Neural population activity for memory: Properties, computations, and codes.

Our ability to perform a variety of difficult tasks everything from reasoning about the best chess move, to shooting a free throw, or finely dicing an onion is due to the coordinated activity of populations of neurons throughout the nervoussystem. And yet, we lack an understanding of how the brain generates the activity appropriate for achieving something as simple as pressing an elevator button. In part, this is because we do not know which neural activity patterns the brain is capable of generating, nor how that activity will change with experience. By exploring the structure and constraints on the activity patterns the brain can express, we move closer to understanding how the brain can generate the activity supportive of such a rich variety of behaviors and adaptations. Presently, in studies of arm or eye movements, we typically don't know the causal relationship between neural activity and behavior. Here we use a brain-computer interface (BCI) paradigm to study learning, because the exact relationship between neural activity and behavior is controlled by the experimenter. To generate proficient behavior, the animal must change the activity of the neurons currently being recorded. This provides us with the means to causally relate any observed structure in neural population activity with animals' performance at the task. The focus of this thesis is to characterize the structure and time course of neural population activity during learning. In the first part of this thesis, we note that just as there is more than one way to win a game of chess, the brain has many differentpatterns of neural activity it can produce to drive the same behavior. Which of these redundant options does the brain prefer? We find that the frequency with which animals used different patterns of neural population activity was remarkably similar before and after learning. This suggests that the brain's ability to take advantage of redundancy may be somewhat limited, at least within the span of a few hours. In the second part of this thesis, we asked how internal states such as our arousal, attention, and motivation interact with how we learn new tasks. We identified large, abrupt fluctuations in neural population activity in motor cortex indicative of arousal-like internal state changes, which we term \neural engagement. We find that stereotyped changes in neural engagement during learning were unrelated to goal-seeking behavior, but nevertheless influenced how quickly different tasks were learned. Overall, this thesis characterizes a variety of different constraints and influences on how populations of neurons change their activity during learning.

Temporally specific patterns of neural activity in interconnected corticolimbic structures during reward anticipation

Traumatic brain injury (TBI) is a leading cause of disability and death and survivors often suffer from long-lasting motor impairment, cognitive deficits, anxiety disorders and epilepsy. Few experimental studies have investigated long-term sequelae after TBI and relations between behavioral changes and neural activity patterns remain elusive. We examined these issues in a murine model of TBI combining histology, behavioral analyses and single-photon emission computed tomography (SPECT) imaging of regional cerebral blood flow (CBF) as a proxy for neural activity. Adult C57Bl/6N mice were subjected to unilateral cortical impact injury and investigated at early (15-57days after lesion, dal) and late (184-225 dal) post-traumatic time points. TBI caused pronounced tissue loss of the parietal cortex and subcortical structures and enduring neurological deficits. Marked perilesional astro- and microgliosis was found at 57 dal and declined at 225 dal. Motor and gait pattern deficits occurred at early time points after TBI and improved over the time. In contrast, impaired performance in the Morris water maze test and decreased anxiety-like behavior persisted together with an increased susceptibility to pentylenetetrazole-induced seizures suggesting alterations in neural activity patterns. Accordingly, SPECT imaging of CBF indicated asymmetric hemispheric baseline neural activity patterns. In the ipsilateral hemisphere, increased baseline neural activity was found in the amygdala. In the contralateral hemisphere, homotopic to the structural brain damage, the hippocampus and distinct cortex regions displayed increased baseline neural activity. Thus, regionally elevated CBF along with behavioral alterations indicate that increased neural activity is critically involved in the long-lasting consequences of TBI.

Mathematics and science are highly integrated disciplines, but the brain association between mathematics and science remains unclear. The current study used functional magnetic resonance imaging (fMRI) scans of 34 undergraduates (17 males, mean age = 20.3±1.64 years old) while they completed mathematical, physical and chemical principles, arithmetic computation, and sentence comprehension. We examined neural activation level, neural activation pattern, and neural connectivity to investigate the neural associations between mathematics and science (including physics and chemistry). The results showed that mathematical, physical, and chemical principles elicited similar neural activation level and neural activation pattern in the visuospatial network (mainly in the middle frontal gyrus and inferior parietal lobule), which were different from those elicited by sentence comprehension; those three principles also elicited similar neural activation level and neural activation pattern in the semantic network (mainly in the middle temporal gyrus, angular gyrus, inferior frontal gyrus, and dorsomedial prefrontal cortex), in contrast to that elicited by arithmetic computation. Effective connectivity analyses showed stronger connectivity between the middle temporal gyrus and inferior parietal lobule for mathematical, physical, and chemical principles than for sentence comprehension. The results suggest that visuospatial and semantic networks were critical for processing both mathematics and science.

