Memory & Cognition1973. Vol. 1. No.2, 106·120

The Stroop phenomenon and its use in the studv ofperceptual, cognitive, and response processes"

FREDERICKN. DYERExperimental Psychology Division. U.S, Army Medical Research Laboratory, Fort Knox, Kentucky 40121

The present review summarizes empirical findings and theoretical views related to the Stroop color-word test.Lxperimental findings were emphasized in contrast to the results of correlational studies, and thebulk of themateriallIas produced since the 1966 review of Jensen and Rohwer. One purpose of the review was to illustrate use of theStroop paradigm as a tool for the study of other psychological processes, The incompleteness. and in some cases theinappropriateness. ofexisting explanations of theStroop phenomenon also were discussed,

In the Stroop color-word test (Stroop, 1935), a largedisruption and delay in color naming occurs for a card ofcolor patches that are shaped to spell incongruent colornames. The most important score on this test is thedifference between the time to name colors on this cardand the time to name colors on a card where the patchesare rectangles. asterisks, or words that are notincongruent color names. Jensen and Rohwer (1966)have provided an extensive review of the Stroopliterature including methodology. research findings, andtheoretical considerations. Much of their review dealswith individual differences in performance as these relateto other performance and personality measures. Thepresent review is concerned primarily with two otherclasses of studies: (I) experiments which were designedto extend knowledge of the Stroop phenomenon itselfand (2) experiments which utilized the phenomenon insome form as a tool to study other problems such asword meaning, semantic satiation, hemisphericdifferences, and bilingual language organization. With afew notable exceptions, these studies have appearedsince the Jensen and Rohwer (1966) review wasprepared, Following those sections of this paper devotedto each of the above topics, a final section attempts tointegrate the various experimental and theoreticalcontributions and thus provide a current perspective onthe locus and mechanism of Stroop interference,

STUDIES OF THE STROOPPHENOMENON

Anomalous situations have long provided much ofman's scientific knowledge, For example, rifts in theearth's crust have been critical for development ofgeology, Similarly, psychopathology has been the basisfor theories of personality, In experimental psychology,a great deal of attention has been focused on opticalillusions in the expectation that their understanding

"The author isgrateful to E. C. Dalrymple-Alford, George S.Harker, and Anne Treisman for their comments on an earlierdraft of thepaper.

would aid our understanding of normal visual andperceptual processes. For this reason, the anomalouscolor naming of the Stroop test should make it animportant means for increasing our understanding of thenormal processes of reading, stimulus identification, andstimulus naming that either disrupt or are disrupted inthe Stroop test. This point of view may not have beenexplicitly stated, but a substantial number ofinvestigations have been conducted to elucidate theconditions that produce the Stroop phenomenon,

Color-Word Configurationsthat Produce Interference

On the critical interference card of the traditionalform of the Stroop test, color patches in the form ofwords denoting color names are printed with the colorof the patch different from the color denoted by theword, A series of patches in different colors appears onthe card, and usually the basic datum is the total timerequired to name the colors of all patches on the card.This format has been highly successful in the productionof interference, i.e., a delay of color naming, andrelatively little deviation from this methodology hasoccurred in the history of the Stroop test. However,onesmall but important departure has been the presentationof single Stroop stimuli and measurement of the latencyof the single color-naming response. This procedureapparently was first utilized by Dalrymple-Alford andBudayr (1966). Sichel and Chandler (1969), using asimilar procedure, found reliable differences in responselatency between conditions with incongruent colornames and with nonword color patches, They alsoincluded congruent combinations of words and colors intheir sequences of individual stimuli and found fasternaming with these than with the incongruentcombinations, The measurement of single responselatencies has allowed a finer analysis of responses and iscritical for some of the uses of the Stroop paradigm instudying other psychological phenomena which aredescribed later in this review.

