[page 104↓]

5  One-trial-Simon Experiments

An even stronger case for a direct influence of response instructions on response coding could be made if it were possible to show that the impact of an irrelevant spatial stimulus attribute (i.e., the Simon effect) depends on response instructions. That is, if non-spatial response instructions reduced the spatial Simon effect, for the following reason.

Remember that the widely accepted dual route models (cf. Chapter 2.2) assume a direct link between overlapping stimulus and response codes. The primary motivation for dual route models to propose a direct route over and above a (controlled or conditionally automatic) translation route has been to parsimoniously account for an impact of irrelevant stimulus attributes. For example, in the standard Simon task in which the relevant S‑R mapping is arbitrary, stimulus position is uninformative with respect to the required responses, and the likelihood of accidental co-translation due to logical recoding is low.

Strong behavioral evidence for the direct response activation account (and against translation accounts of the Simon effect; e.g., Wallace, 1971) comes from studies in which irrelevant spatial information is provided by go/no-go signals that are presented some time after the imperative stimulus. Hence, irrelevant spatial information is presented when imperative stimulus identification and translation can be assumed to be completed on most trials. For example, in experiments conducted by Hommel (1995, 1996c; also see Shiu & Kornblum, 1999) participants were required to press a left or a right key as indicated by spatial precues (i.e., arrows pointing to the left or right; direct mapping), but had to withhold their responses until they saw a go vs. no-go signal (i.e., a green or a red color patch; Hommel, 1995, Exp. 1), or until a go-signal was presented (i.e., a green patch; Hommel, 1996c, Exp. 1) that randomly appeared on the left or the right of the screen after varying intervals. Hommel observed a substantial Simon effect (i.e., a correspondence effect between response location and go-signal location) even when go-signals were presented on 100% of the trials. That is, when subjects presumably had a high motivation to prepare their responses in advance. Hommel concluded that irrelevant response-overlapping stimulus information automatically activates corresponding response (action) codes, thereby influencing (spatial) responding as long as the response has not been executed.

As outlined in Chapter 3, the findings regarding the influence of response instructions on the Simon effect can be considered inconclusive. On the one hand, Hommel (1993a) could [page 105↓]show that variations of spatial response instructions determine the direction of the Simon effect. However, the Hommel experiment manipulated instructions within the spatial dimension. Simon et al. (1976), on the other hand, found a substantial spatial Simon effect when keys were instructed in terms of color and color-(key)-labels changed from trial to trial. However, the latter study did not include a (baseline) condition with spatial response instructions, leaving open the question of whether non-spatial response instructions modulate the Simon effect.

Experiment 4 and 5 extend these findings by using a variant of the Simon task similar to that of the Hommel (1995; 1996c) experiments, and by employing a response-instruction logic similar to that of Experiment 3 (see Chapter 4.3). As in the Hommel (1995) experiment, irrelevant spatial information in Experiments 4 and 5 was provided by go/no-go signals. Go/no-go signals consisted of vertical and horizontal bars presented at different locations, the orientation of which indicated whether the prepared response was to be executed or not. Go/no-go signals followed the imperative stimuli after a considerable delay. Letters served as imperative stimuli. New letters (i.e., letter pairs), and hence, new S‑R pairings, were instructed on successive trials in order to avoid the usual confound between effects of instructions and practice.

Experiments 4 and 5 differ regarding response instructions while stimuli and stimulus presentation remained constant across experiments. That is, in Experiment 4, keys were instructed in terms of location. In contrast, the same keys were instructed in terms of color in Experiment 5. Experiment 4 thus establishes a baseline for a “standard” Simon effect under conditions of changing imperative stimuli. Experiment 5, on the other hand, assesses whether non-spatial response instructions affect the size of the Simon effect.

Both the spatial coding hypothesis and the weak direct coding hypothesis again predict normal and comparable spatial Simon effects in both experiments, either because instructions do not affect response coding (spatial coding hypothesis), or because non-spatial coding does not affect implicit spatial coding (weak direct coding hypothesis). Alternatively, if color coding dominates spatial coding under non-spatial response instructions (i.e., if spatial codes are less strongly weighed and hence contribute less to responding), as assumed by the strong direct coding hypothesis, the (spatial) Simon effect should be strongly reduced under non-spatial as compared to spatial response instructions.


[page 106↓]

5.1  Experiment 4

The main goal of Experiment 4 was to secure that a standard Simon effect shows up when S‑R rules, and hence imperative stimuli, change from trial to trial. Responses were instructed in terms of key location, hence providing a baseline for the Simon effect under non-spatial instructions (Experiment 5). The task was similar to the one used by Hommel (1995, 1996c) in that irrelevant spatial information was provided by go/no-go signals that followed the imperative stimuli after a considerable delay. Unlike Hommel, relevant S‑R mappings were arbitrary instead of compatible. That is, letter stimuli (instead of arrows) were assigned to left and right keys before each trial, thus, according to the dimensional overlap model taxonomy (cf. Kornblum & Lee, 1995), changing the Hommel task from a spatial Stroop task (type 8 ensemble) to a regular Simon task (type 3 ensemble). If the Simon effect obtained by Hommel (also see Shiu & Kornblum, 1999) was indeed due to an overlap between irrelevant stimulus position and response location, a Simon effect should also be observed with arbitrary relevant S‑R mappings.

A further modification of the Hommel (1995, 1996c) task concerns the inclusion of a neutral condition. The neutral (irrelevant position) condition was realized by presenting go/no-go signals at the central screen position. It was included to assess whether potential compatibility and instruction effects obtained with this paradigm are primarily due to interference or to facilitation.

The choice of a delayed-position paradigm with changing S‑R mappings was motivated by several considerations. First, imperative stimuli changed from trial to trial in order to assess the impact of response instructions without any confounding influence of practice. This should not affect direct response activation per se because the direct route is considered independent of whether new imperative stimuli have to be translated on each trial, or whether well-practiced S‑R rules are retrieved. However, with new instructions on every trial, S‑R translation can be assumed to be quite time consuming. This could have been problematic with a standard Simon task (i.e., a task in which imperative stimuli randomly appear at different screen positions) because it has been shown that irrelevant spatial code activation is transient and relatively short-lived. For example, Hommel (1993b) demonstrated that the Simon effect decreases as the difficulty of identifying the relevant stimulus, and hence RT, increases. Therefore, giving the imperative stimulus a head start in my experiments should enhance the likelihood to observe effects of spatial code activation.


[page 107↓]

Moreover, I employed a high percentage of go-trials and a large interval between imperative stimuli and lateralized go/no-go signals. This was done to maximize the likelihood that participants complete stimulus identification and S‑R translation before the go-signal (i.e., irrelevant position information) appears. If any impact of instructions shows with this paradigm (see Exp. 5), they are likely due to influences on response coding, rather than translation.

