As mentioned above, in Experiment 1, the flanker task was embedded in a PP paradigm (Miller & Durst, 2014). In the present adaptation of the PP paradigm, participants are asked to first respond to the center letter, and only respond to the flanker letter if the center letter requires a no-go response. Consequently, any trial that requires a response to the center letter, ends after that response is executed, resembling the classic flanker task with just one overt response per trial. The PP paradigm shares many observable commonalities with the PRP paradigm (Miller & Durst, 2015), and also BCEs have been found in the PP paradigm (e.g., Miller & Durst, 2015; Mittelstädt & Miller, 2017). This therefore allows us to use this paradigm to investigate the BCEs that are typically present in dual-task situations, and try to further locate the sources of this BCE.
One previous study which investigated the BCE in the PP paradigm in more detail was Miller (2017). In this study, lateralized readiness potentials were used to determine the time at which BCEs arise during processing. The results suggested that BCEs are present because T2 stimuli influence T1 response selection (i.e., S2−R1 compatibility effect)—and not because T2 stimuli activate their corresponding R2 (i.e., no evidence for R2−R1 compatibility). The present experiment, then, aims at further shedding light on the nature of BCEs using a dual-task flanker paradigm, and at separating these two kinds of compatibility effects. Specifically, as mentioned above, we used three mapping conditions (i.e., matched: S2−R1 compatible trials were also R2−R1 compatible; reversed: S2−R1 compatible trials were R2−R1 incompatible; and neutral: S2 neutral to R1, and separate R2−R1 compatibility) to separate the contributions S2−R1 and R2−R1 compatibility.
Method
Participants
Participants were 60 University of Otago psychology students (46 women) who took part in the experiment in exchange for course credit. They ranged in age from 17 to 26 (M= 19.8) and they were predominantly right-handed (M = 59.6) as indexed by the Edinburgh Handedness Inventory (Oldfield, 1971). We planned on sample sizes of 20 participants with accuracy above 80% per S-R mapping condition but actually tested one extra participant in the matched and neutral conditions due to the unpredictabilities associated with experimental participation. To obtain equal sample sizes in each condition, we dropped the participant with the lowest accuracy from the matched and neutral groups. We further excluded six additional participants due to low accuracy (i.e., below 80%).
Apparatus and stimuli
The experiment took place in individual test rooms. Stimulus presentation and recording of responses were controlled by an IBM-PC compatible computer using MATLAB with the Psychophysics Toolbox extension (Brainard, 1997; Pelli, 1997). Viewing distance was approximately 60 cm but not restrained. Stimuli were presented vertically and horizontally centered on a 17” screen in a white 35 point font—that is, a center letter was displayed flanked by two outside letters on each side. The letter stimuli were presented in white 35pt font. A centered, white plus sign (+) served as fixation point. Responses were key presses with the left and right index fingers on the “Z” and “?/” keys of a standard computer keyboard.
For each participant the letter stimuli were randomly selected from all consonants, excluding the letters L, R, and Z in order to avoid associations with response side or key. More specifically, two letters each were assigned for any of the three T1 response possibilities (i.e., left/right keypress, no-go). In the matched and reversed conditions, the same letters were used for T2 as for the go-stimuli for T1, with the same or reversed S-R rule, respectively. The T1 no-go letters never appeared as flankers in any condition. In the neutral condition, two additional stimuli were assigned to each possible response (i.e., left/right keypress) for T2, leading to a total of ten different letters used in this condition, and six different letters in the matched and reversed conditions. We only tested trials with different letters for the center and flanker stimuli, at least one of which was assigned to a go response, thus leaving 20 different trial types (we omitted one response-compatible stimulus combination in the neutral condition in order to obtain the same number of trial types in this condition).
Procedure
The single experimental session lasted approximately 45 min. Each subject was tested in one of the three conditions (i.e., matched, reversed, or neutral). The experiment consisted of two practice blocks and eight experimental blocks. The two practice blocks served as single-task training for T1 and T2, respectively, displaying the typical trial sequences but instructing participants to pay attention to only one of the two presented stimuli. In the experimental blocks, subjects were instructed to treat the center letter as the high priority task, and the flanker letters as the low priority task. That is, subjects were instructed to first respond to T1 with left/right index finger presses, respectively, and to only respond to T2 if S1 was the no-go stimulus. In the matched condition, the S-R rule was the same for T2 as for T1. In the reversed condition, the S-R rule for T2 was reversed to the rule for T1—that is, the stimuli which required a left response for T1 required a right response for T2 and vice versa. In the neutral condition, four additional letters served as the stimuli for T2 with two letters each corresponding to left/right responses, respectively. The mapping is illustrated in Table 1.
Each trial type was presented twice in the practice blocks (40 trials). The experimental blocks included 72 trials. In two-thirds of these trials, T1 required a response, and in one-third, T2 required a response. Of the T1 trials, half of the trials were R2−R1 compatible, and the other half were R2−R1 incompatible. In the trials where T1 required no response, half of the responses to T2 required a left-hand response and half required a right-hand response.