Both the traditional card of patches and the individual

106

stirn ulus procedure involve the closest possibleintegration of the color and word aspects with actualformation of the word with the color. Other studies haveaddressed themselves either directly (Kamlet & Egeth,1969; Dyer & Severance, in press; Dyer, 1972b) orincidentally (Grand, 1968) with the question of whetherinterference to color naming occurs with otherword/color configurations. Two investigations usednames printed in a neutral color on colored rectangleswith the task being to name the color of the rectangle.For Kamlet and Egeth (1969), the words were printed inwhite and were names of colors that differed from thecolor of the rectangles. Times to name the rectanglecolor for a series of such stimuli were delayed relative tonaming colored rectangles with a series of white Xsprinted on them. In fact, the large interference foundwas almost exactly equal to that found for a traditionalcolor/word condition. Grand (1968) used black wordsthat were not color names and did not find a delay ofcolor naming, whereas the same words when actuallywritten in color did produce interference. This absenceof interference with black backgrounds might suggestthat contrast direction is critical with implications forthe study of visual processing. Grand's negative findings,however, may instead reflect an inappropriate controlcondition in his study. Instead of having Ss name thecolor of colored rectangles without words or withnonsense syllables, his control condition consisted ofnonsense syllables typed in colored ink. Although thiswas an appropriate control for the (typed) coloredwords, the area of the nonsense syllables would be verysmall compared to his colored rectangle stimuli. Thedifference in area could account for the similarity oflatencies for naming colored rectangles with words inblack and color naming with colored nonsense syllables.

One recent study (Dyer & Severance, in press)assessed the degree of interference that words wouldgenerate to color naming when black words werepresented followed by nonword color patches. The mainpurpose was to develop a paradigm that would allowassessment of interference effects with stimulusdimensions that could not be combined with words.Such a paradigm would allow comparisons betweendimensions of the degree that they are subject to Stroopinterference. Significant delays in color naming of morethan 40 msec occurred when incongruent color names inblack preceded a series of colored Xs. The comparisoncondition was a series of black Vs that preceded thecolor stimulus.

Dyer (l972b) also presented black words and coloredXs to Ss, but instead of successive presentations, theword and color were symmetrically spaced on either sideof a fixation point with the side of the word varyingrandomly from trial to trial. Exposures of 100 msecwere too short to permit eye movements to relocatethese stimuli on the retina. Ss were instructed to ignorethe word and to name the color quickly. Times fornaming colors when the opposite-side word denoted a

THE STROOP PHENOMENON AND ITS USE 107

color different from the color patch averaged nearly50 msec longer than times for a condition where blackXs were paired with the color.

One new version of the Stroop test that might bereferred to as a super-Stroop test because of its highinterference was developed by Daniel (1969). Not onlywere words printed in incongruently colored ink, theywere printed on colored backgrounds that wereincongruent to the color name and different from theincongruent color of the word. The task was to namesuccessively first the color of the background and thenthe color of the word, repeating this doublecolor-naming task for each stimulus on the card. Timesfor this task were more than twice the time to name thecolors of ink on the traditional interference card.

One presentation procedure that has generally failedto produce interference to color naming is an auditorypresentation of the incongruent word stimuli inconjunction with nonword color patches. Thackray andJones (1971) presented individual color patches withsimultaneously spoken words as one of their conditionsand found no delay in color naming compared to acondition without the auditory stimuli. Such auditorystimuli did not increase the interference to color namingeven when the colors were incongruently colored colornames. Simultaneous auditory presentations may notprovide the critical timing (see Timing section below) ofword and color processing that is necessary for Stroopinterference to occur. To test this, Dyer and Severance!presented spoken incongruent color names at severalintervals slightly prior to, during, and after presentationof a rectangular color patch. None of these conditionsproduced more than a trivial increase in the time toname colors over control conditions utilizing spokenneutral words. One implication of this finding (little orno interference from spoken words) is that reading iscritical to the generation of naming interference. In themotor theory of speech perception (Liberman, Cooper,Harris, MacNeilage, & Studdert-Kennedy, 1967), anactive response is claimed to be made by the perceiver'sown vocal apparatus when he perceives a spoken word.The failure of spoken words to interfere with colornaming suggests that this active response bears littlesimilarity to an implicit reading response, if it exists atall.

