Online supplemental material Hoppenbrouwers et al 2015
Methods
Experiments
Experiment 1: bottom-up control of attention
The additional color singleton task was used to measure processing of salient task-irrelevant stimuli (Theeuwes, 1992). In this task, participants search for a diamond among circles. The participants have to indicate whether the line inside the diamond is oriented either horizontally (press the ‘z’-key) or vertically (press the ‘m’- key). The lines in the circles are diagonal (see Figure 1). In half of the trials, all display elements have the same color. In the other half of the trials (36 trials), a color distractor is present: one of the non-target circles has a different color (i.e., red).
This task started with 12 practice trials. Each trial started with a fixation dot presented for 600ms. The lines were white and were presented on a black background. Non-target line-orientations were randomly picked from the following orientations: 22.5, 45, 67.5, 112.5, 135 or 157.5 degrees. The display remained on the screen until a response was made, but not longer than 4 seconds. After an incorrect response, a red fixation dot was shown. After a correct response, a green fixation dot was shown.
Experiment 2: top-down control of attention
This task was administered to assess top-down control of attention. Previous research has shown that reaction times decrease when features that are relevant for target-selection are known before visual search is commenced (Kaptein, Theeuwes, & Van der Heijden, 1995; Treisman & Gelade, 1980). In this task, participants searched for a horizontal or vertical line, which was either located in one of the green or red circles. Before each trial, a written instruction was presented (‘attend’, ‘attend red’ or ‘attend green’). For the latter two instructions, the target was always located in the circle having the color indicated by the instruction (i.e., a 100% valid cue). This task started with 12 practice trials. Each trial started with a written instruction that was presented for 500ms, after which a fixation dot was presented for 600ms. The search display always contained 8 colored circles, of which one circle contained a horizontal or vertical (see Figure 1). The varying subset size consisted of either 2 or 6 elements, one of which would contain the target line segment. In total there were 288 trials (144 non-instructed trials (72 with a red target element and 72 with a green target element); 144 instructed trials (72 for a red target element and 72 for a green target element)). The radius of stimuli was 4.9 degrees. The lines were white and were presented on a black background. Non-target line-orientations were randomly picked from the following orientations: 22.5, 45, 67.5, 112.5, 135 or 157.5 degrees. The display remained on the screen until a response was made, but not longer than 4 seconds. After an incorrect response, a red fixation dot was shown. After a correct response, a green fixation dot was shown.
Data reduction and analyses
For Experiment 1, a repeated measures general linear model (GLM) was conducted to compare the mean reaction time of the condition where the distractor was present with the condition without a distractor. Subsequent repeated measures GLMs were run to check for any influences of degree of psychopathy (PPI Total, PPI-I and PPI-II). For Experiment 2, three repeated measures GLMs with Instruction and Set Size as within-subjects factors and with degree of psychopathy (PPI Total, PPI-I and PPI-II) as a covariate were run. Analyses were conducted using IBM SPSS Statistics 22. Alpha level of significance was set at 0.05.
Results
Experiment 1
Accepting the null hypothesis (i.e., that psychopathy is not related to deficits in bottom-up attention) is a problematic point, especially when using conventional statistics. To address this we use Bayesian statistics (Love et al., 2015) which allows to make inferences about the null hypothesis. The Bayes factor (BF01) is a probabilistic likelihood ratio that indicates whether the null hypothesis or the alternative hypothesis is a more accurate depiction of reality, given the data. The Bayes factor is calculated by dividing the probability of the null-hypothesis being true by the probability of the alternative hypothesis being true (Wetzels, Raaijmakers, Jakab, & Wagenmakers, 2009). In accordance with a general rule of thumb, BF above 10 is considered strong evidence whereas a Bayes factor between 5 and 10 is considered to be substantial evidence.
· Using JASP we ran Bayesian correlation pairs between the attentional capture (i.e., the absolute difference between the conditions ‘distractor present’ and ‘distractor absent’) and the PPI scores.
· The prior hypothesis was that PPI scores would correlate negatively with attentional capture. This was based on a clear prediction that has been made a number of times and is clearly described here:
“psychopaths are impaired in the ability to alter top-down goal-directed behavior to incorporate information from salient bottom-up stimuli (including threat cues). This inability to modulate behavior results directly from a failure to reallocate attention away from the goal-relevant task toward salient, but task-irrelevant, stimuli” (Larson et al 2013).
