This revealed that areas V1-SLVR, V2-SLVR, V4, LIP, 7b, AIP/VIP, F7, 8L and vPFC have significant delay period selectivity that exceeds an incidence of 5% (Fig. 5B). These results suggest that under the conditions of this experiment a subset of areas contribute to short-term memory maintenance. However, because it was impractical to adjust the position and properties of the sample stimuli to optimally activate each unit simultaneously, differences in the absolute incidence of delay period selectivity may be less informative than expected. Therefore, we chose to analyze the incidence of late delay period selectivity, relative to that occurring throughout the task, in order to determine the contribution of each area to short-term memory maintenance. This measure, illustrated in figure 6A, reveals that the relative incidence of stimulus selectivity during the late delay period is substantially higher in vPFC than V2-SLVR, even though the absolute incidence of selectivity during this period is marginally higher in V2-SLVR (see Fig. 5B). To see how this measure varies across cortex, we plotted the relative incidence of delay period selectivity in rank order for those areas/groups with an absolute incidence of delay period selectivity exceeding 5% (Fig. 6B). This revealed that the relative incidence of stimulus selectivity decays rapidly in early visual areas, while it is more sustained in prefrontal and posterior parietal areas, thereby forming a functional short-term memory hierarchy. The inset in Figure 6B shows an anatomically derived hierarchy from the studies by Markov et al. (2014b) and Chaudhuri et al. (2015) for areas that match or are included in the significant areas/groups. The functional hierarchy agrees well with the anatomical hierarchy. Finally, within this functional hierarchy, we find a high incidence of units with 108 responses that are both stimulus selective and task dependent throughout the task (see Fig. S5). These analyses indicate that the incidence of stimulus-selectivity tends to peak during the sample period and then differentially decays during the delay in a manner that reflects the anatomical hierarchy. We suspected that the latter effect might also be due in part to the recruitment of newly stimulus-selective units during the delay period in areas where this decay is less pronounced (see examples in Fig. 4A, III-V, X). To address this question, we identified the first time bin that a unit became stimulus selective. For instance, if a unit responds selectively to the sample stimulus and then remains selective throughout the task, we only count the first time bin during the sample period that the unit is selective and discard the other time bins from the analysis. This provided us with histograms of the newly recruited stimulus selective units. We normalized these histograms and computed the cumulative sum over time for the areas showing >5% significant delay period selectivity (Fig. 6C). The time when the cumulative sum reaches 1 indicates how long into the dMTS task selective units continued to be recruited. In Figure 6C we see that early visual areas peak almost immediately after the sample onset, indicating that no units are recruited during the delay that weren’t already selective during the sample period (e.g., V1- SLVR and V2-SLVR). Areas higher in the hierarchy contain units that become stimulus selective later in the task throughout the delay period. Figure 6D shows the rank ordered time to the last recruited units (cumulative sum = 1). The results from this analysis match the general hierarchical scheme derived from the relative incidence of delay-period 109 selectivity (Fig. 6B), with V2-SLVR and V1-SLVR low in the hierarchy, LIP, 8L and V4 in the middle, and 7b, vPFC, F7, and AIP/VIP at the highest level. Microsaccades Encode Visual Memories These results demonstrate that widespread cortical areas involved in visual processing and perception contribute to short-term mnemonic representations. What is the relationship between perception and memory? Specifically, are mnemonic representations entirely abstract or do they maintain a semblance of the real image (i.e., a shape). Because the animals routinely made microsaccades (typically <1 dva) to different portions of the sample images (Fig. 1C, bottom plots), we posited that if the mnemonic representations maintain a spatial form, then the eye position during the delay period would encode the remembered image. Figure 7A shows an example of the microsaccade endpoints for two stimuli during a recording session in Monkey L. In the presample period, the microsaccade endpoints are overlapping, while during the sample and delay periods they are largely non- overlapping, supporting our hypothesis. To test for this, we used mutual information analysis of the microsaccade endpoints for the same session. This analysis revealed that stimulus-specific microsaccades occur during the sample and late delay periods of the task (Fig. 7B). The overall incidence of this effect across all recording sessions was similar for both monkeys (Fig. 7C), demonstrating that eye position following microsaccades reliably encodes mnemonic representations during the end of the fixed delay period. This suggests that the monkeys were scrutinizing their short-term visual memories of the sample stimuli. These findings, along with the well-established relationship between microsaccades and neural activity (Martinez-Conde et al., 2013), further imply that 110 microsaccadic eye movements may be an integral part of visual short-term memory. Since we found that microsaccades modulate activity in a large number of the units in our sample (Figs. S1, S2; see methods), we sought to determine if these units contribute differentially to stimulus-selectivity during the task. (It is important to note that all significant MSM units were excluded from all of the previous analyses.) To accomplish this, we compared the incidence of stimulus-selectivity between units that were and those that were not modulated by microsaccadic eye movements (Pearson’s chi-squared test of independence, p<.05, FDR corrected). We restricted our analysis to cortical areas V1, V2 and 8L, which contained a minimum of 25 MSM units (Table S1). We did not analyze the MSM SLVR units in V1 and V2 due to low sample sizes (e.g. 10/98 in V1, and 9/107 in V2). This analysis revealed a higher incidence of stimulus-selectivity in the MSM units (Fig. 7D-F). In V1 and 8L, this effect occurs in the sample period and throughout the fixed delay. In V2, the differences occur primarily around the sample offset. These results indicate that MSM units tend to have a higher incidence of stimulus-selectivity that extends into the delay period of the task. Finally, given the higher incidence of stimulus-selectivity in MSM units, we re-ran the stimulus-selectivity analyses with MSM units included in order to determine if this had an effect on the functional hierarchy. We found no significant differences in the overall incidence of delay-period selectivity and no change in the functional hierarchy as described in Figure 6B and D. This is likely due to the fact that the MSM units only compose a small fraction of the total units. Discussion 111 To elucidate the neural circuit dynamics underlying visual short-term memory, we developed a large-scale microelectrode recording device that encompasses an entire cerebral hemisphere (Dotson et al., 2017), and analyzed recordings from a total of 42 cortical areas in two non-human primates performing a feature-based dMTS task and an interleaved visual fixation task. We find that the cortical circuit defined by the differences in activity between these two tasks (task-dependence) is widely distributed, heterogeneous and dominated by population activity during the dMTS task that is either balanced or suppressed relative to the fixation task. Thus, even a simple cognitive task, such as remembering an item during a brief delay, recruits widespread changes in activity throughout the cortex. Embedded within this large-scale circuit, we identified a functional hierarchy for mnemonic representations. The hierarchy is expressed as an increase in the relative incidence of stimulus-selectivity during the delay, and an increase in the latency in which newly selective units are recruited. This hierarchy extends from early visual cortex to high-level association areas in prefrontal and posterior parietal regions, and closely matches the hierarchy derived from anatomical measurements of feed-forward and feedback connections (Markov et al., 2014b; Chaudhuri et al., 2015). Stimulus selective activity within this hierarchy occurs in conjunction with heterogeneous task related information (see Fig. S5). This apparent mixed selectivity may result in high dimensional encoding of all stimulus and task information at each hierarchical level, similar to what has been observed in prefrontal cortex (Riggoti et. al., 2013). It may also endow these areas with varying degrees of distractor resistance, depending on the concentration and types of mixed selective units (Parthasarathy et al., 2017). We also identified a behavioral correlate 112 of visual short-term memory. Microsaccadic eye movements during the memory delay encode the stimuli held in memory, and units modulated by microsaccades are more likely to exhibit stimulus-specific activity. Methodological Caveats While there are clear advantages to performing large-scale simultaneous recordings of neural activity, our approach also introduced several experimental limitations that likely influenced our findings. The first concerns the sampling of activity. We collected less data in Monkey E than Monkey L, due to an improvement in the design and implementation of the recording methods during the time spanning the two experiments (see Dotson et al., 2017). Consequently, our findings for some cortical areas are based on data from one, but not both, animals. Similarly, the sample sizes for some cortical areas required us to combine data among adjacent areas in order to validate our statistical tests. This is not uncommon in some physiological studies of cortex, but does reduce the specificity of some of our findings. Our experimental design also prevented us from performing population decoding analyses (e.g., Mendoza-Halliday et al., 2014; Parthasarathy et al., 2017), which typically rely on combining data across recording sessions that use the same stimuli or high-density recordings of individual cortical areas. We also were unable to reliably measure neural activity from the ventral temporal visual pathway, specifically areas TEO and TE. This was due to limitations in the design of the device implanted on Monkey E, and some failures in the actuator mechanisms in the device used in Monkey L. Therefore, we were unable to characterize neural activity in a 113 major division of the cortical pathway underlying feature-based vision and short-term memory. Additionally, because we sampled neural activity from many different cortical areas simultaneously, it became impractical to tailor the parameters of the experiment (e.g. the location and properties of the stimuli) to each recording site. Thus, many of the recorded neurons may have been unresponsive or weakly responsive to the stimuli and parameters of the task. This likely led us to underestimate the incidence and magnitude of task dependence and stimulus-selectivity. However, our method also enabled us to obtain an unbiased estimate of the distribution of task-dependent and stimulus-selective activity over widespread areas of cortex that otherwise may have gone undetected. Memory Maintenance in V1 and V2 Our findings help to resolve an ongoing debate regarding the contribution of early visual cortical areas to feature-based visual short-term memory. Functional imaging studies in humans have repeatedly demonstrated the ability to decode short-term memory content from primary and early extrastriate areas of visual cortex (Harrison and Tong, 2009; Serences et al., 2009), even though elevated delay-period activity is largely absent in these areas (Riggall and Postle, 2012). These findings support the concept that mnemonic representations in these areas may be mediated by top-down input (Mendoza-Halliday et al., 2014) and/or changes in synaptic strength induced by the sample stimulus (see Serences, 2016 for review). However, studies investigating memory-related unit activity in early areas of visual cortex have reached somewhat conflicting conclusions with respect to this hypothesis (Super et. al., 2001; Bisley et al., 2004; Lee et al., 2005; Zaksas and 114 Pasternak, 2006; Mendoza-Halliday et al., 2014; Van Kerkoerle et. al., 2017). We find evidence of stimulus-specific spiking activity during feature-based visual short-term memory in V1 and V2. These effects occur at firing rates that are near or below the level of activity measured during the interleaved fixation task (Figs. 2, 3), arguing against the need to postulate a sub-threshold storage mechanism. Decoding Items in Memory with Microsaccades Our findings demonstrate that microsaccades provide a behavioral readout of the stimuli held in memory. These results suggest a tight link between perception and short- term memory maintenance. Mnemonic representations may maintain spatial relationships similar to the perceived images, enabling the monkeys to scrutinize these representations. Similar results have been observed in human subjects imagining a previously seen image (Brandt and Stark, 1997; Laeng and Teodorescu, 2002). This oculomotor behavior may result in a sensory-motor feedback loop that facilitates the maintenance of visual working memories. This is supported by our finding of a higher incidence of stimulus-selectivity in microsaccade modulated units in areas V1, V2 and 8L. Interestingly, we find this activity among units in V1 and V2 with receptive fields that likely do not overlap with the sample image (i.e., the non-SLVR units). This may help explain why decoding in functional imaging studies can be done in parts of visual cortex outside of where the sample was presented and even in contralateral areas of cortex (Ester et al., 2009). However, because of our experimental design, we were unable to completely separate the role of delay period selectivity from the activity evoked by microsaccadic eye movements. While it is possible that microsaccades are an integral component of short-term memory, they may also operate 115 in parallel and exert little or no influence on mnemonic representations. Future studies that dissociate these roles will be necessary to determine the full relationship between eye movements and short-term memory. The Balanced Activity State and Short-Term Memory A prominent result in the task-dependence analysis is that the incidence of enhanced and suppressed activity is typically balanced in association areas and suppressed in sensory areas during the delay period. This may be linked to the neural mechanisms underlying stimulus-specific persistent activity. We see that a fundamental difference arises at either end of the hierarchy. In early visual areas the relative incidence of stimulus- specific unit activity decays rapidly and selective unit recruitment ceases after the sample presentation, while high-level association areas maintain a higher relative incidence of stimulus-selectivity and continue to recruit selective units well into the delay period. These differences match the pattern of balanced and suppressed task-dependent activity. Collectively these findings suggest that the balance of enhancement and suppression that occurs during the dMTS task, relative to the interleaved fixation task, is a signature of higher order cortical areas and may contribute to the maintenance of mnemonic representations and short-term memory in general. Stort-term Memories are Maintained in a Hierarchy of Cortical Areas Our analysis of the relative incidence and onset of delay period selectivity enabled us to identify a functional hierarchy for mnemonic representations. What mechanism/s produce the functional hierarchy? Functional imaging (Honey et al., 2012), neurophysiological (Murray et al., 2014), and modeling studies (Chaudhuri et al., 2015) 116 have all identified intrinsic time-scales with a hierarchical ordering. One interpretation of these findings is that they provide a hierarchy of temporal receptive windows of increasing size that enable the accumulation and integration of information over increasing periods of time (Hasson et al., 2016). Our findings support this interpretation and are consistent with a framework in which most cortical areas, within a modality, contribute to mnemonic representations at time scales that increase in a hierarchical manner. This framework provides a parsimonious explanation for any cognitive task that requires information to be gathered, combined, and remembered. 117 Tables Table 1. List of areas contained in each group. The first column indicates the group name and the second column indicates the areas that compose each group. Many of the groups are simply one area. Group Areas Orb. PFC OPRO, 11, 12, 13, 14, 32 24c/24d 24c, 24d vPFC 9/46v, 44, 45A, 45B, 46v dPFC 8B, 9/46d, 9, 46d 8L 8L 8M 8M 8r 8r F2 F2 F7 F7 F1 F1 7op 7op 5/MIP 5, MIP PIP PIP 7a/TPt 7a, TPt AIP/VIP AIP, VIP 7b 7b LIP LIP 1/2/3 1, 2, 3 DP DP V6A V6A MT MT V4 V4 V2 V2 V1 V1 118 Table S1. Sampling distribution of unit activity (related to STAR methods). The first column contains the group names. The second column contains the name of each cortical area. Area names follow Markov et al. (2014a). In the remaining columns there are two rows. The data in the top and bottom rows are from Monkey L and Monkey E, respectively. The number of units (# units) is the total number of units recorded (third column). The number of single units (SUA) is the number of SUA out of the total number of units (fourth column). The number of microsaccade-modulated units (MSM) is the number of microsaccade-modulated units out of the total number of units (fifth column). The number of units with a short latency visual response was only determined for V1 and V2 and the number of units is out of the total number of units (sixth column). 119 Group Area # Units # SUA MSM SLVR 5 0 0 OPRO 0 0 0 16 3 0 11 0 0 0 Orb. 12 10 2 1 PFC 0 0 0 21 3 0 13 0 0 0 7 1 0 14 0 0 0 19 1 0 32 0 0 0 75 4 1 24c 2 0 0 24c/24d 14 0 0 24d 6 1 0 5 1 0 9/46v 4 0 0 vPFC 45 0 0 44 0 0 0 120 14 0 0 45A 0 0 0 14 1 0 45B 2 0 0 40 6 0 46v 0 0 0 131 14 7 8B 5 1 0 34 0 0 9/46d 6 0 0 dPFC 0 0 0 9 19 4 0 54 1 0 46d 1 0 0 126 6 25 8L 8L 7 1 0 119 10 4 8M 8M 1 0 0 8r 8r 48 1 2 121 0 0 0 364 47 22 F2 F2 0 0 0 107 8 4 F7 F7 24 7 1 248 25 10 F1 F1 54 15 3 83 6 0 7op 7op 0 0 0 230 24 3 5 9 3 0 5/MIP 13 1 0 MIP 1 0 0 64 4 0 PIP PIP 0 0 0 7a/TPt 7a 60 2 0 122 3 1 0 16 1 1 TPt 2 0 1 64 9 4 AIP 0 0 0 AIP/VIP 10 0 1 VIP 0 0 0 80 8 1 7b 7b 15 1 0 102 9 5 LIP LIP 0 0 0 60 0 6 1 4 0 0 51 4 4 1/2/3 2 16 1 0 123 3 4 3 7 0 0 DP DP 96 6 13 123 16 2 2 69 4 1 V6A V6A 22 4 0 82 4 3 MT MT 2 0 0 24 1 0 V4 V4 32 1 1 354 4 104 67 V2 V2 27 1 3 5 530 16 98 41 V1 V1 71 4 0 9 124 Figures Figure 1. Behavioral task, distribution of recording sites, and raw data collected on a single trial. (A) Schematic of the feature-based delayed match-to-sample task. The monkeys maintained their gaze within a 3° window (dashed white circle) until the match period. A sample stimulus, randomly drawn from a set of 5 possible images, is presented for 500 ms. Following a variable delay, the fixation target is extinguished, and the match stimulus presented. The match consists of the previous sample image and 1 of the 4 non- sample images presented at 6° eccentricity on either side of the fixation target (see Methods). The monkeys are rewarded by making a saccadic eye movement to the sample image. (B) Cortical flat map showing the distribution of recorded units in each cortical area. The inset shows the same data on an inflated brain. (C) Example of the raw broadband data recorded on a single trial of the dMTS task from 27 separate cortical areas in Monkey L. The names of each cortical area are shown on the left. The vertical and horizontal eye position signals are shown at the bottom. The vertical lines indicate the times of the sample and the match onset, respectively. The arrows during the delay period indicate 2 microsaccadic eye movements. The arrow during the match period indicates the time of the choice. 125 Figure 6. Task-dependent changes in firing rates. (A, I-XII) Example PSTHs from units in 9 different cortical areas illustrating the average neuronal responses during the dMTS (blue) and interleaved fixation tasks (red). Black squares at the top of each plot indicate the timing of significant differences in firing rates between the two tasks. (B) Incidence of significant task-dependent activity from all recordings in vPFC in both monkeys as a function of time. The colored bar at the top shows the same data as a heat map. (C) The same data as in B separated into values in which the responses to the dMTS task are greater (enhanced) or less (suppressed) than those occurring during the fixation task. The colored bar shows the heat map of the response modulation index. There were no significant differences in the incidence of enhanced and suppressed responses in vPFC. The labels P, S, D, and M indicate the presample, sample, delay, and match locked periods, respectively, and denote the same meaning in figures 3-7 and S3-S5. 126 Figure 3. Task-dependent unit activity is widely distributed. (A) Heat maps showing the time course of the incidence of task-dependent unit activity for each of the 24 areas/groups sampled from both monkeys. The bottom two plots show the data for the units in areas V1 and V2 that displayed short latency visual responses (SLVR) to at least one of the sample stimuli. Area/group names and number of units recorded are shown on the left. (B) Heat maps of the rate modulation index for each area/group over the course of the task. Time bins with an incidence of task-dependence <5% are colored black. Positive (warm colors) and negative (cool colors) values indicate that more units have firing rates during the dMTS task that are greater (enhanced) or less (suppressed) than the firing rates occurring during the fixation task, respectively. White plus signs mark the bins when the number of enhanced or suppressed units is significantly different. (C) Flat map showing the general activity pattern during the end of the fixed delay (arrows and gray bars in (B)). 127 Figure 7. Stimulus selective changes in firing rates during the dMTS task. (A, I-XIV) Example PSTHs from 11 different cortical areas illustrating the average neuronal responses to each of the 5 sample stimuli during the dMTS task. The response to each stimulus is plotted in a different color (bottom left). Black squares along the top of each plot mark bins where the firing rates are significantly different across stimuli. (B) Incidence of significant stimulus-selective activity from all recordings in vPFC in both monkeys as a function of time. The colored bar at the top shows the same data plotted as a heat map. 128 Figure 8. Distribution of stimulus selectivity across the sampled cortical areas in both monkeys. (A) Heat maps of the incidence of significant stimulus-selective activity for each area/group over the course of the task. The bottom two plots show the data for the SLVR units in areas V1 and V2. Area/group names and number of units are shown on the left. (B) Median incidence of stimulus selective activity during the 800 ms period marked by the gray shaded region and arrows (top and bottom) in A. The dashed line marks the 5% value. 