5.1.1 Method

Participants

Overall, thirty-one subjects, students and non-students from Berlin (17 female, 14 male, mean age = 24 years), received € 7,- for participation (for exclusion and substitution of participants, see Section Data Analysis, below).

Stimulus Material and Counterbalancing

For the experimental trials, forty-eight letter pairs were constructed by pairing twenty-four letters as follows. Twelve letter pairs were generated for each of the four experimental blocks (see next section) according to several criteria. Letters in a letter pair were required to be at least 6 letters apart in the alphabet (mean distance = 11.2 letters) to avoid associative and/or ordinal (location) priming between neighboring letters. Each letter was paired with three other letters, such that, on each block, it only appeared in one letter pair.

For each block, letters in a letter pair were assigned to left or right responses such that, for half of the letter pairs, left responses were assigned to the letter that occurs “left” in the alphabet (e.g., A: left key; N: right key), whereas right responses to the “left” letter were required on the other half of the letter pairs (e.g., R: left key; C: right key). When re-pairing letters to construct pairs for successive blocks, half of the letters were assigned the same response location as in the previous block, while for the other half the assignment was changed. An attempt was made to ensure that (a) changes of assignment affected ordinally corresponding S‑R assignments (i.e., A: left key; N: right key) approximately as often as non-corresponding assignments, and that (b) response assignment for a particular letter changed at least once across blocks. The latter worked out for 22 out of the 24 letters.

For each of the twelve letter pairs per block, all six go-signal-position (left, middle, right) x response location (left vs. right) conditions were generated. These six go-signal-[page 108↓]position x response location conditions were distributed across three lists by Latin square such that each letter pair appeared only twice within a single list and block. The two repetitions of each letter pair were such that (a) each letter of a given pair was presented only once (i.e., each letter pair appeared in one left-response and one right-response condition), and (b) both instantiations of a specific letter pair realized different compatibility conditions (compatible, neutral, incompatible). The Latin squares used for counterbalancing guaranteed that (a) different combinations of compatibility conditions were determined for different letter pairs on a given list and block, and (b) the same number of trials per stimulus (go-signal) position x response location condition appeared in each block and list. This counterbalancing scheme led to 24 trials per block and list (i.e., 4 of each stimulus position x response location condition in each block, 16 across blocks), whereby, within a given list, different conditions were mostly realized by different letter pairs. Across lists, however, each letter pair appeared in all conditions.

Of the 24 trials within a given list and block, three trials were chosen to be no-go trials (i.e., imperative stimuli were accompanied by horizontal instead of vertical bars at the pre-determined position), amounting to twelve (12.5%) no-go trials per list. The three no-go trials per block and list were determined such that, across blocks, the same number of trials in each stimulus position x response location served as no-go trials (i.e., two trials per condition) and that, within a given block and list, (a) different letter pairs were used for no-go trials, and (b) the excluded responses varied in type (i.e., either two left and one right responses, or one left and two right responses were excluded, never three left or three right responses). Determining no-go trials this way left 14 go-trials (87.5%) per stimulus position x response location condition within each list, that is, 28 go-trials per compatibility condition (i.e., compatible, neutral, incompatible).

One may argue that this is a relatively small number of trials per compatibility condition. The choice to go with such a small number of trials per conditions was based on two considerations. First, there appeared to be a trade-off between reliability of estimates (enhanced with increasing number of trials) and the possibility of carry-over effects if the same letters were presented too often (albeit in different letter pairings). Second, psycholinguistic studies that employ similar counterbalancing schemes as the one used here often only present between 10 and 20 instances per condition as well. Nevertheless, they tend to attain reliable results. However, in order to back up their main results, they often carry out item-analyses [page 109↓](with items instead of, or in addition to, participants as a random factor; e.g., Clark, 1973) to ensure that the restricted sample of selected items represents the larger population of items (in my case letter pairs) sampled from. Therefore, I decided on fewer letter repetitions, but to carry out corresponding item analyses for the present experiments (see Section Data Analysis for details).

From the two letters that were not used in constructing experimental trials (i.e., the letters Z and I), three trials were constructed (one for each compatibility condition, with one trial being a no-go trial) that served as practice trials in the experiment and were not counterbalanced across lists.

Apparatus and Procedure

The experiment was run on Pentium II computers that were connected to two separate keys via an ExKey Logic provided by BeriSoft. Stimulus presentation and response recording was controlled by the Experimental Runtime System software (ERTS©; Beringer, 2000). The ERTS software runs in DOS mode on IBM compatible computers.

Letters appeared in white against black background at the center of the screen. Letter font was the NRC7BIT large-scale bitmap font. The height of the letter stimuli on the screen was 1.2 cm, whereas letter width ranged between 0.6 cm and 1.1 cm, depending on letter identity. The diameter of the line print was about 2 mm. Vertical bars of 3.7 x 1.7 cm (height x width) size served as go-signals, whereas horizontal bars of the same size (width x length) as the vertical bars served as no-go signals. Go- and No-go-signals were also presented in white against black background, and appeared at one of three horizontal positions. The screen center served as the neutral position; the left position was 10 cm on the left of the screen center, and the right position was 10 cm to the right. Viewing distance was approximately 50 cm.

Left and right response keys were two separate external keys. Each key was mounted on a flat metal plate, which was connected to the computers via a so-called ExKey Logic, a system also provided by BeriSoft (Beringer, 2000). The response keys were located to the left and the right of a participant’s body-midline and were aligned with the screen. The distance between the two keys was approximately 20 cm.

After receiving written as well as oral instructions on the task requirements and the sequence of events within a single trial that emphasized both, the need for speed and accuracy, participants worked through five blocks. The first block consisted of three trials that were the same for all subjects. This block served as a practice block that was administered to acquaint [page 110↓]participants with the procedure (see Section Stimulus Material and Counterbalancing, above, for the construction of trials). Participants were then given four experimental blocks consisting of twenty-four trials each. Experimental trials differed across subjects as a function of the list participants were assigned to. The three lists were approximately counterbalanced across participants by Latin square, such that about eight participants out of twenty-four subjects overall saw a particular list (see Section Stimulus Material and Counterbalancing, above, for counterbalancing of trials across lists, and Section Data Analysis, below, for a description of the final sample of subjects)14.

Each trial started with a written instruction stating the mapping of responses (left or right key) to the letters for the upcoming trial. The stimulus-response mappings for a given pair (e.g., R: left key; C: right key) were presented below each other at the center of the screen. For half the trials in each block, the left-key S‑R mapping was presented above the right-key assignment, whereas it was presented below the right-key assignment on the other half of the trials. Care was taken to ensure that the instruction-order counterbalancing applied as often to order-corresponding S‑R mappings (i.e., letters left and right in the alphabet assigned to left and right responses, respectively) as to noncorresponding mappings (i.e., left and right letters assigned to right and left responses, respectively).