A number of other variants of the Stroop test haveused the conventional word-shaped color patch, withresponses other than overt color naming. TecceandHapp (1964) required card sorting on the basis of colorwith Stroop stimuli printed on the cards and found thiscondition to require much more time than was requiredto sort cards with simple rectangles of color printed onthem. Pritchatt (1968) required a keypress response withkeys corresponding to the colors. Stroop stimuliproduced longer latencies for these responses thanrectangular color patches, and the effect was greatestwhen the keys were labeled with words. With coloredrectangles designating the keys. response times were ul1l~

108 DYER

slightly longer for Stroop than for control stimuli. Thisfinding suggests that the Tecce and Happ (1964)card-sorting task would have produced even moreinterference had they labeled their sorting boxes withwords instead of colored rectangles. Treisman andFearnley (1969) required the sorting into same anddifferent piles of cards that carried two stimuli-twowords, two colors, or a word and a color. Word-wordand color-color matches were very fast with little or nointerference, even when one of the words (word-wordmatching) or colors (color-color matching)was a Stroopstimulus, i.e., both a word and a color. When neitherstimulus was a Stroop stimulus, word-color andcolor-word matching were slower than word-word andcolor-color matching. A further large delay in matchingoccurred when one of the stimuliwas a Stroop stimulus.Dyer (1971d) used similar stimuli in word-color andcolor-word matching tasks with verbal same anddifferent response times recorded individually. He foundlarge delays in such responses when "irrelevant"incongruent word and color information was present.Hock and Egeth (1970) required judgments of whetheror not a stimulus color matched a predeterminedmemory set of colors and found that such matching wasdisrupted by Stroop stimuli.

Differences in interference from Stroop stimuli inthese different tasks aid our understanding of the Stroopphenomenon, and these resultswillbe returned to in thefinal section of this paper. Since some of these tasksrequire additional perceptual, cognitive, or motoractivity beyond the color naming of the traditionalversion of the Stroop test, they should be substituted forit with caution.

Asymmetries in Responseto theDual Aspects of the Stroop Stimulus

Since the word aspect of the Stroop stimulus stronglyaffects naming of the color, it is logical to assume that asimilar interference might occur for the reading of thewords as a result of the presence of colors. Stroop(1935) examined this possibility and found that readingwords on the interference card (words in incongruentcolors) was delayed by less than 6% relative to readingthe words on a card with words written in black ink.This difference is sufficiently small that it could be onlythe result of the reduced legibility of the words on theinterference card. The legibility reduction might resultfrom reduced luminance contrast and perhaps fromblurring of one color relative to another because of thechromatic aberration of the eye. However, Stroopprovided evidence that real interference from the colorscan be generated to word reading. He found thatextensive practice on color naming caused a substantialincrease in time for reading words on the interferencecard in a word-reading session that followed thiscolor-naming practice. This effect declined very quickly,however, with reading times almost back to the levels

that existed prior to color-naming practice on the secondtesting of word reading following that practice. This isquite unlike the consistently high interference fromwords to color naming that Stroop found throughouttrials on 8 successive days of strictly color-naming trials.It must be remembered, however, that much otherword-reading practice must have occurred for hisundergraduate Ss.

Gumenik and Glass (1970) appear to have made asimilar finding of interference to word reading fromcolors; however, their procedure did not controladequately for legibility differences that would beexpected between their color-word card and a card withblack words. Their rationale was that reducing the wordlegibility would weaken the reading response and wouldmake this weakened response susceptible to interferencefrom color naming. Besides the problem of a differentialreduction of legibility that would be expected for thecolored and black words, they alsohad all their Ss namecolors just prior to word reading on the interferencecard. The above-mentioned results of Stroop (1935),showing the effect of color-naming practice onsubsequent word reading on the interference card,indicate that this may have produced genuineinterference but of a somewhat artifactual type. Dyerand Severance (1972) attempted to replicate theword-reading portions of the Gumenik and Glass (1970)study with inclusion of two additional controls: (I) aneutral-ward-reading condition which would(presumably) not be subject to interference from colornames and (2) a series of graded achromatic shades forthe control-word-reading card instead of black words ofconstant high contrast. Apparently because of someunknown procedural difference, the word-readingconditions were not delayed to the extent that wordreading was delayed in the Gumenik and Glass (1970)study. However, despite this failure to "weaken" thereading response to the same extent, colors did appear tointerfere, since a significance increase of nearly 20%wasnoted in the time to read color words when they werepresented in incongruent colors over the time to read thesame words when they were printed in shades of gray.This was true despite a longer time required to readneutral words when they were in gray than when theywere in color. It was true also despite a completeabsence of color-naming trials prior to the word-readingconditions for the Ss. Thus, despite a poorly controlledstudy, the conclusions of Gumenik and Glass regardingword reading appear valid.