Differential Neural Activation for Updating Rule versus Stimulus Information in Working Memory

Multivariate pattern analysis (MVPA) is an increasingly popular method of analyzing functional magnetic resonance imaging (fMRI) data1-4. Typically, the method is used to identify a subject's perceptual experience from neural activity in certain regions of the brain. For instance, it has been employed to predict the orientation of visual gratings a subject perceives from activity in early visual cortices5 or, analogously, the content of speech from activity in early auditory cortices6. Here, we present an extension of the classical MVPA paradigm, according to which perceptual stimuli are not predicted within, but across sensory systems. Specifically, the method we describe addresses the question of whether stimuli that evoke memory associations in modalities other than the one through which they are presented induce content-specific activity patterns in the sensory cortices of those other modalities. For instance, seeing a muted video clip of a glass vase shattering on the ground automatically triggers in most observers an auditory image of the associated sound; is the experience of this image in the "mind's ear" correlated with a specific neural activity pattern in early auditory cortices? Furthermore, is this activity pattern distinct from the pattern that could be observed if the subject were, instead, watching a video clip of a howling dog? In two previous studies7,8, we were able to predict sound- and touch-implying video clips based on neural activity in early auditory and somatosensory cortices, respectively. Our results are in line with a neuroarchitectural framework proposed by Damasio9,10, according to which the experience of mental images that are based on memories - such as hearing the shattering sound of a vase in the "mind's ear" upon seeing the corresponding video clip - is supported by the re-construction of content-specific neural activity patterns in early sensory cortices.

Faces contain a variety of information such as one’s identity and expression. One prevailing model suggests a functional division of labor in processing faces that different aspects of facial information are processed in anatomically separated and functionally encapsulated brain regions. Here, we demonstrate that facial identity and expression can be processed in the same region, yet with different neural coding strategies. To this end, we employed functional magnetic resonance imaging to examine two types of coding schemes, namely univariate activity and multivariate pattern, in the posterior superior temporal cortex (pSTS) - a face-selective region that is traditionally viewed as being specialized for processing facial expression. With the individual difference approach, we found that participants with higher overall face selectivity in the right pSTS were better at differentiating facial expressions measured outside of the scanner. In contrast, individuals whose spatial pattern for faces in the right pSTS was less similar to that for objects were more accurate in identifying previously presented faces. The double dissociation of behavioral relevance between overall neural activity and spatial neural pattern suggests that the functional-division-of-labor model on face processing is over-simplified, and that coding strategies shall be incorporated in a revised model.

Different activity patterns for action and language within their shared neural areas: An fMRI study on action observation and language phonology

Perceptual and cognitive content is thought to be represented in the brain by patterns of activity across populations of neurons. In order to test whether a computational model can explain a given population code and whether corresponding codes in man and monkey convey the same information, we need to quantitatively relate population-code representations. Here I give a brief introduction to representational similarity analysis, a particular approach to this problem. A population code is characterized by a representational dissimilarity matrix (RDM), which contains a dissimilarity for each pair of activity patterns elicited by a given stimulus set. The RDM encapsulates which distinctions the representation emphasizes and which it deemphasizes. By analyzing correlations between RDMs we can test models and compare different species. Moreover, we can study how representations are transformed across stages of processing and how they relate to behavioral measures of object similarity. We use an example from object vision to illustrate the method's potential to bridge major divides that have hampered progress in systems neuroscience.

Zebra finches memorize a tutor song as juveniles, and then faithfully reproduce this song for the remainder of their lives. This thesis investigates how the intricate and concerted activity of ensembles of neurons in the zebra finch brain gives rise to this complex behavior. The first section of the thesis demonstrates how neurons in RA, a pre-motor song control nucleus, coordinate their activity to generate the instantaneous spectral and temporal structure seen in the song. Similar sounds are produced by entirely different ensembles of RA neurons. Additionally, rapidly changing patterns of RA neural activity are capable of generating constant acoustic outputs. This is a new form of neural coding, previously unobserved in any other system. This unusual neural representation provides a concise account of an unexplained aspect of song learning. The second part of the thesis develops a new, noninvasive method of altering the auditory feedback heard by singing zebra finches. This type of feedback perturbation causes the normally stable songs of adult birds to deteriorate, indicating that auditory feedback is essential to song control throughout life. Restoration of the normal auditory feedback heard by these birds enables the gradual recovery of their original songs, demonstrating that a memory of the original song persists in the song control system despite the destabilization of the behavior. The song is thus maintained by an active feedback control process. The final project investigates the role of nucleus LMAN in real-time song control via auditory feedback. Neurons in nucleus LMAN were previously believed to compare auditory feedback to a memorized song template. The result of this comparison was thought to be used as an error-correction signal to update the motor control program. By recording from individual LMAN neurons while simultaneously manipulating the auditory feedback heard by the singing bird, it is shown that during singing, rather than auditory feedback, LMAN processes an efference copy of the motor commands used to generate the bird's song. This suggests a different model of the system, in which the efference copy is the basis of the real-time error-correction signal. Taken together, these three projects significantly expand our knowledge of the neural mechanisms responsible for the generation of birdsong.

