Psychological Research (2000) 64: 11±24

Ó Springer-Verlag 2000

ORIGINAL ARTICLE

Robert W. Proctor á Julie Gerred Marble Kim-Phuong L. Vu

Mixing incompatibly mapped location-relevant trials with location-irrelevant trials: effects of stimulus mode on the reverse Simon effect Received: 21 December 1999 / Accepted: 26 April 2000

Abstract When location-relevant trials with an incompatible spatial stimulus-response mapping are mixed with location-irrelevant trials, responses on the latter trials are faster when stimulus and response locations do not correspond than when they do. Experiments 1 and 2 showed that this reverse ``Simon e€ect'' also occurs when the location information is presented verbally or symbolically on both location-relevant and locationirrelevant trials. The reversal was absent, however, in conditions of Experiments 1±3 in which the mode of presentation was di€erent on the location-relevant trials than on the location-irrelevant trials. Experiment 4 demonstrated that di€erences in physical characteristics between the location-relevant and location-irrelevant stimuli were not sucient to eliminate the reverse Simon e€ect. These ®ndings imply that the short-term associations between stimulus location information and responses de®ned for the location-relevant task are relatively mode speci®c.

Introduction In a typical two-choice spatial reaction task, left and right stimulus locations are mapped to left and right keypresses. The usual ®nding is that responses are faster and more accurate when the left stimulus is mapped to the left response and the right stimulus to the right response than when the mapping is reversed (e.g., Dutta & Proctor, 1992; Sha€er, 1965). Moreover, when stimulus location is irrelevant and a non-spatial attribute (e.g., color) conveys the relevant information, responses are

R. W. Proctor (&) á J. Gerred Marble á K.-P. L. Vu Department of Psychological Sciences, Purdue University, West Lafayette, IN 47907-1364, USA Tel.: +1-765-4940784; Fax: +1-765-4961264 e-mail: [email protected]

faster and more accurate when the stimulus and response locations correspond than when they do not (e.g., Lu & Proctor, 1995; Simon, 1990; UmiltaÁ & Nicoletti, 1990). This latter e€ect, called the Simon e€ect, has attracted considerable interest in recent years (e.g., Hommel & Prinz, 1997) because the task requirements provide no reason to activate responses on the basis of stimulus location. Most accounts of the Simon e€ect attribute it to automatic activation of the response that corresponds to the stimulus location. A distinction is often made between task-de®ned, short-term associations and long-term associations (e.g., Barber & O'Leary, 1997; Sto€er & UmiltaÁ, 1997; Zorzi & UmiltaÁ, 1995). The instructions to respond to the relevant stimulus dimension with keypresses are presumed to de®ne short-term associations between the respective stimuli and their assigned responses in working memory. The long-term associations are those between stimulus locations and response locations that are hard-wired or overlearned from years of experience. The automatic activation of the corresponding response that produces the Simon e€ect is attributed to these long-term associations. However, the Simon e€ect is malleable and can be reversed by several manipulations. Hedge and Marsh (1975) provided the ®rst demonstration of such a reversal. They labeled the response keys with colors and found that when the mapping of stimulus colors to response colors was incompatible, the Simon e€ect reversed: Responses were slower when stimulus and response locations corresponded than when they did not. This ``Hedge and Marsh reversal'' has been replicated in numerous other studies (e.g., Hommel, 1995; Lu & Proctor, 1994). The most widely accepted account of this reversal is that of ``logical recoding'', proposed initially by Hedge and Marsh, according to which the ``respond opposite'' mapping rule that is appropriate for the incompatible mapping of stimulus color to response color is also applied to the irrelevant stimulus-location information.

12

Secondly, Proctor and Lu (1999) showed that when subjects practice with an incompatible spatial mapping and then are transferred to the Simon task, a reverse Simon e€ect is evident that persists across at least 600 trials. Tagliabue, Zorzi, UmiltaÁ, and Bassignani (2000) showed similar persistent e€ects of practice with a prior incompatible location mapping and considered whether these e€ects were due to modi®cation of the long-term associations or a long-lasting contribution of the shortterm associations. Based on ®ts of a modi®ed version of Zorzi and UmiltaÁ's (1995) network model to the response time (RT) distributions, they concluded that the e€ect of the prior mapping was due to the short-term associations. In sum, the ®ndings of Proctor and Lu and of Tagliabue et al. imply that the short-term associations between incompatible stimulus and response locations de®ned when stimulus location is relevant continue to a€ect performance when location is no longer relevant. A third way to reverse the Simon e€ect is to intermix location-relevant trials for which the mapping is incompatible with location-irrelevant trials for which a non-spatial stimulus attribute is relevant. Marble and Proctor (in press) had stimuli vary in color and location. When the stimulus was red or green, the assigned keypress response was to be selected on the basis of the color (location-irrelevant trials). When the stimulus was white, one of the same two keypress responses was to be selected on the basis of stimulus location (left or right; location-relevant trials). For conditions in which the location-relevant mapping was incompatible, the location-irrelevant trials showed a reverse Simon e€ect. That is, responses were faster when stimulus and response locations did not correspond than when they did. This reversal of the Simon e€ect was at least as large as the positive Simon e€ect obtained for conditions in which the location-relevant mapping was compatible. Because the long-term associations between corresponding locations should add to the contribution of the short-term associations when the mapping for the location-relevant trials is compatible and subtract from their contribution when the mapping is incompatible, the fact that the reverse Simon e€ect is as large as the positive e€ect suggests that the long-term associations were not contributing to the e€ects. Rather, the reverse Simon e€ect seems to be a result of the short-term associations between noncorresponding locations de®ned for the location-relevant task. Activation via these associations is not simply a consequence of uncertainty about whether location or color will be relevant on a particular trial because the reversal was obtained in Marble and Proctor's Experiment 3 when the trial type (location-relevant or location-irrelevant) was precued by as much as 2400 ms before the imperative stimulus. The results of the mixing experiments conducted by Marble and Proctor (in press) indicate that subjects are unable to prevent activation of the noncorresponding response by way of the short-term location association

for trials on which location is irrelevant. In their experiments, the location information was conveyed by physical location, but such information can also be conveyed by location words and by left- or right-pointing arrows. Similar e€ects of an incompatible mapping of location information to responses on location-irrelevant trials should also occur when the information on both trial types is presented by word or arrow direction. These e€ects would be expected because the short-term associations of location words (or arrow directions) to incompatible responses should produce activation of the noncorresponding response for trials on which the word (or arrow direction) is irrelevant. An interesting issue is whether the reversal of the Simon e€ect on location-irrelevant trials when the location-relevant mapping is incompatible will occur when the mode of the location information di€ers on the two trial types. E€ects similar to those with same-mode presentation are predicted if it is assumed that the shortterm associations producing the activation on locationirrelevant trials are between the concepts left and right and the left and right responses. According to Kornblum's (1992) dimensional overlap model, conceptual similarity is a major factor contributing to dimensional overlap and, hence, stimulus-response (S-R) compatibility. In fact, the task taxonomy based on the model classi®es all tasks in terms of conceptual similarity between the relevant stimulus dimension and the response dimension, the irrelevant stimulus dimension and the response dimension, and the relevant and irrelevant stimulus dimensions. Zhang and Kornblum (1998) developed a network model for the Stroop task based on conceptual similarity in which, prior to the output units, ink colors and color words converge onto an intermediate module, which represents the concept of color. Short-term, task-speci®c associations in the model are between the intermediate module that represents the concept of color and the output module. The implication of this model is that, when the mode by which the location information is presented di€ers on location-relevant and location-irrelevant trials, the irrelevant stimulus mode should activate the same response as that assigned for the relevant stimulus mode. Thus, for spatial location, the irrelevant information should activate the assigned response even if the input mode is verbal for location-relevant trials and nonverbal for locationirrelevant trials, or vice versa. Consequently, a reversal of the Simon e€ect similar to that obtained in samemode conditions should occur when the location information on both location-relevant and location-irrelevant trials is conveyed by di€erent modes. If it is assumed that the short-term associations are speci®c to the stimulus modes, then the prediction is that the reversed Simon e€ect should be evident only in same-mode conditions and not in mixed-mode conditions. Although not implemented in Kornblum's (1992) taxonomy, his de®nition of dimensional overlap includes physical, or mode, similarity. Several studies have shown that mode similarity also contributes to e€ects of irrel-

