Supplemental Materials
Reward Favors the Prepared: Incentive and Informative Cues Interact to Enhance Attentional Control
by K. S. Chiew & T. S. Braver, 2015, JEP: Human Perception and Performance
http://dx.doi.org/10.1037/xhp0000129
Experiment 1: Speed-Accuracy Tradeoff Analyses
As described in Footnote 2 in the main body of the manuscript, we correlated changes in RTs vs. error rates for baseline and non-incentive trials (separately for each task-information condition) to examine for the presence of a speed-accuracy tradeoff (SAT) in Experiment 1. None of these correlation analyses were significant, suggesting against the possibility of a SAT. However, we conducted further analyses to clarify this possibility, which we outline here.
First, we calculated inverse efficiency scores (accuracy-corrected RTs; calculated by dividing median RTs by accuracy scores for each condition; (Bruyer & Brysbaert, 2011)) for performance in Experiment 1. We analyzed these measures as a function of incentive (block- or trial-based), task-information, and congruency following analyzes of RT and error rates outlined in Global Effects. When adjusted for accuracy, block-based incentive was no longer associated with a significant speeding of RTs [F(1,23) = 1.544, p = .226]; however, the main effect of trial-based incentive on RTs remained significant [F(1,23) = 16.803, p < .001], suggesting that trial-based incentive effects did not lead to a SAT.
To further confirm that trial-based incentive was not associated with a speed-accuracy tradeoff, we plotted conditional accuracy functions (Heitz, 2014) specifically for incongruent trials in Experiment 1 (see Figure S1). For these functions, we divided trials in each experimental condition into quintiles by increasing RT, and then plotted mean RT vs. mean accuracy rate (proportion correct) for each quintile. These plots reveal that the function for incentive trials within the reward block is faster (i.e., shifted towards the left, to decreased RTs) without a decrease in accuracy, relative to both baseline and non-incentive trials within reward block, while non-incentive trials show both RT speeding and a decrease in accuracy relative to baseline trials. The conditional accuracy function plots are consistent with analyses of inverse efficiency measures in indicating that a SAT did not occur as a function of trial-based incentive.
Experiment 2: Speed-Accuracy Tradeoff Analyses
Following supplementary analyses examining the possibility of speed-accuracy tradeoff in Experiment 1 data, we also conducted analyses on inverse efficiency measures and plotted conditional accuracy functions for data in Experiment 2.
Inverse efficiency measures for Experiment 2 data were calculated in the same fashion as in Experiment 1 and analyzed as a function of cue timing (Early vs. Late incentive), incentive (block- or trial-based), task-information, and congruency. When adjusted for accuracy, block-based incentive was no longer associated with a significant speeding of RTs [F(1,23 = 2.303, p = .143], but the main effect of trial-based incentive on RTs remained significant [F(1,23) = 18.751, p < .001], suggesting that as in Experiment 1, trial-based incentive effects in Experiment 2 did not lead to a SAT.
Conditional accuracy functions were plotted separately for Early and Late Incentive conditions in Experiment 2 (refer to Figure S2). As in Experiment 1, these plots indicate that RTs in incentive trials were faster than (in the Early Incentive condition) or equivalent to (in the Late Incentive condition) RTs in non-incentive trials without a decrease in accuracy, again indicating that a SAT did not occur as a function of trial-based incentive.
Given that our primary effects of interest in both Experiment 1 and 2 occurred at the trial level, we do not believe that these analyses change overall interpretation of the data, although they do suggest that further investigation of speed-accuracy tradeoffs as a function of incentive and other preparatory factors is warranted.
Figure S1. Conditional accuracy functions (proportion correct vs. mean RT for RT quintiles) for incongruent trials in Experiment 1.
Figure S2. Conditional accuracy functions (proportion correct vs. mean RT for RT quintiles) for incongruent trials in Experiment 2 (a) Early Incentive condition. (b) Late Incentive condition.