Page 161 - Emotions through the eyes of our closest living relatives- Exploring attentional and behavioral mechanisms

Validation of the Pictorial Implicit Association Test
superordinate category images that belonged to the last critical block (block 5). For each of the affected participants, the reaction time of the trial was manually set to 10,001 ms to ensure that it would be discarded in subsequent analyses.
For adults, two trials were removed due to the 10,000 ms cutoff, and one participant was excluded due to a high error rate (Mean error in our sample Merror = 4.91, SD = 6.31). For children, 44 Trials were excluded based on the 10,000 ms criterion (15 trials within one child; the rest divided over 29 children), and one child was excluded based on the 40% error cutoff (Merror = 6.92, SD = 7.23). Remaining reaction time data were then 10% winsorized (Richetin et al., 2015). Note that we did not exclude erroneous trials; inclusion of erroneous trials increases validity and reliability of the scoring method (Richetin et al., 2015). The D-score represents the difference in reaction times (after processing) between critical blocks divided by the standard deviation of the datapoints in both critical blocks. Positive D-scores indicated an association between faces of men of Moroccan descent and negative images, and faces of men of Dutch descent and positive images.
For adults and children we performed two separate one sample t-tests to establish
whether D-scores significantly differed from 0. Furthermore, as presentation order of
the critical blocks and task version may affect D-scores, we also fit two separate linear
models using sum-to-zero coding. We used Congruency (i.e., congruent or incongruent
block first) and Task Version (Version 1, 2 or 3, reflecting the three different stimulus
sets) as fixed effects, with the intercept reflecting the average D-score in our sample.
Next, to assess the consistency of results across all items, we correlated results of the
first half of the trials within the critical blocks with the second half of the trials using 7 the SplitHalf function in IATScores.
Furthermore, the results of sensitivity power analyses for our main hypotheses can be found in the supplements (supplemental Figure S5).
Results
Adult PIAT
For adults we found a significantly positive D-score average of .24 (95% CI [.16, .32], t(127) = 5.94, p < .001, Cohen’s d = .52), indicating that adults respond significantly faster on congruent trials (“Moroccan” + negative and “Dutch” + positive) versus incongruent trials (“Moroccan” + positive and “Dutch” + negative; Figure 3 and Table 1). After controlling for Congruency and Task Version, this effect remained present
159