129 Figure 9. Embedded hierarchy for mnemonic representations. (A) Examples of the incidence of stimulus-specific activity in areas vPFC (top) and V2-SLVR (bottom) (normalized for display purposes). The shaded regions indicate the relative incidence during the late delay period (area under the curve in %). (B) Rank ordered plot of the relative incidence of stimulus-specific activity during the late delay period, split into three categories: low (blue), mid (cyan) and high (red). The inset shows the anatomical hierarchy derived from Markov et al. (2014b) and Chaudhuri et al. (2015). Areas in the hierarchy are colored based on the same scheme. (C) Plots of the cumulative sum of the first occurrence of stimulus-specific activity for each of the areas in B. Arrows at the top indicate the time points where the sum equals 1. The color scheme is the same as in B. (D) Rank ordered plot of the time (vertical axis) when the cumulative sum in C reached 1 for each cortical area. 130 Figure 7. Microsaccades encode the sample stimulus during the delay period. (A) Example of microsaccade endpoints (stimulus 1: red dots; stimulus 2: blue dots) during the presample, sample and delay period for a single recording session from Monkey L. (B) Example of the MI analysis (same data as A, except all 5 stimuli are used). The MI was bias corrected using the mean of the surrogate distribution. The red dashed line indicates the 95th percentile of the surrogate distribution (also bias corrected). (C) Summary of the MI analysis, locked to the earliest possible match for both monkeys (Monkey E: green; Monkey L: blue). (D-F) Incidence of stimulus-specific activity in V1 (D), V2 (E), and 8L (F) for the microsaccade modulated units (blue) and the non- microsaccade modulated units (black) over the course of the task. Red diamonds mark the bins where the incidence values are significantly different. The numbers in parentheses indicate the sample sizes for the two plots. 131 Figure S1. Microsaccade modulation analysis (related to STAR methods and Figure 7). (A, B) Main sequence (magnitude vs. velocity) for all of the microsaccades detected during the fixation trials for Monkey L (A) and Monkey E (B). The x-axis indicates the magnitude (degrees of visual angle, dva) of each microsaccade and the y-axis indicates the velocity (dva/s). Gray diagonal lines show the least-squares fit to the data. R-Squared values are provided at the bottom of each plot. (C-E) Examples of the microsaccade modulation analysis for units in 8L, V2, and V1. Red dashed lines show the 2.5th and 97.5th percentiles of the surrogate distributions for reference only (actual significance testing was performed as described above). 132 Figure S2. Microsaccade modulation analysis results (related to STAR methods and Figure 7). (A) Percentage of units in each area/group in which the unit activity was significantly modulated by microsaccades. (B-D) Plots showing the percentage of units significantly modulated by microsaccades, relative to the time of occurrence of the microsaccade at 0 ms, in areas 8L, V2 and V1 (SLVR units are included). The blue and red lines indicate the timing of increases and decreases in firing rate, respectively. 133 Figure S3. Task-dependence analysis for all 24 areas/groups over the course of the task (related to Figure 3). Each pair of plots shows the incidence of significant task- dependent unit activity on the left and the incidence of enhanced or suppressed unit activity on the right. Vertical gray lines mark the onset and offset of the sample. Vertical dashed gray lines mark the end of the fixed delay period. Red asterisks indicate the bins when the number of enhanced or suppressed units is significantly different. The task- dependence for the short latency visual response units in V1 and V2 (V1-SLVR and V2- SLVR, bottom right) is shown in gray. Light red asterisks at bottom of the V1 and V2 plots indicate the bins when the number of enhanced or suppressed units is significantly different for the SLVR units. The labels P, S, D, and M in the lower left indicate the presample, sample, delay, and match locked periods, respectively. Area/group names are shown at the top of each pair of plots. Sample sizes are given in parentheses. 134 Figure S4. Stimulus-selectivity analysis for all 24 areas/groups over the course of the task (related to Figure 5). Each plot shows the incidence of significant stimulus-selective activity. Vertical gray lines mark the onset and offset of the sample, and the vertical dashed gray lines mark the end of the fixed delay period. Units in V1 and V2 with a short latency visual response are shown in red. The labels P, S, D, and M at the bottom left indicate the presample, sample, delay, and match locked periods, respectively. 135 Figure S5. Incidence of stimulus-selective activity (solid lines) and the incidence of both stimulus-selective and task-dependent activity (dashed lines) for the 9 areas/groups in the hierarchy (related to Figure 6). The incidence of both is simply the overlap between the two analyses. Specifically, for each unit, at each time bin we determined if the activity was both stimulus-selective and task-dependent. Only the V1 and V2 units with a short latency visual response were considered (red). Vertical gray lines mark the onset and offset of the sample, and the vertical dashed gray lines mark the end of the fixed delay period. The labels P, S, D, and M at the bottom left indicate the presample, sample, delay, and match locked periods, respectively. 136 Supplementary Material Experimental Model and Subject Details Subjects: Data was collected from two female macaque monkeys (Monkey E and Monkey L) while microelectrode recordings were performed using a large-scale microdrive system (Dotson et al., 2017). Further details of the recording technique are provided below. All procedures were performed in accordance with NIH guidelines and the Institutional Animal Care and Use Committee of Montana State University. Method Details Behavioral Task: The monkeys were seated in a primate chair (Gray Matter Research, LLC), head fixed using a cranial head post (Gray Matter Research, LLC), and positioned 57 cm from a 19-inch monitor. MonkeyLogic software was used to run the experiment and record eye position data (Asaad and Eskandar, 2008a,b). Eye position data was acquired using an infrared eye-tracking system (240 Hz; ISCAN, Inc.) and converted to degrees of visual angle (dva) in MonkeyLogic. The monkeys were trained to perform a feature-based delayed match-to-sample task (Fig. 1A). A trial begins when the monkey acquires and holds fixation on a small fixation spot (fixation window diameter = 3 dva). At a latency of 500 ms for Monkey E or 800 ms for Monkey L, one of five possible sample images (size: 2.4x2.4 dva) is presented for 500 ms in the center of the screen (obscuring the fixation point). During the sample period the monkey has to maintain it’s gaze in the same 3 dva window used during the fixation period. Sample images were pseudo-randomly chosen from a pool of 40 or 100 images for Monkey E or 137 Monkey L, respectively. The animals were familiar with all images. The sample stimulus is followed by a randomized delay, 800-1200 ms for Monkey E and 1000-1500 ms for Monkey L, in which no stimulus is present. During this time the monkeys must maintain gaze on the central fixation point. At the end of the delay period, the fixation target is extinguished and the matching image and a non-matching image (one of the four other images chosen randomly) appear 5 dva from the center of the screen. For Monkey E, the match and non-match were always placed across from each other on the horizontal plane. The location (left or right) of the match and non-match were randomized on each trial. For Monkey L, the images were aligned either vertically, horizontally or diagonally. The location of the match and non-match and the alignment was randomly chosen on each trial. While the match image is visible, the monkey must make a saccadic eye movement to the matching image and maintain fixation for a brief period of time (200 ms for Monkey E and 500 ms for Monkey L). Correct trials were rewarded with a drop of juice. Approximately 10% of the trials did not include stimuli and the monkey simply had to maintain visual fixation on the central spot throughout the trial to receive a reward. In order to easily compare these trials with the match-to-sample trials, we used the same timing structure, except that there was no sample image presented and no match period. The animals completed >500 correct trials on each session, with behavioral performance typically > 75%. Data used in this report are from 25 and 62 recording sessions from Monkey E and Monkey L, respectively. Recording Techniques: Each animal was implanted with a custom made large-scale recording system containing independently movable microelectrodes spanning the length 138 and width of an entire hemisphere (Dotson et al., 2017). Each system consists of a form- fitting recording chamber and a microdrive composed of a guide array, an actuator block, a printed circuit board (PCB), and a screw guide. Linear actuators (256 for Monkey E, 252 for Monkey L) are housed in the actuator block, with a separation of 2.34 mm. Each actuator consists of a miniature stainless steel lead screw, a threaded brass shuttle, and a compression spring. For Monkey E, each actuator provided 20 mm of microelectrode travel at a resolution of 8 turns/mm. For Monkey L, actuators had 33 mm or 41 mm of microelectrode travel at a resolution of 5 turns/mm. The recording systems remained on the animals throughout the entire experiment. Once the monkeys reached criterion performance on the tasks, we carried out a multi-step implantation sequence and began neural recordings when the animals were fully recovered, healthy and performing the task normally. We gradually moved all of the functioning microelectrodes through the dura and into the cortex over a period of 2-4 weeks. This was done in an incremental manner by advancing a subset of 10-30 microelectrodes each day until unit activity was first detected. Once the recording phase of the experiment began, we made small incremental advancements (~50-500 µm) to a varying subset of electrodes on each recording session. We routinely measured microelectrode impedance and ceased advancing a microelectrode whenever its impedance was < ~50 kΩ or > ~2 MΩ. We attempted to adjust the high impedance microelectrodes and recover the signal. If this failed, we considered the actuator or microelectrode to be damaged and did not move these microelectrodes further. 139 We carried out daily recording sessions 3-5 days/week over a period of ~6 and ~9 months, in monkeys E and L, respectively. Broadband neuronal activity was recorded simultaneously from all viable microelectrodes (0.1 Hz - 9 kHz, sampled at 32 kHz) using a Digital Lynx SX recording system (Neuralynx, Inc.). The reference and ground connections were tied together and connected to the chamber. This created a distributed reference signal. Histology: Recording locations were determined by combining records of the microelectrode depths and histological information. When recordings were completed, small electrolytic lesions were made at the tip of all functioning microelectrodes (10μA DC for 25s). The animals were then euthanized (Pentobarbital, 100mg/kg; i.v.) and perfused through the heart with phosphate buffered saline (PBS) followed by a solution of 5% paraformaldehyde in PBS. After perfusing the animals, we removed the recording systems without retracting the microelectrodes. We used slightly different procedures for Monkey E and Monkey L to perform the reconstructions. For Monkey E, following perfusion, the brain was removed, and sunk in a solution of fixative with 30% sucrose several days before being sectioned (60 µm) and stained for Nissl substance (FD Neurotechnologies, Inc.). To reconstruct the brain, the stained sections were photographed and then imported into Free-D (Andrey and Maurin, 2005). Sections were manually registered and then electrolytic lesions and microelectrode tracks were marked on the images. This information provided a 3D reconstruction of all the microelectrode tracks. We used this information and the record of microelectrode depths 140 to estimate the microelectrode tip position and identify the recording locations of each microelectrode on each recording session. For Monkey L, following perfusion, we removed the top of the skull and then cut the brain in half in the coronal plane at Bregma -15 mm. This ensured that during the sectioning process each slice was in the coronal plane and gave us a crude estimate of the anterior-posterior position of each slice. Each half of the brain was then sunk in a solution of fixative with 30% sucrose for several days before being sectioned (60 µm) and stained for Nissl substance (FD Neurotechnologies, Inc.). During sectioning, we photographed the frozen block-face of the brain in order to preserve the shape of each slice. Stained sections were used to identify electrolytic lesions and microelectrode tracks. The block-face photographs were imported into the Computerized Anatomical Reconstruction and Editing Toolkit (Caret) (Van Essen, 2012). Each photograph was traced and then annotated with information about the location of electrolytic lesions and microelectrode tracks from the Nissl-stained slices. Then, as with Monkey E, we used the 3D reconstruction and the record of microelectrode depths to estimate the microelectrode tip position and identify the recording locations of each microelectrode on each recording session. For both monkeys, we identified the cortical area or subcortical nucleus for each recording site by comparing the histological reconstructions to the atlas published by Markov et al. (2014a). The flat- maps in Figures 1 and 2 were created using the Scalable Brain Atlas website (Van Essen, 2012). Anatomical Hierarchy: The anatomical hierarchy in Figure 6 was derived from Markov et al. (2014b) and Chaudhuri et al. (2015). Areas V1, V2, V4, 8L, 9/46V, 7b, and F7 were 141 ordered based on results reported by Chaudhuri et al. (2015). Area LIP was placed above 8L based on the study by Markov et al. (2014b). Spike Sorting: The technique for spike sorting follows earlier studies (Salazar et al., 2012; Dotson et al., 2014). First, broadband signals (sampled at 32kHz) were highpass filtered (Monkey E: 500 Hz – 9 kHz; Monkey L: 500 Hz – 4 kHz). Second, a threshold of 5 standard deviations of the background signal was used to identify spikes. 32 data points were saved for each spike (11 points before and 21 points after and including the minimum). Waveforms were clustered using KlustaKwik (Rossant et al., 2016). Clusters were merged and artifacts were discarded using MClust (http://redishlab.neuroscience.umn.edu/MClust/MClust.html). To be considered a single unit (SUA), waveforms in the cluster were required to be stable over time, non-overlapping with all other clusters, and have an inter-spike interval histogram with a clear refractory period. Selection of Units: Only units with ≥ 1Hz average firing rates were included in the analysis, in order to ensure a sufficient amount of activity to perform the firing rate analyses. Also, only areas with a large number of units or ones that could be reasonably pooled with adjacent cortical areas were included. Subsequently, here we report on a lower number of cortical areas than were actually recorded from (Dotson et. al., 2017). The average firing rate was calculated for each unit using all correct working memory trials, from the presample period to the match onset. Units were considered to have a short latency visual response (SLVR) if they demonstrated a large change in firing rate within 50-100 142 ms after the sample onset. This indicated that their receptive fields were likely within the region covered by the sample stimulus. This was necessary because recordings were made over large areas of V1 and V2. We analyzed these units separately. See Table S1 for the total number of units and the number of units with a short latency visual response in each area. Quantification and Statistical Analysis Firing Rate Analyses: All data analysis was performed using Matlab with both custom and built-in code. Only correct trials were analyzed. Results were similar across the two monkeys, so data was pooled. Areas with a low number of sampled units were pooled with adjacent areas to form a total of 24 areas or areal groups (Table S1). Firing rate analyses were performed on each unit, over time, using 200 ms time bins (±100 ms from center of bin), stepped every 50 ms. Time bins went from 250 ms prior to the sample onset to 700 ms after the sample offset (bins 1 to 30), and from 300 ms to 100 ms prior to the match onset (bins 31 to 35). For each analysis, we performed a false discovery rate (FDR) correction over all time bins for each unit individually using the Benjamini-Hochberg procedure (Benjamini and Hochberg, 1995). To confirm that the false discovery rate correction was appropriate, we examined the pre-sample period for each area/group. This revealed that during both the task-dependent and stimulus-specific analyses, the incidence of significance was nearly always below 5% at all time bins during the pre-sample period (Figs 3, 5, S3, S4). The task-dependence of each unit, at each time bin, was determined by comparing the firing rates during the match-to-sample trials to the firing rates during fixation trials, 143 using the Wilcoxon rank sum test (p<.05). For each area/group, we calculated the incidence at each time bin by dividing the number of significant observations by the number of units. We determined if the differences in firing rates during the dMTS task were enhanced (increased) or suppressed (decreased) with respect to the fixation task by performing a binomial test on the counts of enhanced and suppressed units (p<.05). Only time bins that had an incidence of task-dependence >5% were tested. To visualize these results, we calculated the percentage difference of enhanced and suppressed units [((number enhanced - number suppressed) / (number enhanced + number suppressed))*100]. We refer to this measure as the rate modulation index (Fig. 2B). The stimulus-selectivity of each unit, at each time bin, was determined by comparing the firing rates across stimuli, using the Kruskal-Wallis test (p<.05). For each area/group, we calculated the incidence at each time bin by dividing the number of significant observations by the number of units. To calculate the normalized incidence we simply divided the incidence at each time bin by the sum across all time bins. We then summed the time bins from 500 to 700ms (bins 26-30) after the sample offset and the match locked period (bins 31-35) to determine the relative incidence during the delay period. To determine the incidence of units that are both task-dependent and stimulus- selective (Fig. S5) we simply found the overlap between the two analyses. Specifically, for each unit, at each time bin we determined if the activity was both stimulus-selective and task-dependent. To identify how long into the task units were recruited, we made a histogram for each cortical area/group of the first time bin that units were stimulus selective. We excluded 144 the presample period (first four time bins) from the analysis. We then calculated the cumulative sum for visualization and to identify the last time bin that new units were recruited. When the cumulative sum reaches one that signifies that all of the units have been recruited (Fig. 6C). To determine if the incidence of stimulus-selectivity in the microsaccade- modulated units was different from the non-microsaccade modulated units, we performed a Pearson’s chi-squared test of independence at each time bin (p<.05). Only a small number of the V1 and V2 microsaccade-modulated units were identified as having a short latency visual response (10/98 in V1, and 9/107 in V2). So, we chose to only analyze the V1 and V2 units without short latency responses. For each area, we performed a false discovery rate correction over all time bins using the Benjamini-Hochberg procedure (Benjamini and Hochberg, 1995). Microsaccade Detection Procedure: We used an eye movement velocity threshold of 10 dva/s to detect microsaccades. The eye position signals were first lowpass filtered (0- 40 Hz) to remove noise. The microsaccades had to last longer than 10 ms and must have occurred at least 50 ms after the previous one (typically they had much longer separations). Since the animals were required to maintain gaze within 1.5 dva of the central fixation dot, the microsaccades were typically < 1dva. Microsaccade Modulation Analysis: To identify microsaccade-modulated units, we computed microsaccade locked peri-event time histograms (±200 ms from microsaccade onset, 20 ms non-overlapping bins) using the fixation trials. Since fixation trials used the same timing structure as the match-to-sample trials, we used the “sample offset” time as a 145 reference point for the analysis. Only microsaccades with an onset time >200 ms and < 600 ms after sample offset were used (Figure S1A, B shows the main sequences for both animals). To determine if neural activity was significantly modulated by the eye movements, we compared the observed average firing rates to surrogate distributions (p<.05, two-tailed test). Surrogate distributions were computed by randomizing the trials with respect to the microsaccade times. Figure S1C-E shows examples of units in areas 8L, V2 and V1 that are modulated by the microsaccadic eye movements. We computed 100 surrogates, and then fit a Poisson function at each time bin in order to estimate p-values less than 0.01 (smallest p-value using just the surrogate distributions is 1/100 = 0.01). Each unit was false discovery rate corrected using the Benjamini-Hochberg procedure (Benjamini and Hochberg, 1995). A unit was considered to be microsaccade modulated if >1 bin was significant. Figure S2A shows the percentage of units in each cortical area/group that are microsaccade-modulated. Figure S2B-D shows when units in areas 8L, V2, and V1 were modulated with respect to a microsaccade, and if the firing rate was above or below the surrogate distribution. The modulations in area 8L occurred around the time of the microsaccade, consistent with this area’s involvement in generating saccadic eye movements. In V1 and V2 there is a suppression followed by an enhancement. These dynamics agree with previous studies (Martinez-Conde et al., 2013) and are likely generated by motor signals to visual cortex rather than retinal input. Microsaccade Pattern Analysis: To determine if the microsaccade patterns that occur during the dMTS task are stimulus specific, we used mutual information (MI) analysis of the microsaccade endpoints. MI was calculated using time bins with a 200 ms window 146 (±100 ms from center of bin), stepped every 50 ms. Time bins went from 250 ms prior to the sample onset to 700 ms after the sample offset (bins 1 to 30). At each time bin, we binned the spatial location of microsaccade endpoints into a 10x10 non-overlapping grid that covered 3x3 dva, and then converted these data into a 1D array. This enabled us to easily calculate the mutual information. Mutual information was calculated as follows: 𝑀𝐼(𝑆; 𝑅) = ∑L,M 𝑃(𝑟)𝑃(𝑠|𝑟)𝑙𝑜𝑔 K(L|M) * K(L) where, P(r) is the probability of observing the response r (microsaccade endpoint), from the response set R (all possible responses); P(s) is the probability of stimulus s being presented, from the set of stimuli S; P(s|r) is the posterior probability that the stimulus s was presented given the response r. For each bin, only one microsaccade was allowed per trial. Since a response did not always occur in every bin and every trial, P(s) was adjusted accordingly. This allowed us to determine the information gained about the sample image based on knowing the eye position. We assessed statistical significance by comparing the observed MI values to surrogate distributions (p<.05). Surrogates were created by randomizing the trial labels and then computing MI. The same number of trials was kept for each label. Each surrogate distribution (1000 surrogates) was fit with a generalized extreme value function in order to estimate p-values less than 0.001 (smallest p-value using just the surrogate distribution is 1/1000 = 0.001). Each individual recording session was false discovery rate corrected using the Benjamini-Hochberg procedure (Benjamini and Hochberg, 1995). 