Fusiform Face Area and Face Perception
"The Fusiform Face Area: A Module in Human Extrastriate Cortex Specialized for Face Perception" -
Summary
Methodology: Used functional magnetic resonance imaging (fMRI).
Subjects: 15 participants were tested.
Discovery: Identified an area in the fusiform gyrus in 12 out of 15 subjects.
Activation: Found this area was notably more active when subjects viewed faces compared to common objects.
Further Tests within the Identified Area:
Subject-Specific Analysis: Each subject's 'face area' was individually defined based on the initial findings.
Additional Tests Conducted:
Five subjects tested: The predefined 'face area' responded more to:
Intact faces versus scrambled two-tone faces.
Full front-view face photos compared to house front-view photos.
Different set of five subjects tested: The 'face area' responded more to: 3. Three-quarter-view face photos (with concealed hair) than photos of human hands.
Showed stronger response during a matching task involving three-quarter-view faces versus hands.
Significance of the Study:
Methodological Solution: Running multiple tests within the same functionally defined region.
Addressing Common Issues in Functional Imaging:
Correcting for multiple statistical comparisons.
Overcoming ambiguity in interpreting studies with few conditions compared.
Rejecting Alternative Explanations:
Exclusion of Other Functions: The data ruled out explanations based on visual attention, subordinate-level classification, or general processing of animate or human forms.
Confirmation of Selective Function: Demonstrated the fusiform face area's selective involvement in face perception.
This study's use of fMRI revealed a specialized region in the brain, the fusiform face area, which responds significantly more to faces compared to other objects. Further tests within this region across different subjects validated its specificity to facial perception, rejecting alternative explanations and highlighting the significance of this specialized brain area in processing facial information.
Introduction
Background Insights:
Various fields—cognitive psychology, computational vision, neuropsychology, and neurophysiology—indicate different processes for recognizing faces and objects (Yin, Bruce, Tanaka, etc.).
Studies in macaques revealed neurons in the superior temporal sulcus (STS) responding selectively to faces.
Human evidence from epilepsy patients with electrodes in the brain showed specific responses to faces in portions of the fusiform and inferotemporal gyri.
Reports of patients with damage to the occipitotemporal region losing the ability to recognize faces further support the existence of specialized neural modules for face perception.
Challenges and Need for Functional Brain Imaging:
Evidence from neurological patients lacks specific anatomical information.
Functional brain imaging offers higher spatial resolution to study cortical specialization in the normal human brain.
Past imaging studies highlighted regions more active during face viewing but did not establish selective involvement in face perception. These activations could be interpreted in various ways (feature extraction, visual attention, specific recognition, or recognition of animate objects).
Study's Approach:
Used functional magnetic resonance imaging (fMRI) to address ambiguity in interpreting brain activations.
Ran multiple tests within the same cortical region in individual subjects to identify specialized face perception areas.
Initially identified a consistent area in the fusiform gyrus more active during face viewing than object viewing across most subjects.
Subsequently, tested this functionally defined fusiform region in individual subjects across four comparisons to determine its specialization in face perception.
Tests Conducted:
Face versus object viewing.
Measurements during subsequent comparisons aimed to address alternative accounts (low-level feature extraction, visual attention, specific recognition of particular exemplars, recognition of animate objects).
Broad Definition of Face Perception:
Includes various higher-level visual processes related to faces, like detecting a face, extracting identity, gaze direction, mood, and gender information.
This study utilized fMRI to delve into the specialized regions in the brain associated with face perception. By running multiple tests within identified cortical areas, the aim was to provide more definitive insights into how the brain processes and recognizes faces, overcoming interpretational challenges faced in previous studies that compared only a few conditions.
Materials and Methods - Study Design and Procedures
General Design:
The study comprised three main parts aimed at identifying specialized areas for face perception in the occipitotemporal region.
Part I: Searched for areas responsive to faces versus assorted objects in each subject, locating candidate "face areas" and defining regions of interest (ROIs) within individual brains. (defined using the face vs objects comparison)
Rationale for Passive Viewing Task: Focused on passive viewing rather than active tasks as foveal face perception is automatic and difficult to volitionally control.
Results of Part I: Revealed consistent activation in the right fusiform gyrus, suggesting specialization for face perception.
Part II and III Comparisons:
Part II: Tested five subjects using different stimulus comparisons from Part I to explore alternatives to the face perception hypothesis.
Compared intact two-tone faces with scrambled two-tone faces.
