Source
- Deci, Koestner & Ryan (1999), LIT-557, read in NOTE-438: CET’s predictions (pp. 628–630), the primary meta-analyses, and the supplemental analysis of informational versus controlling feedback (pp. 652–653).
- Ryan & Deci (2000), LIT-559, read in NOTE-442: CET stated (pp. 70–71).
- Deci & Ryan (2000), LIT-558, read in NOTE-439: “perceived competence is necessary for any type of motivation, perceived autonomy is required for the motivation to be intrinsic” (p. 234).
What was actually shown
The pattern fits. CET predicts that expected rewards contingent on the task undermine intrinsic motivation, that unexpected and task-noncontingent rewards do not, and that positive feedback enhances it. Deci et al. found exactly that across 128 experiments (LIT-557, pp. 653–654; see THEORY-046 for the core effect). Their reading: “the pattern of effects in our meta-analyses provided very strong support for CET in that the effects were virtually all as predicted” (p. 653).
The direct test is small but large in effect. In four studies that gave the same positive feedback in an informational or a controlling way (the controlling version adding, for example, “just as you should”), controlling feedback lowered intrinsic motivation relative to informational, d = −0.78; informational feedback raised it against no feedback (0.66) and controlling feedback lowered it (−0.44) (p. 652). One study did the same with tangible rewards and found the same pattern, not significantly (Ryan et al. 1983, p. 653).
Other events. Threats, deadlines, surveillance, evaluation and imposed goals undermine; choice and acknowledgement of feelings enhance (LIT-559, p. 70; LIT-558, p. 233). Competence feedback enhances only when people feel responsible for their success or it does not “eclipse” their autonomy (LIT-558, p. 234).
What this does not say
- It does not say mediation is shown. “Less well researched has been the concept of perceived self-determination … as a mediator of reward effects … a similar analysis has not been done with respect to reward effects” (LIT-557, pp. 653–654).
- It does not refute the rival account by itself. The behavioural account Deci et al. test (Eisenberger and Cameron’s helplessness and learned industriousness) fails on their data (pp. 654–655). The later account of Cameron, Banko and Pierce (LIT-562, read in NOTE-587; THEORY-154) is different. Re-coding the studies by what participants were told, they find that rewards for surpassing a score do not lower free choice and rewards for exceeding others raise it, where Deci et al.’s pooled performance-contingent category undermined (pp. 16–17). That pattern is the one this account’s refutation clause describes. But their analysis pools control groups given performance feedback with control groups not given it (p. 13), and the clause needs the first. The authors themselves allow that this account could fit if competence information overrides control (p. 26). So it is a rival, not yet a refutation.
- It does not explain the age effects. Positive feedback did not raise children’s free choice, and CET did not predict that (p. 653).
Connections
- THEORY-046 is the effect this account explains.
- Rivals THEORY-154. Cameron, Banko and Pierce explain the same effects by how tightly the stated contingency ties a reward to performance and success. The two accounts agree on rewards for merely doing a task. They part on rewards tied to a standard, which a strict reading of this account says undermine and LIT-562 finds neutral or positive.
- THEORY-054. The same notion of felt autonomy, applied to intrinsic motivation, where that account applies it to internalized regulation.
- Bandura (LIT-545, unread). Self-efficacy is SDT’s perceived competence; CET holds it necessary but not sufficient (LIT-558, p. 234). The rival side explains reward effects through self-efficacy.
Note of 2026-11-03: Ryan, Mims and Koestner 1983, read
- A correction. This entry says, citing LIT-557 p. 653, that Ryan et al. 1983 “found the same pattern, not significantly”. The primary paper (LIT-1970, NOTE-1522, p. 747) reports the informational and controlling performance-contingent groups as significantly different (229 against 117 s of 360; F(1, 89) = 4.25, p < .05). LIT-557’s figure should be checked for which contrast it computed.
- The refutation test. The study has the feedback-matched no-reward control the refutation clause needs. Rewards undermined in both styles (F = 4.34 with the covariate, about p = .07 without), so the refuting outcome did not occur. No mediation test was run, so the promote_when is not met.
- The person’s side. The causality orientations scale (LIT-1955, NOTE-1531) makes an event’s functional significance partly a fact about the person, so a test should measure orientation or randomize it away.
The status is unchanged.