Neural Mechanisms For Simple Decisions PDF

Neural Mechanisms For Simple Decisions 11 January 2024 11:29 Main Ideas Notes Notes Lateral intraparietal area (LIP) cells Types of decisions Perceptual choice - making a decision based on perception Value-based choice - making a decision based on reality - can be based on the value of options ○ Val...

Neural Mechanisms For Simple Decisions 11 January 2024 11:29 Main Ideas Notes Notes Lateral intraparietal area (LIP) cells Types of decisions Perceptual choice - making a decision based on perception Value-based choice - making a decision based on reality - can be based on the value of options ○ Value-based choice may need one to make a perceptual decision Neural basis of perceptual decision-making Investigates the neural basis of the translation between sensory evidence and action Psychophysics The science of investigating the relationship between physical stimuli and perceptual responses Psychophysics experiments often have objectively correct answers - useful test-bed for theories of decision-making Psychophysicists are particularly interested in the speed and accuracy of our decisions in the face of noisy evidence Colby and Goldberg (1999) Notes Human fMRI Humans integrate mixed face/house information Regions identfied that correlate with the difference between FFA and PPA signals Human neuroimaging - predictions ○ need to make an assumption about the relationship between the BOLD signal and the single-cell buildup activity ○ One prediction is that BOLD signals should increase as the choice gets harder, because integration goes on for longer ○ Several brain regions show an enhanced response to difficult vs. easy trials ○ But note this could be due to signalling uncertainty / confidence, rather than involved in the decision itself PFC neu Colour/ One axi Two view on t Sherring Hopfield Causal manipulation of MT/LIP in monkeys When the motion is presented in MT neurons' RF, inactivation impairs decision-making But LIP inactivation does not Integration of motion evidence as a test case Random dot motion task ○ LIP cells respond when attention is directed to a stimulus in their spatial “receptive field” (RF) ○ Different neurons have different RFs Evidence accumulation in area LIP (Gold and Shalten, 2007) Sources of motion evidence ○ Area MT/V5 = motion-sensitive region in extrastriate visual cortex ○ Cells in visual area MT respond to moving bars of light in a directionally selective fashion (Dubner & Zeki, 1971; Born & Bradley, 2005) ○ As motion coherence increases, the firing rate for the cell's preferred direction of motion increases (Newsome et al., 1989; Britten et al. 1992) ○ Neurometric curve = performance expected if making decisions based only on this neuron's firing rates ○ Psychometric curve = for any given level of coherence, likelihood of being correct about motion Rodent models of evidence accumulation Using an auditory version of the random dot test, task is to decide which side has more clicks Inactivation experiments ○ ○ LIP neuron firing rates accelerate towards one or two response options (increases for a decision towards receptive field, decrease for decision away from receptive) ○ Rate of increase is proportional to coherence in the random dot motion stimulus Complex decisions Random dot motion task with colour Motion coherence + colour coherence ○ ○ When the firing rate are aligned to the eye movement response - they reach a common level ○ Consistent with a role in translating perception into action ○ Most neurons show psychophysical performance as good as or better than the whole monkey Neural populations can encode many things at the same time (mixed selectivity) Trajectories of the neural population on a lower dimensional plane can implement categorical decisions Circuit model of integration of motion evidence Area MT provides (noisy) motion direction information “Combining” neurons aggregate this information over time Output neurons control motor actions (e.g. eye movements) ○ Activity in MT and LIP differentiate even when sensory activity is identical ○ MT activity shows a constant, coherence-related signal, but no integration Similar signals also seen in the prefrontal cortex and striatum ○ Kim and Shadlen (1999) ○ Ding and Gold (2010) Human perceptual decision-making EEG and CPP (Centro-parietal positivity) ○ Evidence accumulation signals in primate brain often action-linked (e.g. LIP for an eye movement) ○ CPP in contrast is still observed even when no response is required Summary Sensory signals such as motion direction are represented in dedicated visual areas In many scenarios evidence needs to be accumulated to arrive at a sensible interpretation of the world Downstream “accumulators” in parietal and frontal cortex, together with cortico-basal ganglia loops, support a process of evidence accumulation Together these circuits provide a bridge between perception and action Causal intervention isolates critical nodes of this circuit for different aspects of the decision process (many more to be studied!) A “Sherringtonian” view of this circuit (where neurons implement different parts of a cognitive computation) provides a good model of simple decisions More complex decision problems may be better understood as neural populations moving through representational spaces (a “Hopfieldian” view) PSYC0032 The Brain in Action Page 1 Notes Notes urons act as a population, potentially difficult to gain insight from single units alone motion information always present in the population space - depending on the context, the attractor states differ s of the state space is equivalent to the 'choice' axis discovered by single-unit studies of evidence accumulation the cognitive brain gtonian view dian view PSYC0032 The Brain in Action Page 2 Notes PSYC0032 The Brain in Action Page 3 Notes Notes PSYC0032 The Brain in Action Page 4 PSYC0032 The Brain in Action Page 5