PS3141 Lecture 1: Prefrontal Cortex 2024-2025 PDF

Rules and decisions in the primate brain: The prefrontal cortex Narender Ramnani Department of Psychology Royal Holloway University of London [email protected] @n_ramnani Clinical and Cog...

Rules and decisions in the primate brain: The prefrontal cortex Narender Ramnani Department of Psychology Royal Holloway University of London [email protected] @n_ramnani Clinical and Cognitive Neuroscience, PS3141  How does sensation influence movement? Spinal cord circuitry... Earliest part of the nervous system to receive sensory input Closest part of the nervous system to musculature Rapid, automatic sensory-to-motor transformation, can be independent of the brain How does the motor system access sensory information? Spinal cord circuits are not capable of more sophisticated sensory-to- motor transformations (e.g. when actions are to be computed on the basis of the cognitive, not sensory, content of stimuli). Higher mammals (e.g. primates) have evolved the ability to use the cortex to convert sensation to action with much greater flexibility. The primate cerebral cortex plays a vital role in coordinating more complex behaviours (e.g. over-riding reflexive behaviours) and solving problems related to sophisticated sensory-motor transformations (e.g. decision-making).  Phineas Gage Railroad worker in the 19th century (Cavendish, Vermont). His job was to set off explosive charges in large rocks to break them up. One day, a charge was unexpectedly set off, resulting in a 42 inch long, 1.2 inch wide, metal rod (tamping iron) to be smash into his cheek below the cheek bone and up through his brain, exiting through the top of his skull. Gage survived and suffered little to no pain. He had normal pulse and normal vision, returned to work several days later. He changed significantly as a person… Original skull showing Reconstruction trajectory of iron rod with a typical brain His brain must have sustained very significant damage to the prefrontal cortex  Phineas Gage Gage’s physician noted that… “His contractors, who regarded him as the most efficient and capable foreman in their employ previous to his injury, considered the change in his mind so marked that they could not give him his place again. He is fitful, irreverent, indulging at times in the grossest profanity (which was not previously his custom), manifesting but little deference for his fellows, impatient of restraint or advice when it conflicts with his desires, at times pertinaciously obstinate, yet capricious and vacillating, devising many plans of future operation, which are no sooner arranged than they are abandoned in turn for others appearing more feasible. In this regard, his mind was radically changed, so decidedly that his friends and acquaintances said he was ‘no longer Gage.’ ”  Frontal Lobe Lesions Neuropsychological data now show that patients with frontal lobe lesions often have ‘acquired sociopathy’ and show range of related problems: Emotional processing Decision making Inability to evaluate the consequences of actions Inability to inhibit inappropriate behaviour Impulsivity and risky behaviour ‘Perseveration’ (inability to adapt behaviour to changing circumstances) What is special about the frontal lobes (prefrontal cortex in particular)?  Executive Control “One of the enduring mysteries of brain function concerns the process of cognitive [executive] control. How does complex and seemingly willful behaviour emerge from interactions between millions of neurons?” Earl Miller (2000) What is ‘executive function’? These give us freedom from having to respond to the immediate and local stimulus environment and select actions on the basis of internal plans and long -term goals. There are mainly two types of account that describe executive (or ‘cognitive’) control: Temporal theories – Active maintenance and manipulation of information within working memory – Processing of cross-temporal contingencies (i.e. filling the gap between perception and action) ‘Top-down’ Theories – Attentional supervision (see Shallice and Burgess, 1996) prefrontal cortex is a top-down attentional ‘supervisor’, biasing competition among internal routines for action on the basis of current goals. – Attention to Action (see Passingham,1996): A key function of the PFC; anterior PFC regions exert control over posterior regions, including the premotor cortex.  