Another study (Uleman & Reeves, 1971) also claimeda small but significant amount of "reversed" Stroopinterference. They compared scanning for a particularcolor name when the words on a card were in black andwhen they were in incongruent colors. Lund (1927)showed that scanning for words takes almost twice aslong as scanning for a particular color. Uleman andReeves hypothesized that this faster response to colorthan to words for this scanning task would lead to delays

in scanning for a particular color name when competingcolors were present on the interference card. They didfind that scanning for a word was slower on thecolor-word card than on a card with black words, butthe small difference could have reflected a legibilitydifference as much or more than interference, since theirstudy lacks the same controls as that of the study ofGumenik and Glass (1970). However, Dalrymple-Alfordand Azkoul (1972) report that Dalrymple-Alford, in anunpublished study, replicated the Uleman and Reevesresult with materials that controlled for this legibilitydifference.

Generality of the Stroop Phenomenon

Closely related to the asymmetry in response tothe two aspects of the Stroop stimulus is thequestion of the extent to which naming of otherstimulus dimensions besides color is susceptible tointerference from incongruent dimension namesintegrated with levels of the dimension. Smith and Borg(1964) were interested in producing a parallel form ofthe Stroop test which would allow retesting of Ss toincrease the reliability of their serial scoring technique.They used achromatic shades (white. gray, and black) inincongruent combinations with their correspondingnames. They found this version much less reliable thancomparable color versions and generally unsuitable fortheir purposes, Dyer (1971a) used similar stimuli andfound that interference (time for naming withincongruent word stimuli minus the time for namingwith control stimuli) was about 409C less than that for acomparable color version of the test. However, scores onboth versions correlated at a high level, indicating thatdespite the reduced interference both measuredessentially the same thing.

White (1969) generated an interference paradigmanalogous to the Stroop test by presenting one of thewords north, east, south, and west within a square sothat the word's position was incongruent to the positiondenoted by the word (e.g., the word south at the top ofthe square). The word positions for a series of suchincongruent position-word stimuli were named. and thetime for this naming was found to be 20% greater thanthe time to name the position of nonsense syllables.Although significant, this interference to positionnaming was much less than the 609C increase incolor-naming time for incongruent color names overnonsense syllables found for the other groups in theexperiment. Shor (1970) also explored an analog of theStroop test in which spatial direction was combined withwords denoting spatial directions. A series of arrowspointing up. down. right, or left, each with a worddenoting one of the other directions written within it,were presented in the traditional Stroop format. i.e"cards of arrows with words written within them and acontrol card of arrows without such words. Time toname arrow directions was about laC;;- greater for the

THE STROOP PHENOMENON AND ITS USE 109

interference card than for the control card, Thisdifference was significant. but it is much less than the50%-100% increases found for the comparable colorcards. In a later study, Shor (1971) compared a greatmany different tasks requiring responses to stimulicombining two aspects with at least one aspect normallya word, letter, or digit. Besides color and arrowdirection, he used spatial position and a dimension ofstraightness-crookedness. Naming of all these dimensionswas subject to interference, but only a small amount incomparison to that for the color stimuli.