147 References Andrey, P. and Maurin, Y., 2005. Free-D: an integrated environment for three- dimensional reconstruction from serial sections. Journal of neuroscience methods, 145(1-2), pp.233-244. Asaad, W.F. and Eskandar, E.N., 2008a. Achieving behavioral control with millisecond resolution in a high-level programming environment. Journal of neuroscience methods, 173(2), pp.235-240. Asaad, W.F. and Eskandar, E.N., 2008b. A flexible software tool for temporally-precise behavioral control in Matlab. Journal of neuroscience methods, 174(2), pp.245- 258. Baddeley, A., 2003. Working memory: looking back and looking forward. Nature reviews neuroscience, 4(10), pp.829-839. Benjamini, Y. and Hochberg, Y., 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal statistical society: series B (Methodological), 57(1), pp.289-300. Bisley, J.W., Zaksas, D., Droll, J.A. and Pasternak, T., 2004. Activity of neurons in cortical area MT during a memory for motion task. Journal of neurophysiology, 91(1), pp.286-300. Brandt, S.A. and Stark, L.W., 1997. Spontaneous eye movements during visual imagery reflect the content of the visual scene. Journal of cognitive neuroscience, 9(1), pp.27-38. Chaudhuri, R., Knoblauch, K., Gariel, M.A., Kennedy, H. and Wang, X.J., 2015. A large- scale circuit mechanism for hierarchical dynamical processing in the primate cortex. Neuron, 88(2), pp.419-431. Christophel, T.B., Klink, P.C., Spitzer, B., Roelfsema, P.R. and Haynes, J.D., 2017. The distributed nature of working memory. Trends in cognitive sciences, 21(2), pp.111- 124. Courtney, S.M., Ungerleider, L.G., Keil, K. and Haxby, J.V., 1997. Transient and sustained activity in a distributed neural system for human working memory. Nature, 386(6625), pp.608-611. 148 D’Esposito, M., Postle, B.R. and Rypma, B., 2000. Prefrontal cortical contributions to working memory: evidence from event-related fMRI studies. In Executive control and the frontal lobe: Current issues (pp. 3-11). Springer, Berlin, Heidelberg. D'Esposito, M. and Postle, B.R., 2015. The cognitive neuroscience of working memory. Annual review of psychology, 66. Dotson, N.M., Salazar, R.F. and Gray, C.M., 2014. Frontoparietal correlation dynamics reveal interplay between integration and segregation during visual working memory. Journal of Neuroscience, 34(41), pp.13600-13613. Dotson, N.M., Goodell, B., Salazar, R.F., Hoffman, S.J. and Gray, C.M., 2015. Methods, caveats and the future of large-scale microelectrode recordings in the non-human primate. Frontiers in systems neuroscience, 9, p.149. Dotson, N.M., Hoffman, S.J., Goodell, B. and Gray, C.M., 2017. A large-scale semi- chronic microdrive recording system for non-human primates. Neuron, 96(4), pp.769-782. Ester, E.F., Serences, J.T. and Awh, E., 2009. Spatially global representations in human primary visual cortex during working memory maintenance. Journal of Neuroscience, 29(48), pp.15258-15265. Ester, E.F., Sprague, T.C. and Serences, J.T., 2015. Parietal and frontal cortex encode stimulus-specific mnemonic representations during visual working memory. Neuron, 87(4), pp.893-905. Fuster, J.M. and Alexander, G.E., 1971. Neuron activity related to short-term memory. Science, 173(3997), pp.652-654. Fuster, J.M. and Jervey, J.P., 1981. Inferotemporal neurons distinguish and retain behaviorally relevant features of visual stimuli. Science, 212(4497), pp.952-955. Guo, Z.V., Inagaki, H.K., Daie, K., Druckmann, S., Gerfen, C.R. and Svoboda, K., 2017. Maintenance of persistent activity in a frontal thalamocortical loop. Nature, 545(7653), pp.181-186. Hasson, U., Chen, J. and Honey, C.J., 2015. Hierarchical process memory: memory as an integral component of information processing. Trends in cognitive sciences, 19(6), pp.304-313. Harrison, S.A. and Tong, F., 2009. Decoding reveals the contents of visual working memory in early visual areas. Nature, 458(7238), pp.632-635. 149 Honey, C.J., Thesen, T., Donner, T.H., Silbert, L.J., Carlson, C.E., Devinsky, O., Doyle, W.K., Rubin, N., Heeger, D.J. and Hasson, U., 2012. Slow cortical dynamics and the accumulation of information over long timescales. Neuron, 76(2), pp.423-434. Laeng, B. and Teodorescu, D.S., 2002. Eye scanpaths during visual imagery reenact those of perception of the same visual scene. Cognitive Science, 26(2), pp.207-231. Lara, A.H. and Wallis, J.D., 2015. The role of prefrontal cortex in working memory: a mini review. Frontiers in systems neuroscience, 9, p.173. Lee, H., Simpson, G.V., Logothetis, N.K. and Rainer, G., 2005. Phase locking of single neuron activity to theta oscillations during working memory in monkey extrastriate visual cortex. Neuron, 45(1), pp.147-156. Luck, S.J. and Vogel, E.K., 2013. Visual working memory capacity: from psychophysics and neurobiology to individual differences. Trends in cognitive sciences, 17(8), pp.391-400. Markov, N.T., Ercsey-Ravasz, M.M., Ribeiro Gomes, A.R., Lamy, C., Magrou, L., Vezoli, J., Misery, P., Falchier, A., Quilodran, R., Gariel, M.A. and Sallet, J., 2014a. A weighted and directed interareal connectivity matrix for macaque cerebral cortex. Cerebral cortex, 24(1), pp.17-36. Markov, N.T., Vezoli, J., Chameau, P., Falchier, A., Quilodran, R., Huissoud, C., Lamy, C., Misery, P., Giroud, P., Ullman, S. and Barone, P., 2014b. Anatomy of hierarchy: feedforward and feedback pathways in macaque visual cortex. Journal of Comparative Neurology, 522(1), pp.225-259. Martinez-Conde, S., Otero-Millan, J. and Macknik, S.L., 2013. The impact of microsaccades on vision: towards a unified theory of saccadic function. Nature Reviews Neuroscience, 14(2), pp.83-96. Mendoza-Halliday, D., Torres, S. and Martinez-Trujillo, J.C., 2014. Sharp emergence of feature-selective sustained activity along the dorsal visual pathway. Nature neuroscience, 17(9), pp.1255-1262. Miller, E.K., Li, L. and Desimone, R., 1991. A neural mechanism for working and recognition memory in inferior temporal cortex. Science, 254(5036), pp.1377-1379. Miller, E.K., Li, L. and Desimone, R., 1993. Activity of neurons in anterior inferior temporal cortex during a short-term memory task. Journal of Neuroscience, 13(4), pp.1460-1478. 150 Miller, E.K., Erickson, C.A. and Desimone, R., 1996. Neural mechanisms of visual working memory in prefrontal cortex of the macaque. Journal of neuroscience, 16(16), pp.5154-5167. Miyashita, Y. and Chang, H.S., 1988. Neuronal correlate of pictorial short-term memory in the primate temporal cortexYasushi Miyashita. Nature, 331(6151), pp.68-70. Murray, J.D., Bernacchia, A., Freedman, D.J., Romo, R., Wallis, J.D., Cai, X., Padoa- Schioppa, C., Pasternak, T., Seo, H., Lee, D. and Wang, X.J., 2014. A hierarchy of intrinsic timescales across primate cortex. Nature neuroscience, 17(12), pp.1661- 1663. Owen, A.M., Stern, C.E., Look, R.B., Tracey, I., Rosen, B.R. and Petrides, M., 1998. Functional organization of spatial and nonspatial working memory processing within the human lateral frontal cortex. Proceedings of the National Academy of Sciences, 95(13), pp.7721-7726. Parthasarathy, A., Herikstad, R., Bong, J.H., Medina, F.S., Libedinsky, C. and Yen, S.C., 2017. Mixed selectivity morphs population codes in prefrontal cortex. Nature neuroscience, 20(12), pp.1770-1779. Peng, X., Sereno, M.E., Silva, A.K., Lehky, S.R. and Sereno, A.B., 2008. Shape selectivity in primate frontal eye field. Journal of neurophysiology, 100(2), pp.796- 814. Postle, B.R., Druzgal, T.J. and D'Esposito, M., 2003. Seeking the neural substrates of visual working memory storage. Cortex, 39(4-5), pp.927-946. Rigotti, M., Barak, O., Warden, M.R., Wang, X.J., Daw, N.D., Miller, E.K. and Fusi, S., 2013. The importance of mixed selectivity in complex cognitive tasks. Nature, 497(7451), pp.585-590. Rossant, C., Kadir, S.N., Goodman, D.F., Schulman, J., Hunter, M.L., Saleem, A.B., Grosmark, A., Belluscio, M., Denfield, G.H., Ecker, A.S. and Tolias, A.S., 2016. Spike sorting for large, dense electrode arrays. Nature neuroscience, 19(4), pp.634- 641. Sarma, A., Masse, N.Y., Wang, X.J. and Freedman, D.J., 2016. Task-specific versus generalized mnemonic representations in parietal and prefrontal cortices. Nature neuroscience, 19(1), pp.143-149. Salazar, R.F., Dotson, N.M., Bressler, S.L. and Gray, C.M., 2012. Content-specific fronto-parietal synchronization during visual working memory. Science, 338(6110), pp.1097-1100. 151 Serences, J.T., Ester, E.F., Vogel, E.K. and Awh, E., 2009. Stimulus-specific delay activity in human primary visual cortex. Psychological science, 20(2), pp.207-214. Serences, J.T., 2016. Neural mechanisms of information storage in visual short-term memory. Vision research, 128, pp.53-67. Sereno, A.B. and Maunsell, J.H., 1998. Shape selectivity in primate lateral intraparietal cortex. Nature, 395(6701), pp.500-503. Sreenivasan, K.K., Curtis, C.E. and D’Esposito, M., 2014. Revisiting the role of persistent neural activity during working memory. Trends in cognitive sciences, 18(2), pp.82-89. Super, H., Spekreijse, H. and Lamme, V.A., 2001. A neural correlate of working memory in the monkey primary visual cortex. Science, 293(5527), pp.120-124. Van Essen, D.C., 2012. Cortical cartography and Caret software. Neuroimage, 62(2), pp.757-764. Van Kerkoerle, T., Self, M.W. and Roelfsema, P.R., 2017. Layer-specificity in the effects of attention and working memory on activity in primary visual cortex. Nature communications, 8(1), pp.1-14. Wang, X.J., 2001. Synaptic reverberation underlying mnemonic persistent activity. Trends in neurosciences, 24(8), pp.455-463. 152 CHAPTER FOUR THE CORTICAL LOCAL FIELD POTENTIAL EXHIBITS DISTINCT SPATIAL GRADIENTS THAT VARY WITH FREQUENCY AND TIME DURING VISUAL SHORT-TERM MEMORY Contribution of Authors and Co-Authors Manuscriptin Chapter 4 Author: Steven J. Hoffman Contributions: study design, data collection, data analysis, and preparing manuscript. Co-Author: Nicholas M. Dotson Contributions: data collection, assisted in data analysis, and editing manuscript. Co-Author: Charles M. Gray Contributions: study design, data collection, data analysis and preparing manuscript 153 Manuscript Information Page Steven J. Hoffman, Nicholas M. Dotson, Charles M. Gray Status of Manuscript: __X__ Prepared for submission to a peer-reviewed journal ____ Officially submitted to a peer-review journal ____ Accepted by a peer-reviewed journal __ __ Published in a peer-reviewed journal 154 Summary Decades of research have revealed functional correlates of the cortical local field potential (LFP) and cellular mechanisms underlying its generation. However, the spatial organization of the cortical LFP is much less well understood. Using recent advances in large-scale multi-electrode recording technology (Dotson et al., 2017; 2018), we measured the cortical LFP from 62 identified areas spanning a cerebral hemisphere in two macaque monkeys performing a visual delayed match to sample task (dMTS). Using spectral analysis, we find multiple frequency bands that differ between the two monkeys and which display distinct spatial gradients across the cortex. We also find that anatomically defined cortical areas can be robustly decoded by using a small number of spectral and temporal parameters as features in a machine learning classifier. The decoding performance varies across task epochs and is improved when all task epochs are included. These findings show that cortical areas in the non-human primate have characteristic spectral composition that varies systematically across the cortical mantle in a task dependent manner. Introduction In the early electroencephalographic recordings of brain electrical activity, the frequency band of observed oscillations quickly became a key metric for their description (Berger 1929; Adrian and Matthews 1934; Jasper and Andrews 1936). Subsequent studies, in both humans and macaque monkeys, noted that the predominant frequencies (principally the alpha and beta rhythms, at ~8-14 Hz and ~13-30 Hz, respectively) varied regionally and could be localized to specific cortical areas or regions (Jasper and Andrews 1938; 155 Garvin and Amador 1949; Jasper and Penfield 1949). Over the decades, several additional frequency bands have been identified (e.g., delta 0.5-4 Hz, theta 4-8 Hz, gamma 30-80 Hz), and functional roles have been proposed ranging from perceptual binding (Singer and Gray, 1995), representation of mnemonic content (Salazar et al., 2012; Lundqvist et al., 2016) and attention (Fries 2005; 2015) to inhibitory gating mechanisms (Jensen and Mazaheri 2010). Despite decades of research on the function and mechanisms of oscillatory activity in different frequency bands (Gray, 1994, Buzsáki and Draguhn 2004; Wang, 2010; Pesaran et al., 2018), it is still unclear how the various cortical rhythms are distributed across the array of cortical areas (Markov et al., 2014). Resolving this question is important, given that the spatial distribution of cortical rhythms may be essential to understanding their functions. However, this important category of questions has remained largely unanswered due to technical constraints in making simultaneous recordings of neural activity from the depth and breadth of the brain. To address these questions, we utilized a novel microdrive system to measure intracortical neural activity from 62 identified cortical areas in two macaque monkeys that were engaged in a visual short-term memory task (Dotson et al., 2017; 2018). Using spectral analysis, we characterized the predominant frequencies of oscillatory activity in the local field potential (LFP) during each of the epochs of the task. We used the distribution of detected spectral peaks to designate the frequency bands for subsequent analyses in each monkey. For each frequency band, we found a differing spatial gradient of spectral amplitude across the recorded cortical areas. Using spectral and temporal parameters of the LFP as features in a classifier, we show that the cortical area of origin 156 can be decoded from each epoch of the task. Small, but significant, differences in decoding accuracies occur in all areas between task epochs, but no individual epoch performed better than the others. Incorporating all task epochs into the classifier yielded the highest decoding accuracy. Further analysis revealed that each feature contributed approximately equally to the classification performance and that both variation in spectral features, as well as the similarity in features between adjacent areas, were two aspects that led to lower validation accuracy for particular areas. These findings show that different spectral components of the LFP display distinct spatial gradients across the cortex and that task- dependent changes in spectral power are widespread across the cortex, even for a simple cognitive task. Results We recorded broadband neuronal activity from a total of 62 cortical areas in two NHPs (39 areas in Monkey E and 58 areas in Monkey L) while they performed a feature- based delayed match-to-sample (dMTS) task and an interleaved visual fixation task. The dMTS task required the monkeys to remember a centrally presented sample image (1 of 5 possible images) for a minimum of 800 ms (800-1200 ms in Monkey E; 1000-1500 ms in Monkey L), before making a choice between a matching and a non-matching image. Details of the recording methods, reconstruction of the recording sites and analysis of the task dependence of unit activity have been reported previously (Dotson et al., 2017, 2018). Here we focus on the spatial and spectral organization of the LFP and its task dependence. The LFP was separated from the broadband signal using a bandpass filter and stored for off-line analysis. LFP signals were selected for analysis only if there was detectable unit 157 activity recorded on the same electrode, and if the electrode position had changed from the previous day of recording. All analyses were performed on the correct trials of the dMTS task, obtained from 25 sessions in Monkey E and 61 sessions in Monkey L (minimum 500 correct trials and >75% correct performance on each session). Simultaneous recordings were made from up to 21 and 37 different cortical areas in Monkeys E and L, respectively. The recording locations and sample sizes for each animal are shown in Figure S1 and Table 1. Sparsely sampled and adjacent cortical areas with similar functional properties were merged into groups and included in the analysis if each area/group contained a minimum of 20 recordings. This resulted in a total of 526 recordings from 15 areas merged into 11 area/groups in Monkey E and a total of 1563 recordings from 44 areas merged into 28 area/groups in Monkeys L (Table 1). Spatial Organization of the LFP At the outset of these experiments it was readily apparent that the spectral properties of the LFP varied systematically across the cortex in a task dependent manner. Figure 1 shows a representative example of the raw data, and corresponding power spectra, sampled from a subset of the channels on a single session in monkey L. Prefrontal (9/46d), frontal eye field (8M, 8B, 8r), and premotor (F7, F2) areas tended to display desynchronized fluctuations of low amplitude interspersed with periods of periodic activity in the range of 6-14 Hz. Primary motor and somatosensory areas, F1 and a3, exhibited pronounced oscillations in the 25-35 Hz range, that extended with lower amplitude into premotor areas. Anterior (a2, a5, 7B) and posterior (7A, V6A) parietal areas exhibited higher amplitude fluctuations with salient oscillations in the 6-14 Hz range. Primary visual cortex (V1) was 158 dominated by high amplitude fluctuations at low frequencies. Each of these aspects were clearly apparent in the corresponding power spectra (C), that revealed notable peaks (arrows) and often marked changes in amplitude in different epochs of the task. Some spectra showed a sharp singular peak centered around 10 Hz, while others displayed multiple peaks around 10 Hz and 30 Hz, and in some cases near 65 Hz. Smaller peaks and shoulders were also apparent near 20 Hz. Other recordings, most notably in early visual cortex, displayed a steep fall off in power, with very low amplitudes at higher frequencies. In order to determine the appropriate frequency bands for subsequent analysis, we applied a peak detection algorithm to the average power spectra computed across all correct trials over the full trial length for all recordings in both animals (Figs. S2 and S3). This analysis yielded a histogram of peak frequencies, and their corresponding areas of origin, spanning the full data set in each monkey (Figure 2). We fit each histogram with a probability density function (PDF) and found the local minima in the distributions. This yielded 4 frequency bands in monkey E (0-8 Hz, 8-21 Hz, 21-32 Hz, 32-80 Hz) and 5 bands in monkey L (0-6 Hz, 6-14 Hz, 14-26 Hz, 26-42 Hz, 42-80 Hz). Band-1 included frequencies below 4 Hz, where the peak detection was not applied to avoid spurious peaks resulting from filtering. These data revealed a notable difference in frequency distribution between the two monkeys. Low frequencies (band-1) were sparsely distributed in both monkeys. Spectral peaks in band-2, centered at 14 Hz in monkey E and 10 Hz in monkey L, were widely distributed and present in nearly all areas of cortex. Peak frequencies in band-3 were centered at 25 Hz in monkey E and 21 Hz in monkey L and were frontally distributed. There were more obvious differences between the animals at higher 159 frequencies. Band-4 was very sparse in monkey E, while monkey L showed pronounced narrow-band components in two bands (band-4, band-5) that were concentrated in somatomotor regions and some areas of the prefrontal and parietal cortices. Some high frequency components were also sparsely present in early visual cortex in both animals. Many areas displayed a unimodal distribution, while other areas showed bimodal or even multimodal distributions of spectral peaks. This was most evident in monkey L. We obtained nearly identical results when the analysis was applied to the spectra on a linear scale (Figure S4). These findings further illustrate the diversity of spectral profiles across cortical areas, as well as between animals. Having defined the frequency bands in each monkey, we extracted two values from the spectra computed on each epoch: the percentage of power in each frequency band relative to the entire spectrum, which we refer to as the spectral content (SC), and the peak amplitude (PA) of the power in each band. Figure 3 shows example spectra from each monkey illustrating task dependent changes in SC and PA in the different frequency bands. In order to assess the spatial organization of spectral power across the cortex, we plotted the mean value of SC in the presample epoch (across sessions) on cortical flatmaps for each frequency band (flatmaps of PA were less informative because of the wide range of amplitudes). These maps, shown in figure 4, revealed striking spatial gradients of SC that differ between frequency bands and display both similarities and differences between the two animals. In both monkeys there was a clear posterior to anterior gradient of SC in band- 1, with the greatest concentration of power occurring in visual cortices. Band-2 displayed an anterior shift in the distribution with relative power radiating out from more 160 central/parietal regions. In the higher frequency bands (3 and 4 in Monkey E, 3-5 in Monkey L) there was a progressive anterior shift in the distribution of SC with increasing frequency. A notable difference between animals was again present in band-4 of Monkey L. There was a focal distribution of high amplitude in somatomotor areas (3, F1 and F2) that displayed a declining gradient into premotor, prefrontal and anterior parietal areas. (The rank ordered plots of SC and PA for both monkeys are shown in figures S5, S6, S7 and S8.) In addition to its spectral properties, the LFP also displayed marked variation in temporal structure across different areas of cortex. To quantify this, we computed the Sample Entropy (SE) of each LFP time series (Richman and Moorman, 2000; Delgado- Bonal and Marshak, 2019). SE reflects the stochasticity of a signal, with rhythmic signals having a lower SE than more desynchronized, chaotic signals. Various entropy measures have long been applied to EEG signal analysis and in clinical data SE is often used as a way to estimate the depth of anesthesia (Liang et al., 2015). For each channel on each session, we calculated the average SE across trials on each epoch of the task. An example of the result, applied to the data from figure 1B, is shown in Figure S9. SE was lower in posterior occipital and parietal areas and increased in somatomotor and prefrontal areas. This effect was consistent with the data as a whole. Rank ordered box plots and corresponding flatmaps of SE (presample epoch) for the full data set are shown for both monkeys in Figure 5. This revealed widespread variations among cortical areas and a clear spatial gradient of SE from occipital to prefrontal regions of the cortex. Decoding Cortical Areas 161 Having defined a set of features that reveal systematic spatial variations of the spectral (SC, PA) and temporal (SE) properties of the LFP, we sought to determine if these features could be used to objectively classify the cortical area of origin of the recordings. We used a support vector machine (SVM) classifier algorithm, using a K-fold cross validation scheme (K = 5), and applied it separately to the full data set from each monkey, on each epoch of the task. The analysis included 9 features for Monkey E (SC and PA in 4 frequency bands and SE) and 11 features for Monkey L (SC and PA in 5 frequency bands and SE). In both monkeys, we found average validation accuracies that were well above chance for every area/group in each of the 5 epochs. Figure 6 shows the confusion matrices of the median values (upper plots) and the corresponding distributions of validation accuracies obtained from the diagonal of those matrices (lower plots) for the presample epoch in monkey E (A) and monkey L (B), respectively. The median validation accuracy for every area/group included in the analysis in both animals was well above the theoretical chance level (1/#areas, dashed blue line) as well as the 99% confidence limit (solid red line) computed from the permutation test (see Methods). However, there was a broad distribution of validation accuracies with median values ranging from 35%-85% in Monkey E and 20%-75% in Monkey L. Part of the difference between animals is likely to stem from the larger number of area/groups included in Monkey L. Interestingly, misclassifications in the confusion matrices tended to lie near the diagonal, indicating similarity in the spectral and temporal properties of the LFP among nearby areas within a cortical region. 