Verbal coding and the elimination of stroop interference in a matching task

Before responding to verbal incongruent color word distractors, subjects practiced matching colors to irregular shapes.. Before responding to Stroop distractor stimuli, subjects in the n

Verbal Coding and the Elimination of Stroop Interference in a Matching Task

Author(s): Michael F Mascolo and Stephen C Hirtle

Source: The American Journal of Psychology, Vol 103, No 2 (Summer, 1990), pp 195-215Published by: University of Illinois Press

Stable URL: http://www.jstor.org/stable/1423142

Verbal coding and the elimination of

Stroop interference in a matching task

to process distractors This hypothesis was tested in two experiments In the first experiment, we attempted to eliminate the need for subjects to translate target colors to verbal codes when responding to Stroop stimuli Before responding to verbal incongruent color word distractors, subjects practiced matching colors to irregular shapes It was expected that subjects would use nonverbal codes to mediate responding in this task After practice, subjects continued the matching task in the presence of incongruent color words Stroop interference persisted, contrary to predictions Because subjects re- ported adopting verbal strategies to perform the matching task, Experiment

2 was designed to control the verbal coding strategies that subjects employed Before responding to Stroop distractor stimuli, subjects in the nonsense name group practiced using nonsense names to mediate the matching of shapes to colors; subjects in the actual name group used actual color names

to mediate performance in the matching task When incongruent color word distractors were introduced, Stroop interference was eliminated for subjects

in the nonsense name group, but persisted for subjects in the actual name group The results are interpreted as consistent with an outcome conflict (Navon & Miller, 1987) or a modified translational model of the Stroop effect

In the Stroop (1935) color naming task, persons are asked to name the color of ink in which incongruent color words are printed (e.g., the word RED printed in blue ink) The time required to name the ink color of incongruent color words is delayed relative to the time required to name the ink color of congruent color words, non-color words, rows of X's, or color patches (Dyer, 1973; Klein, 1964; Regan, 1978) Conversely, incongruent ink colors do not impede responding when subjects read the name of Stroop color words In this article,

we explore the conditions under which Stroop interference can be

AMERICAN JOURNAL OF PSYCHOLOGY

Summer 1990, Vol 103, No 2, pp 195-215

C 1990 by the Board of Trustees of the University of Illinois

eliminated or attenuated Data pertaining to the elimination of Stroop interference can provide clues concerning the source of Stroop in- terference

Virzi and Egeth (1985) have proposed a translational model of Stroop interference Based on recent evidence (Dunbar & MacCleod, 1984), Virzi and Egeth rejected models of the Stroop effect that maintain that Stroop interference arises as a simple function of the competition between word and color processing for access to a single centralized decision stage (Dyer, 1973; Morton & Chambers, 1973; Palef & Olson, 1975) Rather, Virzi and Egeth suggested that infor- mation is processed using several processing systems which code and process information using different decision and response stages (cf Palef, 1978) In the traditional Stroop task, a pictorial processing system codes and processes the color input To respond, the pictorial input must be translated to a verbal code Stroop interference results from the conflict between the verbal color name code associated with the target color and the incongruent color name code activated within the verbal system

Virzi and Egeth's (1985) translational model of Stroop interference can be seen as a special case of Navon's outcome conflict model of dual- task interference (Navon, 1984, 1985; Navon & Miller, 1987) Navon suggests that dual-task interference often results from conflicts be- tween outputs of parallel processes, rather than from competition for limited processing resources (e.g., as in Kahneman, 1973) Outcome conflicts are analogous to interference caused by cross-talk among telephone lines, even when the number of telephone calls does not exceed the number of lines available In testing the outcome conflict model, Navon and Miller (1987) asked subjects to search for boys' names in one channel of a stimulus display, and for city names in another channel of the display When both search tasks were per- formed simultaneously, responses were delayed when nontargets on one channel belonged to or were related to the target category on the other channel No such effects were found when each search task was performed alone, despite identical stimulus displays These results suggest that part of the dual-task decrement observed resulted from the conflict between the outcomes of processing nontargets on one channel and targets on the other channel