In a similar paradigm to that of Shor (1970), Dyer(1972a) combined direction names with the dimensionof movement direction by moving the words up, down,right, and left and also a series of four Xs in one of thedirections up, down, right, or left on the oscilloscopeface of a laboratory computer. Some facilitation ofdirection naming occurred for conditions wheredirections and words were combined congruentlyrelative to the control (Xs) condition, but only a verysmall amount of interference was found for namingdirections with the incongruent combinations. Theinterference effect was less than a fourth of that foundfor comparable color work with individual Stroopstimuli (Dyer, 1971c). The results were thus similar toShor's (1970) finding of only low interference despitethe closer integration of words and directions with themoving word display relative to his words written withinarrows. Low interference prevailed even with astationary preexposure of the word prior to itsmovement, which had been expected to increaseinterference (see Timing below).

The general failure of dimensions of arrow direction.spatial position, achromatic shade, movement direction,and form to provide a strong analog of the Stroop testsuggests that processing of color information is in someway different from these other dimensions. Cramer(1967) found that it was more difficult for preschoolchildren to name colors with the patches a mixture ofnameable forms than it was to name forms with theforms printed in different nameable colors. The samestimulus plate was used for both tasks. If the formdominance that this task apparently illustrated in thesechildren occurs throughout life, her result suggests thatnames are assigned with more difficulty to colors than toforms and perhaps to other dimensions as well. This mayaccount for higher interference from irrelevant words tonaming the dimension of color than to naming of otherdimensions.

Windes (1968) perhaps first showed the potential ofnumerals to interfere with counting when the numeralswere the counted objects. Many of the tasks reported byShor (1971) also involved numerosity responseswith theobjects enumerated being a group of digits with thedigits incongruent to their number. Numerosity does notintuitively appear classifiable as a stimulus dimension.but substantial interference to numerosity namingoccurred in these tasks. Even more removed from a

110 DYER

stimulus dimension are letters and numbers. the namingof which was considerably delayed (though not to theextent of naming colors) when the letters and numbersto be named were used to outline other letters andnumbers. The stimulus dimension for these latter tasksmight be regarded as form.

Morton (1969) also obtained data on a set of taskslargely involving enumeration, Or perhaps more correctlysubitizing, of various letter and digit stimuli. He foundconsiderable interference from the irrelevant digits andletters. These results were conceptualized as equivalentor at least closely related to the Stroop test results.Morton also found interference from spoken digits in theenumeration of visually presented stimuli and also theenumeration of a series of distinct auditory events. Thisis in contrast to the previously discussed failure ofspoken words to interfere with color naming (Thackray& Jones, 1971).

The fairly high interference found by Windes (1968),Morton (1969), and Shor (1971) for enumeration taskssuggests some comparability between color andnumerosity dimensions. As presently conceptualized,these dimensions are quite unrelated if numerosity canbe classified as a stimulus dimension at all. Sameness anddifferentness is another such "dimension," the namingof which Egeth, Blecker, and Kamlet (1969) showed canbe interfered with to much the same extent as color andnumerosity. They found that judging whether tworectangles of color were the same or different wasgreatly slowed when words spelling the words same anddifferent were written on the rectangles.

One other failure to generate an analog of Stroopinterference should be mentioned. Dyer and Mosk0 2

required Ss to name voices with the single word voicedon a trial being the name of another speaker whose voicewas used in the experiment (the control condition was aneutral word). This task reproduces in the auditorymode the conditions of the Stroop stimulus. Voice andirrelevant word are closely integrated, and both arecompletely auditory in nature. Times for naming voicessaying incongruent or congruent voice names werealmost identical to times for naming voices sayingneutral word stimuli.

The failure of these conditions to change namingtimes for an auditory dimension appears to contrast witha recent study by Hamers and Lambert (1972). Theylooked at times to name the pitch (high or low) of avoice saying the words high or low and found over100msec difference between congruent and incongruentpitch-word combinations. Unfortunately, no controlwords were included to determine whether the effectwas interference from incongruent combinations orfacilitation of pitch naming with congruentcombinations. One other difficulty for interpretation oftheir results was a 30% rate oferrors for the incongruentpitch-word combinations, whereas only 6% of thecongruent combinations were erroneously named. Thisdifference indicates a strong tendency for the Ss to

simply echo the irrelevantword aspect of the pitch-wordstimulus. Such "correct" responding with congruentcombinations might account for their faster responsetimes. Further research is required to establish thatauditory dimensions can provide a naming task withinterference from irrelevant words which is analogous tothe Stroop test. Such an effort would be highlyworthwhile, since an auditory Stroop paradigm could beused to study speech recognition and other aspects ofauditory perception.