162 We applied the same analysis to each of the remaining epochs of the task in each animal (sample, delay1, delay2, delayM) and found similar results. To visualize the results, we plotted the distribution of validation accuracies for all 5 epochs in each area/group in Monkey E (Figure S10) and monkey L (Figure S11). These data showed that the median value in all area/groups in all epochs of both animals were significantly above the expected theoretical level and above the 99% confidence limit in all but one epoch in one area (DP, delay1) in Monkey E (Figure S10). Task-dependent changes in the validation accuracy between epochs were also apparent in every area/group. For example, some areas displayed an increase in validation accuracy during the delay period (DP and V2 in Monkey E, Fig. S10; 7b and V6A in Monkey L, Fig. S11), others displayed the opposite pattern (a2 in Monkey E, Fig. S10; a1 and a2 in Monkey L, Fig. S11), and many others displayed differences between epochs with no discernable pattern across areas. In fact, every area/group in both animals showed a significant difference in validation accuracy across epochs (Kruskal-Wallis test, P<<10-5, FDR corrected), indicating widespread changes in the spectral and temporal properties of the LFP during the task that span all of the cortical areas measured. Next we asked if the inclusion of information from all the task epochs would improve validation accuracy. For this, we included the same set of features from all five task epochs (presample, sample, delay1, delay2, delaym) into the decoding analysis (45 features for Monkey E, 55 features for Monkey L). This analysis, shown in Figure 7, resulted in a modest, but consistent increase in the validation accuracy when compared to any of the other epochs. To visualize this result, we plotted the distribution of differences 163 in validation accuracy between the presample epoch and all epochs combined in Figure 8. In Monkey E, 9 out of 11 area/groups showed a significant increase in validation accuracy, one area showed no difference (a2), and one group decreased (dPFC) (P<<10-5, Wilcoxon ranksum test, FDR correction). The mean change across area/groups was an increase of 8%. In Monkey L, 25 out of 28 area/groups showed a significant increase in validation accuracy, and three area/groups showed no difference (OrbPFC, a1, a2) (P<<10-5, Wilcoxon ranksum test, FDR correction). The mean change across area/groups was an increase of 10%. These results, combined with the analysis of changes between epochs (Figures S10 and S11), demonstrate task dependent changes in the LFP that span all areas of the cortex measured in both animals. Feature Importance We next wanted to discern whether specific features were responsible for successful classification. This question proved difficult to address, because some features are informative for certain areas, but not others. For example, spectral power in band-4 of Monkey L may be useful in distinguishing somato-motor areas, but largely uninformative in areas where the power in this band is very low (Figures 2 and 4). In order to address this question, we investigated the decrease in validation accuracy that occurs when the values of a particular feature are randomized (Dreyfus and Guyon, 2006). Randomizing informative features would lead to a decrease in validation accuracy while uninformative features would result in little or no change when randomized. This question is further complicated by the fact that many of the features have high covariance. A recording that has a high SC value in a particular frequency band will likely have a high peak amplitude 164 in that frequency band as well. In figure 9 we show the change in mean validation accuracy across all area/groups after randomization of each feature individually. There was a ~5% decrease for each feature in both monkeys as compared to the baseline, with Band-3 SC and SE showing slightly larger effects in both monkeys. Areal Misclassification What is responsible for the wide range of validation accuracies across area/groups? We considered two interrelated factors that might introduce confusion in the classification analysis. First, the spectral and temporal features of the LFP could vary widely within an area/group, making the data difficult to separate. Second, as indicated by the flatmaps of SC, there could be a high degree of feature similarity between areas, particularly if they are close to one another. We examined the first question by calculating the correlation coefficient between validation accuracy and the standard deviation of each feature used in the classifier. Examples of this result are shown in figure 10 for SC in band-2 of the presample epoch in each monkey. This revealed a significant negative correlation between feature variance and validation accuracy. We applied the same analysis to all features in both animals and the results are shown in table 2. A significant negative correlation was present in 3 of the 9 features in Monkey E and in 4 of the 11 features in Monkey L. These effects occurred at the lower frequencies, bands 1 and 2, but were absent or insignificant in the higher frequency bands. Although not statistically significant, 8 out of 9 of the features for Monkey E, and 8 out of 11 for Monkey L had negative trends in their correlation, suggesting that higher variability in the feature is correlated with lower validation accuracy. 165 To determine if validation accuracy was the result of misclassification between area/groups, we calculated a “pairwise confusability” metric from the confusion matrices computed from all task epochs combined (Figure 7). This metric is the sum of the off- diagonal classification errors between each pair of area/groups in the confusion matrix. The resulting values were used to create a similarity matrix where all pairwise combinations of area/groups are represented, and each value indicates the percentage of incorrect classifications between each pair. We then used this similarity matrix as input into a multi- dimensional scaling (MDS) algorithm to characterize the classification errors between each of the cortical areas/groups. This provided a unique way to visualize the off-diagonal misclassifications in the confusion matrix where the distance between points represents the magnitude of separation among the spectral and temporal features of all the area/groups (Figure 11). Interestingly, the data from both monkeys appeared to segregate into regional clusters (early visual, extrastriate visual and fronto-parietal in Monkey E, and occipito- parietal, somatomotor and premotor-prefrontal in Monkey L). This provides further evidence that the spectral and temporal features exhibit spatial gradients with nearby areas displaying similar profiles. Discussion We employed large-scale, simultaneous measurements of intracortical neural activity in two monkeys performing a visual short-term memory task (Dotson et al., 2017) to investigate the spatial organization of the LFP across a significant fraction of the anatomically identified areas of the cortical map in non-human primates (Markov et al., 2014). Analysis of the peak frequencies in the LFP power spectra revealed multiple narrow 166 frequency bands in both animals. These distributions overlapped, but also differed in some key respects, demonstrating that the spectral composition of the cortical LFP can vary significantly between subjects and does not always fall neatly into classically defined canonical frequency bands. These results, along with differences in the recording locations and sample sizes, required us to confine the subsequent analyses to each monkey individually. We found that spatial maps and rank-ordered plots of two spectral parameters (SC, PA) revealed striking spatial gradients across the cortical map that differed markedly between frequency bands. These gradients were similar in the two monkeys, with the exception of Band-4 in Monkey L that was largely absent in Monkey E. Low frequency power (Band-1) displayed a pronounced occipital to prefrontal gradient. Power in Band-2 displayed a more gradual gradient, radiating from somatomotor and parietal areas to occipital and prefrontal regions. Power at higher frequencies was much lower in amplitude and more frontally distributed in both animals. A separate analysis of the temporal structure of the LFP, using the measure of Sample Entropy (SE), revealed a similarly striking spatial gradient across the cortical map in both monkeys. Together, these analyses demonstrate clear areal differences in the spectral and temporal properties of the LFP that exhibit multiple, distinct spatial gradients across the cortical map. These gradients are reminiscent of the earliest qualitative reports in humans and monkeys (Jasper and Andrews 1938; Garvin and Amador 1949; Jasper and Penfield 1949) and exhibit similarities to recent reports in humans (Groppe et al., 2013; Frauscher et al., 2018; Keitel and Gross, 2016). Using these measures of spectral power and temporal structure, we could reliably classify cortical areas, or small groups of areas, well above the 99% confidence limit 167 derived from surrogate distributions that randomized areal assignment. The classification performance varied across epochs of the task and was improved when all epochs were included in the analysis. However, no single feature incorporated in the analysis stood out as particularly informative, and validation accuracies exhibited a wide range in both monkeys. This suggested that variance in the parameter distributions, or similarities in those distributions among nearby areas, could have led to degraded classification performance. We found evidence supporting both of these conjectures. Variance in some features was negatively correlated with validation accuracy and analysis of the confusion matrices revealed that anatomically close areas were often mis-classified with respect to one another. Together these analyses demonstrate widespread task-dependent changes in the LFP that span a wide range of frequencies and all areas of the map that were recorded. The spectral and temporal structure of the LFP can distinguish between many individual areas, but there is also a considerable degree of overlap among adjacent areas, indicating that the properties of the LFP vary at a regional level that incorporates multiple areas with similar functional properties. Methodological Considerations Our approach represents a significant advance over other methods such as ECoG that are limited to surface measurements with lower resolution (Bosman et al., 2012; Bastos et al., 2015; Dubey and Ray, 2019). However, the study also had several methodological limitations. First, we were unable to record from most ventral areas of the cortex, particularly in the temporal lobe. Second, the cortical areas we recorded from, and the size of the sample measured from each area, differed substantially between monkeys and 168 between areas within each monkey. Part of this was due to an improvement in the methods used in Monkey L (Dotson et al., 2017). This sampling problem could have biased our results and interpretations. For example, we obtained a large sample of recordings in area F2 in Monkey L, but only two recordings in Monkey E (Table 1). We made efforts to mitigate this problem by analyzing the data from each monkey separately and by sub- sampling the data from each area/group in each iteration of the decoding analysis. Third, although we were able to recover the areal location of each recording, we could not identify the cortical layer at each recording site. Given that the LFP is known to vary in amplitude and frequency across the cortical layers (Murthy and Fetz, 1996; Spaak et al., 2012; Bastos et al., 2018), our inability to identify the cortical layer of each recording could have introduced significant variance or biased our results. Fourth, variations in arousal level are known to have pronounced influences on the spectral properties of the LFP (Magoun, 1961), and these were clearly present in our data, particularly when the animals became drowsy or disinterested in the task (see Figure 12). We attempted to reduce these effects by restricting our analysis to correct trials during periods of high performance on the task. Finally, the type of cognitive task, and the stimuli used in the task, are likely to have a significant influence on the properties of the LFP. These influences may vary in ways that depend on cortical area. There were two other factors that could have affected our ability to discriminate differences in the LFP between cortical areas. First, because we used monopolar recording methods, volume conducted signals could have blurred the differences between areas. However, multiple lines of evidence indicate that the LFP is highly local (Gray and Singer, 169 1989; Katzner et al., 2009; Spaak et al., 2012) and the vast majority of our recordings were taken from sites that greatly exceed the minimum separation distance of 2.4 mm. The reference channel in our recordings was tied to the large titanium chamber implanted on the animals (Dotson et al., 2017). This is expected to lead to a quiet reference potential by integrating signals from widespread areas of the skull. A separate analysis of the spectral coherence between simultaneously recorded LFPs revealed many instances of non- significant coherence at all frequencies (data not shown), suggesting that the reference signal is indeed quiet. Second, although our analysis of spectral peaks utilized the entire trial duration (2100 ms), the 400 ms duration of the epoch-based spectral analysis, and the AC-coupling of our recording system, limited our ability to evaluate signals less than 2.5 Hz. This was apparent in the lack of DC power in our recordings (Figure 1.) Therefore, our analysis of low frequency was limited to values >2.5 Hz. Frequency Differences Outside of the low frequencies in Band-1, we found several notable differences in the distribution of spectral peaks in the two monkeys. In Band-2 the signals were widely distributed across the cortical areas but centered at 14 Hz in monkey E and 10 Hz in monkey L. Peak frequencies in Band-3 were centered at ~25 Hz in Monkey E and ~20 Hz in Monkey L and they were frontally distributed in both animals. The clustering of spectral peaks in bands 4 and 5 of Monkey L appeared to be completely absent in Monkey E. What can account for these differences between animals? Part of the difference may stem from the sampling problem discussed above. But perhaps the simplest explanation is that the occurrence and frequency distribution of narrow band oscillatory activity can differ widely 170 between individuals and do not necessarily fall into the classically defined canonical frequency bands. This conclusion is supported by studies in both monkeys and humans (Kilavik et al., 2012; van Pelt et al., 2012; Confais et al., 2020), where narrow band oscillatory activity can differ significantly between individuals. Another notable finding in our data was the very sparse distribution of gamma-band (30-80 Hz) activity that is characteristic of primary and extrastriate areas of the visual cortex (Gray and Singer, 1989; Singer and Gray, 1995; Friedman-Hill et al., 2000; Fries et al., 2001; Bosman et al., 2012). This type of activity is known to depend on the use of appropriate visual stimuli that are tailored to the receptive field properties of the cortical neurons in those areas. While we did observe clear instances of visually-evoked gamma- band oscillations in V1 and V2, this was largely by chance, since we made no attempt to measure cellular receptive fields and adjust the sample stimuli to optimally activate the neurons at those recording sites. The absence of gamma-band activity from most of our recording sites argues against a general role for this form of activity as a general information carrier in the neocortex (Basar, 2013). Anatomical Correlations Given our ability to decode cortical area based on the properties of the LFP, we wondered how the histological features of the cortex, which form the basis for areal identification (Brodman 1909; Kornmüller 1933; von Bonin and Bailey, 1947; Garvin and Amador 1949), correlated with the different frequency bands in the LFP. We used recently published histological data in order to investigate possible cytoarchitectonic correlations with the spectral features (Beul et al., 2017). Using the published measures of cell density, 171 provided by the work of Beul et al. (2017), we plotted the relationship between cell density and the mean spectral content derived from our analyses. Example plots are shown in figure 13 illustrating the striking positive correlation between spectral content in Band-1 and cell density across the cortical map. Significant correlations occur in the other frequency bands and these are listed in Table 3. Interestingly the trend in correlations was negative for the higher frequency bands, indicating neuron density and spectral content are anti-correlated at higher frequencies. Detailed Methods Detailed descriptions of the experimental methods for behavioral training, neural recording, and preprocessing of the data have been described in two previous reports (Dotson et al., 2017; 2018). We provide brief descriptions of these methods here. Behavioral Task Two female macaque monkeys were trained to perform an object-based delayed match to sample task (dMTS; MonkeyLogic software: Asaad and Eskandar, 2008a, 2008b). A trial began when the animal acquired and fixated on a small fixation spot (presample period; fixation window = 3dva). After 500ms for Monkey E, and 800 ms for Monkey L the fixation dot was replaced with one of five randomly selected sample images for 500ms (size: 2.4x2.4 dva). The five daily images used were drawn from a larger pool of 92 images. During the sample period the monkey had to maintain its gaze within a 3 dva window encompassing the image. Next the sample image was extinguished and replaced with the fixation dot for a variable delay period (800-1200ms Monkey E ;1000-1500ms Monkey 172 L). At the end of the delay period, the fixation target is extinguished and the matching image and a non-matching image (one of the four other images) appear at 5 dva from the center of the screen. For Monkey E, the match and non-match images were always randomly oriented across from each other on a horizontal plane. For Monkey L, the images were randomly aligned either vertically, horizontally or 45 degrees diagonally across from each other. Finally, while the match image was visible, the monkey had to make a saccadic eye movement to the matching image and maintain fixation for a brief period (200 ms for Monkey E, and 500 ms for Monkey L). Correct trials were rewarded with a drop of juice. Electrophysiological Recording Broadband recordings (0.1 Hz – 9 kHz, sampled at 32 kHz) were made across the breadth of a cerebral hemisphere in the two monkeys as described in Dotson et al., 2017. Briefly, a hemisphere-wide, large-scale microdrive was implanted in two female macaque monkeys with the capability of simultaneously recording from up to 256 independently moveable micro-electrodes (inter-electrode spacing = 2.5mm). Neural activity was sampled from an overlapping set of 62 cortical areas over the course of 6 and 9 months for Monkey E and L, respectively. The broadband signal was lowpass filtered at 0.1-250Hz and resampled at 1KHz in order to obtain the LFP. Anatomical designations (using the nomenclature of Markov et al., 2014) were achieved through reconstruction of each electrode’s track through histological sections (Dotson et al., 2017). Only recordings that showed evidence of neural unit activity were used for further analysis. To avoid the over- sampling of activity, only LFP signals where the electrode position differed from the previous day of recording were used in this analysis. 