13

evant information on performance (e.g., Lu, 1997; Lu & Proctor, in press), suggesting that the processing of visuo-spatial stimuli is at least partially distinct from that of verbal stimuli (e.g., Virzi & Egeth, 1985; Wickens, 1984). The stimulus mode distinction is captured in a parallel distributed processing model of the Stroop color-naming e€ect developed by Cohen, Dunbar, and McClelland (1990). The model contains separate pathways for color information and color-word information that converge only at output units that represent the responses. This model implies that the associations between verbal and nonverbal stimulus properties and responses are distinct. Extending the model to the concept of spatial information, the implication is that a locationrelevant mapping de®ned for one stimulus mode should have little impact on the response activation produced by the other stimulus mode on location-irrelevant trials. Hence, an incompatible location mapping should have little e€ect on location-irrelevant trials when the stimulus modes for the two trial types di€er. The purpose of the present study thus was to evaluate the roles played by conceptual and mode similarity in the reversal of the Simon e€ect that occurs when location-irrelevant trials are mixed with incompatibly mapped location-relevant trials. Experiment 1 was conducted in part to replicate the results of the incompatible mixing condition of Marble and Proctor (in press) in which the location information on both location-relevant and location-irrelevant trials was conveyed by physical locations. In addition, a condition was examined in which the location information was conveyed by words on both location-relevant and location-irrelevant trials to determine whether the reverse Simon e€ect would generalize to verbal location information. Conditions were also conducted in which the two modes of presentation (location words and physical locations) were mixed such that if the location-relevant mode was verbal, the location-irrelevant mode was physical location, and vice versa. These mixed conditions enabled evaluation of whether presenting the relevant location information in one mode and the irrelevant location information in another eliminates the reverse Simon effect. Experiment 2 was similar to Experiment 1, except that left- and right-pointing arrows were used instead of physical locations. This allowed the generalizability of the results to symbolic stimuli to be evaluated, as well as determination of whether a mode distinction for the relevant and irrelevant location information is sucient to eliminate the reverse Simon e€ect when all stimuli occur in the same location. In Experiment 3, mixed mode conditions involving arrows and physical locations were examined to determine whether the reverse Simon e€ect would be eliminated when both relevant and irrelevant location modes were nonverbal. Finally, Experiment 4 considered the possibility that the reverse Simon e€ect would be eliminated by any change that decreased the physical similarity of the location-relevant and location-irrelevant stimuli, not just di€erences in stimulus mode.

Experiment 1 In this experiment, we replicated the condition from Marble and Proctor's (in press) Experiment 1 in which the location dimension was physical on both locationrelevant (with an incompatible mapping) and locationirrelevant trials. A second condition was also conducted in which the location dimension was verbal on both location-relevant and location-irrelevant trials (i.e., the words ``right'' and ``left'' were mapped to the left and right responses, respectively). Because the ``left'' and ``right'' words are conceptually similar to the left and right keypress responses, we expected to replicate the reverse Simon e€ect for the physical condition and to ®nd a similar e€ect for the verbal condition. The correspondence e€ect for location words and keypresses will be referred to as a Simon e€ect, consistent with the fact that both the word and location versions of the task are classi®ed as Type 3 ensembles in Kornblum's (1992) taxonomy, although the label sometimes is restricted to the case where the irrelevant information is physical location. Because the set of location words is less compatible with the set of keypress responses than is the set of physical locations (Proctor & Wang, 1997; Wang & Proctor, 1996), the tendency to activate the noncorresponding response when the location information is irrelevant may be less for the location words than for the physical locations, thus producing a weaker reverse Simon e€ect. In the remaining two conditions, the mode of presentation (position and word) for the location information was di€erent for the location-relevant trials than for the location-irrelevant trials. In one condition, the location-relevant information was physical location and location-irrelevant information was location word, and in the other condition the mode relationship was the opposite. If the reversal of the Simon e€ect still occurs when the location information di€ers in mode on location-relevant and location-irrelevant trials, then it will suggest that the short-term associations between stimulus location information and responses are conceptual in nature. In contrast, if the reversal of the Simon e€ect is eliminated, then this will suggest that the short-term associations that translate location information into activation of the opposing response are dependent on the mode in which this information is presented. Method Subjects. Sixty-four students enrolled in Introductory Psychology at Purdue University participated in the experiment in partial ful®llment of course credit. All subjects had normal or correctedto-normal vision and were naive to the purpose of the study. Apparatus and stimuli. The experiment was conducted in a dimly lit room. An IBM-compatible microcomputer with a 14-in. VGA color monitor was used for stimulus generation. Response registration and timing were controlled by Micro Experimental

14 Laboratory (MEL). The subjects sat directly in front of the monitor, at a viewing distance of approximately 55 cm. The stimuli were red, green, or white (MEL color codes 4, 2 and 15, respectively) circles, 5-mm (0.52°) diameter, presented in left or right locations (60 mm from center at 6.23° viewing angle) on the computer screen, and red, green, or white words ``left'' or ``right'' (approximately 12 mm ´ 5 mm and 15 mm ´ 5 mm, visual angles of 1.56° ´ 0.52° and 1.24° ´ 0.52°, respectively), presented at the center of the screen. Responses were made by pressing the ``/'' and ``z'' keys on the computer keyboard, located at the extreme right and left positions of the bottom row, with the index ®ngers of each hand. Procedure. Sixteen subjects were tested in each of the four conditions. In the PositionI/PositionR condition, the location information was conveyed by physical position on both locationirrelevant (I) and location-relevant (R) trials. In this condition, the stimuli were presented in the left or right locations, being red or green circles on half of the trials and white circles on the other half, randomly intermixed. Subjects were to respond with a right or left keypress according to the color of the red and green circles or to the location of the white circles. The WordI/WordR condition was similar to the PositionI/PositionR except that both the colored and white stimuli were location words presented in the center of the screen. In the PositionI/WordR condition, the location irrelevant trials were red or green circles presented in left or right positions, and the location-relevant trials were white location words presented in the center of the screen. In the WordI/PositionR condition, the location irrelevant trials were red or green location words presented in the center of the screen, and the location-relevant trials were white circles presented in left or right positions. For all conditions, the spatial mapping of the white stimuli to response keys was incompatible, that is, the left response was made to the right stimulus and the right response to the left stimulus. Each subject was tested in a single session consisting of 400 trials, 200 of which were location-irrelevant trials for which the relevant stimulus dimension was color and 200 of which were location-relevant trials. The order of trial types was randomized in all conditions. For the red and green stimuli, subjects were instructed to ignore the location information provided by the stimulus when selecting the correct response. Half of the subjects in each condition were to respond to a red stimulus with the left response and a green stimulus with the right response, and the other half were to respond with the opposite mapping. When the white stimuli were presented, subjects were to respond to the stimulus location (or location word) by pressing the noncorresponding response key. Subjects were instructed to respond as quickly and accurately as possible without making too many mistakes. The stimulus remained present until a response was made. The intertrial interval was 1 s, and it was initiated immediately if the response was correct or after a 400-Hz, 500-ms ``error'' tone if the response was incorrect.

Results Mean RT for correct responses was determined as a function of the subconditions. RTs less than 200 ms or greater than 2000 ms (<1% in this experiment and in the remaining experiments) were excluded from analysis. Proportion of error (PE) was analyzed separately. The ®rst 20 trials were considered practice and excluded from analysis. Location-irrelevant trials Table 1 contains the RT and PE data for the locationirrelevant trials. For both RT and PE, a 2 (Stimulus

Table 1 Mean RT and PE (in parentheses) for the location-irrelevant task in Experiment 1 as a function of condition (PositionI/ PositionR, PositionI/WordR, WordI/PositionR, WordI/WordR), stimulus location, and response location Stimulus location

Response location Left

Right

Left Right

PositionI/PositionR 646 (0.105) 582 (0.029)

603 (0.021) 667 (0.137)

Left Right

PositionI/WordR 569 (0.044) 574 (0.034)

576 (0.017) 573 (0.044)

Left Right

WordI/PositionR 611 (0.032) 644 (0.060)

649 (0.052) 618 (0.031)

Left Right

WordI/WordR 776 (0.067) 737 (0.022)

706 (0.015) 750 (0.064)

