The congruency sequence effect emerges when the distracter precedes the target

Share Embed


Descripción

Acta Psychologica 156 (2015) 8–21

Contents lists available at ScienceDirect

Acta Psychologica journal homepage: www.elsevier.com/ locate/actpsy

The congruency sequence effect emerges when the distracter precedes the target Daniel H. Weissman a,⁎, Tobias Egner b, Zoë Hawks a, Jacqueline Link a a b

Department of Psychology, University of Michigan, USA Center for Cognitive Neuroscience, Duke University, USA

a r t i c l e

i n f o

Article history: Received 10 October 2014 Received in revised form 31 December 2014 Accepted 6 January 2015 Available online xxxx PsycINFO Codes: 2340 Cognitive Processes Keywords: Conflict adaptation Gratton effect Sequential modulations Response inhibition

a b s t r a c t The congruency effect in distracter interference tasks is typically smaller when the previous trial was incongruent as compared to congruent, suggesting the operation of a control process that minimizes the influence of irrelevant stimuli on behavior. However, both the conditions under which this congruency sequence effect (CSE) can be most easily observed without the typical learning and memory confounds, and the control process underlying it, remain controversial. We therefore tested a recent hypothesis that the CSE is most easily observed without the typical confounds when the distracter is processed before the target. In line with this “distracter head start” hypothesis, in Experiments 1 and 2 the CSE was larger when the distracter appeared before, relative to with, the target. Further, in Experiment 3, we observed a negative congruency effect after incongruent trials when a long interval separated the distracter from the target, consistent with a modulation of the response engendered by the distracter but not with a shift of attention toward the target. These findings reveal an important determinant of CSE magnitude when the typical learning and memory confounds are absent and new insights into the nature of control processes that contribute to this phenomenon. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Researchers often employ distracter interference tasks to investigate how selective attention minimizes distraction from irrelevant stimuli. In such tasks, study participants are instructed to identify a target stimulus while ignoring one or more distracters. For example, in the classic flanker task, participants are asked to identify a central target letter (e.g., H or S) that is positioned between two identical distracters (Eriksen & Eriksen, 1974). In congruent trials, the target and distracter letters match (HHH or SSS) and therefore engender the same response. In incongruent trials, the target and distracter letters differ (HSH or SHS) and therefore engender conflicting responses. Across a wide variety of distracter interference tasks (e.g., Stroop, flanker, and Simon), performance is slower and less accurate in incongruent than in congruent trials (MacLeod, 1991). This phenomenon, known as the congruency effect, indicates that selective attention often does not eliminate distraction from irrelevant stimuli. The efficiency of selective attention, however, appears to vary considerably from one moment to the next. For example, the congruency effect is typically smaller when the previous trial was incongruent as compared to congruent. This congruency sequence effect (CSE) has been observed in numerous distracter interference tasks, including the ⁎ Corresponding author at: Department of Psychology, 530 Church Street, Ann Arbor, MI 48109, USA. Tel.: +1 734 763 3321; fax: +1 734 647 9440. E-mail address: [email protected] (D.H. Weissman).

http://dx.doi.org/10.1016/j.actpsy.2015.01.003 0001-6918/© 2015 Elsevier B.V. All rights reserved.

flanker task (Gratton, Coles, & Donchin, 1992), the Stroop task (Kerns et al., 2004), and the Simon task (Sturmer, Leuthold, Soetens, Schroter, & Sommer, 2002). There is an ongoing debate, though, about whether the CSE is driven by (1) a cognitive control process that minimizes the influence of irrelevant stimuli or (2) learning and memory processes that are confounded with the CSE in the overwhelming majority of distracter interference tasks (Egner, 2007; Schmidt, 2013). We begin by describing two cognitive control accounts of the CSE – the attentional shift account and the response modulation account – before moving on to discuss learning and memory accounts and a recent hybrid cognitive control/learning memory account. 1.1. Cognitive control accounts The attentional shift account posits that the CSE reflects a control process that changes the distribution of attention to distracter and target stimuli as a function of whether the previous trial was incongruent or congruent. Four distinct variants of this account appear most prevalent in the literature. First, the perceptual expectation hypothesis posits that participants expect the forthcoming trial to resemble the previous trial, and adapt their attentional strategy accordingly. For instance, in the flanker task, the focus of spatial attention would widen after a congruent trial to include both the target and the distracter, but would narrow after an incongruent trial to include the target but exclude distracter (Gratton et al., 1992). Second, the conflict monitoring hypothesis asserts that participants pay more attention to the target and/or less attention

D.H. Weissman et al. / Acta Psychologica 156 (2015) 8–21

to the distracter in the current trial when the previous trial engenders high levels of response conflict (e.g., incongruent trials) as compared to low levels of response conflict (e.g., congruent trials) (Botvinick, Braver, Barch, Carter, & Cohen, 2001). Third, in tasks wherein the distracter temporally precedes the target, the temporal attention hypothesis posits that, via a change in the task representation, participants allocate more or less attention to the moment in time at which the distracter appears, respectively, depending on whether the previous trial was congruent or incongruent (Hazeltine, Lightman, Schwarb, & Schumacher, 2011). Fourth, the negative affect hypothesis posits that participants experience the effort associated with processing an incongruent stimulus in the previous trial as aversive (cf. Botvinick, 2007), which leads them, in a form of avoidance learning, to increase attention to the target and/or decrease attention to the distracter in the current trial, thereby avoiding the re-occurrence of negative affect (Dreisbach & Fischer, 2012). Critically, all four variants of the attentional shift account posit that the CSE indexes a relative increase of attention toward the target and away from the distracter when the previous trial was incongruent relative to congruent. The response modulation account posits that the CSE indexes a control process that modulates the response engendered by the distracter when the distracter is able to activate a response before the target. There are two distinct variants of this account. First, the activation–suppression hypothesis posits a control mechanism that inhibits the response evoked by the distracter. To account for the CSE, the model assumes that this mechanism is more efficient on the current trial when it had to be engaged on the previous trial; thus, the distracter response is inhibited more effectively before the target is identified when the previous trial was incongruent as compared to congruent (Ridderinkhof, 2002a). Second, in two-alternative-forced-choice (2-AFC) tasks, the response expectation hypothesis posits that participants “pre-activate” the response signaled by the distracter if the previous trial was congruent, leading to a relatively large congruency effect, or the opposite response if the previous trial was incongruent, leading to a relatively small congruency effect (Logan, 1985). While this hypothesis was originally formulated in the context of biased proportions of congruent and incongruent trials (Logan & Zbrodoff, 1979), such a strategy may occur even when congruent and incongruent stimuli appear equally often, because participants appear to expect congruency repetitions across consecutive trials more than they expect congruency alternations (Duthoo, Wuhr, & Notebaert, 2013; Gratton et al., 1992). In sum, the response modulation account posits that the CSE occurs when the distracter is processed before the target, such that control processes have time to modulate the distracter response before the target is identified. 1.2. Learning and memory accounts In contrast to the cognitive control accounts of the CSE above, learning and memory accounts posit that the CSE reflects learning and memory processes that are confounded with congruency sequence in the vast majority of distracter interference tasks. There are two main variants of the learning and memory account. First, the feature integration hypothesis posits that the CSE indexes the influence of stimulus and/or response feature repetitions that are typically confounded with congruency sequence in distracter interference tasks (Hommel, Proctor, & Vu, 2004; Mayr, Awh, & Laurey, 2003). Specifically, especially in distracter interference tasks with small stimulus sets, congruency repetitions across consecutive trials (i.e., congruent–congruent [cC] and incongruent– incongruent [iI] trials) are composed of either complete stimulus and response feature repetitions or complete stimulus and response feature alternations. In contrast, congruency alternations across consecutive trials (i.e., congruent–incongruent [cI] and incongruent–congruent [iC] trials) consist entirely of “partial repetitions”, wherein some stimulus and response features change while others remain the same. Since it is well-established that partial repetitions evoke a form of memoryretrieval conflict, and are associated with slower and more error-prone responses (Hommel, 1998), the feature integration hypothesis posits

9

that the CSE indexes different distributions of complete feature repetitions, complete feature alternations, and partial feature repetitions across the four possible congruency sequences (Hommel et al., 2004). Second, in tasks with more than two stimuli and responses, the contingency learning hypothesis posits that the CSE indexes a stronger association between each distracter and its congruent response than between each distracter and any of the multiple possible incongruent responses (Schmidt & De Houwer, 2011). This strengthened association, or “high contingency”, occurs because each distracter is presented more frequently with the congruent target than with each possible incongruent target, a procedure that is frequently employed to equate the number of congruent and incongruent trials (Schmidt, 2013). Since responding in a high contingency (congruent) trial occurs more quickly following a previous high contingency trial, it has been suggested that the CSE reflects a contingency (rather than congruency) sequence effect (Schmidt & De Houwer, 2011). Critically, in line with the learning and memory account, the CSE often vanishes in the manual Stroop, flanker, and Simon tasks when feature integration and contingency learning confounds are removed (Mordkoff, 2012; Schmidt & De Houwer, 2011). Thus, some researchers have suggested that the CSE reflects nothing more than learning and memory confounds (Mayr et al., 2003; Schmidt, 2013; Schmidt & De Houwer, 2011). 1.3. Hybrid accounts Some researchers have suggested that cognitive control and learning and memory processes interact to engender a CSE. For example, the adaptation by binding model posits that response conflict in incongruent trials leads a performance-monitoring system housed in the posterior medial frontal cortex (pMFC) to increase arousal via interactions with the locus coeruleus (Verguts & Notebaert, 2008). This increase in arousal strengthens associations between task-relevant stimulus features and the current task representation, which reduces the congruency effect in the next trial by enabling the correct response in incongruent trials to be more quickly retrieved. In more recent versions of the model (Blais & Verguts, 2012), the association between a given stimulus feature and the current task representation is strengthened in incongruent trials to a greater degree when the stimulus was more recently presented (and thus has a higher level of activation) than when the stimulus was less recently presented (and thus has a lower level of activation). This assumption fits with several recent findings indicating that the CSE is reduced or eliminated when the current trial contains no feature repetitions from the previous trial (Blais & Verguts, 2012; Mayr et al., 2003; Mordkoff, 2012; Schmidt & De Houwer, 2011). Such findings clearly indicate that learning and memory processes contribute to the CSE, either independently or via interactions with cognitive control. 1.4. Recent support for an independent contribution of cognitive control to the CSE Consistent with accounts in which cognitive control is able to contribute to the CSE independently of learning and memory processes, we recently reported that the CSE can be observed without feature integration or contingency learning confounds (Schmidt & Weissman, 2014). For example, in each trial of a prime–probe word task, three vertically-stacked distractor words (left, right, up, or down) were followed by a target word (left, right, up, or down). Participants were asked to identify the direction indicted by the target word (e.g., left) as quickly as possible without making mistakes by pressing one of four spatially-compatible keys. To prevent stimulus repetitions in consecutive trials, we divided the 4-AFC prime–probe word task into a pair of 2-AFC choice tasks – a “left–right” task and an “up–down” task – each of which contained two congruent stimuli and two incongruent stimuli (Mayr et al., 2003). We then alternated between these tasks on every trial (Jimenez & Mendez, 2013; Mayr et al., 2003). To prevent response repetitions in consecutive trials, participants responded

10

D.H. Weissman et al. / Acta Psychologica 156 (2015) 8–21

in the “left–right” task using two fingers of the left hand and responded in the “up–down” task using two fingers of the right hand. Finally, to avoid contingency learning biases, we presented the two congruent and two incongruent stimuli in each task equally often. Critically, we observed a CSE in the prime–probe word task, conceptually replicating prior findings from a prime–probe arrow task in which mental rotation confounds could have influenced the CSE (Kunde & Wuhr, 2006). We therefore concluded that a CSE can be observed without the typical confounds in the prime–probe word task, consistent with an independent contribution of cognitive control processes to this phenomenon.