Timing

Manipulation of the processing time for color relativeto words can greatly reduce interference to color namingfrom irrelevant incongruent words andmay also produceincreases in such interference. In the portion of theirstudy concerned with color naming, Gumenik and Glass(1-970) showed that reducing the legibility of theincongruent names bya mask that minimally changedthe visibility of the colors greatlyreduccd theinterference of the incongruent color names to colornaming. Dyer (1970) manipulated the backgroundluminance of constant luminance color words and foundhigh Stroop interference when, as a result of thismanipulation, color naming was slow and word readingwas fast. Interference was generally least for theconditions of minimal luminance contrast between thecolored word and its white background. One problem ofthe study was that the background manipulationproduced larger changes in color visibility despiteconstant color luminance than the changes in wordvisibility (color naming was delayed more than wordreading). This concurrent variation of color and wordprocessing rates allowed only a correlational analysis ofthis relationship of Stroop interference to theseprocessing rates.

Klein (1964), in a little-discussed second experimentof his famous paper, had Ss read the word then name thecolor on the interference card and found that times forthis double-response condition wereonly slightlygreaterthan for a condition with color naming alone on thiscard. In addition, Ss reported that the task involved littleof the strain that is characteristic of the performance onthis interference card (e.g., Jensen & Rohwer, 1966).Taken together, the results of the Gumenik and Glass(1970), Dyer (1970), and Klein (1964) studies wouldpredict that some optimal processing rate for colorsrelative to words would maximize interference 10 colornaming from the incongruent words. Relatively fastcolor processing would be expected to produce theGumenik and Glass (1970) condition of lowinterference. Fast word processing relative to colorprocessing would also be expected to reduce interferencebecause the reading response is "gotten out of the way"as it was for Klein's (1964) dual-response condition. Totest this prediction, Dyer (1971 c) used an individualstimulus technique and preexposed words in black for

various intervals prior to coloration, presumablyadvancing word processing relative to color processing.Color-naming latencies were measured from the time ofcoloration of the black word. These latencies increasedslightly then decreased sharply as black wordpreexposures varied from 0 to 500 msec. Maximuminterference occurred at 40 msec preexposure with asharp drop in interference for preexposures greater than60 msec. Some interference to color naming occurredeven with the longest preexposures. Congruentcombinations of words and colors were included toprevent the word from serving as a cue to reduce theresponse set.

The importance of relative processing rates of wordsand colors suggested an explanation of the reducedinterference found by Dyer (1971a) for achromaticshades compared to a chromatic version of the Strooptest. Control achromatic shade rectangles were namedslightly faster than control colored rectangles, and wordprocessing times were approximately equal for the twoconditions. It was thus assumed that the Gumenik andGlass (1970) low-interference condition of fastdimension processing relative to word processing alreadyexisted not as a result of slowing word processing,but asa result of inherently faster processing of achromaticshades. Given the apparent success of this explanation, itwas assumed to also account for the even lowerinterference conditions found for direction naming byShor (1970) and the similar results of pilot work for astudy (Dyer. 1972a) with moving word stimuli on thecomputer oscilloscope. It was assumed that thedimension of direction was processed even more rapidlythan achromatic shade and that this was the basis of thelow interference for direction naming. A similarexplanation would probably apply to all noncolordimensions producing low interference. To test thisexplanation, an analogous experiment to the study ofcolor naming with black preexposures (Dyer. 1971c) wasconducted. Word stimuli were presented for variousintervals prior to movement on the center of theoscilloscope face, then moved in one of four directions.It was assumed that some stationary preexposure of aparticular length would maximize interference and. infact. bring this interference to the levels obtained inexperiments with color. The results of two experiments(Dyer. 1972a) indicated that. although stationarypreexposures of the word for 200 msec or more reducedthe small amount of interference to even lower levels.not one of eight preexposures between 0 and 200 mseccaused any appreciable increase in interference todirection naming. Ironically. this apparentdisconfirrnation of the hypothesis that the processingrate of a dimension relative to words is the criticalfeature is itself contradicted by data from the samestudy that indicated that interference to directionnaming was proportional to the amount of time it tookto name a particular direction. Direction naming incontrol conditions was fastest for up. followed by dnwn.

