Charles S. Carver & Michael F. Scheier (1990), Psychological Review 97(1):19–35 — DOI-10.1037/0033-295X.97.1.19
Key takeaways
Registered, not read. What follows is what Carver’s own later restatement says of it (LIT-1449), not a reading of this paper.
- Affect is a rate signal. Affect is not the size of the discrepancy between a goal and the present state. It tracks how fast that discrepancy is being reduced. The 2018 restatement describes valence as neutral at a reference value, with everything “seen in terms of velocity of goal attainment rather than states”.
- Either system, either sign. An approach loop and an avoidance loop can each do well or badly, so each can produce positive or negative affect. The 2018 text cites “Carver & Scheier, 1990, 1998” for this.
- Not read here: the meta-monitoring loop as the 1990 paper states it, the reference rate and how it is set, and what the paper says about conflict between goals.
Standing in the record
Filed on 2026-10-23 for the record’s proposal that hedonic valence is a regime in which one ordering of a system’s own states prevails in the negotiation among its perspectives (CLAIM-440). That claim already names “Carver and Scheier’s rate-of-progress account” as prior art for its sign clause, and gives LIT-604 as the place it is registered. This is the paper that states the affect half of the account, so the claim’s citation should move here. LIT-604 is the general control-theory review.
Two points from the 2018 restatement cut against the proposal as it is worded. They should be checked against this paper once it is read:
- Neutral is not “no movement”. On Carver and Scheier’s account valence is neutral when progress matches a reference rate. A system standing still where it expects to advance should feel bad. The proposal’s “Neutral is potential with no movement” makes zero velocity the neutral point.
- Valence belongs to each goal loop, not to one ordering. Every loop that monitors its own rate of progress has an affect of its own. The proposal asks for one ordering, ranking states by their bearing on maintenance, to prevail.
No instruction for machine-learning practice; nothing here belongs in the anthology.