173 Areal Grouping Data from adjacent cortical areas with few recordings and similar functional properties were pooled to form small groups. Recordings in visual areas V1 and V2 with short-latency visual responses in the unit activity (SLVR) were analyzed separately (i.e. V1-SLVR, V2-SLVR). The SLVR designation was assigned if the unit activity displayed a significant change in firing rate within 50-100ms following the sample presentation. After these pooling procedures, there were a total of 25 and 33 different cortical areas or small groups of areas in Monkeys E and L, respectively (Table 1). Identification of Spectral Bands For each recording the power spectral density, averaged across all correct trials in a session, was calculated from 0.1 – 100Hz (Field Trip toolbox; Oostenveld et al., 2011). A multitaper smoothing window of 1Hz was used. In Monkey E residual 60Hz line noise was removed with a notch filter. Spectra were calculated over a time window of 2100ms encompassing the task period from presample until match onset. Due to the variable length of the delay period, trials that were less than 2100 ms in length were zero padded in order to achieve this length. Longer trials were truncated to fit the 2100 ms time window. The frequency resolution of the power spectra was 0.5Hz. We used an empirical method to designate the frequency band limits for further analysis. We applied a peak detection algorithm to the normalized power spectra from 3- 100 Hz, in both linear and semi-log coordinates, to identify narrow band oscillatory activity. Local maxima <3Hz were excluded, due to artifactual peaks introduced by the multi-taper smoothing filter. In order to limit small, spurious peaks, a peak prominence 174 criterion (as defined by the Matlab findpeaks.m function) was used. The prominence threshold was set at 1% of the maximum prominence found across all spectra (see Figures S2 and S3). Values exceeding the prominence threshold were used to generate a histogram of peak frequencies across all recordings in each animal. The histograms were fit with a probability density function (PDF) using a Gaussian kernel. Frequency band limits were designated as the local minima in the PDFs (Figure 2). This analysis yielded nearly identical band limits when applied to the linear and semi-log spectra, with deviations in frequency bands being less than 1.5 Hz. Results from the two approaches were combined by averaging the local minima designations from the two methods (Figure S4). Classifier Features Using the band limits defined for each monkey, we computed the spectral content (SC, the percentage of power within a frequency band relative to the entire spectrum) and the peak amplitude (PA) of the power spectra within each frequency band and epoch of the task for the entire data set in each animal. To quantify the temporal properties of the LFP, we calculated the Sample Entropy (SE) on each task epoch for each recording (Richman and Moorman 2000), by taking the average SE across trials. Together these analyses resulted in 9 features in each of 5 epochs in Monkey E (SC and PA in bands 1-5, and SE) and 11 features in each of 5 epochs in Monkey L (SC and PA in bands 1-5, and SE). Machine Learning Classifier Each of the features were included in cubic Support Vector Machine (SVM) classifier. Cortical areas with small numbers of recordings were grouped with those in adjacent areas with similar functional properties surrounding areas. To avoid bias (in both 175 the training and validation phases) towards areas with substantially more recordings, an equal number of recordings (n=20) from each area/group were randomly drawn from the larger pool of data. The classifier was trained and validated using k-fold cross validation (k=5, 4/5th training, 1/5th cross validation). The training-validation process was iteratively repeated 1000 times, with each iteration using a different set of 20 randomly selected recordings for an area/group drawn from the data pool (Table 1). The theoretical level of chance classification is 1/# areas (1/11 Monkey E, 1/28 Monkey L). In order to further assess classification accuracy and avoid biases arising from the use of the theoretical chance level (Combrisson and Jerbi 2015), the mean and 99% confidence interval for chance classification was calculated from a permutation distribution. This distribution was created by randomly shuffling areal/group labels before the classification process, then repeating over 1000 iterations. 176 Tables Table 1. Recording counts for all areas and areal groupings in Monkeys E and L. Area and areal group names are given in the first and second columns, respectively. Columns labeled as “Flatmaps” show the recording count reported in flatmaps of SC and SE. Values in red (n<3) in Monkey E are not reported in the flatmaps. Columns labeled as “Decoding” show the area/groups that are included in the classification analysis. Areas A rea/Group Monkey E Monkey E Monkey L Monkey L Flatmaps Decoding Flatmaps Decoding OPRO, 11,12,13,14,32 OrbPFC 42 42 24c, 24d a24 14 40 40 9/46v, 46v vPFC 14 20 20 9/46d, 9, 46d dPFC 40 40 32 32 8B 8B 10 60 60 8L 8L 9 55 55 8M 8M 6 54 54 8r 8r 25 25 44, 45, 45B a44/45 5 33 33 F1 F1 90 90 102 102 F2 F2 2 120 120 F6, F7 F6/F7 24 24 49 49 1 1 17 41 41 2 2 43 43 31 31 3 3 17 79 79 Insula Ins 16 7op 7op 1 16 c5, MIP a5/MIP 15 74 74 PIP PIP 1 22 22 7a 7a 7 30 30 STPc, TPt STPc/TPt 12 15 AIP, VIP AIP/VIP 35 35 7b 7b 23 23 41 41 LIP LIP 37 37 DP DP 30 30 56 56 V6A V6A 44 44 35 35 MT, MST MT/MST 12 59 59 V4, V4t V4/t 54 54 26 26 V3 V3 6 12 V2SLVR V2SLVR 6 23 23 V2 V2 51 51 117 11 7 V1SLVR V1SLVR 27 27 16 V1 V 1 100 10 0 225 22 5 Recording count 793 526 1654 1563 Areas 39 15 58 44 Area/groups 25 11 33 28 177 Table 2. Correlation coefficients, and corresponding P values, between the standard deviation of classification features and validation accuracy for the Presample epoch of the task. SC – Spectral Content; PA – Peak Amplitude, SE – Sample Entropy. Numerical values 1-5 indicate the separate frequency bands for each monkey. Monkey E SC1 SC2 SC3 SC4 PA1 PA2 PA3 PA4 SE Correlation -.75 -.66 -.28 .28 -.84 -.16 -.44 -.54 -.05 P value .008 .029 .41 .41 .001 .63 .18 .09 .89 Monkey L SC1 SC2 SC3 SC4 SC5 PA1 PA2 PA3 PA4 PA5 SE Correlation -.48 -.45 .16 .02 .14 -.56 -.61 -.35 -.06 -.20 -.11 P value .01 .017 .42 .91 .47 .002 .0006 .07 .75 .30 .59 Table 3. Correlation coefficients, and corresponding P values, between the average spectral content in the frequency bands and neuron density (adopted from Beul et al., 2017). Numerical values 1-5 indicate the separate frequency bands for each monkey. Monkey E SC1 SC2 SC3 SC4 Correlation .64 .49 -.49 -.41 P value .004 .03 .10 .08 Monkey L SC1 SC2 SC3 SC4 SC5 Correlation .78 -.52 -.71 -.50 -.67 P value <<.01 .014 <.01 .019 <.01 178 Figures Figure 1. Spatial and task-dependent variation in the spectral organization of the LFP across the cortex during a single recording session. (A) Schematic of the recording sites in monkey L. The outline of the left hemisphere and major cortical sulci, drawn from a photograph, are shown in red. Each circle shows the entry location of an electrode that recorded neural unit activity at some point during the 9 months of the experiment. Filled circles indicate electrode locations that recorded neural activity in this session (n=94). Cyan filled circles mark the electrode locations that correspond to the signals shown in B. (B) Broadband (0.1Hz-9kHz) raw data recorded on a single trial from 14 separate cortical areas. The areal name (left of each trace) follows the nomenclature of Markov et al. (2014). The bottom two traces show the vertical and horizontal components of the eye position signal. The black vertical lines, from left to right, mark the onset and offset of the sample stimulus and the onset of the match stimulus, respectively. The colored horizontal bars at the top indicate the time and duration (400 ms) of the Presample (black), Sample (green), Delay1 (blue), Delay2 (red) and DelayM (magenta) epochs, respectively. A saccadic eye movement, occurring ~200 ms following the onset of the match, indicates the monkey’s behavioral choice. (C) Plots of LFP power as a function of frequency (0-100 Hz), averaged across all correct trials for each of the 5 epochs, for each area indicated in A and B. Spectral traces are plotted in the color corresponding to each epoch shown in B. Clear differences in the spectral power of the LFP and its task dependence are apparent between different areas of cortex. Black arrows mark notable peaks or shoulders in the power spectra. Peaks occurring below 3 Hz are due to the absence of a DC component in the filtered signals. Abbreviations: PS - Principal Sulcus; AS – Arcuate Sulcus; CS – Central Sulcus; IPS – Intraparietal Sulcus; LS – Lunate Sulcus. 179 Figure 2. Distributions of Peak Frequencies. Distributions of peak frequencies obtained from semi-log power spectra on all channels and sessions in Monkey E (A) and Monkey L (B). The top plots in A and B show the cumulative histograms of all peak frequencies (3- 100 Hz) that exceeded the peak prominence threshold. The continuous red lines show the probability density function (PDF) computed with a Gaussian kernel fit to each distribution. The local minima in each PDF define the boundaries between selected frequency bands for each animal (dashed vertical lines). Four bands were chosen for Monkey E and five bands were chosen for Monkey L. The lower plots show the normalized counts of peak frequencies for each cortical area/group. This revealed a rough spatial organization of peak frequencies across the sampled cortical areas. 180 Figure 3. Spectral variables and their task dependence. (A, B) Example plots of the LFP power spectra for each of the 5 epochs from a single session in Monkey E (A) and Monkey L (B). The data in A and B were from sampled areas 7B and F2, respectively. The inset plots show the task dependence of the spectral content (upper) and peak amplitude (lower) within each of the defined frequency bands labeled B1-B4 in A and B1-B5 in B. 181 Figure 4. Cortical Flatmaps of Spectral Content. Cortical flatmaps of mean spectral content in each area/group, averaged across all sessions, during the presample epoch for each of the frequency bands in Monkey E (A) and Monkey L (B). Areal boundaries and nomenclature follow that of Markov et al (2014). V1-SLVR and V2-SLVR refer to the subset of recordings in areas V1 and V2, respectively, that displayed short latency responses of spiking activity to one or more of the sample stimuli presented on each session. 182 Supplemental Figure 5. Variation of Sample Entropy (SE) Across The Cortex for Both Monkeys. Variation of sample entropy (SE) across the cortex for both monkeys. Rank ordered box plots, and spatial flatmaps of SE for Monkeys E (A) and L (B). The number of recordings in each area/group corresponds to the counts shown in Figure S1 and Table 1. Areas with fewer than 3 recordings are not included. 183 Figure 6. Baseline Epoch Areal Classification. Results of the areal classification analysis from the presample epoch in Monkey E (A) and Monkey L (B). The upper plots show the confusion matrices of the median classification accuracy. The lower graphs show box plots of the distribution of values for each cortical area/group along the diagonal of the confusion matrices (validation accuracies, n=1000). In each box plot the red line shows the median, the blue box displays the interquartile range, the whiskers show the 5th and 95th percentiles, and the red asterisks show outliers. The dashed horizontal blue line and solid red line show the mean and 99th percentile along the diagonal for the surrogate confusion matrices computed after randomization of the cortical area assignments. 184 Figure 7. Combined Epochs Areal Classification. Results of the areal classification analysis in Monkey E (A) and Monkey L (B) applied to all epochs combined. Descriptions and conventions are the same as figure 6. 185 Figure 8. Validation Accuracy Improvement with Inclusion of All Epochs. Validation accuracy of cortical area/groups improves in both monkeys with the inclusion of all task epochs. (A, B) Box plots of the difference in validation accuracy between the presample epoch and all epochs combined for monkey E (A) and monkey L (B). Red asterisks at the top of each box plot denote significant difference (p<.01, FDR corrected, n=1000). Positive values indicate that validation accuracy is greater for all epochs combined. Box plot conventions are the same as those in figures 6 and 7. 186 Figure 9. Feature Importance. Contribution of features to validation accuracy. Box plots of the median validation accuracy across all values (areas/groups) along the diagonal of the confusion matrices from the presample epoch when each feature is separately randomized. The plots in A and B show the results for monkeys E and L, respectively. The control distribution (baseline) is shown in the leftmost box of each plot. A drop in classification accuracy reflects the importance of that feature for correct classification. 187 Figure 10. Correlations in Validation Accuracy and Variation. Scatter plots of validation accuracy vs the standard deviation of spectral content (SC) in band-2 during the presample epoch for Monkey E (A) and Monkey L (B). The correlation coefficient and corresponding P value are shown at the top of each plot. Figure 11. Multidimensional Scaling of the Pairwise Confusability. Multidimensional scaling maps of the pairwise confusability in Monkey E (A) and Monkey L (B). The distance between points represents the magnitude of separation among the spectral and temporal features of all the area/groups. 188 Figure 12. State Dependent Oscillations. Plots of a 3-second segment of broadband raw data recorded during a period of rest with the room lights turned off. The data is shown for the same channels on the same session as the plots in figures 1 and S9. A rapid-onset sleep spindle occurs halfway into the segment with high amplitude in prefrontal, premotor, and parietal areas. 189 Figure 13. Neuron Density Correlated to Spectral Content. Scatter plots of the mean areal neuron density (Data adopted from Beul et al., 2017) vs the mean spectral content in Band-1 of Monkey E (A) and Monkey L (B). 190 Supplemental Figure 1. Flatmaps of the Recording Count in Cortical Area/Groups In Both Monkeys. Flatmaps of the recording counts in cortical area/groups in monkey E (Top) and L (Bottom). Areas with less than 3 recordings are shaded grey in A. Areal groupings and recording counts are listed in table 1. 191 Supplemental Figure 2. Peak Detection Results in Monkey E. Peak detection results in Monkey E. (A, C) Examples of normalized average power spectra computed on the full trial duration (2.1s, see Methods) using all correct trials in a single session and displayed in linear (A) and semi-log (C) coordinates. Detected peaks are indicated by a filled triangle and the prominence and width of each peak are marked by red vertical and horizontal lines, respectively. Labels at the top of each plot indicate the cortical area from which the data were obtained. (B, D) Distributions of peak prominence taken from all channels in all sessions obtained from the power spectra in linear (B) and semi-log (D) coordinates, respectively. Peaks <4Hz were excluded from the analysis. The red horizontal line in each plot shows the peak prominence threshold used to exclude small peaks at the noise level. 192 Supplemental Figure 3. Peak Detection Results in Monkey L. Peak detection results in Monkey L. Descriptions and conventions are the same as those in supplementary figure 2. 193 Supplemental Figure 4. Distributions of Spectral Peaks Obtained From All Channels and Sessions. Distributions of spectral peaks obtained from all channels and sessions in Monkey E (A, C) and Monkey L (B, D) in linear (A, B) and semi-log (C, D) coordinates, respectively. The descriptions and conventions are the same as those in figure 2. There were small differences in the distribution of detected peaks derived from the linear and semi-log spectra, but these had no effect on the selection of frequency bands. 194 Supplemental Figure 5. Boxplots of the Spectral Content for Monkey E. Rank ordered box plots of spectral content (SC) for each frequency band in Monkey E during the presample epoch of the task. The circle within each box shows the median, the box displays the interquartile range, the whiskers show the 5th and 95th percentiles, and the open circles show outliers. 195 Supplemental Figure 6. Boxplots of the Peak Amplitude for Monkey E. Rank ordered box plots of the peak amplitude (PA) obtained from the power spectra in each frequency band in Monkey E during the presample epoch of the task. Plotting conventions are the same as figure S5. 196 Supplemental Figure 7. Boxplots of the Spectral Content for Monkey L. Rank ordered box plots of spectral content (SC) for each frequency band in Monkey L during the presample epoch of the task. Plotting conventions are the same as figure S5. 197 Supplemental Figure 8. Boxplots of the Peak Amplitude for Monkey L. Rank ordered box plots of the peak amplitude (PA) obtained from the power spectra in each frequency band in Monkey L during the presample epoch of the task. Plotting conventions are the same as figure S5. 198 Supplemental Figure 9. Entropy Example. Broadband data taken from Figure 1B and corresponding mean Sample Entropy values (right) computed across trials and epochs on the LFP signal on the same channels. Areal names are shown on the left of each trace. 199 Supplemental Figure 10. Distribution of Validation Accuracies as a Function of Task Epoch for Each Area/Group in Monkey E. Distribution of validation accuracies as a function of task epoch for each area/group in Monkey E. The bottom left plot shows the cumulative distributions across all area/groups. Box plot conventions are the same as described in figure 6. The dashed blue line and solid red line in each plot show the theoretical expected value and the 99th percentile confidence limit derived from the surrogate distribution. 200 Supplemental Figure 11. Distribution of Validation Accuracies as a Function of Task Epoch for Each Area/Group in Monkey L. Distribution of validation accuracies as a function of task epoch for each area/group in Monkey L. Description and conventions are the same as figure S10. 201 References Adrian, E.D. and Matthews, B.H., 1934. The interpretation of potential waves in the cortex. The Journal of Physiology, 81(4), p.440. Asaad, W.F. and Eskandar, E.N., 2008a. A flexible software tool for temporally precise behavioral control in Matlab. Journal of neuroscience methods, 174(2), pp.245- 258. Asaad, W.F. and Eskandar, E.N., 2008b. Achieving behavioral control with millisecond resolution in a high-level programming environment. Journal of neuroscience methods, 173(2), pp.235-240. Başar, E., 2013. A review of gamma oscillations in healthy subjects and in cognitive impairment. International Journal of Psychophysiology, 90(2), pp.99-117. Bastos AM, Vezoli J, Bosman CA, Schoffelen JM, Oostenveld R, Dowdall JR, De Weerd P, Kennedy H, Fries P. 2015. Visual areas exert feedforward and feedback influences through distinct frequency channels. Neuron, 85:390-401. Bastos, A.M., Loonis, R., Kornblith, S., Lundqvist, M., Miller, E.K., 2018. Laminar recordings in frontal cortex suggest distinct layers for maintenance and control of working memory. PNAS, 115: 1117-1122. Berger, H., 1929. Über das elektroenkephalogramm des menschen. Archiv für psychiatrie und nervenkrankheiten, 87(1), pp.527-570. Beul, S.F., Barbas, H. and Hilgetag, C.C., 2017. A predictive structural model of the primate connectome. Scientific reports, 7(1), pp.1-12. Bosman CA, Schoffelen JM, Brunet N, Oostenveld R, Bastos AM, Womelsdorf T, Rubehn B, Stieglitz T, De Weerd P, Fries P., 2012. Attentional stimulus selection through selective synchronization between monkey visual areas. Neuron, 75:875-88. Brodmann K., 1909. Vergleichende Lokalisationslehre der Großhirnrinde in ihren Prinzipien dargestellt auf Grund des Zellenbaues. Barth, Leipzig. Buzsáki, G. and Draguhn, A., 2004. Neuronal oscillations in cortical networks. science, 304(5679), pp.1926-1929. Combrisson, E. and Jerbi, K., 2015. Exceeding chance level by chance: The caveat of theoretical chance levels in brain signal classification and statistical assessment of decoding accuracy. Journal of neuroscience methods, 250, pp.126-136. 202 Confais, J., Malfait, N., Brochier, T., Riehle, A., Kilavik, B.E., 2020. Is there an Intrinsic Relationship between LFP Beta Oscillation Amplitude and Firing Rate of Individual Neurons in Macaque Motor Cortex? Cerebral Cortex Communications, 1: 1-15. Delgado-Bonal, A. and Marshak, A. 2019. Approximate Entropy and Sample Entropy: A Comprehensive Tutorial. Entropy, 21, 541 Dotson, N.M., Hoffman, S.J., Goodell, B. and Gray, C.M., 2017. A large-scale semi- chronic microdrive recording system for non-human primates. Neuron, 96(4), pp.769-782. Dotson, N.M., Hoffman, S.J., Goodell, B. and Gray, C.M., 2018. Feature-based visual short-term memory is widely distributed and hierarchically organized. Neuron, 99(1), pp.215-226. Dreyfus, G. and Guyon, I., 2006. Assessment Methods. In: Feature Extraction: Foundations and Applications. Guyon, I., Gunn, S, Nikravesh, M. and Zadeh, L., pgs. 65-88, Eds. Springer-Verlag, Berlin. Dubey, A., Ray, S., 2019. Cortical Electrocorticogram (ECoG) Is a Local Signal. J. Neurosci., 39: 4299–4311. Frauscher, B., von Ellenrieder, N., Zelmann, R., Doležalová, I., Minotti, L., Olivier, A., Hall, J., Hoffmann, D., Nguyen, D.K., Kahane, P., Dubeau, F., Gotman, J., 2018. Atlas of the normal intracranial electroencephalogram: neurophysiological awake activity in different cortical areas. Brain, 141:1130-1144. Friedman-Hill, S.R., Maldonado, P.E., Gray, C.M., 2000. Dynamics of Striate Cortical Activity in the Alert Macaque: I. Incidence and Stimulus-Dependence of Gamma- Band Neuronal Oscillations. Cerebral Cortex, 10:1105-1116. Fries, P., Reynolds, J.H., Rorie, A.E., Desimone, R., 2001. Modulation of oscillatory neuronal synchronization by selective visual attention. Science, 291: 1560–1563. Fries, P., 2005. A mechanism for cognitive dynamics: neuronal communication through neuronal coherence. Trends in cognitive sciences, 9(10), pp.474-480. Fries, P., 2015. Rhythms for cognition: communication through coherence. Neuron, 88(1), pp.220-235. Garvin, J.S. and Amador, L.V., 1949. Electrocorticograms of the cytoarchitectural areas of macaca mulatta. Journal of neurophysiology, 12(6), pp.425-433. 203 Gray, C.M., Singer, W., 1989. Stimulus-specific neuronal oscillations in orientation columns of cat visual cortex. PNAS, 86:1698–1702. Gray CM (1994) Synchronous Oscillations in Neuronal Systems: Mechanisms and Functions. J. Comp. Neurosci., 1:11-38. Groppe, D.M., Bickel, S., Keller, C.J., Jain, S.K., Hwang, S.T., Harden, C., Ashesh D Mehta, A.D., 2013. Dominant frequencies of resting human brain activity as measured by the electrocorticogram. Neuroimage, 79:223-33. Jasper, H.H. and Andrews, H.L., (1936) Human Brain Rhythms: I. Recording Techniques and Preliminary Results, J. Gen. Psychol., 14:98-126. Jasper, H.H. and Andrews, H.L., 1938. Electro-encephalography: III. Normal differentiation of occipital and precentral regions in man. Archives of Neurology & Psychiatry, 39(1), pp.96-115 Jasper, H. and Penfield, W., 1949. Electrocorticograms in man: effect of voluntary movement upon the electrical activity of the precentral gyrus. Archiv für Psychiatrie und Nervenkrankheiten, 183(1-2), pp.163-174. Jensen, O. and Mazaheri, A., 2010. Shaping functional architecture by oscillatory alpha activity: gating by inhibition. Frontiers in human neuroscience, 4, p.186. Katzner, S., Nauhaus, I., Benucci, A., Bonin, V., Ringach, D.L., Carandini, M., 2009. Local origin of field potentials in visual cortex. Neuron, 61:35– 41. Keitel, A., Gross, J., 2016. Individual Human Brain Areas Can Be Identified from Their Characteristic Spectral Activation Fingerprints. PLoS Biol., 14: e1002498. Kilavik, B.E., Ponce-Alvarez, A., Trachel, R., Confais, J., Takerkart, S., Riehle, A., 2012. Context-Related Frequency Modulations of Macaque Motor Cortical LFP Beta Oscillations. Cerebral Cortex, 22: 2148-2159. Kornmüller, A.E., 1933. Bioelektrische Erscheinungen architektonischer Felder. Eine Methode der Lokalisation auf der Großhirnrinde. Deutsche Zeitschrift für Nervenheilkunde, 130(1-4), pp.44-60. Liang, Z., Wang, Y., Sun, X., Li, D., Voss, L.J., Sleigh, J.W., Hagihira, S. and Li, X., 2015. EEG entropy measures in anesthesia. Frontiers in computational neuroscience, 9, p.16. Magoun, H.W., 1961. Brain Mechanisms for Wakefulness. Brit. J. Anaesth., 33: 183-193. 