The trial sequence started with a fixation cross for 500 ms. The center letter and the flankers were displayed with an SOA of 100 ms. Stimuli remained on screen for a maximum of 2 s or until a response was given. Erroneous responses resulted in an error feedback screen display of 3 seconds. After responses faster than 200 ms or slower than 2 s, subjects were informed that they had responded too fast or too slow, respectively, for 3 s. If subjects made more than three consecutive errors, they were shown the instructions again as a reminder. Trials were followed by an intertrial interval of 2 s.
Results
Practice blocks and the first experimental block were excluded from any further analysis as training. Erroneous trials were removed from the RT analyses (8.7%). One and 34 trials were removed from the analyses as RT outliers based on lower and upper RT cutoffs of 200 ms and 2 s, respectively. The analyses comparing the matched and reversed conditions focus on stimulus compatibility. In the neutral mapping condition, there is no S2−R1 compatibility as the flanker letters are always drawn from a different letter set than the center letters—thus, for this condition, we refer to the R2−R1 compatibility.
Primary task: RT1 and PE1
Matched vs. reversed conditions
Figure 2A shows the means of RT1 for the matched and reversed mapping conditions as a function of S2−R1 compatibility. We ran an ANOVA with the between-subject factor condition (i.e., matched, reversed) and the within-subject factor of stimulus (i.e., flanker) compatibility for RT1. This ANOVA yielded a significant main effect of stimulus compatibility, F(1,38) = 51.119, p < .001, \(\eta ^{2}_{p}=0.574\), with faster responses in stimulus compatible (721 ms) than in stimulus incompatible (765 ms) trials. Interestingly, no other effect was significant (p s >.241), indicating no differences in compatibility effects between the matched and reversed conditions.
We ran parallel analyses on the percentage of erroneous responses for the primary task (PE1). Figure 2B visualizes the results of this ANOVA. This ANOVA revealed a significant main effect of stimulus compatibility, F(1,38) = 45.690, p < .001, \(\eta ^{2}_{p}=0.546\), with fewer erroneous responses in the stimulus compatible (4.3%) than in the stimulus incompatible (8.3%) trials. Interestingly, the condition x stimulus compatibility interaction was significant, \(F(1,38) = 15.972, p < .001, \eta ^{2}_{p}=0.296\), indicating a larger stimulus compatibility effect in the matched (6.5%) than in the reversed (1.7%) condition, with the latter effect still significant, \(F(1,19) = 7.175, p=.015, \eta ^{2}_{p}=0.274\). The main effect of condition was not significant, p =.519.
However, the interpretation of the PE data is not as straightforward in the PP paradigm, because R1 and R2 were both performed with the same response keys. It is therefore not possible to distinguish which of the two tasks the participant actually aimed to respond to, and there are most likely some trials where a participant incorrectly responded to T2 where s/he should have responded to T1. In the matched condition, the responses in these trials were correct in the S2−R1 compatible trials and incorrect in the reversed condition. Thus, these “wrong task” trials increase the S2−R1 BCE on PE in the matched condition. In the reversed condition, however, the opposite happens, and the “wrong task” trials decrease the compatibility effect. It might therefore be possible that this task confusion effect underlies the significant interaction reported above.
We checked whether the task (i.e., T1 vs. T2) responded to on trial n-1 played a role here, because it is reasonable to assume that this task confusion happens more often following a background task response. We thus added the task on trial n-1 as an additional factor in the ANOVA reported above. Figure 3 visualizes the separate ANOVAs for each mapping condition when including the n-1 task factor. This ANOVA produced a significant three-way interaction of condition, stimulus compatibility, and n-1 task, \(F(1,38) = 14.096, p=.001, \eta ^{2}_{p}=0.271\). We followed this up by separately running ANOVAs for each mapping condition. In the matched condition, there was a significant interaction between n-1 task and stimulus compatibility, \(F(1,19) = 39.463, p < .001, \eta ^{2}_{p}=0.675\). In the reversed condition, this interaction was not significant, \(F(1,19) = 0.955, p=.341, \eta ^{2}_{p}=0.048\). As can be seen in Fig. 3A and B, it seems as if the larger S2−R1 compatibility effect just stems from the trials following a background task response. Thus, it seems plausible that this task confusion was responsible for the difference in the sizes of the S2−R1 compatibility effects between the matched and reversed conditions.
Neutral condition
Figure 2C shows the means of RT1 for all groups (i.e., matched, reversed, neutral) as a function of R2−R1 compatibility, thus also showing the means of the neutral condition. In the neutral condition, we did not observe any R2−R1 compatibility effect, \(F(1,19) < 0.01, p=.939, \eta ^{2}_{p}=0\), with no difference between R2−R1 compatible (732 ms) and incompatible (732 ms) trials. Separate ANOVAs with each pair of conditions showed that this null effect was significantly different from the compatibility effects obtained in both the matched condition, \(F(1,38) = 20.275, p < .001, \eta ^{2}_{p}=0.348\), and the reversed condition, \(F(1,38) = 10.486, p=.002, \eta ^{2}_{p}=0.216\).