Location: left or right) ´ 2 (Response Location: left or right) ´ 2 (Location-Relevant Mode: position or word) ´ 2 (Location-Irrelevant Mode: position or word) ANOVA was conducted. Stimulus and response locations were within-subjects factors, and location-relevant and location-irrelevant modes were between-subjects factors. For the RT data, the main e€ect of location-irrelevant mode was signi®cant, F(1, 60) ˆ 14.57, p < 0.001, MSE ˆ 33,756, as was the interaction of irrelevant-location mode and relevant-location mode, F(1, 60) ˆ 12.62, p < 0.001, MSE ˆ 33,756. Responses were faster when the irrelevant location information was conveyed by physical position (M ˆ 599 ms) than by word (M ˆ 687 ms). When the irrelevant dimension was physical position, responses were faster if the location-relevant stimuli were words (M ˆ 573 ms) than if they were physical positions (M ˆ 624 ms). However, when the irrelevant dimension was location word, responses were faster if the location-relevant stimuli were physical positions (M ˆ 631 ms) than if they were words (M ˆ 742 ms). Stimulus location interacted with response location, F(1, 60) ˆ 7.08, p ˆ 0.010, MSE ˆ 7793, for RTs. This interaction re¯ects an overall reversal of the Simon effect, with responses being slower when the stimulus and response locations corresponded than when they did not (652 ms and 634 ms, respectively). However, the four-way interaction of Stimulus Location ´ Response Location ´ Location-Irrelevant Mode ´ Location-Relevant Mode was also signi®cant, F(1, 60) ˆ 28.74, p < 0.001, MSE ˆ 7793. The PositionI/PositionR and WordI/WordR conditions showed reverse Simon e€ects of 64 ms (M ˆ 656 ms, corresponding versus 592 ms, noncorresponding), F(1, 15) ˆ 25.28, p < 0.001, MSE ˆ 2610, and 41 ms (763 versus 722 ms), F(1, 15) ˆ 4.75, p ˆ 0.046, MSE ˆ 5877, respectively. In contrast, the PositionI/WordR condition showed no signi®cant di€erence between corresponding and

15

noncorresponding locations (571 versus 575 ms), F(1, 15) < 1.0, and the WordI/PositionR condition showed a positive Simon e€ect of 31 ms (615 versus 646 ms), F(1, 15) ˆ 7.80, p ˆ 0.014, MSE ˆ 2055. For the error data, the main e€ect of location-relevant mode was signi®cant, F(1, 60) ˆ 4.14, p ˆ 0.046, MSE ˆ 0.0063, with responses more accurate when the relevant stimulus information was conveyed by words (PE ˆ 0.038) than by physical positions (PE ˆ 0.058). The interaction of location-irrelevant mode and location-relevant mode did not quite attain the 0.05 level, F(1, 60) ˆ 3.37, p ˆ 0.071, MSE ˆ 0.0063, but the error data showed a similar pattern as the RT data. When physical position was irrelevant, responses were more accurate if the location-relevant stimuli were words (M ˆ 0.035) than if they were physical positions (M ˆ 0.073). However, when location word was irrelevant, there was no di€erence in accuracy when the location-relevant stimuli were physical positions (M ˆ 0.043) and when they were words (M ˆ 0.041). The error data also showed a signi®cant main e€ect of stimulus location, F(1, 60) ˆ 6.04, p ˆ 0.017, MSE ˆ 0.0073, in which responses were more accurate to the left location (M ˆ 0.044) than the right location (M ˆ 0.052). The two-way interaction of Stimulus Location ´ Response Location and three-way interaction of Stimulus Location ´ Response Location ´ LocationIrrelevant mode were also signi®cant, Fs(1, 60) ˆ 28.21 and 12.77, ps < 0.001, MSE ˆ 0.0027. However, these interactions were quali®ed by a signi®cant four-way interaction of the three factors with location-relevant mode, F(1, 60) ˆ 33.55, p < 0.001, MSE ˆ 0.0027. The PositionI/PositionR and WordI/WordR conditions showed reverse Simon e€ects of 0.096 (0.12, corresponding versus 0.025, noncorresponding), F(1, 15) ˆ 27.76, p < 0.001, MSE ˆ 0.0053, and 0.047 (0.065 versus 0.018), F(1, 15) ˆ 27.93, p < 0.001, MSE ˆ 0.0013, respectively. In contrast, the PositionI/ WordR condition showed no signi®cant di€erence between corresponding and noncorresponding locations (0.044 versus 0.025), F(1, 15) ˆ 2.69, p ˆ 0.122, MSE ˆ 0.0021, and the WordI/PositionR condition showed a positive Simon e€ect of 0.025 (0.031 versus 0.056), F(1, 15) ˆ 5.03, p ˆ 0.040, MSE ˆ 0.002. Location-relevant trials For both the RT and PE data, a 2 (Response Location: left or right; stimulus location is the opposite of response location) ´ 2 (Location-Relevant Mode: position or word) ´ 2 (Location-Irrelevant Mode: position or word) ANOVA was conducted. Response location was a within-subjects factor and location-relevant and location-irrelevant modes were between-subjects factors. For the RT data (Table 2), responses were faster when the location-relevant stimuli were physical positions (M ˆ 554 ms) than when they were words (M ˆ 642 ms), F(1, 60) ˆ 12.57, p < 0.001, MSE ˆ 19,507.

Table 2 Mean RT and PE (in parentheses) for the location-relevant task in Experiments 1±4 as a function of response location and mixing condition Mixing condition

Response location Left

Right

PositionI/PositionR PositionI/WordR WordI/PositionR WordI/WordR

Experiment 1 584 (0.044) 482 (0.040) 517 (0.026) 790 (0.037)

596 472 519 823

(0.051) (0.034) (0.019) (0.046)

ArrowI/ArrowR ArrowI/WordR WordI/ArrowR WordI/WordR

Experiment 2 681 (0.083) 724 (0.082) 592 (0.043) 823 (0.089)

659 669 565 773

(0.064) (0.076) (0.034) (0.074)

ArrowI/PositionR PositionI/ArrowR

Experiment 3 558 (0.021) 546 (0.037)

558 (0.024) 551 (0.026)

PositionI/PositionR ArrowI/ArrowR WordI/WordR

Experiment 4 640 (0.021) 610 (0.024) 837 (0.040)

624 (0.015) 608 (0.014) 806 (0.035)

The PositionI/PositionR, ArrowI/ArrowR, and WordI/WordR conditions in Experiment 4 include a form distinction that was not present in the corresponding conditions of Experiments 1±3

In addition, responses were faster when the locationirrelevant stimuli were physical positions (M ˆ 533 ms) than when they were words (M ˆ 663 ms), F(1, 60) ˆ 27.34, p < 0.001, MSE ˆ 19,507. However, both of the main e€ects were quali®ed by a Location-Relevant Mode ´ Location-Irrelevant Mode interaction, F(1, 60) ˆ 65.40, p < 0.001, MSE ˆ 19,507. When the location-relevant stimuli were physical positions, responses were faster when they were mixed with location-irrelevant words (M ˆ 520 ms) than with locationirrelevant physical positions (M ˆ 590 ms). When the location-relevant stimuli were words, responses were faster when they were mixed with location-irrelevant physical positions (M ˆ 477 ms) than with locationirrelevant words (M ˆ 806 ms). The RTs also showed a three-way interaction of Response Location ´ Location-Relevant Mode ´ Location-Irrelevant Mode, F(1, 60) ˆ 4.99, p ˆ 0.029, MSE ˆ 1140. When the modes of presentation for the location-relevant and location-irrelevant stimuli were the same, left responses were faster than right responses [Mean di€erence (MD) ˆ 13 ms for the PositionI/PositionR condition and 33 ms for the WordI/WordR condition]. When the modes of presentation for the location-relevant and location-irrelevant stimuli were di€erent, left responses were slower than right responses for the PositionI/WordR condition (MD ˆ )10 ms) and of similar magnitude for the WordI/PositionR condition (MD ˆ 2 ms). For the PE data, there was also a three-way interaction of Response Location ´ Location-Relevant Mode ´ Location-Irrelevant Mode, F(1, 60) ˆ 4.32, p ˆ 0.042, MSE ˆ 0.00033. When the modes of presentation

