Brain and Cognition 63 (2007) 260–270 www.elsevier.com/locate/b&c

Response inhibition in motor and oculomotor conXict tasks: DiVerent mechanisms, diVerent dynamics? Jasper G. Wijnen a,¤, K. Richard Ridderinkhof a,b a

Amsterdam Center for the Study of Adaptive Control in Brain and Behaviour, Department of Psychology, University of Amsterdam, Amsterdam, The Netherlands b Department of Psychology, Leiden University, Leiden, The Netherlands Accepted 7 September 2006 Available online 27 October 2006

Abstract Previous research has shown that the appearance of task-irrelevant abrupt onsets inXuences saccadic eye movements during visual search and may slow down manual reactions to target stimuli. Analysis of reaction time distributions in the present study oVers evidence suggesting that top-down inhibition processes actively suppress oculomotor or motor responses elicited by a salient distractor, in order to resolve the conXict that arises when reXex-like and deliberate responses are in opposition. Twenty-six participants carried out a variation of the oculomotor capture task. They were instructed to respond with either a saccade toward or with a button press at the side of the hemiWeld in which a target color singleton appeared. A distractor stimulus could appear either in the same or in the opposite hemiWeld. Delta plots revealed competition between reXex-like and deliberate response activation, and highlighted selective inhibition of automatic responses: While participants generally responded more slowly in incongruent compared to congruent situations, this eVect diminished and even reversed in the slowest speed quantiles. These eVects were present in both the oculomotor and motor response-mode conditions. © 2006 Elsevier Inc. All rights reserved. Keywords: Attentional capture; Saccadic eye movements; Motor control; Response inhibition; Distributional analysis

1. Introduction In the distinction between automatic and controlled processes, Wrst proposed by William James (1890), and revived by ShiVrin and colleagues (Atkinson & ShiVrin, 1968; ShiVrin & Schneider, 1977), the term automatic (exogenously controlled or stimulus-driven) refers to those actions that are initiated indeliberately without awareness as a necessary condition for doing so. An example would be orienting to a loud noise or bright Xash in the periphery of the visual Weld. In contrast, controlled (endogenously controlled or goal-directed) refers to processes which require, in the words of James: “an additional conscious element in the shape of a Wat, mandate, or expressed consent”. These acts

*

Corresponding author. Fax: +31206390279. E-mail address: [email protected] (J.G. Wijnen).

0278-2626/$ - see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.bandc.2006.09.003

are “willed” and call for some eVort to execute. An eye movement away from a salient stimulus, as is required in for example the antisaccade task (Hallet, 1978), necessitates a controlled response as well as inhibition of the automatic response towards the stimulus (for a recent comprehensive review see Reuter & Kathmann, 2004). Ignoring highly salient information in the environment while attending to objects that do not automatically capture attention places similar demands on human information processing. The outcomes of controlled and automatic processes may clash, and the resulting conXict needs to be resolved in favor of the controlled response. This paper aims to examine and compare the dynamics of activation and inhibition during such conXicts in the motor domain on the one hand and the oculomotor domain on the other. SpeciWcally, we seek to investigate the role of inhibitory processes in conXict resolution. A task was designed to allow the use of a common method for both domains and to investigate whether

J.G. Wijnen, K.R. Ridderinkhof / Brain and Cognition 63 (2007) 260–270

processes underlying motor control and oculomotor control exhibit similar dynamics. 1.1. Oculomotor capture and competitive integration Various saccadic eye-movement tasks have been developed to investigate the relationship between goal-directed and automatic responses in the oculomotor domain. Theeuwes, Kramer, Hahn, and Irwin (1998) presented a display of six gray circles to their subjects, Wve of which changed to a red color after a set period of time. Subjects were instructed to make an eye-movement towards the one circle that did not change color. On some trials a new red circle was introduced between two of the existing circles. The novel object was found to disrupt the programming and execution of the saccade towards the target (gray) circle, even though the new circle was irrelevant to the task. In many instances the eye Wrst moved toward the distractor and paused there brieXy before moving on toward the target. The automatic attraction of the eyes towards a salient stimulus in the environment was termed oculomotor capture. Since the brief Wxation durations at the location of the distractor stimulus were shorter than the time typically needed to program a saccade, Theeuwes et al. concluded that the saccades toward the target and toward the distractor must have been programmed, at least partly, in parallel. The question remained whether the resulting saccades were simply those programmed by the Wrst pathway to complete its processing or whether the pathways interact and compete even before a saccade is executed. Godijn and Theeuwes (2002), building in part on Trappenberg, Dorris, Munoz, and Klein (2001) and others, proposed a competitive integration model. In this model, the activation and inhibition signals for exogenous and endogenous saccades converge on a common saccade map which is presumably located in the superior colliculus (SC). On the saccade map, where information is represented retinotopically, the saccade goal is determined by a competitive process. In general, activation in a certain location will inhibit all distant locations, while having an excitatory eVect on adjacent locations. In the model, saccades are executed once the activation in a certain location on the saccade map reaches a threshold level. A modiWed version of the oculomotor capture paradigm was used by Godijn and Theeuwes (2002) to test predictions made by the competitive integration model with regard to the timing of eye movements (saccade latencies). As described above, the oculomotor capture paradigm requires participants to overtly search for a speciWc target stimulus, which is presented at the same time as a highly salient distractor stimulus. In their rendering of this experimental paradigm, Godijn and Theeuwes varied the distances between the locations at which the distractor and stimulus appeared. Since activation at distant locations on the saccade map will result in the mutual inhibition of both locations, it was expected that in this situation activation levels would need longer to reach the threshold level. On the other hand when