Compare front-view face photos with front-view house photos.
Part III: Involved a new set of five subjects, further exploring the specialization of the identified face areas.
Compared three-quarter-view face photos with photos of human hands.
Conducted a matching task using the same stimuli to test attentional mechanisms.
Subject Details:
15 subjects (9 women, and
6 men) participated, reporting normal or corrected-to-normal vision and no neurological history.
Some subjects participated in multiple parts of the study across different scanning sessions.
Stimuli:
Stimuli were grayscale photographs (or photo-like images) of faces, objects, and houses.
Various stimulus sets were used to test different aspects of face perception, including face recognition based on viewpoints and external/internal features.
MRI Acquisition:
Scans were conducted using a 1.5 T MRI scanner with echo-planar imaging, covering the occipital and most of the temporal lobes.
Functional data were obtained using an asymmetric spin echo sequence with a voxel size of 3.25 x 3.25 x 6 mm.
Data Analysis:
15 subjects' data were analyzed after omitting subjects with excessive head motion or other artifacts.
Motion correction was not applied to the remaining subjects.
Data were statistically analyzed using Kolmogorov–Smirnov tests and Talairach coordinates to identify regions responding more to faces than objects.
Regions of interest (ROIs) in the right fusiform gyrus were defined based on signal intensity differences between face and object epochs.
Analyzing Results:
ANOVA was used to test signal intensity differences between different conditions across subjects.
Time courses of signal intensity were converted into percent signal change to measure the selectivity of face ROIs to specific stimulus contrasts.
This study used fMRI to conduct multiple tests on individual subjects to identify specialized brain areas for face perception. It employed various stimuli and comparisons across different parts of the study to investigate the neural mechanisms involved in face processing
Results - Part I: Identification of Face-Sensitive Brain Areas
Observations:
Examined brain areas more active during face versus object viewing.
Identified a region in the right fusiform gyrus significantly activated during face presentation for most subjects.
Another bilateral and medial area in the right and left hemispheres showed higher activation for objects than faces, forming a double dissociation.
Right fusiform activation was consistent across subjects, becoming the focus for detailed investigation in subsequent parts.
Fusiform Activation and Subject Variability:
In 12 of 15 subjects, fusiform activations for faces were observed; in three subjects, no significant activation was found.
Consistency: Despite variability, fusiform activation was consistent in location and Talairach coordinates across subjects.
Right hemisphere activation was more prevalent in right-handed subjects; bilateral activation was also observed.
Activation size: Average size for right hemisphere activation in right-handed subjects was 1 cm³, and for left hemisphere activation, 0.5 cm³.
Reliability: Test-retest reliability shown in subject S1 over multiple testing sessions spread across six months.
Variability in Hemispheric Lateralization:
Right hemisphere activation was predominant in right-handed subjects, while left hemisphere and bilateral activations were observed in left-handed subjects.
Additional activation: Some subjects also exhibited activation in the middle temporal gyrus/superior temporal sulcus (ST) of the right hemisphere for faces compared to objects.
Other activations: While most subjects showed additional activations, these were not consistent across subjects.
The study identified a consistent activation in the right fusiform gyrus during face viewing, demonstrating its potential specialization for face perception across subjects. This activation was further investigated in subsequent parts of the study.
Results - Part II: Testing Face Stimulus Specificity in Fusiform Activation
Objective:
To ascertain if the observed activation in the fusiform gyrus during face viewing was specific to faces rather than general visual stimulus differences.
Method:
Involved five subjects from Part I, scanned in the same session, using the previous face Regions of Interest (ROIs).
Analyzed responses in the right fusiform face ROI for different stimulus conditions: faces versus objects, intact faces versus scrambled faces, and faces versus houses.
Observations:
Clear activation pattern: Raw data from each subject’s face ROI showed higher activation for faces compared to nonface stimuli.
Statistical analysis (three-way ANOVA) across subjects confirmed significantly higher signal intensity during face epochs than control stimulus epochs.
Pairwise comparisons for each test independently reached significance.
Conversion of ROI-averaged time course data into percentage signal change showed distinct peaks during face epochs compared to nonface epochs.
Selectivity Ratio: Ratio of average percentage signal change during face epochs to nonface epochs ranged from 2.8 to 6.6, indicating high stimulus selectivity in face ROIs.
Comparison with Previous Study: The selectivity ratios were comparable to reported ratios for responses in visual area MT to moving versus stationary displays by Tootell et al. (1995).