The mapping instructions remained on the screen for at least 2 seconds. After two seconds had passed, subjects could commence the trial when they felt they had remembered the instructions by pressing either the left or the right key, depending on their counterbalancing condition. Half the subjects in each list condition (i.e., four participants on each list) pressed the left key to initiate a trial, whereas the other half pressed the right key. Pressing the key triggered a fixation cross at the central screen position that remained on the screen for 500 ms. Simultaneously with the offset of the fixation cross, one of the letters of the instructed letter pair appeared at the same position for 500 ms. Letter offset was followed by a 1200 ms blank interval after which a go-signal (vertical bar) or a no-go signal (horizontal bar) was presented for 150 ms at one of the three horizontal positions (left, middle, right) that indicated whether the (prepared) response should be executed or not. Response recording was initiated at go/no-go signal onset and was terminated when a response was made or after 950 ms had passed. If participants made the wrong response (i.e., pressed the wrong key on go-trials), or if their go-[page 111↓]responses were either too slow (i.e., slower than the 950 ms recording duration) or too fast (i.e., when a response was made before the go/no-go signal appeared) they received a written error feedback in red against black background on the lower part on the screen for 1000 ms, followed by a 500 ms blank interval after which the mapping instruction for the next letter pair (i.e., the next trial) appeared on the screen. When participants incorrectly pressed a key on no-go trials, a 1500 Hz warning tone was presented for 500 ms via the internal PC speaker, followed by a blank interval of 500 ms until the next trial started. In the case of correct (non-) responses, the next mapping instructions were presented after a blank interval of 500 ms that followed the response or the maximum recording duration.

Trials were presented in a quasi-random pre-determined order, which ensured that (a) at least six trials/letter pairs intervened between the two repetitions of a single letter pair within a block (mean number of intervening trials = 10.4 on each list), (b) trials realizing the same stimulus (go-/no-go signal) position x response location condition did not follow each other more than once in a row, and (c) no-go trials neither occurred on the first nor the last position within a block and did not follow each other immediately.

5.1.2 Results

Data Analysis

Six participants were excluded from the analyses because they produced errors or misses on more than 10% of the trials. This relatively strict exclusion criterion seemed warranted given the low number of trials overall, leading to unreliable RT estimates when participants make too many errors. One additional participant was excluded because of very slow responses (i.e., his or her average RT across conditions exceeded the overall group mean by more than 2.5 standard deviations), indicating consistent postponement of S‑R translation until go-signal presentation. However, inspection of the data of excluded participants revealed a qualitatively similar pattern of results as that observed for included participants.

The raw RTs of correct go-responses of the remaining twenty-four participants (14 female, 10 male, mean age = 23.1 years) were trimmed by excluding trials that were more than 2.5 standard deviations away from a participant’s overall mean across compatibility conditions. This screening procedure was applied in order to eliminate individual RT outliers that may have unduly influenced the RT estimates, given the small number of trials used in the present experiments (cf. Ratcliff, 1993, who argued that data trimming helps to stabilize RT [page 112↓]estimates). Overall, 2.6% correct responses were screened, and the pattern across compatibility conditions resembled that of the errors (see Section 4.2.2). No-go errors (i.e., false alarms) occurred on 32% of the no-go trials, that is, on about four of the twelve no-go trials, and were not analyzed further.

For the main analysis (i.e., the subject analyses), median RTs for each participant were computed on trimmed correct raw go-RTs as a function of compatibility15. The percentage of incorrect or missing go-responses (PI) was determined accordingly. However, because of the rather strict error-based exclusion criterion (i.e., 10% across conditions, see above), the errors in the present set of experiments are less conclusive than in the dual task experiments reported above (see Chapter 4). Therefore, only overall (omnibus) PI analyses were run to ensure that the error pattern did not contradict RTs. Data were aggregated on compatibility conditions instead of stimulus position x response location primarily because of the low number of trials per condition. That is, aggregating on compatibility conditions seems to provide more reliable estimates. Moreover, this procedure seems more comparable to the aggregation procedure used in the dual task experiments presented in the previous chapter.

Again, in all analyses presented in the result section of Experiment 4 (and Experiment 5), reported p – values for effects involving factors with more than two within-‘subjects’ conditions are based on Greenhouse-Geisser corrected degrees of freedom. This also applies to the item analyses.

In addition to the subject analysis, an item analysis was carried out to test whether the same pattern of results emerges when items (instead of participants) are treated as a random factor. This type of analysis is akin to the item-analyses regularly found in psycholinguistic studies (e.g., Clark, 1973). Like in psycholinguistic studies, an item analysis seems warranted in the present experiment because (a) for a single subject, different compatibility conditions were realized by different items, and (b) a rather restricted number of items (i.e., letter pairs) realized each condition. Therefore, the interpretation of the results from the subject analysis rests on the assumption that the same results are obtained regardless of the specific item instances in a particular condition. If this assumption is correct, then similar results as in the subject analysis should be observed when items are treated as “subjects”. For the item analy[page 113↓]sis median RTs16 for correct go-responses, and PIs for incorrect go-responses, were calculated as a function of letter pair (i.e., letters occurring together in a given S‑R instruction) and compatibility, averaging across participants. Four items (letter pairs) had to be excluded from the item analysis because they served as no-go trials for the same compatibility condition on more than one list, and therefore did not provide data for all compatibility conditions. After excluding these four items, data from forty-four items, based on a maximum of sixteen observations (participants) per compatibility condition, were entered into the analysis. The motivation for treating letter pairs instead of single letters of a pair as items was similar to that of aggregating on compatibility instead of stimulus position x response location in the subject analysis, namely to maximize the number of observations per item and condition.

Main Results

The group means of the individual median RTs and the PIs for the final sample of subjects (N = 24) are presented in Table 11. As is evident from Table 11, there was a 19 ms Simon effect in RTs that was due to interference on incompatible trials.

Table 11: Mean Median Reaction Times (RT in ms) and percent invalid (PI) as a Function of Compatibility between Go-Signal Position and Response Location in Experiment 4.

 

S-R compatibility

 

Compatible

 

Neutral

 

Incompatible

 

Δ

        

RT

332

 

332

 

351

 

19 ms

PI

1.9

 

3.0

 

4.3

 

2.4%

Note: The Column labeled Δ indicates the size of the Simon effect (incompatible minus compatible).

