1Max Planck Institute for Human Development, Berlin 14195, Germany, 5Department of Education and Psychology, and 6Dahlem Institute for the Neuroimaging of Emotion, Freie Universität Berlin, Berlin D-14195
Stimuli and task. A, Example event-related fMRI trial. Subjects decided whether a noisy stimulus represented a house or a face, without knowing how they would indicate their response. Following a delay, subjects received a specific motor preparation instruction for either an eye or hand movement. Star and house symbols indicated faces versus houses, respectively (counterbalanced across subjects). During the response stage, subjects indicated their face or house decision by either saccading to the remembered target location in eye trials (up, down, left, or right), or pressing the corresponding button on a diamond-shaped button box in hand trials. A red fixation cross indicated the start of the response period. Subjects maintained fixation at all times of the trial, except during eye movement responses. Fixation was used as baseline in-between trials (jittered duration, see Materials and Methods). B, Example face and house stimuli at high or low levels of sensory evidence (low or high levels of noise, respectively). Note that these levels were adjusted for each subject individually, based on performance during training and after each run in the scanner. Face and house stimuli and levels of sensory evidence were pseudo-randomized across trials. C, Possible motor plans for the motor preparation stage. Eye and Hand trials as well as target locations were pseudo-randomized across trials. F_Hi, H_Hi, High sensory evidence for faces and houses, respectively; F_Lo, H_Lo, low sensory evidence for faces and houses, respectively.
Accuracy (percentage correct classification) for high and low levels of sensory evidence for faces and houses. Subjects were sensitive to the amount of sensory evidence available for their perceptual decision, with significantly higher accuracy for higher levels of sensory evidence. Error bars represent the SEM. Abbreviations are as in Figure 1.
BOLD activation contrasts between faces and houses, and between high and low sensory evidence. A, Red to yellow: activations greater for face stimuli than house stimuli (across both high and low sensory evidence). Face stimuli activated the fusiform gyrus, including the fusiform face area, significantly more than house stimuli. Dark blue to light blue: the parahippocampal place area shows greater activation for house stimuli than face stimuli. Fusiform face and parahippocampal house regions of interest were selected in each subject based on Face>House and House>Face contrasts. B, Multiple brain regions are modulated by the amount of sensory evidence available for a decision. Red to yellow: greater activations for high sensory evidence than low sensory evidence (across both faces and house) can be found in multiple brain regions, including prefrontal and inferior parietal cortices. Dark blue to light blue: greater activations for low sensory than high sensory evidence can be found in multiple cortical and subcortical areas, including prefrontal and posterior parietal cortex. All activations are shown at voxelwise p < 0.001, corrected to p < 0.05 with AlphaSim (see Materials and Methods).
PPI results. A, Several brain regions increase their effective connectivity with face and house regions specifically during the face–house decision stage (Table 1). Activations are shown at voxelwise p < 0.001, cluster corrected to p < 0.05 (see Materials and Methods). The percent BOLD signal change of these areas is shown in Figure 5. The seed time series for the PPI was the absolute difference between face and house voxels selected in the fusiform face gyrus and parahippocampal gyrus (see Fig. 3A; see Materials and Methods). B, PPI results surviving voxelwise p < 0.005, corrected to p < 0.05. This lower threshold reveals additional brain areas that increase their effective connectivity with face and house regions during the decision stage, including the left pre-central sulcus and right intraparietal sulcus (also see Table 2). None of these additional areas showed a response profile consistent with the accumulation of sensory evidence for perceptual decision making. The percent BOLD signal change of these regions is shown in Figure 6. a, Anterior; post., posterior; inf. pre-CS, inferior pre-CS; sup. par. gyr., superior parietal gyrus, postCG, post-central gyrus. Also see Fig. 3 legend. Left is left in coronal and axial images.
Percent BOLD signal change for each area revealed by the PPI in Figure 4A. The left inferior frontal sulcus is the only region that (1) shows a modulation by the amount of sensory evidence (here: Low>High); (2) shows greater activation during the perceptual decision stage than during the rest of the trial; and (3) is positively activated during the perceptual decision stage. Other areas are deactivated (left anterior IFG), not significantly different from baseline during the decision (right inferior pre-CS, left posterior SFS, fundus of left IFS), or are not modulated by evidence levels during the sensory stage. Error bars represent the SEM (vs baseline). Note HI versus LO Sensory error bars are versus baseline, not for the difference between HI and LO. A paired two-tailed t test between HI and LO Sensory activations was significant (p < 0.05). HI, High sensory evidence (both faces and houses); LO, low sensory evidence (both faces and houses).