16

for the location-relevant and location-irrelevant stimuli were the same, left responses were more accurate than right responses (MD ˆ 0.005 for PositionI/PositionR and 0.009 for WordI/WordR). When the modes of presentation for the location-relevant and location-irrelevant stimuli were di€erent, left responses were less accurate than right responses (MDs ˆ 0.005 for PositionI/WordR and WordI/PositionR). Discussion The reversal of the Simon e€ect found by Marble and Proctor (in press) was replicated for the PositionI/PositionR condition, in which both tasks used stimuli in left and right locations. Moreover, a numerically smaller reversal was also obtained for the WordI/WordR condition in which the location information was conveyed by the location word. Thus, a location-relevant mapping has an e€ect on location-irrelevant trials regardless of whether the location information on both trial types is spatial or verbal. The reverse Simon e€ect was not evident, however, for the PositionI/WordR and WordI/PositionR conditions in which the location information was presented in di€erent modes on location-relevant and locationirrelevant trials. For the WordI/PositionR condition a positive Simon e€ect was obtained for the locationirrelevant trials, implying that the incompatible mapping of stimulus locations to responses had little or no impact on the e€ect produced by an irrelevant location word. However, for the PositionI/WordR condition there was neither a negative nor positive Simon e€ect. To evaluate whether mixing eliminated the in¯uence of the location-relevant mapping entirely, it is necessary to know whether a Simon e€ect would be obtained when the stimuli are similar to those in the mixed conditions of Experiment 1, but the location information is irrelevant on all trials. Therefore, we conducted an experiment in which stimulus color was relevant for all trials. A colored circle occurred in the left position on 25% of the trials and the right position on 25% of the trials, and a colored word occurred in the centered position on the remaining trials (25% the word ``left'' and 25% the word ``right''). Sixteen subjects were tested, and the total number of trials was the same as in Experiment 1. Positive Simon e€ects were obtained for both the circles and words, being 25 ms for the former and 23 ms for the latter. A similar experiment was also conducted using centered colored circles on 50% of the trials, rather than the colored words. In this experiment, a colored circle occurred in the left position on 25% of the trials and the right position on 25% of the trials, and in the center position on 50% of the trials. Results showed a positive Simon e€ect of 32 ms. The results for these two control experiments suggest that mixing modes entirely eliminates the in¯uence of a relevant mapping of physical locations on responses when location word is irrelevant

but only reduces the e€ect of a relevant location-word mapping when physical location is irrelevant. For both the location-irrelevant and location-relevant tasks, responses were faster and tended to be more accurate when the location mode was di€erent on the two trial types than when it was the same. The results for both of the trial types thus imply that the locationirrelevant task can be distinguished more readily from the location-relevant task when one involves centered location words and the other circles in left and right locations. The reduced diculty of the tasks when the mode in which the location information is presented is di€erent on location-relevant and location-irrelevant trials, along with the elimination of the reverse Simon e€ect, is consistent with the hypothesis that the shortterm associations de®ned for the location-relevant task are mode speci®c. Subjects had particular diculty when both the location-relevant and location-irrelevant trials involved location words (WordI/WordR condition). In fact, mean RT for the location-relevant trials almost doubled when the location-irrelevant trials used location words rather than physical locations. We can only speculate as to why it is so dicult to perform when both tasks involve location words. The diculty may be linked speci®cally to word stimuli: When it is known that the content of the location word will always be relevant, the shortterm association between the word meaning and the assigned incompatible keypress response can be maintained in working memory, thus allowing rapid generation of the appropriate response when the word is identi®ed. However, when there is uncertainty about whether the content of the location word will be relevant or irrelevant, the word is named while the decision regarding the color dimension is being made. This word name produces response activation that must be inhibited before the appropriate incompatible response can be selected. Although the lack of a reverse Simon e€ect in the PositionI/WordR and WordI/PositionR conditions could be due to the di€erence in mode for the location stimuli (verbal, nonverbal), another possibility is that the difference between the positions in which the location-relevant words occurred and the circles occurred when color was relevant provided a basis for easily distinguishing the two tasks. Thus, Experiments 2±4 evaluated whether the mode distinction or the position distinction is the crucial factor.

Experiment 2 Experiment 2 was similar to Experiment 1, except that arrows rather than physical positions were used. As in Experiment 1, there were two conditions in which the location information on location-relevant and locationirrelevant trials was in the same mode. In one condition the mode of presentation was verbal (word: ``left'' or ``right''), and in the other it was symbolic (arrow

17

direction: left or right). The WordI/WordR condition is the same as in Experiment 1 and should replicate the reverse Simon e€ect for location-irrelevant trials found in that experiment. The ArrowI/ArrowR condition allows us to determine whether the Simon e€ect is reversed by an incompatible location-relevant mapping when the presentation mode is nonverbal and symbolic. If the short-term associations between the location information and responses are responsible for the reversal, it should be evident for the arrow stimuli as it was for physical locations and location words in Experiment 1. In the previous experiment, a reversal of the Simon e€ect was not obtained when the location-irrelevant circles were intermixed with the location-relevant words ``left'' and ``right,'' and vice versa. Experiment 2 pursued this ®nding further using mixed mode conditions in which words conveyed the location information for one task and arrows for the other task. These conditions should provide evidence regarding whether the lack of reversal depends on using di€erent stimulus modes for the two tasks or having the stimuli occur in distinct locations. If the ®rst factor is crucial, the reversal should not be evident in this experiment when stimulus sets are mixed, but if the second factor is crucial, the reversal should be evident. In summary, by varying factorially whether the relevant and irrelevant information is conveyed by word meaning or arrow direction, it is possible to evaluate the generalizability of the ®ndings that (a) mixing location-relevant and location-irrelevant trials of the same stimulus modes produces a reverse Simon e€ect and (b) mixing trials of di€erent stimulus modes does not. Method Subjects. Sixty-four new undergraduates from the same subject pool as in Experiment 1 participated. Sixteen were assigned randomly to each of four conditions. Procedure. The procedure was similar to the previous experiment. Two types of stimuli conveying spatial information were used, the words ``left'' or ``right,'' and right- or left-pointing arrows. All stimuli were presented in the center of the screen. There were four conditions, varied between subjects, in all of which the locationrelevant stimuli were mapped incompatibly to the responses. The location-irrelevant stimuli were presented in red or green, while all location-relevant stimuli were presented in white. For two conditions, the stimuli for the location-irrelevant trials were location words (same size and visual angle as in Experiment 1); for one of these conditions the stimuli for the location-relevant trials were the also location words (the WordI/WordR condition), while for the second, the stimuli were right- or left-pointing arrows (the WordI/ArrowR condition). The arrows were constructed from two equal signs and a carat (<ˆˆ or ˆˆ>, which were approximately 10 mm ´ 5 mm, or 1.04° ´ 0.52° of visual angle). For the remaining two conditions, the stimuli for the location-irrelevant trials were the arrows presented in red or green; in one condition the location-relevant stimuli were arrows presented in white (the ArrowI/ArrowR condition), while in the other they were the location words presented in white (the ArrowI/WordR condition).

Results The RT and PE data were analyzed in the same manner as in Experiment 1. In addition, the same criteria for exclusion of outliers were used. Location-irrelevant trials The RT data (Table 3) showed a trend toward a main e€ect of location-irrelevant mode, F(1, 60) ˆ 3.62, p ˆ 0.062, MSE ˆ 50,728, with responses being slower when the location-irrelevant stimuli were words (M ˆ 713 ms) than when they were arrows (M ˆ 667 ms). There was also a signi®cant main e€ect of response location, F(1, 60) ˆ 4.56, p ˆ 0.037, MSE ˆ 4036, with left responses (M ˆ 682 ms) being faster than right responses (M ˆ 699 ms). Only two interactions were signi®cant, that of Location-Irrelevant Mode ´ Location-Relevant Mode, F(1, 60) ˆ 6.40, p ˆ 0.014, MSE ˆ 50,728, and the four-way interaction of those variables with stimulus location and response location, F(1, 60) ˆ 18.73, p < 0.0001, MSE ˆ 4119. The two-way interaction indicates that responses were slower overall when the relevant and irrelevant stimulus modes were the same (M ˆ 722 ms) than when they were di€erent (M ˆ 659 ms), as in Experiment 1. The four-way interaction re¯ects di€erent result patterns for the same and mixed mode conditions. A reverse Simon e€ect was evident for the same mode conditions (56 ms for the ArrowI/ArrowR condition and 26 ms for the WordI/WordR condition), being signi®cant for the ArrowI/ArrowR condition, F(1, 15) ˆ 50.87, p < 0.001, MSE ˆ 991, but not for the WordI/WordR condition, F(1, 15) ˆ 1.38, p ˆ 0.26, MSE ˆ 7824, respectively. In contrast, a positive Simon e€ect was obtained for the mixed mode conditions (33 ms for the ArrowI/WordR condition and 12 ms for the WordI/ArrowR condition), Table 3 Mean RT and PE (in parentheses) for the location-irrelevant task in Experiment 2 as a function of condition (ArrowI/ ArrowR, ArrowI/WordR, WordI/ArrowR, WordI/WordR), stimulus location, and response location Stimulus location