This observation received support from the omnibus RT ANOVA with S‑R compatibility (compatible, neutral, incompatible) as a within-subjects factor, F(2,46) = 10.02, p < .001, MSe = 270.01. Planned comparisons testing all three ‘components’ separately, that is, the overall Simon effect (i.e., compatible vs. incompatible), its interference component (i.e., neutral vs. incompatible), and the facilitation component (i.e., neutral vs. compatible), showed that the Simon effect, F(1,23) = 22.24, p < .001, MSe = 370.12, and the interference component, F(1,23) = 13.27, p < .01, MSe = 602.29, were significant, whereas the facilitation component was not, F(1,23) < 1, MSe = 647.65.

The compatibility main effect in the overall PI ANOVA just missed significance, F(1,46) = 2.8, p < .08, MSe = 12.19. However, PIs showed a similar pattern as RT, as indi[page 114↓]cated by an overall 2.4% Simon effect and the MANOVA that yielded a significant compatibility effect, F(4,20) = 8.4, p < .001.

Additional Analyses

As for the dual task experiments (cf. Chapter 4), a distribution analysis was carried out to explore the temporal dynamics of the Simon effect. However, unlike in the dual task experiments reported above and in the study by Hommel (1996c, Exp. 1), rank-ordered RTs were segregated into only two instead of five bins for each participant and condition. The reason to go with median splits (i.e., fast vs. slow RTs for each participant and compatibility condition) instead of a quintile analysis was that, for the former, the maximum number of observations per bin and condition was already reduced to 14, whereas the latter would be based on only 5-6 observations each, likely rendering the RT estimates highly unreliable. Figure 10 shows the group means of the median RTs for each participant, RT bin and compatibility condition.

Figure10: Mean median RTs for slow and fast responses as a function of compatibility between Go-Signal Position and Response Location in Experiment 4.

Figure 10 indicates that (a) the Simon effect was present for both fast and slow responses, but that (b) the effect (i.e., incompatible minus compatible) tended to be slightly larger for slow than for fast RTs, and (c) that the Simon effect for slow responses tended to be more symmetric with respect to interference (9 ms) and facilitation (13) ms than for fast re[page 115↓]sponses (14 ms vs. 1 ms, respectively). However, the 2 (RT-bin) x 3 (compatibility) within-subjects ANOVA only yielded significant main effects of RT-bin, F(1,23) = 260.34, p < .001, MSe = 1,089.47, and compatibility, F(2,46) = 5.82, p < .01, MSe = 720.28, whereas the interaction of RT-bin and compatibility was not significant, F(2,46) = 1.19, p > .3, MSe = 400.89.

Finally, I also carried out an item-analysis to assess whether the same pattern as in the main analysis emerges when items instead of subjects are treated as a random factor. To this end, I determined the median RTs and PIs for each of the forty-four letter pairs for which data points existed for all compatibility conditions (cf. Section Data Analysis, above) as a function of compatibility (see Table 12 for means of medians across items). The within-‘subjects’ ANOVA of item RTs showed that the compatibility conditions significantly differed from each other, F(2,86) = 6.85, p < .01, MSe = 1,257.67, thus corroborating the results obtained in the subject analysis. Interestingly, in the item analysis, the Simon effect seemed to be more symmetric regarding the neutral condition than in the subject analysis. However, planned comparisons only revealed a significant Simon effect, F(1,43) = 18.44, p < .001, MSe = 1,868.07, whereas the interference and the facilitation component missed significance (F(1,43) = 2.62, p > .11, MSe = 3,117.23; and F(1,43) = 3.54, p >.06, MSe = 2,560.74, for interference and facilitation, respectively).

Importantly, the distribution of the Simon effect across items was clearly unimodal, and order correspondence (e.g., A: left key, N: right key as opposed to R: left key, C: right key) did not interact with compatibility, indicating that the Simon effect in the present experiment is not restricted to the subset of items with order-corresponding S‑R assignments.

Table 12: Mean Median Reaction Times (RT) and percent invalid (PI) as a Function of Compatibility between Go-Signal Position and Response Location in the Item Analysis of Experiment 4.

 

S-R compatibility

 

Compatible

 

Neutral

 

Incompatible

 

Δ

        

RT

328

 

342

 

356

 

28 ms

PI

1.9

 

3.3

 

4.2

 

2.3%

Note: The Column labeled Δ indicates the size of the Simon effect (incompatible minus compatible).

The pattern of PI across compatibility conditions closely resembled that in the subject analysis, and again missed significance, F(2,86) = 2.42, p > .10, MSe = 24.35. However, as in the subject analysis, the MANOVA on RT and PI showed a highly significant compatibility effect, F(4,40) = 5.94, p < .001, indicating that PI did not contradict RT in the item analysis.


[page 116↓]

5.1.3  Discussion

Experiment 4 extends the findings obtained by Hommel (1995, 1996c) as well as Shiu and Kornblum (1999) by demonstrating that a Simon effect of normal size (cf. Proctor & Lu, 1999, p. 67, who state that Simon effects observed with visual stimulus material are typically about 25 ms or less) occurs in go/no-go Simon tasks, even when the imperative stimuli are arbitrarily mapped to left and right responses and when S‑R mappings change from trial to trial. This result implies that (a) the go-/no-go Simon effect does not depend on S‑S congruency between a spatial precue and go-signal position (i.e., between go-position and direction of arrows in the Hommel experiments, or between go-position and location words in the Shiu & Kornblum, 1999, study), and that (b) the Simon effect in this task is not restricted to highly overlearned S‑R mappings, but is also obtained when new stimuli need to be translated to their assigned spatially instructed responses on each trial. The results from the item analysis support the main results, indicating that the Simon effect observed in this Experiment is relatively stable despite the low number of items used in the experiment.

Numerically, the Simon effect tended to be slightly (7 ms) larger for slow than for fast responses, paralleling results obtained by Hommel (1996c, Exp. 1). Hommel proposed that faster responses exhibit a smaller Simon effect because at the time fast responses are emitted, the irrelevant location code has not yet been formed. In contrast, spatial codes are formed ‘in time’ to affect slow responses. However, unlike in the Hommel (1996c) experiment, the interaction between RT-bin and compatibility was not significant in the present study. Possible explanations are lack of power, and the relatively coarse segmentation of responses (i.e., 2 vs. 5 bins). The latter might have led to median RTs that differ less between bins than the average RTs in the extreme bins of Hommel’s quintile analyses. Moreover, RTs in my experiment were somewhat slower on average than in the Hommel experiment (this applies to medians as well as means), possibly resulting in more overlap in code activation in my experiment.

