Abstract
Using fMRI to investigate visual processing of facelike objects is intriguing in two aspects: first, its name implies the interplay between several category-related (e.g., face and object-selective) regions-of-interest (ROIs); second, since fMRI is relatively slow in temporal resolution, whether it is possible to show the transient face-related signatures (as N170 in ERP or M170 in MEG) in fMRI, and under what task context and temporal resolution, will also reveal the strength (and limitations) of different methodologies. Previously, we have found that under two different task contexts: face/object classification (FOC), or facelikenss judgment (FLJ), face-like objects (or FLO) showed two different patterns of time courses in face- (e.g., FFA) and object- (e.g., LOC) selective areas: in FOC condition, FLO drove the LOC, but flat FFA responses; whereas in FLJ condition, FLO drove the FFA in earlier times (< 6s), but LOC increased later (after 6s). To further mitigate the FFA and LOC interaction (and also with other brain regions), Granger Causality Mapping (GCM) was applied to assess relative dGCM by either FFA or LOC. We found that (a) under both FOC and FLJ tasks, LOC Granger-caused many related areas: inferior temporal areas, primary visual cortex, anterior cingulate, dlPFC, and r-Precuneus; whereas the rFFA was more influenced by surround IT areas, especially more so in FLJ context. In addition, MVPA searchlight analyses on FLO under two tasks revealed that the only significant region to reliably classify was thalamus, a region implicated in input-output mapping contingencies. Taken together, these results suggested that FLO was flexibly modulated by the task instruction, and the effective connectivity between FFA and LOC, revealed by GCM, showed not only the distributed nature of LOC and FFA under different contexts, but also the influencing and being influenced to upstream and by downstream areas, via the gating control in thalamus.