Response location Left

Right

Left Right

ArrowI/ArrowR 702 (0.125) 649 (0.037)

661 (0.039) 720 (0.105)

Left Right

ArrowI/WordR 633 (0.054) 667 (0.080)

668 (0.085) 637 (0.066)

Left Right

WordI/ArrowR 652 (0.071) 645 (0.130)

698 (0.107) 666 (0.073)

Left Right

WordI/WordR 762 (0.087) 744 (0.052)

753 (0.059) 787 (0.105)

18

being signi®cant for the ArrowI/WordR condition but not for the WordI/ArrowR condition, Fs(1, 15) ˆ 8.45 and 0.38, ps ˆ 0.011 and 0.547, MSEs ˆ 2028 and 6433. The PE data (Table 3) showed only two signi®cant e€ects, the three-way interaction of Location-Irrelevant Mode ´ Stimulus Location ´ Response Location and the four-way interaction of these variables with locationrelevant mode, Fs(1, 60) ˆ 5.05 and 50.97, ps < 0.029 and 0.0001, MSE ˆ 0.00277. As in the RT data, the four-way interaction re¯ects that a reverse Simon e€ect for errors was evident in the same mode conditions [0.077 for the ArrowI/ArrowR condition and 0.040 for the WordI/WordR condition, Fs(1, 15) ˆ 30.74 and 6.39, ps ˆ 0.00006 and 0.023, MSEs ˆ 0.00308 and 0.00407], whereas a positive Simon e€ect was evident in the mixed mode conditions [0.023 for the ArrowI/WordR condition and 0.047 for the WordI/ArrowR condition; Fs(1, 15) ˆ 4.13 and 20.54, ps ˆ 0.060 and 0.0004, MSEs ˆ 0.00223 and 0.00169]. The three-way interaction re¯ects that, averaged across location-relevant mode, the irrelevant arrows show a reverse Simon e€ect but the irrelevant words show no e€ect. Location-relevant trials The RT and PE data are shown in Table 2. The RT data showed a signi®cant main e€ect of location-relevant mode, F(1, 60) ˆ 30.02, p < 0.001, MSE ˆ 19,040, in which responses were faster when arrows were relevant (Ms ˆ 670 and 579 ms for the ArrowI/ArrowR and WordI/ArrowR conditions, respectively) than when location word was relevant (Ms ˆ 798 and 697 ms for the WordI/WordR and ArrowI/WordR conditions, respectively). There was also a signi®cant main e€ect for response location, F(1, 60) ˆ 19.15, p < 0.001, MSE ˆ 1722, in that left responses (M ˆ 705 ms) were slower than right responses (M ˆ 667 ms). In addition, the RT data showed two signi®cant interactions, LocationIrrelevant Mode ´ Location-Relevant Mode, F(1, 60) ˆ 14, 65, p < 0.001, MSE ˆ 19,040, and LocationIrrelevant Mode ´ Response Location, F(1, 60) ˆ 4.13, p ˆ 0.047, MSE ˆ 1722. The former, which is most important, is due to responses being faster when the location-irrelevant and location-relevant modes di€ered than when they were the same (Ms ˆ 638 and 734 ms, respectively). The latter is that left responses were relatively slower than right responses when the irrelevant stimulus mode was verbal than when it was symbolic. For the PE data, there was a marginally signi®cant main e€ect of location-relevant mode, F(1, 60) ˆ 3.92, p ˆ 0.052, MSE ˆ 0.00483. Responses tended to be more accurate when arrows were relevant (Ms ˆ 0.073 and 0.038, for the ArrowI/ArrowR and WordI/ArrowR conditions, respectively) than when location words were relevant (Ms ˆ 0.082 and 0.079, for the WordI/WordR and ArrowI/WordR conditions, respectively). The main e€ect for response location was also marginally signi®-

cant, F(1, 60) ˆ 3.57, p ˆ 0.064, MSE ˆ 0.00132, in which left responses (M ˆ 0.074) tended to be less accurate than right responses (M ˆ 0.062). There were no signi®cant interactions for the PE data. Discussion The reversal of the Simon e€ect on the location-irrelevant trials was replicated for the WordI/WordR condition (26 ms in RT and 0.040 in PE). The ArrowI/ArrowR condition showed an even stronger reversal of the Simon e€ect (56 ms in RT and 0.077 in PE). This result demonstrates that reversal of the Simon e€ect occurs when the location information on both trial types is symbolic, as well as when it is physical or verbal. The fact that the magnitude of the reverse Simon e€ect for arrows was larger than that for words and similar to that for physical locations obtained in Experiment 1 is not surprising. Arrows produce S-R compatibility e€ects similar to those obtained with physical positions (e.g., Eimer, 1995; Lu & Proctor, in press; Wang & Proctor, 1996), and are relatively more compatible with keypress responses than are location words. Mixing stimulus modes, such that the mode by which the location information was conveyed di€ered when it was relevant versus irrelevant, made the performance of the individual tasks easier overall, as re¯ected in faster and more accurate responding. The advantage for mixed mode presentations of the location words was not as large in this experiment, in which the alternative mode was arrow direction, as in Experiment 1, in which it was physical location. This outcome implies that part of the diculty for the WordI/WordR condition arises from having to discriminate locationrelevant and location-irrelevant stimuli presented at the same location. The fact that responses to word stimuli are slower and less accurate in the WordI/WordR condition than in the ArrowI/WordR and WordI/ArrowR conditions indicates that having both types of centered stimuli be location words adds an additional diculty to the task. The two mixed mode conditions showed positive Simon e€ects: ArrowI/WordR condition, 33 ms in RT and 0.023 in PE; WordI/ArrowR condition, 12 ms in RT and 0.047 in PE. The e€ects for the two conditions are of similar magnitude, as indicated by the fact that the Simon e€ect was signi®cant in the RT data and not the PE data for the ArrowI/WordR condition, but signi®cant for the PE data and not the RT data for the WordI/ArrowR condition. Thus, the e€ect of the incompatible location-relevant mapping was reduced if not eliminated when the location information for the two tasks was in di€erent modes, even though the stimuli were presented in the same centered location on the display screen. This outcome supports the view that the short-term associations between location information and responses de®ned for the location-relevant task are mode speci®c.

19

Overall RT was slower on location-relevant trials, and the Simon e€ect was smaller on location-irrelevant trials, when the stimulus mode was verbal rather than symbolic direction. These outcomes are consistent with the fact that location words are less strongly associated with keypress responses than are arrow stimuli (Wang & Proctor, 1996) and tend to produce only small Simon e€ects when their meaning is irrelevant to the task (e.g., Lu & Proctor, 1995; in press).

Experiment 3 The results from Experiments 1 and 2 conform with the view that a distinction between the modes by which the location information is conveyed for the location-irrelevant and location-relevant trials is crucial to the lack of a reversal of the Simon e€ect in the mixed mode conditions. That is, the conditions under which no reversal was evident were those in which location words were intermixed with either physical location stimuli or arrow stimuli. This outcome implies that associations between verbal location information and responses and between nonverbal location information and responses are relatively distinct. This is in agreement with many accounts of stimulus mode e€ects that propose a distinction in processing for verbal and nonverbal stimuli (e.g., Lu & Proctor, in press; Virzi & Egeth, 1985; Wickens, 1984). An implication of such accounts is that it is necessary for one stimulus mode to be verbal and the other nonverbal if the e€ects of an incompatible location-relevant mapping are to be minimized. Experiment 3 tested this implication by mixing two nonverbal stimulus modes, arrow direction and physical location. If a distinction between verbal and nonverbal modes is necessary to eliminate the reverse Simon e€ect, then the e€ect should be present in Experiment 3 when the location-relevant information is arrow direction and the location-irrelevant information is physical location, and vice versa. However, if only a di€erence in stimulus modes is necessary, and not the verbal-nonverbal distinction, then the reverse Simon e€ect should not be obtained. Method Sixty-four undergraduates enrolled in Introductory Psychology at Purdue University participated in the experiment. None had participated in past experiments. Thirty-two subjects were randomly assigned to each of two conditions. The procedure was similar to those of the mixed mode conditions of Experiments 1 and 2. In the ArrowI/PositionR condition, the location-irrelevant stimuli were red and green, left- and right-pointing arrows presented in a centered location and the location-relevant stimuli were white circles presented in left or right locations. In the PositionI/ArrowR condition, the location-irrelevant stimuli were red and green circles presented in left and right locations and the location-relevant stimuli were centered, leftand right-pointing white arrows. As in Experiments 1 and 2, the S-R mapping for the location-relevant stimuli was spatially incompatible.

