Cognitive Control

cognitive control (executive function)

adjusting how our brain processes info to best achieve our current goal(s)

doesn’t create (transmit), it modifies (modulates) existing pathways/representations

to modify appropriately, control requires info about the current context and goals

automatic behavior

habitual, obligatory, rapid, efficient, and stimulus-driven

often engaged w/o consideration of future consequences

persists under high cognitive workload

goal-directed/controlled behavior

internally guided, slow, deliberative, consistent w/ goals and broader context

harder to do under load/fatigue (e.g., reading vs. comprehension; repeating vs. processing)

typically feels effortful

PFC and Goals

is the most widely interconnected part of cortex

is most recent evolutionary expansion in primates (especially white matter)

regions of brain that continue expanding latest in development (adolescence)

PFC is needed to exert attentional control

PFC is involved in suppressing associations between concepts

Early researchers on the role of the frontal lobes

Were uncertain and showed skepticism to the vital role of the frontal lobes → caused scientists to believe that cutting this part of the brain would not affect the brain much → frontal lobotomies began being used as treatment in the early to mid-20th century

dysexecutive syndrome

frontal lobe damage symptoms include environmental dependency, preservation, poor planning abilities, disconnect between knowledge and action, loss of goal-directedness, and socially inappropriate behavior

most metrics of cognitive ability intact: vocab., lang. function, semantic and episodic memory, perceptual processing, motor function, and IQ score in typical range

symptom: environmental dependency → behavior becomes more immitative and stimulus-driven; imitation, utilization

symptom: show preservation → trouble stopping oneself from repeating an action (see Wisconsin sorting task)

symptom: poor planning abilities → patients also have trouble/take longer completing tasks that require planning multiple steps ahead (see multiple errands task)

symptom: disconnect between knowledge and action → can recall a rule, but have trouble executing the appropriate action or inhibiting the wrong action

symptom: loss of goal-directedness

symptom: socially inappropriate → lack of appreciation and disregard of social rules resulting in impulsivity, social disinhibition, and short periods of anger or rage; blunted affect, prone to outbursts, difficulty interacting w/ others, but can state appropriate rules for behavior; show remorse → understand behavior is wrong but have trouble inhibiting it

Model to Explain Dysexecutive Syndrome

patients are being driven by automatic behavior; PFC’s role to evaluate action in relation to context (i.e., what is appropriate in the given environment), so when it’s damaged, you aren’t sensitive to the role of context and behavior does not get rerouted or adjusted → results in automatic behavior

Wisconsin Card Sorting Task

used to study perseveration; tasks participants w/ sorting cards by a rule (color, shape, or number) and feedback from experimenter indicates whether the rule is correct

rule changes ~10 trials after correct response is given

found that PFC-damaged patients can learn a new rule, but perseverate or have difficulty shifting to that rule

Multiple Errands task

patients given an allowance and set of errands to do around London

found that PFC-damaged patients weren’t able to complete any of the assigned tasks

Mechanisms of cognitive control

1. attentional control

2. response inhibition

3. thought suppression

multiple-demand network

a neural correlate of a variety of control-demanding tasks; includes regions of lateral PFC, posterior parietal cortex, and dorsal medial PFC

attentional control can be used to either enhance relevant info or suppress irrelevant info within posterior cortex (e.g., FFA, PPA)

damage to PFC impairs filtering of sensory signals (e.g., greater auditory response in ignored ear)

TMS to PFC leads to less attention-related enhancement in neural activity (color-related P100 for attuned vs. ignore color) and performance

sources of attention

direct attention to relevant info (e.g., frontal, parietal)

sites of attention

targets for modulating relevant info processing (e.g., VI, V4, MT)

flanker task

measures attentional control

shows involvement of PFC to enhance/suppress relevant/irrelevant info; PFC is needed to exert attentional control

response inhibition

the ability to stop, interrupt, or abort inappropriate responses

damage to inferior frontal gyrus impairs stopping

Go/No-Go task

a method of testing response inhibition where participants are asked to push a button when certain visual stimuli appear (Go trials) and withhold response to other stimuli (No-Go trials)

Stop-signal task

measures response inhibition; respond as quickly as possible to a stimulus, unless there’s another signal (e.g., auditory tone) soon after, in which case stop yourself

found that unsuccessful stoping involves interactions between inferior frontal gyrus (IFG), pre-SMA, subthalamic nucleus, and motor cortex

thought suppression

suppressing word associations during encoding → neural correlate of cognitive control → leads to (PFC-driven) decrease in hippocampal activity and poorer recall

Think/No-think paradigm

measures thought suppression; words in teh think condition are remembered than the no-think condition

suggests people can actively suppress retrieval of memory trace when told to do so

shows involvement of PFC to suppress associations between concepts

Other forms of cognitive control

1. Planning

2. Task-switching

3. Regulating one’s emotions/impulses

Working memory (WM) and Cognitive control

the mind’s “blackboard” → is needed to hold goals in mind, like relevant pieces of info together to make inferences, direct attention and action towards the goal, and connect across time (e.g., strategize, plan)

memory component = representation of info that is not present (generally recently or retrieved from long-term knowledge)

working component = activation/maintenance of info relevant to a goal (e.g., shopping list) and updating/adjusting that info (e.g., add item)

