Experiment 1 in the context of the previous studies suggests that biasing the presentation of a salient distractor towards a specific location is necessary to induce spatial distractor suppression. However, it is still possible that such a manipulation alone is not sufficient for distractor suppression to occur. Indeed, it may be that only by combining a bias in target presentation away from one location and a bias in distractor presentation towards the same location, this location becomes suppressed such that both target selection is impaired and capture by the distractor is reduced at this location.
To investigate whether biasing distractor presentation alone is sufficient to elicit spatial suppression, we employed the same manipulation as Wang and Theeuwes (2018a). That is, the distractor was presented much more often in one location than in all other locations in Experiment 2. Crucially, however, we also ensured that the target was equally likely to appear in each location in both distractor-present and -absent trials. If presenting the distractor singleton much more often in one location than in all other locations results in a suppression of that location even when the target is equally likely to appear in any of the other locations, we would expect to find the same results as Wang and Theeuwes (2018a). If differences in the probability of presenting the target were critical for observing the reduced capture effect in the previous studies, we would expect no suppression of the location that is likely to contain the distractor.
Methods
Another set of 24 healthy adults (14 females, age M=23.9) participated in Experiment 2. The stimuli, procedure, and experimental design of this experiment were similar to Experiment 1 except that we dropped the first awareness questions. That is, all participants had to indicate a regularity at the end of the experiment whether they reported to be aware of any or not.
Another, yet critical, difference to Experiment 1 was a change in the manipulation of target and distractor probabilities. In general, the conditions regarding the spatial position of the target and the distractor were more akin to Wang and Theeuwes (2018a) in that the distractor was more likely to appear in one specific location. In distractor-present trials of Experiment 2, the distractor appeared in 65% of the trials, or ~43.3% of all trials in the experiment, in one specific location (“high-probability distractor location”; counterbalanced across participants) and in 5% of the trials, or ~3.3% of all trials, in each of the seven remaining locations (“low-probability distractor locations”). Crucially different to Wang and Theeuwes (2018a), however, the target was equally likely to appear in any location in distractor-present as well as distractor-absent trials. That is, the target appeared in 12.5% of all distractor-present trials and 12.5% of all distractor-absent trials in any given location.
Results
Applying the same data exclusion criteria as in Experiment 1 led to the exclusion of around 3.5% of the data from the analyses.
Attentional capture effect
In order to assess whether spatial suppression for a high-probability distractor location is observed when the target is equally likely to appear in each location, RT and error rate data were first sorted according to the location of the distractor. An ANOVA on mean RT with distractor position (high-probability distractor location, low-probability distractor location, and distractor absent) as within-subject factor showed a main effect for distractor position, F(2,46)=58.82, p<.001, ηp2=.72. Planned comparisons revealed that, relative to when there was no distractor (996 ms), RTs were significantly slower when the distractor appeared in either the high-probability distractor location (1,070 ms), t(23)=6.28, p<.001, d=0.23, or the low-probability distractor location (1,126 ms), t(23)=9.14, p<.001, d=0.23. This indicates attentional capture by the distractor. Importantly, and in line with findings from previous studies, interference in RT was significantly lower when the distractor appeared in the high-probability compared to when it appeared in the low-probability distractor location, t(23)=5.81, p<.001, d=0.17 (Fig. 3A). This shows that for this high-probability location attentional capture by the distractor was reduced, suggesting spatial suppression of the high-probability distractor location.
The results of the analysis on error rates resembled those on RT. The ANOVA on error rate with distractor position (high-probability distractor location, low-probability distractor location, and distractor absent) as within-subject factor showed a main effect for distractor, F(2,46)=12.5, p<.001, ηp2=.35. Planned comparisons showed that error rate was higher when there was a distractor present in either the high-probability distractor location (5.9%), t(23)=2.81, p=.01, d=0.43, or any of the low-probability distractor locations (9.2%), t(23)=3.7, p=.001, d=0.3, relative to when there was no distractor present (4.3%). Importantly, error rate was significantly lower for when the distractor appeared in the high-probability distractor location relative to the low-probability locations, t(23)=3.46, p=.002, d=0.69 (Fig. 3A). This indicates that there were no speed-accuracy trade-offs.
