To control for outliers, we applied a standard trimming procedure similar to the one suggested by Chang et al. (
2011). In a first step, for each participant, all reproduced durations that were more than ±2
SDs from that participant's mean reproduced duration for a given experimental condition were considered invalid trials and, therefore, not included in further data analysis. By applying this criterion, 4.4% of all trials were removed from data analysis. Next, each participant's remaining reproduced durations were submitted to a one-way ANOVA with target intervals (800, 1000, and 1200 ms) as three levels of a repeated-measures factor. The lack of a significant main effect as well as any nonsignificant differences among the three factor levels would provide an indication of an individual's inability to follow the instruction to reproduce the target intervals. None of our participants, however, had to be excluded on the basis of this criterion.
Analysis of error rates on the two versions of the stimulus-attribute relevance condition yielded faultless performance with error rates of 0.00. This outcome indicated that all participants conformed to the instructions and directed their attention to either stimulus size or numerical value depending on task requirements. Means of reproduced durations for each target interval and across all target intervals as a function of stimulus size, numerical value, and stimulus attribute relevance are given in
Table 1.
Four-way ANOVA revealed statistically significant main effects of target duration, F(2, 58) = 206.94, p < 0.001, ηp2 = 0.877, and stimulus size, F(1, 29) = 42.47, p < 0.001, ηp2 = 0.594. The significant main effect of target duration indicated longer MRDs with increasing duration of the target intervals. Subsequent post hoc tests revealed that MRDs of all three target intervals differed significantly from each other (p < 0.001). Furthermore, MRDs differed significantly as a function of stimulus size. The significant main effect of stimulus size on MRD indicated that digits presented at a large image size were reproduced longer than digits presented at a small image size; MRDs (± SD) were 925 ± 143 ms and 975 ± 151 ms for the small and the large stimulus size, respectively. With respect to numerical digit value, there was no statistically significant main effect of numerical value, F(1, 29) = 0.32, p = 0.58, ηp2 = 0.011; MRDs were 949 ± 145 ms and 952 ± 147 ms for digits of low and high numerical magnitude, respectively. Also no main effect of stimulus attribute relevance on MRD could be established, F(1, 29) = 0.10, p = 0.75, ηp2 = 0.004; MRD was 952 ± 147 ms when participants' attention was directed to stimulus size and 949 ± 150 ms when participants' attention was directed to the numerical value of the digits presented during the target interval.
A statistically significant interaction between target duration and stimulus size,
F(2, 58) = 3.50,
p < 0.05,
ηp2 = 0.108, indicated that, although large stimuli were reproduced longer than small stimuli for all target durations, this effect was much stronger for the 1000- and 1200-ms target durations compared to the 800-ms target duration (see
Figure 2).
Also the interaction between numerical value and stimulus attribute relevance became statistically significant,
F(1, 29) = 4.79,
p < 0.05,
ηp2 = 0.142. Although post hoc Tukey tests failed to show any significant differences, this interaction suggested that, when participants' attention was focused on numerical digit value rather than on physical digit size, they tended to perceive digits of a large numerical magnitude to last longer than digits of smaller numerical magnitude presented for the same duration; MRDs were 942 ± 150 ms for low and 955 ± 152 ms for high numerical values, respectively. However, when participants were instructed to pay attention to physical digit size, MRDs tended to be longer for digits with low (956 ± 147 ms) than for digits with high (948 ± 149 ms) numerical value (see
Figure 3). No other interactions reached the 5% level of statistical significance.