Prefrontal Cortex: Anatomy and Function Beyond Neuropsychology: We can now ask questions about neurons in the prefrontal cortex more meaningfully… Where do they receive information from? How do they process the information? Where does the processed information go? Cytoarchitecture: How much functional diversity is there? Connectivity: Ability to process information at its most abstract level Activity: What are the characteristics of neuronal activity in PFC? The Frontal Lobes: Gross Anatomy Superior frontal gyrus (SFG) Superior frontal sulcus Middle frontal gyrus (MFG) Inferior frontal sulcus Inferior frontal gyrus (IFG) SFG Motor strip MFG IFG  Cytoarchitecture Within The Prefrontal Cortex Orbital views Medial views As with any other area of cerebral cortex, the prefrontal cortex also contains zones that differ in terms of organisation at a cellular level. Human Understanding this organisation in non-human primates is essential for understanding how it works in humans: We need to understand how these neurons work in non-human primates before we can extend our understanding to the human brain. Monkey But investigators sometimes disagree about issues in cross- species ‘homology’ of prefrontal areas Connectivity… From last year, you will remember that the brain is organised in hierarchical networks. The prefrontal Refresher from PS2061 cortex sits at the apex of these hierarchies Visual Hierarchy Auditory Hierarchy Motor Hierarchy Supramodal Ventral prefrontal Superior Superior Dorsal prefrontal Cortex cortex temporal temporal cortex sulcus gyrus Infero- Parietal temporal cortex cortex Parabelt Premotor cortex: Unimodal V4 PMd PMv Cortex PO Belt SMA CMAd/v/r V3 MT / V5 Primary V2 Cortex V1 Core Primary motor cortex Thalamic area LGN MGN Motor thalamus Sensory/motor Retina Cochlea Spinal cord neurons  The Prefrontal Cortex Sensation / perception leads to action: Actions can be internally generated (independently of stimuli), but must often be guided by stimuli. Perception Action  The Prefrontal Cortex Sensation / perception leads to action: Actions can be internally generated (independently of stimuli), but must often be guided by stimuli. How does sensory information reach the motor Perception system? The sensory and motor systems operates at several levels of abstraction. Each level of the motor system accepts Action information from a comparable level of the sensory hierarchies  The Prefrontal Cortex Sensation / perception leads to action: Actions can be internally generated (independently of stimuli), but must often be guided by stimuli. How does sensory information reach the motor Perception system? The sensory and motor systems operates at several levels of abstraction. Each level of the motor system accepts Action information from a comparable level of the sensory hierarchies Actions lead to sensory consequences The brain evaluates the sensory The perception-action consequences of actions These are used to determine whether the cycle action was executed properly Errors signalled by sensory feedback are minimised by adjusting actions  The Prefrontal Cortex Less abstract More abstract Sensory Hierarchy in the brain Sensory input  The Prefrontal Cortex Less abstract More abstract Sensory Hierarchy in the brain Sensory input Motor output Motor Hierarchy in the brain  The Prefrontal Cortex Less abstract More abstract Sensory Hierarchy in the brain Sensory input ? Motor output Motor Hierarchy in the brain At which hierarchical levels does sensory information influence actions? Defining the problem space in terms of anatomy makes this problem tractable. Where in the brain do the sensory and motor systems converge?  The Prefrontal Cortex Less abstract More abstract Sensory Hierarchy in the brain Sensory input Motor output Motor Hierarchy in the brain Anatomical studies show that convergence occurs in many locations Transformations occur at many points in the hierarchy Fuster’s hierarchical model The anatomy of the perception-action cycle… Perception in Action in posterior cortex anterior cortex  Fuster’s hierarchical model The perception-action cycle… Posterior association Prefrontal Stratified anatomy cortex cortex Ranges across the neuraxis Unimodal (spinal cord to neocortex) Premotor sensory cortex cortex Sensory structures connect with motor structures Primary Primary sensory motor Sensory areas project to cortex cortex motor areas of a comparable hierarchical rank Sensory Sp. Motor Sp. cord cord Sensory input Motor output  Fuster’s hierarchical model The perception-action cycle… Posterior Stratified anatomy association Prefrontal cortex cortex Ranges across the neuraxis (spinal cord to neocortex) Unimodal Premotor sensory Sensory structures connect with cortex cortex motor structures Sensory areas project to motor Primary Primary areas of a comparable hierarchical sensory motor rank cortex cortex Feedback Sensory Sp. Motor Sp. “Efference copy” cord cord “Corollary discharge” Environmental change Sensory input Motor output caused by action  Cascading Cognitive Control Koechlin et al. (2003) “The Architecture of Cognitive Control in the Human Prefrontal Cortex”, Science, 302: 1181-1185 Koechlin and Summerfield (2007) An information theoretical approach to prefrontal executive function. Trends Cogn Sci. 11(6):229-35. Even the most complex executive functions can be decomposed into simple routines for selecting actions (or thoughts) Branching Control The demand of executive control can be measured as the amount of Episodic information required for action selection. Control “…the model proposes that action Contextual Control selection is guided by hierarchically ordered control signals, processed in a network of brain regions organized Sensory Control along the anterior–posterior axis of the lateral prefrontal cortex.” Koechlin and Summerfield (2007)  Cascading Cognitive Control Main features of model: Temporal proximity: The fractionation of Immediate cognitive control according to the temporal Low level framing of actions and events involved in Essential selection (immediate, anterior; distant, Distant posterior) High level Subsidiary Functional proximity: Higher levels of control are predicted to involve more anterior prefrontal regions Branching Subsidiarity: Higher levels of the hierarchy Control are only recruited when lower levels do not Different sets of provide enough information to control action rules (low-order to Episodic high-order) that selection. Control govern action selection If at all higher regions are recruited Contextual Control (see review for proactively, their signals can bias and details) overrule prepotent sensorimotor associations Sensory or task sets. Control  Anterior Prefrontal Cortex (aPFC) Also called area 10, BA 10 and fronto-polar cortex There are a number of models that attempt to explain the nature of information processing in area 10 (see Ramnani and Owen (2004) for review) Lateral prefrontal cortex Branching Control Episodic Control Contextual Control Sensory Control Anatomical features that make area 10 / aPFC special compared to other prefrontal areas: Connectivity: Appears to be the only area of PFC connected only with other ‘supramodal’ areas of cortex, with no input or output to sensory and motor areas of the cortex. Evolution: Has evolved more rapidly than any other, and is the largest in the human brain Synaptic integration: Contains a lower cell density, but a higher density of ‘dendritic spines’ than other prefrontal areas Ramnani and Owen (2004): “We propose that the aPFC is engaged when problems involve more than one discrete cognitive process: that is, when the application of one cognitive operation (such as a rule) on its own is not sufficient to solve the problem as a whole, and the integration of the results of two or more separate cognitive operations is required to fulfil the higher behavioural goal”  Prefrontal Cortex: The top of the hierarchy… The primate brain is remarkably flexible in the way that it converts sensation into action Dealing with novel situations Primates can take a completely novel problem and determine strategies for solving problems Bridging temporal gaps Primates can receive an instruction and execute appropriate actions at a later time Contingencies and rules Primates can learn rules about Antecedents to actions (e.g. instructions that predict rewards) Actions (e.g. complex behaviour that links antecedents to consequences) Consequences of actions (e.g. rewards)  The prefrontal cortex and working memory One of the most important concepts behind explaining the functions of the prefrontal cortex is the delayed response task This involves the ability to apply a rule and actively represent information that is not currently present Example: Delayed matching-to-sample task  The prefrontal cortex and working memory One of the most important concepts behind explaining the functions of the prefrontal cortex is the delayed response task This involves the ability to apply a rule and actively represent information that is not currently present Example: Delayed matching-to-sample task  The prefrontal cortex and working memory One of the most important concepts behind explaining the functions of the prefrontal cortex is the delayed response task This involves the ability to apply a rule and actively represent information that is not currently present Example: Delayed matching-to-sample task Sample  The prefrontal cortex and working memory One of the most important concepts behind explaining the functions of the prefrontal cortex is the delayed response task This involves the ability to apply a rule and actively represent information that is not currently present Example: Delayed matching-to-sample task Delay….  The prefrontal cortex and working memory One of the most important concepts behind explaining the functions of the prefrontal cortex is the delayed response task This involves the ability to apply a rule and actively represent information that is not currently present Example: Delayed matching-to-sample task Choice?  The prefrontal cortex and working memory One of the most important concepts behind explaining the functions of the prefrontal cortex is the delayed response task This involves the ability to apply a rule and actively represent information that is not currently present Example: Delayed matching-to-sample task Choice?  The prefrontal cortex and working memory One of the most important concepts behind explaining the functions of the prefrontal cortex is the delayed response task This involves the ability to apply a rule and actively represent information that is not currently present Example: Delayed matching-to-sample task Reward  The prefrontal cortex and working memory The critical features are that the monkey i) understands the contingencies ii) continues to ‘represent’ the sample ‘online’ for the duration of the delay in working memory Lesions to the prefrontal cortex (area 9/46) severely impair working memory tasks with delays, but leave similar control tasks without delays unaffected. The prefrontal cortex… “…closes the cycle at the summit, integrating in the time domain cognitive representations of perception and of action as required in goal-diracted behaviour.” Fuster (2001), “The prefronal cortex – an update: Time is of the essence”, Neuron 30:319-333  The prefrontal cortex and working memory Neuronal activity… cue trigger ‘Working memory’ delay (Constantinides, et. al. J. Neurosci. (2001), 21:3646-3655) Delayed response tasks activate ‘memory fields’ in the prefrontal cortex.  