1.5. Tentative support for an important determinant of CSE magnitude Other recent findings from our group indicate that a CSE can be observed without the typical confounds more readily in some tasks than in others. Namely, we observed a CSE in the Simon task1 but not in the manual Stroop or flanker tasks (Weissman, Jiang, & Egner, 2014). To explain why a CSE is more readily observed in the prime–probe and Simon tasks than in the manual Stroop and flanker tasks, we considered the likely temporal dynamics of response activation in these four tasks. While there is evidence to suggest that the distracter response becomes active before the target response in all four tasks (Eimer & Schlaghecken, 1998; Gratton, Coles, Sirevaag, Eriksen, & Donchin, 1988; Sturmer, Ouyang, Zhou, Boldt, & Sommer, 2013; Szucs, Soltesz, Bryce, & Whitebread, 2009), this is the case to a greater degree in the prime–probe and Simon tasks than in the manual Stroop and flanker tasks. In the prime–probe task, the distracter is presented before the target, which gives activation of the distracter response a large “head start” relative to activation of the target response. Similarly, although the distracter and target are presented simultaneously in the manual Stroop, flanker, and Simon tasks, recent findings from an event-related potential (ERP) study indicate that the distracter (spatial position) in the Simon task activates a response more than 100 ms before the distracter (item identity) in the flanker task (Mansfield, van der Molen, Falkenstein, & van Boxtel, 2013). The authors of this study concluded that it takes less time to translate spatial information in the Simon task into a spatially-compatible response than to translate symbolic information in the flanker task into an arbitrary response. Since the distracter (word identity) in the manual Stroop task is also symbolic, the degree to which the distracter activates a response before the target is also likely greater in the Simon task than in the manual Stroop task. Given these considerations, we suggested that the degree to which the distracter activates a response before the target plays an important role in determining CSE magnitude, regardless of whether the distracter and target appear sequentially (in the prime–probe task) or simultaneously (in the Simon, flanker, and manual Stroop tasks). In the present article, we refer to this suggestion as the distracter head start hypothesis. This suggestion appears more consistent with the response modulation account of the CSE than with attentional shift account (Weissman et al., 2014). Indeed, while the response modulation account posits that CSE magnitude should depend on the degree to which the distracter activates a response before the target, the attentional shift account does not make this assertion, unless giving the distracter a head start also happens to increase conflict or trial difficulty (Botvinick et al., 2001; Dreisbach & Fischer, 2012). In fact, since greatly reducing the temporal overlap of response activations engendered by the distracter and target can at times strongly attenuate response conflict (i.e., reducing or even abolishing the congruency effect; e.g., Eriksen & Schultz, 1979), the attentional shift account should at times predict a reduced CSE when the distracter is given a large head start. We therefore suggested that the CSE is more 1

Our finding of a CSE in the Simon task may appear inconsistent with a recent finding indicating that the CSE in the Simon task vanishes when feature repetition and contingency learning confounds are removed (Mordkoff, 2012). However, our protocol likely provided more statistical power for observing a CSE than this prior study because it included many more trials and participants.

consistent with the response modulation account than with the attentional shift account. Drawing a firm conclusion on this point, however, would be premature for two reasons. First, differences among the prime–probe, Simon, flanker, and manual Stroop tasks that are unrelated to the relative timing with which distracter and target responses become active could have influenced the pattern of CSEs in our prior study. These include differences in the task-relevant stimulus dimension, the taskirrelevant stimulus dimension, the stimulus–response mapping, and so forth. Thus, it remains unclear whether the distracter head start hypothesis describes an important determinant of CSE magnitude when the typical learning and memory confounds are absent. Second, the response modulation account makes a unique prediction that we did not test in our prior study (Weissman et al., 2014). Namely, this account predicts a reverse, or negative, congruency effect following incongruent trials when a long interval separates the distracter and the target in each trial. Under such conditions, the overall congruency effect should be minimal and there should be ample time to (a) inhibit the response signaled by the current-trial distracter (Burle, van den Wildenberg, & Ridderinkhof, 2005; Machado, Wyatt, Devine, & Knight, 2007) or (b) pre-activate either the response signaled by the distracter if the previous trial was congruent or the opposite response if the previous trial was incongruent (Gratton et al., 1992; Logan, 1985).2 Critically, as we describe more fully in Experiment 3, either process should lead to a negative congruency effect after incongruent trials. In contrast, even shifting all of one's attention to the target following incongruent trials might eliminate, but not reverse, the congruency effect. Thus, observing a negative congruency effect after incongruent trials would indicate that the CSE is more consistent with the response modulation account than with the attentional shift account.

1.6. The present study The present study had two goals. First, we wished to conduct a stronger test of the distracter head start hypothesis than we conducted in our prior study (Weissman et al., 2014). Second, we wished to determine whether the CSE is more consistent with the response modulation account than with the attentional shift account. To achieve these goals, we varied the inter-stimulus interval (ISI) between the distracter and the target in a prime–probe task (Eriksen & Schultz, 1979). In Experiment 1, we investigated whether the CSE is larger when the distracter appears shortly before the target (sequential trials) relative to simultaneously with the target (simultaneous trials). We reasoned that such a finding would provide novel support for the distracter head start hypothesis. In Experiment 2, we conducted a fine-grained test of this hypothesis by determining whether the CSE is larger when the distracter appears shortly before the target in the previous trial, the current trial, or both. In Experiment 3, we investigated whether the CSE is associated with a negative congruency effect following incongruent trials when a relatively long ISI separates the distracter and the target. As explained earlier, such an effect is predicted by the response modulation account but not by the attentional shift account. Crucially, in all three experiments we measured the CSE without feature integration or contingency learning confounds using the methods described earlier and in our prior work (Schmidt & Weissman, 2014; Weissman et al., 2014). Finally, we note that although the congruency effect is typically reduced when a long ISI separates the distracter and the target (Eriksen & Schultz, 1979; Machado et al., 2007), a variety of factors influence

2 Since the CSEs in our experiments come from 4-AFC tasks, one might question whether the response expectation hypothesis – which was developed to account for performance in 2-AFC tasks – could ever account for them. However, since we divided each 4AFC task into a pair of 2-AFC tasks and alternated between these two tasks on every trial, it is possible that the response expectation hypothesis might still explain the CSEs we observed.

D.H. Weissman et al. / Acta Psychologica 156 (2015) 8–21

whether it is maximal when the distracter appears shortly before the target or simultaneously with the target (Appelbaum, Boehler, Won, Davis, & Woldorff, 2012; Donohue, Appelbaum, Park, Roberts, & Woldorff, 2013; Glaser & Glaser, 1982; Mattler, 2003; Wendt, Kiesel, Geringswald, Purmann, & Fischer, 2014). Although investigating the sources of this variability was not the aim of our study, we made use of it when considering whether certain variants of the attentional shift account, which posit a strong relationship between the CSE and the congruency effect, could explain our findings. 2. Experiment 1 The goal of Experiment 1 was to conduct a stronger test of the distracter head start hypothesis than in our prior study (Weissman et al., 2014). Thus, unlike in our prior study, we tested this hypothesis while holding the task constant, thereby strengthening our ability to draw conclusions about the role that processing the distracter before the target plays in driving the CSE. Specifically, we varied the ISI in a prime–probe arrow task (Kunde & Wuhr, 2006), wherein participants indicated the direction in which a small target arrow pointed while ignoring a large distracter arrow that pointed in either the same or the opposite direction. In some blocks of trials, the distracter and target arrows were separated by a 33 ms ISI (sequential trials). In others, these stimuli appeared at the same time (simultaneous trials), with the small arrow appearing inside the large arrow. We reasoned that the degree to which the distracter can activate a response before the target should be higher in sequential than in simultaneous trials. Thus, if this factor plays an important role in determining CSE magnitude, as posited by the distracter head start hypothesis, then the CSE should be larger in sequential than in simultaneous trials. 2.1. Methods 2.1.1. Participants Twenty-four undergraduate students from the University of Michigan (7 males, 17 females; mean age, 18.88 years, age range, 18–25 years) completed Experiment 1. All participants reported normal or correctedto-normal vision and hearing, and no history of head trauma, seizures, or neuropsychiatric illness. 2.1.2. Stimuli A fixation cross (0.8° × 0.8°) appeared at the center of the screen for two seconds at the beginning of each block. The distracter in each trial was a large, hollow arrow at the center of the screen that pointed in one of four directions: left, right, up, or down (horizontal orientation, 2.19° × 5.81°; vertical orientation, 5.81° × 2.19°). The target in each trial was a small, hollow arrow at the center of the screen that pointed left, right, up, or down (horizontal orientation, 0.83° × 1.87°; vertical orientation, 1.87° × 0.83°). By combining each of the four distracter arrows with only the two targets from the same orientation category (horizontal or vertical; e.g., the large leftward pointing distracter arrow was combined only with the small leftward and small rightward pointing target arrows), we were able to create eight unique distracter– target pairings. Four were congruent (left–left, right–right, up–up, down–down) and four were incongruent (left–right, right–left, up– down, down–up). All stimuli were created using Adobe Photoshop and appeared in white on a black background. We used PsychToolbox (Brainard, 1997) to present the stimuli and record participants' responses. 2.1.3. Task and design In each trial (duration, 2 s), one of the eight distracter–target pairings described earlier appeared at the center of the screen (see Fig. 1). Participants were instructed to identify the direction in which the small target arrow pointed as quickly and as accurately as possible. Specifically, participants were instructed to press “F” with their left middle finger to indicate a leftward-facing arrow, “G” with their left

11

index finger to indicate a rightward-facing arrow, “J” with their right middle finger to indicate an upward-facing arrow, and “N” with their right index finger to indicate a downward-facing arrow. In the event that a participant failed to correctly identify the small arrow (i.e., pressed an incorrect key) or neglected to respond within 1500 ms of target onset, the word “Error” appeared at the center of the screen for 200 ms. There were two types of blocks, which differed with regard to the timing of events in each trial. In sequential blocks (Fig. 1, top row), the sequence of events in each trial was as follows: distracter (133 ms), blank screen (33 ms), target (133 ms), and blank screen (1700 ms). In simultaneous blocks (Fig. 1, bottom row), the sequence of events in each trial was as follows: distracter plus target (133 ms) and blank screen (1867 ms). There were four 96-trial blocks for each of these “presentation modes” (i.e., sequential and simultaneous). We presented the trials in each block in a first-order counterbalanced sequence, which would ideally result in 24 congruent trials preceded by congruent trials (cC trials), 24 congruent trials preceded by incongruent trials (iC trials), 24 incongruent trials preceded by congruent trials (cI trials), and 24 incongruent trials preceded by incongruent trials (iI trials). However, one of these four trial types had only 23 trials because the first trial in each block was not preceded by a trial and was thus excluded from analyses of CSEs. The underrepresented trial type varied randomly across blocks. Finally, the experimental design controlled for two common learning and memory confounds: feature integration confounds and contingency learning biases. To prevent feature integration confounds, we excluded the possibility of feature repetitions in consecutive trials by alternating the dimension along which the arrows were oriented (horizontal or vertical) on every trial (Schmidt & Weissman, 2014; Weissman et al., 2014). To prevent contingency learning biases (Schmidt & De Houwer, 2011), we presented each congruent and each incongruent stimulus an equal number of times (i.e., 12 times) in each block. 2.1.4. Procedure Upon arriving at the laboratory, each participant gave informed written consent, filled out a brief screening form, and was assigned to an individual testing chamber. Participants were seated such that the distance between their eyes and the computer screen was 55 cm. Head position was stabilized with a chinrest. Next, participants were given verbal instructions about how to perform the task (see “Task and design”). Written instructions were also provided on the computer screen. Each presentation mode (sequential and simultaneous) consisted of a single 48-trial practice session and four 96-trial test blocks. Half of the participants performed the sequential blocks first and the simultaneous blocks second while the other half performed these blocks in the

Fig. 1. The prime–probe arrow task employed in Experiments 1 and 2. In some trials, the prime (large distracter arrow) and the probe (small target arrow) appeared simultaneously (Sim, top row). (b) In other trials, they appeared sequentially (Seq, bottom row). In the actual experiments, the arrows were drawn in white and appeared on a black background. The numbers beneath the boxes indicate the durations of individual trial components.