Table 4 Mean RT and PE (in parentheses) for location-irrelevant stimuli in Experiment 3 as a function of condition (ArrowI/PositionR, PositionI/ArrowR), stimulus location, and response location Stimulus location

Response location Left

Right

Left Right

ArrowI/PositionR 560 (0.044) 560 (0.039)

559 (0.067) 570 (0.041)

Left Right

PositionI/ArrowR 635 (0.051) 640 (0.037)

641 (0.058) 641 (0.044)

Results Location-irrelevant trials For both RT and PE, a 2 (Stimulus Location: left or right) ´ 2 (Response Location: left or right) ´ 2 (Condition: ArrowI/PositionR or PositionI/ArrowR) ANOVA was conducted. Stimulus and response locations were within-subjects factors, and condition was the betweensubjects factor. As shown in Table 4, there was a main e€ect for type of location-irrelevant stimuli, with responses to the location-irrelevant arrows being signi®cantly faster than responses to the location-irrelevant circles (Ms ˆ 562 ms versus 639 ms, respectively), F(1, 62) ˆ 6.43, p ˆ 0.014, MSE ˆ 58,819. The interaction of Response Location ´ Stimulus Location was not signi®cant, F(1, 62) < 1.0, nor was that of Condition ´ Stimulus Location ´ Response Location, F(1, 62) < 1.0. Thus, neither condition showed a signi®cant Simon e€ect, Fs < 1.0, with the e€ect being )5 ms for the ArrowI/ PositionR condition and +2 ms for the PositionI/ArrowR condition. The PE data showed no signi®cant e€ects. Of most relevance, the Simon e€ect was not signi®cant for either the ArrowI/PositionR condition (MD ˆ +0.011) or the PositionI/ArrowR condition (MD ˆ 0.000), Fs < 1.0. Location-relevant trials A 2-way ANOVA was conducted with response location as a within-subject factor and condition as a betweensubject factor. For the location-relevant trials (see Table 2), responses tended to be faster in the PositionI/ ArrowR condition (M ˆ 549 ms) than in the ArrowI/ PositionR condition (M ˆ 558 ms). However, this difference was well within the range of chance, F(1, 62) < 1.0, and no other e€ects were signi®cant for the RT or PE data. Discussion Neither the PositionI/ArrowR condition nor the ArrowI/ PositionR condition showed a signi®cant reverse Simon

20

e€ect on the location-irrelevant trials. When physical stimulus location was irrelevant, the impact of the incompatible mapping for arrows was much less than that for physical locations in Experiment 1 (2 ms versus 64 ms). Similarly, when arrow direction was irrelevant, the impact of the incompatible mapping of physical locations was much less than that for arrows in Experiment 2 (5 ms versus 56 ms). Like the PositionI/WordR condition in Experiment 1, mixing arrows and locations apparently did not entirely eliminate the e€ect of the irrelevant mapping of location information, however, since the mixed mode conditions did not show a positive Simon e€ect. In summary, the results of Experiment 3 indicate that the impact of the incompatible mapping for location information on location-irrelevant trials is reduced substantially when a mode distinction exists. Because the two modes in the present experiment, spatial and symbolic, were both nonverbal, it appears that the verbal/ nonverbal distinction is relatively unimportant. The short-term associations between relevant location information and responses seem to be speci®c to the mode of presentation even when neither mode is verbal.

Experiment 4 In the previous experiment, as well as in the others, the two types of stimuli used for the mixed mode conditions di€ered in form as well as color, whereas those for the same mode conditions di€ered in color only. Thus, the possibility remains that rather than a di€erence in stimulus mode being the crucial factor that minimizes the intrusion of the location-relevant mapping into performance on location-irrelevant trials, it is simply the greater dissimilarity between the stimuli for the two tasks. This latter possibility was evaluated in Experiment 4. In this experiment, the location information was conveyed in the same stimulus mode (spatial, verbal, or symbolic) for both location-relevant and location-irrelevant trials. However, the stimuli for the two types of trials were distinguished not only by color but also by a global form property. For physical locations, two easily discriminated geometric shapes (circles and squares) were used to indicate whether location was relevant or irrelevant, along with the white versus color distinction. For arrows, in addition to color, arrow size distinguished location-relevant from location-irrelevant trials. For words, upper- and lower-case letters were used to indicate whether location was relevant or irrelevant. If a discrimination based on stimulus dissimilarity is sucient to eliminate the e€ect of the incompatible mapping on the location-irrelevant trials, then we would expect the same results as when stimulus modes are mixed, that is, little evidence of a reverse Simon e€ect. However, if reduction or elimination of the reverse Simon e€ect requires a distinction between stimulus modes, then the reverse Simon e€ect should be seen.

Method Forty-eight undergraduates from the same pool as the previous experiments participated. None had participated in the previous experiments. Sixteen subjects were assigned to each of the three conditions. The apparatus was identical to the previous experiments. One of two forms, physically dissimilar but of the same stimulus mode, was assigned to the location-relevant task and the other to the location-irrelevant task, counterbalanced across subjects. In the PositionI/PositionR condition, stimuli were circles and squares, approximately 10 mm in diameter (1.04° viewing angle), presented in left and right locations. In the ArrowI/ArrowR condition, stimuli were large and small left-and right-pointing arrows. The arrows were ®lled symbols ( ) presented in the center of the computer screen. The small arrows were 1.75 cm (1.82°) long, with the width of the arrowhead being 1.43 cm (1.49°) and the width of the tail being 0.80 cm (0.83°) and the large arrows were 2.54 cm (2.64°) long, with the width of the arrowhead being 2.86 cm (2.98°) and the width of the tail being 1.27 cm (1.32°)], presented in the center of the computer screen. The words were displayed in upper or lowercase, in the same font and size as the previous experiments.

Results Location-irrelevant trials For both RT and PE, a 2 (Stimulus Location: left or right) ´ 2 (Response Location: left or right) ´ 3 (Stimulus Mode: physical locations, arrows, or words) ANOVA was conducted. Stimulus and response locations were within-subjects factors, and stimulus mode was a between-subjects factor. Mean RTs for the three stimulus modes are shown in Table 5. The main e€ect of mode was signi®cant, F(2, 45) ˆ 6.05, p ˆ 0.005, MSE ˆ 29,739, with responses being fastest to arrows (M ˆ 626 ms), followed by physical locations (M ˆ 672 ms) and words (M ˆ 731 ms). There was a signi®cant main e€ect for response location, F(1, 45) ˆ 4.07, p ˆ 0.050, MSE ˆ 1,801, with right responses (M ˆ 670 ms) being faster Table 5 Mean RT and PE (in parentheses) for location-irrelevant stimuli in Experiment 4 as a function of condition (PositionI/ PositionR, ArrowI/ArrowR, WordI/WordR), stimulus location, and response location Stimulus location

Response location Left

Right

Left Right

PositionI/PositionR 702 (0.064) 637 (0.008)

643 (0.017) 707 (0.053)

Left Right

ArrowI/ArrowR 652 (0.066) 609 (0.021)

599 (0.016) 643 (0.064)

Left Right

WordI/WordR 761 (0.046) 734 (0.019)

712 (0.026) 718 (0.042)

The PositionI/PositionR, ArrowI/ArrowR, and WordI/WordR conditions include a form distinction that was not present in the corresponding conditions of Experiments 1±3

