Contribution
The first experimental demonstration, the authors say, of both halves of the control-process prediction for running ahead: less on the goal that is ahead (coasting) and more on a competing goal (shifting). Goal progress was manipulated by false feedback, and performance, not self-report, was the outcome, which answers the common-method objection to Fulford et al. (2010). A second experiment adds a manipulation aimed at affect.
Key insight
With two goals that trade off, news that one is ahead moves performance to the other, and the move seems to run through feeling: tell people to ignore their feelings and it stops. Being ahead does not just relax effort; it reorders which goal performance serves.
Assumptions
- The two-loop model (Introduction): a rate loop whose output “is affective in nature: Negative affect is proposed to occur when progress is slower than expected or needed and positive affect when progress is faster than expected, needed, or desired … Negative affect then leads to investing more effort (pushing) and positive affect leads to investing less effort (coasting).”
- Carver 2003 for shifting (Introduction): “Carver (2003) argued that people generally pursue many goals at once, even with respect to a single task. He suggested that fast goal progress toward one goal can lead to an openness toward alternative goals where progress is slower (shifting).” And (The Present Research): “Carver (2003 …) suggests that coasting may occur mainly when one can shift to another goal.”
- The task’s two goals are negatively interdependent (speed–accuracy trade-off), and participants were told so (General Discussion).
- Performance stands for effort. Speed and accuracy were measured, not effort (General Discussion, first limitation).
Key results
- Experiment 1 design. 151 MTurk workers (of 173), 2 progress (overshoot, undershoot) × 2 dimension (speed, accuracy) plus an on-target control. 200 lexical-decision trials with a target of 120 fast (< 550 ms) and 120 correct responses for a 50-cent bonus; false bar-graph feedback after 100 trials (90 against 30 against 60 on the manipulated dimension, 60 on the other). Mood (−3 to 3) and planned effort (0–6) rated before and after feedback.
- Experiment 1, accuracy. The planned contrast on Round 2 errors, covarying Round 1, F(1,145) = 15.67, p < .001, ηp² = .098. Overshoot on accuracy gave more errors than undershoot on accuracy (coasting), F = 7.01, p = .005, ηp² = .046; overshoot on speed gave fewer errors than undershoot on speed (shifting), F = 11.85, p < .001, ηp² = .076.
- Experiment 1, speed. Contrast F(1,145) = 7.29, p = .004, ηp² = .048. Overshoot on speed gave slower Round 2 responses (coasting), F = 3.27, p = .037, ηp² = .022; overshoot on accuracy gave faster ones (shifting), not significant, F = 2.02, p = .079. The control condition’s regression slope differed; without it the contrast held, F(1,118) = 5.07, p = .013.
- Experiment 1, mood and effort. Overshoot left mood higher than undershoot at Round 2, covarying Round 1, T = −4.11, p < .001. In the raw means (Table 1) the change was −0.09 and −0.06 after overshoot, −0.96 and −0.69 after undershoot, and −0.04 in the control: being told one was behind lowered mood, and being told one was ahead did not raise it. Self-reported effort did not differ (p = .161), near ceiling (about 5.7 of 6) throughout.
- Experiment 2 design. 222 CMU students, 2 progress (accuracy overshoot, undershoot) × 2 plan. Before feedback, an emotion-regulation plan (“When I’m working on the word recognition task, then I will ignore my feelings”) or a control plan (”… I will ignore the pictures”), each typed three times after viewing Rothko paintings. Speed feedback on target in both. No mood or effort ratings.
- Experiment 2, coasting. Contrast (weights −1, 3, −1, −1) on Round 2 errors, F(1,217) = 5.80, p = .009, ηp² = .026. With the control plan, overshoot gave more errors than undershoot, F = 5.89, p = .008; with the emotion-regulation plan, F = 1.10, p = .148.
- Experiment 2, shifting. Contrast on Round 2 reaction times, F(1,217) = 18.92, p < .001, ηp² = .080. With the control plan, overshoot on accuracy gave faster responses, F = 22.61, p < .001, ηp² = .094; with the emotion-regulation plan the difference was “smaller and non-significant”, F = 2.67, p = .052.