12

D.H. Weissman et al. / Acta Psychologica 156 (2015) 8–21

opposite order. The researcher remained in the room during the initial practice block to ensure that each participant could perform the task accurately. In the rare case that a participant misunderstood the instructions as indicated by numerous errors, a second practice block was administered. Participants were told they could rest between blocks. Following the study, participants were debriefed and received course credit as compensation. 2.1.5. Data analysis Before analyzing the mean RT and mean error rate data, we excluded several types of trials. First, practice trials and the first trial of each block were excluded from both analyses. Second, outliers (i.e. trials with RTs greater than 3 standard deviations away from their conditional mean), errors (incorrect responses and response omissions), and trials following errors were excluded from analyses of mean RT. Third, outliers and trials following errors were excluded from analyses of mean error rate (errors were retained for these analyses as mean error rate was the dependent measure). On average, 1.44% of the trials were outliers and 3.74% were errors. Following these exclusions, mean correct RT and mean percentage error rate were calculated for cC, cI, iC, and iI trials, separately in the sequential and simultaneous trial blocks. Separate repeated measures analyses of variance (ANOVAs) with three withinparticipants factors – presentation mode (simultaneous, sequential), current congruency (incongruent, congruent), and previous congruency (incongruent, congruent) – were then conducted on mean RT and mean error rate. Table 1 lists the mean RT, congruency effect, and congruency sequence effect (in ms) in the main conditions of Experiments 1–3. 2.2. Results 2.2.1. Mean RT There were two significant main effects. First, as expected, there was a main effect of current congruency, F(1,23) = 162.478, p b 0.001, η2p = 0.876, because mean RT was slower in incongruent trials (510 ms) than in congruent trials (452 ms). Second, there was a main effect of previous congruency, F(1, 23) = 4.784, p b 0.05, η2p = 0.172, because mean RT was slower when the previous trial was incongruent (482 ms) relative to congruent (479 ms), in line with previous reports of “postconflict slowing” (Ullsberger, Bylsma, & Botvinick, 2005). There were also three significant two-way interactions. First, as expected, there was an interaction between previous congruency and current congruency, F(1, 23) = 66.515, p b 0.001, η2p = 0.743, because the congruency effect was larger following congruent trials (70 ms) than following incongruent trials (47 ms). Second, there was an interaction between presentation mode and current congruency, F(1, 23) = 6.024, p b 0.05, η2p = 0.208: the congruency effect was larger in sequential

Table 1 Descriptive statistics (in ms) from Experiments 1–3. Mean RT

I–C

CSE

Experiment 1 Seq Sim

482 480

69 49

43 1

Experiment 2 SeqSeq SeqSim SimSeq SimSim

465 460 520 504

86 97 25 37

43 21 0 10

Experiment 3 33 ms ISI 1000 ms ISI

559 524

78 0

42 29

Notes: Seq = sequential trials (33 ms ISI). Sim = simultaneous trials (0 ms ISI). I–C = incongruent–congruent. CSE = congruency sequence effect, calculated as: (cI − cC) − (iI − iC).

(69 ms) than in simultaneous (49 ms) trials. Third, there was an interaction between previous congruency and presentation mode, F(1, 23) = 10.846, p b 0.005, η2p = 0.320, because the effect of previous congruency on mean RT (i.e., post-conflict slowing) was greater in sequential trials (8 ms) than in simultaneous trials (1 ms). Finally, consistent with the distracter head start hypothesis, there was a significant three-way interaction between presentation mode, previous congruency, and current congruency, F(1, 23) = 74.361, p b 0.001, η 2p = 0.764 (Fig. 2). This interaction occurred because the CSE was larger in sequential trials (43 ms; F(1, 23) = 85.882, p b 0.001, η2p = 0.789; Fig. 2, left) than in simultaneous trials (1.0 ms; F(1, 23) = .281, p b 0.601, η2p = 0.012; Fig. 2, right). Tests of simple effects for sequential trials revealed that mean RT was faster in cC (433 ms) than in iC (462 ms) trials, F(1, 23) = 76.316, p b 0.001, η2p = 0.768, and faster in iI trials (509 ms) than in cI (523 ms) trials, F(1, 23) = 21.670, p b 0.005, η2p = 0.485. No other effects were significant. 2.2.2. Mean error rate There were three main effects. First, as expected, there was a main effect of current congruency, F(1, 23) = 21.199, p b 0.001, η2p = 0.480, because mean error rate was higher in incongruent trials (4.8%) than in congruent trials (1.4%). Second, there was a main effect of previous congruency, F(1, 23) = 19.045, p b 0.001, η2p = 0.453, because there was a post-conflict reduction of mean error rate. In particular, mean error rate was lower when the previous trial was incongruent (2.3%) relative to congruent (3.9%). In combination with our finding of postconflict slowing in mean RT, this result suggests that participants traded speed for accuracy following incongruent trials, because they responded both more slowly and more accurately after incongruent trials than after congruent trials. Third, there was a main effect of presentation mode, F(1, 23) = 6.856, p b 0.025, η2p = 0.230, because mean error rate was higher in sequential trials (3.7%) than in simultaneous trials (2.6%). There were also three significant two-way interactions. First, as predicted, there was an interaction was between previous congruency and current congruency, F(1, 23) = 16.955, p b 0.001, η2p = 0.424: the congruency effect in mean error rate was smaller when the previous trial was incongruent (2.1%) relative to congruent (4.7%). Second, there was an interaction between presentation mode and current congruency, F(1, 23) = 8.755, p b 0.01, η2p = 0.276: the congruency effect was larger in sequential (4.9%) than in simultaneous (2.0%) trials. Third, there was an interaction between previous congruency and presentation mode, F(1, 23) = 6.450, p b 0.025, η2p = 0.219: the effect of previous congruency on mean error rate (i.e., the post-conflict reduction in mean error rate) was larger in sequential (2.4%) than in simultaneous trials (0.80%). Finally, consistent with the distracter head start hypothesis, there was a significant three-way interaction between presentation mode, previous congruency, and current congruency, F(1, 23) = 4.383, p b 0.05, η2p = 0.160. This interaction occurred because the CSE was larger in sequential trials (4.0%; F(1, 23) = 11.968, p b 0.005, η2p = 0.342) than in simultaneous trials (1.2%; F(1, 23) = 3.974, p = 0.058, η2p = 0.147). Tests of simple effects for sequential trials revealed that mean error rate did not differ between cC (1.5%) and iC (1.0%) trials, F(1, 23) = 2.046, p = 0.166, η2p = 0.082, but was higher in cI (8.4%) than in iI (3.9%) trials, F(1, 23) = 16.053, p = 0.001, η2p = 0.411. No other effects were significant. 2.3. Discussion The results of Experiment 1 support the distracter head start hypothesis, which posits that the degree to which the distracter can activate a response before the target is an important determinant of CSE magnitude. First, there was a robust CSE when the distracter appeared before the target in sequential trials but not when these stimuli appeared at the same time in simultaneous trials. Second, the CSE was

D.H. Weissman et al. / Acta Psychologica 156 (2015) 8–21

13

Fig. 2. Congruency sequence effects in Experiment 1. (a) In sequential trials, there was highly significant congruency sequence effect. (b) In simultaneous trials, we did not observe a congruency sequence effect.

significantly larger in sequential than in simultaneous trials. These findings reveal an important determinant of CSE magnitude when the typical learning and memory confounds are absent. They may also help to explain why the CSE is easier to observe in the prime–probe and Simon tasks than in the Stroop and flanker tasks. Indeed, as we explained in the introduction, the degree to which the distracter can activate a response before the target is higher in the former tasks than in the latter ones. The response modulation account can readily explain why the CSE was larger in sequential than in simultaneous trials. As noted earlier, both variants of this account assume that the CSE reflects a control process that modulates the response engendered by the distracter before the target response reaches threshold. The CSE should therefore be larger when this control process has sufficient time to modulate the response engendered by the distracter before the target response reaches threshold (sequential trials) than when it does not (simultaneous trials). Can the attentional shift account also explain why the CSE was larger in sequential than in simultaneous trials? The answer to this question depends on which variant of this account is considered. We therefore begin with the single variant that cannot explain this finding. The perceptual expectation hypothesis posits that participants widen or narrow spatial attention, respectively, after congruent and incongruent trials to include both the target and the distracter or only the target (Gratton et al., 1992). Since the distracter and target arrows occupied the same spatial locations in sequential and simultaneous trials, however, such adjustments of the attentional spotlight were equally possible in these trial types. It is therefore unclear why such adjustments would lead to a CSE in sequential trials but not in simultaneous trials. The remaining three variants of the attentional shift account can explain some or all of our findings in Experiment 1. The negative affect hypothesis posits that incongruent stimuli engender negative affect, which induces participants to expend greater effort toward optimizing performance (Dreisbach & Fischer, 2012). To explain our findings, this hypothesis would need to assume that incongruent stimuli evoke more negative affect in sequential than in simultaneous trials, which could have been the case because the congruency effect was larger in sequential than in simultaneous trials. It is unclear, however, why the CSE was completely absent in simultaneous trials, wherein a sizable congruency effect was also observed. The conflict monitoring hypothesis posits that experienced response conflict in the previous trial, as indexed by the size of the congruency effect, drives the CSE (Botvinick et al., 2001; Wendt et al., 2014; Yeung, Cohen, & Botvinick, 2011). Thus, this hypothesis is able to explain why the CSE was larger when the previous trial involved sequential as compared to simultaneous presentation. As with the negative affect hypothesis, though, it is unclear why a CSE was not also observed in simultaneous trials. Finally, the temporal attention hypothesis posits that participants allocate more or less attention to the moment at which the distracter appears, respectively, depending on whether the previous trial was congruent or incongruent.