In Virzi and Egeth's (1985) translational model of Stroop interfer- ence, the outcome of the linguistic processing of the incongruent color word conflicts with the outcome of the translation of the target color into a linguistic color name code The translational model differs from the more general outcome conflict model only by specifying that the outcome conflict results from an obligatory translation process

ELIMINATION OF STROOP INTERFERENCE

between two independent processing systems (the pictorial and the linguistic) If Stroop interference results from this sort of outcome conflict, then it follows that Stroop interference can be eliminated or attenuated by eliminating the need for the translational stage when performing the Stroop task For example, if subjects could be induced

to use nonverbal codes exclusively to mediate target responding in the Stroop task, then Stroop interference should be eliminated, be- cause translation of the target color to the verbal system-the system which codes distractors-would be eliminated

To test the translational model of Stroop interference, Virzi and Egeth (1985) asked subjects to sort Stroop color word stimuli (and control stimuli) into bins which were labeled with either a color word

or with a color patch When subjects were asked to sort the cards on the basis of the color of ink, incongruent color words produced greater response delays (relative to controls) when bins were labeled with color words than when bins were labeled with color patches Alter- natively, when subjects were asked to sort cards on the basis of the meaning of the color words, incongruent color words produced greater delays (relative to controls) when bins were labeled with color patches than when bins were labeled with color words Thus, in cases in which interference resulted, the codes activated by target stimuli were trans- lated to the coding system that processed incongruent distractor stim- uli When no such translation occurred, no interference was observed Similar results were reported by Dalrymple-Alford and Azkoul (1972) Several other studies have reported data which concur with Virzi and Egeth's (1985) findings McClain (1983b) presented the words HIGH and LOW in high and low pitches Subjects were asked to identify either the pitch or word using either a verbal response or a pitched hum McClain found that interference occurred when the response mode and the target stimulus dimension conflicted (e.g., pitch-verbal, word-hum), but not when the response mode and stimulus dimension were compatible (e.g., pitch-hum, word-verbal) These data concur with a translational account When subjects responded to pitch by humming, processing was restricted to a tonal system Because tones were not translated to the verbal system, interference from verbal distractor codes was avoided

Treisman and Fearnley (1969) asked subjects to sort cards contain- ing either two words, two colors, or a word and a color into same or different piles Within-dimension sorts (word-word and color-color) proceeded more quickly than between-dimension sorts (word-color), even when one of the words (for word-word matching) or colors (for color-color matching) constituted a Stroop color word Finally, Prit- chatt (1968) and McClain (1983a) eliminated Stroop interference when

197

subjects identified the ink colors of Stroop color words by pressing keys labeled with color patches, but not when keys were labeled with color words The results from these studies suggest that when target stimuli are processed using coding systems that are distinct from the coding systems that process distractors, interference by incongruent distractors can be eliminated

Several studies have demonstrated that Stroop inteference can be created when subjects are induced to process (normally innocuous) incongruent distractors within the coding systems which mediate tar- get responding Recall that incongruent ink colors normally do not delay the reading of base words of Stroop color stimuli From a translational model, word reading never requires activation of non- verbal color codes Therefore, irrelevant colors cannot conflict with the verbal color name codes that mediate responding to base words However, if subjects were to code irrelevant colors verbally, Stroop interference might result Glaser and Glaser (1982, Exp 2) presented irrelevant color patches at various intervals prior to the presentation

of color words To induce subjects to encode colors using verbal labels, congruent color words appeared with a probability of 80 The re- searchers found that incongruent colors impeded reading of target color words when colors appeared 300 to 500 ms before target words Thus, when encoded using the same coding system that processes target stimuli, irrelevant distractors can impede target responding Virzi and Egeth (1985) reported similar results using a spatial variant

