423 M-AM
Learning modulates indicators of relational complexity in the occipitoparietal lobe
Steven Phillips, Kazuhisa Niki
National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Japan

Relational complexity theory explains cognitive difficulty in terms of the relationships between task items (Halford, Wilson, and Phillips, 1998), and a number of studies have reported increased prefrontal and parietal activity with relational complexity (Christoff, et al., 1998; Kroger et al, 2002; Phillips and Niki, 2003). Learning and expertise also modulate the effect of complexity inherent in task design. For example, complexity may be reduced by learning to recode two items as a single item chunk. Alternatively, complexity may be increased by employing a ternary relation between three items where a binary relation would suffice. In previous work (Phillips and Niki, 2003), we identified a bilateral region in the occipitoparietal lobe sensitive to changes in relational complexity. Here, we report an extension to this work where we investigated the effect of learning on sensitivity to differences in relational complexity.

Eight Japanese subjects were scanned before, during and after a 2-week training period during which they learned to recognized shape pairs. For each scanning trial, subjects determined whether a probe pair appeared in a previous three-pair study list. Lists were either AB CD EF (low relational complexity), or AB AD CB (high relational complexity) of novel or learned planar shapes. Study pairs were presented serially for 2170 ms separated by a 2000 ms blank screen (encoding phase). Following the encoding phase, there was an 8000 ms blank screen delay period (retention phase) after which the probe pair was presented for 3000 ms, followed by a “+” symbol (5000 ms) to indicate the end of the trial (Figure 1). Scanning was performed on a 3.0-T MRI Scanner (GE 3T Signa) with EPI capability. 18 axial slices (5.3 mm thick, interleaved) were set to cover the entire brain. A T2* weighted gradient echo EPI was employed. The imaging parameters were TR=2 sec, TE=30 ms, FA=70°, FOV=20x20 (64x64 mesh). Using SPM99, data were preprocessed and fitted by a general linear model, where 11 (i.e., 2 [low, high complexity] x 2 [novel, learned shapes] x 2 [encoding, retention phase] + 3 [target, distractor, error trials]) event conditions were defined (Phillips and Niki, in press, for details).

Example shapes and trial sequence
Figure 1. Example shapes and trial sequence.

Regions of interest (ROI) analysis, based on ROIs from our previous study, revealed a significant interaction between complexity and learning (P<.009) for a bilateral region with centers at Brodmann co-ordinates [-22 -70 37] and [22 -78 41], which border on the occipitoparietal sulcus (see Figure 2 for whole-brain contrast). Further analysis suggested that subjects treated low complexity lists as high complexity lists before training, possibly due to difficulty maintaining distinct memory traces for novel shapes. These results suggest that the occipitoparietal region provides a window into effective relational complexity. That is, complexity as imposed on cognitive brain processes, not as prespecified by task design.

Complexity by learning contrast
Figure 2. Complexity by learning contrast (p<.01, uncorrected).

References

Christoff, et al. (1998). NeuroImage, 14(5), 1136-1149.
Kroger, et al. (2002). Cerebral Cortex, 12, 477-485.
Halford, Wilson, and Phillips (1998). Behavioral and Brain Sciences, 21(6), 803-831.
Phillips and Niki (2003). NeuroImage, 20(2), 1226-1235.
Phillips and Niki (in press). NeuroImage.