Consistent with this view, we observed a CSE only when this temporal strategy was available in sequential trials. In sum, the perceptual expectation hypothesis is unable to explain the CSEs we observed, but the negative affect, conflict monitoring, and temporal attention hypotheses can explain some or all of these effects. Finally, we note that our findings are at least partially consistent with the adaptation by binding model. This model proposes that the CSE is driven by conflict-triggered increases of arousal, which strengthen associations between task-relevant stimulus features and the current task representation (Verguts & Notebaert, 2008). Given this assertion, the CSE should be greater when the congruency effect (an index of conflict) is relatively large as compared to relatively small. Our finding that the CSE was larger in sequential trials (wherein there was a relatively large congruency effect) than in simultaneous trials (wherein there was a relatively small congruency effect) is therefore consistent with the adaptation by binding model. However, this interpretation has difficulty explaining why a CSE was completely absent in simultaneous trials, wherein we still observed a significant congruency effect. 3. Experiment 2 In Experiment 2, we conducted a fine-grained test of the distracter head start hypothesis. Specifically, we investigated whether giving the distracter a head start in processing is helpful for (a) “triggering” control processes that engender a CSE in the previous trial, (b) “applying” control processes that engender a CSE in the current trial, or (c) both. To this end, we determined whether the CSE is larger when sequential, as compared to simultaneous, presentation is employed in the previous trial, the current trial, or both. To reveal the potentially distinct contributions of previous- and current-trial presentation mode (sequential, simultaneous) to the CSE, we manipulated the presentation mode on a trialby-trial basis. We reasoned that any outcome would be helpful for better understanding how processing the distracter before the target engenders a CSE. At the same time, observing an influence of sequential (versus simultaneous) presentation on CSE magnitude in both the previous and the current trial would further support the distracter head start hypothesis by highlighting the generality of the effect. 3.1. Methods 3.1.1. Participants Thirty-three undergraduate students from the University of Michigan (7 males, 25 females; mean age, 18.69, age range, 18–23) completed Experiment 2. One participant was excluded for responding accurately in less than 70% of the trials. As in Experiment 1, all participants reported normal or corrected-to-normal vision and hearing, and no history of head trauma, seizures, or neuropsychiatric illness. 3.1.2. Stimuli The stimuli were identical to those in Experiment 1.

14

D.H. Weissman et al. / Acta Psychologica 156 (2015) 8–21

3.1.3. Task and design The task and design were identical to those of Experiment 1 with only a few exceptions. First, we varied presentation mode (sequential, simultaneous) within – rather than between – blocks. Second, presentation mode was counterbalanced with congruency to create 16 trial types in each 96-trial block. Specifically, each of the four congruency sequences – cC, cI, iC, iI – occurred equally often in sequential followed by sequential trials (SeqSeq trials), sequential followed by simultaneous trials (SeqSim trials), simultaneous followed by sequential trials (SimSeq trials), and simultaneous followed by simultaneous trials (SimSim trials). Third, given the increased number of conditions, we included ten 96-trial test blocks, rather than eight. 3.1.4. Procedure The procedure was identical to that of Experiment 1, with the exception that participants completed just a single 32-trial practice block before starting the test trials. 3.1.5. Data analysis The data analysis was identical to that in Experiment 1 with the exception that there were 16 trial types. On average, 3.97% of the trials were errors and 1.40% of the trials were outliers. Mean RT and mean percent correct were analyzed using separate repeated measures analyses of variance (ANOVAs) with four within-participants factors: previous presentation mode (sequential, simultaneous), current presentation mode (sequential, simultaneous), previous congruency (congruent, incongruent) and current congruency (congruent, incongruent). 3.2. Results 3.2.1. Mean RT There were four significant main effects. First, there was a main effect of current congruency, F(1,31) = 260.474, p b 0.001, η2p = 0.894, because mean RT was slower in incongruent (518 ms) than in congruent (457 ms) trials. Second, there was a main effect of previous congruency, F(1, 31) = 28.180, p b 0.001, η2p = 0.476, because mean RT was slower when the previous trial was incongruent (490 ms) relative to congruent (484 ms), consistent with post-conflict slowing (Ullsberger et al., 2005). Third, there was a main effect of current presentation mode, F(1, 31) = 48.032, p b 0.001, η2p = 0.608, because mean RT was slower in sequential (492 ms) than in simultaneous (482 ms) trials. Finally, there was a main effect of previous presentation mode, F(1, 31) = 307.570, p b 0.001, η2p = 0.908: mean RT was faster when the previous trial involved sequential (462 ms) relative to simultaneous (512 ms) presentation. There were also five significant two-way interactions. First, as expected, there was an interaction between previous congruency and current congruency, F(1, 31) = 86.344, p b 0.001, η2p = 0.736, because the congruency effect was larger after congruent (70 ms) relative to incongruent (52 ms) trials. Second, there was an interaction between current presentation mode and current congruency, F(1, 31) = 23.844, p b 0.001, η2p = 0.435: unlike in Experiment 1, the congruency effect was larger in simultaneous (67 ms) than in sequential (55 ms) trials. Third, there was an interaction between previous presentation mode and current congruency, F(1, 31) = 107.129, p b 0.001, η2p = 0.776, because the congruency effect was larger when the previous trial involved sequential (91 ms) as compared to simultaneous (31 ms) presentation. Fourth, there was an interaction between previous presentation mode and current presentation mode, F(1, 31) = 9.719, p b 0.005, η2p = 0.239: participants responded more slowly in sequential than in simultaneous trials to a lesser extent when the previous trial involved sequential (5 ms) as compared to simultaneous (16 ms) presentation. Finally, there was an interaction between current presentation mode and previous congruency, F(1, 31) = 18.476, p b 0.001, η2p = 0.373: the effect of previous congruency on mean RT (i.e., post-conflict slowing) was greater in sequential (11 ms) than in simultaneous (2 ms) trials.

Finally, in line with the view that giving the distracter a head start is important for both triggering and applying control processes that engender a CSE, two higher-order interactions revealed that both previous and current presentation mode influenced CSE magnitude. First, there was a significant three-way interaction among previous presentation mode, previous congruency, and current congruency, F(1, 31) = 77.029, p b 0.001, η2p = 0.713: the CSE was larger when the previous trial involved sequential (32 ms) relative to simultaneous (5 ms) presentation. Second, there was a significant four-way interaction among previous presentation mode, current presentation mode, previous congruency, and current congruency, F(1, 31) = 15.357, p b 0.001, η2p = 0.331 (Fig. 3), because CSE magnitude varied with an interaction between previous presentation mode and current presentation mode. When the previous trial involved sequential presentation (Fig. 3, right), the CSE was larger in current sequential (43 ms; F(1, 31) = 114.524, p b 0.001, η2p = 0.787) than in current simultaneous (21 ms; F(1, 31) = 32.218, p b 0.001, η2p = 0.510) trials, F(1, 31) = 16.649, p b 0.001, η2p = 0.349. In contrast, when the previous trial involved simultaneous presentation (Fig. 3, left), the CSE did not differ between current sequential (0 ms; F(1, 31) = 0.001, p = 0.975, η2p = 0.00) and current simultaneous (10 ms; F(1, 31) = 5.224, p b 0.05, η2p = 0.144) trials, F(1, 31) = 2.946, p = .096, η2p = 0.087. In short, the CSE was larger when the previous trial involved sequential (versus simultaneous) presentation, and this effect was magnified when the current trial also involved sequential (versus simultaneous) presentation. No other effects were significant. 3.2.2. Mean error rate Mirroring the RT data, we observed four main effects. First, there was a main effect of current congruency, F(1, 31) = 47.201, p b 0.001, η2p = 0.604, because mean error rate was higher in incongruent (5.9%) than in congruent (1.1%) trials. Second, there was a main effect of previous congruency, F(1, 31) = 41.781, p b 0.001, η2p = 0.574. As in Experiment 1, there was a post-conflict reduction of mean error rate, because mean error rate was lower when the previous trial was incongruent (2.8%) relative to congruent (4.1%). Given that participants also responded more slowly when the previous trial was incongruent relative to congruent, this result suggests that participants traded speed for accuracy following incongruent trials. Third, there was a main effect of current presentation mode, F(1, 31) = 8.817, p b 0.01, η2p = 0.221: mean error rate was higher in simultaneous (3.9%) than in sequential (3.0%) trials. Finally, there was a main effect of previous presentation mode, F(1, 31) = 46.445, p b 0.001, η2p = 0.600, because mean error rate was higher when the previous trial involved sequential (5.6%) as compared to simultaneous (1.4%) presentation. There were five significant two-way interactions, the first four of which were also observed in the RT data. First, there was an interaction

Fig. 3. Congruency sequence effects in Experiment 2. The congruency sequence effect varied with an interaction between previous-trial presentation mode (SEQ, SIM) and current-trial presentation mode (SEQ, SIM). When the previous trial involved sequential presentation (right), the congruency sequence effect was larger when the current trial involved sequential as compared to simultaneous presentation. When the previous trial involved simultaneous presentation (left), the congruency sequence effect did not vary with whether the current trial involved sequential as compared to simultaneous presentation.

D.H. Weissman et al. / Acta Psychologica 156 (2015) 8–21

between previous congruency and current congruency, F(1, 31) = 36.373, p b 0.001, η2p = 0.540, because the congruency effect was smaller when the previous trial was incongruent (3.5%) relative to congruent (6.0%). Second, there was an interaction between current presentation mode and current congruency, F(1, 31) = 9.396, p b 0.005, η2p = 0.233: the congruency effect was larger in simultaneous (5.7%) than in sequential (3.9%) trials. Third, there was an interaction between previous presentation mode and current congruency, F(1, 31) = 43.652, p b 0.001, η2p = 0.585, because the congruency effect was larger when the previous trial involved sequential (9.3%) relative to simultaneous (0.20%) presentation. Fourth, there was an interaction between previous presentation mode and current presentation mode, F(1, 31) = 6.315, p b 0.025, η2p = 0.169, because the degree to which participants responded more accurately in sequential trials than in simultaneous trials was reduced when the previous trial involved simultaneous (0.40%) relative to sequential (1.5%) presentation. Finally, there was an interaction between previous presentation mode and previous congruency, F(1, 31) = 16.878, p b 0.001, η2p = 0.353: the post-conflict reduction of mean error rate was larger when the previous trial involved sequential (2.2%) as compared to simultaneous (0.40%) presentation. There were also two significant three-way interactions that were not critical for testing the distracter head start hypothesis, which were not observed in the RT data. First, there was an interaction among previous presentation mode, current presentation mode, and current congruency, F(1, 31) = 7.487, p = 0.01, η2p = 0.195, because the extent to which the congruency effect was larger in simultaneous than in sequential trials was greater when the previous trial involved sequential (3.1%) as compared to simultaneous (0.30%) presentation. Second, there was a three-way interaction among previous presentation mode, current presentation mode, and previous congruency, F(1, 31) = 4.415, p b 0.05, η2p = 0.125. When the previous trial involved sequential presentation, there was a larger post-conflict reduction of mean error rate when the current trial involved sequential (2.8%) relative to simultaneous (1.6%), presentation. In contrast, when the previous trial involved simultaneous presentation, then there was a smaller post-conflict reduction of mean error rate when the current trial involved sequential (0.30%) relative to simultaneous (0.50%) presentation. Finally, in line with the view that giving the distracter a head start is important both for triggering and applying control processes that engender a CSE, two significant higher-order interactions, which were also present in the RT data, revealed that both previous and current presentation mode influenced CSE magnitude. First, there was a three-way interaction among previous presentation mode, previous congruency, and current congruency, F(1, 31) = 21.774, p b 0.001, η2p = 0.413, because the CSE was larger when the previous trial involved sequential (4.4%) relative to simultaneous (0.40%) presentation. Second, there was a four-way interaction among previous presentation mode, current presentation mode, previous congruency, and current congruency, F(1, 31) = 7.149, p b 0.025, η2p = 0.187. When the previous trial involved sequential presentation, the CSE was larger in current sequential (6.6%; F(1, 31) = 24.716, p b 0.001, η2p = 0.444) than in current simultaneous (2.5%; F(1, 31) = 6.134, p b 0.025, η2p = 0.165) trials, F(1, 31) = 6.052, p = 0.02, η2p = 0.163. In contrast, when the previous trial involved simultaneous presentation, the CSE did not differ between current sequential (0.20%; F(1, 31) = 0.229, p = .636, η2p = 0.007) and current simultaneous (0.70%; F(1, 31) = 5.486, p b 0.03, η2p = 0.150) trials, F(1, 31) = 1.233, p = .275, η2p = 0.038. Thus, as in the RT data, the CSE was larger when the previous trial involved sequential (versus simultaneous) presentation, and this effect was enhanced when the current trial also involved sequential (versus simultaneous) presentation. No other effects were significant. 3.3. Discussion The results of Experiment 2 further reveal the importance of giving the distracter a head start on the ability to observe a CSE without the