21

than the left responses (M ˆ 683 ms). In addition, response location interacted with mode, F(1, 45) ˆ 3.22, p ˆ 0.049, MSE ˆ 1,801, with the right response faster than the left response for arrows (MD ˆ 10 ms) and words (MD ˆ 32 ms), but slower for shapes (MD ˆ )5 ms). The interaction of Stimulus Location ´ Response Location was signi®cant, F(1, 45) ˆ 37.56, p < 0.001, MSE ˆ 2190, and the three-way interaction of these variables with stimulus mode was also signi®cant, F(1, 45) ˆ 4.29, p ˆ 0.020, MSE ˆ 2190. RTs were faster when stimulus and response locations did not correspond (M ˆ 656 ms) than when they did (M ˆ 697 ms), that is, they showed a reverse Simon e€ect. The reverse Simon e€ect was largest for the physical location stimuli (65 ms), F(1, 15) ˆ 53.74, p < 0.001, MSE ˆ 1243, followed by the arrow stimuli (44 ms), F(1, 15) ˆ 22.75, p < 0.001, MSE ˆ 1315 and word stimuli (17 ms), F(1, 15) ˆ 1.06, p ˆ 0.31 MSE ˆ 4009. Analysis of the PE data supported the ®ndings of the RT data. The Stimulus Location ´ Response Location interaction was signi®cant, F(1, 45) ˆ 65.16, p < 0.001, MSE ˆ 0.0011, indicating less accurate responses when stimulus and response locations corresponded (PE ˆ 0.056) than when they did not (PE ˆ 0.018). Similar to the RT data, this interaction was quali®ed by a marginally signi®cant three-way interaction of these variables with stimulus mode, F(1, 45) ˆ 3.01, p ˆ 0.059, MSE ˆ 0.0011. The reverse Simon e€ect was signi®cant for physical locations, arrows, and words, Fs(1, 15) ˆ 46.10, 22.75, and 5.56, ps < 0.05, MSEs ˆ 0.0084, 0.0013, and 0.0013, respectively, indicating that responses opposite the stimulus location were signi®cantly more accurate than responses on the same side as the stimulus location. Location-relevant trials A 2-way ANOVA was conducted with response location as a within-subject factor and stimulus mode as a between-subject factor. The RT and PE data showed only signi®cant main e€ects of mode, Fs(1, 45) ˆ 24.03 and 9.07, ps < 0.001, MSEs ˆ 18,137 and 0.0004. Responses were fastest and most accurate with arrow stimuli (M ˆ 609 ms, 0.020) followed by physical location stimuli (M ˆ 632 ms, 0.021) and word stimuli (M ˆ 823 ms, 0.038). Discussion There was a reverse Simon e€ect of 65 ms (0.046) for the physical location stimuli, 44 ms (0.046) for the arrow stimuli, and 17 ms (0.022) for the word stimuli. Although the reverse Simon e€ect was not signi®cant in the RT data for the word stimuli, it was in the PE data. Moreover, when the RT data for the WordI/WordR condition in Experiment 4 were compared to those of

Experiments 1 and 2, the reverse Simon e€ect was signi®cant, F(1, 45) ˆ 6.39, p ˆ 0.015, MSE ˆ 5904, and did not interact with experiment, F(2, 45) ˆ 0.45, p ˆ 0.64, MSE ˆ 5904. Thus, for physical locations, arrows, and words, a form distinction between the stimuli used in the location-relevant and location-irrelevant tasks was not sucient to eliminate the reverse Simon e€ect on location-irrelevant trials. This outcome provides support for the hypothesis that a shift in the mode of the stimulus is necessary to eliminate the reverse Simon e€ect when location-irrelevant trials are intermixed with incompatibly mapped location-relevant trials. Together, the results of Experiments 1±4 imply that the short-term associations between location information and responses are mode speci®c. When the location information is in the same mode on location-relevant and location-irrelevant trials, the location-relevant mapping a€ects performance on the location-irrelevant trials. When the location information is in di€erent modes on the two trial types, the e€ect of the locationrelevant mapping on the location-relevant trials is reduced substantially.

General discussion Intermixing location-irrelevant trials with incompatible location-relevant trials produced a reverse Simon e€ect when the same stimulus mode was used for both trial types. Experiment 1 replicated the reversal previously reported by Marble and Proctor (in press) for the condition in which both location-relevant and locationirrelevant stimuli occurred in left and right positions. Experiment 1 extended this ®nding to the centered words ``left'' and ``right,'' and Experiment 2 extended it to left- and right-pointing arrows. Thus, within a stimulus mode, the incompatible mapping of the location information to response locations a€ected performance on those trials for which stimulus location was irrelevant to determining the correct response. In all cases, there was a tendency to make the response that was opposite that of the location information conveyed by the stimulus. Although all of the stimulus modes showed a reverse Simon e€ect, the e€ect size di€ered in magnitude across the modes. The largest e€ect was for physical positions (M ˆ 65 ms), next largest for arrow directions (M ˆ 50 ms), and least for words (M ˆ 28 ms). This ordering matches that of the relative compatibility of these stimulus modes with keypress responses (Wang & Proctor, 1996). The implication is that the degree to which the incompatibly mapped response is activated on location irrelevant trials is a direct function of the compatibility between the stimulus and response sets. Even though the reverse Simon e€ect for words was signi®cant when collapsed across Experiments 1, 2, and 4, it was not in Experiments 2 and 4 alone. One reason why the e€ects did not attain the 0.05 level in those experiments was that the error variability for the WordI/

22

WordR condition was much larger than that for the ArrowI/ArrowR and PositionI/PositionR conditions. Examination of Simon e€ect sizes for the WordI/WordR condition of individual subjects in the three experiments revealed a bimodal distribution. Of the 48 subjects, only 8 had e€ect sizes within ‹20 ms, indicating that most subjects showed either a large positive (16 subjects) or negative e€ect (24 subjects). This bimodal distribution suggests that a signi®cant portion of the subjects were activating the corresponding response for locationirrelevant words even though the location-relevant mapping for those words was to the noncorresponding responses. The subjects who showed the reverse Simon e€ect were likely directly activating the noncorresponding response prior to determination of which trial type, location relevant or irrelevant, was being signaled. The subjects who showed the positive Simon e€ect were likely adopting a strategy of naming the word and performing a transformation to the noncorresponding response only if it was determined that the trial was location relevant. This latter strategy would involve an extra step on location-relevant trials relative to the former strategy, which implies that RTs on location-relevant trials should be slower for those subjects showing a positive Simon e€ect than for those showing a negative e€ect. An analysis of the location-relevant RTs for the WordI/WordR condition pooled across Experiments 1, 2, and 4 showed this to be the case: Mean RT for the subjects showing a positive Simon e€ect (M ˆ 852 ms) was slower than that for subjects showing a negative Simon e€ect (M ˆ 780 ms), F(1, 38) ˆ 4.88, p ˆ 0.033, MSE ˆ 20,776. When the tasks used stimuli of di€erent modes, the reverse Simon e€ect was eliminated. A reverse Simon e€ect was not evident in Experiment 1 for the conditions in which physical location stimuli were intermixed with location words, regardless of which stimulus mode was used for location-relevant trials and which for locationirrelevant trials. Likewise, the reversal also was not evident when centered words and arrows were mixed in Experiment 2, indicating that it is not necessary that the relevant and irrelevant location information for the di€erent modes be presented in distinct spatial locations. The verbal/nonverbal distinction is also not necessary for elimination of the reversal. In Experiment 3, there was no signi®cant reverse Simon e€ect when arrows were mixed with physical locations, again regardless of which stimulus mode conveyed the relevant location information and which the irrelevant. Dissimilarity of the forms for the location-relevant and location-irrelevant trials did not eliminate the reverse Simon e€ect in Experiment 4 when the location information was conveyed by the same mode on both trial types for physical location, arrow direction, or location word. Apparently, then, the reverse Simon e€ect occurs when the location information is in the same mode on both types of trials, and this reversal is eliminated when the stimulus modes di€er for the location-relevant and location-irrelevant trials.