- Interpretation (General Discussion). “The present two studies thus provide experimental evidence of the coasting and the shifting hypotheses, as well as the role that affect plays in their occurrence.” Shifting is called adaptive. The authors explain the discrepancy with Gollwitzer and Rohloff, who found no coasting with a single goal, by the secondary goal: quoting them, “only when other important goals have to be served at the same time might one observe a decrease in effort”.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Feedback that a goal is ahead of target lowers later performance on it (coasting) | moderate | Exp. 1 accuracy and speed, Exp. 2 accuracy; ηp² .022–.046 for the simple effects, one-tailed |
| C2 | It raises performance on a competing goal (shifting) | moderate | Exp. 1 accuracy (p < .001), Exp. 2 speed (p < .001); Exp. 1 speed not significant |
| C3 | Ahead-of-target feedback produces better mood than behind-target feedback | moderate as a contrast | Exp. 1; but the raw means show the difference comes from undershoot lowering mood, not overshoot raising it |
| C4 | Affect is causal in coasting and shifting | weak | Exp. 2: an “ignore my feelings” plan removes the effects; no manipulation check, and the plan was not tested against the no-plan simple effects directly |
| C5 | Coasting needs a second goal to shift to | weak | reconciliation with Gollwitzer and Rohloff, not tested within these studies |
Concepts
- Coasting — “unexpectedly fast goal progress leads to reduced effort at that goal” (Abstract); measured here as lower performance on that goal.
- Shifting — “shifting focus toward other goals” (Abstract); measured as higher performance on the alternative goal.
- Pushing — more effort after slower-than-expected progress.
Connections
It extends Fulford et al. (2010, LIT-1860), which it cites as the “only one published study” testing the hypotheses directly, and Louro, Pieters and Zeelenberg (2007). It takes Carver and Scheier’s control-process model (1990, 1998; LIT-1441, read in the record through LIT-1449 and NOTE-1102) and Carver 2003 (LIT-1445) as its theory. Phan and Beck (in press) is cited for feedback’s effect on affect. Carver died while it was under revision (acknowledgements).
Bearing on the record
- CLAIM-468: partly. The objection’s positive half needs evidence that positive affect lowers a goal’s priority, and waits on Carver 2003 for it. This paper states the thesis’s shifting form from Carver 2003 and tests it: running ahead lowers the overshot goal’s hold on performance and moves performance to the rival goal, and a plan to ignore feelings removes both, which puts affect in the path. Two cautions keep it to partly. Exp. 1’s affect difference is negative affect after undershoot, not positive affect after overshoot: in Table 1 mood changed by −0.09 and −0.06 after overshoot, against −0.04 in the on-target control, so being told one was ahead did not raise it; and Exp. 2 shows that an emotion-regulation plan matters, not that positive affect was present and was what it regulated. Carver 2003 itself stays unread, so the claim’s attribution to that paper should now cite the reading of this one.
- CLAIM-440: partly, against clause 3 for positive valence. Clause 3 makes intensity how far the negotiation’s decisions follow the prevailing ordering. Here, the state the theory calls positive goes with the overshot goal ceding performance to a rival, which is the difficulty the claim records against itself. Whether that is a weaker hold of the prevailing ordering, or that ordering moving resources to the goal now furthest behind, is for the claim to decide: on Carver’s satisficing reading the shift serves the goals as a set, which the second reading fits. The paper gives no graded measure of affect, so it says nothing about intensity as such, only sign.
- No instruction for machine-learning practice; nothing for the anthology.
Limitations
- Performance, not effort, and self-rated effort did not move (General Discussion; Table 1).
- Demand. Participants were told the goals trade off (General Discussion).
- Small effects, one-tailed planned contrasts, and one follow-up contrast not significant in each experiment.
- No manipulation check in Experiment 2. Mood was not measured, so the plan’s effect on affect is assumed from prior work.
- Exp. 1’s control condition had a different regression slope, possibly because “on target” was read as good by some and bad by others.
- Online samples and a laboratory task; the supplementary analyses were not read here.
Open questions
- Does overshoot raise positive affect above baseline, and does that rise, graded, predict the size of coasting and shifting? Exp. 1’s means suggest the affect contrast is driven by the undershoot side.
- Would the effects survive without telling participants the goals trade off, and with a direct measure of effort?