For the analyses of average response times (RTs), incorrect and responses faster than 200 ms (< 1% in both experiments) were discarded.
Experiment 1
Attentional capture
To demonstrate that the distractor interfered with target search and, crucially, that the interference varied with its salience, we first submitted RT data to an ANOVA with distractor presence (absent vs. high-salience distractor vs. low-salience distractor) as factor. This effect was reliable (F(2,46)=57.87, p<.001, ηp2=.716). Planned comparisons revealed that both distractors interfered with target search (absent vs. high-salience distractor: M=825 ms ± SD=92 vs. 856 ms ± 95, t(23)=10.437, p<.001, Cohen’s d=2.130; absent vs. low-salience distractor: 825 ms ± 92 vs. 844 ms ± 92 t(23)=7.256, p<.001, d=1.481). Crucially, the high-salience distractor caused larger interference than the low-salience distractor (t(23)=3.785, p<.001, d=.773). Similar results were obtained for error rates with all effects being congruent with (i.e., the same direction as) the RT effects (p<.01), which excludes an alternative explanation in terms of a speed-accuracy trade-off. This demonstrates that our salience manipulation was successful, suggesting that while both distractors captured attention, there was more capture by the high- than by the low-salience distractor.
Attentional suppression
To investigate the impact of our probability manipulation on the distractor’s interference, we assessed whether search times differed depending on where any of the distractors appeared. Relative to distractor-absent trials, RTs were significantly slower when a distractor appeared in the high-probability location of the high-salience distractor (t(23)=3.071, p=.005, d=1.060) or the low-salience distractor (t(23)=6.953, p<.001, d=1.419; see Table 1 for RT and error rates), or one of the low-probability locations (t(23)=12.854, p<.001, d=2.624; Fig. 2, left). Both high-probability distractor locations showed evidence for suppression since target search was significantly faster when a distractor appeared in one of the high-probability locations relative to when it appeared in a low-probability location (vs. high-salience distractor location: t(23)=10.532, p<.001, d=2.150; vs. low-salience distractor location: t(23)=5.337, p<.001, d=1.089). Crucially however, target search was even faster when a distractor appeared in the high-probability location of the high-salience distractor relative to when it appeared in the high-probability location of the low-salience distractor (t(23)=2.656, p=.014, d=.542). There was no evidence for a speed-accuracy trade-off (all comparisons either p>.1 or congruent with the RT effects when significant). These findings suggest that the high-probability distractor locations were suppressed differently: the high-probability location of the high-salience distractor was suppressed more strongly than the high-probability location of the low-salience distractor.
Table 1 Mean response times (RTs) and error rates of both experiments Next, we examined whether the pattern of suppression differed depending on which of the two distractors was present. An ANOVA on mean RT distractor salience (high vs. low) and distractor position (high-probability location of high-salience distractor vs. high-probability location of low-salience distractor vs. low-probability location) as factors showed a main effect of distractor salience (F(2,23)=22.999, p<.001, ηp2=.500), distractor position (F(2,46)=42.187, p<.001, ηp2=.647), as well as a significant interaction (F(2,46)=5.459, p=.007, ηp2=.192). In line with our previous analysis, presenting either distractor in either its high-probability location or the high-probability location of the other distractor significantly reduced its interference with target search relative to when it appeared in a low-probability location (all comparisons p<0.001; Fig. 2, right). However, only when the high-salience distractor appeared in its specific high-probability location, was interference in target search even further reduced (high-salience distractor trials: high-probability location of high-salience distractor vs. high-probability location of low-salience distractor: t(23)=3.665, p=.001, d=.748). No such difference was observed for the low-salience distractor (low-salience distractor trials: high-probability location of high-salience distractor vs. high-probability location of low-salience distractor: t(23)=.180, p=.859). Although interference by the two distractors did not differ for the high-probability location of the high-salience distractor (t(23)=.879, p=.389), interference by the low-salience distractor when it appeared at this location was so far reduced as to be statistically indistinguishable from search performance in distractor-absent trials (t(23)=1.503, p=.146). Finally, there was no evidence for a speed-accuracy trade-off as all comparisons were either non-significant (p>.1) or congruent with RT effects when significant.