Cochlear implants provide good speech discrimination ability despite highly limited amount of information they transmit compared with normal cochlea. Noise vocoded speech, simulating cochlear implants in normal hearing listeners, have demonstrated that spectrally and temporally degraded speech contains sufficient cues to provide accurate speech discrimination. We hypothesized that neural activity patterns generated in the primary auditory cortex by spectrally and temporally degraded speech sounds will account for the robust behavioral discrimination of speech. We examined the behavioral discrimination of noise vocoded consonants and vowels by rats and recorded neural activity patterns from rat primary auditory cortex (A1) for the same sounds. We report the first evidence of behavioral discrimination of degraded speech sounds by an animal model. Our results show that rats are able to accurately discriminate both consonant and vowel sounds even after significant spectral and temporal degradation. The degree of degradation that rats can tolerate is comparable to human listeners. We observed that neural discrimination based on spatiotemporal patterns (spike timing) of A1 neurons is highly correlated with behavioral discrimination of consonants and that neural discrimination based on spatial activity patterns (spike count) of A1 neurons is highly correlated with behavioral discrimination of vowels. The results of the current study indicate that speech discrimination is resistant to degradation as long as the degraded sounds generate distinct patterns of neural activity.

OPINION article Front. Integr. Neurosci., 18 August 2011 Volume 5 - 2011 | https://doi.org/10.3389/fnint.2011.00040

Which motives drive the decision of a voter to approve or reject a policy proposal? The Public Choice literature distinguishes between instrumental and expressive voting motives. We investigated the importance of these motives by analyzing the patterns of neural activity in different voting situations. We conducted a functional magnetic resonance imaging experiment that investigates neural activation at the moment of voting and used the altruism scale proposed by Tankersley, Stowe, and Huettell (2007) to differentiate between altruists and nonaltruists. Nonaltruists showed neural activation patterns that were consistent with expressive voting motives. Among nonaltruists, we also observed activation patterns that point at egoistic instrumental motives. Both results are in line with the corresponding Public Choice literature. On the other hand, we found no evidence for expressive voting motives among altruists. Their neural activation pattern was generally much less conclusive with respect to the underlying motives.

During normal development of the nervous system, a major reduction occurs in the initially excessive number of neurons and synapses. This "pruning" process is heavily influenced by patterns of electrical activity in the synaptic circuits being pruned. Many of the cell biological and molecular mechanisms involved in this activity-dependent modification of nervous system structure and function have been explicated, and the area is one of intense study. Similarly, an explosive increase has occurred in knowledge about the molecular pathogenesis of Alzheimer disease (AD). There are significant mechanistic commonalities between the normal neurodevelopmental process and development of AD. We hypothesize that abnormalities in neural activity patterns, or in the coupling between neural activity and maintenance of neurons and synaptic circuits, may be a key determinant in the pathogenesis of AD that is late in onset, sporadic in nature, and in which the genes for the presenilins and the beta amyloid precursor protein are normal. Behavioral data suggests that an active, socially engaged life-style may be associated with a reduced risk for AD. If so, mechanisms linking neural activity with synaptic circuit integrity are probably involved and provide a target for ameliorative pharmacological intervention.

Objective. Stimulation of neural activity is an important scientific and clinical tool, causally testing hypotheses and treating neurodegenerative and neuropsychiatric diseases. However, current stimulation approaches cannot flexibly control the pattern of activity in populations of neurons. To address this, we developed a model-free, adaptive, closed-loop stimulation (ACLS) system that learns to use multi-site electrical stimulation to control the pattern of activity of a population of neurons. Approach. The ACLS system combined multi-electrode electrophysiological recordings with multi-site electrical stimulation to simultaneously record the activity of a population of 5–15 multiunit neurons and deliver spatially-patterned electrical stimulation across 4–16 sites. Using a closed-loop learning system, ACLS iteratively updated the pattern of stimulation to reduce the difference between the observed neural response and a specific target pattern of firing rates in the recorded multiunits. Main results. In silico and in vivo experiments showed ACLS learns to produce specific patterns of neural activity (in ∼15 min) and was robust to noise and drift in neural responses. In visual cortex of awake mice, ACLS learned electrical stimulation patterns that produced responses similar to the natural response evoked by visual stimuli. Similar to how repetition of a visual stimulus causes an adaptation in the neural response, the response to electrical stimulation was adapted when it was preceded by the associated visual stimulus. Significance. Our results show an ACLS system that can learn, in real-time, to generate specific patterns of neural activity. This work provides a framework for using model-free closed-loop learning to control neural activity.