At least in the subject analysis, the compatibility effect in RTs was entirely due to interference (i.e., slower responses on incompatible than on neutral trials). No facilitation effect showed in the overall (subject) RTs. In contrast, most of the (relatively few) studies using a standard Simon task (in which the imperative stimuli appear at varying locations) that included a neutral condition found facilitation effects. Unfortunately, none of the experiments that used a go/no-go variant of the Simon task included a neutral condition. Thus, at present, it cannot be determined whether my results are akin to delayed stimulus-position type of tasks [page 117↓]in general, or whether they are due to the specific neutral condition used in the present experiment. According to the former possibility, no facilitation effect might be observed in delayed-position experiments because facilitation only occurs with relatively slow responses (note that choice reactions are typically much slower, i.e., about 500 ms, than RTs in the present experiment), whereas interference is less affected by response speed or even more pronounced for fast responses. Some, albeit weak, evidence in support of this explanation comes from the RT-bin analysis that showed a (nonsignificant) facilitation effect for slow, but not for fast responses. This possibility is theoretically interesting because it may imply that response preparation, but not response initiation of fully prepared responses benefits from irrelevant corresponding location information.

However, there also was a tendency for a facilitation effect in the overall item RTs that cannot be explained in terms of differential response speed because overall RTs in the subject and the item analysis were almost identical. Hence, an alternative explanation of the present results is that there might have been much variability in how subjects treated the neutral condition, leading to inconsistent facilitation effects in the subject analysis when aggregating across items for individual subjects, but not necessarily so when averaging across subjects (i.e., in the item analysis). Clearly, further research is required to resolve this issue. One way of addressing this question would be to vary the SOA between imperative stimuli and go/no-go signals, possibly comparing different neutral conditions. If only unprepared responses benefit from corresponding trials, then larger facilitation effects should be observed at short than at long SOAs, regardless of how the neutral condition is realized.

A point related to the question of whether facilitation effects only occur for slow, unprepared responses concerns the locus of the Simon effect obtained with delayed location presentation. On the one hand, (a) the high proportion of go-trials (87.5%), (b) the fact that location information was presented 1700 ms after the onset of the imperative stimuli, and (c) the finding of comparatively short overall RTs, it seems safe to conclude that (imperative) stimulus identification and S‑R translation had been completed on most trials when the go/no-go signals appeared, thus rendering a translation explanation of the Experiment 4 results unlikely. Instead, the Simon effect observed in this task strongly favors a direct response activation account. On the other hand however, one may ask whether the source of the Simon effect differs across tasks requiring unprepared vs. prepared responses. According to one view (e.g., Hommel, 1997), response selection is accomplished by activating the corresponding [page 118↓]codes, and stimulus processing and response processing overlap in time. This view implies that

(Logan and Gordon) as long as the response is not carried out, any response-congruent or conflicting stimulus information may facilitate or hamper responding. Consequently, response uncertainty should not play a major role(Logan and Gordon). (Hommel, 1997, p. 298)

Accordingly, this view does not make a principled distinction between the locus of the Simon effect observed in situations where the response has to be prepared (i.e., when location information appears during S‑R translation) and situations in which location information affects the initiation of a prepared response. However, one may disagree with this view if one adheres to more traditional, stage-like models. More specifically, it could be argued that the go/no-go Simon effect differs from the Simon effect observed when the position of the relevant stimulus varies in that the former reflects interference at initiating a prepared response, whereas the latter measures online interference during response selection (and, perhaps, response initiation; see, e.g., Shiu & Kornblum, 1999, for a discussion). I will return to this issue in the General Discussion section (Chapter 5.3).

In sum, Experiment 4 established a Simon effect for a 1-trial Simon task involving spatial response instructions while avoiding the confounding of instruction and practice effects typically present in other experiments on response coding. However, this experiment does not yet permit any conclusions about whether response instructions affect the Simon effect because spatial response instructions do not allow to discriminate between the alternative coding hypotheses (see Chapter 3). In order to address the question of whether response instructions affect response coding, response instructions in Experiment 5 were changed to non-spatial.

5.2 Experiment 5

Experiment 5 was largely identical to Experiment 4, except that responses were no longer instructed in terms of location. Rather, keys were instructed in terms of color, as in Experiments 2 and 3. That is, participants were instructed to press the green or the blue key, depending on letter identity. As in Experiment 4, a new letter pair was instructed on each trial, this time specifying letter to color (key) mappings (e.g., A: blue key, N: green key, on trial n; and R: blue key, C: green key, on trial n+1). Unlike the dual task Experiments 2 and 3, color-to-key assignment remained constant within a given block of trials, and was changed after a block had been completed. This modification was motivated by two considerations. First, [page 119↓]keeping color-to-key assignment constant within a given block seemed to render the procedures in Experiments 4 and 5 more comparable. Second, it appeared to provide a fairer test of the different coding hypotheses because a constant color-to-(key) location assignment within a block can be assumed to facilitate recoding of responses in terms of location.

The predictions were similar to those of Experiment 3. If instruction determines response coding, that is, if participants include and weigh response codes in their response representations as instructed, then instructing responses in terms of color should deemphasize spatial response codes, and hence reduce the influence of irrelevant stimulus location. Therefore, the strong version of the direct coding hypothesis predicts a reduced Simon effect (reduced compared to the Simon effect with spatial response instructions).

In contrast, according to both the spatial and the weak version of the direct coding hypothesis, the spatial Simon effect under color instructions should not differ from that observed in Experiment 4. This is expected because both hypotheses assume that spatial coding is unaffected by instructions.

5.2.1 Method

Participants

Thirty-two students and non-students from Berlin (18 female, 14 male, mean age = 23 years) received € 7,- for participation (see Section Data Analysis, below, for a description of the final sample).

Stimulus Material and Counterbalancing

The same stimulus material and counterbalancing scheme as in Experiment 4 was used, with the following exceptions.

The S‑R mapping instructions for each letter pair were changed such that letters were assigned to keys referred to by the color names BLAU and GRÜN, the German words for blue and green, respectively (e.g., R: blue key; C: green key). Color-to-key assignment was changed from block to block. On the practice block, as well as on Blocks 2 and 4, the green key was the one located on the left, and the blue key the one on the right. On the remaining blocks the assignment was reversed. This was true for all lists. Thus, letters that had been assigned left vs. right key responses in Experiment 4 were now assigned blue or green key responses, depending on block, keeping the physical letter-to-key (location) assignment con[page 120↓]stant across experiments. Because color-to-key assignment was changed between blocks, changes of letter-color assignments for a given letter on successive blocks (i.e., in new letter pairs) occurred when the location of the response remained the same. For 22 out of the 24 letters the color-assignment changed at least once across blocks.

