Source
Ryan, Mims and Koestner (1983), LIT-1970, JPSP 45, 736–750, pp. 744–748; read in NOTE-1522.
What was actually shown
Design (pp. 744–745). 96 students, randomly assigned, n = 16 per cell. Rewarded subjects were promised $3 “if you do well” (informational) or “if you perform up to our standards” (controlling). All feedback groups got the same positive feedback after each puzzle, with “just as you should” added in the controlling style. Unrewarded groups got the same feedback in the same two styles. A no-feedback, no-reward group and a no-feedback, task-contingent group completed the design.
Free choice, seconds of 360 (Table 1, p. 746): informational feedback, no reward 282.6, reward 229.4; controlling feedback, no reward 195.8, reward 117.3; no feedback, no reward 164.1, task-contingent 113.4.
Tests (ANCOVA, initial interest covariate). Rewarded vs feedback-matched unrewarded, F(1, 89) = 4.34, p < .05. Informational vs controlling performance-contingent reward, F(1, 89) = 4.25, p < .05. Controlling vs informational across the four feedback cells, F(1, 89) = 6.11, p < .02 (pp. 746–747).
Checked. From Table 1, with a pooled within-cell SD of 143.5 and no covariate: framing among the rewarded F ≈ 4.88 (d ≈ 0.78); reward vs matched controls F ≈ 3.37 (p ≈ .07; d ≈ −0.37 informational, −0.55 controlling). The reported tests agree with the table; the reward contrast needs the covariate to reach p < .05.
The manipulation reached felt pressure. Pressure–tension 3.61 controlling vs 2.91 informational, F(1, 90) = 9.19, p < .005 (p. 748).
This could have come out otherwise: an informationally given reward could have matched or beaten the unrewarded group with the same feedback, as the reviewed studies of Karniol and Ross and Rosenfield et al. found against positive-feedback controls (p. 741).
What this does not say
- That felt control mediates. No mediation analysis; the pressure effect is reported for the information–control dimension across cells.
- That it generalizes. One task, one session, 16 per cell, college students; the self-report interest measure did not reach significance (p. 748).
- Anything about rewards tied to a stated score or to beating others. The criterion was “do well”, unstated; THEORY-154’s distinction is not tested.
Connections
- THEORY-049. This is the primary study its promote_when names, and evidence for it: the same reward’s effect changes with its framing, and the clause’s refuting outcome (rewards tied to performance raising free choice against feedback-matched controls) did not occur. It is one study without mediation, so it does not promote THEORY-049.
- THEORY-154, only indirectly. Cameron, Banko and Pierce classify by what participants were told about the standard; this experiment held the standard vague and varied only the style.