Neural Networks and Cognitive Control (3) Flashcards
What is schizophrenia?
- a chronic and severe mental disorder that affects how a person thinks, feels, and behaves
- people with schizophrenia may seem like they have lost touch with reality
- affects cognition, feelings and behaviour
What are some schizophrenia facts?
- about 1% of Canadians
- more frequent in males
- late adolescence and early adulthood are peak onset periods (25 m and 27 f)
What are positive symptoms?
- psychotic behaviours not generally seen in healthy people
- hallucinations, delusions, though disorders, movement disorders
What are negative symptoms?
- disruptions to normal emotions and behaviours
- “flat affect”, reduced feelings of pleasure, difficulty beginning and sustaining activities, reduced speaking
What are the cognitive deficits of schizophrenia?
- attention: impairments in endogenous selective attention
- working memory: especially with high load, distraction or manipulation
- episodic memory: impaired use of contextual cues to organize information at encoding or retrieval
- executive function: impairments in set switching, planning, and dual-task coordination
What is Braver, Barch and Cohen (1999)’s hypothesis?
- perhaps many of the cognitive deficits in schizophrenia can be accounted for by a single underlying impairment of cognitive control: impaired dopamine fating of working memory
What are the participants?
- 16 matched controls
- 16 participants with schizophrenia: neuroleptic-naive (experiencing first hospitalization for psychotic symptoms and not treated with medication yet)
What is the cognitive control task?
- AX-CPT
- see a sequence of letters (cues and probes)
- target is an X that followed an A
What is the model?
- cognitive control model with dopamine gating
- impairment: noisy dopamine gating unit
What is the model fitting?
- account for performance of controls with intact model
- account for performance of participants with schizophrenia with impaired model
What are the two conditions?
- short delay condition: cue, short delay, probe (or target if X)
- long delay condition: cue, long delay, probe (or target if x)
What is context sensitivity?
- comparing correctly pushing X and incorrectly pushing X
- AX hits vs BX false alarms
- ability to use context (A or B) to guide correct response to X
What is context cost?
- compare reaction time of AY and BY trials
- should be able to response fast on BY trials because B indicates that even with X, target response is not necessary
- cost of context (A or B) on speed of response to Y
What does the gating model look like?
- cue: needs to be stored in prefrontal cortex/working memory
- gating unit (DA): decides if new information should be let into working memory or not
- context (PFC): working memory maintenance that applies top-down control to probe and output
What is affected in an impaired gating model?
- the gating unit is noisy
- it can be modeled this way
What does the behavioural data for context sensitivity look like?
- patients have a lower sensitivity to context
- fewer AX hits and more BX false alarms
- interaction with delay: even lower sensitivity at long delay
What does the behavioural data for context cost look like?
- patients have a smaller context cost: meaning the difference in reaction time between AY and BY trials is smaller
- interaction with delay: even smaller context cost at long delay
- (more cost means you are keeping track of things well)
What does the model data show?
- use of intact model and noisy gating model
- the main effect and interaction are similar compared to the human data
What does normal function of the DA gating unit look like?
- low tonic (stable) activity: keeps gate closed and maintains WM contents
- high phasic (pulse) activity: opens gate, updates WM contents
What does the impaired function of the DA gating unit look like?
- higher tonic activity: gate may open randomly, WM contents may be lost
- lower phasic activity: gate may fail to open, WM contents may not be updated
How are the results explained by the impaired function?
- cost isn’t shown because patients fail to store for context (A or B)
- interaction effects result because information is more likely to leak out over a longer period of time after the probe
How does this help us understand schizophrenia?
- a single unified explanation for a collection of seemingly disparate symptoms, including impairments of attention, working memory, episodic memory, executive function
- a proposed neural impairment that accounts for cognitive deficits
Why is computational psychiatry important and useful?
- neural network models provide an explicit, mechanistic account of cognitive processes in terms of their neural implementation
- can explore hypotheses about the neural basis of mental illness by creating focal impairments in models and investigating consequences
- leads to improved diagnosis, treatment and management