261

the distractor and target are close, activation in both locations will be strengthened and a threshold will be reached sooner. In accordance with these predictions, saccade latencies were shorter in the latter condition. 1.2. Motor capture and multiple routes to response activation To study the dynamics of attentional and response conXicts in the motor domain, researchers have developed a variety of experimental speeded response time tasks which induce conXicting processing. Well-known examples of such conXict tasks are the Stroop task (Stroop, 1935), the Simon task (Simon, 1969), and the Eriksen Xanker task (Eriksen & Eriksen, 1974). All of these tasks include a relevant stimulus or relevant stimulus aspect on which the participant has to base his or her response. On the other hand there are irrelevant stimuli or stimulus aspects which are often highly salient and which may or may not suggest a response opposite to the response induced by the relevant stimulus. The condition in which irrelevant stimulus features manage to more or less automatically activate the corresponding motor response can tentatively be referred to as motor (or ideomotor) capture (analogous to oculomotor capture), the tendency of the motor system to engage in automatic, stimulus-driven acts despite task instructions that specify otherwise. In conXict tasks those trials where relevant and irrelevant dimensions of the stimuli suggest responses that are opposed are usually called incongruent whereas trials in which the dimensions suggest the same response are called congruent. The mean diVerence in reaction time (RT) or accuracy between incongruent and congruent trials consistently found in conXict tasks is taken as a measure of the time needed to overcome automatic stimulus-driven activation between diVerent processing streams and is called the congruence eVect. In recent years, dual-process conceptions of how sensory information leads to activation of the correct response have become increasingly popular. In such conceptions, perception–action coupling can be established via two parallel routes, one controlled and deliberate, the other fast, direct, and more or less automatic. Kornblum, Hasbroucq, and Osman (1990) set the stage with their dual-route model for congruence eVects. Although dual-route models had been formulated previously (e.g., Frith & Done, 1986; Sanders, 1967), the Kornblum et al. model has served as a signiWcant impetus for subsequent research into congruence eVects. Conceived on the basis of theoretical considerations, the model contains a number of discrete stages of processing, arranged partly in parallel. Upon identiWcation, a stimulus is thought to deliberately activate an instruction-based response via the controlled route, and to activate an automatic, instruction-independent response via the direct route. If the two response codes match, the motor program already activated via the direct route can be carried out quickly; if they mismatch, this motor program must be aborted in favor of the alternative motor program, whose retrieval and execution costs extra time.

262

J.G. Wijnen, K.R. Ridderinkhof / Brain and Cognition 63 (2007) 260–270

The rudimentary dual-route architecture of the Kornblum et al. model has been embraced by many authors in the Weld (e.g., Eimer, Hommel, & Prinz, 1995; de Jong, Liang, & Lauber, 1994; Munoz & Everling, 2004; Proctor, Lu, Wang, & Dutta, 1995; Ridderinkhof, van der Molen, & Bashore, 1995), and has been implemented in formal connectionist models capable of simulating many of the key phenomena reported in the conXict-task literature (e.g., Cohen, Dunbar, & McClelland, 1990). Interestingly, this type of multipleprocess model for motor capture is apparently compatible with the competitive integration model for oculomotor capture (in which the saccade map is also aVected by two separate processing streams). 1.3. Bottom-up activation and top-down control Strong habitual connections between sensory stimuli and corresponding responses may aVord rapid execution of natural responses without demanding much attention. However, these responses need to be overruled when prepotent stimulus–response mappings are inappropriate and more controlled goal-directed action selection is required. This top-down bias is especially important when the pathways leading to the desired action compete for expression in behavior with concurrent, more habitual pathways (Miller & Cohen, 2001). Crucial in this architecture is that the information-processing system “knows” when top-down control is required without recourse to a homunculus. According to the conXict monitoring hypothesis (Botvinick, Braver, Barch, Carter, & Cohen, 2001), the response system is monitored for the occurrence of conXict. When response conXict is detected, this information is used by other mechanisms to implement appropriate top-down control, as described in more detail below. Inhibitory control has been postulated as one of the mechanisms by which top-down bias may be implemented to optimize behavior (for a review see Munoz & Everling, 2004; Ridderinkhof, van den Wildenberg, Segalowitz, & Carter, 2004a). Response inhibition can be deWned descriptively as “the mechanism or set of processes that results in the containment of prepotent behavioral responses when such actions are reXex-like, premature, inappropriate, or incorrect” (Ridderinkhof, van den Wildenberg, Wijnen, & Burle, 2004b). Thus, response inhibition can be used as an instrument to suppress exogenous direct activation of motor programming structures to prevent them from resulting in overt behavior. Several experiments have pointed to a role for inhibitory processes in conXict tasks. For instance, Eimer and Schlaghecken (1998) asked participants to react to the orientation of a target-stimulus (an arrow) and found performance costs when target stimuli were preceded by a masked prime in the same orientation. BeneWts were found when the prime suggested the opposite response. The eVect reversed when the interval between prime and target was shortened. The pattern suggests that participants were actively inhibiting the non-predictive

prime, but failed to do so in time when the interval between prime and target was relatively short. To accommodate a wider range of empirical phenomena, models for oculomotor and motor capture may be supplemented to incorporate such inhibitory mechanisms. The competitive integration model for instance may be extended so that the distractor is inhibited soon after both target and distractor have been identiWed. 1.4. A method for studying inhibitory control To accommodate a mechanism that might perform the selective response suppression functions described above, at least conceptually, Ridderinkhof (2002) argued for an addition to the dual-process model in which an activation-suppression mechanism, that would be triggered in the event of conXict between the two routes and its operations, would aVect the exogenous “direct-activation” route (or the response induced by that route). The eVect of such a mechanism would be to suppress exogenous inXuences (such as motor capture) and to allow endogenous processing to eVectively compete and gain control over the response. On the assumption that selective response suppression takes some time to evolve during the course of response activation and thus would not be operational immediately after presentation of the stimulus giving rise to this response activation, the eVects of response inhibition would be manifest especially in relatively slow responses, in which response suppression would have more time to build up than in responses that happen to be fast. Hence, responses at the slower tail of the RT distribution are more sensitive to the manifestations of selective suppression than responses at the fast end. Thus, in case of incongruent trials, in which selective suppression of the inappropriate response should beneWt the activation of the correct response, slow responses should beneWt more than fast trials, and this is indeed what is often observed in distribution analyses of RTs in conXict tasks (e.g. Bub, Masson, & Lalonde, 2006; Burle, Possamaï, Vidal, Bonnet, & Hasbroucq, 2002; Eimer, 1999; de Jong et al., 1994; Ridderinkhof, 2002; for an overview see Ridderinkhof et al., 2004b). Analyses of RT distributions typically focus on cumulative density functions (CDFs), which represent the cumulative probability of responding as a function of response speed. In analyzing CDFs from conXict tasks, delta plots can be abstracted from vincentized CDFs (RatcliV, 1979; Vincent, 1912) for illustrational purposes. Delta plots represent the congruence eVect (the diVerence in RT between incongruent and congruent trials) as a function of response speed. This technique involves dividing the RT distributions of congruent and incongruent trials into a number of equal-size RT quantiles, subtracting mean RTs for congruent trials, quantile by quantile, from those of incongruent trials and then plotting these diVerences, again quantile by quantile, against the average RT of the congruent and incongruent speed quantiles.