Conclusion:
The fusiform gyrus region responding strongly to faces also demonstrated higher activation for intact versus scrambled faces and faces versus houses.
Limited data for the MT gyrus/ST sulcus activation restricted the evaluation of its selectivity due to weak or absent responses for faces versus objects in most subjects.
These findings reinforced the specificity of the fusiform gyrus activation to face perception rather than general visual stimulus differences, supporting its role in face processing.
Results - Part III: Further Examination of Fusiform Activation
Objective:
Determine if the fusiform gyrus activation for faces observed in Part I was attributable to several factors including animate vs. inanimate objects, visual attention, or specific cognitive tasks.
Method:
Utilized five subjects (including two from Part II) in the same session for Parts I and III.
Analyzed fusiform face ROIs from Part I to test responses during new tasks: passive viewing of three-quarter faces vs. hands, and 1-back matching task for faces vs. hands.
Face ROIs are defined separately for each using the average of two faces vs object
Observations:
Individual raw data for each subject showed higher signal intensities in the face ROI during the new face compared to nonface tests.
Statistical analysis (three-way ANOVA) revealed a significant main effect of higher signal intensity for face stimuli than nonface stimuli across subjects.
Independent analyses confirmed significant differences for each test (faces vs. objects, faces vs. hands passive, faces 1-back vs. hands 1-back).
Percentage signal change analysis demonstrated clear peaks during face epochs compared to nonface epochs, indicating high selectivity for faces in the fusiform ROI.
Limited data regarding the MT gyrus/ST sulcus activation showed activations in this region for faces versus objects in only two subjects (S10 and S11). These subjects also showed greater signal intensities for faces versus hands, suggesting partial selectivity for faces.
Additional Information:
Absence of behavioral responses from subjects during the 1-back task due to technical limitations. Experimenters confirmed task performance by monitoring subject responses and stimuli in real-time.
Subjects performed well above chance in both tasks, with subsequent behavioral measurements showing similar performance levels in both tasks in a different group of subjects in a laboratory setting. However, subjects reported greater difficulty with the hands task than the faces task.
Conclusion:
The fusiform gyrus region exhibiting stronger responses to faces compared to objects also showed increased activation during passive viewing of three-quarter faces versus hands and a 1-back matching task for faces versus hands.
Limited data for the MT gyrus/ST sulcus activation suggested partial selectivity for faces, which requires further confirmation with more subjects.
Both tasks engaged general attentional mechanisms, with subjects finding the hands task at least as challenging as the faces task, suggesting comparable attentional engagement in both tasks.
Discussion
Identification of Fusiform Gyrus Activation:
Fusiform gyrus activation in 12 out of 15 subjects responded more to passive face viewing compared to object stimuli.
Identified and used specific fusiform face ROIs within subjects for further face selectivity tests.
Conducted multiple tests:
Response to intact faces vs. scrambled versions to rule out luminance differences.
Greater response to faces than houses, indicating high stimulus selectivity.
Response to three-quarter-view faces vs. human hands, demonstrating response specificity.
Activation during a matching task for faces vs. hands, suggesting task specificity.
Specificity of Fusiform Face Area (Area FF):
Confirmed fusiform activation specifically linked to face perception, ruling out general processes related to visual attention or categorization.
Area FF responded to various face stimuli including different orientations and appearances, showing generalization to face perception.
Responded more strongly to faces despite external features (like hair) being concealed, indicating involvement in face recognition over general head recognition.
Variability and Localization of Area FF:
Test-retest reliability observed within subjects across multiple sessions, indicating consistency in within-subject results.
Area FF mostly found in the fusiform gyrus or adjacent cortical areas in right-handed subjects.
More activation in the right than left fusiform, consistent with earlier studies.
Variation observed across individuals might reflect actual individual differences in the fusiform face area's localization.
Comparison with Other Brain Areas:
Additional activation observed in the middle temporal gyrus/STS for some subjects but not consistently across all.
Comparison to macaque studies suggests the homology of human fusiform activation to inferotemporal regions involved in face perception.
Implications and Conclusions:
Establishes the existence of a fusiform region specifically responsive to faces, distinct from general object processing areas.
Demonstrates cortical specialization through testing the responsiveness of the same cortex region to various stimuli.
Suggests the need for distinct computational processes for face recognition compared to other objects.
Opens avenues for future studies to explore holistic processing in face recognition compared to other stimuli and the role of visual experience in the development of specialized cortical modules like area FF.