The reverse inference problem fMRI: L10

Question 1

define reverse inference

Answer 1

drawing conclusions about cognitive processes from the presence of activation

Question 2

what do psychologists actually want to understand?

Answer 2

• the mental architecture underlying functions

- gather evidence for engagement of mental/cognitive processes in a particular task

Question 3

what is “wrong” with the following logic:

when cognitive process X, brain region Z is active

Answer 3

• not exclusive

- > brain region Z may be active for many tasks, not just in X

Question 4

what have researchers concluded about more 1. anterior regions (towards front)
2. posterior regions

Answer 4

1. represent more abstract information

2. more specific content

Question 5

1. dorsal axis

2. ventral axis

Answer 5

1. abstractness of rules

2. abstractness of memory, representations/retrieval

Question 6

what did Duncan argue?

Answer 6

• frontal cortex shows relative but not absolute specialisation
• > prefrontal regions might be recruited “more strongly” if the task at hand becomes more difficult

Question 7

What did Duncan argue about other brain regions?

-> problem

Answer 7

• this was true for other regions as well (recruited if task becomes difficult)
• > if we find an activated region, part of this “multiple-demand” network, we don’t know what each part is contributing

Question 8

what is the multi-demand network?

Answer 8

• remembering word/non-word strings, arithmetic, spatial working memory, verbal working memory, and three versions of resisting response conflict

Question 9

What did Poldrack say about reverse inference?

Answer 9

• if the experimental setup fails to manipulate the cognitive process of interest, it cannot provide useful information about the process

Question 10

Problem highlighted by Poldrack

Answer 10

how good task A is for understanding cognitive process X

-> if the task measures more than one cognitive function we dont learn much

Question 11

how did Poldrack express these problems?

Answer 11

• in probabilistic terms

Question 12

The probability that we learn from our fMRI results that cognitive process X is involved depends on? (2)

Answer 12

1. the quality of the task to measure the cognitive process

2. the specificity of region fro this cognitive process

Question 13

1. P(COGTASK)
2. P(ACTCOG)
   = we can use this design to?

Answer 13

1. the probability of the task measure the cognitive process (unobservable) -> cognitive process
2. the probability of activation occurring in the brain for a particular cognitive process (observable) -> fMRI activation
   = make better experiments and inferences

Question 14

a problem highlighted by Kamitani & Tong?

Answer 14

-overinterpretation of null results

Question 15

what does it mean if you find no region was significantly stronger activated for task A vs task B?
-> what should we always avoid doing?

Answer 15

• we dont know
• statistical tests aren’t designed to make Ho difficult, so we can’t interpret null results
• out method might not be sensitive enough to detect small differences
• > concluding that brain regions are not involved in a cognitive process