J.G. Wijnen, K.R. Ridderinkhof / Brain and Cognition 63 (2007) 260–270

The typical pattern for the sigmoid-shaped CDFs is an increase of RT mean diVerences with slower response speeds. This is a consequence of the proportional diVerence between conditions remaining constant as the average RT increases, resulting in an increase of absolute diVerences between conditions. In delta plots, this is reXected in a positive slope. However, if selective suppression of the inappropriate response in incongruent trials beneWts the activation of slow correct responses more than fast responses, then slow responses should beneWt more than fast trials, and this will be expressed in the delta plot as a leveling oV or even turning negative of the delta plot at slower quantiles. Indeed, in the Simon task, where subjects typically have to press a left or right hand button according to stimulus color, whereas the stimulus itself may be presented to the left or right of a Wxation point (creating compatible and incompatible stimulus response relations), delta plots typically run negative, with strongly reduced or even reversed congruence eVects at the slowest response quantiles. However, when the exact same Simon task is intermixed with a small number of trials in which location is the target stimulus dimension, subjects can no longer suppress the location-driven response a priori, even though location remains irrelevant in the Simon task itself. In such a context, immediate selective suppression of the location-driven response is prevented, and consequently the delta plot for the Simon eVect is found to run much more positive (Ridderinkhof, 2002). Several studies have shown similar eVects using this method. Burle et al. (2002) used electromyographic recordings in a Simon task to separate trials in which participants made partial errors (as indicated by subthreshold activation of muscles involved in the incorrect response) from correct trials. These partial-error trials were associated with reductions of the congruence eVect with increasing response speed, seen in a more downward-going delta plot compared to the delta plots for fully correct responses. The high conXict in these partial-error trials elicited stronger selective suppression of exogenously induced activation which made for a more eYcient inhibition of the erroneous response.

263

target”) requiring a discriminative response. In such a procedure task-irrelevant cues are suYciently similar to targets to capture the attentional or response systems). For all response modalities, it was found that under these circumstances (masked cues) only motor activation processes triggered by foveal visual stimuli are subject to inhibition. Given these similarities in oculomotor and motor control we sought to explore whether diVerences still exist in the dynamics of inhibitory control for both response modalities, using the distribution-analyses method described earlier. To this end we developed a variation of the oculomotor capture paradigm suitable for both saccadic and manual responses, which includes congruent and incongruent trials (as opposed to the distractor and no-distractor trials typically used in the oculomotor capture paradigm), thus allowing us to chart congruence eVects. Participants were asked to respond to one of four circles that did not change color while being simultaneously distracted by a displacement of one of the circles. They were instructed to indicate the side of the screen which contained the target either by making an eye movement towards that side or by pressing one of two buttons corresponding to the relevant side. Congruent trials were deWned as those trials in which both target and distractor appeared on the same half of the screen; other trials were labeled incongruent. We expected a leveling oV of the delta plot (similar to that found in earlier studies using manual conXict tasks) for both oculomotor and manual response modes. In order to test the commonalities of motor and oculomotor control processes more speciWcally, we varied the interval between the distractor and the target (this interval is often called the Stimulus Onset Asynchrony or SOA). An eVect of SOA on delta-plot slopes was found by Burle, van den Wildenberg, and Ridderinkhof (2005) using a cue-priming procedure. Delta-plot slopes were found to decline more steeply for larger SOAs. We investigated whether a comparable eVect of SOA could be found for the current task.and, crucially, whether such an eVect of SOA on delta-plot slopes would be similar across response modalities. 2. Method

1.5. Common dynamics of inhibition and activation for diVerent response modalities Although diVerent brain circuits are involved in the activation and inhibition of responses in the motor and oculomotor systems we anticipated that the respective mechanisms of selective response suppression ‘behave’ similarly on a functional level, and thus will be expressed in similar ways in CDFs and delta plots of manual and saccadic RTs. Evidence for some commonalities in the mechanisms involved in inhibition and activation for manual, saccadic and vocal response modalities was found in a study by Eimer and Schlaghecken (2001). In this study, centrally and peripherally positioned masked cues were presented to subjects in a cue priming procedure (i.e. a conXict task whereby subjects are presented with two consecutive signals, one of which (“the cue”) is task-irrelevant and the other (“the

2.1. Participants Twenty-six undergraduate students (17 female) from the University of Amsterdam took part in the experiment and obtained course credits for their participation. Mean age was 21.31 (SD: 3.87, Min: 18, Max: 36). All participants reported to be healthy and to have normal or corrected-tonormal vision (however persons wearing spectacles were excluded from participation due to possible interference with the eyetracker software). Subjects were tested individually in a quiet and dimly lit university chamber. 2.2. Stimuli and apparatus Participants were seated in a comfortable chair with response-buttons attached to both arm-rests, at a distance