The same trials as in Experiment 4 served as no-go trials, again “eliminating” two trials in each stimulus-position x response-location condition, leaving fourteen trials per condition (28 per compatibility condition) for the go-trial analyses. Because the same no-go trials were used as in Experiment 4, and because color-to-key assignment changed between blocks, the number of blue and green no-go trials was not equal on all lists. More specifically, whereas the same number of blue and green trials was (accidentally) excluded on Lists 1 and 3, on List 2 there were 8 green no-go trials (4 left and 4 right responses; 2 green trials on each block) and only 4 blue no-go trials (2 left and 2 right; 1 blue trial on each block).

Apparatus and Procedure

Apparatus and procedure were the same as in Experiment 4 with the following exceptions. In the introductory instructions, the position of the response keys was not mentioned. Rather, participants were told that their task would be to press either the blue or the green key, depending on the letter, and that they would be informed at the beginning of each block which key would be the blue viz. green key in that block.

Accordingly, each block started with a presentation of color patches of 3.5 x 3.5 cm size that simultaneously appeared at the left and right screen positions (i.e., the same lateral positions at which the go/no-go signals appeared). As noted above, on the practice block, as well as the second and fourth experimental block, the green patch appeared on the left, and the blue key appeared on the right, indicating that the left key had to be pressed when a green response was required and vice versa. On the first and third experimental blocks, the color-to-key assignment was reversed. When participants saw the color patches for a given block on the screen, the experimenter arranged color labels on the keys accordingly. These color labels consisted of 2.6 x 3.8 cm (height x width) paper color patches that lay behind the keys on the metal plates on which the keys were mounted. Extra color-labels on keys were provided to ensure that participants would not screw up a complete block of trials just because they had forgotten the color-to-key assignment.


[page 121↓]

Counterbalancing of lists17, the sequence of trials, and the sequence of events within a trial were identical to Experiment 4, except that the mapping instructions (i.e., instructions assigning letters to responses) now instructed green vs. blue as opposed to left vs. right key responses. Substituting the spatial mapping instructions with color instructions resulted in presenting the green-key S‑R mapping above the blue-key assignment on half of the trials in each block, whereas the order was reversed for the other half, again considering ordinal correspondence of letter-to-key location assignment.

5.2.2 Results

Data Analysis

Again, several participants from the total sample were excluded according to the same pre-experimentally defined criteria as in Experiment 4. Six participants produced more than 10% errors or misses. One additional participant was excluded because his or her overall RT exceeded the group mean by more than 4 standard deviations. Finally, one further participant had to be excluded because of a program error. However, the pattern of results for the excluded subjects did not statistically deviate from the results for included participants (see main results below). That is, although there seemed to be a hint of a Simon effect in PIs (but not in RTs), this tendency did not reach significance when data from excluded participants of Experiments 4 and 5 were combined (N = 15). On average, false alarms on no-go trials occurred on about three of the twelve no-go trials (25.4%) and were not analyzed further.

Trimming of correct go-RTs and aggregation of RTs and PIs for the remaining twenty-four participants (13 female, 11 male; mean age = 23 years) was the same as in Experiment 4. Screened correct responses were again excluded. They occurred on 2.6% of the trials and were evenly distributed across compatibility conditions. The item analysis was based on the same 44 items (i.e., letter pairs) as in Experiment 4 and was carried out accordingly.

Main Results

For the data of the remaining N = 24 participants, median RTs and PIs were computed for each participant and compatibility condition (see Table 13 for group means).


[page 122↓]

Table 13: Mean Median Reaction Times (RT in ms) and percent invalid (PI) as a Function of Compatibility between Go-Signal Position and Response Location in Experiment 5.

 

S-R compatibility

 

Compatible

 

Neutral

 

Incompatible

 

Δ

        

RT

332

 

331

 

337

 

5 ms

PI

1.6

 

1.0

 

2.5

 

0.9%

Note: The Column labeled Δ indicates the size of the Simon effect (incompatible minus compatible).

Table 13 shows that the effects in both, RT and PI were numerically very small and extremely reduced as compared to the 19 ms Simon effect under spatial response instructions (i.e., in Experiment 4) The overall RT level was comparable across experiments (333.3 ms in Experiment 5 vs. 338.3 ms in Experiment 4).

These observations were supported by the ANOVAs. The compatibility effect in RTs did not reach significance, F(2,46) = 1.25, p > .29, MSe = 226.61, nor did any of the planned pairwise comparisons (F(1,23) = 1.14, p > .29, MSe = 534.30, for the Simon effect; and F(1,23) < 1, MSe = 436.15, and F(1,23) = 2.66, p > .11, MSe = 389.19, for facilitation and interference, respectively).

Compatibility did not reach significance in the omnibus PI ANOVA either, F(2,46) = 1.88, p > .16, MSe = 7.25. Although the PI pattern across compatibility conditions seemed to generally follow the RT pattern, additional correlation analyses and a MANOVA were carried out to ensure that the null results were not masked by some kind of speed-accuracy tradeoff. Neither analysis revealed a hint for a tradeoff. The correlation between the Simon effects (i.e., the Δs) in RTs and PIs was nonsignificant, r = ‑.08, p > .7. Similarly, the doubly multivariate MANOVA that considered all compatibility conditions again supported the RT analysis by not revealing a significant compatibility effect either, F(4,20) = 1.19, p > .34.

Further evidence for a reduction of the Simon effect under non-spatial compared to spatial response instructions comes from a comparison of the Simon effects in Experiments 4 and 5. The 2 (experiment) x 2 (compatible vs. incompatible) ANOVA on RTs revealed a significant main effect of compatibility, F(1,46) = 14.7, p < .01, MSe = 226.10, that was qualified by an interaction between compatibility and experiment, F(1,46) = 4.83, p < .05, indicating that the Simon effect was significantly larger in Experiment 4 than in Experiment 5. Moreover, as already noted, overall RT level was comparable across experiments, as indicated by a nonsignificant main effect of experiment, F(1,46) < 1, MSe = 5,651.33.


[page 123↓]

Additional Analyses

As in Experiment 4, a median-split analysis was carried out to assess whether the effect differed as a function of response speed. Figure 11 shows the group means of the medians of each participant, compatibility condition, and RT-bin.

Figure 11: Mean median RTs for slow and fast responses as a function of compatibility between Go-Signal Position and Response Location in Experiment 5.

No compatibility effect was obtained at either RT-bin in Figure 11. The Simon effect was 2 ms for fast responses, and 4 ms for slow responses, indicating that the (null-) effect obtained in the main analysis did not vary with response speed. That is, no reversal of the effect at either RT-bin masked an effect at the other bin, as also indicated by the ANOVA that only yielded a main effect of RT-bin, F(1,23) = 192.17, p < .001, MSe = 1,336.87. Neither the compatibility main effect nor the interaction of compatibility and RT-bin reached significance, both F’s < 1.