204 Markov, N.T., Ercsey-Ravasz, M.M., Ribeiro Gomes, A.R., Lamy, C., Magrou, L., Vezoli, J., Misery, P., Falchier, A., Quilodran, R., Gariel, M.A. and Sallet, J., 2014. A weighted and directed interareal connectivity matrix for macaque cerebral cortex. Cerebral cortex, 24(1), pp.17-36. Murthy, V.N., Fetz, E.E., 1996. Oscillatory activity in sensorimotor cortex of awake monkeys: synchronization of local field potentials and relation to behavior. J. Neurophysiol., 76:3949-3967. Oostenveld, R., Fries, P., Maris, E. and Schoffelen, J.M., 2011. FieldTrip: open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data. Computational intelligence and neuroscience, 2011. Richman, J.S. and Moorman, J.R., 2000. Physiological time-series analysis using approximate entropy and sample entropy. American Journal of Physiology-Heart and Circulatory Physiology, 278(6), pp.H2039-H2049. Singer, W., Gray, C.M., 1995. Visual feature integration and the temporal correlation hypothesis. Ann. Rev. Neurosci. 18: 555–586. Spaak, E., Bonnefond, M., Maier, A., Leopold, D.A., Jensen, O., 2012. Layer-Specific Entrainment of Gamma-Band Neural Activity by the Alpha Rhythm in Monkey Visual Cortex. Current Biology 22: 2313–2318. van Pelt, S., Boomsma, D.I., Fries, P., 2012. Magnetoencephalography in Twins Reveals a Strong Genetic Determination of the Peak Frequency of Visually Induced Gamma- Band Synchronization. J. Neurosci., 32: 3388-3392. von Bonin, G. and Bailey, P. 1947. The Neocortex of Macaca mulatta. Allen RB, Kampmeier OF, Schour I, Serles ER (eds): Illinois Monographs in the Medical Sciences, Vol.5. The University of Illinois Press, Urbana, pp. 1-163. Wang, X.J., 2010. Neurophysiological and computational principles of cortical rhythms in cognition. Physiological reviews, 90(3), pp.1195-1268. 205 CHAPTER FIVE GENERAL DISCUSSION Summary of Results In Chapter 2 we introduce a novel recording technology that allowed us to record intra-cortical activity from broad parts of a macaque hemisphere. This large-scale semi- chronic Microdrive system will be essential for the studying of cognitive function, as even simple cognitive tasks rely on the coordinated and dynamic activity of a network of brain areas (Mesulam, 1990; Gray, 1994; Bressler, 1995; Friston, 1997; Bressler and Kelso 2001; Siegel et al., 2012). We next utilized this recording device to conduct the analyses in the subsequent chapters. In Chapter 3 we demonstrate that in a short-term visual working memory task dependent changes in firing rate are wide-spread, yet resolve into minimally suppressed or a largely balanced state of increases and decreases. This shows that even this relatively simple cognitive task of remembering an object for a short delay invokes widespread changes in neural firing across cortex. Within this circuit of areas that showed task dependent changes in firing rate we found a subset of areas that displayed stimulus specific patterns of firing rate during the delay period. Further, these areas fell into a hierarchy of the relative incidence of stimulus selectivity, and an increase in the time-point during the delay that newly selective units became involved. We additionally found that microsaccadic eye movements could be used to decode the stimulus that was previously shown, and that units that were modulated by the microsaccades were more likely to be stimulus-specific. 206 Finally, in Chapter 4 we implemented a spectral based analysis on the recorded LFPs in order to investigate the oscillatory make-up of the broad-spread cortical areas. We found that we could use several aspects of the LFP spectra as features to decode the cortical area of origin. This finding is of relevance, as it suggests that as cortical areas have been shown to vary as a function of their cytoarchitectonic make-up (a fact that has driven the many various anatomical classification schemes), each cortical area also seems to display individual cortical spectral profiles. We further found that incorporation of these features from the other task epochs led to an overall increase in decoding performance. Suggesting that these cortical spectral profiles also display stereotyped responses as a function of the task epoch. Future Directions The previous chapters entailed analyses of both firing rate and spectral properties from simultaneous recordings made across a large portion of cortex. These analyses did not take into account interactions between the recordings. There are several measures that explore these types of interactions (often termed functional connectivity analyses) at either the level of the spiking activity, or at the level of the LFP. One thing to note with any pair- wise calculations is that computation time does not scale linearly. For instance if on a particular recording session if activity was recorded from 100 electrodes, the number of possible pairs between these 100 electrodes would be 4,940 (𝑤ℎ𝑒𝑟𝑒 𝑡ℎ𝑒 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑝𝑜𝑠𝑠𝑖𝑏𝑙𝑒 𝑝𝑎𝑖𝑟𝑠 𝑖𝑠: 𝑛 × (𝑛 − 1)]2 ). 207 Spiking activity could be investigated in how it correlates with spiking activity in another recording location via simple cross-correlation analyses. This analysis can also incorporate correlations over different time points by “lagging” the correlation over previous time point. Spiking activity could also be investigated in how it correlates with the simultaneously recorded LFP. One measure that probes this is that of Spike-Field- Coherence, where spiking activity is assessed with relation to the phase of the LFP in a particular frequency band of interest. There are many functional connectivity analyses that are often applied to LFP data (Sakkalis, 2011). These include non-directed measures such as simple correlation analyses such as coherence, as well as measures such as phase locking value, the pairwise phase consistency, and mutual information (Bastos and Schoffelen, 2016). These functional connectivity analyses are often performed as a function of time, by either using a sliding window approach, or by breaking the task into different epochs, as well as a function of frequency (again either through a sliding window of frequencies, or by breaking up into different frequency bands). These can include directed measures as well, which aim to assert a direction influence or causality from say recording 1 to recording 2. The notion of causality invokes the temporal domain, in that cause precedes effect. Meaures that attempt to analyze these types of connections include those that use auto-regressive models such as Granger Causality, or phase slope index (Wiener, 1956; Granger, 1969; Bressler and Seth, 2011; Seth et al., 2015) or model free measures of directed interactions such as transfer entropy. 208 The recording device described in Chapter 2, is almost optimally designed to produce the types of data that are a target of these analyses. This being because it results in many recordings made simultaneously from disparate regions of cortex. These types of analyses are indeed already being conducted, although they bring with them their own set of difficulties in both interpretation and added computational complexity. However, they will be the target of future analyses and focus. As was the case with Chapters 3 (with firing rate) and 4 (spectral power), there is already much research in terms of individual brain areas, however there is little research in the broad-spread hemisphere-wide interactions during cognitive processes. Concluding Remarks The field of brain research finds itself in a golden era. New technologies are arising nearly daily with capabilities in not only recording various aspects of its activity, but in allowing perturbations which will allow for causal inferences (e.g. Optogenetics; DREADD receptors). With respect to recording technologies the focus seems to be on developing probes that can make recordings at tens (Townsend et al., 2002) to even thousands of recording sites along the shank of the probe (Raducanu et al., 2017). One can see that multiple of these multi-site probes could then be incorporated into a larger recording device to allow greater spatial coverage. Some are even suggesting that such high channel count technologies are on the horizon of being implanted in humans (Musk, 2019)! Speaking from experience I will caution the data hungry researcher, with these new capabilities come new problems. Simple data preprocessing steps such as filtering, histological categorization, and spike sorting become months or even years-long 209 endeavors. Data storage and management become no small feat when working with data sets on the order of hundreds of terabytes or greater. Computational tasks require supercomputer clusters, and knowledge of parallel computing techniques (and even then calculations can take months to resolve). Then, once all these hurdles have been met there comes the final issue of interpretation of a very complex and often messy result. Invariably new analyses techniques will need to be developed to assist in interpretation of these data sets. Perhaps spectral analyses will eventually become antiquated, a sentiment reflected by Ted Bullock nearly two decades ago: “Our descriptions of brain waves […] have usually been in terms of the power spectrum. This may be about as adequate as describing an opera in terms of its power spectrum” (Bullock, 2002). Despite all these challenges, it is through these technologies that we will have any hope of understanding cognition. 210 References Bastos, A.M. and Schoffelen, J.M., 2016. A tutorial review of functional connectivity analysis methods and their interpretational pitfalls. Frontiers in systems neuroscience, 9, p.175. Bressler, S.L., 1995. Large-scale cortical networks and cognition. Brain Research Reviews, 20(3), pp.288-304. Bressler, S.L. and Kelso, J.S., 2001. Cortical coordination dynamics and cognition. Trends in cognitive sciences, 5(1), pp.26-36. Bressler, S.L. and Seth, A.K., 2011. Wiener–Granger causality: a well established methodology. Neuroimage, 58(2), pp.323-329. Bullock, T.H., 2002. Biology of Brain waves: natural history and evolution of an information-rich sign of activity. In Advances in Electrophysiolgy in Clinical Practice and Research. Kjellberg, Inc.. Friston, K.J., 1997. Another neural code?. Neuroimage, 5(3), pp.213-220. Granger, C.W., 1969. Some recent developments in a concept of causality," Journal of Econometrics, vol. 39, 1988, pp. 199—211; and. Investigating causal relations by econometric models and cross-spectral methods,” Econometrica, 37, pp.424-38. Gray, C.M., 1994. Synchronous oscillations in neuronal systems: mechanisms and functions. Journal of computational neuroscience, 1(1-2), pp.11-38. Mesulam, M.M., 1990. Large‐scale neurocognitive networks and distributed processing for attention, language, and memory. Annals of Neurology: Official Journal of the American Neurological Association and the Child Neurology Society, 28(5), pp.597-613. Musk, E., 2019. Neuralink. An integrated brain-machine interface platform with thousands of channels. Raducanu, B.C., Yazicioglu, R.F., Lopez, C.M., Ballini, M., Putzeys, J., Wang, S., Andrei, A., Rochus, V., Welkenhuysen, M., Helleputte, N.V. and Musa, S., 2017. Time multiplexed active neural probe with 1356 parallel recording sites. Sensors, 17(10), p.2388. Sakkalis, V., 2011. Review of advanced techniques for the estimation of brain connectivity measured with EEG/MEG. Computers in biology and medicine, 41(12), pp.1110-1117. 211 Seth, A.K., Barrett, A.B. and Barnett, L., 2015. Granger causality analysis in neuroscience and neuroimaging. Journal of Neuroscience, 35(8), pp.3293-3297. Siegel, M., Donner, T.H. and Engel, A.K., 2012. Spectral fingerprints of large-scale neuronal interactions. Nature Reviews Neuroscience, 13(2), pp.121-134. Townsend, G., Peloquin, P., Kloosterman, F., Hetke, J.F. and Leung, L.S., 2002. Recording and marking with silicon multichannel electrodes. Brain Research Protocols, 9(2), pp.122-129. Wiener, N., 1956. The theory of prediction. Modern mathematics for engineers. 212 CUMULATIVE REFERENCES Adams, D., 2002. The salmon of doubt: Hitchhiking the universe one last time (Vol. 3). Harmony. Adey, W.R., 1988. Do EEG-like processes influence brain function at a physiological level?. In Dynamics of Sensory and Cognitive Processing by the Brain (pp. 362- 367). Springer, Berlin, Heidelberg. Adrian, E.D. and Matthews, B.H., 1934. The interpretation of potential waves in the cortex. The Journal of Physiology, 81(4), p.440. Adrian, E.D. and Moruzzi, G., 1939. Impulses in the pyramidal tract. The Journal of physiology, 97(2), pp. 153-199. Adrian, E.D., 1942. Olfactory reactions in the brain of the hedgehog. The Journal of physiology, 100(4), p.459. Amzica, F. and Steriade, M., 1998. Electrophysiological correlates of sleep delta waves. Electroencephalography and clinical neurophysiology, 107(2), pp.69-83. Andrey, P. and Maurin, Y., 2005. Free-D: an integrated environment for three-dimensional reconstruction from serial sections. Journal of neuroscience methods, 145(1), pp.233-244. Antzoulatos, E.G. and Miller, E.K., 2014. Increases in functional connectivity between prefrontal cortex and striatum during category learning. Neuron, 83(1), pp.216- 225. Asaad, W.F. and Eskandar, E.N., 2008a. Achieving behavioral control with millisecond resolution in a high-level programming environment. Journal of neuroscience methods, 173(2), pp.235-240. Asaad, W.F. and Eskandar, E.N., 2008b. A flexible software tool for temporally-precise behavioral control in Matlab. Journal of neuroscience methods, 174(2), pp.245- 258. Asaad, W.F., Rainer, G. and Miller, E.K., 2000. Task-specific neural activity in the primate prefrontal cortex. Journal of Neurophysiology, 84(1), pp.451-459. Babiloni, C., Vecchio, F., Bultrini, A., Luca Romani, G. and Rossini, P.M., 2006. Pre-and poststimulus alpha rhythms are related to conscious visual perception: a high- resolution EEG study. Cerebral cortex, 16(12), pp.1690-1700. 213 Baddeley, A., 1992. Working memory. Science, 255(5044), pp.556-559. Baddeley, A., 2003. Working memory: looking back and looking forward. Nature reviews neuroscience, 4(10), pp.829-839. Başar, E., Güntekin, B., Atagün, I., Gölbaşı, B.T., Tülay, E. and Özerdem, A., 2012. Brain’s alpha activity is highly reduced in euthymic bipolar disorder patients. Cognitive neurodynamics, 6(1), pp.11-20. Başar, E., 2012. A review of alpha activity in integrative brain function: fundamental physiology, sensory coding, cognition and pathology. International Journal of Psychophysiology, 86(1), pp.1-24. Başar, E., 2013. A review of gamma oscillations in healthy subjects and in cognitive impairment. International Journal of Psychophysiology, 90(2), pp.99-117. Bassett, D.S., Bullmore, E., Verchinski, B.A., Mattay, V.S., Weinberger, D.R. and Meyer-Lindenberg, A., 2008. Hierarchical organization of human cortical networks in health and schizophrenia. Journal of Neuroscience , 28(37), pp.9239- 9248. Bastian, M., Heymann, S. and Jacomy, M., 2009. Gephi: an open source software for exploring and manipulating networks. ICWSM, 8, pp.361-362. Bastos, A.M., Loonis, R., Kornblith, S., Lundqvist, M., Miller, E.K., 2018. Laminar recordings in frontal cortex suggest distinct layers for maintenance and control of working memory. PNAS, 115: 1117-1122. Bastos, A.M., Vezoli, J., Bosman, C.A., Schoffelen, J.M., Oostenveld, R., Dowdall, J.R., De Weerd, P., Kennedy, H. and Fries, P., 2015. Visual areas exert feedforward and feedback influences through distinct frequency channels. Neuron, 85(2), pp.390-401. Bastos, A.M. and Schoffelen, J.M., 2016. A tutorial review of functional connectivity analysis methods and their interpretational pitfalls. Frontiers in systems neuroscience, 9, p.175. Battaglia, F.P., Kalenscher, T., Cabral, H., Winkel, J., Bos, J., Manuputy, R., van Lieshout, T., Pinkse, F., Beukers, H. and Pennartz, C., 2009. The Lantern: an ultra-light micro-drive for multi-tetrode recordings in mice and other small animals. Journal of neuroscience methods, 178(2), pp.291-300. Beck, A., 1891. Oznaczenie lokalizacyi w mózgu i rdzeniu za pomoczjawisk elektrycznych [The determination of localizations in the brain and spinal cord 214 with the aid of electrical phenomena]. Rozprawy Akademii Umiejętności, wydział matematyczno-przyrodniczy II(1): 187–232. Benchenane, K., Peyrache, A., Khamassi, M., Tierney, P.L., Gioanni, Y., Battaglia, F.P. and Wiener, S.I., 2010. Coherent theta oscillations and reorganization of spike timing in the hippocampal-prefrontal network upon learning. Neuron, 66(6), pp.921-936. Benjamini, Y. and Hochberg, Y., 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal statistical society: series B (Methodological), 57(1), pp.289-300. Berger, H., 1929. Über das elektroenkephalogramm des menschen. Archiv für psychiatrie und nervenkrankheiten, 87(1), pp.527-570. Berry, R.B., Brooks, R., Gamaldo, C.E., Harding, S.M., Marcus, C. and Vaughn, B.V., 2012. The AASM manual for the scoring of sleep and associated events. Rules, Terminology and Technical Specifications, Darien, Illinois, American Academy of Sleep Medicine, 176, p.2012. Beul, S.F., Barbas, H. and Hilgetag, C.C., 2017. A predictive structural model of the primate connectome. Scientific reports, 7(1), pp.1-12. Binnie, C.D. and Stefan, H., 1999. Modern electroencephalography: its role in epilepsy management. Clinical Neurophysiology, 110(10), pp.1671-1697. Bishop, G.H., 1936, January. The interpretation of cortical potentials. In Cold Spring Harbor Symposia on Quantitative Biology (Vol. 4, pp. 305-319). Cold Spring Harbor Laboratory Press. Bisley, J.W., Zaksas, D., Droll, J.A. and Pasternak, T., 2004. Activity of neurons in cortical area MT during a memory for motion task. Journal of neurophysiology, 91(1), pp.286-300. Bonnefond, M., Kastner, S. and Jensen, O., 2017. Communication between brain areas based on nested oscillations. eneuro, 4(2). Bosman CA, Schoffelen JM, Brunet N, Oostenveld R, Bastos AM, Womelsdorf T, Rubehn B, Stieglitz T, De Weerd P, Fries P., 2012. Attentional stimulus selection through selective synchronization between monkey visual areas. Neuron, 75:875-88. Brandt, S.A. and Stark, L.W., 1997. Spontaneous eye movements during visual imagery reflect the content of the visual scene. Journal of cognitive neuroscience, 9(1), pp.27-38. 215 Brazier, M.A., 1961. A history of the electrical activity of the brain: The first half- century. Brazier, M.A., 1984. Pioneers in the discovery of evoked potentials. Electroencephalography and Clinical Neurophysiology/Evoked Potentials Section, 59(1), pp.2-8. Bremer, F., 1949. Considérations sur l'origine et la nature des⪡ ondes⪢ cérébrales. Electroencephalography and clinical neurophysiology, 1(1-4), pp.177- 193. Bressler, S.L., 1984. Spatial organization of EEGs from olfactory bulb and cortex. Electroencephalography and clinical neurophysiology, 57(3), pp.270-276. Bressler, S.L., 1995. Large-scale cortical networks and cognition. Brain Research Reviews, 20(3), pp.288-304. Bressler, S.L. and Freeman, W.J., 1980. Frequency analysis of olfactory system EEG in cat, rabbit, and rat. Electroencephalography and clinical neurophysiology, 50(1- 2), pp.19-24. Bressler, S.L. and Kelso, J.S., 2001. Cortical coordination dynamics and cognition. Trends in cognitive sciences, 5(1), pp.26-36. Bressler, S.L. and Seth, A.K., 2011. Wiener–Granger causality: a well established methodology. Neuroimage, 58(2), pp.323-329. Brockhaus-Dumke, A., Mueller, R., Faigle, U. and Klosterkoetter, J., 2008. Sensory gating revisited: relation between brain oscillations and auditory evoked potentials in schizophrenia. Schizophrenia research, 99(1-3), pp.238-249. Brodmann K., 1909. Vergleichende Lokalisationslehre der Großhirnrinde in ihren Prinzipien dargestellt auf Grund des Zellenbaues. Barth, Leipzig. Brovelli, A., Ding, M., Ledberg, A., Chen, Y., Nakamura, R. and Bressler, S.L., 2004. Beta oscillations in a large-scale sensorimotor cortical network: directional influences revealed by Granger causality. Proceedings of the National Academy of Sciences, 101(26), pp.9849-9854. Brown, E.N., Lydic, R. and Schiff, N.D., 2010. General anesthesia, sleep, and coma. New England Journal of Medicine, 363(27), pp.2638-2650. 216 Bullmore, E.T., & Bassett, D.S., 2011. Brain graphs: graphical models of the human brain connectome. Annual review of clinical psychology , 7 , pp.113-140. Bullock, T.H., 1945. Problems in the comparative study of brain waves. The Yale journal of biology and medicine, 17(5), p.657. Bullock, T.H., 1997. Signals and signs in the nervous system: the dynamic anatomy of electrical activity is probably information-rich. Proceedings of the National Academy of Sciences, 94(1), pp.1-6. Bullock, T .H., 2002. Biology of Brain waves: natural history and evolution of an information-rich sign of activity. In Advances in Electrophysiolgy in Clinical Practice and Research. Kjellberg, Inc.. Buschman, T.J., Denovellis, E.L., Diogo, C., Bullock, D. and Miller, E.K., 2012. Synchronous oscillatory neural ensembles for rules in the prefrontal cortex. Neuron, 76(4), pp.838-846. Buschman, T.J. and Miller, E.K., 2007. Top-down versus bottom-up control of attention in the prefrontal and posterior parietal cortices. science, 315(5820), pp.1860-1862. Buzsáki, G. and Draguhn, A., 2004. Neuronal oscillations in cortical networks. science, 304(5679), pp.1926-1929. Buzsaki, G., 2006. Rhythms of the Brain. Oxford University Press. Buzsáki, G., Anastassiou, C.A. and Koch, C., 2012. The origin of extracellular fields and currents—EEG, ECoG, LFP and spikes. Nature reviews neuroscience, 13(6), pp.407-420. Caton, R., 1875. The electric currents of the brain. Br Med J, 2, p.278. Caton, R., 1877. Interim report on investigations of the electric currents of the brain. Br Med J, Suppl. L: 62-65. Chapeton, J.I., Haque, R., Wittig Jr, J.H., Inati, S.K. and Zaghloul, K.A., 2019. Large- Scale communication in the human brain Is rhythmically modulated through alpha coherence. Current Biology, 29(17), pp.2801-2811. Chaudhuri, R., Knoblauch, K., Gariel, M.A., Kennedy, H. and Wang, X.J., 2015. A large- scale circuit mechanism for hierarchical dynamical processing in the primate cortex. Neuron, 88(2), pp.419-431. 