Intertrial priming
To assess whether suppression can be explained in terms of short-lived, intertrial location priming instead of learning about the statistical regularities regarding the distractor positions, we compared trials in which the distractor position of a given trial was identical to the previous trial with trials in which it had changed (same position vs. different position). Indeed, there was evidence for intertrial location priming as RT (t(23)=8.006, p<.001, d=1.634) and error rate (t(23)=3.352, p=.003, d=1.124) in same-position trials was lower than in different-position trials. Importantly though, intertrial location priming could neither explain the overall suppression effects nor the differences in suppression between the two distractors in particular. After excluding all trials in which the distractor position was repeated between trials (i.e., same-position trials), an ANOVA on RT with distractor salience (high vs. low) and distractor position (high-probability location of high-salience distractor vs. high-probability location of low-salience distractor vs. low-probability location) replicated all major findings. There was no evidence for a speed accuracy trade-off (all comparisons non-significant (ns) at p>.1 or congruent with the RT effects).
Using a similar approach, we also investigated whether the suppression effects can be explained in terms of intertrial feature priming. To this end, we compared trials in which the distractor type of a given trial was identical to the distractor type in the previous trial (same-distractor vs. different-distractor type). This difference reached significance for RT (t(23)=2.227, p=.036, d=.455) but not for error rate (t(23)=1.023, p=.317, d=.209). Importantly though, after excluding all trials in which the distractor type repeated between trials, the ANOVA replicated all major findings, which shows that intertrial feature priming had no influence on the observed suppression effects. There was no evidence for a speed accuracy trade-off (all error rate comparisons p>.1).
Spatial gradient of suppression
We also analyzed the spatial profile of suppression by assessing how interference by the distractor changed as a function of the distractor’s distance to the high-probability distractor locations (indexed by the number of positions). An ANOVA on RT with distractor salience (high vs. low) and distance of the distractor to the high-probability location matching its salience (from distance 0, or high-probability location salience match, to distance 4, or high-probability location salience mismatch) as factors showed a main effect of distractor salience (F(1,23)=30.647, p<.001, ηp2=.571) and distance (F(4,92)=22.969, p<.001, ηp2=.500) as well as a significant interaction (F(4,92)=2.515, p=.047, ηp2=.099). Figure 3 (right panel) shows a clear spatial gradient pattern for both distractor types mirroring the key features of the previous analyses. The spatial gradient for the distractors followed a quadratic trend (t(23)=9-234, p<.001), albeit the gradient for the high-salience distractor showed a steeper rise from its salience-matching high-probability location. There was no evidence of a speed-accuracy trade-off in the error rates (all comparisons p>.1 or congruent with the RT effects). These results provide strong evidence for a spatial gradient in the suppression of high-probability distractor locations: the closer a given distractor to a high-probability location, the stronger the suppression.
Awareness
Out of 24 participants, six indicated in the implicit learning questionnaire that they had noticed some relationship. Two participants correctly identified both high-probability locations (chance level ~1.6% or zero participants), although these two participants had indicated that they had noticed no relationship. Removing those two participants had no significant influence on the main findings.
Experiment 2
Attentional capture
An ANOVA on RT with distractor presence (absent vs. high-salience distractor vs. low-salience distractor) as factor showed a significant main effect (F(2,46)=56.24, p<.001, ηp2=.710). Both distractors interfered with target search (absent vs. high-salience distractor: 726 ms ± 104 vs. 769 ms ± 111, t(23)=8.367, p<.001, d=1.708; absent vs. low-salience distractor: 726 ms ± 104 vs. 753 ms ± 110, t(23)=7.890, p<.001, d=1.611), but interference by the high-salience distractor was stronger (t(23)=4.564, p<.001, d=.932). There were no significant differences in error rate, which excludes the possibility of a speed-accuracy trade-off (all p>.1). This suggests that the distractor captured attention and that capture by the high-salience distractor was larger.
