Questions: EEG, Event-Related Potentials, and Neural Timing

Question 1

A researcher wants to determine whether semantic processing of a word occurs before or after syntactic parsing. Which method is best suited to answer this question, and why?

• A. fMRI, because it localizes activity to Broca's and Wernicke's areas with millimeter precision
• B. EEG/ERP, because it can resolve cognitive processes to within tens of milliseconds and track their temporal sequence
• C. fMRI, because its superior signal-to-noise ratio makes it more reliable for language research
• D. EEG/ERP, because it directly measures the firing of individual language neurons

Answer: (B)

The question is about *when* processes occur relative to each other — a temporal sequencing question. EEG has millisecond temporal resolution and ERPs can reveal, for example, that syntactic violations elicit an ELAN at 100–200 ms while semantic violations elicit an N400 at ~400 ms, demonstrating that syntax precedes semantics. fMRI's hemodynamic response unfolds over 4–6 seconds — it cannot distinguish processes separated by hundreds of milliseconds. Option D is wrong: EEG records summed postsynaptic potentials from populations, not individual neuron firing.

Question 2

An ERP experiment presents 200 trials: in 160 trials, a sentence ends with a predictable word; in 40 trials, a semantically unexpected word appears. What will happen to the N400 amplitude as you average more trials together?

• A. N400 amplitude will decrease with more trials because more averaging introduces noise
• B. N400 amplitude will be stable but spatial resolution will improve with more trials
• C. N400 amplitude will become more reliably detectable because the ERP component is time-locked to the event and averages constructively while random background EEG averages toward zero
• D. N400 amplitude reflects individual differences and cannot be recovered by averaging

Answer: (C)

The N400 occurs at a consistent latency (~400 ms) after the unexpected word on every trial — it is time-locked. When you average, this consistent component adds up coherently (√N improvement in SNR with N trials). Background EEG, being random in phase relative to the stimulus, cancels out. This is the fundamental logic of ERP extraction: the neural response is signal (coherent across trials), everything else is noise (incoherent). This is why ERP studies require many trials and why single-trial ERP analysis is technically difficult.

Question 3

EEG primarily records action potentials from individual cortical neurons because action potentials are the main output of neural computation.

• T. True
• F. False

Answer: False

EEG records slow, graded *postsynaptic potentials* (EPSPs and IPSPs) from large, synchronized populations of cortical pyramidal neurons — not action potentials. Action potentials are extremely brief (~1 ms), poorly synchronized across neurons, and have dipole fields that cancel across the population. Postsynaptic potentials last much longer (tens to hundreds of ms) and, when thousands of aligned pyramidal neurons receive synchronous input, their summed electrical fields are large enough to be detected at the scalp. This is a crucial distinction: EEG reflects inputs to cortical regions and local processing, not the all-or-nothing output signals.

Question 4

The P300 ERP component is typically larger when the target stimulus is rare and task-relevant than when it is common.

• T. True
• F. False

Answer: True

The P300 amplitude is inversely related to target probability — rarer targets (fewer than ~20% of stimuli) elicit larger P300s. This reflects greater attentional resource allocation and more substantial working memory updating when something surprising and relevant occurs. The classic 'oddball' paradigm exploits this: frequent 'standard' tones produce no P300; rare 'deviant' tones do. The P300 is thought to index the neural cost of updating a mental model of the current context when an unexpected, relevant event occurs.

Question 5

Explain why EEG and fMRI are better understood as complementary methods than competing ones, and give an example of a research question where each would be preferred.

Model answer: EEG/ERP offers millisecond temporal resolution but poor spatial resolution (scalp signals are spatially smeared by skull and tissue). fMRI offers millimeter spatial resolution but only seconds-scale temporal resolution (hemodynamic BOLD response). They answer different questions: EEG is preferred when you need to know *when* something happens (e.g., 'does the brain detect syntactic errors before semantic ones?'); fMRI is preferred when you need to know *where* activity occurs (e.g., 'which hippocampal subregion is active during memory encoding?'). The trade-off is fundamental — no single current method achieves both simultaneously.

This complementarity is not a limitation to be engineered away but reflects a genuine physical constraint: electrical signals propagate almost instantaneously (giving EEG its temporal edge), while fMRI measures blood flow changes that take seconds to develop. Some researchers combine both methods to leverage the strengths of each — EEG for timing, fMRI for localization — in the same experiment.