Using fMRI to test models of complex cognition

John R. Anderson, Cameron S Carter, Jon Fincham, Yulin Qin, Susan Ravizza, Miriam Rosenberg-Lee

Research output: Contribution to journalArticle

43 Citations (Scopus)

Abstract

This article investigates the potential of fMRI to test assumptions about different components in models of complex cognitive tasks. If the components of a model can be associated with specific brain regions, one can make predictions for the temporal course of the BOLD response in these regions. An event-locked procedure is described for dealing with temporal variability and bringing model runs and individual data trials into alignment. Statistical methods for testing the model are described that deal with the scan-to-scan correlations in the errors of measurement of the BOLD signal. This approach is illustrated using a "sacrificial" ACT-R model that involves mapping 6 modules onto 6 brain regions in an experiment from Ravizza, Anderson, and Carter (in press) concerned with equation solving. The model's visual encoding predicted the BOLD response in the fusiform gyrus, its controlled retrieval predicted the BOLD response in the lateral inferior prefrontal cortex, and its subgoal setting predicted the BOLD response in the anterior cingulate cortex. On the other hand, its motor programming failed to predict anticipatory activation in the motor cortex, its representational changes failed to predicted the pattern of activity in the posterior parietal cortex, and its procedural component failed to predict an initial spike in caudate. The results illustrate the power of such data to direct the development of a theory of complex problem solving, both at the level of a specific task model as well as at the level of the cognitive architecture.

Original languageEnglish (US)
Pages (from-to)1323-1348
Number of pages26
JournalCognitive Science
Volume32
Issue number8
DOIs
StatePublished - 2008

Fingerprint

Cognition
Magnetic Resonance Imaging
Parietal Lobe
Gyrus Cinguli
Motor Cortex
Brain
Temporal Lobe
Prefrontal Cortex
Functional Magnetic Resonance Imaging
Statistical methods
Chemical activation
Testing
Power (Psychology)
Experiments

Keywords

  • ACT-R
  • Cognitive modeling
  • fMRI
  • Model evaluation
  • Problem solving

ASJC Scopus subject areas

  • Language and Linguistics
  • Artificial Intelligence
  • Cognitive Neuroscience
  • Experimental and Cognitive Psychology

Cite this

Anderson, J. R., Carter, C. S., Fincham, J., Qin, Y., Ravizza, S., & Rosenberg-Lee, M. (2008). Using fMRI to test models of complex cognition. Cognitive Science, 32(8), 1323-1348. https://doi.org/10.1080/03640210802451588

Using fMRI to test models of complex cognition. / Anderson, John R.; Carter, Cameron S; Fincham, Jon; Qin, Yulin; Ravizza, Susan; Rosenberg-Lee, Miriam.

In: Cognitive Science, Vol. 32, No. 8, 2008, p. 1323-1348.

Research output: Contribution to journalArticle

Anderson, JR, Carter, CS, Fincham, J, Qin, Y, Ravizza, S & Rosenberg-Lee, M 2008, 'Using fMRI to test models of complex cognition', Cognitive Science, vol. 32, no. 8, pp. 1323-1348. https://doi.org/10.1080/03640210802451588
Anderson JR, Carter CS, Fincham J, Qin Y, Ravizza S, Rosenberg-Lee M. Using fMRI to test models of complex cognition. Cognitive Science. 2008;32(8):1323-1348. https://doi.org/10.1080/03640210802451588
Anderson, John R. ; Carter, Cameron S ; Fincham, Jon ; Qin, Yulin ; Ravizza, Susan ; Rosenberg-Lee, Miriam. / Using fMRI to test models of complex cognition. In: Cognitive Science. 2008 ; Vol. 32, No. 8. pp. 1323-1348.
@article{9e7f343e8ea644d48e41c3ffb8f8b274,
title = "Using fMRI to test models of complex cognition",
abstract = "This article investigates the potential of fMRI to test assumptions about different components in models of complex cognitive tasks. If the components of a model can be associated with specific brain regions, one can make predictions for the temporal course of the BOLD response in these regions. An event-locked procedure is described for dealing with temporal variability and bringing model runs and individual data trials into alignment. Statistical methods for testing the model are described that deal with the scan-to-scan correlations in the errors of measurement of the BOLD signal. This approach is illustrated using a {"}sacrificial{"} ACT-R model that involves mapping 6 modules onto 6 brain regions in an experiment from Ravizza, Anderson, and Carter (in press) concerned with equation solving. The model's visual encoding predicted the BOLD response in the fusiform gyrus, its controlled retrieval predicted the BOLD response in the lateral inferior prefrontal cortex, and its subgoal setting predicted the BOLD response in the anterior cingulate cortex. On the other hand, its motor programming failed to predict anticipatory activation in the motor cortex, its representational changes failed to predicted the pattern of activity in the posterior parietal cortex, and its procedural component failed to predict an initial spike in caudate. The results illustrate the power of such data to direct the development of a theory of complex problem solving, both at the level of a specific task model as well as at the level of the cognitive architecture.",
keywords = "ACT-R, Cognitive modeling, fMRI, Model evaluation, Problem solving",
author = "Anderson, {John R.} and Carter, {Cameron S} and Jon Fincham and Yulin Qin and Susan Ravizza and Miriam Rosenberg-Lee",
year = "2008",
doi = "10.1080/03640210802451588",
language = "English (US)",
volume = "32",
pages = "1323--1348",
journal = "Cognitive Science",
issn = "0364-0213",
publisher = "Wiley-Blackwell",
number = "8",