217 Christophel, T.B., Klink, P.C., Spitzer, B., Roelfsema, P.R. and Haynes, J.D., 2017. The distributed nature of working memory. Trends in cognitive sciences, 21(2), pp.111-124. Christopher deCharms, R., Blake, D.T. and Merzenich, M.M., 1999. A multielectrode implant device for the cerebral cortex. Journal of neuroscience methods, 93(1), pp.27-35. Colrain, I.M., Crowley, K.E., Nicholas, C.L., Afifi, L., Baker, F.C., Padilla, M., Turlington, S.R. and Trinder, J., 2010. Sleep evoked delta frequency responses show a linear decline in amplitude across the adult lifespan. Neurobiology of aging, 31(5), pp.874-883. Colrain, I.M., Turlington, S. and Baker, F.C., 2009. Impact of alcoholism on sleep architecture and EEG power spectra in men and women. Sleep, 32(10), pp.1341- 1352. Combrisson, E. and Jerbi, K., 2015. Exceeding chance level by chance: The caveat of theoretical chance levels in brain signal classification and statistical assessment of decoding accuracy. Journal of neuroscience methods, 250, pp.126-136. Confais, J., Malfait, N., Brochier, T., Riehle, A., Kilavik, B.E., 2020. Is there an Intrinsic Relationship between LFP Beta Oscillation Amplitude and Firing Rate of Individual Neurons in Macaque Motor Cortex? Cerebral Cortex Communications, 1: 1-15. Constantinidis, C. and Procyk, E., 2004. The primate working memory networks. Cognitive, Affective, & Behavioral Neuroscience, 4(4), pp.444-465. Courtney, S.M., Ungerleider, L.G., Keil, K. and Haxby, J.V., 1997. Transient and sustained activity in a distributed neural system for human working memory. Nature, 386(6625), pp.608-611. Csicsvari, J., Henze, D.A., Jamieson, B., Harris, K.D., Sirota, A., Barthó, P., Wise, K.D. and Buzsáki, G., 2003. Massively parallel recording of unit and local field potentials with silicon-based electrodes. Journal of neurophysiology, 90(2), pp.1314-1323. da Silva, F.L., 1991. Neural mechanisms underlying brain waves: from neural membranes to networks. Electroencephalography and clinical neurophysiology, 79(2), pp.81-93. da Silva, F.L., 2013. EEG and MEG: relevance to neuroscience. Neuron, 80(5), pp.1112- 1128. 218 Danilevsky, V.Y., 1877. Investigations into the physiology of the brain. Doctoral Thesis, University of Kharkov. Dawson, G.D., 1951. A summation technique for detecting small signals in a large irregular background. J physiol, 115, p.2. Delgado-Bonal, A. and Marshak, A. 2019. Approximate Entropy and Sample Entropy: A Comprehensive Tutorial. Entropy, 21, 541 Desmedt, J.E. and Tomberg, C., 1994. Transient phase-locking of 40 Hz electrical oscillations in prefrontal and parietal human cortex reflects the process of conscious somatic perception. Neuroscience letters, 168(1-2), pp.126-129 D’Esposito, M., Postle, B.R. and Rypma, B., 2000. Prefrontal cortical contributions to working memory: evidence from event-related fMRI studies. In Executive control and the frontal lobe: Current issues (pp. 3-11). Springer, Berlin, Heidelberg. D'Esposito, M. and Postle, B.R., 2015. The cognitive neuroscience of working memory. Annual review of psychology, 66. Dickey, A.S., Suminski, A., Amit, Y. and Hatsopoulos, N.G., 2009. Single-unit stability using chronically implanted multielectrode arrays. Journal of neurophysiology, 102(2), pp.1331-1339. Donoghue, J.P., Sanes, J.N., Hatsopoulos, N.G. and Gaál, G., 1998. Neural discharge and local field potential oscillations in primate motor cortex during voluntary movements. Journal of neurophysiology, 79(1), pp.159-173. Dotson, N.M., Goodell, B., Salazar, R.F., Hoffman, S.J. and Gray, C.M., 2015. Methods, caveats and the future of large-scale microelectrode recordings in the non-human primate. Frontiers in systems neuroscience, 9, p.149. Dotson, N.M., Hoffman, S.J., Goodell, B. and Gray, C.M., 2017. A large-scale semi- chronic microdrive recording system for non-human primates. Neuron, 96(4), pp.769-782. Dotson, N.M., Hoffman, S.J., Goodell, B. and Gray, C.M., 2018. Feature-based visual short-term memory is widely distributed and hierarchically organized. Neuron, 99(1), pp.215-226. 219 Dotson, N.M., Salazar, R.F. and Gray, C.M., 2014. Frontoparietal correlation dynamics reveal interplay between integration and segregation during visual working memory. Journal of Neuroscience, 34(41), pp.13600-13613. Dreyfus, G. and Guyon, I., 2006. Assessment Methods. In: Feature Extraction: Foundations and Applications. Guyon, I., Gunn, S, Nikravesh, M. and Zadeh, L., pgs. 65-88, Eds. Springer-Verlag, Berlin Dubbelink, K.T.O., Felius, A., Verbunt, J.P., Van Dijk, B.W., Berendse, H.W., Stam, C.J. and Delemarre-van de Waal, H.A., 2008. Increased resting-state functional connectivity in obese adolescents; a magnetoencephalographic pilot study. PLoS One, 3(7), p.e2827. Dubey, A., Ray, S., 2019. Cortical Electrocorticogram (ECoG) Is a Local Signal. J. Neurosci., 39: 4299–4311. Du Bois-Reymond, E., 1848. Untersuchungen über thierische Elektricität, Berlin, Reimer, Vol. I. Eckhorn, R., Bauer, R., Jordan, W., Brosch, M., Kruse, W., Munk, M. and Reitboeck, H.J., 1988. Coherent oscillations: A mechanism of feature linking in the visual cortex?. Biological cybernetics, 60(2), pp.121-130. Edelman, G.M. and Mountcastle, V.B., 1978. The mindful brain: cortical organization and the group-selective theory of higher brain function. Massachusetts Inst of Technology Pr. Einevoll, G.T., Kayser, C., Logothetis, N.K. and Panzeri, S., 2013. Modelling and analysis of local field potentials for studying the function of cortical circuits. Nature Reviews Neuroscience, 14(11), pp.770-785. Eliades, S.J. and Wang, X., 2008. Chronic multi-electrode neural recording in free-roaming monkeys. Journal of neuroscience methods, 172(2), pp.201-214. Engel, A.K. and Fries, P., 2010. Beta-band oscillations—signalling the status quo?. Current opinion in neurobiology, 20(2), pp.156-165. Engel, A.K., Friston, K., Kelso, J.A., König, P., Kovács, I., MacDonald III, A., Miller, E.K., Phillips, W.A., Silverstein, S.M., Tallon-Baudry, C. and Triesch, J., 2010. Coordination in behavior and cognition. Engel, A.K., König, P., Kreiter, A.K. and Singer, W., 1991a. Interhemispheric synchronization of oscillatory neuronal responses in cat visual cortex. Science, pp.1177-1179. 220 Engel, A.K., König, P., Kreiter, A.K. and Singer, W., 1991b. Interhemispheric synchronization of oscillatory neuronal responses in cat visual cortex. Science, pp.1177-1179. Engel, A.K., Kreiter, A.K., König, P. and Singer, W., 1991. Synchronization of oscillatory neuronal responses between striate and extrastriate visual cortical areas of the cat. Proceedings of the National Academy of Sciences, 88(14), pp.6048- 6052. Erickson, C.A. and Desimone, R., 1999. Responses of macaque perirhinal neurons during and after visual stimulus association learning. Journal of Neuroscience, 19(23), pp.10404-10416. Ester, E.F., Serences, J.T. and Awh, E., 2009. Spatially global representations in human primary visual cortex during working memory maintenance. Journal of Neuroscience, 29(48), pp.15258-15265. Ester, E.F., Sprague, T.C. and Serences, J.T., 2015. Parietal and frontal cortex encode stimulus-specific mnemonic representations during visual working memory. Neuron, 87(4), pp.893-905. Fee, M.S. and Leonardo, A., 2001. Miniature motorized microdrive and commutator system for chronic neural recording in small animals. Journal of neuroscience methods, 112(2), pp.83-94. Feingold, J., Desrochers, T.M., Fujii, N., Harlan, R., Tierney, P.L., Shimazu, H., Amemori, K.I. and Graybiel, A.M., 2012. A system for recording neural activity chronically and simultaneously from multiple cortical and subcortical regions in nonhuman primates. Journal of neurophysiology, 107(7), pp.1979-1995. Fell, J., Klaver, P., Lehnertz, K., Grunwald, T., Schaller, C., Elger, C.E. and Fernández, G., 2001. Human memory formation is accompanied by rhinal–hippocampal coupling and decoupling. Nature neuroscience, 4(12), pp.1259-1264. Fernandez, T., Harmony, T., Gersenowies, J., Silva-Pereyra, J., Fernández-Bouzas, A., Galán, L. and Díaz-Comas, L., 2002. Sources of EEG activity during a verbal working memory task in adults and children. In Supplements to Clinical Neurophysiology (Vol. 54, pp. 269-283). Elsevier. Ferrier, D., 1874. The localization of function in the brain. Proceedings of the Royal Society of London, 22(148-155), pp.228-232. 221 Fleischl von Marxow, E., 1890. Mittheilung, betreffend die Physiologie der Hirnrinde. Centralblatt fur Physiologie, 4, pp.537-540. Frauscher, B., von Ellenrieder, N., Zelmann, R., Doležalová, I., Minotti, L., Olivier, A., Hall, J., Hoffmann, D., Nguyen, D.K., Kahane, P., Dubeau, F., Gotman, J., 2018. Atlas of the normal intracranial electroencephalogram: neurophysiological awake activity in different cortical areas. Brain, 141:1130-1144. Freeman, W.J., 1975. Mass action in the nervous system (Vol. 2004). Academic Press, New York. Freeman, W.J., 2007. Definitions of state variables and state space for brain-computer interface. Cognitive Neurodynamics, 1(1), pp.3-14. Freeman, W.J. and van Dijk, B.W., 1987. Spatial patterns of visual cortical fast EEG during conditioned reflex in a rhesus monkey. Brain research, 422(2), pp.267- 276. Friedman-Hill, S.R., Maldonado, P.E., Gray, C.M., 2000. Dynamics of Striate Cortical Activity in the Alert Macaque: I. Incidence and Stimulus-Dependence of Gamma- Band Neuronal Oscillations. Cerebral Cortex, 10:1105-1116. Fries, P., 2005. A mechanism for cognitive dynamics: neuronal communication through neuronal coherence. Trends in cognitive sciences, 9(10), pp.474-480. Fries, P., 2015. Rhythms for cognition: communication through coherence. Neuron, 88(1), pp.220-235. Fries, P., Reynolds, J.H., Rorie, A.E., Desimone, R., 2001. Modulation of oscillatory neuronal synchronization by selective visual attention. Science, 291: 1560–1563. Fries, P., Roelfsema, P.R., Engel, A.K., König, P. and Singer, W., 1997. Synchronization of oscillatory responses in visual cortex correlates with perception in interocular rivalry. Proceedings of the National Academy of Sciences, 94(23), pp.12699- 12704. Friston, K.J., 1997. Another neural code?. Neuroimage, 5(3), pp.213-220. Fuster, J.M. and Alexander, G.E., 1971. Neuron activity related to short-term memory. Science, 173(3997), pp.652-654. Fuster, J.M. and Bressler, S.L., 2012. Cognit activation: a mechanism enabling temporal integration in working memory. Trends in cognitive sciences, 16(4), pp.207-218. 222 Fuster, J.M. and Jervey, J.P., 1981. Inferotemporal neurons distinguish and retain behaviorally relevant features of visual stimuli. Science, 212(4497), pp.952-955. Garvin, J.S. and Amador, L.V., 1949. Electrocorticograms of the cytoarchitectural areas of macaca mulatta. Journal of neurophysiology, 12(6), pp.425-433. Galvani, L., 1791. De viribus electricitatis in motu musculari: Commentarius. Bologna: Tip. Istituto delle Scienze, 1791; 58 p.: 4 tavv. ft; in 4.; DCC. f. 70. Gonzalez, S.L., de Peralta, R.G., Thut, G., Millán, J.D.R., Morier, P. and Landis, T., 2006. Very high frequency oscillations (VHFO) as a predictor of movement intentions. Neuroimage, 32(1), pp.170-179. Goodell A.B, Gray C.M., 2015. Next generation of large-scale, semi-chronic micromanipulator systems for measuring neuronal activity in distributed neural circuits in behaving, non-human primates. Soc. Neurosci. Abs., 267.06. Granger, C.W., 1969. Some recent developments in a concept of causality," Journal of Econometrics, vol. 39, 1988, pp. 199—211; and. Investigating causal relations by econometric models and cross-spectral methods,” Econometrica, 37, pp.424-38. Grastyan, E., Lissak, K., Madarasz, I. and Donhoffer, H., 1959. Hippocampal electrical activity during the development of conditioned reflexes. Electroencephalography and clinical neurophysiology, 11(3), pp.409-430. Gray, C.M., 1994. Synchronous oscillations in neuronal systems: mechanisms and functions. Journal of computational neuroscience, 1(1-2), pp.11-38. Gray, C.M. and Goodell, B., 2007. A high-density, large-scale, distributed recording system for semi-chronic monitoring of cortical and sub-cortical neuronal activity in alert monkeys. In Soc Neurosci Abstr (Vol. 33, No. 624.8). Gray, C.M. and Goodell, B., 2011. Spatiotemporal dynamics of synchronous activity across multiple areas of the visual cortex in the alert monkey. The Dynamic Brain: An Exploration of Neuronal Variability and Its Functional Significance, pp.233-254. Gray, C.M., Goodell, B. and Lear, A., 2007. Multichannel micromanipulator and chamber system for recording multineuronal activity in alert, non-human primates. Journal of neurophysiology, 98(1), pp.527-536. Gray, C. M. & Singer, W. (1987) Soc. Neurosci. Abstr. 404.3. 223 Gray, C.M. and Singer, W., 1989. Stimulus-specific neuronal oscillations in orientation columns of cat visual cortex. Proceedings of the National Academy of Sciences, 86(5), pp.1698-1702. Gray, C.M., 1999. The temporal correlation hypothesis of visual feature integration: still alive and well. Neuron, 24(1), pp.31-47. Grey Walter, W., 1963. The living brain. Groppe, D.M., Bickel, S., Keller, C.J., Jain, S.K., Hwang, S.T., Harden, C., Ashesh D Mehta, A.D., 2013. Dominant frequencies of resting human brain activity as measured by the electrocorticogram. Neuroimage, 79:223-33. Guo, Z.V., Inagaki, H.K., Daie, K., Druckmann, S., Gerfen, C.R., and Svoboda, K. (2017). Maintenance of persistent activity in a frontal thalamocortical loop. Nature 545, 181-186. Haider, B. and McCormick, D.A., 2009. Rapid neocortical dynamics: cellular and network mechanisms. Neuron, 62(2), pp.171-189. Harmony, T., 2013. The functional significance of delta oscillations in cognitive processing. Frontiers in integrative neuroscience, 7, p.83. Harmony, T., Fernández, T., Silva, J., Bernal, J., Díaz-Comas, L., Reyes, A., Marosi, E., Rodríguez, M. and Rodríguez, M., 1996. EEG delta activity: an indicator of attention to internal processing during performance of mental tasks. International journal of psychophysiology, 24(1-2), pp.161-171. Harrison, S.A. and Tong, F., 2009. Decoding reveals the contents of visual working memory in early visual areas. Nature, 458(7238), pp.632-635. Hartline, H.K., 1940. The receptive fields of optic nerve fibers. American Journal of Physiology-Legacy Content, 130(4), pp.690-699. Hasson, U., Chen, J. and Honey, C.J., 2015. Hierarchical process memory: memory as an integral component of information processing. Trends in cognitive sciences, 19(6), pp.304-313. Hebb, D.O., 1949. John Wiley & Sons; 1949. The organization of behavior: A neuropsychological approach. Hernández, A., Nácher, V., Luna, R., Alvarez, M., Zainos, A., Cordero, S., Camarillo, L., Vázquez, Y., Lemus, L. and Romo, R., 2008. Procedure for recording the simultaneous activity of single neurons distributed across cortical areas during 224 sensory discrimination. Proceedings of the National Academy of Sciences, 105(43), pp.16785-16790. Herreras, O., 2016. Local field potentials: myths and misunderstandings. Frontiers in neural circuits, 10, p.101. Helmholtz, H., 1850. Über die Fortpflanzungsgeschwindigkeit der Nervenreizung. Bericht über die Bekanntmachung geeigneten Verhandlungen der Königlichen Preussischen Akademie der Wissenschaften zu Berlin, 14, pp.14-15. Hodgkin, A.L. and Huxley, A.F., 1945. Resting and action potentials in single nerve fibres. The Journal of physiology, 104(2), p.176. Hodgkin, A.L. and Huxley, A.F., 1953. Movement of radioactive potassium and membrane current in a giant axon. The Journal of physiology, 121(2), p.403. Hodgkin, A.L., Huxley, A.F. and Katz, B., 1952. Measurement of current‐voltage relations in the membrane of the giant axon of Loligo. The Journal of physiology, 116(4), pp.424-448. Hoerzer, G.M., Liebe, S., Schloegl, A., Logothetis, N.K. and Rainer, G., 2010. Directed coupling in local field potentials of macaque v4 during visual short-term memory revealed by multivariate autoregressive models. Frontiers in computational neuroscience, 4, p.14. Hoffman, K.L. and McNaughton, B.L., 2002. Coordinated reactivation of distributed memory traces in primate neocortex. Science, 297(5589), pp.2070-2073. Hogan, M.J., Swanwick, G.R., Kaiser, J., Rowan, M. and Lawlor, B., 2003. Memory- related EEG power and coherence reductions in mild Alzheimer's disease. International Journal of Psychophysiology, 49(2), pp.147-163. Honey, C.J., Thesen, T., Donner, T.H., Silbert, L.J., Carlson, C.E., Devinsky, O., Doyle, W.K., Rubin, N., Heeger, D.J. and Hasson, U., 2012. Slow cortical dynamics and the accumulation of information over long timescales. Neuron, 76(2), pp.423-434. Howard, M.W., Rizzuto, D.S., Caplan, J.B., Madsen, J.R., Lisman, J., Aschenbrenner- Scheibe, R., Schulze-Bonhage, A. and Kahana, M.J., 2003. Gamma oscillations correlate with working memory load in humans. Cerebral cortex, 13(12), pp.1369-1374. Hubel, D.H. and Wiesel, T.N., 1959. Receptive fields of single neurones in the cat's striate cortex. The Journal of physiology, 148(3), pp.574-591. 225 Ito, J., Maldonado, P.E. and Grün, S., 2013. Cross-frequency interaction of the eye- movement related LFP signals in V1 of freely viewing monkeys. Frontiers in systems neuroscience, 7, p.1. Jacob, S.N., Hähnke, D. and Nieder, A., 2018. Structuring of abstract working memory content by fronto-parietal synchrony in primate cortex. Neuron, 99(3), pp. 588- 597. Jacomy, M., Venturini, T., Heymann, S. and Bastian, M., 2014. ForceAtlas2, a continuous graph layout algorithm for handy network visualization designed for the Gephi software. PloS one, 9(6), p.e98679. Jasper, H.H. and Andrews, H.L., (1936) Human Brain Rhythms: I. Recording Techniques and Preliminary Results, J. Gen. Psychol., 14:98-126. Jasper, H.H. and Andrews, H.L., 1938. Electro-encephalography: III. Normal differentiation of occipital and precentral regions in man. Archives of Neurology & Psychiatry, 39(1), pp.96-115. Jasper, H. and Penfield, W., 1949. Electrocorticograms in man: effect of voluntary movement upon the electrical activity of the precentral gyrus. Archiv für Psychiatrie und Nervenkrankheiten, 183(1-2), pp.163-174. Jensen, O., Gelfand, J., Kounios, J. and Lisman, J.E., 2002. Oscillations in the alpha band (9–12 Hz) increase with memory load during retention in a short-term memory task. Cerebral cortex, 12(8), pp.877-882. Jensen, O. and Mazaheri, A., 2010. Shaping functional architecture by oscillatory alpha activity: gating by inhibition. Frontiers in human neuroscience, 4, p.186. Jensen, O. and Tesche, C.D., 2002. Frontal theta activity in humans increases with memory load in a working memory task. European journal of Neuroscience, 15(8), pp.1395-1399. Jung, R. and Kornmüller, A.E., 1938. Eine methodik der ableitung iokalisierter potentialschwankungen aus subcorticalen hirngebieten. Archiv für Psychiatrie und Nervenkrankheiten, 109(1), pp.1-30. Kanta, V., Pare, D. and Headley, D.B., 2019. Closed-loop control of gamma oscillations in the amygdala demonstrates their role in spatial memory consolidation. Nature communications, 10(1), pp.1-14. 226 Karakaş, S. and Barry, R.J., 2017. A brief historical perspective on the advent of brain oscillations in the biological and psychological disciplines. Neuroscience & Biobehavioral Reviews, 75, pp.335-347. Katzner, S., Nauhaus, I., Benucci, A., Bonin, V., Ringach, D.L. and Carandini, M., 2009. Local origin of field potentials in visual cortex. Neuron, 61(1), pp.35-41. Keitel, A., Gross, J., 2016. Individual Human Brain Areas Can Be Identified from Their Characteristic Spectral Activation Fingerprints. PLoS Biol., 14: e1002498. Kilavik, B.E., Ponce-Alvarez, A., Trachel, R., Confais, J., Takerkart, S. and Riehle, A., 2012. Context-related frequency modulations of macaque motor cortical LFP beta oscillations. Cerebral cortex, 22(9), pp.2148-2159. Kim, T., Troyk, P.R. and Bak, M., 2006, August. Active floating micro electrode arrays (AFMA). In Engineering in Medicine and Biology Society, 2006. EMBS'06. 28th Annual International Conference of the IEEE (pp. 2807-2810). IEEE. Kipke, D.R., Vetter, R.J., Williams, J.C. and Hetke, J.F., 2003. Silicon-substrate intracortical microelectrode arrays for long-term recording of neuronal spike activity in cerebral cortex. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 11(2), pp.151-155. Klimesch, W., 1999. EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis. Brain research reviews, 29(2-3), pp.169-195. Klimesch, W., Doppelmayr, M., Russegger, H., and Pachinger, T., 1996. Theta band power in the human scalp EEG and the encoding of new information. Neuroreport, 7, pp.1235-1240. Kloosterman, F., Davidson, T.J., Gomperts, S.N., Layton, S.P., Hale, G., Nguyen, D.P. and Wilson, M.A., 2009. Micro-drive array for chronic in vivo recording: drive fabrication. JoVE (Journal of Visualized Experiments), (26), pp.e1094-e1094. Knyazev, G.G., 2007. Motivation, emotion, and their inhibitory control mirrored in brain oscillations. Neuroscience & Biobehavioral Reviews, 31(3), pp.377-395. Koch, C., 2004. The quest for consciousness. Engineering and Science, 67(2), pp.28-34. König, P., 1994. A method for the quantification of synchrony and oscillatory properties of neuronal activity. Journal of neuroscience methods, 54(1), pp.31-37 König, P., Engel, A.K. and Singer, W., 1996. Integrator or coincidence detector? The role of the cortical neuron revisited. Trends in neurosciences, 19(4), pp.130-137. 227 Kornmüller, A.E., 1933. Bioelektrische Erscheinungen architektonischer Felder. Eine Methode der Lokalisation auf der Großhirnrinde. Deutsche Zeitschrift für Nervenheilkunde, 130(1-4), pp.44-60. Krupa, D.J., Wiest, M.C., Shuler, M.G., Laubach, M. and Nicolelis, M.A., 2004. Layer- specific somatosensory cortical activation during active tactile discrimination. Science, 304(5679), pp.1989-1992. Lachaux, J.P., Rodriguez, E., Martinerie, J. and Varela, F.J., 1999. Measuring phase synchrony in brain signals. Human brain mapping, 8(4), pp.194-208. Laeng, B. and Teodorescu, D.S., 2002. Eye scanpaths during visual imagery reenact those of perception of the same visual scene. Cognitive Science, 26(2), pp.207- 231. Lara, A.H. and Wallis, J.D., 2015. The role of prefrontal cortex in working memory: a mini review. Frontiers in systems neuroscience, 9, p.173. Lee, H., Simpson, G.V., Logothetis, N.K. and Rainer, G., 2005. Phase locking of single neuron activity to theta oscillations during working memory in monkey extrastriate visual cortex. Neuron, 45(1), pp.147-156. Le Van Quyen, M., Martinerie, J., Navarro, V., Boon, P., D'Havé, M., Adam, C., Renault, B., Varela, F. and Baulac, M., 2001. Anticipation of epileptic seizures from standard EEG recordings. The Lancet, 357(9251), pp.183-188. Lewicki MS. A review of methods for spike sorting: the detection and classification of neural action potentials. Network: Computation in Neural Systems. 1998 Jan 1;9(4):R53-78. Liang, Z., Wang, Y., Sun, X., Li, D., Voss, L.J., Sleigh, J.W., Hagihira, S. and Li, X., 2015. EEG entropy measures in anesthesia. Frontiers in computational neuroscience, 9, p.16. Liebe, S., Hoerzer, G.M., Logothetis, N.K. and Rainer, G., 2012. Theta coupling between V4 and prefrontal cortex predicts visual short-term memory performance. Nature neuroscience, 15(3), p.456. Lilly, J.C., 1950. A method of recording the moving electrical potential gradients in the brain: the 25 channel bavatron and electroiconograms. In Conference on Electronic Instrumentation in Nucleonics and Medicine. New York: American Institute of Electronic Engineering (pp. 37-3). 