Lecture 6: Introduction to the Human Brain

What this lecture covers

The lecture answers a practical question: which research methods can actually show that a brain region causes a behavior, rather than merely correlating with it? It wraps up a review of cognitive neuroscience methods started the previous week, closing with transcranial magnetic stimulation (TMS) and animal research, then pivots to the vocabulary and logic of experimental design that students will need to build their own imaging studies.

After watching, you can explain why fMRI and EEG/MEG cannot establish that neural activity causes behavior, describe how TMS and patient lesion studies fill that gap, and define the core experimental-design terms (independent/dependent variable, hypothesis, prediction, confound, contrast, minimal pair) well enough to design a controlled fMRI experiment.

Key ideas

• Two kinds of causality: showing a stimulus causes neural activity is easy to test with fMRI or EEG; showing that neural activity causes behavior is not, because you'd need to manipulate the neural activity itself.
• Temporal vs. spatial resolution trade-off: fMRI has good spatial resolution but a slow, blurred hemodynamic response (peaking ~5-6 seconds after a stimulus), while scalp recordings (EEG/MEG) have fine timing but poor spatial resolution.
• Double dissociation: when one patient can do task A but not B, and another can do B but not A, that is much stronger evidence of separate underlying mechanisms than either deficit alone.
• Transcranial magnetic stimulation (TMS): a brief, strong magnetic pulse through a coil held near the scalp induces an electric field that disrupts a patch of cortex for under a millisecond, letting researchers test the causal role of a region in normal, uninjured subjects.
• TMS limits: it only reaches tissue near the skull (so the deep fusiform face area is out of reach, unlike the more superficial occipital face area), its spatial resolution is roughly one to two centimeters, and it rarely produces positive percepts, mostly just disruption.
• Subtraction logic and minimal pairs: because fMRI only measures differences between conditions, isolating one mental process requires two tasks that are identical except for that process.
• Confound: any difference between conditions other than the intended one, which offers an alternative explanation for a result (for example, snake images having grassier backgrounds than control images).
• Baseline vs. control condition: a control condition (e.g., non-snake images) isolates the process of interest by contrast; a baseline (e.g., fixating a dot) approximates a minimal, near-resting state and shows how selective a response really is.

Walkthrough

Recap of causal inference across methods (0:10)

The lecture opens by reviewing why behavioral methods, fMRI, and EEG/MEG each have limits for establishing causality. Behavioral measures only capture output, so internal processing must be inferred. fMRI has strong spatial resolution but cannot show that a given neural response causes behavior, since it only tracks correlations between stimuli and activity. The key distinction drawn is between "stimulus causes neural activity" (testable by manipulating the stimulus) and "neural activity causes behavior" (not testable with fMRI or EEG alone).

Temporal resolution and invasive recording (3:12)

Using the BOLD response curve, the lecture shows that neural activity in primary visual cortex happens in under a tenth of a second, while the fMRI signal peaks five to six seconds later and is temporally smeared. Direct recording from the cortical surface (in patients undergoing surgery) offers both spatial and temporal precision but is highly invasive and limited to rare clinical cases; it still cannot on its own link neural responses to behavior.

Causal inference from patients and stimulation (7:14)

Studying patients with focal brain damage gives a direct causal link: losing a brain region and losing an ability. Double dissociations strengthen this evidence further, and the fusiform face area stimulation case (where direct electrical stimulation produced a hallucinated face) is presented as a rare, striking example of manipulating neural activity and observing a behavioral/perceptual consequence.

Transcranial magnetic stimulation (9:15)

The lecture explains how TMS works physically (a brief strong current in a coil induces an electric field in underlying cortex), its history and crude early apparatus, its safety profile, and contraindications such as seizure risk. It covers TMS's limited reach (it cannot get at the deep fusiform face area) and describes David Pitcher's experiment zapping the more superficial occipital face area during a same/different face-matching task. Stimulating that region during a narrow window (roughly 60-100 milliseconds after the second face appeared) reduced accuracy compared to stimulating a control site, showing the region is causally engaged in face perception during that time window, though the lecture stresses this doesn't prove the effect is face-specific rather than general to visual perception.

Animal research and its advantages (24:24)

The lecture lists deeper questions that human methods cannot answer well: what exactly is represented in a region, what the neural code is, how regions are anatomically connected, and how face-selective regions develop. After a discussion of the ethics of animal research, including regulation, welfare efforts, and how to weigh costs against long-lasting scientific benefit, the lecture describes what invasive work in monkeys makes possible: direct electrode recordings from hundreds of neurons in face-selective patches, tracking how population codes change over time, and mapping anatomical connections between the roughly six face patches found in monkey cortex, which turn out to be interconnected with each other but not with intervening cortex.

Experimental design vocabulary (33:27)

The lecture defines independent variable (what the experimenter manipulates), dependent variable (what is measured), hypothesis, prediction, confound, and contrast. It stresses that fMRI signal intensity is only meaningful as a difference between conditions, so every imaging experiment needs at least two conditions, and that isolating a mental process requires "subtraction logic": two tasks differing in only the process of interest, ideally forming a minimal pair.

Applying the concepts: designing a snake-perception study (44:34)

Students discuss control conditions for a snake-perception experiment (snakes vs. worms or horns, superimposing objects on a matched background, comparing rigid vs. slithering motion), what task subjects should perform in the scanner without introducing new confounds, and the purpose of baseline conditions such as fixation, which show how selective a region's response really is rather than just whether it differs between two active conditions.

Before you watch

• Review the previous lecture's coverage of behavioral methods, fMRI, and EEG/MEG, since this lecture assumes that context and builds on it.
• Be familiar with the fusiform face area and face-inversion findings from earlier lectures, as they are referenced throughout.
• Some basic familiarity with electromagnetism (the right-hand rule) helps with the TMS explanation, though the lecture recaps it.

Check your understanding

1. Why can fMRI establish that a stimulus causes neural activity, but not that neural activity causes behavior?
2. What makes a double dissociation more informative than a single dissociation?
3. Why can TMS reach the occipital face area but not the fusiform face area, and what does this limit imply for interpreting negative TMS results?
4. What is the difference between a confound and a "sub-optimal" design choice, using the snake-picture background example?
5. Why does an experiment need a baseline condition in addition to an experimental and control condition?