Attentional suppression
The distractor interfered significantly with target search no matter where it appeared (Fig. 4, left; absent vs. high-probability location of high-salience distractor, t(23)=4.913, p<.001, d=1.003, vs. high-probability location of low-salience distractor, t(23)=7.931, p<.001, d=1.619, vs. low-probability location, t(23)=10.632, p<.001, d=2.170; see Table 1 for RT and error rates). As expected, when the distractor appeared in either high-probability location it interfered less compared to when it appeared in a low-probability location (vs. high-salience distractor location: t(23)=8.380, p<.001, d=1.711; vs. low-salience distractor location: t(23)=4.771, p<.001, d=.974). When any distractor appeared in the high-probability location of the high-salience distractor, interference by the distractor was even further reduced (high vs. low-salience distractor location: t(23)=2.160, p=.041, d=.441). All comparisons on error rate were either non-significant (p>.1) or congruent with RT effects. Similar to Experiment 1, these findings suggest that suppression was different for the two high-probability distractor locations. The high-probability location containing the high-salience distractor was suppressed more strongly than the high-probability location containing the low-salience distractor.
An ANOVA on RT with distractor salience (high vs. low) and distractor position (high-probability location of high-salience distractor vs. high-probability location of low-salience distractor vs. low-probability location) showed a main effect of distractor salience (F(2,23)=40.402, p<.001, ηp2=.637), distractor position (F(2,46)=29.159, p<.001, ηp2=.559), and a significant interaction (F(2,46)=4.742, p=.025, ηp2=.171). Confirming the previous analysis, when a distractor appeared either in its high-probability location or the high-probability location of the other distractor, its interference on target search was significantly reduced (all comparisons p<.01; Fig. 4, right). However, similar to Experiment 1, only when the high-salience distractor appeared in its high-probability location was interference even further reduced (high-salience distractor trials: high-probability location of high-salience distractor vs. high-probability location of low-salience distractor: t(23)=3.254, p=.003, d=.664). This was not the case for the low-salience distractor (low-salience distractor trials: high-probability location of high-salience distractor vs. high-probability location of low-salience distractor: t(23)=.162, p=.872). Interference at the high-probability location of the high-salience distractor did not differ between the two distractors (t(23)=1.313, p=.202), but target search was significantly impaired relative to distractor-absent trials for both distractors (both p<.01). Finally, there was no evidence for a speed-accuracy trade-off (all comparisons either ns at p>.1 or congruent with RT effects at p<.05).
Intertrial priming
Similar to Experiment 1, we also found evidence for intertrial location priming when comparing RT (t(23)=8.823, p<.001, d=1.801) and error rate (t(23)=2.198, p=.038, d=.449) of same-position against different-position trials. Nonetheless, after removing priming (i.e., same-position) trials, an ANOVA on RT with distractor salience (high vs. low) and distractor position (high-probability location salience match vs. high-probability location salience mismatch vs. low-probability location) replicated all previous effects and showed no evidence of a speed-accuracy trade-off (all error rate comparisons p>.1).
In Experiment 2, we did not find any evidence for intertrial feature priming when comparing either RT (same vs. different distractor type: t(23)=1.615, p=.12, d=.33) or error rate (t(23)=.585, p=.564, d=.119). Removing trials in which the distractor type repeated between trials also had no influence on the major findings of this experiment, as revealed by the ANOVA on RT with distractor salience and distractor position as factors. There was also no evidence for a speed-accuracy trade-off (all error rate comparisons p>.1).
Spatial gradient of suppression
The analysis on the spatial gradient of suppression showed a main effect of distractor salience (F(1,23)=49.936, p<.001, ηp2=.685) and distance (F(4,92)=18.647, p<.001, ηp2=.448) as well as a marginally significant interaction (F(4,92)=2.198, p=.075, ηp2=.087). There was a spatial gradient pattern for both distractor types that mirrored the key aspects of the previous analyses (Fig. 3, right). Similar to Experiment 1, the spatial gradient for the distractors followed a quadratic trend (t(23)=8.145, p<.001) with yet again a steeper rise for the high-salience distractor condition from the distractor’s salience-matching high-probability location. There were no significant differences in error rates (all p>.1). In short, there is clear evidence for a spatial gradient of suppression for the two high-probability distractor locations with suppression centered around the high-probability locations and gradually falling off the further away from those locations.
Awareness
Out of 24 participants, ten indicated in the implicit learning questionnaire that they had noticed some relationship. However, not a single participant correctly identified both high-probability locations (chance level ~1.6% or zero participants).