Reciprocal relations between cognitive neuroscience and formal cognitive models: opposites attract?

Abstract

Cognitive neuroscientists study how the brain implements particular cognitive processes such as perception, learning, and decision-making. Traditional approaches in which experiments are designed to target a specific cognitive process have been supplemented by two recent innovations. First, formal cognitive models can decompose observed behavioral data into multiple latent cognitive processes, allowing brain measurements to be associated with a particular cognitive process more precisely and more confidently. Second, cognitive neuroscience can provide additional data to inform the development of formal cognitive models, providing greater constraint than behavioral data alone. We argue that these fields are mutually dependent; not only can models guide neuroscientific endeavors, but understanding neural mechanisms can provide key insights into formal models of cognition.

Figure 1

The “model-in-the-middle” paradigm [1]unifies three different scientific disciplines. The horizontal dashed arrow symbolizes experimental psychology, a field that studies cognitive processes using behavioral data; the diagonal dotted arrows symbolize the field of cognitive neuroscience, a field that studies cognitive processes using brain measurements and constraints from behavioral data; the top two solid arrows symbolize the field of mathematical psychology, a field that studies cognitive processes using formal models of cognitive processes constrained by behavioral data (see also box 1). The bi-directional red arrow symbolizes the symbiotic relationship between formal modeling and cognitive neuroscience and is the focus of this review (see also Box 3).

Figure 2

Flowchart for the assessment of model validity.

Figure 3

The LBA model for response time and accuracy.

Figure 4

(A) Influence of rewards earned n trials in the past on the log odds of choosing one of two options on the current trial (a clockwise or counter-clockwise rotated grating), where each option was stochastically rewarded at an independent rate. (B) Similar to A, but depicts influence of prior choices on current choice: a prior choice decreases the probability of the same choice being made on the current trial because the task included a ‘baiting’ scheme to encourage switching between alternatives (see refs. –47). (C) The estimated influence of prior rewards and choices (panels A,B) can be combined to generate a trial-by-trial estimate of the probability that one of the options will be selected (in this case, the probability of the clockwise grating being selected is shown by the black line after the application of a causal Gaussian filter to smooth the data, see refs. –47). The solid green line depicts the expected choice probabilities on each block of trials based on the relative reward probability assigned to each choice alternative. For this subject, the estimated choice probability (black line) closely tracked the expected probabilities, supporting the notion that the estimated choice probability can serve as a stand-in for the subjective value of each alternative. Note, however, that there are local trial-by-trial fluctuations away from the expected choice probabilities, consistent with momentary changes in the subjective value of an option given the stochastic reward structure of the task. Data based on ref. .