of the Stroop task

Flowers and his colleagues (Compton & Flowers, 1977; Flowers & Dutch, 1976; Flowers & Stoup, 1977) have reported evidence that leads to similar conclusions For example, Flowers and Stoup (1977) asked subjects to sort shapes according to a grouping rule Either incongruent shape names (names of shapes from the opposite group)

or rows of X's were printed on the shapes When a single pictorial attribute common to all shapes in a group (e.g., curved vs edged) could be used to discriminate the groups, incongruent shape names did not impede sorting However, when subjects used grouping rules which did not yield pictorial attributes that could readily discriminate the groups, incongruent shape names impeded sorting, presumably because subjects now encoded shapes verbally As long as the irrelevant distractors were processed using a different system from that used to process target stimuli, sorting proceeded without delay

To garner additional support for a translational or outcome conflict model, we attempted to eliminate interference in a Stroop-like task

by inducing subjects to use nonverbal codes to mediate responding

to colors The study proceeded in two phases In Phase 1, subjects

ELIMINATION OF STROOP INTERFERENCE

learned to match color patches to irregular shapes We hypothesized that because neither colors nor irregular shapes constitute linguistic input, subjects would not employ verbal color name codes to mediate responding In Phase 2, subjects matched colors to shapes in the presence of incongruent color words, congruent color words, non- color words, and rows of X's If subjects adopt nonverbal codes to mediate the matching of shapes to colors, then the incongruent verbal color name codes activated by distractors could not impede respond- ing Thus, we predicted that incongruent color words would not delay responding relative to control conditions

to adopt nonverbal response codes to mediate responding to Stroop stimuli

In the distractor phase, subjects continued to perform the discrimination task

in the presence of various distractor stimuli

Subjects

Twenty undergraduates, 14 females and 6 males, attending the State University of New York at Albany, participated Subjects were paid $3.00/

hr

Stimuli and apparatus

The entire experiment was conducted in a soundproof chamber An Apple 2e microcomputer controlled stimulus presentation and data collection All stimuli were presented on a 28 x 21-cm Amdek screen A black border was placed around the computer screen, limiting the subject's view to a 17.5 x 17.5-cm viewport Subjects responded using a response box containing two buttons positioned 3 cm apart

Nonfocal colors were used in the study to increase the likelihood that subjects would encode color patches using imaginal rather than verbal color name codes The hues used were brown, orange, gray, and pink The hues corresponded to the Apple 2e low-resolution color codes 8, 9, 10, and 11, respectively The names of each of these hues served as color word distrac- tors Non-color word distractors included most, deal, wage, and best Rows

of X's served as control stimuli

The four shapes were constructed by arranging 13 squares into a shape such that there were no solid blocks of four squares and no blank areas enclosed by filled squares The four shapes are shown in Figure 1

199

Figure 1 Irregular figures used in the matching task

In the learning phase, the stimulus display consisted of one of four color bars, whose lower edge was depicted 3.25 cm above the center of the view- port The color bar was 14.2 x 2 cm One of the four shapes was centered

3 cm below the color bar The average dimensions of the shapes were 3 x

2 cm In the distractor phase, the stimulus display consisted of a distractor positioned in between two color bars, each depicting the identical hue One

of the four irregular shapes appeared 1.5 cm below the lower color bar The color bars and distractors were positioned on the screen such that the plus sign was always replaced with the distractor stimulus The color bars were 12.2 x 1 cm The color words averaged 9.2 x 2.3 cm; the row of X's was 14.5 x 2.3 cm, Figure 2 shows the arrangement of the stimuli on the computer screen for the distractor phase

Design

Four distractor conditions were employed: incongruent, e.g., the color word PINK surrounded by two brown color bars; congruent, e.g., the color word PINK surrounded by two pink color bars; non-color word, e.g., the word BEST surrounded by two color bars; and rows of X's (XXX condition)