THE STROOPPHENOMENON AND ITS USE III

left. and then right. Correspondingly, interference todirection naming was least for up, followed by down,left, and then right. Some validity to the hypothesis of acorrelation of high interference with fast wordprocessing relative to dimension processing thus may stillexist, although stationary preexposures of the word donot produce the "fast word processing relative todimension processing." The result argues that the studyof color naming with black preexposures (Dyer, 1971c)may not be a true study of the effects of differentrelativeword and color processing rates either.

Sequence Effects

Although the individual stimulus presentation methoddoes provide important benefits for the analysis ofStroop processes and allows applications of it for studyof other variables, the amount of interference isconsiderably higher for the more traditional versionwhere total naming time is obtained for a card ofincongruently colored color words. Response times forthe incongruent name condition in individual stimulusversions are usually less than 25% greater than forcontrol conditions. whereas interference card times intraditional versions are usually from 50% to 1000/C longerthan color-naming times for the card of control patches.At least two factors appear to account for the greatertimes with the cards of multiple stimuli. One is a delayof responding to a color stimulus produced by responsecompetition due to the presence of other color stimuli.The other is the effect of suppression of the irrelevantword response and color naming for one Stroop stimuluson the suppression and naming responses to the nextstimulus. Sichel and Chandler (1969) found that colornaming for the first of two stimuli took over 200 mseclonger than color naming for a single Stroop stimulus.They attribu ted this increased response latency tocompeting response tendencies generated by thepresence of the second Stroop stimulus. Anotherpossible explanation is that Ss adopted a strategy ofprocessing both stimuli prior to responding to either,since they were instructed to name both. If they hadbeen instructed to respond to only the first of the twostimuli, they might not have shown this delay. Somesupport for this comes from the fact that therelationship of the second stimulus characteristics tothose of the first was a critical factor in the amount oftime required to name the first stimulus. In fact. theylater discussed this delay of naming with two stimuli asreflecting the effects of responses to the firststimulus-both suppression of the irrelevant response tothe word and making the appropriate response to thecolor-on responding to the second of the stimuli. Thiswould seem to imply a fairly high level of processing ofboth stimuli prior to the response to the first colorstimulus.

The most protracted color-naming response in theSichel and Chandler (1969) study occurred for a pair of

112 DYER

stimuli in which the color of the first corresponded tothe word of the second and conversely the word of thefirst corresponded to the color of the second (e.g.. theword red in green followed by the word green in red).For such a combination. the first suppressed response tothe irrelevant aspect becomes the appropriate responseto the second and the recently completed response is theresponse that must be suppressed on the next stimulus.Because of the high level of interference found for this"suppress-say/say-suppress" combination. a usefulversion of the Stroop test was created by these authorswhich was made up largely of such pairs of stimuli.

Dalrymple-Alford and Budayr (1966) also exploredthe effects of particular sequences of Stroop stimuli oncolor-naming times. Traditional versions of the Strooptest such as the version recommended by Jensen andRohwer (1966) present the series of stimuli on the cardsubject to the constraints that succeeding words andsucceeding colors differ. Dalrymple-Alford and Budayrstudied the effects of three different sequentialconditions. In one. the colors differed on succeedingstimuli and. in addition, the color nevercorresponded tothat denoted by the previous irrelevant word. For a"say-suppress" condition, the irrelevant word alwayscorresponded (after the first stimulus) to the precedingcolor. For a "suppress-say" condition, the color alwayscorresponded to the preceding irrelevant word. Unlikethe latter two conditions. the first does not involveeither saying a previously suppressed word orsuppressing a previously said word. The "say-suppress"condition did not differ from the condition wheresucceeding stimuli bore no relationship betweenirrelevant words and colors. On the other hand, the"suppress-say" condition required 20C7c more time forresponding than either of these two.

