Feldman, J. M. & Lynch, J. G. (1988). Journal of Applied Psychology 73(3):421–435. doi:10.1037/0021-9010.73.3.421.
Key takeaways
From the reading, NOTE-957.
- Measurement creates cognitions. “If the researcher asks a question about attitude (or belief, intention, etc.), and no such cognition already exists, the measure creates one” (p. 431). Evidence cited: Fazio, Lenn and Effrein (1984), where filling in an attitude measure made later evaluative answers faster, read as the formation of “retrievable attitudes” (p. 423).
- The created answer is carried forward. “Once the answer has been computed, it can be retrieved to answer subsequent questions” (p. 422), and it can drive behaviour: the survey brings “normally automatic behavior under conscious control” (p. 423).
- Which earlier answer is used. Perceived diagnosticity (asymmetric, itself a judgement), accessibility (time, interference, rehearsal) and the accessibility of more diagnostic alternatives (pp. 424–430). Reanalysed data: Schuman and Ludwig’s order effects, 90% against 66% and 36% against 73% (p. 425).
- Moderators. Direct experience, expertise and trait involvement, which supply formed and rehearsed judgements; experts lose the protection when choosing between non-comparable options (p. 430).
- Scope. Theoretical synthesis, “necessarily … speculative” by its own account (p. 432). Mechanisms all classical. The authors hold that measurement is one case of a general fact: “all real-world experiences with an object have the potential to shape subsequent responses to it” (p. 431).
Why this is here
Registered for CLAIM-341 as the empirical literature the claim lacked. The reading confirms that the paper states the claim’s diachronic form, an earlier answer becoming input to a later one, and that the arrow it draws runs from the measurement context into the interpreter’s state, the route TERM-039 and CLAIM-127 leave open. It also proposes a sequential, diagnosticity-based test that would separate that route from direct influence on the response (p. 425).
Three qualifications from the reading. The paper reports no data of its own; its support is a synthesis of others’ studies. The created state it cares most about is one that shapes behaviour, so it does not support the claim’s exclusion of action from the loop, and it gives judgement no special place among the experiences that shape later responses. And its moderators are not evidence routed around the interpretation, as CLAIM-342’s anchor is: they are prior, rehearsed judgements, formed, on the paper’s account, through direct experience and environments with strong outcome contingencies. That bears on CLAIM-339 and QUESTION-121 as much as on CLAIM-342. It carries no instruction for machine-learning practice.