Examples of novel objects. Thirty-two different objects were created for study. Across three 1.5-h training sessions, subjects performed timed trials of functional tasks using 16 of the objects, with sets counterbalanced across subjects.

Group results showing learning-related changes in hemodynamic response for novel object matching, relative to scrambled image matching in lateral cortical regions. Group activation maps are displayed on the structural magnetic resonance image from a single subject. Regions in red were more active, and regions in blue were less active, after training than before training. Regions in yellow, which overlap with regions in red, were more active for trained (T) objects than for not-trained (NT) objects during Session 2 but not Session 1. (A) Sagittal section (Talairach and Tournoux coordinate, x = −50) showing the location of learning-related activity in the left middle temporal gyrus/inferior temporal sulcus (red region, local maxima in Brodmann area [BA] 37, −50, −60, +5; yellow region, BA 37/21, −53, −51, −3). (B) Axial section (z = +47) showing the location of emergent learning-related activity in the left premotor/prefrontal (red region, BA 6, −36, +2, +40; yellow region, BA 6/8, −43, +10, +40) and intraparietal cortices (red region, BA 7/40, −32, −59, +36; yellow region, BA 7/40, −42, –43, +38). (C, D, E) Group-averaged activity expressed as the difference between novel object matching and scrambled image matching for all voxels in the middle temporal gyrus, left premotor, and intraparietal ROIs, respectively. Red bars represent brain regions that showed increased activity for object matching after but not prior to training; yellow bars represent regions that demonstrated greater activity for trained objects than not-trained objects after but not prior to training. Error bars represent standard error of the mean.

Group results showing learning-related changes in ventral temporal cortex. As is Figure 3, regions in red were more active and regions in blue were less active, after training than before training. (A) Coronal section (y = −62) showing changes in the fusiform gyrus. The medial portion of the left fusiform gyrus showed increased activity (red) (local maxima, Brodmann area [BA] 37, −27, −60, −9), whereas the more lateral region showed decreased activity (blue) (with 2 local maximas, BA 37, extending anteriorly from −38, −63, −11 to −38, −44, −21) after training. Increased activity in the left inferior temporal sulcus that extended into the middle temporal gyrus (see Fig. 3) can also be seen. (B, C) Group-averaged activity expressed as the difference between novel object matching and scrambled image matching for all voxels in the medial (red bars) and lateral (blue bars) fusiform ROIs. Error bars represent standard error of the mean.

Sample trials for the perceptual matching task. (A, B) During object-matching blocks, subjects indicated for each stimulus pair whether the 2 photographs were of the same object photographed from different views (A) or of two different objects (B). (C, D) During blocks of scrambled image matching, subjects indicated whether each pair depicted identical scrambled images (C) or 2 different scrambled images (D). (E) Increased efficiency in using the novel objects to perform functional tasks during the training sessions. The average time to complete each training trial for all objects in the training set is plotted along the y axis. (F) Reaction times during functional brain scanning for matching to-be-trained (or trained) (T) and not-to-be-trained (or not-trained) (NT) novel objects (red bars) and matching scrambled images of the T and NT novel objects (white bars) prior to and after the training sessions. Error bars represent standard error of the mean.

Hemodynamic response in a single subject prior to and after training. Shown is a single, 1-mm-thick magnetic resonance image of the left hemisphere (x = −47). Regions in red were significantly more active when performing the matching task with the novel objects relative to scrambled object matching. (A) Prior to training, the object-matching task was associated with activity in ventral occipital cortex, which continued anteriorly along the fusiform gyrus (not visible in this lateral slice). (B) After training, activity emerged in the left middle temporal gyrus and left premotor and intraparietal cortices (circled regions). (C) Direct comparison of the pre- and posttraining scans revealed significantly more activity in these regions after training (all P < 0.001).