Task 8: Trying to make it work Flashcards
A1: Describe Cisek’s distinction between action selection & action specification
Both
- > require external sensory information
- > require internal info about current behavioural needs
Old hypothesis:
- serial process: Sensation -> representation of objects -> judgments -> decide action -> plan -> execute
New evidence:
- both occur simultaneously
- more integrative activation
e. g. activity in PPC for sensory, motor & cognitive info
A1: Describe Cisek’s Affordance Competition model
- Movement Specification
3 neural populations that encode for movement specification
1) Visual cortex -> encoding potential visual targets
- > modulated by attentional selection
2) Parietal cortex -> encoding potential actions
3) activity in PreMC
- -> fronto-parietal cortex for action competition
A1: Describe Cisek’s Affordance Competition model
- fronto-parietal cortex & input from Action Selection area
- Fronto-parietal cortex is biased by action selection input from BG & PFC
- > biasing influences many nuclei across cortex
BG:
- activity to movement parameters & decision variables
PFC/lPFC:
- advanced action selection
- PFC decisions evolve through collection of “votes” for selecting one action
ventral stream (IT): - detecting stimulus combinations relevant for action selection
A1: Describe Cisek’s Affordance Competition model
- Action execution
- overt feedback through environment
- internal predictive feedback through CB
A1: Explain why action selection & specification occur continuously & simultaneously.
-> behaviour is a constant competition between internal representations of potential actions
A1: Why is the parietal cortex so important for competition & how does it bias competition?
Parietal cortex:
- transforming visual input into representations of potential actions
- > only most promising actions are represented in parietal cortex
- > biases competition via attentional modulation & propagation of decision-signals
Fronto-parietal system:
- place of competition
- reciprocally interconnected
- > mixing of sensory, cognitive & motor variables (kinda like votes?)
- activity to one action increases while others are inhibited
- > based on decision factor (e.g. utility, affective value, …)
A1: Explain the phrase: “selective attention is seen as an early mechanism for action selection”
- Action selection includes areas related to visual processing which are places affected by selective attention
- > it reduces the amount of information represented along the dorsal stream -> increasing clarity on action to be done
A2: Compare Cisek’s view on functional organization of the parietal “affordance” path with Binkofsky & Buxbaum’s view in task 1
- Similarities
Similarities
- online modulation of actions
- Functions of dorsal & ventral stream
- > ventral: object identification (here part of action selection)
- > dorsal: spatial info, visually-guided movement (MIP, AIP)
- > increasing influence of attentional modulation
A2: Compare Cisek’s view on functional organization of the parietal “affordance” path with Binkofsky & Bixbaum’s view in task 1
- Differences
Cisek
- High crosstalk -> differentiated anatomical dorsal pathways but one integrated system
Binkofski
- Little crosstalk between dorsal pathways
A3: Where in the brain do decisions take place according to Cisek?
-> fronto-parietal system
Cortico-cortical connections are bidirectional:
- if a decision begins to emerge in one region -> will propagate outward into other regions
- > this interconnectivity causes decision to emerge
A3: Give an example for fronto-parietal decision-making
e. g. banana vs. pineapple
- sensory input: I see banana & pineapple -> ventral stream
- representing possible actions (=affordance): I can grab banana or apple
- > dorsal streams
- competition between potential actions: I think I want to get banana
- > fronto-parietal system, mutual inhibition
- (sub)cortical biases/decision factors: I want sugar, banana has sugar, i should get banana
- > BG
- task rules + object selection: I’ll get banana
- > PFC & FEF
- attention mechanisms to prioritize processing of one object: I’m looking for a banana
- > lateral intraparietal area
- action selection + movement specification: I’ll grab this banana
- > SMA + dPReMC/M1
A4: Describe the filter-view on the dorsal attention system, its selectivity & priority maps
Attentional modulation
- Central principle of filtering
- > there is competition for representation
- > enhanced representation for some stimuli at expense of others -> filter
- > representations of objects become more sparse along the stream/away from the striate cortex -> info is filtered
–> only most important targets “make it” to the parietal cortex
Priority maps
- created by BG/PFC biasing signals from action selection
A4: Explain how priority maps serve the larger good of goal-direct behaviour (vs. simply filtering sensory info)
Classic view: Attention as filter
Cisek: Attentional biasing from BG/PFC
- propagated via reciprocal cortical connections
Krauszlis: Attention is by-product/effect of BG’s state estimation
- related to motivation/cognitive loop –> motor goal
- Sensory input -> prior knowledge -> internal state
- > state estimation (which one matches input best?) -> decision policy
- > weighting sensory inputs based on policy
- competition between actions (vs. competition between stimuli)
A4: Present evidence for Krauzlis’ state estimation framework
Lesion in SC
- > according to old theory should disrupt neural activity associated with attentional biasing
- > no more attention related activity -> but still behaviour!
