Neural Correlates of Reasoning by Exclusion

Abstract

When asked to select a label for a novel object from a given group of labels that includes both novel and familiar labels, one tends to choose a novel label. Nonhuman animals robustly fail to demonstrate the same tendency, although this tendency called “exclusion” that can bias human behavior may seem quite natural. The functional magnetic resonance imaging study described here investigated the neural correlates of this bias. The subjects were trained on two sample-to-comparison associations. In the scanner, they were shown a novel sample and were asked to choose between a trained comparison and a novel comparison. The subjects readily chose the novel comparison and rejected the trained one, thus demonstrating exclusion. Significant activation was observed in the prefrontal cortex (PFC) and inferior parietal lobule (IPL) during exclusion. Medial frontal activation was also observed when the novel stimuli appeared. These results suggest that the medial frontal cortex is associated with novelty detection and that the PFC and IPL are involved in rejecting the defined comparison.

Appendices

Appendix 1: Image Acquisition and Analysis

The brain images were collected using a 4 T Varian Unity Inova MRI system. The BOLD signal was measured using a T2*-weighted echo planar imaging sequence (TR = 2,600 ms, TE = 25 ms, FA = 40°). Twenty-five axial slices (thickness = 5.0 mm, gap = 0 mm, FOV = 240 × 240 mm, matrix = 64 × 64) were acquired per volume. A set of high-resolution T1-weighted structural images was obtained by magnetization-prepared 3D FLASH (TR = 110 ms, TE = 6.2 ms, FA = 11°, matrix = 256 × 256 × 180, voxel size = 1 × 1 × 1 mm³).

The functional and structural images were analyzed with Brain Voyager QX. The functional images for each subject were preprocessed, including slice time correction, three-dimensional motion correction, spatial smoothing with a Gaussian filter (FWHM = 6 mm) and high-pass filtering (0.01 Hz). The structural image was transformed into the standard Talairach space. The functional images were transformed into the standard Talairach space by normalizing and resizing them to the transformed structural image. BOLD signals were modeled using a synthetic hemodynamic response function composed of two gamma functions. Random-effects analysis was performed on the functional data to reveal significant activation.

Appendix 2: S− Condition

In S−, a novel comparison replaced the relevant trained comparison, whereas in S+, a novel comparison replaced the trained irrelevant comparison. The S− condition was not used in the analysis because eight subjects did not respond as expected to the S− condition. Four of the subjects chose the defined comparison (C2 was chosen in the presence of S1), whereas the other four subjects responded randomly. To respond accurately, it was necessary to reject the defined comparison and choose the novel one, based on the negative relations of ‘if S1, then not C2’ and ‘if S2, then not C1’. There are two possible explanations for this failure: one is that the training was insufficient, and the other is that the subjects could not reject the defined comparison stimulus. The latter possibility is not plausible, because the subjects rejected the defined comparison stimulus in Exclusion. These results suggest that the subjects did not learn the negative relations sufficiently.