Delay-period activity… Delay activity has four important properties… 1. Absent after a similar stimulus with no instructional value 2. Absent in the mere expectation of reward 3. Correlated with performance 4. Can be diminished or ‘switched off’ by distractors presented during the delay  Prefrontal Cortex: Working Memory or action selection? Rowe et al. (2000): Spatial working memory to test controversial hypothesis using delayed response Selection (BA 46) 1 %0.5 19 BOLD % BOLD0 0 15 -1 0 9 Delay 18 Length Time (s) 27 10 36 Activity time-locked to selection event at the end trial, not delay period  Prefrontal Cortex: Working Memory or action selection? Working memory (BA 8) 1 % 19 BOLD 0 15 -1 0 Delay 9 18 Length Time (s) 27 10 36 Activity changes in proportion to the length of delay, not time-locked (Rowe to selection et. al. (2000) Science...)  Distractor-resistant memory and the prefrontal cortex Sakai, Rowe and Passingham (2002), “ Active maintenance in prefrontal area 46 creates distractor-resistant memory”, Nature Neuroscience 5, 479 – 484. Memory task: Distractor task: Memory test: Remember spatial After a delay period, remember spatial Remember spatial locations of a series locations of blue dots, and say whether locations of a series of squares the position of an asterisk matched one of squares of the locations of the blue dots series of squares Some trials were given with and others without the distractor, so that errors could be induced (hence, error- related and no-error-related trials could be compared) Distractor-resistant memory and the prefrontal cortex Delay period activity in area Delay period 46 is present in correct trials but not in incorrect trials. Other areas with delay activity are indifferent to this effect [How does this study contradict Sakai et. al?] Correction to handout: Rowe, not Sakai!  Distractor-resistant memory and the prefrontal cortex Delay period Delay period activity for trials with distractor Resistance to distraction increases with area 46 delay activity: Trails in which memory was sustained across the delay (distractor- resistant), was associated with activity sustained across the delay. Further reading: What relationships do these areas have with each other?  But, “…in the constant chatter of the brain, a brief scream is heard better than a constant whisper…” Miller, Lundqvist and Bastos (2018): Lundqvist et al., (2018), Gamma and beta bursts during Working Memory 2.0. Neuron 100(2):463- working memory readout suggest roles in its volutional 475. control. Nature Communications, 9(1):394. Newer ideas by Miller and colleagues challenge these traditional views. Monkeys were trained to release the bar after the second test Evidence for persistent spiking comes from object only if both the identity and the order of the test averaging spiking over time and across objects matched the sample sequence. trials Gamma and beta showed different dynamics for different Averaging makes spiking appear persistent, types of match/non-match decisions (identity, order) and did even though it might be sparse under some so in a way that predicted different types of errors the animals circumstances (e.g., on single trials) could make. There are lots of theoretical reasons for sparse spiking to be the preferred model Evidence from single-trial analyses: Spiking is sparse in real time “we are not suggesting that the classic model of persistent spiking is wrong…It is just that the story is more complex than previously suspected” How else could it work? Working memories could be held between spiking by spiking- induced changes in synaptic weights.  The abstraction of rules… Why do we need to learn rules at all? “A brain limited to storing an independent record of each experience would require a prodigious amount of storage and burden us with unnecessary details….by extracting the essential elements from our experiences, we can generalize to future situations that share some elements which may, on the surface, appear very different.” “We have evolved the ability to identify commonalities among experiences and store them as abstract concepts, general principles and rules”. “This is an efficient way to deal with a complex world and allows the navigation of many different situations with a minimal amount of storage. It also allows us to deal with novelty.” Miller (2002)  Prefrontal Cortex: Encoding Rules Intelligent primate behaviour that permits flexibility and adaptability depends on the abstraction of general rules so that they can be applied in a variety of situations. In this way, these rules are not tied down to any specific stimulus or response Can neurons in the prefrontal cortex encode abstract rules? Does this activity reflect the generalisation of these rules to new situations?  