A di€erence in stimulus mode seems to provide a coding distinction that enables the short-term S-R associations de®ned for the two tasks to be kept relatively separate. Responses were faster and more accurate for both the location-relevant and location-irrelevant tasks when there was a mode distinction between the stimuli for the respective tasks. The implication that stimulus mode provides a coding feature is in agreement with network models of the Stroop color-naming e€ect that postulate distinct pathways for the word and nonword attributes of stimuli (e.g., Cohen et al., 1990). However, it is in con¯ict with those models that postulate convergence of the pathways on an intermediate conceptual module (e.g., Zhang & Kornblum, 1998). In other words, the short-term associations between the relevant location information and the noncorresponding keypress responses are restricted primarily to codes speci®c to the mode in which that information is presented. Consequently, when a colored stimulus occurs for which the irrelevant location information is in a di€erent mode, that stimulus produces little activation of the noncorresponding response. Although the reverse Simon e€ect was eliminated in all mixed mode conditions examined in Experiments 1± 3, a positive Simon e€ect of normal magnitude was not always present. For the conditions in which the location information on relevant and irrelevant trials was verbal for one and nonverbal for the other, a positive Simon e€ect was found in three of the four cases: When (1) the irrelevant stimulus information was location word and the relevant information physical position, (2) the irrelevant information was arrow direction and the relevant information location word, and (3) the irrelevant information was location word and the relevant information arrow direction. These results indicate that when the irrelevant location information is conveyed by location word, a positive Simon e€ect of approximately normal magnitude occurs regardless of whether the location information on location-relevant trials is conveyed by physical position or arrow direction. This outcome suggests that the location word produces no activation of the incompatible response by way of the short-term associations de®ned for the location-relevant task. That is, the activation is restricted to the long-term associations between the location words and their corresponding responses. Similarly, the positive Simon e€ect for irrelevant arrow direction when the information on location-relevant trials is location word implies that the arrow stimulus produces no activation by way of the short-term associations of the words to the noncorresponding responses. The one case in which the verbal-nonverbal distinction did not lead to a complete reinstatement of the positive Simon e€ect was when the irrelevant stimulus information was physical location and the relevant stimulus information location word. The fact that a normal Simon e€ect is not obtained when the irrelevant stimulus dimension is physical location and the relevant location information is location word can be explained

23

by assuming that some verbal coding occurs on the location-irrelevant trials. This coding would produce some activation of the noncorresponding response via the short-term word-response associations that would o€set any activation of the corresponding response produced via long-term position-response associations. Neither condition for which the location mode distinction was nonverbal produced a statistically reliable Simon e€ect, either positive or negative. This was the case when (1) the irrelevant information was arrow direction and the relevant information physical position, and (2) the irrelevant information was physical location and the relevant information arrow direction. The absence of a normal Simon e€ect for these conditions is open to alternative interpretations. One possibility is that there is a tendency for some verbal coding when arrow direction and physical position are mixed, regardless of which mode is location relevant and which location irrelevant. A second, and more likely, possibility is that because both arrow direction and physical position are visuo-spatial in nature, the stimuli of each type produce some activation by way of the incompatible associations de®ned for the other dimension, without verbal coding playing any role. In summary, when the mode in which irrelevantlocation information is presented is the same as that for incompatibly mapped relevant-location information on intermixed trials, the incompatible response is activated through the short-term associations de®ned for the location-relevant task. This activation produces a reverse Simon e€ect. These associations are apparently not between an abstract location code and the response because the reverse Simon e€ect is not evident when the stimulus modes di€er on location-relevant and locationirrelevant trials. This implies that models such as that of Cohen et al. (1990), which have distinct associative pathways between di€erent modes of the same abstract information and the responses, are basically correct. However, the distinction is not simply one of verbal and nonverbal pathways, although pathways for di€erent nonverbal modes may not be as strongly di€erentiated as those between verbal and nonverbal modes. Acknowledgements We would like to thank Carlo UmiltaÁ, Marco Zorzi, and an anonymous reviewer for helpful comments on a previous version of the paper. We would also like to thank Amanda Foster, Ben Neureiter, Marland V. Pittman, Brian Turner, Jim Whanger, Sarah Wilcox, and Marc Winterbottom for assistance in collecting the data for Experiments 2 through 4.

References Barber, P., & O'Leary, M. (1997). The relevance of salience: towards an activation account of irrelevant stimulus-response compatibility e€ects. In B. Hommel & W. Prinz (Eds.), Theoretical issues in stimulus-response compatibility (pp. 135±172). Amsterdam: North-Holland. Cohen, J. D., Dunbar, K., & McClelland, J. L. (1990). On the control of automatic processes: a parallel distributed processing account of the Stroop e€ect. Psychological Review, 97, 332±361.

Dutta, A., & Proctor, R. W. (1992). Persistence of stimulus-response compatibility e€ects with extended practice. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 801±809. Eimer, M. (1995). S-R compatibility and automatic response activation: evidence from psychophysiological studies. Journal of Experimental Psychology: Human Perception and Performance, 21, 837±854. Hedge, A., & Marsh, N. W. A. (1975). The e€ect of irrelevant spatial correspondences on two-choice response time. Acta Psychologica, 39, 427±439. Hommel, B. (1995). Stimulus-response compatibility and the Simon e€ect: toward an empirical clari®cation. Journal of Experimental Psychology: Human Perception and Performance, 21, 764±775. Hommel, B., & Prinz, W. (Eds.) (1997). Theoretical issues in stimulus-response compatibility. Amsterdam: North-Holland. Kornblum, S. (1992). Dimensional overlap and dimensional relevance in stimulus-response and stimulus-stimulus compatibility. In G. E. Stelmach & J. Requin (Eds.), Tutorials in motor behavior, II (pp.743±777). Amsterdam: North-Holland. Lu, C.-H. (1997). Correspondence e€ects for irrelevant information in choice-reaction tasks: characterizing the stimulus-response relations and processing dynamics. In B. Hommel & W. Prinz (Eds.), Theoretical issues in stimulus-response compatibility (pp. 85±117). Amsterdam: North-Holland. Lu, C.-H., & Proctor, R. W. (1994). Processing of an irrelevant location dimension as a function of the relevant stimulus dimension. Journal of Experimental Psychology: Human Perception and Performance, 20, 286±298. Lu, C.-H., & Proctor, R. W. (1995). The in¯uence of irrelevant location information on performance: a review of the Simon e€ect and spatial Stroop e€ects. Psychonomic Bulletin & Review, 2, 174±207. Lu, C.-H., & Proctor, R. W. (in press). In¯uence of irrelevant information on human performance: e€ects of S-R association strength and relative timing. Quarterly Journal of Experimental Psychology. Marble J. G., & Proctor, R. W. (in press). Mixing location-relevant and location-irrelevant choice-reaction tasks: in¯uences of location mapping on the Simon e€ect. Journal of Experimental Psychology: Human Perception and Performance. Proctor, R. W., & Lu, C.-H. (1999). Processing irrelevant location information: practice and transfer e€ects in choice-reaction tasks. Memory & Cognition, 27, 63±77. Proctor, R. W., & Wang, H. (1997). Di€erentiating types of setlevel compatibility. In B. Hommel & W. Prinz (Eds.), Theory of stimulus-response compatibility (pp. 11±37). Amsterdam: NorthHolland. Sha€er, L. H. (1965). Choice reaction with variable S-R mapping. Journal of Experimental Psychology, 70, 284±288. Simon, R. J. (1990). The e€ects of an irrelevant directional cue on human information processing. In R. W. Proctor & T. G. Reeve (Eds.), Stimulus-response compatibility: An integrated perspective (pp. 31±86). Amsterdam: North-Holland. Sto€er, T. H., & UmiltaÁ, C. (1997). Spatial stimulus coding and the focus of attention in S-R compatibility and the Simon e€ect. In B. Hommel & W. Prinz (Eds.), Theoretical issues in stimulusresponse compatibility (pp. 181±208). Amsterdam: North-Holland. Tagliabue, M., Zorzi, M., UmiltaÁ, C., & Bassignani, F. (2000). The role of LTM links and STM links in the Simon e€ect. Journal of Experimental Psychology: Human Perception and Performance, 26, 648±670. UmiltaÁ, C., & Nicoletti, R. (1990). Spatial stimulus-response compatibility. In R. W. Proctor & T. G. Reeve (Eds.), Stimulusresponse compatibility: an integrated perspective (pp. 89±116). Amsterdam: North-Holland. Virzi, R. A., & Egeth, H. E. (1985). Toward a translational model of Stroop interference. Memory & Cognition, 13, 304±319. Wang, H., & Proctor, R. W. (1996). Stimulus-response compatibility as a function of stimulus code and response modality.

24 Journal of Experimental Psychology: Human Perception and Performance, 22, 1201±1217. Wickens, C. D. (1984). Processing resources in attention. In R. Parsuraman & D. R. Davies (Eds.), Varieties of attention (pp. 63±102). New York: Academic Press. Zhang, H., & Kornblum, S. (1998). The e€ects of stimulus-response mapping and irrelevant stimulus-response and stimulus-stimu-

lus overlap in four-choice Stroop tasks with single-carrier stimuli. Journal of Experimental Psychology: Human Perception and Performance, 24, 3±19. Zorzi, M. & UmiltaÁ, C. (1995). A computational model of the Simon e€ect. Psychological Research, 58, 193±205.