Earl K. Miller - US grants

In vision, there is a major division between an occipitotemporal pathway that processes information about features that identify an object ('what') and an occipitoparietal pathway that processes information about spatial relationships ('where'). This raises the important question of how information about disparate attributes of an object are integrated. A candidate for a role in integration is the prefrontal (PF) cortex. A major function of PF cortex is working memory (WM), the active holding of information "on--line," in order to guide behavior. The aim of this project is to determine how 'what' and 'where' might be integrated in the WM functions of PF cortex. Neuronal activity from the lateral PF cortex of monkeys will be recorded from multiple electrodes while monkeys perform two WM tasks. One task requires monkeys to separately retain 'what' information and 'where' information in WM. The other task requires monkeys to integrate 'what' and 'where' in WM ('what/where'). The functional topography of the lateral PF cortex will be mapped in order to determine whether the individual PF neurons are activated when monkeys perform both 'what' and 'where' WM tasks and whether there are distinct neural representations of 'what,' 'where,' and 'what/where.' The activity of simultaneously recorded neurons will be examined to search for synchronized activity between cells selective for object features and cells selective for object locations. Data from this project may yield insight into solving the "binding problem" by uncovering general mechanisms involved in forming a unified representation from disparate stimulus attributes. Furthermore, this project will yield important insights into the mechanisms underlying WM. This has also become important as recent research has established a link between neuropsychiatric disorders, WM, and PF function.

Poggio
9872936

The ability to learn to categorize and recognize objects is a key feature of the visual system of humans and higher animals. Yet, how the representation underlying these powerful visual abilities is organized and acquired is still largely unknown. The investigator and his colleagues tackle the problem of how the visual cortex learns to represent and recognize novel objects and object classes through a combination of computational, physiological and psychophysical approaches. The physiological experiments on alert monkeys rely on multielectrode recordings for which the investigators develop a tool kit of appropriate data mining techniques, based in part on their own work on learning and classification algorithms. In particular, the investigators undertake a multi-disciplinary research project consisting of four interacting components: i) Computational modeling of inferotemporal (IT) cortical neurons, extending their previous work on representations of single objects in IT; ii) cortical physiology using multiple electrodes in awake, behaving monkeys trained on between- and within-class classification tasks on novel classes of stimuli; iii) new data mining techniques for processing multiple electrode data, including classification and learning techniques; iv) visual psychophysics including fMRI studies in humans and monkeys, allowing to relate the findings from monkey physiology to object learning in the human brain.

Understanding learning in the human brain means understanding the very core of intelligence. Not only is this one of the remaining fundamental challenges in science but it is also one area where even small steps forward will have significant implications for understanding neurological diseases and disorders, and also for the future of computing and machine intelligence. However, despite enormous progress in the last decade or two, science does not yet know what various areas of the cortex do and how. Because understanding the brain, the most complex system we know, is a huge endeavor, the present project focuses on understanding a part of cortex, involved in a key and very difficult task in everybody's daily life -- even if subjectively very easy: learning to categorize and recognize visual objects such as faces or cars. Understanding how brain cells come to represent objects will be a major breakthrough for neuroscience and also for eventually designing machines capable of achieving human-like performance. More importantly, any significant progress in the specific problem of object recognition will have a major impact on the goals of the KDI program, because it will open the door to understanding broader issues of learning and intelligence in brains and machines.