264

J.G. Wijnen, K.R. Ridderinkhof / Brain and Cognition 63 (2007) 260–270

downwards from above) these starting conWgurations were presented more frequently than conWgurations in which the open locations were above or below the Wxation point on both sides (with a ratio of 3:2). For each of the starting conWgurations, all possible combinations of color changes and circle displacements occurred once for incongruent trials, or twice for congruent trials. Given that the target was constrained not to appear on the circle that moved, the number of options for the target location is limited to one when both target and distractor appear in the same half of the screen (congruent trials), while two options are available to incongruent trials. To keep the ratio between congruent and incongruent trials at 1:1 all possible congruent target–distractor combinations appeared twice. The sequence of starting conWgurations and distractor and target positions was randomised for each experimental block. As a consequence congruent and incongruent trials were presented in randomised order. In one half of the trial blocks, subjects indicated their response by pressing one of the two response keys with their index Wngers. In the other half, no manual response was requested and participants responded by merely making an eye movement to the target stimulus. Saccades were recorded using an infrared-based iView eyetracker (SMI, Berlin, Germany) with 50-Hz temporal resolution and a <0.1° spatial resolution. The head was stabilized with a comfortable chin-rest.

of 55 cm in front of a computer screen on which stimuli where shown against a black background. A small white cross (subtending a visual angle of approximately 0.5° £ 0.5°) served as a Wxation point during the experiment. On each trial four green circles (diameter 1.6°) were presented appearing in four out of six possible locations. The conWguration of circles was constrained in such a way that at least two circles were presented in each half (left/right) of the screen, leaving one possible location open. Vertical distance from the Wxation point was either 0° or 2.5° of visual angle above or below the Wxation point. Horizontal distance from the Wxation point was 10.8° of visual angle to the left or the right of the Wxation point. During each trial, three of the four circles changed from a green to an equiluminant blue color. In addition one of the circles changed position to the empty location in the same half of the screen (if the empty spot was situated above or below the Wxation point, the circle changing position was always the one closest to it). The circle that changed position was never the one that retained its original color. Fig. 1 depicts an experimental trial. On each trial, participants were asked to respond to the circle that did not change color (sustained color-singleton target). The circle displacement was irrelevant for the execution of the task (oVset–onset distractor). Trials in which the target and the distractor were presented within the same hemiWeld are designated congruent trials; in all other cases the trial is labelled incongruent. The target and distractor could either be presented simultaneously (a SOA of 0 ms) or the distractor could be presented 120 or 240 ms in advance of the target. In each experimental block, all possible starting conWgurations of circles appeared with one exception. The possible conWguration where initially both open locations were located at the same height as the Wxation point on both sides was excluded to limit the total duration of an experimental block. Since more distractor options are available when the starting conWguration of circles features an open location on the same height as the Wxation point (the distractor on that side can move either upwards from below or

2.3. Design Each of the experimental blocks contained 240 trials, 120 of which were congruent and 120 incongruent. Each of the congruence conditions contained 40 trials per SOA condition. A total of four experimental blocks were completed, two of which were manual response blocks and two were oculomotor response blocks. The sequence of response mode blocktypes was counterbalanced over subjects with the constraint that a manual block was always followed by an oculomotor block and vice versa.

Tim e Fig. 1. A possible congruent experimental trial (left) with a Stimulus Onset Asynchrony (SOA) of 0 ms and an incongruent experimental trial (right) with a SOA of greater than 0 ms. In the congruent example, the circle positioned center-right moves to the bottom-right position, thus forming the distractor. In the incongruent example, the circle positioned top-left moves to the center-left position. In both examples, the circle positioned top-right is the only one not to change to an equiluminant blue color, forming the target.

J.G. Wijnen, K.R. Ridderinkhof / Brain and Cognition 63 (2007) 260–270

2.4. Procedure The experiment involved a single session which lasted t60 min. Participants Wxated four series of six calibration targets at the start of the session. Calibration targets appeared at the locations which would later be occupied by the experimental stimuli. After calibration participants practiced the experimental task. Practice ended when at least 20 trials were responded to correctly. Four experimental blocks were completed in this fashion, each lasting t12 min. In blocks where the subject was asked to make a manual response, the instruction was given to push the button corresponding to the side of the screen in which the target was shown. Eye-movements were not recorded in manual blocks. Blocks where a saccade was requested were accompanied by an instruction to move the eyes to the target as soon as it was presented on the screen. Instructions to the participant were given verbally and were reiterated in written form on an instruction screen at the start of each experimental block. Speed was emphasized over accuracy. It was stressed to the participant that the eyes should be directed to the Wxation point at the start of each trial and should remain Wxated until a target was presented on the screen. The starting conWguration of circles was presented on the screen for 600 ms. After that the distractor appeared simultaneously with or quickly followed by the target according to the SOA condition. Target and distractor were shown for an additional period of 1200 ms after the presentation of the distractor. All stimuli then disappeared and a black screen with only the Wxaton point visible was displayed for 1000 ms until the start of the next trial. 2.5. Data analysis In the analysis of the manual response data, trials in which the RT was below 100 ms were excluded (these were assumed be fast guesses). In addition responses slower then 1000 ms were excluded (misses). This resulted in a loss of 0.40% of the trials, all of which were fast guesses. For the oculomotor response data, an identical time-window existed outside of which saccadic RTs were excluded from further analysis (loss 2.27% for fast guesses; 1.03% for misses). Movement artefacts or eye blinks occurring within this time-window were also grounds for excluding the trial from further analysis (loss 1.59%). Furthermore, trials were excluded if the eyes were not directed at the Wxation point at the time the target was presented on the screen (loss 4.11%). To determine whether the eyes were at Wxation at that time, a small time-window was deWned beginning 40 ms before target appearance and extending 40 ms after target appearance (5 samples). If the average position signal in this window did not exceed 2° of visual angle relative to the Wxation position, the eyes were considered to be at Wxation at target presentation. Outliers were removed from the data using a recursive procedure adopted from McCormick (1997), which involved temporarily removing the fastest