Percent BOLD signal change for PPI areas shown in Figure 4B. The majority of these areas show negative BOLD responses during the perceptual decision stage or the entire trial (vs fixation baseline). The left pre-central sulcus is activated across all events, with no significant modulation by amount of sensory evidence. The right posterior IPS and a second IPS region also show no modulation by amount of sensory evidence for the decision. Abbreviations are as in Figures 4 and 5.
Modulation by sensory evidence during motor preparation. A, Red to yellow: putative bilateral LIP (lateral intraparietal area) shows greater activation for high sensory evidence compared with low sensory evidence, but only during the eye movement preparation stage (i.e., after the motor plan is known). B, Dark blue to light blue: areas in posterior parietal cortex, including medial parietal cortex, do not show greater activation for high compared with low sensory evidence during hand movement preparation. In contrast, motor areas such as the left pre-SMA and right paracentral lobule do show greater activation for hand movement preparation in high sensory versus low sensory trials. C, LIP shows a reversal of High>Low and Low>High modulations from the sensory stage to the eye movement preparation stage. During the decision (sensory) stage, both left and right LIP show a Low>High modulation (**paired two-tailed p = 0.008 and 0.005; t(15) = 3.03 and 3.28 for left and right LIP, respectively). In contrast, during the eye movement preparation stage, both left and right LIP show a High>Low modulation in a High>Low activation contrast (+p < 0.05, corrected at whole-brain level; within the ROI: paired two-tailed p = 0.007 and 0.009, t(15) = 3.14 and 3.02 for left and right LIP, respectively). Note that although the ROI was selected based on the latter contrast, High>Low during eye movement preparation, the contrast is significant at the whole-brain-level, corrected, before the selection of any ROI. Error bars represent the SEM.
Overlay between PPI activations (p < 0.05, uncorrected) and Eye Preparation HI>LO activations (p < 0.05, corrected with AlphaSim cluster thresholding) around the IPS. A, Stringent thresholds and clustering might accidentally miss a possible IPS activation that might show a modulation by sensory evidence during the decision (sensory period). To ensure that we did not miss any IPS PPI activations that may be indicative of sensory integration, we lowered the threshold to p < 0.05 (uncorrected) for PPI activations. Two activations appeared lateral and medial of the IPS, bilaterally (in red to yellow). We overlaid the LIP activation from Figure 7 (in blue: Eye Preparation HI > LO, p < 0.05, corrected with AlphaSim cluster thresholding) on the same image. The motor LIP activation (blue) did not overlap with any of the uncorrected IPS PPI activations (red). B, None of the uncorrected IPS PPI activations showed a significant modulation by sensory evidence during the perceptual decision, as required for a sensory integration role during the decision (sensory) stage. All two-tailed paired t tests (HI vs LO during the Sensory period) were nonsignificant. The many regions and response profiles around the IPS suggest functional heterogeneity, even within saccade-responsive regions.
An SFS area previously identified as posterior DLPFC does not integrate sensory evidence. A, Red to yellow: a region located in the left SFS (see crosshairs) shows significantly greater activation (voxelwise p < 0.0001) for high compared with low sensory evidence (faces and houses). B, A percent BOLD signal change analysis in this region shows that it is deactivated versus fixation baseline during both high and low sensory evidence stimuli. Further, it is not significantly above baseline during other parts of the trial. The positive difference in signal is due to the significantly less negative activation during high compared with low sensory evidence, consistent with previous literature (see Discussion). This region did not increase its effective connectivity with face and house regions during the decision period (Fig. 4). Abbreviations are as in Figure 5 legend.
Brain areas involved in a psychophysiological interaction with fusiform face and parahippocampal house voxels during the decision period, thresholded at voxelwise p < 0.001, cluster corrected to p < 0.05 with AlphaSim. L, Left; R, right; BA, Brodmann area.
Brain areas involved in a psychophysiological interaction with fusiform face and parahippocampal house voxels during the decision period, thresholded at voxelwise p < 0.005, cluster corrected to p < 0.05 with AlphaSim. The top seven regions highlighted in bold are the same as the regions listed in Table 1, surviving a voxelwise p < 0.001, clustercorrected to p < 0.05. Several of the PPI clusters from Table 1 (left IFS and left anterior IFG; posterior SFS and MFG) merge at the lower threshold shown here. BA, Brodmann area.