Finally, the item analysis again revealed that the pattern of results obtained for subjects generalizes to items, as indicated by the group means (of item medians) presented in Table 14 and the item ANOVAS.


[page 124↓]

Table 14: Mean Median Reaction Times (RT) and percent invalid (PI) as a Function of Compatibility between Go-Signal Position and Response Location in the Item Analysis of Experiment 5.

 

S-R compatibility

 

Compatible

 

Neutral

 

Incompatible

 

Δ

        

RT

335

 

331

 

336

 

1 ms

PI

1.7

 

0.9

 

2.4

 

0.7%

Note: The Column labeled Δ indicates the size of the Simon effect (incompatible minus compatible).

The item RT ANOVA revealed that compatibility did not reach significance, F(2,86) < 1, MSe = 454.74. Nor did the Simon effect,or any of its components (all F’s < 1) when analyzed separately in planned contrasts. As in Experiment 4, the distribution of the Simon effect in item RTs was unimodal, and compatibility did not significantly interact with order correspondence, again suggesting that order correspondence has a negligible effect in the present task.

The PI pattern in the item analysis closely resembled that of the subject analysis, and neither the PI ANOVA nor the MANOVA indicated any tradeoff between RT and PI (compatibility effects of F(2,86) = 2.45, p > .1, MSe = 10.54; and F(4,40) = 1.55, p > .2, for the PI ANOVA and the MANOVA, respectively).

5.2.3 Discussion

When responses were instructed in terms of color, the effect of corresponding viz. noncorresponding irrelevant position information on responding was small (5 ms) and not significant. Moreover, the Simon effect with non-spatial response instructions was significantly smaller than the 19 ms Simon effect observed with spatial response instructions (Experiment 4).

The results from the experimental comparison, as well as the fact that the statistical error term for the compatibility effect in the present experiment did not exceed that of Experiment 4 argues against a power explanation of the null-effect. Furthermore, the item analysis was consistent with the subject analysis. The item-RT distribution was again unimodal, and compatibility did not interact with alphabetic ordering, suggesting that the null-effect observed in Experiment 5 is “real” and cannot be attributed to a reversed effect in a subset of items (i.e., order-noncorresponding letter pairs).


[page 125↓]

Unlike the dual-task Experiment 3 reported above, color-to-key assignment remained constant throughout a complete block of trials, that is, recoding instructions and/or responses in terms of location was comparatively easy.

The Simon effect did not increase with increasing RT level, as indicated by the distribution analysis. This finding suggests that even less “prepared” responses, that is, responses for which S‑R translation may not have been completed when the go-signal was presented, were hardly affected by irrelevant stimulus location under non-spatial response instructions.

Consequently, Experiment 5 corroborates the Experiment 3 results and provides converging evidence in favor of the strong direct coding hypothesis. It suggests that participants arbitrarily coded their responses in terms of color as instructed, weighing instructed codes more strongly than uninstructed spatial codes.

5.3 General Discussion Experiments 4-5

The experiments presented in Chapter 5 corroborate the dual task results reported in Chapter 4 by providing converging evidence for an impact of response instructions on response coding. More specifically, in Experiment 4 and 5, responses on a Simon-like task with delayed position presentation were either instructed spatially or in terms of color. Any effect of stimulus position and of response instructions on the Simon effect obtained with this task cannot easily be explained by translation accounts. Consistent with the intentional weighing hypothesis, which assumes that response codes can be weighed according to instructions and hence predicts a reduced Simon effect under non-spatial instructions, Experiments 4 and 5 showed that the 19 ms Simon effect observed under spatial response instructions was significantly reduced to 5 ms under color instructions.

The reduction of the Simon effect under color instructions contradicts dual route models such as the DO model proposed by Kornblum and colleagues (e.g., Kornblum et al., 1990; Kornblum et al., 1999; Zhang et al., 1999), which postulate that “when there is correspondence between the stimulus code and the response code, the latter is automatically activated, regardless of their relevance to the task” (Azuma et al., in press). Therefore, according to the DO model, which can be considered an example of the weak direct coding hypothesis, one would have expected a normal Simon effect under color instructions.

As a consequence, the results also contradict models that restrict activation via the direct route to the spatial dimension, and hence propose instruction-independent (pure) spatial response coding (e.g., De Jong et al., 1994).


[page 126↓]

Empirically, my results with non-spatial response instructions extend Hommel (1993a) who found a reversal of the Simon effect when response instructions emphasized different spatial aspects of the response array (i.e., either key location or contralaterally presented response effects). Unlike the Hommel (1993a) experiment, no salient visible action effects were presented in Experiment 5. Instead, color-plates were shown at the beginning of each block, and non-spatial response coding was evoked by verbally instructing the responses in terms of color before each trial. This finding indicates that instruction suffices to implement the intention to make color responses, that is, to implement arbitrary response codes in the action representation and persuade participants to use these codes in the control of responding.

The present findings also extend the results by Simon et al. (1976, Exp. 2; see Chapter 3.1.4) by providing a spatial response-instruction baseline for assessing the impact of non-spatial response instructions on the Simon effect. However, whereas the Simon effect in Experiment 5 was close to non-existent, Simon et al. observed a substantial (36 ms) effect even when color-to-key assignment changed from trial to trial. Several procedural differences may be responsible for this discrepancy. First, Simon et al. used an auditory variant of the Simon task that typically leads to much larger Simon effects (i.e., around 60 ms) than visual Simon tasks. Thus, as argued above, color instructions may have reduced the effect in Simon at al.’s varied label groups.

Moreover, auditory stimulus presentation, combined with the fact that lights attached to the keys served as color-labels in the Simon et al. experiment, enabled (or even required) subjects to fixate the response arrangement during stimulus processing. In contrast, participants could not simultaneously fixate the go/no-go signals (i.e., the screen) and the labels mounted on the response keys in my Experiment 5, presumably resulting in a higher likelihood of effector-position (or key location) coding in the former study. Finally, in the Simon et al. experiment, the relevant S‑R mapping (i.e., pitch to color) was instructed only once at the beginning of the experiment, whereas new S‑R mappings were instructed on each trial in Experiment 5. That is, in the latter but not in the former experiment, left and right keys were repeatedly referred to by color names, possibly priming the color dimension and reducing the likelihood of spatial re-coding.

Figure 12 illustrates how instructions in Experiments 4 and 5 might have affected response coding, and hence, the Simon effect, on incompatible trials.


[page 127↓]

Figure 12: An illustration of the hypothetical activation flow on incompatible trials in Experiment 4 (Panel A) and Experiment 5 (Panel B).