DESCRIPTION: (Adapted from the Application) To understand fully the role of memory in high-order cognition, we must identify mechanisms involved in the reactivating (recalling) of information stored in long-term memory. While much research effort has been devoted to studying the neural mechanisms in the primate brain that hold a recent sensory event temporarily "in mind" (i.e., in what is known as "working memory"), almost nothing is known about the mechanisms that hold recalled information. Individual Project #3 will contribute to the Center's research objectives by identifying neural mechanisms involved in the recall and storage of long-term memories in the prefrontal and hippocampal system neurons. Elucidation of prefrontal and hippocampal system memory mechanisms is a major target of investigation for the Center. Monkeys are advantageous for studying memory and cognition because they can be trained to perform very complex tasks while rodents are advantageous for genetic manipulations that identify cellular mechanisms. We will test whether recalled long-term memories are maintained by the same mechanisms (and indeed the same neurons) involved in maintaining recent sensory information. This will be achieved by training monkeys to alternate between tasks that can be solved by maintaining sensory information in working memory and tasks that require recall of long-term memories. We will also examine neural activity as monkeys learn associations between stimuli that predicate the ability to recall them from long-term memory. By examining how neural activity evolves during this learning, we will be able to identify structures and mechanisms involved in forming these associations.

EIA-0218693
Poggio, Tomaso
MIT

Collaborative Research: CRCNS: Detection and Recognition of Objects in Visual Cortex

A three way collaboration between the laboratories of Profs. T. Poggio at MIT, D. Ferster at Northwestern University and C. Koch at Caltech is exploring and evaluating the hypotheses that the cortical organization and the neural mechanisms of visual recognition can be explained by a coherent theoretical framework built on two existing computational models for recognition and attention and, secondly, that a combination of physiological work on monkeys and cats, together with visual psychophysics can be used to test and refine the theory. The research is organized into three main projects. The work at MIT is guided by a quantitative hierarchical model of recognition, probing the relations between identification and categorization and the properties of selectivity and invariance of the neural mechanisms in IT cortex. The work at Northwestern University is testing a key prediction of the model about the nature of the pooling operation (a max operation vs. a linear sum) performed by complex cells in V1. The experiments are done in the anesthetized cat, intracellularly, to allow for a characterization of the underlying circuit and biophysical mechanisms. Finally, work at Caltech is extending the basic model of recognition by integrating it with a saliency-based attentional model. The computational component of this work, centered around the development of a quantitative model of visual recognition, constitutes the primary tool to enforce interactions between the investigators: the model suggests experiments and guides planning and interpreting new experiments.

DESCRIPTION (provided by applicant): Categories are a foundation of cognition. Imagine if we could not abstract the essence of experiences and had to learn anew about every unique object and situation. We would probably have dysfunctions like those seen in neuropsychiatric disorders like autism and schizophrenia, which are marked by an impaired ability to generalize and extract meaning from experience. While a great deal is known about the cortical organization of the processing of bottom-up sensory information, virtually nothing is known about the respective roles of different cortical areas in top-down processing, particularly categorization. This is because few neurophysiological investigations have directly compared neural correlates of categories across brain areas and, importantly, no neurophysiological study has manipulated the attributes that determine the level of categorization. We will do so while directly comparing neural activity in the prefrontal cortex (PFC) and lateral intraparietal area (LIP), two cortical areas that human and monkey studies indicate are engaged during visual categorization. We will test the hypothesis that the PFC plays the central role in extracting learned categories or that either the PFC or LIP will play the leading role in categorization depending on the level of category demand. We will record simultaneously two brain regions known to have neural correlates of categories the PFC and LIP in a task known to activate both areas in humans (dot pattern categorization). Monkeys will classify category exemplars formed by distorting prototypes of arbitrary dot patterns. Because dot patterns can be parametrically varied, we can manipulate fundamental category properties (abstractness, complexity, and number of alternative category decisions). Because virtually nothing is known about the how these manipulations affect the neural correlates of categories, any pattern of results will be informative and provide insight into the fundamental mechanisms by which the brain adds meaning to the world.

