This 2005 study by Schwarzlose, Baker, and Kanwisher is a landmark paper in cognitive neuroscience that addresses a fundamental debate: is the human visual system organized into highly specialized "modules" for specific categories, or does it use a more distributed, overlapping code?
TL;DR
By pushing the limits of fMRI resolution, researchers discovered that the "Fusiform Face Area" (FFA) and the "Fusiform Body Area" (FBA) are actually distinct, adjacent cortical regions. Previous reports of the FFA responding to bodies were likely an artifact of low-resolution imaging blurring these two specialized neighbors together.
What Problem Does This Paper Address?
For years, the Fusiform Face Area (FFA) was the poster child for "domain-specific" processing—the idea that the brain has a dedicated "app" for faces. However, this view was challenged by studies showing that the FFA also responds quite strongly to images of headless bodies.
This led to a major theoretical rift. One camp argued that the FFA is actually a "face-and-body" area, supporting a "graded and overlapping" model of object recognition. The authors of this paper proposed an alternative: perhaps the FFA is perfectly selective for faces, but it sits right next to a body-selective area. At standard fMRI resolutions, a single "pixel" (voxel) might contain parts of both regions, leading to a blurred signal that looks like dual selectivity—a phenomenon known as partial volume effects.
Methods
The researchers tested this "blurring" hypothesis by comparing standard-resolution fMRI with high-resolution fMRI.
- Resolution Contrast: They compared standard resolution (3.125 x 3.125 x 4.0 mm) against high resolution (1.4 x 1.4 x 2.0 mm). The high-res voxels were roughly 10 times smaller in volume than the standard ones.
- Two-Step Experimental Design:
- Localizer Scans: Participants viewed blocks of faces, bodies, and objects to identify the FFA and FBA in each individual.
- Event-Related Scans: Using independent data, they measured the response of these identified regions to specific stimuli (faces, headless bodies, body parts, and cars).
- ROI Refinement: In a clever move, they created "pure" regions of interest (called FFA and FBA**) by mathematically removing any voxels that showed overlapping selectivity in the localizer scans.
Datasets
The study involved two experiments:
- Experiment 1 (Standard Res): 7 participants.
- Experiment 2 (High Res): 10 participants (9 analyzed due to head motion).
- Stimuli: A custom set of images including faces, headless bodies, body parts (like hands or feet), cars, and assorted everyday objects.
Key Insights and Ideas
- The "Blur" is Real: At standard resolution, the FFA and FBA overlapped significantly (about 27% of their total volume). At high resolution, this overlap dropped to 18%.
- Anatomical Neighbors: The study confirmed the existence of the Fusiform Body Area (FBA), a region on the ventral surface that is distinct from the more famous Extrastriate Body Area (EBA). Crucially, the FFA is consistently located medial (closer to the midline) to the FBA.
- Resolution Restores Selectivity: As imaging resolution increased, the "accidental" response to bodies in the FFA decreased.
- Modular Purity: The most striking "aha moment" came from the refined FFA and FBA regions. When the overlapping voxels were removed, the FFA showed zero response to bodies above the level of control objects (like cars). Similarly, the FBA showed no special response to faces. This suggests that the underlying neural populations are likely 100% specialized; the "overlap" seen in previous studies was a measurement error, not a biological reality.
Key Results
- Standard Res: FFA responded significantly more to bodies than to cars ($p < 0.005$), appearing to support the "overlapping representations" theory.
- High Res (Refined): In the FFA*, the response to headless bodies was statistically indistinguishable from the response to cars or mixed objects ($p > 0.3$).
- Spatial Separation: The centers of the FFA and FBA were separated by an average of only 3.1 mm. This explains why standard 4mm-wide voxels would easily catch both regions.
- Body Parts vs. Bodies: The FFA's response to bodies wasn't just "face-inference" (the brain imagining a face on a headless body). The FFA responded even more strongly to body parts (where a face is unlikely to be inferred) than to cars, further proving that the signal was simply "bleeding" over from the nearby FBA.
Limitations and Future Work
- Functional Differences: While the study proves the FBA is separate from the FFA, it doesn't fully explain how the FBA differs from the EBA (the other body-selective area).
- The Overlap Region: The authors acknowledge that while they've shown some parts of the fusiform are purely selective, they haven't ruled out the possibility that a small population of "interleaved" neurons truly does respond to both categories. Distinguishing this would require even higher resolution or fMRI adaptation techniques.
- Clinical Questions: The authors suggest that future research should look at patients with prosopagnosia (face blindness). If the FFA and FBA are so close, do these patients also have undiagnosed deficits in body perception?
This paper serves as a cautionary tale for neuroimaging: sometimes the "distributed" nature of the brain is actually just a limitation of our cameras.
Summary updated