- > BG actually biases behaviour (after it’s whole loopy thing)
- SC & BG form different parts of spatial attention and/or SC influences attention after BG
Clinical:
- Spatial neglect
- damage to PPC/subcortical structures/striatum -> e.g. hallucinations (erroneous state estimations)
- might better explain ADHD
Evolutionary:
- animals without cortex also have attentional mechanisms
A4: What are advantages & constraints of Krauzlis’ new model of attention?
Advantages:
- Better explains learning
- > weights are learned not innate (S-R via DA)
- > Mechanisms for switching how salient events are handled
- > Instead of performing habitual responses to new events -> attention circuits open window of opportunity (new states/actions)
Constraints:
- expanding behavioural repertoire requires adding new states (via PFC-striatum projections)
- Decision policy biases behaviour but does not dictate outcomes
- not all instances of sensory-motor selection require this flexible control but still depend on BG circuits (e.g. habits)
Example of decision-making via Krauzlis framework
chocolate vs. berries
Visual input:
- chocolate & berries
Prior knowledge & internal status:
- chocolate made me vomit once
- I want something sweet
- > coupled with visual input
- > input to BG via SC
Motivation (S-R) + cognition (A-O) + motor loops (S-A):
- Berries are sweet and I want something yummy
- If I get berries I can have something yummy
- If I see sweet berried, I’ll get them
- > BG loops
- > state estimations -> interact to resolve competition between possible actions
Action selection:
- I will buy berries!
Weighted state:
- focus on berries
- ignore the chocolate
- > weights are assigned to sensory/nonsensory inputs
- > biased behaviour/decision policies
A5: What is a state?
Current condition of subject & environment
- > involves interpreting features of external world + internal status of subject (e.g. prior knowledge, current needs, …)
- > goal-direct decision making involves properly identifying state & associated actions
A6: Compare the views of Cisek & Krauzlis
- Competition
Cisek:
- sensory data is collected
- simultaneously actions compete for processing
- sensory info biases action competition until one response “wins”
- > constant competition between affordances
- > fronto-parietal network
Krauzlis:
- state estimates are made (external + internal state)
- each has a different weight & best-matching will win & decide about action
- > BG
A6: Compare the views of Cisek & Krauzlis
- Priority maps
Cisek:
- Biasing influence from BG & PFC
- fronto-parietal networks are modulated by decision policy (e.g. salience, attention)
- > influence competition
Krauzlis:
- Sensory & nonsensory information input is weighted based on best-matching state estimation in BG (which in itself is affected by frontal, parietal regions & SC)
- > bias action competition
- changes in environment -> shift in attention -> transitions in state (BG)
A6: Compare the views of Cisek & Krauzlis
- subordinate role of attention
Cisek:
- Selective attention mechanism -> selective sensory processing -> already elimination of many actions -> only most promising reach parietal cortex
Krauzlis:
- Attentional bias is an effect, not agent
- it’s a consequence of value-based decision-making in BG
- strength of contribution to current state (weight) -> decides about how much attention is paid to processing particular sensory inputs & types of knowledge
A6: Compare the views of Cisek & Krauzlis
- difference to classical theories
Cisek:
- more integrated view
- parallel processing of sensory info/cognition/action
Krauzlis:
- mainly criticises the traditional role of attention as filter
- attention is effect/by-product
A6: Compare the views of Cisek & Krauzlis
- link to evolution
Cisek:
- reject idea of new structures on top of old ones
- instead: differentiation & specialization of already existing functions
- > shifting projections not adding new structures
Krauzlis:
- important contributions from old subcortical structures
- neocortex doesn’t play fundamental role, it rather outsources other areas (BG)
A6: Compare the views of Cisek & Krauzlis
- role of PFC
Cisek:
- no only concerned with new cognitive functions
- relevant for advanced decision-making
- can be interpreted in the same way as subcortical structures
- > has similar biasing effect as BG
Krauzlis:
- neocortex is not fundamental in attention but there are important connections between BG & PFC
A6: Compare the views of Cisek & Krauzlis
- core steps of goal-directed decision-making
Cisek:
- Action specification & action selection occur continuously & simultaneously
- many different actions are prepared at the same time to be able to switch plans
Krauzlis:
- Internal & external state estimation (BG)
- similar to action specification
- > action/state estimate competition & biasing signals in fronto-parietal network/BG