15

typical learning and memory confounds. More specifically, they indicate that processing the distracter before the target to a high degree is important in (a) the previous trial for “triggering” control processes that engender a CSE and (b) the current trial for “applying” control processes that engender a CSE. First, the CSE was larger when the distracter appeared before, relative to with, the target in the previous trial. Second, this effect was greater when the current trial also involved sequential (relative to simultaneous) presentation. In short, sequential presentation in the previous trial interacted with sequential presentation in the current trial to engender the largest CSE. 3.3.1. Response modulation, attentional shift, or adaptation by binding? Is the interaction between previous and current presentation mode that we observed predicted by either variant of the response modulation account? At first blush, the interaction appears consistent with the activation–suppression hypothesis. Relative to simultaneous presentation, sequential presentation in the previous trial should allow more time for incorrect response activation to build before it is ultimately inhibited, which is important for “triggering” enhanced suppression (Ridderinkhof, 2002a). Sequential presentation in the current trial should then provide more time for control processes to inhibit the distracter response before it influences the target response (Burle et al., 2005). Thus, as we observed, the CSE should be largest when both the previous and the current trial involve sequential presentation. The interaction we observed may or may not be consistent with the response expectation hypothesis. Sequential presentation in the current trial should increase the CSE, because pre-activating a response (i.e., response priming), which leads to a CSE in this account, takes time (Vorberg, Mattler, Heinecke, Schmidt, & Schwarzbach, 2003). Whether sequential presentation in the previous trial should also increase the CSE, however, is less clear. On the one hand, a memory of whether the previous trial was congruent or incongruent, which helps to engender a CSE, might be stronger or more accurate after sequential trials, wherein the distracter and target can be separately attended and encoded, than after simultaneous trials, wherein the distracter and target must compete for attention and memory encoding processes. On the other hand, such a memory might be equally strong and accessible regardless of whether the previous trial involved sequential or simultaneous presentation. Thus, additional studies will be needed to determine whether the response expectation hypothesis can explain the interaction we observed. Can the attentional shift account explain the CSEs we observed? As in Experiment 1, the perceptual expectation hypothesis is unable to explain why there was a larger CSE in sequential than in simultaneous trials. Unlike in Experiment 1, however, the negative affect and conflict monitoring hypotheses are also unable to explain this effect, because the congruency effect in Experiment 2 was larger in simultaneous as compared to sequential trials. Notably in this regard, these hypotheses assume the size of the congruency effect in the previous trial is an index of negative affect (Dreisbach & Fischer, 2012) or response conflict (Yeung et al., 2011), respectively, which have been proposed to drive the CSE. Given these assumptions, the CSE should have been larger after simultaneous trials, wherein the congruency effect was relatively large, than after sequential trials, wherein the congruency effect was relatively small. Finally, as in Experiment 1, the temporal attention hypothesis is able to explain why the CSE was larger in sequential than in simultaneous trials. If one assumes that employing the same strategy on consecutive trials aids the implementation of that strategy (Rogers & Monsell, 1995), then this hypothesis might also explain why the CSE was largest when both the previous and the current trial involved sequential presentation. Thus, the temporal attention hypothesis is the only variant of the attentional shift account that appears able to explain the CSEs we observed in both Experiments 1 and 2. It is also important to consider whether the adaptation by binding model can explain the CSEs we observed. Recall that this model asserts that the CSE should increase with the size of the congruency effect

16

D.H. Weissman et al. / Acta Psychologica 156 (2015) 8–21

(Verguts & Notebaert, 2008). As mentioned above, however, we observed the opposite result. Specifically, the CSE was larger when the previous trial involved sequential as compared to simultaneous presentation, even though the congruency effect was larger in simultaneous than in sequential trials. This pattern is difficult to reconcile with the adaptation by binding interpretation of the CSE. 3.3.2. Other potential explanations of the CSEs we observed We next considered whether the CSEs we observed fit with the task representation view of the CSE, in which a CSE is observed only when two consecutive trials involve the same task set (Hazeltine et al., 2011). If participants perceive sequential and simultaneous trials as different tasks, then the CSE should be larger when the presentation mode (sequential, simultaneous) remains constant across two consecutive trials than when it varies. In line with this view, the CSE was larger when a sequential trial preceded another sequential trial than when a sequential trial preceded a simultaneous trial. However, the CSE was not significantly larger when a simultaneous trial preceded another simultaneous trial than when a simultaneous trial preceded a sequential trial. The CSE also was not significantly larger when a simultaneous trial preceded another simultaneous trial than when a sequential trial preceded a simultaneous trial. Finally, in Experiment 1, we did not observe a significant CSE in blocks containing only simultaneous trials, even though every trial in such blocks should have been perceived as being part of the same task. Thus, even if participants viewed sequential and simultaneous trials as different tasks, the task representation view of the CSE could not entirely explain our findings. We also considered whether the pattern of CSEs we observed indexed episodic retrieval processes that are sensitive to whether the presentation mode remains constant or varies across two consecutive trials. In the episodic retrieval view of the CSE (Spape & Hommel, 2008), the repetition of a stimulus feature enhances CSE magnitude, at least in part, by cueing the nature of control processes that are expected to optimize performance in an upcoming trial. As in the task representation account of the CSE described earlier, this view predicts that the CSE will be larger when the presentation mode (a salient stimulus feature) repeats across two consecutive trials than when it does not. Therefore, the findings above indicating that the task representation view cannot fully explain our findings also weigh against the episodic retrieval view. Still, randomly intermixing sequential and simultaneous trials in Experiment 2 clearly influenced CSE magnitude. For example, unlike in Experiment 1, a small CSE was observed even when both the previous and the current trial involved simultaneous presentation. We therefore acknowledge that the task representation and/or episodic retrieval view might partially (although not completely) explain the CSEs we observed in Experiment 2. Future studies might formally investigate these possibilities. 4. Experiment 3 The results of Experiments 1 and 2 ruled out three variants of the attentional shift account. However, the fourth variant – the temporal attention hypothesis – could not be distinguished from the two variants of the response modulation account. Therefore, the goal of Experiment 3 was to more definitively tease apart the predictions of the temporal attention hypothesis from those of the response modulation account. To this end, we investigated whether a negative, or reverse, congruency effect can be observed after incongruent trials when a 1 s ISI separates the distracter and target.3 When employing such a long ISI, the overall congruency effect should be minimal and there should be ample time to (a) inhibit the response signaled by the current-trial distracter 3 Although negative congruency effects have previously been observed following incongruent trials (Ridderinkhof, 2002a; Wylie et al., 2010), to our knowledge this has always been in the context of feature repetition and/or contingency learning confounds that preclude firm conclusions about the contributions of cognitive control to the CSE.

(Burle et al., 2005; Machado et al., 2007) or (b) pre-activate either the same response if the previous trial was congruent or the opposite response if the previous trial was incongruent (Gratton et al., 1992; Logan, 1985). As we explain below, either process should lead to slower responses in congruent than in incongruent trials, resulting in a negative congruency effect after incongruent trials. In contrast, no variant of the attentional shift account predicts a negative congruency effect after incongruent trials. Indeed, even allocating all of one's attention to the target and none to the distracter after a previous incongruent trial might eliminate the congruency effect but not reverse it. Experiment 3 therefore provided a crucial test of whether the CSE is better explained by the response modulation or by the attentional shift account. We now explain why the response modulation account predicts a negative congruency effect after incongruent trials when a long ISI separates the distracter and the target and the overall congruency effect is minimal. First, consider the activation–suppression hypothesis (Burle, Spieser, Servant, & Hasbroucq, 2013; Ridderinkhof, 2002a, 2002b; van den Wildenberg et al., 2010; Wylie, Ridderinkhof, Bashore, & van den Wildenberg, 2010). Inhibiting the response associated with the distracter (e.g., “Left”) slows performance in congruent trials, because responding correctly to the congruent target (e.g., “Left”) requires overcoming inhibition of this response (Burle et al., 2005). In contrast, inhibiting the response associated with the distracter (e.g., “Left”) does not slow performance in incongruent trials, because responding correctly to the incongruent target (e.g., “Right”) does not require overcoming inhibition of this response. Thus, participants should respond more slowly in congruent than in incongruent trials, leading to a negative congruency effect. Second, consider the response expectation hypothesis (Logan, 1985). Here, participants pre-activate the response opposite to the one signaled by the distracter after incongruent trials. Thus, they should respond more quickly in incongruent than in congruent trials, leading to a negative congruency effect. In sum, unlike the four variants of the attentional shift account, both variants of the response modulation account predict a negative congruency effect after incongruent trials when a long ISI separates the distracter and the target. Finally, it is important to note that we employed letters as stimuli in Experiment 3 rather than arrows. We did so because it has been suggested that mental rotation confounds unique to arrows might engender a CSE in the prime–probe arrow task (Kunde & Wuhr, 2006). Although the lack of a CSE in simultaneous trials of Experiment 1 weighs against this possibility, we wished to employ a paradigm in which potential mental rotation confounds were completely absent. We therefore employed a prime–probe letter task because, unlike with arrows, it is impossible to rotate one letter of the alphabet (e.g., A) to make it look a different letter (e.g., Z).4 4.1. Methods 4.1.1. Participants Thirty-two undergraduate students from the University of Michigan (13 males, 19 females; mean age, 18.72 years, age range, 18–21 years) participated in Experiment 3. All participants reported normal or corrected-to-normal vision and hearing, and no history of head trauma, seizures, or neuropsychiatric illness. 4.1.2. Stimuli The stimuli were the letters A, B, Y, and Z. The distracters were relatively large versions of these four letters: A (2.08° × 2.29°), B (2.08° × 2.08°), Y (2.08° × 2.08°), and Z (2.08° × 1.67°). The targets 4 Recent findings indicate that the inhibitory mechanisms posited by the activation– suppression account also operate in the Eriksen letter flanker task (Burle et al., 2013), even though the distracter in this task is not “automatically” mapped to a response as envisioned by a traditional dual-route architecture (Kornblum & Lee, 1995). Thus, similar to the indirect (i.e., target) route in this architecture, the direct (i.e., distracter) route may be able to translate stimuli into responses using arbitrary, task-dependent stimulus– response mappings (Ridderinkhof, van der Molen, & Bashore, 1995).

