fMRI-research and modeling of spatial and temporal relational reasoning

- no english description available -

This interdisciplinary project intends to (i) determine the neural correlates by using fMRI in the human brain, (ii) model this by a formal and computational framework, and (iii) identify differences between determinate and indeterminate tasks and therefore gain additional insights into cognitive complexity. The starting point was a computational model for human spatial relational reasoning which has been developed in a terminated project (R2-BackSpace). This model generates mental models from spatial information and manipulates these by executing specific operations. By assigning unit costs to each model operation, empirical differences (and phenomena like the indeterminacy effect) could be explained. Previous work (Fangmeier et al., 2006; Fangmeier and Knauff, 2009) found areas that were engaged during the processing of determinate problems. For each of the described phases which Johnson-Laird and Byrne (1991) described distinct activated areas could be found. One important objective for the current project is to identify the neural correlates during deductive reasoning with indeterminate problems. In comparison to determinate problems there are some differences during the construction and variation/validation of models. The question is if indeterminate problems activate the same, other or additional networks in comparison to determinate problems. In comparison to the spatial domain only very few experiments have been done for temporal reasoning problems (cf. Schaeken et al., 1996a, 1996b; Schaeken and Johnson-Laird, 2000). An underlying assumption from the Mental Model Theory (MMT) is that the mental models constructed out of the given spatial or temporal descriptions use the same underlying cognitive processes. The MMT is considered in the community to best explain the behavioral data. This implicit assumption – that humans use almost the same neural networks and that there is no main distinction between determinate and indeterminate tasks (since an assumption of the preferred MMT is that humans tend to construct one initial model and start to operate and reason with this one) however, has never been tested. Obviously, only a negative answer (that different networks are involved) would indicate a theoretical distinction. The fMRI investigation must be backed up with an integration of our results in findings of the literature. However, a main theory is missing. There is neither a computational nor a cognitive model which makes clear predictions which productions or mental processes correspond with which activation. The latest version of ACT-R 6.0 allows modeling the BOLD- function and will be a sensible platform for a concise modeling of our results and the results of the literature. We aim for such a model.

contact person: Ragni M

Start of project: 01.09.2009
End of project: 31.12.2010

Albert-Ludwigs-University Freiburg
Ragni M, Strube G

Actual Research Report

• Fangmeier T