Mental rotation of complex three-dimensional (3D) figures is a well-studied example of mental imagery but the existing functional imaging data are very limited. Here we report the use of fMRI to localize cortical areas activated during this task. A pilot study of this kind has already appeared elsewhere in abstract form (1).
Subjects: 10 normal right handed subjects participated, under guidance of the MGH sub-committee on human studies. Task: The figures were the same as those reported by Shepard & Meltzer in their 1971 study (2), and were provided by R. Shepard. Each trial consisted of 3 conditions: A) Rest, with a fixation point; B) Control, in which subjects indicated by a key press whether a pair of figures (the same used in the Task condition), were identical or different; C) Task, in which subjects indicated whether a pair of rotated figures were the same or mirror images. Rest periods alternated with stimulus conditions in blocks of 30 s duration. Each trial consisted of three blocks of control and three blocks of task conditions. Subject response latency was recorded for all studies. Imaging: Scanning was performed in a 1.5T GE Signa MRI modified for echo-planar imaging (EPI) by Advanced NMR. Prior to behavioral studies, T1, T2 & flow-weighted images were acquired for anatomical registration. During trials, EPI scans were acquired continuously, covering 7 contiguous slices aligned parallel to the calcarine fissure, every 3.2 s. Data Analysis: Each voxel's signal intensity time course was compared for the various conditions was compared using Student's t-test for unequal variance. Pixels with t
3.62 were included for analysis, and activation maps, indicating in color the log of the calculated p values were overlayed onto the high resolution scans. Pixels shown in the flow images to contain macroscopic vessels were excluded. Activated regions were then classified by anatomical landmarks including the major sulci & gyri, based on the atlases of Ono and Talairach. Brodmann's areas were determined from the latter.
Technically adequate studies (complete, and free from motion artifacts) were obtained in 7/10 subjects. Comparison of the task+control conditions to fixation showed strong, consistent, activations in the area of the calcarine fissure (areas 17,18, 19), inferior occipital lobe (areas 37 & 19), pre- & postcentral gyrus (areas 4,3,1 & 2), and precuneus (diffusely through area 7). In comparing the task with control conditions, primary visual cortex (esp. 17 & 18) showed little signal difference. The strongest activations in all subjects included the frontal eye fields (area 8), precuneus (area 7) and lateral occipital sulcus in the region of the 19/39 border. Signal from frontal eye fields was weaker in control than in fixation. Repeated (N=6/7), activations were seen in the neighborhood of areas 46 & 44 in the inferior frontal lobe, and in postcentral regions (areas 3-1-2) (N=4/7).
The task of comparing two-dimensional drawings of rotated 3D figures is particularly interesting in the context of mental imagery, as the psychometric data (2) suggest that subjects might "mentally rotate" the figures at nearly constant velocity. Further, subjects often describe their strategy in imagery terms. We hypothesized that, to the extent that imagery engages neural mechanisms similar to direct perception, we should see activation during this task in more primary areas of the visual system. Our control condition therefore used the same figures as the rotation task, so that any signal difference could be attributed to the task per se. Further, our slice locations were aligned with visual cortex. We observed little difference in activity in striate cortex between control and task, suggesting that calcarine cortex, and indeed the cuneus, are not engaged strongly by this task. While the first reported fMRI study (1) showed area 7 activity, our results also show consistent activation in frontal eye fields, which we attribute to the subjects scanning the objects to "solve" the task. Indeed, the control task required so little visual scanning that these areas were more active during fixation, which may require more intentional control. Most intriguing was the activation in the lateral occipital sulcus, which we preliminarily assign to human V5/MT (3) . A plausible theory for the neural mechanism of the mental rotation task is that imagined object rotation engages motion-sensitive regions of the brain, while area 7 might be more involved with object identification. The apparent lack of primary visual cortex activity suggests that this task does not require descending input to these regions.