265

and slowest RTs from each condition for each participant. After removal the mean and standard deviation for the remaining data was calculated. If either of the two removed data-points fell outside an interval bounded by 4 SDs from the mean, that data-point was removed permanently. If the data-point fell within the interval it was returned to the data set. This procedure continued until no more datapoints were removed permanently. In the manual blocks 1.26% of the data was removed as outlier, in the saccade blocks 0.88 % of the data was removed. In oV-line analysis, saccadic reaction times (SRTs) were deWned as the time at which the velocity of the eye movement exceeded 100°/s, but only if the movement resulted in a horizontal position-signal exceeding 2.5° of visual angle relative to the Wxation position. For saccade accuracy purposes, trials were scored as an error trial if the Wrst eye movement that satisWed the above criteria with regards to velocity and position moved away from the target on the horizontal axis. Note that for both manual and oculomotor blocks a reaction that is based solely on the location of the distractor is scored as correct when the trial is congruent. For each subject mean RT and overall accuracy were computed for congruent and incongruent trials aggregated over blocks. Two-way analyses of variance using SPSS’s GLM repeated measures feature were conducted with congruence and SOA as within-subjects factors and RT and accuracy as the dependent variables. This was done separately for manual and oculomotor blocks. To capture the dynamics of the congruence eVect, the RTs for each subject were rank-ordered for each SOA and congruence condition. Only correct trials were used for this purpose. The RTs were then divided into Wve equal-size speed bins, thus comprising vincentized CDFs. For each of these bins the congruence eVect was computed and plotted against the average RT collapsed across congruence conditions for that speed bin and SOA condition. 3. Results All reported p values were corrected using the Greenhouse-Geisser adjustment of the degrees of freedom when necessary. 3.1. Overall performance 3.1.1. Manual response speed and accuracy Fig. 2 shows the mean RTs for correct responses and accuracy levels in the manual response condition. Congruent trials were responded to signiWcantly faster in comparison with incongruent trials. ANOVA revealed this diVerence to be signiWcant (F(1, 25) D 25.02, p < 0.001). Subjects made more errors when distractor and target appeared at opposite halves of the screen (F(1, 25) D 6.63, p D 0.016). The length of the SOA had a clear eVect on both RT (F(2, 50) D 145.73, p < 0.001) and accuracy (F(2, 50) D 18.77, p < 0.001). Participants responded faster at the longer SOAs but also were more prone to make an error. For the longer

J.G. Wijnen, K.R. Ridderinkhof / Brain and Cognition 63 (2007) 260–270

Ideomotor

32%

450

400

350

4%

RT (ms)

12%

8%

500 E% Congruent E% Incongruent RT Congruent RT Incongruent

36%

300

Percentage of Errors

E% Congruent E% Incongruent RT Congruent RT Incongruent

16%

Percentage of Errors

Oculomotor 40%

500

450

28% 24%

400

20% 16%

350

RT (ms)

266

12% 8%

300

4% 0%

250 SOA0 ms

SOA 120msS

OA 240ms

0%

250 SOA0 ms

SOA 120msS

OA 240ms

Fig. 2. Mean reaction times (RTs; line graphs) and error rates (bar graphs) in congruent (gray) and incongruent (black) trials for the three stimulus onset asynchrony (SOA) conditions, displayed separately for the manual (left) and oculomotor (right) blocks. Error bars represent the standard error of the mean.

SOAs diVerences in reaction speeds between congruence conditions were reduced (F(2, 50) D 6.82, p D 0.004). No interaction between congruence and SOA condition was found on accuracy (F(2, 50) D 1.88, p D 0.18).

signiWcantly between the SOA conditions (interaction between SOA and congruence (F(2, 50) D 2.27, p D 0.11)).

3.1.2. Oculomotor response speed and accuracy In contrast to the manual response-mode blocks, oculomotor reaction speed did not seem to be hampered by the distractor appearing in a position opposite to that of the target. The main eVect of congruence on SRTs was not signiWcant (F(1, 25) D 1.13, p D 0.30). Longer SOAs however did enhance subjects reaction speeds (F(2, 50) D 187.93, p < 0.001). The eVect of congruence on SRTs did not diVer signiWcantly between SOA conditions (F(2, 50) D 1.04, p D 0.36). Accuracy was aVected by congruence with more errors being made on incongruent trials (F(1, 25) D 8.81, p D 0.007). Longer intervals between the target and the distractor also elicited more errors (F(2, 50) D 187.93, p < 0.001). The congruence eVect on accuracy did not diVer

CDFs, which plot the probability of responding as a function of response speed, were constructed for Wve speed bins. From these delta plots were extracted, which plot diVerence scores (incongruent minus congruent) for each of the Wve speed bins against the average of both conditions in that speed bin. This was done for each subject separately and then averaged across subjects (see Fig. 3). The CDFs and delta plots allow us to examine whether the presence and absence of congruence eVects (in ideo- and oculomotor versions of the task, respectively) remained stable in magnitude across the entire RT distribution. The plots for the manual response mode depict a positive congruence eVect for all of the speed-bins in the 0 and 120 ms SOA conditions, and a positive eVect in the 240 ms SOA condition for

3.2. Distributional analysis

Fig. 3. Cumulative density functions (CDFs) and delta plots (top-left corner) for the three stimulus onset asynchrony (SOA) conditions, displayed separately for the manual (left) and oculomotor (right) blocks. Separate CDFs are shown for the congruent (CG; thick lines) and incongruent (IC, thin striped lines) conditions. Black diamonds are used for the shortest SOA, gray diamonds are used for the intermediate SOA (120 ms), white diamonds are used for the longest SOA (240 ms).