· The beta-width was set to 1.
For Experiment 1, BF01 for the correlation between bottom-up attention and total PPI score was 13.45, in favor of the null-hypothesis. This means that the data is roughly 13 times more likely to be an accurate depiction of reality under the null hypothesis than under the alternative hypothesis. For PPI-I and PPI-II BF01 is 6.39 and 15.16, respectively, also in favor of the null-hypothesis.
As this study is a replication and extension of an earlier study (Hoppenbrouwers, Van der Stigchel, Slotboom, Dalmaijer, & Theeuwes, 2015), we have also calculated BF01 for those data to provide further evidence in favor of the null hypothesis (regarding bottom-up attention). In a well-controlled sample of violent offenders (i.e., no medication, no Axis I disorders, not older than 65 years of age and controlled for the influence of IQ), we showed that psychopathy as measured with the PCL-R is not related to bottom-up attention. In relation to bottom-up attention, for the total PCL-R score, Factor 1 and Factor 2, BF01 was 7.4, 7.8 and 5.1 respectively. These outcomes provide substantial evidence, also in favor of the null hypothesis.
Experiment 2
For the non-instructed condition 92.4% of trials were left, whereas for the instructed conditions (‘attend green’ and ‘attend red’) 97.2% and 97.1% of all trials were left, respectively.
Task manipulation
Two paired samples t-tests were used to check for differences between the two instructed conditions (“attend red” and “attend green”) for both Set Sizes. There were no significant differences, p’s> .291. The two instructed conditions were therefore averaged.
A repeated measures GLM with Instruction (instructed vs non-instructed) and Set Size as within subjects variables, showed a main effect of Instruction, F(1,61)= 118.665, p< .001, partial η2= .660, indicating that participants were faster in the condition in which a specific instruction was given (e.g., attend red, or attend green). A main effect of Set Size was also observed, (F1,61)= 122.766, p< .001, partial η2= .668, indicating that participants were faster when the target was in the smaller set (i.e., two circles in which the target could be) compared to the larger set (i.e., six circles in which the target could be) (See Figure 2). In addition, there was an interaction between Instruction and Set Size was observed, F(1,61)= 38.762, p< .001, partial η2= .389. Reaction times differences between the instructed and non-instructed condition in the smaller set size were significantly larger than between the instructed and non-instructed condition for the larger set size, t(61)= 6.226, p< .001. Together, these analyses showed that the instruction was used and that set size modulated search times indicating that task manipulation was successful.
Figure 1. Reaction time as a function of set size for conditions in which participants received an instruction (e.g., attend red, or attend green) or received no specific instruction. Error bars reflect SEM.
Accuracy
Two paired samples t-tests were used to check for differences between the two instructed conditions (“attend red” and “attend green”) for both Set Sizes. There were no significant differences, p’s> .326. The two instructed conditions were therefore averaged.
A repeated measures GLM with Instruction and Set Size as within subjects variable showed a significant main effect for Instruction F(1,61)= 94.766, p< .001, partial η2= 0.608. This indicated that participants made significantly less errors (~5%) in the instructed condition. No main effect for Set Size was also observed, F(1,61)= .614, p= .436, partial η2= 0.01. There was no interaction between Instruction and Set Size, F(1,61)= .198, p= .658, partial η2= 0.03.
Inter-trial priming
Median-split analyses
To further explore the significant interactions between inter-trial priming and psychopathy, a median-split on the degree of psychopathy (PPI Total, PPI-I and PPI-II) was conducted. For PPI Total, a paired samples t-test showed that the high-psychopathy group was significantly slower when the previous target response was different than the current, t(29)= 2.161. p= .039, while this was not the case for the low-psychopathy group, t(31)= 0.631, p= .533 (See Figure 4). For PPI-I, a paired samples t-test showed that the high-psychopathy group was significantly slower when the previous target response was different than the current, t(31)= 2.553, p= .016, while this was not the case for the low-psychopathy group, t(29)= 0.871, p= .391. For PPI-II, a paired samples t-test showed that the high-psychopathy group was not significantly slower when the previous target response was different than the current, t(29)= 1.518, p= .140, which was also the case for the low-psychopathy group, t(31)= 0.025. p= .980.
Figure 2. Reaction times for current response (same previous response versus different previous response) for participants with low and high psychopathic traits are shown. Error bars reflects SEM.