Mechanisms include maintenance and updating → dilemma between maintaining (gating out) vs. updating (gating in) info in WM: stability vs. flexibility → this gating problem may be solved in part by similar gating mechanisms as motor responses (i.e., basal ganglia)

is argued to be a core requirement for cognitive control

WM → maintenance

our ability to hold memory contents

PFC activity persists during WM delay after external stimulus is removed

activity increases w/ load, stops when the item is no longer needed

persistent activity predicts WM performance in humans

PFC-lesioned monkeys are impaired at recalling location of food that’s out of sight, unless there’s a learned cue and/or the room is darkened during delay (i.e., less distraction)

WM representation in PFC can be complex and abstract (e.g., context, rules, strategies) independent of specific sensory and motor elements

PFC’s support for WM may be why PFC is central to executive function → it’s argued that WM is a core requirement to cognitive control

WM → updating

manipulation of WM contents

dilemma between maintaining (gating out) vs. updating (gating in) info in WM: stability vs. flexibility

there’s an adaptive feature to either updating or not in WM

this gating problem may be solved in part by similar gating mechanisms as motor responses (i.e., basal ganglia)

N-back task

measures working memory maintenance and manipulation; participants are shown a series of stimuli (e.g., letters) and based on the n-back condition they’re in, they’re asked if the current stimuli matches the n-back one (e.g, “Does this letter match the letter 2 letters back?)

shows involvement of ventral lPFC to actively maintain info and involvement of dorsal lPFC to manipulate info

When might controlled processing fail us?

“overthinking” or under-relying on well-learned behavior

coming up w/ atypical solutions to problems, which benefits more from lower-level control

Control is directed through

monitoring and adjustment

Stroop task

there are two separate automatic pathways for color vs. word itself which interact in this task → they don’t have equal number of votes → the pathway for reading words is more automatic than reading color

stimulus features that support conflicting responses (incongruent) can result in poor performance (slower, less accurate responses) than response-congruent stimuli

control (lateral PFC) can support processing of the task-relevant feature → boosts the less automatic pathway (color) when needed for the task

helps us understand why we need cognitive control in the first place

the effect of incongruency also depends on how automatic the current task is (asymmetric interference)

identifying colors is more “control demanding” than word reading

asymmetric interference

when incongruent stimuli have different levels of control demand, causing slower, less accurate performance in task

performance depends on how automatic the current task is

homunculus problem

who is controlling the controller?

homunculus = the hand controlling control

control is directed through monitoring and adjustment

possible solutions → monitoring for:
feedback
errors
conflict
surprise

anterior cingulate cortex (ACC)

region most correlated w/ monitoring; refers to many areas, but in the context of monitoring and cognitive control, is typically dorsal regions of ACC (dACC) or anterior mid-cingulate cortex (aMCC)

dACC is widely connected to sensory, motor, affective, and control-related regions (e.g., lPFC)

one of the most debated regions of the brain, but w/ common view that it links monitoring/evaluation and action/control

feedback-related negativity (FRN)

ERG signal that monitors feedback; occurs ~270ms after feedback and reflects the degree to which the outcome is better or worse than expected (e.g,. in a gambling task)

source is localized around medial PFC/ACC

can be used as learning signal to update predictions and adjust performance accordingly…but feedback often isn’t available when we need it!

error-related negativity (ERN)

ERG signal that monitors error; is the neural signal that occurs after an error is made

occurs ~100ms after the response (error) is made (prior to/without feedback)

source is localized around rostral ACC (also error-related fMRI signals)

amplitude increases the larger the error and under conditions in which response accuracy is emphasized (vs. speed)

ERN followed by error-related positivity (Pe), more closely associated w/ error awareness and signaling levels of confidence in response

stimulating/inhibiting ACC w/ tDCS increases/decreases ERN and post-error increases in caution

conclusion: if you increase activity in ACC, you increase focus and are less likely to make another error

error-related adaptation

errors often lead to increased caution on subsequent trials (slower and/or more accurate responses) and can also lead a person to focus their attention more, resulting in better performance (e.g., fast, more accurate, less interference)

results from interaction between ACC and motor regions (e.g,. MI) and feature-specific regions (e.g., V4)

an orchestration of changes in the brain in order to slow a person down and/or focus their attention for the task they made an error on to improve performance

limitation: adapting to errors requires us to have (a) made a response and (b) know that it was wrong

monitoring for conflict

many situations that require us to engage control involve conflicting response → so what if we just monitor conflict levels to determine when to increase control!

has the benefit of using info already available before making a response (early error detection) and w/o needing feedback

consistent w/ neural activity measured within/around dACC w/ fMRI, EEG (N2), human neuronal recordings, and more

adapting to conflict

ppl often perform better on an incongruent trial if the previous trial was also incongruent (il) vs. if it was congruent (cl)

these conflict adaptations have been tied to interactions between ACC and lPFC across trials, resulting in increased control and decreased conflict for il trials

some evidence of reduced conflict adaptation w/ surgical lesions to ACC (cingulotomies)…but other mixed findings from lesion data and a LOT of debate over whether/when/why ACC signals conflict → takeaway: ACC is involved, but there may be other mechanisms involved as well

monitoring for suprise

surprise is another general-purpose signal where control may be needed

surprise is a common denominator across situations that evoke ACC activity, and result in some form of pause in action or cognition

proactive vs. reactive monitoring/adaptation

rather than relying on signals during a task (e.g., feedback, errors, conflict, surprise) to reactively adjust control, it often makes more sense to allocate control proactively

proactive control can be tailored to expected demands and incentives

e.g., expected demands → proportion incongruence → shows better conflict adaptation when expecting more conflict

e.g., expected rewards → better performance seen w/ higher incentives