J.G. Wijnen, K.R. Ridderinkhof / Brain and Cognition 63 (2007) 260–270

the faster responses. The congruence eVect gradually declines or remains roughly constant in all of the SOA conditions when responses become slower. In the longest SOA condition, this even leads to a negative congruence eVect. To investigate the signiWcance of mean diVerences in the CDFs we conducted three-way ANOVAs with congruence, SOA, and quantile as within-subject factors and mean RT as the dependent variable. This was done for both the motor and the oculomotor CDFs. For the motor CDFs, a signiWcant interaction was found between SOA and quantile (F(8, 200) D 11.69, p < .001), which appears to reXect that the variance in the RT distribution associated with the longest SOA is higher compared to the distributions of the other two SOA conditions. No interaction was observed between quantile and congruence condition (F(4, 100) D 0.10, p D .84) or between all three factors (F(8, 200) D 0.77, p D .52). This pattern indicates that the slopes of the delta plots do not deviate from zero, in any of the SOA conditions. For the oculomotor CDFs, a signiWcant interaction was found between SOA and quantile (F(8, 200) D 14.82, p < .001), again pointing to the pattern that the variance in the RT distribution associated with the longest SOA is higher than in the other two SOA conditions. Delta-plot slopes for this response modality showed a strong decline in the eVect of congruence when subjects responded slower. A signiWcant two-way interaction between quantile and congruence condition (F(8, 200) D 16.58, p < .001) conWrmed this pattern. The sharpest decline in congruence eVects was seen in the longest SOA condition; the shortest SOA condition showed a similar decline but less steep; and, unexpectedly, the middle SOA condition slopes downward sharply for the Wrst two speed bins but then levels out for the remaining bins. Delta-plots for the diVerent SOA conditions diVered signiWcantly from each other as indicated by a three-way interaction between all of the factors (F(8, 200) D 3.96, p D .013). Delta-plot slopes for the oculomotor data seemed more sharply negative-going for the oculomotor data compared to the motor condition. A four-way ANOVA with congruence, SOA, quantile and response modality as within-subject factors and mean RT as the dependent variable was conducted to examine whether these diVerences between the response modalities could have been generated by chance. The analysis revealed a signiWcant three-way interaction between response modality, congruence and quantile (F(4, 100) D 9.57, p < .001), conWrming the more pronounced decline in the congruence eVect for the oculomotor data as compared to the motor data, which is more stable across the RT distribution. 4. Discussion The aim of this study was to explore response activation and selective response suppression in motor and oculomotor versions of a conXict task. More speciWcally, we set out to verify whether the dynamics of response suppression

267

would be expressed in similar ways in motor and oculomotor RT distributions. To that end, a conXict task was designed that allowed us to record manual and saccadic responses from the same subjects using identical task and stimulus formats. An additional element of our approach was to include an experimental manipulation that is known to inXuence the expression of response inhibition in RT distribution functions (in this case SOA, or the interval between distractor and target onset; Eimer, 1999; Burle et al., 2005), and to examine whether this manipulation had similar eVects on the distributions of saccadic RTs. The results can be summarized as follows. Congruent trials elicited faster and more accurate manual responses than incongruent trials, and this congruence eVect tended to decrease with longer SOAs. For oculomotor responses these patterns were less pronounced or absent, at least at the level of overall performance. Closer inspection of the performance data using distribution analysis, however, revealed diVerent patterns. As anticipated, the congruence eVects on RT did not increase monotonically as a function of response speed. For manual responses, the eVect remained stable in magnitude across the entire RT distribution in each SOA condition. For oculomotor responses, the eVect in fact decreased as a function of response speed, and more so at long than at short SOAs. In general, these patterns are in accordance with the predictions derived from the activation–suppression model (Burle et al., 2002, 2005; Eimer, 1999; Ridderinkhof, 2002). This type of model assumes that distractor stimuli may initially activate a response, but that this response activation is subsequently inhibited by a process of selective suppression. This selective suppression may result in relative facilitation of the response driven by the target, to such an extent that in some cases responses to incongruent stimuli are in fact faster than responses to congruent stimuli (Burle et al., 2005; Eimer, 1999). Selective suppression is slow to build up, such that responses at the slow tail of the RT distribution beneWt more from this inhibition than responses at the faster end. The results of the distribution analyses suggest that selective response suppression is at play in the oculomotor version of the conXict task as much as in the motor version. Even though a congruence eVect was not detectable when looking at overall saccadic RTs, such an eVect was revealed in the distribution of RTs. Fast saccadic responses, presumably being less inXuenced by a slowly building inhibition process, showed a relatively large congruence eVect, but for slower responses the congruence eVect declined and reversed in all three SOA conditions. We take this pattern to indicate the evolving presence of active top-down inhibition processes, which selectively suppresses automatic oculomotor responses to the distractor. If anything, suppression manifested itself more prominently in oculomotor than manual responses, as the delta plots ran more negatively for saccadic RTs than for manual RTs. In fact, for saccadic responses, the suppression was so strong as to produce reversed congruence eVects in the slowest speed

