MIT OpenCourseWare 1105: Methods in Human Cognitive Neuroscience with a Focus on Face Perception

Overview

This lecture from MIT OpenCourseWare surveys the methods used in human cognitive neuroscience to study face perception, showcasing how behavioral data, fMRI, EEG/ERP, MEG, and intracranial recordings contribute to our understanding of specialized brain regions and representations.

Key insights

• Faces are treated as a distinct visual category with specialized brain mechanisms, evidenced by selective responses and inversion effects.
• Functional MRI provides great spatial resolution but is correlational, while EEG/ERP and MEG offer precise timing to reveal fast processing dynamics.
• Direct intracranial recordings yield high spatial and temporal precision, and stimulation can reveal causal roles in face processing.
• Double dissociations between face and object recognition strengthen claims about domain-specific brain circuits.

Introduction and Course Context

The presentation introduces methods in human cognitive neuroscience, using face perception as a richly informative domain to illustrate how different techniques address questions about processing inputs, outputs, and representations in the brain. The lecturer emphasizes that methods exist to answer questions, not to impress peers, and begins with a speed through behavioral data and functional MRI before expanding to additional methods and a quiz.

From Behavior to Brain: Core Questions

Centering on Marr's computational theory level, the discussion frames questions about what problem faces solve, what the inputs and outputs are, and how much information is carried by the input. A central behavioral finding is that familiar faces show more invariant representation across images than unfamiliar faces, illustrating that invariant coding is not completely image independent. The inversion effect is introduced as strong evidence that face processing may rely on specialized mechanisms in the brain, a claim historically supported by Yin's classic stopwatch study comparing upright and inverted faces.

Strengths and Limits of Behavioral Methods

Behavioral methods are cheap and useful for characterizing internal representations and dissociating mental phenomena, but they provide only sparse data and lack direct brain measurements. This motivates the use of brain imaging and recording techniques to link behavior to neural processes, while remaining mindful of interpretational limits and the need for causal evidence.

Imaging Methods: fMRI and Localizers

Functional MRI is highlighted as the best noninvasive method for examining brain region selectivity in normal subjects. Early work suggested a face-selective region (often localized in the fusiform gyrus, referred to as the fusiform face area, or FFA). A key methodological point is the use of subject-specific localizers to identify this region in each participant, acknowledging that exact location varies across individuals. The lecture also stresses that observing a region that responds more to faces than objects does not prove it is necessary for face recognition, motivating further causal tests and consideration of alternative accounts.

Temporal Dynamics: EEG, ERP, and MEG

EEG/ERP and MEG complement fMRI by providing millisecond timing. The ERP data reveal a face-sensitive response around 170 milliseconds after stimulus onset (the N170 component), with stronger right-hemisphere involvement. MEG data parallel EEG findings, confirming early face processing and offering improved spatial localization relative to EEG alone, though still less precise than fMRI. The methods emphasize that timing matters for understanding the sequence of processing stages and the nature of computations (feedforward versus feedback, rapid discrimination versus iterative processing).

Causal Probes: Intracranial Recording and Stimulation

The most compelling causal evidence comes from intracranial recordings in neurosurgical patients and direct electrical stimulation. Intracranial electrodes provide high spatial and temporal resolution, allowing recording from individual or small populations of neurons in regions like the fusiform face area. Stimulating those regions can selectively evoke or distort face perception, offering strong causal claims about the region’s role. The talk also notes limitations, including clinical constraints and the difficulty of generalizing from a single patient to broader theories of face processing.