D.H. Weissman et al. / Acta Psychologica 156 (2015) 8–21

were relatively small versions of these four letters: A (1.25° × 1.56°), B (1.25° × 1.25°), Y (1.25° × 1.25°), and Z (1.25° × 1.04°). 4.1.3. Task and design The task and design were similar to those in Experiment 1 with five exceptions. First, the leftward, rightward, upward, and downward pointing arrows in Experiments 1 and 2 were replaced, respectively, with the letters A, B, Y, and Z. Thus, rather than alternating across trials between “left–right” and “up–down” stimulus sets as in Experiments 1 and 2, participants alternated between “A–B” and “Y–Z” stimulus sets. Second, rather than presenting simultaneous and sequential trials in different blocks, we presented two types of sequential trials in different blocks. These two types of sequential trials were distinguished by the ISI separating the distracter and the target (33 ms versus 1000 ms). Third, as the 1000 ms ISI was quite long, trial duration was increased to 3 s in both ISI conditions to provide participants with enough time to respond and prepare for the next trial. Fourth, to minimize fatigue, we included 64 trials per block (rather than 96). Fifth, we increased the stimulus duration from 133 ms to 200 ms to precisely match both the stimulus duration and the ISI in the 1000 ms ISI condition to those from a prior study in which robust effects of response expectancies on performance were observed (Gratton et al., 1990). Given the task parameters above, each ISI condition involved a sequence of four events (Fig. 4). In the 33 ms ISI condition, participants viewed the distracter for 200 ms, a blank screen for 33 ms, the target for 200 ms, and a second blank screen for 2567 ms. In the 1000 ms ISI condition, participants viewed the distracter for 200 ms, a blank screen for 1000 ms, the target for 200 ms, and a second blank screen for 1600 ms. Finally, as in Experiments 1 and 2, we presented the trials in each of the 64-trial test blocks in a counterbalanced order to ensure roughly equal numbers of cC trials, iC trials, cI, and iI trials in each block. More specifically, there were always 16 trials in three of these four conditions and 15 trials in the fourth condition. As in Experiments 1 and 2, one condition was underrepresented because the first trial in each block was not preceded by a trial and was therefore excluded from analyses of CSEs. The underrepresented condition varied randomly across blocks. 4.1.4. Procedure The procedure was identical to that of Experiment 1 with three exceptions. First, participants completed only 24 practice trials prior to the set of five test blocks in each ISI condition. Second, given that we reduced the number of test trials per block from 96 to 64 (see above), we included five – rather than four – test blocks for each ISI condition. Third, analogous to how we counterbalanced the order of sequential and simultaneous trials in Experiment 1, half of the participants

Fig. 4. The prime–probe letter task employed in Experiment 3. In some blocks, the prime (large letter) and the probe (small letter) were separated by a 33 ms inter-stimulus-interval (ISI) (top row). (b) In other blocks, they were separated by a 1000 ms ISI (bottom row). In the actual experiment, the letters were drawn in white and appeared on a black background. The numbers beneath the boxes indicate the durations of individual trial components.

17

performed the 33 ms ISI blocks first and the 1000 ms ISI blocks second while the other half performed these blocks in the opposite order. 4.1.5. Data analysis The data analysis was analogous to that in Experiment 1. On average, 4.09% of the trials were errors and 1.32% of the trials were outliers. Separate repeated-measures ANOVAs were employed to analyze the mean RT and mean percent error data with the following within-participants factors: ISI (33 ms, 1000 ms), current congruency (incongruent, congruent), and previous congruency (incongruent, congruent). 4.2. Results 4.2.1. Mean RT There were two significant main effects. First, as expected, there was a main effect of current congruency, F(1,31) = 49.037, p b 0.001, η2p = 0.613, because mean RT was slower in incongruent trials (560 ms) than in congruent trials (521 ms). Second, there was a main effect of ISI, F(1, 31) = 18.259, p b 0.001, η2p = 0.371, because mean RT was slower in the 33 ms ISI condition (558 ms) than in the 1000 ms ISI condition (524 ms). There were also two significant two-way interactions. First, there was an interaction between ISI and current congruency, F(1, 31) = 80.285, p b 0.001, η2p = 0.721: as expected, the congruency effect was larger in the 33 ms ISI condition (78 ms) than in the 1000 ms ISI condition (0 ms) trials. Second, and also expected, there was an interaction between previous congruency and current congruency, F(1, 31) = 60.816, p b 0.001, η2p = 0.662, because the congruency effect was larger following congruent trials (57 ms) than following incongruent trials (21 ms). Tests of simple effects revealed that mean RT was faster in cC (512 ms) than in iC (531 ms) trials, F(1, 31) = 14.089, p = 0.001, η 2p = 0.312, and faster in iI (552 ms) than in cI (569 ms) trials, F(1, 31) = 32.411, p b 0.001, η2p = 0.511. As illustrated in Fig. 5a and b, the two-way interaction between previous congruency and current congruency was significant in both the 33 ms ISI condition, F(1, 31) = 52.278, p b 0.001, η2p = 0.628, and the 1000 ms ISI condition, F(1, 31) = 21.733, p b 0.001, η2p = 0.412. We therefore analyzed the data from each ISI condition separately. As expected, tests of simple effects in the 33 ms ISI condition revealed that mean RT was faster in cC (508 ms) than in iC (530 ms) trials, F(1, 31) = 25.536, p b 0.001, η2p = 0.432, and slower in cI (607 ms) than in iI (587 ms) trials, F(1, 31) = 19.762, p b 0.001, η2p = 0.389. Furthermore, there was a positive congruency effect after both congruent trials (99 ms; F(1, 31) = 120.218, p b 0.001, η2p = 0.795) and incongruent trials (57 ms; F(1, 31) = 37.650, p b 0.001, η2p = 0.548). Also as expected, tests of simple effects in the 1000 ms ISI condition revealed that mean RT was faster in cC (516 ms) than in iC (532 ms) trials, F(1, 31) = 6.510, p b 0.02, η2p = 0.174, and slower in cI (530 ms) than in iI (517 ms) trials, F(1, 31) = 12.712, p = 0.001, η2p = 0.291. Further, in line with the response modulation account, there was a positive congruency effect after congruent trials (14 ms; F(1, 31) = 5.324, p b 0.05, η2p = 0.147) and a negative congruency effect after incongruent trials (−15 ms; F(1, 31) = 7.239, p b 0.02, η2p = 0.189). Critically, an interaction contrast revealed that the congruency effect after incongruent trials was more negative in the 1000 ms ISI condition (−15 ms) than in the 33 ms ISI condition (57 ms), F(1, 31) = 70.785, p b 0.001, η2p = 0.695. Thus, as predicted, the negative congruency effect was both significantly larger in the 1000 ms ISI condition than in the 33 ms ISI condition and specific to the 1000 ms ISI condition. No other effects were significant. 4.2.2. Mean error rate There were two significant two-way interactions. First, as expected, there was an interaction between ISI and current congruency, F(1, 31) = 8.116, p b 0.01, η2p = 0.207, because the congruency effect was larger in the 33 ms ISI condition (11.0%) than in the 1000 ms ISI condition (−6.0%). Second, and also expected, there was an interaction

18

D.H. Weissman et al. / Acta Psychologica 156 (2015) 8–21

Fig. 5. Congruency sequence effects in Experiment 3. We observed a robust congruency sequence effect whose magnitude did not differ between the (a) 33 ms and (b) 1000 ms ISI conditions. In the 1000 ms ISI condition, we observed a positive congruency effect after congruent trials (left side of panel b) and a negative congruency effect after incongruent trials (right side of panel b).

between previous congruency and current congruency, F(1, 31) = 4.563, p b 0.05, η2p = 0.128, because the congruency effect was smaller when the previous trial was incongruent (− 3.0%) relative to congruent (7.0%). The two-way interaction between previous congruency and current congruency was not significant for either ISI condition in isolation (33 ms ISI: F(1, 31) = 2.012, p = 0.166, η2p = 0.061; 1000 ms ISI : F(1, 31) = 2.885 = 0.101, η2p = 0.085). Therefore, we did not conduct tests of simple effects in these conditions.

4.3. Discussion The results of Experiment 3 confirm an important prediction of the response modulation account. Specifically, in the 1 s ISI condition, we observed both (a) no overall congruency effect and (b) a significant negative congruency effect following incongruent trials. As described earlier, a negative congruency effect following incongruent trials is consistent with both variants of the response modulation account but is not predicted by any variant of the attentional shift account. Indeed, even if participants allocated all of their attention to the target after incongruent trials, the result would be a lack of a congruency effect, rather than a negative congruency effect. Our findings in Experiment 3 therefore indicate that the CSE is more consistent with the response modulation account than with the attentional shift account. Finally, we note that the CSEs we observed in Experiment 3 also cannot be explained by the adaptation by binding model. Specifically, although the model posits that the CSE stems from a conflict-triggered increase of arousal (Verguts & Notebaert, 2008), we observed a significant CSE in the 1 s ISI condition in the complete absence of an overall congruency effect. It therefore appears that a conflict-triggered increase of arousal is not necessary for a CSE to emerge in the present tasks.

5. General discussion The present study had two goals. First, we wished to conduct a stronger test of the distracter head start hypothesis than we conducted in our prior study (Weissman et al., 2014). Second, we wished to investigate whether the CSE, when measured without the typical learning and memory confounds, is more consistent with the response modulation account than with the attentional shift account. The results of Experiments 1 and 2 provided novel support for the distracter head start hypothesis while the results of Experiment 3 revealed that the CSE is more consistent with the response modulation account than with the attentional shift account. The results of Experiments 2 and 3 also weighed against an adaptation by binding interpretation of the CSEs we observed. These findings have important implications for current theories of the CSE and, more broadly, for our understanding of how cognitive control processes minimize distraction from irrelevant stimuli.

5.1. The CSE is greater with sequential versus simultaneous presentation of distracters and targets In a previous study, we suggested that the degree to which the distracter can activate a response before the target may be an important determinant of whether a CSE is observed without the typical learning and memory confounds (Weissman et al., 2014). To test this hypothesis in Experiment 1, we manipulated across blocks whether the distracter arrow in a prime–probe arrow task appeared before the target arrow (sequential trials) or at the same time as the target arrow (simultaneous trials). We reasoned that the degree to which the distracter can activate a response before the target should be higher in sequential than in simultaneous trials. We therefore predicted that the CSE would be larger in sequential than in simultaneous trials. As predicted by the distracter head start hypothesis, the CSE was significantly larger in sequential than in simultaneous trials. Moreover, we observed a significant CSE in sequential but not in simultaneous trials. These findings may help to explain why the CSE is often difficult to observe in the Stroop and flanker tasks when feature repetition and contingency learning confounds are absent (Mayr et al., 2003; Schmidt & De Houwer, 2011). Indeed, the degree to which the distracter is processed before the target is lower in these tasks than in the prime– probe task, because only in the prime–probe task does the distracter appear before the target. Our findings also inform prior results indicating the absence of a CSE when participants alternate between tasks employing horizontal and vertical stimuli, especially when distracters and targets appear simultaneously (Lee & Cho, 2013). Specifically, our findings suggest the absence of a CSE in such tasks stems from an inability to process the distracter before the target. This view contrasts with the suggestion that distinct control processes underlie CSEs for horizontal and vertical stimuli, such that an incongruent stimulus in one task (e.g., horizontal) cannot engender a reduced congruency effect in the other task (e.g., vertical) (Kim & Cho, 2014; Lee & Cho, 2013). In Experiment 2, we investigated whether varying the degree to which the distracter can activate a response before the target in the previous trial, the current trial, or both influences CSE magnitude. To this end, we randomly mixed sequential and simultaneous trials within the same blocks. Critically, the CSE was larger when the previous trial involved sequential as compared to simultaneous presentation, and this effect was amplified when the current trial also involved sequential as compared to simultaneous presentation. These findings provide further support for the distracter head start hypothesis by showing that processing the distracter before the target is important for both “triggering” and “applying” control processes that engender the CSE. 5.2. Support for the response modulation account of the “control-driven” CSE While the results of Experiments 1 and 2 weighed against three out of four variants of the attentional shift account and against the adaptation

