Source
- Deci, Koestner & Ryan (1999), LIT-557, read in NOTE-438 (from an OCR of the authors’ deposit): the engagement-contingent analysis (p. 641), the timing analysis (p. 650), the interesting-versus-dull analysis (p. 651), and Appendix A’s comparison with Cameron and Pierce.
- Cameron, Banko & Pierce (2001), LIT-562, read in NOTE-587: the results for rewards offered for doing the task (pp. 16–18), durability (pp. 23–24) and magnitude (pp. 24–25).
What was actually shown
An engagement-contingent reward is one offered for working on the task, whatever the person completes or how well (p. 628). Across 55 experiments with a free-choice measure, such rewards lowered the time people later spent on the task when free to do anything (d = −0.40, CI −0.48 to −0.32). Children: −0.43 after four outliers (k = 39); college students: −0.21 (k = 12). On self-reported interest the effect was smaller, −0.15 (k = 35) (p. 641).
The effect is specific. Unexpected rewards (0.01) and rewards not tied to the task (−0.14, not significant) did not lower free choice (p. 640). With tasks experimenters selected as dull, rewards did not lower it (p. 651). Among children, it was present immediately (−0.40) and after more than a week (−0.53) (p. 650).
The dispute, as Deci et al. report it. Cameron and Pierce (1994, LIT-556) concluded that rewards do not, overall, harm intrinsic motivation. But by Deci et al.’s account they too found significant undermining of free choice by task-contingent rewards (−0.23 after outliers, p. 649), and Eisenberger and Cameron’s performance-independent category, which Deci et al. take to be equivalent to engagement-contingent rewards or nearly so, gave −0.29 on free choice (p. 641). The rival account even predicted this case (helplessness from rewards that do not depend on performance, p. 630). The disagreement in Appendix A is over coding, control groups, dull tasks and the other contingencies.
The other side, read. Cameron, Banko and Pierce’s 2001 reply (LIT-562) merges both databases (145 studies), adopts Deci et al.’s split by initial interest, and codes contingencies by procedure. Rewards offered for doing an interesting task, with no requirement to finish or do well, lowered free choice: K = 57, d = −0.35 (CI −0.43 to −0.27) before outliers, and still significantly negative once seven were removed (pp. 16–17). Self-reported interest fell too, −0.13, or −0.22 without imputed zeros (p. 18). Their estimate for rewards for doing or doing well is about −0.30, roughly 47 seconds less in an 8-minute free period (pp. 24–25). So the effect is found by the analysis built to show that rewards do not generally harm intrinsic motivation, and the dispute between the two sides is about everything around it.
What this does not say
- It does not say rewards in general undermine intrinsic motivation, nor that rewards tied to performance standards do. Those are disputed.
- It does not say why. Cognitive evaluation theory reads the effect as felt control (THEORY-049). Cameron, Banko and Pierce read it as a reward loosely tied to performance, which carries no competence information (THEORY-154). This account is the effect only.
- It does not say the effect is large or lasting in adults. Delayed measures exist only for children, and repeated-reward studies with controls are two, with nine rewarded participants in all (p. 650). Cameron, Banko and Pierce read the delayed measures the other way. Of seven studies of rewards for doing the task with a second free-choice measure weeks later, two showed undermining at the second. Five single-subject studies without controls showed none after reward ended (LIT-562, pp. 23–24). They count significant results rather than pooling them, so this is not the multi-session refutation the record would need. But it leaves durability open, not settled.
- It does not cover dull tasks or self-reported interest beyond the numbers given.
Connections
- THEORY-049 is the explanation SDT gives for this effect.
- THEORY-154 is the rival explanation, from the reply that confirmed the effect.
- Bandura (LIT-545, unread). The social-cognitive account the rival side draws on; the reply cites Bandura (1986), not this paper.