}

TY - JOUR

T1 - Using fMRI to test models of complex cognition

AU - Anderson, John R.

AU - Carter, Cameron S

AU - Fincham, Jon

AU - Qin, Yulin

AU - Ravizza, Susan

AU - Rosenberg-Lee, Miriam

PY - 2008

Y1 - 2008

N2 - This article investigates the potential of fMRI to test assumptions about different components in models of complex cognitive tasks. If the components of a model can be associated with specific brain regions, one can make predictions for the temporal course of the BOLD response in these regions. An event-locked procedure is described for dealing with temporal variability and bringing model runs and individual data trials into alignment. Statistical methods for testing the model are described that deal with the scan-to-scan correlations in the errors of measurement of the BOLD signal. This approach is illustrated using a "sacrificial" ACT-R model that involves mapping 6 modules onto 6 brain regions in an experiment from Ravizza, Anderson, and Carter (in press) concerned with equation solving. The model's visual encoding predicted the BOLD response in the fusiform gyrus, its controlled retrieval predicted the BOLD response in the lateral inferior prefrontal cortex, and its subgoal setting predicted the BOLD response in the anterior cingulate cortex. On the other hand, its motor programming failed to predict anticipatory activation in the motor cortex, its representational changes failed to predicted the pattern of activity in the posterior parietal cortex, and its procedural component failed to predict an initial spike in caudate. The results illustrate the power of such data to direct the development of a theory of complex problem solving, both at the level of a specific task model as well as at the level of the cognitive architecture.

AB - This article investigates the potential of fMRI to test assumptions about different components in models of complex cognitive tasks. If the components of a model can be associated with specific brain regions, one can make predictions for the temporal course of the BOLD response in these regions. An event-locked procedure is described for dealing with temporal variability and bringing model runs and individual data trials into alignment. Statistical methods for testing the model are described that deal with the scan-to-scan correlations in the errors of measurement of the BOLD signal. This approach is illustrated using a "sacrificial" ACT-R model that involves mapping 6 modules onto 6 brain regions in an experiment from Ravizza, Anderson, and Carter (in press) concerned with equation solving. The model's visual encoding predicted the BOLD response in the fusiform gyrus, its controlled retrieval predicted the BOLD response in the lateral inferior prefrontal cortex, and its subgoal setting predicted the BOLD response in the anterior cingulate cortex. On the other hand, its motor programming failed to predict anticipatory activation in the motor cortex, its representational changes failed to predicted the pattern of activity in the posterior parietal cortex, and its procedural component failed to predict an initial spike in caudate. The results illustrate the power of such data to direct the development of a theory of complex problem solving, both at the level of a specific task model as well as at the level of the cognitive architecture.

KW - ACT-R

KW - Cognitive modeling

KW - fMRI

KW - Model evaluation

KW - Problem solving

UR - http://www.scopus.com/inward/record.url?scp=57549088109&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=57549088109&partnerID=8YFLogxK

U2 - 10.1080/03640210802451588

DO - 10.1080/03640210802451588

M3 - Article

C2 - 21585456

AN - SCOPUS:57549088109

VL - 32

SP - 1323

EP - 1348

JO - Cognitive Science

JF - Cognitive Science

SN - 0364-0213

IS - 8

ER -