D.H. Weissman et al. / Acta Psychologica 156 (2015) 8–21

by binding model, the temporal attention hypothesis remained indistinguishable in its predictions from the two variants of the response modulation account. In Experiment 3, we therefore sought to more definitively distinguish between the response modulation and attentional shift accounts of the CSE. To this end, we investigated whether the CSE is associated with a negative congruency effect after incongruent trials when a relatively long (i.e., 1 s) ISI separates the distracter and the target. Based on prior work (Burle et al., 2005; Machado et al., 2007), we predicted that the overall congruency effect would be eliminated under such conditions. Therefore, a negative congruency effect following incongruent trials would be consistent with both variants of the response modulation account. In contrast, no variant of the attentional shift account predicts a negative congruency effect, because even focusing all of one's attention on the target and none on the distracter might eliminate the congruency effect but not reverse it. Critically, we observed a significant CSE in the 1 s ISI condition, which was associated with a positive congruency effect after congruent trials and with a negative congruency effect after incongruent trials. These results – particularly the negative congruency effect after incongruent trials – indicate that the CSE is more consistent with the response modulation account than with the attentional shift account. It is important to note that other studies of the CSE have also reported a negative congruency effect following incongruent trials (Ridderinkhof, 2002a; Wylie et al., 2010). To our knowledge, however, such an effect has not been reported in the absence of feature repetition and contingency learning confounds that are known to influence the CSE. One study that came close to demonstrating such an effect without these confounds involved the Simon task (Lee & Cho, 2013). The authors observed a positive congruency effect following congruent trials and a negative congruency effect following incongruent trials in the complete absence of stimulus repetitions, response repetitions, and contingency learning confounds. However, since there were only two response categories (“Left” and “Right”), congruency repetitions across consecutive trials were either complete repetitions or complete alternations at the categorical level (i.e., the distracter and target categories both repeated or both alternated) while congruency alternations across consecutive trials were always partial repetitions (i.e., either the target category or the distracter category alternated, but not both). Thus, the negative congruency effect may have resulted from categorical repetition confounds. Notably, our study contained no such confounds as the stimulus category (e.g., “Left”, “Right”, “Up”, or “Down”) alternated on every trial. It is also important to consider whether the reversal of the congruency effect that we observed in the 1000 ms ISI condition of Experiment 3 reflects the same process or processes that engender a CSE at shorter ISIs (e.g., 33 ms) in the prime–probe task. Consistent with the recruitment of similar processes, CSE magnitude in Experiment 3 did not differ between the 33 ms and 1000 ms ISI conditions, even though the negative congruency effect after incongruent trials was specific to the latter condition. It would therefore appear that our use of a relatively long ISI revealed important information about the nature of the “controldriven” CSE without fundamentally altering the processes that contribute to this phenomenon. Consistent with this view, manipulating the ISI between the distracter and the target in distracter interference tasks has revealed a great deal about the nature of congruency effects, even when the ISI was relatively long (Appelbaum et al., 2012; Eriksen & Schultz, 1979; Machado et al., 2007). Our findings above may appear to contradict evidence from fMRI suggesting that an attentional shift toward the target and/or away from the distracter occurs in incongruent relative to congruent trials (Egner & Hirsch, 2005; Polk, Drake, Jonides, Smith, & Smith, 2008; Weissman, Warner, & Woldorff, 2004). One possible explanation is that, in the prime–probe task, an attentional shift facilitates conflict resolution within a single trial, but does not contribute to the CSE. Another is that an attentional shift is one of many processes that contribute to the CSE when the distracter is presented before the target. Future brain imaging studies could test this hypothesis by determining whether

19

activity in regions of the sensory cortex that underlie perceptual aspects of distracter processing is enhanced when the previous trial was congruent as compared to incongruent, using a design that is modeled on the confound-minimizing protocol we employed here. Whatever the outcome, our findings indicate that the attentional shift account cannot fully explain the CSE. For example, as discussed earlier, it cannot explain the negative congruency effect in Experiment 3. 5.3. A potential alternative explanation of the present findings While the present findings appear more consistent with the response modulation account than with the attentional shift account, it is important to consider whether an alternative account of the CSE might also explain our findings. According to the temporal learning hypothesis (Schmidt, 2013), the CSE reflects expectations about when to respond, which differ as a function of whether RT in the previous trial was relatively fast (as in congruent trials) or relatively slow (as in incongruent trials). More specifically, participants expect to respond in the current trial at around the same time they responded in the previous trial and temporarily reduce the threshold for responding at this time. This brief reduction of the threshold leads to a relatively fast response if evidence for a decision has accumulated to a level that is close to the reduced threshold. This will often be the case when the congruency of the current trial matches the congruency of the previous trial. In contrast, such a benefit will not usually occur when previous- and current-trial congruency mismatch. Mean RT will therefore be faster when congruency repeats (i.e., in cC and iI trials) than when it does not repeat (i.e., in iC and cI trials), thereby engendering a CSE. Of importance, our findings in Experiment 3 weigh against this alternative interpretation of the CSE in the prime–probe task. Specifically, in the 1 s ISI condition, we observed a CSE in the complete absence of a difference in mean RT between incongruent and congruent trials, which should be necessary for congruency-based temporal learning to occur. Further, CSE magnitude in the 1 s ISI condition did not differ from that in the 33 ms ISI condition, in which there was a robust 78 ms congruency effect. Thus, inconsistent with the temporal learning hypothesis, CSE magnitude was not closely linked to the size of the congruency effect. 5.4. Implications for future studies The present findings have two important implications for future behavioral and brain imaging studies investigating the CSE. First, complementing our recent prior work (Schmidt & Weissman, 2014; Weissman et al., 2014), they indicate an approach for observing robust CSEs without feature integration or contingency learning confounds. As discussed in several prior reviews of the literature (Egner, 2007; Larson, Clayson, & Clawson, 2014; Schmidt, 2013), the vast majority of previous behavioral, fMRI, EEG, and lesion studies of the CSE contain one or both of these confounds. Therefore, by providing a protocol for observing robust CSEs without such confounds, our work may aid future attempts to isolate control processes that minimize the influence of irrelevant stimuli on performance, including their underlying neural mechanisms. Second, our findings provide new insights into the nature of control processes that drive the CSE. Specifically, they suggest that the CSE is driven by a control process that modulates the response signaled by the distracter, which is recruited most effectively when the distracter is processed before the target. Further investigations of the response modulation account may therefore provide important insights into how people cope with distraction. 5.5. Limitations The present study has three main limitations. First, although our findings appear most consistent with the response modulation account of the CSE, the behavioral data were not accompanied by corresponding

20

D.H. Weissman et al. / Acta Psychologica 156 (2015) 8–21

neural or electromyographic (EMG) measures of response activation. Such measures would provide converging support for the view that CSE magnitude is influenced by the degree to which the distracter activates a response before the target when the typical learning and memory confounds are absent. We note, however, that prior work has already established that the distracter response is activated before the target response in the prime–probe task (Eimer & Schlaghecken, 1998). Second, the present data do not allow us to distinguish between the activation–suppression and response expectation variants of the response modulation account. Thus, additional studies will be needed to more precisely identify the control processes that produced the CSEs we observed. Third, while the distracter head start hypothesis describes an important determinant of CSE magnitude, it may not describe a critical boundary condition for observing the CSE. Indeed, we observed a small CSE in Experiment 2 even when both the previous and the current trial involved simultaneous presentation of distracter and target stimuli. Similarly, recent findings indicate that a CSE can be observed in a color flanker task with simultaneous presentation and without the typical learning and memory confounds if two tasks share the same response mode (Kim & Cho, 2014). Thus, additional studies will be needed to flesh out other facilitating conditions and potential control processes that engender a CSE in the absence of the typical learning and memory confounds.

6. Conclusion The present findings indicate that the CSE is more easily observed without the typical confounds when the degree to which a distracter can activate a response before a target is relatively high as compared to relatively low. They also indicate that the CSE is more consistent with a modulation of the response engendered by the distracter than with a shift of attention toward the target. Future studies investigating the role that processing the distracter before the target plays in producing the CSE, as well as different possible accounts of the CSE, may further elucidate the cognitive control processes that minimize distraction from irrelevant stimuli.

References Appelbaum, L.G., Boehler, C.N., Won, R., Davis, L., & Woldorff, M.G. (2012). Strategic allocation of attention reduces temporally predictable stimulus conflict. Journal of Cognitive Neuroscience, 24(9), 1834–1848. http://dx.doi.org/10.1162/jocn_a_00209. Blais, C., & Verguts, T. (2012). Increasing set size breaks down sequential congruency: Evidence for an associative locus of cognitive control. Acta Psychologica, 141(2), 133–139. http://dx.doi.org/10.1016/j.actpsy.2012.07.009 (S0001-6918(12)00121-7 [pii]). Botvinick, M.M., Braver, T.S., Barch, D.M., Carter, C.S., & Cohen, J.D. (2001). Conflict monitoring and cognitive control. Psychological Review, 108(3), 624–652. Botvinick, M.M. (2007). Conflict monitoring and decision-making: reconciling two perspectives on anterior cingulate function. Cognitive, Affective, and Behavioral Neuroscience, 7(4), 356–366. Brainard, D.H. (1997). The psychophysics toolbox. Spatial Vision, 10(4), 433–436. Burle, B., Spieser, L., Servant, M., & Hasbroucq, T. (2013). Distributional reaction time properties in the Eriksen task: Marked differences or hidden similarities with the Simon task? Psychonomic Bulletin and Review. http://dx.doi.org/10.3758/s13423013-0561-6. Burle, B., van den Wildenberg, W.P.M., & Ridderinkhof, K.R. (2005). Dynamics of facilitation and interference in cue-priming and Simon tasks. European Journal of Cognitive Psychology, 17, 619–641. Donohue, S.E., Appelbaum, L.G., Park, C.J., Roberts, K.C., & Woldorff, M.G. (2013). Crossmodal stimulus conflict: The behavioral effects of stimulus input timing in a visual– auditory Stroop task. PloS One, 8(4), e62802. http://dx.doi.org/10.1371/journal.pone. 0062802 (PONE-D-12-26881 [pii]). Dreisbach, G., & Fischer, R. (2012). Conflicts as aversive signals. Brain and Cognition, 78, 94–98. Duthoo, W., Wuhr, P., & Notebaert, W. (2013). The hot-hand fallacy in cognitive control: Repetition expectancy modulates the congruency sequence effect. Psychonomic Bulletin and Review, 20(4), 798–805. http://dx.doi.org/10.3758/s13423-013-0390-7. Egner, T. (2007). Congruency sequence effects and cognitive control. Cognitive, Affective, & Behavioral Neuroscience, 7(4), 380–390. Egner, T., & Hirsch, J. (2005). Cognitive control mechanisms resolve conflict through cortical amplification of task-relevant information. Nature Neuroscience, 8(12), 1784–1790.