The learning phase consisted of 408 trials Half the trials were distributed equally among the four types of correct color/shape pairs, and the remaining half were distributed among the 12 types of incorrect color/shape pairs The distractor phase consisted of 192 trials The incongruent condition contained 48 trials, the congruent condition contained 16 trials (for a total

of 64 stimulus items containing color words), and the non-color word and XXX conditions contained 64 trials each Half the trials required a correct response and half required an incorrect response

STROOP INTERFERENCE

Figure 2 Stimulus display for distractor phases of Experiments 1 and 2 (subjects were asked to match the previously assigned shape to the color bars while ignoring the word)

peared sequentially on the computer screen for 1 s Subsequently, the first trial began with the presentation of a plus sign in the middle of the screen The subject was instructed to use the plus sign as a fixation point After Is,

a color bar and a shape appeared on the computer screen The subject was required to determine whether or not the shape was the correct associate

of the color bar The subject responded yes or no using a binary button press

If the subject responded incorrectly, the word error appeared on the screen for the first second of a 3-s intertrial interval After every block of 12 trials, the subject was presented with cumulative and block accuracy scores Subjects were instructed to use the error feedback to learn the correct color/shape pairs

Each trial in the distractor phase began with the presentation of the plus sign in the middle of the screen After 1 s, the plus sign disappeared, and

a color/shape/distractor combination appeared on the computer screen The subject was instructed to decide whether or not the color matched the shape, while ignoring the distractor word The subject was instructed to respond as quickly but as accurately as possible, using a button press, as in the learning phase Reaction time was recorded Error messages appeared

as in the learning phase

RESULTS AND DISCUSSION

Error analysis

For each subject, the percentage error rate for each condition was calculated Error rates were 025 for incongruent distractors, 003

201 ELIMINATION OF

for congruent distractors, 005 for non-color word distractors, and 008 for the XXX condition A one-way repeated measures analysis

of variance (ANOVA) was significant, F(3, 51) = 6.52, MS, = 00028,

p = < 01 Bonferroni t tests indicated that the error rate for the incongruent condition differed significantly from the remaining con- ditions No remaining comparisons proved significant

Reaction time (RT) analysis

The mean RTs for correct yes responses for each distractor con- dition were collapsed across subjects (RTs in milliseconds are shown

in Table 1) A one-way repeated measures ANOVA indicated a sig- nificant effect of distractor, F(3, 51) = 11.73, p < 001 Post-hoc tests using the Bonferroni procedure indicated that subjects required sig- nificantly more RT to respond in the incongruent condition (M = 935) than in the remaining distractor conditions The RTs for the congruent (M = 778), non-color word (M = 805), and XXX (M = 785) distractor conditions did not differ significantly from each other Thus, contrary to our hypothesis, incongruent color word distrac- tors inhibited responding relative to non-color words, congruent color words, or rows of X's Using the XXX condition as a baseline, no evidence of inhibition of non-color words nor of facilitation of con- gruent color words was adduced The finding that incongruent color words inhibited responding whereas non-color words produced no inhibition is consistent with the results of other Stroop studies that employ a keypress response mode (Keele, 1972; McClain, 1983a; Redding & Gerjets, 1977) The finding that congruent color words did not facilitate responding relative to an XXX baseline is consistent with the findings of McClain (1983a), but not with the findings of Redding and Gerjets

Informal interviews with subjects were conducted at the completion

of the experiment When asked about the strategies used to perform the task, subjects generally reported that they used actual color names

to encode colors in the experiment Thus, subjects did not appear spontaneously to adopt nonverbal strategies to match colors to shapes Instead, subjects might have used verbal color name codes to mediate Table 1 Reaction times (RTs, in ms) for Experiment 1

ELIMINATION OF STROOP INTERFERENCE

responding in the task As such, incongruent color words delayed target responding Experiment 2 was designed to control the verbal codes that subjects used to encode the color/shape displays