Prefrontal Neurons Encode Rules Wallis, Andersson and Miller (2001) “Single neurons in prefrontal cortex encode abstract rules”, Nature, 411:953-956 Monkeys were trained to switch flexibly between two rules while the activity of several monkey prefrontal neurons was recorded Delayed matching to sample rule: Identify the object that matches with the cue by releasing a lever Release lever Cue …Delay… Test Response  Prefrontal Neurons Encode Rules Wallis, Andersson and Miller (2001) “Single neurons in prefrontal cortex encode abstract rules”, Nature, 411:953-956 Monkeys were trained to switch flexibly between two rules while the activity of several monkey prefrontal neurons was recorded Delayed non-matching to sample rule: Identify the object that does not match with the cue by releasing a lever Release lever Cue …Delay… Test Response Prefrontal Neurons Encode Rules Rule Abstraction: The monkeys were able to generalise these rules across different situations “The monkeys performed this task with new pictures, thus showing that they had learned two general principles that could be applied to stimuli that they had not yet experienced. The most prevalent neuronal activity observed in the PFC reflected the coding of these abstract rules.” Rule-selective neurons Total neurons recorded: 492 Rule-selective neurons (n=200, 41%) were the most common Prefrontal neuron selective for ‘match’ rule  The Prefrontal Cortex And Perceptual Categories Categorisation: The application of rules to different instances Categories are acquired by experiencing multiple exemplars of these categories Earl Miller and colleagues trained monkeys to distinguish between two categories of animal: Cats and Dogs. They parametrically blended images of 3 species of cat and 3 breeds of dog along different dimensions… Freedman, D.J., Riesenhuber, M., Poggio, T. and Miller, E.K. (2001) Science, 291:312-316 Freedman, D.J., Riesenhuber, M., Poggio, T. and Miller, E.K. (2002) J. Neurophysiology, 88:914-928. Freedman, D.J., Riesenhuber, M., Poggio, T. and Miller, E.K, (2003) J. Neuroscience, 23:5235-5246. They trained monkeys to categorize a set of computer-generated stimuli as “cats” and “dogs”. There were three prototype cats and three prototype dogs. A morphing program generated morphs that were blends of different amounts of cats and dogs. So, if you look left to right along each horizontal line, you can see that cats gradually morph into dogs. 100% Cat 80% Cat 60% Cat 60% Dog 80% Dog 100% Dog (prototypes) Morphs Morphs Morphs Morphs (prototypes) Cat Dog species 1 breed 1 Dog Cat breed 2 species 2 Dog Cat breed 3 species 3 Monkeys were trained using thousands of unique morphs based on all combinations of the six prototypes. After training, they could categorize novel morphs even when they were close to the category boundary and looked similar to the other category. Category boundary “Cats” “Dogs”  Delayed match to category task. RELEASE. (Category Match). Fixation 500 ms.. (Match) Sample 600 ms. Delay 1000 ms.. HOLD (Category Non-match) Test (Nonmatch) Monkeys saw two pictures, separated by a memory delay. If the pictures were from the same category, monkeys indicated “yes” by releasing a lever. If the pictures were from different categories, monkeys indicated “no” by continuing to hold the lever. Test object is a “match” if it the same category (cat or dog) as the sample  Recording sites How did prefrontal neurons behave in this task? in PFC Activity of a single prefrontal neuron ~1/3 of randomly selected neurons in the lateral prefrontal cortex showed activity that reflected the category of the stimuli (not their exact appearance). This neuron distinguished between cats and dogs, but did not distinguish between different levels of ‘catness’ and ‘dogness’. So, this neuron is communicating information about category membership and not exact physical appearance.  These captive-bred monkeys must have learned the categorisations because they had never seen cats or dogs before. What happens to these neurons when category boundaries change? Miller and colleagues re-trained one of their monkeys along new category boundaries PFC neural activity shifted to reflect the new boundaries and no longer reflected the old boundaries 100% Cat 80% Cat 60% Cat 60% Dog 80% Dog 100% Dog (prototypes) Morphs Morphs Morphs Morphs (prototypes) Cat Dog species 1 breed 1 Dog Cat breed 2 species 2 Dog Cat breed 3 species 3 Points for further consideration… Models of PFC function… How do other models explain PFC information processing? Coding in PFC neurons How flexible is coding in PFC neurons? Are these neurons so flexible that there is in fact no functional distinction between prefrontal areas? Categorisation Other brain areas interconnected with PFC have also been investigated. Which areas are they, and how do their neurons behave during categorisation? Number processing is a form of ‘relational’ categorisation. Can monkeys hold representations of numbers? How do PFC neurons represent numbers?