Go/no-go position (Sp) activates its corresponding location codes in either case. However, with non-spatial response instructions (Panel B) location codes are less strongly weighed, leading to less interference on incompatible trials. Note: Slet = (representation of) the imperative letter stimulus.

When responses are instructed in terms of location, as was the case in Experiment 4, spatial response codes are highly weighed and linked to/activated by the letter stimuli (Slet). Stimulus position (Sp) activates competing – highly weighed - location codes and thus leads to relatively strong response competition (Figure 12, Panel A). In contrast, when responses are instructed in terms of color, as was the case in Experiment 5, color codes are weighed more strongly than location codes that only play a minor role in response coding and selection. Consequently, even if stimulus position activates location codes (Figure 12, Panel B), this activation leads to less interference, and hence, reduced response competition.

Such an explanation of the reduced Simon effect under color instructions seems to bear some similarity with Magen and Cohen’s (2002) notion of ‘response-based input-selection’. In the experiments reported by Magen and Cohen, task-irrelevant flankers (and Stroop distractors) that were not part of the target set did or did not overlap with the outcome specification of verbal responses. These experiments showed distractor interference effects only when the distractors matched (or mismatched) the required verbal responses. Magen and Cohen [page 128↓]explained these results by suggesting that the output specification primes specific dimensions, opening the door for irrelevant stimuli that overlap regarding the relevant response dimension. Similarly, in Experiments 4 and 5, a Simon effect was only observed when the irrelevant stimulus attribute overlapped with the instructed manual response dimension.

As already discussed (see Chapter 5.1.3), one peculiarity about the Simon effect obtained with spatial response instructions in Experiment 4 was that the RT effect was entirely due to interference, at least according to the subject analysis. Two possible reasons for this outcome have been discussed. According to one, this effect is an artifact produced by how subjects treated the neutral condition used here, whereas, according to the other explanation, an interference-dominant pattern of results may be systematically obtained with delayed-position type tasks. Although some researchers (e.g., Hommel, 1997) do not make a principled distinction between the Simon effects with vs. without response uncertainty, it is nevertheless possible that prepared responses (i.e., those for which S‑R translation is completed at presentation of location information) are differently affected by irrelevant location information than yet-to-be-prepared responses. If so, one might argue that the observed instruction effect may be restricted to situations where fully programmed responses have to be initiated, but would not be as pronounced if irrelevant position information were presented during S‑R translation.

While a definite answer to this question can only be provided by experiments that vary the SOA between the imperative stimuli and the go/no-go signals (see Section 5.1.3), the present results do not support this view. That is, if unprepared responses were more affected by stimulus position under non-spatial instructions, then there should have been a hint of a Simon effect for slow RTs in Experiment 5 because slow responses can be assumed to be less prepared on average. However, the distribution analysis of the Experiment 5 data did not indicate such a tendency – the Simon effect was 4 ms for slow as opposed to 2 ms for fast responses in Experiment 5. Therefore, it seems unlikely that the observed reduction of the effect was due to the specific task used here.

A final argument against an interpretation of the instruction effect reported in Chapter 5 in terms of response coding concerns recent observations that the use of the direct route might be – at least partially – under intentional control. For instance, Stürmer et al. (2002) demonstrated that the size of the Simon effect depends on (a) the proportion of compatible vs. incompatible trials, and (b) whether the preceding trial (i.e., trial n-1) was compatible or not. [page 129↓]Stürmer et al. found a reversed Simon effect when the overall percentage of incompatible trials was high. Moreover, they reported a regular Simon effect only for those trials following compatible trials, and this effect was independent of the overall proportion of compatible/incompatible trials. They interpreted the latter (sequence) effect as indicating suppression of the direct route after encountering conflict (i.e., upon incompatible trials), suggesting that the direct route may not be as unconditionally automatic as commonly assumed (but see Hommel, Proctor, & Vu, in press, for an alternative interpretation of such sequential effects).

Depending on how much intentional control (e.g., suppression) can be exerted on the direct route, one could argue that the reduction of the Simon effect in Experiment 5 was due to participants’ deliberate decision not to use the direct route. However, this possibility seems questionable because Experiments 4 and 5 were identical in terms of the number of trials in each compatibility condition (all compatibility conditions were equally frequent), as well as regarding the sequence of trials. Thus, if participants had as much control over the direct route as implied by this view, the question arises as to why a Simon effect was observed in Experiment 4. More generally speaking, the question would be why an effect of irrelevant stimulus position (i.e., a Simon effect) usually shows with equiprobable compatibility conditions at all.

Nevertheless, in order to back up my conclusion that participants indeed coded their responses in terms of color, I am currently planning an experiment similar to Experiment 5, but with irrelevant color instead of location information, following the logic of Experiment 2. More specifically, responses will again be instructed in terms of color, and go/no-go signals determine whether a response is to be executed or not. However, unlike Experiment 5, go/no-go signals (vertical and horizontal bars) will be presented centrally and will randomly vary in terms of color instead of position. The (irrelevant) color of the go-signals either corresponds or does not correspond with the instructed key color, or is neutral with respect to the required response. If participants indeed arbitrarily code their responses as instructed, then irrelevant color information should automatically activate the highly weighed color codes, leading to interference on incompatible conditions. That is, a color-based Simon effect should be observed. Such an effect would also provide evidence against a strategic direct-route-suppression account of the reduced Simon effect in Experiment 5.

In sum, the Experiments presented in Chapter 5 extend existing findings and corroborate the dual-task results presented in Chapter 4 by showing that non-spatial response instructions can reduce the impact of irrelevant stimulus location on manual responding. In the pre[page 130↓]sent paradigm, irrelevant location information was presented considerably after the onset of the relevant (letter) stimulus, suggesting that the effect of instruction (i.e., the Simon effect in Experiment 4 and the lack thereof in Experiment 5) was not due to S‑R translation in the conditional route. Rather, the findings seem to imply that instructions directly affected response coding. Thus, the present results provide additional evidence for the strong version of the direct coding hypothesis, which holds that codes referred to in the instructions are weighed more strongly and thus govern responding.


Footnotes and Endnotes

14 Due to experimenter error, one participant that should have received a particular list according to the counterbalancing scheme was mistakenly given another list.

15 The pattern of results for trimmed medians closely resembled the numerical pattern of medians based on unscreened RTs. Similarly, analyses on trimmed means led to comparable results as the analyses on trimmed medians reported here.

16 Data aggregation for the item analysis was based on untrimmed raw RTs.

17 Due to experimenter error, two participants who should have received one list according to the counterbalancing scheme were mistakenly given another list.



© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.
DiML DTD Version 3.0Zertifizierter Dokumentenserver
der Humboldt-Universität zu Berlin
HTML generated:
02.09.2004