"Divide and conquer" seems to dominate many neural analyses: There are specialized systems for analyzing different types of information. Cognition requires synthesizing their results. To plan and execute complex, goal-directed behaviors we must learn "the rules of the game": predictive relationships between disparate sensory events, environmental context, the possible actions and consequences. This depends on brain systems specialized for learning and memory: the prefrontal cortex (PFC), basal ganglia (BG) and hippocampal systems (HS). Damage to any of these systems, or their disconnection, impairs rule learning. Previous studies have shown that neural correlates of acquisition and/or representation of concrete (specific) rules and higher-level abstract rules (general principles) are prevalent in the PFC, a brain region central to rule-based behaviors. But our understanding is limited by our lack of knowledge about the respective contributions of, and PFC interactions with, the other critical systems: the BG and HS. The main goal of this project is to provide that knowledge. We plan to simultaneously study neural activity from up to 28 electrodes implanted these systems while monkeys larn and follow concrete rules )conditional visuomoter associations between an object and a saccade direction) and follow abstract rules (matching and non-matching rules applied to new stimuli). This will afford a precise assessment of the respective contributions of the PFC, BG, and GS to complex goal-directed behaviors and insight into the underlying neural circuitry. Our specific aims are: 1. To compare and contrast the neural representation of concrete rules in the PFC with anatomically and functionally-related systems (BG and HS). 2. To assess the relative contributions of PFC, BG and HS to rule acquisition by comparing neural correlates of their learning. 3. To compare and contrast the neural representation of abstract rules in the PFC with the BG, and HS. As rule learning is fundamental to all higher-order behavior, data from this project has the potential to impact on our understanding of a wide range of behaviors and human and human disorders. The ability to glean rules and principles from experience is disrupted in a variety of neuropsychiatric disorders such as autism and schizophrenia. By identifying brain structures important for these abilities, discerning their relative roles, and uncovering their neural mechanisms, we can open a path to drug therapies designed to alleviate their dysfunction.

During the previous funding period, our laboratory has made great progress identifying and characterizing neural correlates of the recall of stored memories. But an important facet of memory remains almost wholly unknown: the "executive control" over memory recall. Having identified candidate neural structures and mechanisms, we are now ready to address this issue. In our proposed project, we aim to address the most fundamental question about executive control of recall, its functional organization. To do so, we will compare and contrast the time-course of neural correlates of the recall of task demands with a behavioral measure of recall: switch costs. Switch costs are seen when subjects must flexibly alternate between performing different, often incompatible, judgments and thus must repeatedly recall and update their internal representations of task demands. Switch costs are increases in reaction time (RT) and errors whose magnitude depends on the time between a cue to switch tasks and task initiation. Thus, switch costs reflect the time needed for a set of rules to be recalled from memory and supplant a set of rules already held "on-line". By comparing neural dynamics of recall with switch costs, we can determine the relationship between this activity and its impact on behavior. We will examine neural activity in brain areas where we have previously identified neural correlates for the recall of task demands: the prefrontal cortex, premotor cortex, and basal ganglia. We found differences in latencies and other neural properties between these areas that suggest different roles in recall. Comparing these properties to a behavioral measure of recall will allow us to tease apart their respective contributions. Monkeys will switch between discriminating the color and orientation; we have already found switch costs in monkeys trained under a similar paradigm. Because executive control over recall and other brain processes is central to intelligent behavior, data from this project has the potential to impact on a wide range of behaviors and human disorders. Executive control is disrupted in a variety of neuropsychiatric disorders such as autism and schizophrenia. By identifying brain structures important for these abilities, discerning their relative roles, and uncovering their neural mechanisms, we can open a path to drug and behavioral therapies designed to alleviate dysfunctions of executive control.