EXPERIMENT 2

In Experiment 1, we reasoned that if the processing of target colors

in the Stroop task could proceed exclusively within the pictorial pro- cessing system, then verbal color name codes activated by incongruent color words could not impede target processing However, in Exper- iment 1, subjects appeared to use verbal, rather than nonverbal codes

to mediate the matching of shapes to colors Thus, Experiment 2 was designed to control the verbal codes that subjects used to mediate the matching process In so doing, we attempted to induce subjects

to use nonsense names to mediate the matching of colors and shapes Experiment 1 relied upon the assumption that pictorial codes differ qualitatively from verbal codes, and are constructed and used by separate processing systems Experiment 2 relies upon the further assumption that distinctions can be made among different processing codes within the verbal system itself We assume that verbal processing codes differ quantitatively according to the similarity of the semantic, phonetic, or orthographic information which they represent (Keele, 1972; Klein, 1964) Our decision to induce subjects to adopt nonsense names as verbal mediates in the matching task proceeded from the assumption that unfamiliar nonsense names share the lowest degree

of semantic, phonetic, or orthographic similarity possible with the verbal color name codes activated by incongruent color words We hypothesized that if subjects could be induced to use nonsense names

to mediate processing of colors, then the verbal codes activated by incongruent color words would not conflict with the verbal color name codes used to match colors to shapes In so doing, Stroop interference would be eliminated

To test this hypothesis, two groups of subjects were recruited Sub- jects in the nonsense name group were trained to encode the colors and shapes using nonsense names as verbal mediators Subjects in the actual name group were trained to encode colors and shapes using actual color names as verbal mediators Subsequently, subjects in both groups matched colors to shapes in the presence of various distractor stimuli

We hypothesized that incongruent color word distractors would pro- duce little or no delay in matching shapes to colors for subjects who encoded the colors and shapes using nonsense names We expected that incongruent color word distractors would inhibit the matching

203

of colors to shapes for subjects who were encouraged explicitly to use actual color names to mediate their responses

Several researchers have examined the effect of modifying the form

of the verbal mediator in the Stroop task Stirling (1979) required subjects to name colors using different arbitrarily assigned letters Using this procedure, incongruent color words still delayed color naming, but to a lesser degree than did incongruent letter strings Glaser and Dolt (1977) asked participants to associate a set of color patches with one set of nonsense names, and to associate the corre- sponding set of color names with another set of nonsense names Subsequently, subjects were asked to provide the nonsense name for the colors of Stroop color words However, the RT to name the color

of incongruent color words was still delayed in comparison to the RT

to name colors of control stimuli (rows of I's)

In studies of Stroop interference in bilinguals (Chen & Ho, 1986; Pyer, 1971; Magiste, 1984; Preston & Lambert, 1969), subjects were asked to use one language to name the color of ink, while the incon- gruent word was printed in either the same language (within-language)

or in the other language (between-language) In general, these studies indicate that within-language Stroop interference is usually greater than between-language Stroop interference Chen and Ho reported evidence that between-language Stroop interference is greater when the two languages involved are similar rather than dissimilar Thus, studies that examine the effects of modifying the verbal mediators employed in the Stroop task suggest that Stroop interference can be attenuated by decreasing the similarity between the verbal color code representations activated by target and distractor stimuli in the Stroop task

The present experiment differs from the studies reviewed above in several ways First, in the present experiment, subjects were not re- quired to name colors; rather, subjects matched colors to shapes and responded via button press, as in Experiment 1 Second, subjects engaged in a greater amount of practice in the use of nonsense names

to match shapes to colors than did subjects in the Glaser and Dolt (1977) or in the Stirling (1979) studies Through practice, we intended

to encourage subjects in the nonsense name group to translate target color information to a nonsense name code, and then to use the nonsense name code to mediate the matching of colors to shapes To the extent that the nonsense name code shared little similarity to the incongruent color name codes activated by incongruent color words,

we predicted no outcome conflict between the matching of colors to shapes and the processing of incongruent color words Thus, in the present experiment, we predicted the elimination of Stroop interfer-