228 Lilly, J.C., Hughes, J.R., Alvord Jr, E.C. and Galkin, T.W., 1955. Brief, noninjurious electric waveform for stimulation of the brain. Science. Lilly, J.C., 1958. Electrode and cannulae implantation in the brain by a simple percutaneous method. Science. Lilly, J.C., 1997. The Scientist: A metaphysical autobiography. Ronin Publishing. Luck, S.J. and Vogel, E.K., 2013. Visual working memory capacity: from psychophysics and neurobiology to individual differences. Trends in cognitive sciences, 17(8), pp.391-400. Lundqvist, M., Rose, J., Herman, P., Brincat, S.L., Buschman, T.J. and Miller, E.K., 2016. Gamma and beta bursts underlie working memory. Neuron, 90(1), pp.152- 164. Magoun, H.W., 1961. Brain Mechanisms for Wakefulness. Brit. J. Anaesth., 33: 183-193. Maris, E., Fries, P. and van Ede, F., 2016. Diverse phase relations among neuronal rhythms and their potential function. Trends in neurosciences, 39(2), pp.86-99. Markov, N.T., Ercsey-Ravasz, M.M., Ribeiro Gomes, A.R., Lamy, C., Magrou, L., Vezoli, J., Misery, P., Falchier, A., Quilodran, R., Gariel, M.A. and Sallet, J., Gamanut, R., Huissoud, C., Clavagnier, S., Giroud, P., Sappey-Marinier, D., Barone, P., Dehay, C., Toroczkai, Z., Knoblauch, K., Van Essen, D.C., and Kennedy, H., 2014. A weighted and directed interareal connectivity matrix for macaque cerebral cortex. Cerebral cortex, 24(1), pp.17-36. Markov, N.T., Vezoli, J., Chameau, P., Falchier, A., Quilodran, R., Huissoud, C., Lamy, C., Misery, P., Giroud, P., Ullman, S. and Barone, P., 2014b. Anatomy of hierarchy: feedforward and feedback pathways in macaque visual cortex. Journal of Comparative Neurology, 522(1), pp.225-259. Markowitz, D.A., Wong, Y.T., Gray, C.M. and Pesaran, B., 2011. Optimizing the decoding of movement goals from local field potentials in macaque cortex. Journal of Neuroscience, 31(50), pp.18412-18422. Marshall, W.H., Woolsey, C.N. and Bard, P., 1937. Cortical representation of tactile sensibility as indicated by cortical potentials. Science, 85(2207), pp.388-390. Martinez-Conde, S., Otero-Millan, J. and Macknik, S.L., 2013. The impact of microsaccades on vision: towards a unified theory of saccadic function. Nature Reviews Neuroscience, 14(2), pp.83-96. 229 Mathalon, D.H. and Sohal, V.S., 2015. Neural oscillations and synchrony in brain dysfunction and neuropsychiatric disorders: it’s about time. JAMA psychiatry, 72(8), pp.840-844. Maurer, K. and Dierks, T., 2012. Atlas of brain mapping: topographic mapping of EEG and evoked potentials. Springer Science & Business Media. Maynard, E.M., Hatsopoulos, N.G., Ojakangas, C.L., Acuna, B.D., Sanes, J.N., Normann, R.A. and Donoghue, J.P., 1999. Neuronal interactions improve cortical population coding of movement direction. Journal of Neuroscience, 19(18), pp.8083-8093. Mendoza, G., Peyrache, A., Gámez, J., Prado, L., Buzsáki, G. and Merchant, H., 2016. Recording extracellular neural activity in the behaving monkey using a semichronic and high-density electrode system. Journal of Neurophysiology, 116(2), pp.563- 574. Mendoza-Halliday, D., Torres, S. and Martinez-Trujillo, J.C., 2014. Sharp emergence of feature-selective sustained activity along the dorsal visual pathway. Nature neuroscience, 17(9), p. 1255. Mesulam, M.M., 1990. Large‐scale neurocognitive networks and distributed processing for attention, language, and memory. Annals of Neurology: Official Journal of the American Neurological Association and the Child Neurology Society, 28(5), pp.597-613. Miller, E.K., Erickson, C.A. and Desimone, R., 1996. Neural mechanisms of visual working memory in prefrontal cortex of the macaque. Journal of neuroscience, 16(16), pp.5154-5167. Miller, E.K., Li, L. and Desimone, R., 1991. A neural mechanism for working and recognition memory in inferior temporal cortex. Science, 254(5036), pp.1377- 1379. Miller, E.K., Li, L. and Desimone, R., 1993. Activity of neurons in anterior inferior temporal cortex during a short-term memory task. Journal of Neuroscience, 13(4), pp.1460-1478. Miller, E.K. and Wilson, M.A., 2008. All my circuits: using multiple electrodes to understand functioning neural networks. Neuron, 60(3), pp.483-488. Milner, P.M., 1974. A model for visual shape recognition. Psychological review, 81(6), p.521. 230 Mitzdorf, U., 1985. Current source-density method and application in cat cerebral cortex: investigation of evoked potentials and EEG phenomena. Physiological reviews, 65(1), pp.37-100. Miyashita, Y. and Chang, H.S., 1988. Neuronal correlate of pictorial short-term memory in the primate temporal cortexYasushi Miyashita. Nature, 331(6151), pp.68-70. Mizumori, S.J.Y., Perez, G.M., Alvarado, M.C., Barnes, C.A. and McNaughton, B.L., 1990. Reversible inactivation of the medial septum differentially affects two forms of learning in rats. Brain research, 528(1), pp.12-20. Moran, R.J., Campo, P., Maestu, F., Reilly, R.B., Dolan, R.J. and Strange, B.A., 2010. Peak frequency in the theta and alpha bands correlates with human working memory capacity. Frontiers in human neuroscience, 4, p.200. Mountcastle, V., 1978. An organizing principle for cerebral function: the unit module and the distributed system. The mindful brain. Mountcastle, V.B., Reitboeck, H.J., Poggio, G.F. and Steinmetz, M.A., 1991. Adaptation of the Reitboeck method of multiple microelectrode recording to the neocortex of the waking monkey. Journal of neuroscience methods, 36(1), pp.77-84. Mulliken, G.H., Bichot, N.P., Ghadooshahy, A., Sharma, J., Kornblith, S., Philcock, M. and Desimone, R., 2015. Custom-fit radiolucent cranial implants for neurophysiological recording and stimulation. Journal of neuroscience methods, 241, pp.146-154. Munk, M.H., Linden, D.E., Muckli, L., Lanfermann, H., Zanella, F.E., Singer, W. and Goebel, R., 2002. Distributed cortical systems in visual short-term memory revealed by event-related functional magnetic resonance imaging. Cerebral Cortex, 12(8), pp.866-876. Murray, J.D., Bernacchia, A., Freedman, D.J., Romo, R., Wallis, J.D., Cai, X., Padoa- Schioppa, C., Pasternak, T., Seo, H., Lee, D. and Wang, X.J., 2014. A hierarchy of intrinsic timescales across primate cortex. Nature neuroscience, 17(12), pp.1661-1663. Murthy, V.N. and Fetz, E.E., 1992. Coherent 25-to 35-Hz oscillations in the sensorimotor cortex of awake behaving monkeys. Proceedings of the National Academy of Sciences, 89(12), pp.5670-5674. Murthy, V.N., Fetz, E.E., 1996. Oscillatory activity in sensorimotor cortex of awake monkeys: synchronization of local field potentials and relation to behavior. J. Neurophysiol., 76:3949-3967. 231 Musallam, S., Bak, M.J., Troyk, P.R. and Andersen, R.A., 2007. A floating metal microelectrode array for chronic implantation. Journal of neuroscience methods, 160(1), pp.122-127. Musk, E., 2019. Neuralink. An integrated brain-machine interface platform with thousands of channels. Nguyen, D.P., Layton, S.P., Hale, G., Gomperts, S.N., Davidson, T.J., Kloosterman, F. and Wilson, M.A., 2009. Micro-drive array for chronic in vivo recording: tetrode assembly. JoVE (Journal of Visualized Experiments), (26), pp.e1098-e1098. Nicolelis, M.A., Ghazanfar, A.A., Faggin, B.M., Votaw, S. and Oliveira, L.M., 1997. Reconstructing the engram: simultaneous, multisite, many single neuron recordings. Neuron, 18(4), pp.529-537. Nicolelis, M.A., Dimitrov, D., Carmena, J.M., Crist, R., Lehew, G., Kralik, J.D. and Wise, S.P., 2003. Chronic, multisite, multielectrode recordings in macaque monkeys. Proceedings of the National Academy of Sciences, 100(19), pp.11041- 11046. Niedermeyer, E., 2005. The normal EEG of the waking adult. Electroencephalography: Basic principles, clinical applications, and related fields, 167, pp.155-164. Nordhausen, C.T., Maynard, E.M. and Normann, R.A., 1996. Single unit recording capabilities of a 100 microelectrode array. Brain research, 726(1), pp.129-140. Nunez, P.L., Srinivasan, R., Westdorp, A.F., Wijesinghe, R.S., Tucker, D.M., Silberstein, R.B. and Cadusch, P.J., 1997. EEG coherency: I: statistics, reference electrode, volume conduction, Laplacians, cortical imaging, and interpretation at multiple scales. Electroencephalography and clinical neurophysiology, 103(5), pp.499- 515. Nunez, P.L. and Srinivasan, R., 2006. Electric fields of the brain: the neurophysics of EEG. Oxford University Press, USA. O'Keefe, J. and Dostrovsky, J., 1971. The hippocampus as a spatial map: Preliminary evidence from unit activity in the freely-moving rat. Brain research. O'Keefe, J., 1976. Place units in the hippocampus of the freely moving rat. Experimental neurology, 51(1), pp.78-109. O'Keefe, J. and Recce, M.L., 1993. Phase relationship between hippocampal place units and the EEG theta rhythm. Hippocampus, 3(3), pp.317-330. 232 Oostenveld, R., Fries, P., Maris, E. and Schoffelen, J.M., 2011. FieldTrip: open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data. Computational intelligence and neuroscience, 2011. Owen, A.M., Stern, C.E., Look, R.B., Tracey, I., Rosen, B.R. and Petrides, M., 1998. Functional organization of spatial and nonspatial working memory processing within the human lateral frontal cortex. Proceedings of the National Academy of Sciences, 95(13), pp.7721-7726. Parthasarathy, A., Herikstad, R., Bong, J.H., Medina, F.S., Libedinsky, C., and Yen, S.C. (2017). Mixed selectivity morphs population codes in prefrontal cortex. Nat. Neuro.20, 1770-1779. Peng, X., Sereno, M.E., Silva, A.K., Lehky, S.R., and Sereno, A.B. (2008). Shape selectivity in primate frontal eye field. J. Neurophys. 100, 796-814. Pesaran, B., Pezaris, J.S., Sahani, M., Mitra, P.P. and Andersen, R.A., 2002. Temporal structure in neuronal activity during working memory in macaque parietal cortex. Nature neuroscience, 5(8), pp.805-811. Pesaran, B., Vinck, M., Einevoll, G.T., Sirota, A., Fries, P., Siegel, M., Truccolo, W., Schroeder, C.E. and Srinivasan, R., 2018. Investigating large-scale brain dynamics using field potential recordings: analysis and interpretation. Nature neuroscience, 21(7), pp.903-919. Pfurtscheller, G., 1977. Graphical display and statistical evaluation of event-related desynchronization (ERD). Electroencephalography and clinical neurophysiology, 43(5), pp.757-760. Pfurtscheller G, Aranibar A. Evaluation of event-related desynchronization (ERD) preceding and following voluntary self-paced movement. Electroencephalography and clinical neurophysiology. 1979 Feb 1;46(2):138-46. Pfurtscheller, G. and Da Silva, F.L., 1999. Event-related EEG/MEG synchronization and desynchronization: basic principles. Clinical neurophysiology, 110(11), pp.1842- 1857. Pfurtscheller, G., Stancak Jr, A. and Neuper, C., 1996. Event-related synchronization (ERS) in the alpha band—an electrophysiological correlate of cortical idling: a review. International journal of psychophysiology, 24(1-2), pp.39-46. Piccolino, M., 1998. Animal electricity and the birth of electrophysiology: the legacy of Luigi Galvani. Brain research bulletin, 46(5), pp.381-407. 233 Pilon, M., Zadra, A., Joncas, S. and Montplaisir, J., 2006. Hypersynchronous delta waves and somnambulism: brain topography and effect of sleep deprivation. Sleep, 29(1), pp.77-84. Pipa, G., Städtler, E.S., Rodriguez, E.F., Waltz, J.A., Muckli, L., Singer, W., Goebel, R. and Munk, M.H., 2009. Performance-and stimulus-dependent oscillations in monkey prefrontal cortex during short-term memory. Frontiers in integrative neuroscience, 3, p.25. Postle, B.R., Druzgal, T.J. and D'Esposito, M., 2003. Seeking the neural substrates of visual working memory storage. Cortex, 39(4-5), pp.927-946. Postle, B.R., 2006. Working memory as an emergent property of the mind and brain. Neuroscience, 139(1), pp.23-38. Power, J.D., et al., 2011. Functional network organization of the human brain. Neuron 72(4), pp.665-678. Pravdich-Neminsky, V.V., 1913. Experiments on the registration of the electrical phenomena of the mammalian brain. Zbl. Physiol, 27, pp.951-960. Prut, Y., Vaadia, E., Bergman, H., Haalman, I., Slovin, H. and Abeles, M., 1998. Spatiotemporal structure of cortical activity: properties and behavioral relevance. Journal of neurophysiology, 79(6), pp.2857-2874. Purushothaman, G., Scott, B.B. and Bradley, D.C., 2006. An acute method for multielectrode recording from the interior of sulci and other deep brain areas. Journal of neuroscience methods, 153(1), pp.86-94. Raducanu, B.C., Yazicioglu, R.F., Lopez, C.M., Ballini, M., Putzeys, J., Wang, S., Andrei, A., Rochus, V., Welkenhuysen, M., Helleputte, N.V. and Musa, S., 2017. Time multiplexed active neural probe with 1356 parallel recording sites. Sensors, 17(10), p.2388. Raghavachari, S., Kahana, M.J., Rizzuto, D.S., Caplan, J.B., Kirschen, M.P., Bourgeois, B., Madsen, J.R. and Lisman, J.E., 2001. Gating of human theta oscillations by a working memory task. Journal of Neuroscience, 21(9), pp.3175-3183. Ramón, S. and Cajal, Y., 1904. Textura del Sistema Nervioso del Hombre y de los Vertebrados. Imprenta y Librería de Nicolás Moya, Madrid (1899–1904). Richman, J.S. and Moorman, J.R., 2000. Physiological time-series analysis using approximate entropy and sample entropy. American Journal of Physiology-Heart and Circulatory Physiology, 278(6), pp.H2039-H2049. 234 Riddle, J., Scimeca, J.M., Cellier, D., Dhanani, S. and D’Esposito, M., 2020. Causal evidence for a role of theta and alpha oscillations in the control of working memory. Current Biology. Rigotti, M., Barak, O., Warden, M.R., Wang, X.J., Daw, N.D., Miller, E.K. and Fusi, S., 2013. The importance of mixed selectivity in complex cognitive tasks. Nature, 497(7451), pp.585-590. Roelfsema, P.R., Engel, A.K. and Singer, W., 1997. Visuomotor integration is associated with zero time-lag synchronization among cortical areas. Nature, 385(6612), p.157. Rossant, C., Kadir, S.N., Goodman, D.F., Schulman, J., Hunter, M.L., Saleem, A.B., Grosmark, A., Belluscio, M., Denfield, G.H., Ecker, A.S. and Tolias, A.S., 2016. Spike sorting for large, dense electrode arrays. Nature neuroscience, 19(4), pp.634-641. Rousche, P.J., Pellinen, D.S., Pivin, D.P., Williams, J.C., Vetter, R.J. and Kipke, D.R., 2001. Flexible polyimide-based intracortical electrode arrays with bioactive capability. IEEE Transactions on biomedical engineering, 48(3), pp.361-371. Roux, F. and Uhlhaas, P.J., 2014. Working memory and neural oscillations: alpha– gamma versus theta–gamma codes for distinct WM information?. Trends in cognitive sciences, 18(1), pp.16-25. Roux, F., Wibral, M., Singer, W., Aru, J. and Uhlhaas, P.J., 2013. The phase of thalamic alpha activity modulates cortical gamma-band activity: evidence from resting- state MEG recordings. Journal of Neuroscience, 33(45), pp.17827-17835. Rubinov, M. and Sporns, O., 2010. Complex network measures of brain connectivity: uses and interpretations. Neuroimage, 52(3), pp.1059-1069. Sakkalis, V., 2011. Review of advanced techniques for the estimation of brain connectivity measured with EEG/MEG. Computers in biology and medicine, 41(12), pp.1110-1117. Salazar, R.F., Dotson, N.M., Bressler, S.L. and Gray, C.M., 2012. Content-specific fronto- parietal synchronization during visual working memory. Science, 338(6110), pp.1097-1100. Sarma, A., Masse, N.Y., Wang, X.J. and Freedman, D.J., 2016. Task-specific versus generalized mnemonic representations in parietal and prefrontal cortices. Nature neuroscience, 19(1), pp.143-149. 235 Sauseng, P., Klimesch, W., Schabus, M. and Doppelmayr, M., 2005. Fronto-parietal EEG coherence in theta and upper alpha reflect central executive functions of working memory. International journal of Psychophysiology, 57(2), pp.97-103. Schroeder, C.E. and Lakatos, P., 2009. Low-frequency neuronal oscillations as instruments of sensory selection. Trends in neurosciences, 32(1), pp.9-18. Schürmann, M., Basar-Eroglu, C. and Basar, E., 1997. Gamma responses in the EEG: elementary signals with multiple functional correlates. Neuroreport, 8(7), pp.1793- 1796. Schwarz, D.A., Lebedev, M.A., Hanson, T.L., Dimitrov, D.F., Lehew, G., Meloy, J., Rajangam, S., Subramanian, V., Ifft, P.J., Li, Z. and Ramakrishnan, A., 2014. Chronic, wireless recordings of large-scale brain activity in freely moving rhesus monkeys. Nature methods, 11(6), pp.670-676. Serences, J.T., 2016. Neural mechanisms of information storage in visual short-term memory. Vision research, 128, pp.53-67. Serences, J.T., Ester, E.F., Vogel, E.K. and Awh, E., 2009. Stimulus-specific delay activity in human primary visual cortex. Psychological science, 20(2), pp.207- 214. Sereno, A.B. and Maunsell, J.H., 1998. Shape selectivity in primate lateral intraparietal cortex. Nature, 395(6701), pp.500-503. Seth, A.K., Barrett, A.B. and Barnett, L., 2015. Granger causality analysis in neuroscience and neuroimaging. Journal of Neuroscience, 35(8), pp.3293-3297. Siegel, M., Donner, T.H. and Engel, A.K., 2012. Spectral fingerprints of large-scale neuronal interactions. Nature Reviews Neuroscience, 13(2), pp.121-134. Siegel, M., Donner, T.H., Oostenveld, R., Fries, P. and Engel, A.K., 2007. High- frequency activity in human visual cortex is modulated by visual motion strength. Cerebral Cortex, 17(3), pp.732-741. Siegel, M., Warden, M.R. and Miller, E.K., 2009. Phase-dependent neuronal coding of objects in short-term memory. Proceedings of the National Academy of Sciences, 106(50), pp.21341-21346. Singer, W. and Gray, C.M., 1995. Visual feature integration and the temporal correlation hypothesis. Annual review of neuroscience, 18(1), pp.555-586. 236 Skarda, C.A. and Freeman, W.J., 1987. How brains make chaos in order to make sense of the world. Behavioral and brain sciences, 10(2), pp.161-173. Spaak, E., Bonnefond, M., Maier, A., Leopold, D.A., Jensen, O., 2012. Layer-Specific Entrainment of Gamma-Band Neural Activity by the Alpha Rhythm in Monkey Visual Cortex. Current Biology 22: 2313–2318. Sreenivasan, K.K., Curtis, C.E. and D’Esposito, M., 2014. Revisiting the role of persistent neural activity during working memory. Trends in cognitive sciences, 18(2), pp.82-89. Steriade, M. and McCarley, R.W., 2005. Synchronized brain oscillations leading to neuronal plasticity during waking and sleep states. Brain control of wakefulness and sleep, pp.255-344. Steriade, M., McCormick, D.A. and Sejnowski, T.J., 1993b. Thalamocortical oscillations in the sleeping and aroused brain. Science, 262(5134), pp.679-685. Steriade, M., Nuñez, A. and Amzica, F., 1993a. Intracellular analysis of relations between the slow (< 1 Hz) neocortical oscillation and other sleep rhythms of the electroencephalogram. Journal of Neuroscience, 13(8), pp.3266-3283. Super, H., Spekreijse, H. and Lamme, V.A., 2001. A neural correlate of working memory in the monkey primary visual cortex. Science, 293(5527), pp.120-124. Sutton, S., Braren, M., Zubin, J. and John, E.R., 1965. Evoked-potential correlates of stimulus uncertainty. Science, 150(3700), pp.1187-1188. Swadlow, H.A., Bereshpolova, Y., Bezdudnaya, T., Cano, M. and Stoelzel, C.R., 2005. A multi-channel, implantable microdrive system for use with sharp, ultra-fine “Reitboeck” microelectrodes. Journal of neurophysiology, 93(5), pp.2959-2965. Szabo, I., Czurko, A., Csicsvari, J., Hirase, H., Leinekugel, X. and Buzsáki, G., 2001. The application of printed circuit board technology for fabrication of multi-channel micro-drives. Journal of neuroscience methods, 105(1), pp.105-110. Tallon-Baudry, C., Bertrand, O., Delpuech, C. and Pernier, J., 1996. Stimulus specificity of phase-locked and non-phase-locked 40 Hz visual responses in human. Journal of Neuroscience, 16(13), pp.4240-4249. Tallon-Baudry, C., Bertrand, O., Peronnet, F. and Pernier, J., 1998. Induced γ-band activity during the delay of a visual short-term memory task in humans. Journal of Neuroscience, 18(11), pp.4244-4254. 237 Tallon-Baudry, C., Mandon, S., Freiwald, W.A. and Kreiter, A.K., 2004. Oscillatory synchrony in the monkey temporal lobe correlates with performance in a visual short-term memory task. Cerebral cortex, 14(7), pp.713-720. Tass, P., Rosenblum, M.G., Weule, J., Kurths, J., Pikovsky, A., Volkmann, J., Schnitzler, A. and Freund, H.J., 1998. Detection of n: m phase locking from noisy data: application to magnetoencephalography. Physical review letters, 81(15), p.3291. Tolias, A.S., Ecker, A.S., Siapas, A.G., Hoenselaar, A., Keliris, G.A. and Logothetis, N.K., 2007. Recording chronically from the same neurons in awake, behaving primates. Journal of neurophysiology, 98(6), pp.3780-3790. Townsend, G., Peloquin, P., Kloosterman, F., Hetke, J.F. and Leung, L.S., 2002. Recording and marking with silicon multichannel electrodes. Brain Research Protocols, 9(2), pp.122-129. Tuladhar, A.M., Huurne, N.T., Schoffelen, J.M., Maris, E., Oostenveld, R. and Jensen, O., 2007. Parieto‐occipital sources account for the increase in alpha activity with working memory load. Human brain mapping, 28(8), pp.785-792. Van Diepen, R., Foxe, J.J. and Mazaheri, A., 2019. The functional role of alpha-band activity in attentional processing: the current zeitgeist and future outlook. Current opinion in psychology. Van Essen, D.C., 2012. Cortical cartography and Caret software. Neuroimage, 62(2), pp.757-764. Van Kerkoerle, T., Self, M.W. and Roelfsema, P.R., 2017. Layer-specificity in the effects of attention and working memory on activity in primary visual cortex. Nature communications, 8(1), pp.1-14. van Pelt, S., Boomsma, D.I. and Fries, P., 2012. Magnetoencephalography in twins reveals a strong genetic determination of the peak frequency of visually induced gamma-band synchronization. Journal of Neuroscience, 32(10), pp.3388-3392. Venkatachalam, S., Fee, M.S. and Kleinfeld, D., 1999. Ultra-miniature headstage with 6- channel drive and vacuum-assisted micro-wire implantation for chronic recording from the neocortex. Journal of Neuroscience Methods, 90(1), pp.37-46. Vetter, R.J., Williams, J.C., Hetke, J.F., Nunamaker, E.A. and Kipke, D.R., 2004. Chronic neural recording using silicon-substrate microelectrode arrays implanted in cerebral cortex. IEEE transactions on biomedical engineering, 51(6), pp.896-904. 238 von Bonin, G. and Bailey, P. 1947. The Neocortex of Macaca mulatta. Allen RB, Kampmeier OF, Schour I, Serles ER (eds): Illinois Monographs in the Medical Sciences, Vol.5. The University of Illinois Press, Urbana, pp. 1-163. Von Der Malsburg, C., 1985. Nervous structures with dynamical links. Berichte der Bunsengesellschaft für physikalische Chemie, 89(6), pp.703-710. Von Stein, A. and Sarnthein, J., 2000. Different frequencies for different scales of cortical integration: from local gamma to long range alpha/theta synchronization. International journal of psychophysiology, 38(3), pp.301-313. Vos, B.P., Wijnants, M., Taeymans, S. and De Schutter, E., 1999. Miniature carrier with six independently moveable electrodes for recording of multiple single-units in the cerebellar cortex of awake rats. Journal of Neuroscience Methods, 94(1), pp.19-26. Wang, X.J., 2001. Synaptic reverberation underlying mnemonic persistent activity. Trends in neurosciences, 24(8), pp.455-463. Wang, X.J., 2010. Neurophysiological and computational principles of cortical rhythms in cognition. Physiological reviews, 90(3), pp.1195-1268. Walter, W.G., 1953. The living brain. Wiener, N., 1956. The theory of prediction. Modern mathematics for engineers. Wilson, M.A. and McNaughton, B.L. (1993) Dynamics of the hippocampal ensemble code for space. Science, 261(5124), pp.1055-1058. Winson, J., 1978. Loss of hippocampal theta rhythm results in spatial memory deficit in the rat. Science, 201(4351), pp.160-163. Womelsdorf, T., Schoffelen, J.M., Oostenveld, R., Singer, W., Desimone, R., Engel, A.K. and Fries, P., 2007. Modulation of neuronal interactions through neuronal synchronization. science, 316(5831), pp.1609-1612.