Questions: Causal Inference in Neural Research

Question 1

An fMRI study shows that brain region X activates significantly more during face recognition than during object recognition. A researcher concludes that region X is necessary for face processing. What is wrong with this reasoning?

• A. fMRI cannot detect neural activity, only blood flow changes
• B. Activation demonstrates association, not necessity — a TMS or lesion study is needed to show the region is causally required
• C. The study should have used a different baseline condition
• D. Region X may simply be the only face-processing region, making the result trivially true

Answer: (B)

Neuroimaging reveals correlations — region X co-varies with face processing — but cannot establish that the region is necessary or sufficient. The region might be active because of attention, emotional salience, or proximity to the true face-processing circuit. Only interventional methods that perturb the region (TMS, lesions) and observe behavioral consequences can establish necessity. This is the core causal inference problem in neuroimaging research.

Question 2

Patient A has damage to region X and loses function Y but retains function Z. Patient B has damage to region Z and loses function Z but retains function Y. This pattern is called a double dissociation and it demonstrates what?

• A. That function Y is more cognitively demanding than function Z
• B. That functions Y and Z depend on genuinely independent neural systems
• C. That region X and region Z are the only brain areas involved in these functions
• D. That lesion studies are more reliable than TMS studies

Answer: (B)

A double dissociation rules out the 'harder task' confound: if Y were simply harder than Z, any brain damage might impair Y while sparing Z. The double dissociation shows that either function can be selectively preserved depending on lesion location, which is only possible if they depend on distinct neural substrates. This is why double dissociations are the gold standard for inferring functional independence in neuropsychology.

Question 3

If a TMS pulse delivered over brain region X produces no measurable impairment on task Y, this result proves that region X is not involved in task Y.

• T. True
• F. False

Answer: False

A null TMS result shows that region X is not *necessary* for task Y as tested, but it does not prove the region is entirely uninvolved. The TMS timing might have missed the critical window, the spatial targeting might be imprecise, or the region might contribute redundantly with other areas. Null results are genuinely informative — they rule out necessity under specific conditions — but they cannot establish that a region is completely uninvolved. This is an important asymmetry in causal inference from interventional neuroscience.

Question 4

A double dissociation provides stronger causal evidence for functional independence between two cognitive processes than a single dissociation alone.

• T. True
• F. False

Answer: True

A single dissociation (damage to X impairs only Y) can be explained by task difficulty differences — harder tasks are always more vulnerable to any brain damage. The double dissociation eliminates this confound because it shows each function can be selectively preserved while the other is impaired, depending only on which region is damaged. The pattern is only coherent if the two functions genuinely rely on separate neural systems, not just if one is harder than the other.

Question 5

Why can neuroimaging alone not establish that a brain region is necessary for a cognitive function, and what kind of evidence is required?

Model answer: Neuroimaging shows correlation — a region is active when a function is engaged — but correlation does not imply necessity. A region could be active due to associated processes (attention, effort, emotional response) rather than the function itself. Establishing necessity requires interventional evidence: showing that when the region is disrupted (via TMS, lesion, or pharmacology), the function is impaired. TMS provides virtual lesions in healthy participants; permanent lesion studies provide complementary evidence. Ideally, converging evidence from both imaging and intervention gives the strongest causal claim.

The logic parallels causal inference generally: observational co-variation generates hypotheses, but only experimental manipulation — holding everything else constant while perturbing the variable of interest — can establish causation. In neuroscience, the 'experiment' is an intervention on brain function, not just measuring it.