A neuroscientist observes robust BOLD activation in prefrontal cortex during a working memory task and concludes that the prefrontal cortex is necessary for working memory. What is the critical flaw in this reasoning?
AThe hemodynamic response function is too slow to capture working memory processes
BfMRI is a correlational method — activation shows that a region is associated with a task, not that it is causally required for task performance
CBOLD signals in prefrontal cortex are unreliable due to magnetic field artifacts
fMRI activation establishes correlation, not causation. A region may activate during a task as a downstream consequence, as part of a control network responding to difficulty, or because of co-occurring processes — without being necessary for the core computation. Establishing necessity requires a causal manipulation, such as TMS disruption or lesion studies. The 'dead salmon study' illustrated the opposite problem: without proper statistical correction, spurious activations appear everywhere, including in a deceased fish.
Question 2 Multiple Choice
Why does the BOLD signal increase when neurons in a brain region become more active?
AActive neurons directly pump oxygenated blood into nearby capillaries via a motor protein mechanism
BNeural activity consumes oxygen, increasing deoxyhemoglobin, which is paramagnetic and amplifies the MRI signal
CBlood flow to active regions overshoots metabolic demand, flushing out deoxyhemoglobin and reducing local magnetic field distortions — increasing the BOLD signal
DNeurons release iron ions during firing, which enhance local magnetic resonance
The mechanism is counterintuitive: neural activity increases local blood flow *more* than the neurons actually consume. This surplus of oxygenated blood pushes out deoxyhemoglobin (paramagnetic) and replaces it with oxyhemoglobin (diamagnetic). Less deoxyhemoglobin means less field distortion, and the MRI scanner reads this reduced distortion as an increased BOLD signal. This vascular response — not the neural activity itself — is what fMRI measures. It is a proxy that happens to correlate with neural activity.
Question 3 True / False
fMRI's temporal resolution is fundamentally limited to seconds rather than milliseconds because the hemodynamic response unfolds on that timescale, regardless of the speed of the underlying neural event.
TTrue
FFalse
Answer: True
The hemodynamic response function (HRF) rises over 4–5 seconds after a neural event and returns to baseline over 10–15 seconds. This is a property of cerebrovascular physiology, not scanner hardware. Even if neural processing completes in 50 milliseconds, the BOLD ripple from that processing takes 15+ seconds to fully pass. From a Fourier perspective, the HRF acts as a low-pass filter on the neural signal, smearing all rapid events into slow, overlapping bumps.
Question 4 True / False
If the BOLD signal in a brain region peaks 6 seconds after a stimulus, this indicates that the relevant neural processing begins 6 seconds after the stimulus.
TTrue
FFalse
Answer: False
The BOLD peak at 6 seconds reflects the hemodynamic response function, not the onset of neural processing. Neural activity may begin within milliseconds of the stimulus; the vascular response that the scanner detects simply takes 5–6 seconds to peak. This delay is why fMRI is poorly suited to questions about *when* cognitive events occur — the hemodynamic lag obscures the true neural timeline.
Question 5 Short Answer
Explain why a brain region showing increased BOLD activation during a cognitive task may not be causally necessary for performing that task.
Think about your answer, then reveal below.
Model answer: BOLD activation indicates that a region is more metabolically active during the task condition compared to a baseline, but it does not reveal why that activity occurs or whether it contributes to task performance. A region might activate because it receives input from the core processing network, because it monitors task difficulty, because it participates in incidental processes that co-occur with the task, or because of attentional engagement. None of these constitute causal necessity. Demonstrating that a region is necessary requires showing that disrupting it — via TMS, pharmacology, or lesions — impairs behavior. fMRI identifies candidate regions for causal investigation; it cannot itself provide causal evidence.
This is the central interpretive limitation of fMRI as a method. Correlation between activation and task performance is the beginning of an investigation, not its end. The history of cognitive neuroscience includes many cases where confidently identified 'regions for X' turned out to be neither necessary nor sufficient when tested with causal methods.