Visual recognition of objects, scenes and people around us depends on the visual cortex, one of the largest and most complex parts of the primate brain. This area also presents one of the major puzzles in cognitive sciences: What is the role of the cortical back-projections, that is, links from the secondary areas back into the primary areas? In this project, Drs. Tomaso Poggio and Earl Miller try to interpret the neural computations underlying complex visual recognition tasks. Their studies will help to transfer knowledge gained from animal models of visual perception toward the understanding of higher cognitive visual processes in humans. A very recent computational theory of the primate object recognition system agrees with a variety of physiological findings in different visual areas, such as V1 (primary visual cortex), V4 (visual area 4), IT (inferotemporal cortex), and PFC (prefrontal cortex). Even more surprisingly, it mimics human behavioral performance on rapid categorization of complex natural images, and performs, as well as several state-of-the-art computer vision systems on difficult recognition tasks. Considering that the model is able to account for rapid object recognition, and that it currently only uses feedforward processing, a significant puzzle concerns the computational and functional role of the abundant anatomical back-projections known to exist in cortex. The proposed project takes a two-pronged approach toward finding their function. First, experiments with the computational model of the ventral stream will provide useful insights for understanding the possible functions of the back-projections. Second, experiments with multi-unit recordings in macaques will characterize top-down effects and their timing in several different recognition tasks at the level of IT and PFC. The analysis will make use of a recently developed automatic classification technique that relates brain activity with individual visual objects. By combining results from the modeling and experimentation, computational explanations for the cognitive role of the feedback processing can be tested.

Understanding the function of back-projections in vision will help us understand the neural basis of vision itself, but it will also help us understand the global design of the brain. The intricacy of this organ continues to amaze us, and getting a handle on its mechanism through a relatively well-understood area like vision is a promising avenue for expanding our appreciation of the whole system. A further goal is to show that the interaction between computational theories, in particular of vision, and experiments will make it easier to comprehend brain functions. Such advances help us appreciate the normal function of the brain and allow us better means of helping when it does not function normally, of making machines see better, and of bringing new approaches to robotics.

DESCRIPTION (provided by applicant): Ordered thought and behavior rests on the brain's capacity to absorb the structure of experience associated with favorable outcomes, i.e., rules. This allows us to imagine possible goals and provides mental navigation to reach them. These mechanisms are dysfunctional or lost following frontal lobe stroke and in schizophrenia and Parkinson's disease. They are also co-opted by drugs of addiction. During the last funding period, we used multiple electrodes (up to 50) to compare and contrast brain areas in the frontal lobe and temporal lobe of monkeys, areas critical for goal-directed behavior and implicated in neuropsychiatric disorders and drug addiction, while monkeys performed disparate rule-guided behaviors. We now aim to use our lab's expertise in multiple electrode recording and complex behaviors to test major hypotheses of their role in goal-directed rule learning. Our results so far are in line with models suggesting that the basal ganglia is specialized for fast acquisition of simple associations and that it helps train slower acquisition of more elaborate rules in the executive prefrontal cortex: Thus, is critical to know how ubiquitous or rare these effects are. Finding the same pattern for different types of learning would strongly support these hypotheses, while determining its limitations would provide insight into functional specializations. We will record simultaneously from the prefrontal cortex and striatum during different types of stimulus-response (SR) learning, compare SR and stimulus-stimulus learning, and determine the influence of reversal learning, which is known to be prefrontal cortex-dependent. We will also test the hypothesis that the prefrontal cortex and striatum are specialized for elaborate versus simple rules, respectively. Our monkeys will switch between applying three different rules, the basic logical operations, AND, OR, and XOR, while we record from the prefrontal and striatum. These rules bridge a critical distinction (linear versus non-linear) yet share a logical relation (one can be built from the other two). This allows head-to-head comparison between rules that have different complexities but have the same formal requirements. Because rule-learning is a fundamental cognitive function, data from this project has the potential to impact on a wide range of human behaviors and disorders and by doing so open a path to drug and behavioral therapies that will alleviate them.