We again ran parallel analyses on PE1, and Fig. 2D shows the means of PE1 for all three conditions as a function of R2−R1 compatibility. In the ANOVA of the neutral condition, responses in R2−R1 compatible trials were significantly less erroneous (4.7%) than in R2−R1 incompatible trials (8.0%), F(1,19) = 18.875, p < .001, \({\eta }^{2}_{p}=0.498\). Again, in the ANOVA comparing the neutral and matched conditions, there was a significant interaction of R2−R1 compatibility and condition (i.e., neutral vs. matched), F(1,38) = 6.315, p = .016, \(\eta ^{2}_{p}=0.142\), indicating a smaller compatibility effect in the neutral (3.3%) than in the matched (6.5%) condition. In the ANOVA comparing the neutral and reversed conditions, the interaction of compatibility and mapping condition was not significant, F(1,38) = 2.705, p = .108, \(\eta ^{2}_{p}=0.066\).
As was mentioned above, it is not possible to clearly distinguish which task a participant intended to respond to in the PP paradigm—possibly allowing task confusion effects in the error data to produce effects which look like compatibility effects. We therefore ran a parallel analysis including the n-1 task (i.e., T1 vs. T2) as an additional factor to the ANOVA reported above. The corresponding means of this ANOVA are displayed in Fig. 3C. There was a significant interaction of n-1 task and R2−R1 compatibility in the neutral condition, F(1,19) = 19.369, p < .001, \(\eta ^{2}_{p}=0.505\), indicating that the R2−R1 compatibility effect for PE reported above is mainly due to trials following a background task response. Because there was no R2−R1 compatibility effect in the RT data, we believe that this task confusion effect is the most likely explanation for the R2−R1 compatibility effect in the PE data.
Background task: RT2 and PE2
An ANOVA with the between-subject factor mapping condition was run for RT2. We found a significant main effect of condition, F(2,57) = 16.422, p < .001, \(\eta ^{2}_{p}=0.366\), with RT2 being fastest in the matched condition (828 ms), followed by the neutral condition (956 ms), and the slowest T2 responses in the reversed condition (1087 ms). Post hoc pairwise comparisons revealed significant differences between all three conditions (p s <.014). Note that it is not possible to examine effects on R2 of the match between R1 and R2 (i.e., forward compatibility) in the PP paradigm, because any trial with an R2 in the PP paradigm necessarily followed a no-go S1. Such effects will be examined in Experiment 2 using the PRP paradigm, however.
A parallel analysis was carried out for the percentage of erroneous responses for the background task (PE2). For PE2, a similar result pattern emerged. That is, the ANOVA revealed a significant main effect of condition, F(2,57) = 8.214, p = .001, \(\eta ^{2}_{p}=0.224\). Post-hoc pairwise comparisons revealed a significant effect between the matched condition (9.9%) and the reversed condition (18.1%), p < .001, and between the neutral (12.6%) and the reversed condition, p = .015. The comparison between the matched and the neutral condition was not significant, p = .151.
Discussion
The main findings of Experiment 1 can be summarized as follows. In both the matched and the reversed conditions, we found a significant S2−R1 compatibility effect, with little change in this effect across the two mapping conditions, replicating the general existence of BCEs in the PP paradigm (e.g., Miller, 2017; Miller and Durst, 2015; Mittelstädt & Miller, 2017). In the neutral condition—which featured only R2−R1 compatibility—we found no compatibility effect.Footnote 1 Even though there was some evidence for an R2−R1 compatibility effect in the error data, we believe that this is mostly due to task confusion effects. This argument seems particularly plausible when considering that these R2−R1 compatibility effects stem only from trials that followed trials requiring a background task response—and it seems logical that this kind of task confusion (i.e., responding to T2 instead of T1) happens more often after responding to the background task.
The finding of a stimulus-based BCE on RT aligns well with previous findings in the PP paradigm. That is, it seems that the BCE is mainly based on S2−R1 compatibility, and not R2−R1 compatibility, aligning well with the findings of Miller (2017). S2−R1 compatibility seems to be the main source of the BCE in Experiment 1, as (a) the S2−R1 compatibility effect was also observed in the reversed condition, where S2−R1 compatible trials were R2−R1 incompatible, (b) this S2−R1 compatibility effect was not significantly decreased in the reversed compared to the matched condition, and (c) we did not observe any compatibility effect in the neutral condition, which only featured R2−R1 compatibility, but not S2−R1 compatibility. Moreover, there was also a main effect of mapping condition on RT2. Obviously, one would expect this effect because of mapping complications in the reversed and neutral conditions compared to the matched condition—however, our main concerns were regarding the BCEs in the different conditions.
However, as the PP paradigm only requires one overt response per trial, it is not clear whether that might have played a role in finding no evidence for an R2−R1 compatibility effect (or at least a decreased S2−R1 compatibility effect) in the reversed condition. Moreover, we also did not observe any R2−R1 compatibility effect in the neutral condition (which featured only R2−R1 but not S2−R1 compatibility), and this could also possibly have been due to the nature of the PP paradigm, where only one overt response per trial is required, with strong prioritization of T1. Therefore, Experiment 2 used a PRP paradigm—where on every trial two overt responses are necessary.