To examine the spatial distribution of the reduction in distractor interference as was done in the previous reports on distractor suppression (e.g., Wang & Theeuwes, 2018a, 2018b, 2018c), RT was examined after sorting data according to the distance of the distractor to the high-probability distractor location (with dist-0 being the high-probability distractor location and dist-5 being the location on the opposite side of the imaginary circle). A repeated measures ANOVA showed a significant main effect for distance, F(4,92)=4.08, p=.004, ηp2=.15. Figure 4A (left panel) shows that RT plateaued at the dist-2 location. To describe the nature of the trend, a linear function was fitted for the data from dist-0, i.e., the high-probability distractor location, to dist-2, the plateau of the distribution. The slope (28.38 ms per point of distance) was significantly larger than zero, t(23)=6.0, p<.001, d=1.57, which shows that the reduction in distractor interference became smaller with increasing distance of the location of the distractor from its high-probability location. Identical analyses on error rates revealed similar results. The repeated measures ANOVA showed a significant main effect of distance, F(4,92)=2.83, p=.029, ηp2=.11, and the slope of the linear function (1.22% per display element) was marginally larger than zero (Fig. 4A, right panel), t(23)=1.86, p=.075, d=0.57. Post hoc inspection of the gradient suggests that the pattern observed here might not be best described by a linear function. Nonetheless, in line with previous studies, there is convincing evidence for a spatial gradient of suppression.
We also assessed the possibility that inter-trial location priming drove the suppression (e.g., Maljkovic & Nakayama, 1994). Indeed, the location of the distractor could repeat from one trial to another allowing for location priming to occur. To exclude this alternative explanation, we excluded all trials in which the location of the distractor repeated from one trial to the next. The ANOVA on mean RTs with distractor position (high-probability distractor location, low-probability distractor location, and distractor absent) as within-subject factor showed a main effect, F(2, 46)=48.05, p < .001, ηp2=.68. Relative to when there was no distractor (996 ms), RTs were significantly slower when the distractor appeared in either the high-probability distractor location (1,082 ms), t(23)=6.25, p<.001, d=0.26, or the low-probability distractor location (1,126 ms), t(23)=9.14, p<.001, d=0.27. Crucially, and in line with previous findings, interference by the distractor was reduced whenever it appeared in the high-probability location, t(23)=3.45, p=.002, d=0.13. This indicates that the previously reported effect is not the consequence of inter-trial location priming.
Target at the high-probability distractor location
To determine the efficiency of target selection in the absence of a distractor, we first analysed RT in distractor-absent trials. A paired-samples t-test showed that participants were significantly slower when the target appeared in the high-probability distractor location (1,035 ms) relative to when it appeared in one of the low-probability distractor locations (991 ms), t(23)=2.67, p=.013, d=0.14 (Fig. 3B). There was no effect on error rates (5.1% vs. 4.1%), t(23)=1.23, p=.232, d=0.21 (Fig. 3B).
To assess the spatial distribution of this interference in RT, distractor-absent trials were sorted according to distance of the target position to the high-probability distractor location (as was done for the other analyses). The repeated measures ANOVA on distance (dist-0 through dist-5) showed a marginally significant main effect, F(4,92)=2.39, p=.056, ηp2=.09, which plateaued at the dist-2 location (Fig. 4B). Fitting a linear function for the data on the first three locations revealed that the gradient in the distractor-absent condition was reversed relative to the one in the distractor-present condition. Interference reduced gradually with increasing distance of the target location to the high-probability distractor location. The slope was significantly different from zero (-21.83 ms per display element), t(23)=3.08, p=.005, d=0.92. There was no effect on error rates, t<1 (Fig. 4B).
In Experiment 2, the target could also appear in the high-probability distractor location in distractor-present trials. We therefore compared trials in which the target appeared in the high-probability distractor location with trials in which it appeared in any of the low-probability distractor locations. In line with the notion of spatial suppression of high-probability distractor locations, a paired-samples t-test showed that RT was significantly slower when the target appeared in the high-probability distractor location (1,134 ms vs. 1,083 ms), t(23)=3.63, p=.001, d=0.15. There were no significant differences in error rate.
Awareness assessment
Only nine participants correctly identified the high-probability distractor location with a relatively low average confidence (31.1% sure). The remaining participants (i.e., 15 participants) indicated an incorrect regularity. Since more participants correctly identified the high-probability location than would be expected by chance, we analyzed whether awareness had a significant effect on the suppression effect. However, an ANOVA on mean RT with awareness (aware vs. unaware) and distractor position (high-probability distractor location, low-probability distractor location, and distractor absent) as factors showed no significant interaction between awareness and distractor position, F(2,44)=.21, p=.808, ηp2=.09. To quantify the evidence against the interaction effect, we compared the model with both main effects against the model with the main effects and the interaction term. This comparison showed moderate evidence against the model that included the interaction (BF = 3.79). Analogously, comparing the model with the interaction against all other models provided evidence against the interaction (BFm = .473). In other words, there was no evidence that awareness affected suppression.
Discussion
The current experiment replicates all findings of Wang and Theeuwes (2018a, 2018b), even when the target is equally likely to appear in all locations. As in Wang and Theeuwes (2018a, 2018b), we showed reduced capture by the singleton distractor when it appeared in the high-probability distractor location and less efficient selection when the target happened to be presented in that location. Moreover, we found a spatial gradient effect of suppression around that location, evident both in distractor suppression and in less efficient selection of the target.