Lecture 5: Cognitive Neuroscience Methods II

What this lecture covers

Continuing directly from the previous lecture, this session reviews behavioral and fMRI evidence for a specialized face-recognition system, then works through methods that add what fMRI cannot provide: speed (EEG and MEG) and causal proof (patient lesions and direct brain stimulation). Face perception remains the running example throughout, culminating in a same-electrode stimulation video showing a face literally appearing to distort on a patient's own face during stimulation.

After watching, you should be able to compare the spatial and temporal resolution trade-offs of EEG, MEG, and intracranial recording, explain the logic of a functional localizer, and describe what double dissociation evidence (prosopagnosia versus object agnosia) establishes that single case studies cannot.

Key ideas

• Face inversion effect: people recognize upright faces far better than upside-down ones, and this cost is disproportionately larger for faces than for other object categories (Robert Yin's original stopwatch-and-paper study), suggesting face recognition works differently from general object recognition.
• Functional localizer: because the exact location of a face-selective region varies between individuals, researchers first identify it in each subject's own brain with a functional scan before asking further questions of it.
• N170 / M170: a face-selective electrical (EEG) or magnetic (MEG) response appearing around 170 milliseconds after a face appears, showing the brain discriminates faces from other stimuli very quickly.
• Temporal versus spatial resolution trade-off: EEG and MEG have excellent timing but poor spatial precision; fMRI has good spatial resolution but poor timing; intracranial recording in neurosurgery patients is the rare method offering both.
• Correlation versus causation: a brain region responding more to faces than objects (fMRI, EEG, single neurons) shows correlation; only damage, stimulation, or similar interventions can show the region is causally necessary.
• Double dissociation: finding one patient who can't recognize faces but can recognize objects (prosopagnosia), and another who can't recognize objects but can recognize faces (patient CK), together rule out the explanation that face recognition is simply "harder" object recognition.
• Electrical stimulation: directly stimulating a face-selective electrode in a neurosurgery patient visibly and specifically distorts perception of a face (but not of other objects) shown at the same moment, offering strong causal evidence.

Walkthrough

Review and the face inversion effect (0:10)

Kanwisher briefly reviews the previous lecture's Dutch politicians findings, then introduces Robert Yin's classic finding that people are much worse at recognizing previously studied faces when they are shown upside down, and that this inversion cost is larger for faces than for houses or stick figures, an early clue that face processing may rely on distinct machinery.

Strengths and weaknesses of behavioral methods (9:18)

She summarizes that behavioral data are cheap and good for dissociating mental phenomena, but sparse: they only give accuracy and reaction time, without revealing intermediate processing stages, motivating the move to brain-based methods.

Functional MRI and the localizer logic (10:19)

Revisiting the faces-versus-objects fMRI experiment, Kanwisher explains why researchers localize the face-selective region in each subject individually rather than using group-averaged coordinates, since its exact location varies from person to person. Using an inverted two-tone face image that most people cannot recognize as a face until told, she shows the region responds much more strongly when subjects consciously perceive a face in the identical image, ruling out low-level explanations based on spatial frequency or contrast alone.

EEG and MEG: speed over location (25:28)

The lecture introduces scalp EEG and event-related potentials (ERPs), comparing their spatial blur to a microphone in a stadium detecting a touchdown without knowing who scored. Data from the 1996 Bentin study show a face-selective ERP response (N170) emerging around 170 milliseconds. MEG, invented at MIT, offers similar timing with slightly better localization by detecting magnetic fields from cortical activity in sulci, shown with a matching M170 response.

Intracranial recording: the rare method with both (39:42)

In neurosurgery patients being monitored for epilepsy, electrodes placed directly on or in the brain give both good spatial and temporal resolution. Kanwisher shows data from an electrode grid in Japan revealing a highly face-selective response peaking near 170 milliseconds, and later single-neuron recordings from a patient's fusiform face area showing individual face-selective neurons firing to faces but not objects, tools, or patterns.

Causal evidence: lesions and stimulation (54:49)

The lecture turns to causality. A patient with a lesion overlapping the typical FFA location cannot recognize faces but recognizes objects normally, while patient CK shows the reverse pattern, unable to recognize objects but fully able to recognize faces; together these form a double dissociation that rules out the idea that face recognition is simply harder object recognition. Finally, a video shows a neurosurgery patient's face-selective electrode being electrically stimulated: while looking at a colleague's face, the patient reports the face appearing to distort, but stimulation during viewing of non-face objects produces no such distortion, offering direct causal evidence that the region is specifically involved in face perception.

Before you watch

• Watch Lecture 4 first; this lecture continues its review of behavioral and fMRI evidence for face-selective brain regions.
• Basic familiarity with fMRI and the BOLD signal, introduced in the previous lecture, is assumed.

Check your understanding

1. Why is the face inversion effect (worse recognition of upside-down faces) treated as evidence for specialized face-processing machinery?
2. Why do researchers localize a face-selective brain region separately in each individual subject instead of using averaged group coordinates?
3. What are the main resolution trade-offs between EEG/MEG and fMRI, and what advantage does intracranial recording offer over both?
4. Why does finding a patient with prosopagnosia alone not prove that a brain region is specialized only for face recognition, and how does patient CK's case address that gap?
5. What did the electrical stimulation demonstration show that the earlier fMRI and EEG evidence could not?