A popular efficiency tip says to read the question stem before the stimulus, so you know what you're looking for. For some question types this genuinely helps — for others, it can quietly backfire, priming you to see what you expect rather than what's actually there.
For questions with a narrow, mechanical task — like identifying a specific flaw type, or finding a parallel structure — knowing the task in advance lets you read the stimulus with a targeted lens, which can be a real efficiency gain.
For Must Be True or Most Strongly Supported questions, knowing the stem in advance can cause you to read the stimulus looking for confirmation of an answer you haven't even seen yet — subtly filling in gaps or over-weighting details that seem to point toward a plausible-sounding inference, before you've actually processed the argument on its own terms.
Confirmation bias doesn't require bad intentions — it's a natural byproduct of reading with an expectation already in mind. Reading the stimulus cold, then the question, forces you to build a neutral understanding first, before any specific task biases your attention.
This isn't an argument for always reading one way or the other — it's a reason to be deliberate about WHICH order you use for which question type, rather than defaulting to a single habit across every question type without considering the tradeoff.