DESCRIPTION (provided by applicant): NIH Broad Challenge Area 15 - Translational Science NIMH Specific Challenge Topic 15-MH-109 - Prefrontal cortex regulation of higher brain function and complex behaviors. Abstraction is the ability to detect and store the commonalities across experiences while ignoring irrelevant details. It is a critical cognitive function because imbues the world with meaning and orders thought by endowing generalization and prediction. Without this crucial faculty, experiences would be fragmented and unrelated. Sensory inputs would seem strange and unfamiliar because they differ in appearance from previous experience. In other words, we would have the disordered thought of schizophrenics or the inability to generalize seen in autistics. While a variety of recent studies have examined the distribution and other properties of neural correlates across a range of brain areas, these studies have all examined categories that were already familiar to the animal. Virtually nothing is known about how and where the brain acquires new categories. Our laboratory will do so by using a test of category learning widely employed in humans and by using our unique approach of recording from many electrodes simultaneously in different areas of the monkey brain. By recording from three brain areas known to play a role in categorization (the prefrontal cortex, posterior parietal cortex, and basal ganglia), we will determine the how and where of how categories are learned, such as where they first arise, how they are formed from learning about individual exemplars. We will test the hypothesis that new, arbitrary categories are first acquired by the prefrontal cortex and the hypothesis that the prefrontal cortex forms categories by putting together exemplars learned by the striatum. This will help aid the U.S. economic recovery in several ways. The funds will be used to retain employment of scientific personnel and will also aid the economy via increased part-time labor for the manufacturing of microdrives and corresponding purchases of supplies from U.S. based companies. Our lab has been acutely impacted by the current economic crisis. We suddenly lost a source of funding when The Picower Foundation collapsed in the Bernie Madoff Ponzi scandal. The funds will also be used to retain the services of a postdoctoral fellow who will otherwise have to be laid off this summer due to this loss of funding. Finally, and perhaps most importantly, we have (and will continue to) freely share our microdrive design and methodology with other laboratories, several of which have adopted our system. Thus, the economic (and scientific) results of our supplement are likely to be "viral", inspiring others to adopt multiple-electrode technology and thus make similar purchases. PUBLIC HEALTH RELEVANCE: We will use multiple-electrode technology to determine the functional circuitry of category learning in the monkey brain. By recording from multiple brain areas simultaneously, we will determine where, when, and how categories are acquired by the brain.

There is a dynamic and delicate balance between top-down vs bottom-up control. We all know that sometimes we are more on task and other times we are less focused and more distractible. The balance is critical: A relative weakening of top-down attention is readily apparent in ADHD. More severe and global losses of top-down control can explain the weak central coherence of autism and the disordered thought of schizophrenia (e.g., Uhlhaas & Singer, 2012). Our previous Aims employed many-electrode recording to trace top-down and bottom-up functional circuits across wide stretches of cortex for the first time at the neural level. The results suggest that oscillatory coherence between cortical areas regulates top-down and bottom-up communication. Now, we aim to leverage this understanding to determine how oscillatory dynamics fine-tunes these circuits for different balances of top-down vs bottom-up control. Our previous tasks switched between different types of top-down and bottom-up signals. We will now employ a task that varies the relative degree of top-down vs bottom-up control. Corresponding changes in neural dynamics across will powerful support for a role of oscillatory coherence in regulating top-down/bottom-up balance. Further support will come from non-invasive stimulation that modifies these neural dynamics and changes the balance of cognition. RELEVANCE (See instructions): There is mounting evidence for a role in oscillatory coherence in regulating neural communication and for its dysfunction in neuropsychiatric disorders like ADHD, autism, and schizophrenia. This project will shed new light on the role of oscillatory dyanmics in balancing cortical processing and can directly point to theraputic interventions.

PROJECT SUMMARY (See instructions): This project addresses the role of functional circuitry of how memories are formed using a non-human primate model. Despite the great deal of evidence that both the prefrontal cortex (PFC) and medial temporal lobe (MTL) are critical for normal memory, there has been remarkably little experimental effort toward understanding how these two systems interact This is, in part, because neurophysiological studies have largely focused on either the PFC or MTL individually. As a result, much of what we know about them is based on comparisons between different animals with different training histories, often on different tasks, and/or different levels of experience. This confounds their comparison and precludes examination of the precise timing of their activity that gives insight into network properties and signal flow. Our goal is a more integrated understanding of the PFC and MTL. We will accomplish this by recording simultaneously from multiple electrodes in multiple subregions of the PFC and MTL while monkeys form and recall new context guided associative memories and while they make inferences based on those memories. By comparing the relative neural latencies for memories to be formed and recalled, we will determine where memories are formed, how they are recalled, and how memory-related signals flow between the PFC and MTL.