268

J.G. Wijnen, K.R. Ridderinkhof / Brain and Cognition 63 (2007) 260–270

quantiles. For manual responses the congruence eVect did not seem to decrease strongly in any of the SOA conditions. The resulting delta plots which look more or less “Xat” should not be taken to indicate an absence of response inhibition. Such absence would be reXected in a monotonous increase of the congruence eVect with increasing RT and thus a positively sloped delta plot. A linear increase in RT diVerences between conditions is a standard statistical property of RT distributions, following directly from the fact that the variance increases with the mean (Wagenmakers, Grasman, & Molenaar, 2005). Departures from this pattern such as Xat or negative-going delta plots, appear only when response-inhibition is brought into the equation. The Wnding that the eVects of congruence decrease with increasing SOA length concur with other studies that introduced an SOA between the appearance of irrelevant and relevant stimuli or stimulus aspects. In the inhibition of return (IOR) paradigm (Posner & Cohen, 1984; Posner, Rafal, Choate, & Vaughan, 1985; for a review, see Klein, 2000), a task-irrelevant cue is presented prior to the presentation of a target. As long as the SOA between cue and target is relatively short, subjects beneWt from the cue if it appears at the location of subsequent target appearance. When SOAs are increased, beneWts in RT turn into costs, thus decreasing and reversing eVects of congruence between cue and target. Eimer and Schlaghecken (1998; Eimer, 1999; Schlaghecken and Eimer, 2000) observed similar performance beneWts for (central and subliminally presented) cues that validly point out the response required by the imminent target (somewhat similar to our congruent trials), as long as the interval between the two was relatively short. When the interval was lengthened the beneWts decreased and even turned into performance costs. This pattern was interpreted to reXect initial activation (based in direct perceptuomotor links) followed by active inhibition of the initial response tendency. The decrease of the congruence eVect with longer SOAs suggests that the eVectiveness of inhibition is a function of the degree of temporal overlap between the responses activated by distractor and target. When the target is presented while inhibition is still in its early stages, the system can take little advantage from it, resulting in long RTs for incongruent trials. When the target is presented later, at a time when inhibition has more fully developed, incongruent trials beneWt more from the distractor-related inhibition. Not only responses to incongruent stimuli, but to a lesser extent also responses to congruent stimuli beneWted from a longer SOA. Schlaghecken and Eimer (2000) found a similar eVect when (subliminal) arrow-cues where presented peripherally instead of centrally. They concluded that while centrally presented cues give rise to an activation-followed-by-inhibition eVect, peripheral cues produce an activation-only eVect, with the cue causing a burst of response activation that dissipates passively over time. However, passive dissipation cannot explain the decrease and reversal of the congruence eVect that we see

in our data. Thus, the small beneWt for congruent trials appears to reXect a more general preparation eVect associated with longer SOAs. The present SOA manipulation appeared to result in a speed/accuracy tradeoV: speed beneWts from the longer SOAs were accompanied by accuracy eVects in the opposite direction (i.e., more errors). Importantly, however, the Wnding that increasing SOAs are associated with increasing congruence eVects on accuracy but with decreasing congruence eVects on RT likely reXects details of the RT distributions rather than speed/accuracy trade-oV. On the one hand, long SOAs were shown to be associated with the strongest response inhibition, explaining the decrease of congruence eVects on RT. On the other hand, long SOAs are accompanied by relatively many fast responses (cf. Fig. 2), many of which may be distractor-driven and, therefore, result in error. Strong response inhibition in this condition might not be able to prevent these fast errors because, in line with the activation–suppression hypothesis, the inhibition takes some time to develop and would exert its inXuence (be it on RT or accuracy) only on trials within the slower tail of the RT-distribution. Although the patterns in the delta plots were largely in line with the predictions based on the activation–suppression hypothesis, and were not predicted by alternative frameworks, one remaining concern is the possibility that the presently observed patterns are not typical for just any kind of manipulation within our task set-up, but instead are speciWc to our critical experimental manipulation of congruence. In a separate study (Wijnen & Ridderinkhof, submitted for publication), two experimental tasks were designed to be as similar to each other and to the present one as possible. One version included a conXict element (presumably invoking the process of selective response suppression, as in the present task), whereas the other task version lacked this conXict element. For the conXict task we observed negativegoing delta plots, as in the present case, but for the control task we established that, as much as in any non-conXict RT task, eVect size increased (rather than decreased) with slower responses. This pattern, which was observed for both the manual and oculomotor response modalities, attests to the speciWcity of the RT distribution eVect as an eVect ensuing from inhibitory processes triggered by the need to overcome automatic stimulus-driven activation, and strengthens further our interpretation in terms of activation–suppression models. In conclusion, despite massive diVerences in the neural circuitry underlying motor control and oculomotor control, the present study indicates that the processes of direct response activation and selective response suppression have several functional and dynamical characteristics in common. While the expression of response inhibition in the RT distributions was more pronounced for saccadic responses, this pattern appears to reXect gradual diVerences rather than fundamental discrepancies. This is an important observation, since it implies that various global functional features of the response inhibition system, as modeled in

J.G. Wijnen, K.R. Ridderinkhof / Brain and Cognition 63 (2007) 260–270

the dual-process architecture and accompanying activation–suppression model, can be studied in the motor and oculomotor system alike. This is not to say that results (e.g., clinical individual diVerences) observed in one modality can readily be generalized to the other; but eVects seen in one modality at the very least may directly inform hypotheses about whether and how such eVects might come to expression in the other. Another implication is that relatively simple RT distribution analysis techniques may be applied in the Weld of oculomotor control to examine the dynamics of oculomotor inhibition. This supplementary technique, combined with the available (reasonably detailed) knowledge about the anatomical and physiological properties of the oculomotor system, may prove valuable in advancing our understanding of the nature of response inhibition processes. Acknowledgment The present study was supported by grants from the Netherlands Organisation for ScientiWc Research (NWO) to the second author. References Atkinson, R. C., & ShiVrin, R. M. (1968). Human memory: a proposed system and its control processes. Psychology of Learning and Motivation, 2, 89–195. Botvinick, M. M., Braver, T. S., Barch, D. M., Carter, C. S., & Cohen, J. D. (2001). ConXict monitoring and cognitive control. Psychological Review, 108(3), 624–652. Bub, D. N., Masson, M. E. J. & Lalonde, C. E. (2006). Cognitive control in children: Stroop interference and suppression of word reading. Phychological Science, 17(4), 351–357. Burle, B., Possamaï, C.-A., Vidal, F., Bonnet, M., & Hasbroucq, T. (2002). Executive control in the Simon eVect: an electromyographic and distributional analysis. Psychological Research, 66, 324–339. 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(5), 619–641. Cohen, J. D., Dunbar, K., & McClelland, J. L. (1990). On the control of automatic processes: a parallel distributed processing model of the Stroop eVect. Psychological Review, 97(3), 332–361. de Jong, R., Liang, C., & Lauber, E. (1994). Conditional and unconditional automaticity: a dual-process model of eVects of spatial stimulus– response correspondence. Journal of Experimental Psychology: Human Perception and Performance, 20(4), 731–750. Eimer, M. (1999). Facilitatory and inhibitory eVects of masked prime stimuli on motor activation and behavioural performance. Acta Psychologica, 101(2–3), 293–313. Eimer, M., Hommel, B., & Prinz, W. (1995). S–R compatibility and response selection. Acta Psychologica, 90, 301–313. Eimer, M., & Schlaghecken, F. (1998). EVects of masked stimuli on motor activation: behavioral and electrophysiological evidence. Journal of Experimental Psychology: Human Perception and Performance, 24, 1737–1747. Eimer, M., & Schlaghecken, F. (2001). Response facilitation and inhibition in manual, vocal and oculomotor performance; evidence for a modality-unspeciWc mechanism. Journal of Motor Behaviour, 33(1), 16–26. Eriksen, B. A., & Eriksen, C. W. (1974). EVects of noise letters upon the identiWcation of target-letters in a non-search task. Perception & Psychophysics, 16, 249–263. Frith, C. D., & Done, D. J. (1986). Routes to action in reaction time tasks. Psychological Research, 48, 169–177.