Eimer, M., & Schlaghecken, F. (1998). Effects of masked stimuli on motor activation: Behavioral and electrophysiological evidence. Journal of Experimental Psychology: Human Perception and Performance, 24, 1737–1747. Eriksen, B.A., & Eriksen, C.W. (1974). Effects of noise letters upon the identification of a target letter in a nonsearch task. Perception & Psychophysics, 16(1), 143–149. Eriksen, C.W., & Schultz, D.W. (1979). Information processing in visual search: A continuous flow conception and experimental results. Perception & Psychophysics, 25(4), 249–263. Glaser, M.O., & Glaser, W.R. (1982). Time course analysis of the Stroop phenomenon. Journal of Experimental Psychology: Human Perception and Performance, 8(6), 875–894. Gratton, G., Bosco, C.M., Kramer, A.F., Coles, M.G.H., Wickens, C.D., & Donchin, E. (1990). Event-related potentials as indices of information extraction and response priming. Electroencephalography and Clinical Neurophysiology, 75, 419–432. Gratton, G., Coles, M.G.H., & Donchin, E. (1992). Optimizing the use of information: Strategic control of activation and responses. Journal of Experimental Psychology: General, 4, 480–506. Gratton, G., Coles, M.G.H., Sirevaag, E.J., Eriksen, C.W., & Donchin, E. (1988). Pre- and poststimulus activation of response channels: A psychophysiological analysis. Journal of Experimental Psychology: Human Perception and Performance, 14, 331–344. Hazeltine, E., Lightman, E., Schwarb, H., & Schumacher, E.H. (2011). The boundaries of sequential modulations: Evidence for set-level control. Journal of Experimental Psychology: Human Perception and Performance, 37(6), 1898–1914. http://dx.doi.org/ 10.1037/a0024662 (2011-14564-001 [pii]). Hommel, B. (1998). Automatic stimulus-response translation in dual-task performance. Journal of Experimental Psychology: Human Perception and Performance, 24(5), 1368–1384. Hommel, B., Proctor, R.W., & Vu, K.P. (2004). A feature-integration account of sequential effects in the Simon task. Psychological Research, 68(1), 1–17. http://dx.doi.org/10. 1007/s00426-003-0132-y. Jimenez, L., & Mendez, A. (2013). It is not what you expect: Dissociating conflict adaptation from expectancies in a Stroop task. Journal of Experimental Psychology: Human Perception and Performance, 39(1), 271–284. http://dx.doi.org/10.1037/a0027734 (2012-06948-001 [pii]). Kerns, J.G., Cohen, J.D., MacDonald, A.W. r., RY, C., Stenger, V.A., & Carter, C.S. (2004). Anterior cingulate conflict monitoring and adjustments in control. Science, 303(5660), 1023–1026. Kim, S., & Cho, Y.S. (2014). Congruency sequence effect without feature integration and contingency learning. Acta Psychologica, 149, 60–68. http://dx.doi.org/10.1016/j. actpsy.2014.03.004 (S0001-6918(14)00079-1 [pii]). Kornblum, S., & Lee, J.W. (1995). Stimulus–response compatibility with relevant and irrelevant stimulus dimensions that do and do not overlap with the response. Journal of Experimental Psychology: Human Perception and Performance, 21, 855–875. Kunde, W., & Wuhr, P. (2006). Sequential modulations of correspondence effects across spatial dimensions and tasks. Memory and Cognition, 34(2), 356–367. Larson, M.J., Clayson, P.E., & Clawson, A. (2014). Making sense of all the conflict: A theoretical review and critique of conflict-related ERPs. International Journal of Psychophysiology, 93(3), 283–297 (doi: S0167-8760(14)00141-X [pii] 10.1016/j. ijpsycho.2014.06.007). Lee, J., & Cho, Y.S. (2013). Congruency sequence effect in cross-task context: Evidence for dimension-specific modulation. Acta Psychologica, 144(3), 617–627. http://dx.doi. org/10.1016/j.actpsy.2013.09.013 (S0001-6918(13)00218-7 [pii]). Logan, G.D. (1985). Executive control of thought and action. Acta Psychologica, 60, 193–210. Logan, G.D., & Zbrodoff, J.N. (1979). When it helps to be misled: Facilitative effects of increasing the frequency of conflicting stimuli in a Stroop-like task. Memory & Cognition, 7(3), 166–174. Machado, L., Wyatt, N., Devine, A., & Knight, B. (2007). Action planning in the presence of distracting stimuli: An investigation into the time course of distractor effects. Journal of Experimental Psychology: Human Perception and Performance, 33(5), 1045–1061 (doi: 2007-14662-004 [pii] 10.1037/0096-1523.33.5.1045). MacLeod, C.M. (1991). Half a century of research on the Stroop effect: An integrative review. Psychological Bulletin, 109(2), 163–203. Mansfield, K.L., van der Molen, M.W., Falkenstein, M., & van Boxtel, G.J. (2013). Temporal dynamics of interference in Simon and Eriksen tasks considered within the context of a dual-process model. Brain and Cognition, 82(3), 353–363. http://dx.doi.org/10.1016/ j.bandc.2013.06.001 (S0278-2626(13)00081-X [pii]). Mattler, U. (2003). Delayed flanker effects on lateralized readiness potentials. Experimental Brain Research, 151(2), 272–288. http://dx.doi.org/10.1007/s00221003-1486-5. Mayr, U., Awh, E., & Laurey, P. (2003). Conflict adaptation effects in the absence of executive control. Nature Neuroscience, 6, 450–452. Mordkoff, J.T. (2012). Observation: Three reasons to avoid having half of the trials be congruent in a four-alternative forced-choice experiment on sequential modulation. Psychonomic Bulletin and Review, 19(4), 750–757. http://dx.doi.org/10.3758/s13423012-0257-3. Polk, T.A., Drake, R.M., Jonides, J.J., Smith, M.R., & Smith, E.E. (2008). Attention enhances the neural processing of relevant features and suppresses the processing of irrelevant features in humans: A functional magnetic resonance imaging study of the Stroop task. Journal of Neuroscience, 28(51), 13786–13792. http://dx.doi.org/10.1523/ JNEUROSCI.1026-08.2008 (28/51/13786 [pii]). Ridderinkhof, K.R. (2002a). Micro- and macro-adjustments of task set: Activation and suppression in conflict tasks. Psychological Research, 66(4), 312–323. http://dx.doi. org/10.1007/s00426-002-0104-7. Ridderinkhof, K.R. (Ed.). (2002). A dual-route processing architecture for stimulus–response correspondence effects. Vol. 19. (pp. 494–519). Oxford: Oxford University Press.

D.H. Weissman et al. / Acta Psychologica 156 (2015) 8–21 Ridderinkhof, K.R., van der Molen, M.W., & Bashore, T.R. (1995). Limits on the application of additive factors logic: Violations of stage robustness suggest a dual-process architecture to explain flanker effects on target processing. Acta Psychologica, 90, 29–48. Rogers, R.D., & Monsell, S. (1995). Costs of a predictable switch between cognitive tasks. Journal of Experimental Psychology: General, 124(2), 207–231. Schmidt, J.R. (2013). Questioning conflict adaptation: Proportion congruent and Gratton effects reconsidered. Psychonomic Bulletin and Review. http://dx.doi.org/10.3758/ s13423-012-0373-0. Schmidt, J.R., & De Houwer, J. (2011). Now you see it, now you don't: Controlling for contingencies and stimulus repetitions eliminates the Gratton effect. Acta Psychologica, 138(1), 176–186. http://dx.doi.org/10.1016/j.actpsy.2011.06.002 (S0001-6918(11)00125-9 [pii]). Schmidt, J.R., & Weissman, D.H. (2014). Congruency sequence effects without feature integration or contingency learning confounds. PloS One, 9(7). Spape, M.M., & Hommel, B. (2008). He said, she said: Episodic retrieval induces conflict adaptation in an auditory Stroop task. Psychonomic Bulletin & Review, 15(6), 1117–1121. http://dx.doi.org/10.3758/PBR.15.6.1117 (15/6/1117 [pii]). Sturmer, B., Leuthold, H., Soetens, E., Schroter, H., & Sommer, W. (2002). Control over location-based response activation in the Simon task: Behavioral and electrophysiological evidence. Journal of Experimental Psychology: Human Perception and Performance, 28(6), 1345–1363. Sturmer, B., Ouyang, G., Zhou, C., Boldt, A., & Sommer, W. (2013). Separating stimulusdriven and response-related LRP components with Residue Iteration Decomposition (RIDE). Psychophysiology, 50(1), 70–73. http://dx.doi.org/10.1111/j.1469-8986.2012. 01479.x. Szucs, D., Soltesz, F., Bryce, D., & Whitebread, D. (2009). Real-time tracking of motor response activation and response competition in a Stroop task in young children: A lateralized readiness potential study. Journal of Cognitive Neuroscience, 21(11), 2195–2206. http://dx.doi.org/10.1162/jocn.2009.21220.

21

Ullsberger, M., Bylsma, L.M., & Botvinick, M.M. (2005). The conflict adaptation effect: It's not just priming. Cognitive, Affective, and Behavioral Neuroscience, 5(4), 467–472. van den Wildenberg, W.P., Wylie, S.A., Forstmann, B.U., Burle, B., Hasbroucq, T., & Ridderinkhof, K.R. (2010). To head or to heed? Beyond the surface of selective action inhibition: A review. Frontiers in Human Neuroscience, 4, 222. http://dx.doi.org/10. 3389/fnhum.2010.00222. Verguts, T., & Notebaert, W. (2008). Hebbian learning of cognitive control: Dealing with specific and nonspecific adaptation. Psychological Review, 115(2), 518–525. http:// dx.doi.org/10.1037/0033-295X.115.2.518 (2008-04236-010 [pii]). Vorberg, D., Mattler, U., Heinecke, A., Schmidt, T., & Schwarzbach, J. (2003). Different time courses for visual perception and action priming. Proceedings of the National Academy of Sciences of the United States of America, 100(10), 6275–6280. http://dx.doi.org/10. 1073/pnas.0931489100 (0931489100 [pii]). Weissman, D.H., Jiang, J., & Egner, T. (2014). Determinants of congruency sequence effects without learning and memory confounds. Journal of Experimental Psychology: Human Perception and Performance, 40(5), 2022–2037. http://dx.doi.org/10.1037/a0037454 (2014-31981-001 [pii]). Weissman, D.H., Warner, L.M., & Woldorff, M.G. (2004). The neural mechanisms for minimizing cross-modal distraction. Journal of Neuroscience, 24(48), 10941–10949. Wendt, M., Kiesel, A., Geringswald, F., Purmann, S., & Fischer, R. (2014). Attentional adjustment to conflict strength: Evidence from the effects of manipulating flanker-target SOA on response times and prestimulus pupil size. Experimental Psychology, 61(1), 55–67. http://dx.doi.org/10.1027/1618-3169/a000227 (1834816088485432 [pii]). Wylie, S.A., Ridderinkhof, K.R., Bashore, T.R., & van den Wildenberg, W.P. (2010). The effect of Parkinson's disease on the dynamics of on-line and proactive cognitive control during action selection. Journal of Cognitive Neuroscience, 22(9), 2058–2073. http:// dx.doi.org/10.1162/jocn.2009.21326. Yeung, N., Cohen, J.D., & Botvinick, M.M. (2011). Errors of interpretation and modeling: A reply to Grinband et al. NeuroImage, 57(2), 316–319. http://dx.doi.org/10.1016/j. neuroimage.2011.04.029 (S1053-8119(11)00428-9 [pii]).

Lihat lebih banyak...

Comentarios

Copyright © 2017 DATOSPDF Inc.