Why it is a question
THEORY-335 holds that the selective-accessibility account fits the moderations Strack and Mussweiler found, and that its one direct test, repeated five times by Harris et al., found nothing with tasks that also missed direct picture priming. So the accessibility step is inferred, not measured. CLAIM-315 asserts it, and CLAIM-298 rests on CLAIM-315.
What would answer it
THEORY-335’s former promote_when, unchanged:
A design whose measure of accessibility is first shown sensitive, by detecting direct priming of the same knowledge in the same setting, used between the comparison and the estimate with anchor-consistent and anchor-inconsistent knowledge measured separately. It promotes the account if anchor-consistent knowledge is measurably more accessible after the comparison and that difference goes with the size of the estimate shift; it refutes the accessibility step if the sensitive measure shows no change while anchoring is present. A design that separates a recalled candidate answer from accessible knowledge as the carrier also bears on it. More moderation by applicability or plausibility, latency patterns, coded free reports, or another lexical-decision null with an insensitive task cannot settle it: they infer accessibility, or cannot detect it.