269

Godijn, R., & Theeuwes, J. (2002). Programming of endogenous and exogenous saccades: evidence for a competitive integration model. Journal of Experimental Psychology: Human Perception and Performance, 28, 1039–1054. Hallet, P. E. (1978). Primary and secondary saccades to goals deWned by instructions. Vision Research, 18, 1279–1296. James, W. (1890). The principles of psychology. New York: Dover Publications. Klein, R. M. (2000). Inhibition of return. Trends in Cognitive Sciences, 4, 138–147. Kornblum, S., Hasbroucq, T., & Osman, A. (1990). Dimensional overlap: Cognitive basis for stimulus–response compatibility—A model and taxonomy. Psychological Review, 97, 253–270. Miller, E. K., & Cohen, J. D. (2001). An integrative theory of prefrontal cortex function. Annual review of Neuroscience, 24, 167–202. McCormick, P. A. (1997). Orienting attention without awareness. Journal of Experimental Psychology: Human Perception and Performance, 23(1), 168–180. Munoz, D. P., & Everling, S. (2004). Look away: the anti-saccade task and the voluntary control of eye-movement. Nature Reviews Neuroscience, 5, 218–228. Posner, M. I., & Cohen, Y. (1984). Components of visual orienting. In H. Bouma & D. G. Bouwhuis (Eds.), Attention and performance X (pp. 531–556). Hove, UK: Lawrence Erlbaum. Posner, M. I., Rafal, R. D., Choate, L. S., & Vaughan, J. (1985). Inhibition of return: neural basis and function. Cognitive Neuropsychology, 2, 211–238. Proctor, R. W., Lu, C-H., Wang, H., & Dutta, A. (1995). Activation of response codes by relevant and irrelevant stimulus information. Acta Psychologica, 90, 275–286. RatcliV, R. (1979). Group reaction time distributions and an analysis of distribution statistics. Psychological Bulletin, 86(3), 446–461. Reuter, B., & Kathmann, N. (2004). Using saccade tasks as a tool to analyze executive dysfunctions in schizophrenia. Acta Psychologica, 115, 255–269. Ridderinkhof, K. R. (2002). Activation and suppression in conXict tasks: empirical clariWcation through distributional analyses. In W. Prinz & B. Hommel (Eds.), Common mechanisms in perception and action. Attention and performance (Vol. XIX, pp. 494–519). Oxford: Oxford University Press. Ridderinkhof, K. R., van den Wildenberg, W. P. M., Segalowitz, S., & Carter, C. S. (2004a). Neurocognitive mechanisms of cognitive control: the role of prefrontal cortex in action selection, response inhibition, performance monitoring, and reward-based learning. Brain and Cognition, 56(2), 129–140. Ridderinkhof, K. R., van den Wildenberg, W. P. M., Wijnen, J. G., & Burle, B. (2004b). Response inhibition in conXict tasks is revealed in delta plots. In M. Posner (Ed.), Cognitive Neuroscience of Attention (pp. 369– 377). New York: Guilford Press. 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 Xanker eVects on target processing. Acta Psychologica, 90, 29–48. Sanders, A. F. (1967). Some aspects of reaction processes. Acta Psychologica, 27, 115–130. Schlaghecken, F., & Eimer, M. (2000). A central–peripheral asymmetry in masked priming. Perception & Psychophysics, 62(7), 1367–1382. ShiVrin, R. M., & Schneider, W. (1977). Controlled and automatic human information processing: II. perceptual learning, automatic attending, and a general theory. Psychological Review, 84, 127–190. Simon, J. R. (1969). Reactions toward the source of stimulation. Journal of Experimental Psychology, 81, 174–176. Stroop, J. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18, 643–662. Theeuwes, J., Kramer, A. F., Hahn, S., & Irwin, D. E. (1998). Our eyes do not always go where we want them to go: capture of the eyes by new objects. Psychological Science, 9, 379–385.

270

J.G. Wijnen, K.R. Ridderinkhof / Brain and Cognition 63 (2007) 260–270

Trappenberg, T. P., Dorris, M. D., Munoz, D. P., & Klein, R. M. (2001). A model of saccade initiation based on the competitive integration of exogenous and endogenous signals in the superior colliculus. Journal of Cognitive Neuroscience, 13, 256–271. Vincent, S. B. (1912). The function of the vibrissae in the behavior of the white rat. Behavioral Monographs, 1, 1–82.

Wagenmakers, E.-J., Grasman, R. P. P. P., & Molenaar, P. C. M. (2005). On the relation between the mean and the variance of a diVusion model response time distribution. Journal of Mathematical Psychology, 49, 195–204. Wijnen, J. G. & Ridderinkhof, K. R. Top-down control over attentional and oculomotor capture as elicited by salient color-changes, submitted for publication.