DESCRIPTION (provided by applicant): This project addresses the role of functional circuitry of the most fundamental and striking fact about cognition, its limited capacity (e.g., it is difficult to write an email while talking on the phone). Despite the remarkable power and flexibility of human cognition, the online workspace that most cognitive mechanisms depend upon is surprisingly limited, capable of representing only a few items simultaneously. Understanding the neurobiology of capacity limitations is critical because a reduction in capacity has been tied to a host of neuropsychiatric diseases such as schizophrenia and ADHD. In fact, training designed to increase capacity of working memory in children diagnosed with ADHD has been suggested to alleviate symptoms and may be able to improve fluid intelligence. This raises the possibility that therapeutic improvement to one bedrock aspect of cognition could lead to improvements in a wide range of what may prove to be symptomatic, rather than primary, ills, such as attention deficit disorder, poor executive function, etc. However, while capacity limitations are well-studied in humans (it may be the most well-studied cognitive phenomenon), it has never been investigated in the animal brain. Thus, fundamental questions about its neural basis have not yet been addressed. Our laboratory will do so by using a test of capacity limitations in humans and by using our unique approach of recording from many electrodes simultaneously in different areas of the monkey brain. This will allow us to determine the how, where, and why of capacity limitations, such as where it arises in cortical processing, how items are lost from memory after capacity is reached, and why neural coding leads to a capacity limitation. We will test the two major theories of capacity limitations (slot model vs information-load model) and target cortical areas most associated with working memory capacity limitations in humans: the prefrontal cortex, posterior parietal cortex, and mid-level visual cortex (i.e., area V4). By comparing the relative neural latencies for information loss between them, we can determine where capacity limitations arise in cortical processing and whether it is a bottom-up or top-down phenomenon. PUBLIC HEALTH RELEVANCE: We will use multiple-electrode technology to record from multiple areas of the macaque cortex during a visual, working memory capacity task. We will determine the neural substrates of where, when, and how capacity limitations arise in cortical processing by evaluating where, when, and how neural information is degraded when capacity is exceeded. Understanding the neural substrates of working memory capacity limitations will provide important insights towards both normal (e.g., IQ performance & fluid intelligence) and abnormal (e.g., ADHD & schizophrenia) cognitive functions that have not yet been explored at the neurophysiological level.

Top-down functions (e.g., attention and working memory) go hand-in-hand with consciousness. Yet we know more about the former, because it is hard to define consciousness in most contexts. There is, however, a way in. General anesthesia (GA) provides transition through different stages of consciousness. We aim to determine which aspects of cortical processing are critical for these changes by using a surgical anesthetic, propofol, in monkeys. Propofol increases cortical-wide delta oscillations and alpha in frontal cortex. One hypothesis posits that the increased frontal alpha disrupts top-down cortical feedback from frontal cortex. Another points to the widespread delta oscillations that, unlike sleep, are decoupled across cortex. This could fragment long-range communication across cortex or, as some suggest, loss of consciousness (LOC) could be due to frontal-parietal cortex disconnectivity per se. The thalamus is likely to play a major role. Mounting evidence suggests it controls top-down cortical processing, perhaps by modulating cortical oscillations. Abnormal thalamic oscillations could entrain different cortical areas or the changed dynamics could be largely a cortical phenomenon. Previous primate studies of GA lack the combination of wide scope and single-neuron precision to directly test these hypotheses. We will use chronic intracranial multiple electrodes in monkeys to record simultaneously from a wide range of critical brain structures (prefrontal cortex, posterior parietal cortex, auditory cortex, and thalamus). We will also test if thalamic stimulation can restore consciousness and wakeful cortical dynamics.