Source
Cameron, Banko & Pierce (2001), LIT-562, read in NOTE-587: the procedural coding (Table 1, p. 12), the results by contingency (pp. 16–20, Figures 1–4), the discussion (pp. 21–26) and Appendix B.
What was actually shown
Cameron, Banko and Pierce re-coded each study by what participants were told the reward was for, in seven categories, instead of in cognitive evaluation theory’s four. On interesting tasks, expected tangible rewards then split.
What undermines free choice (pp. 16–17):
- rewards offered for doing the task: K = 57, −0.35 before outliers, still negative after them;
- rewards offered for doing well, with no standard stated;
- rewards offered per unit solved, but only where some participants earned less than the maximum: −0.22 (K = 14), against −0.03 (K = 6) where all earned it in full (pp. 19–20).
What does not:
- rewards unrelated to the task;
- rewards for finishing;
- rewards for surpassing a score.
What enhances free choice: rewards for meeting or exceeding others’ performance. On self-reported interest, every contingency except task-noncontingent rewards and rewards for doing the task was positive (p. 18).
The authors’ summary is that “positive effects are obtained when the rewards are explicitly tied to performance standards and to success; negative effects are produced when rewards signify failure or are loosely tied to behavior” (p. 23). They read it through social cognitive theory: rewards tied to a standard the person meets raise self-efficacy, and with it interest and time on task (p. 26).
This could have come out otherwise. Cognitive evaluation theory predicts that performance-contingent rewards are on balance controlling and undermine. Had the score- and norm-based categories come out negative, as Deci, Koestner and Ryan’s pooled performance-contingent category did (NOTE-438), this account would have failed on the paper’s own data.
What this does not say
- It does not say rewards are harmless. Rewards for doing an interesting task undermine free choice on this analysis as on its rival’s. That effect is THEORY-046, and this account includes it as its “loosely tied” case.
- It does not show the mechanism. Self-efficacy and competence are inferred from which contingencies help; nothing in the paper measures them.
- It does not isolate the key contrast. Control groups told the performance objective and given feedback were pooled with those that were not (p. 13). The comparison that separates this account from CET, a performance-tied reward against a control given the same feedback, is reported only from Eisenberger, Pierce and Cameron (1999, LIT-754), which is filed but unread.
- It does not establish the size of the positive effects. The per-contingency values after outliers are in figures with no text layer, and the authors call every effect small (p. 25).
- It does not cover low-interest tasks, where rewards raised free choice on this analysis (0.28). That is a separate claim.
Connections
- Rivals THEORY-049. Cognitive evaluation theory explains the same reward effects by whether an event is felt as controlling or as information about competence, and a strict reading predicts that performance-contingent rewards undermine. The paper’s finding that “rewards specifically tied to level of performance (surpassing a score, exceeding others) do not undermine measures of intrinsic motivation is incompatible with the claims of these theories” (LIT-562, p. 26). The authors allow that CET could be modified to let competence information override control, so this is a rival rather than a correction. Neither account has the mediation test that would settle it.
- THEORY-046 is the effect both accounts explain, and on which both meta-analyses agree.