A combination condition, Le., suppress-say!say-suppress. such as that used in the stim­ulus pairs by Sichel and Chandler (1969) cannoteffectively be extended in a pure form for more thantwo stimuli. since there would only be two stimuli whichwould alternate. In this respect, even theDalrymple-Alford and Budayr conditions are somewhatsuspect. On cards utilizing only the "say-suppress"condition or the "suppress-say" condition, the S canrespond correctly to the colors by reading the words.This reading response is correct for the preceding colorstimulus in the "say-suppress" condition and for thefollowing color stimulus in the "suppress-say" condition.However, no indication of Ss' adopting such a strategyoccurred for the Dalrymple-Alford and Budayr study.

USE OF THE STROOP PHENOMENONTO STUDY OTHER PROBLEMS

The studies and their findings described in this sectionof the review were primarily concerned with otherpsychological processes and used the Stroopphenomenon as a tool to explore these processes. Much

information about the Stroop phenomenon itself isavailable in these results. however, and many of themcould as easily have been included in the earlier section.The present section is demarcated from the first in anattempt to orient the reader to the rich potential thatthe Stroop phenomenon offers for studying otherprocesses.

Interference to Color Namingas an Index of WordMeaning

Perhaps the most important Stroop paper sinceStroop's originalwork was a study of the late G. S. Klein(1964). Using a traditional format of cards of colorpatches, he illustrated that interference to color namingappeared for words other than incongruent color names.His six different word types consisted of nonsensesyllables. rare words, common words, words such asgrass and sky that "implicate the colors in theirmeaning," color names from a set of colors differentthan those being named, and the traditional incongruentcolor-name plate. The percentage increase in namingtimes for these different word types over a controlcondition where the color patches were asterisks was12% for nonsense syllables, 17% for rare words. 27% forcommon words, 35% for color-related words. 41% fordistant color names. and 85% for the traditional closecolor-name condition. These results clearly indicate thepotential of color naming of word-shaped color patchesto provide an objective indication of differences in themeaning of words, even though "meaning" isconfounded with color-relatedness. Both the differencesillustrated to exist between classes of words unrelated tocolor and the differences between classes of words whichare color-related have general importance for theories ofword meaning.

Scheibe, Shaver. and Carrier (1967), in a very similarstudy to Klein's. showed that interference to colornaming was directly related to the frequency with whichthe words used to interfere were associated with thecolors to be named. It is interesting that in both studiesthe close color-name condition provides over twice theinterference of the distant color-name condition.Hochman manipulated the rate at which individualcolored word stimuli from different word classes werepresented and recorded as the dependent variable, thenumber of erroneous responses in adults (1967) andyoung Ss (1969). In both studies, he found that whenthe words were incongruent names, they produced themost errors with the error rate dropping systematicallyas the words changed through the Klein categories tononsense syllables. More recently, Fox, Shor. andSteinman (1971) showed that Klein's basic finding alsoapplied to other dimensions besides color. They foundthat spatial direction and numerosity naming weredelayed as the semantic aspects of the stimuli rangedfrom incongruent names to nonsense words and figures.

Dalrymple-Alford 0972a) used an individual stimulus

procedure and also showed that words related to colorsdelayed color naming more than unrelated words. Aunique aspect of this study was the inclusion ofcolor-related words with the related color congruent tothe word's color. Color naming for these stimuli wassignificantly faster than for unrelated words. In anotherstudy. Dalrymple-Alford (l972b) showed that soundsimilarity between words and color names matchessemantic similarity in the production of delays of colornaming. Words with similar first sounds (e.g .. run, blot,grown) or similar last sounds (e.g.. bed, true, queen) tothe names of the colors produced more interference thanunrelated words when used as Stroop stimuli.

Bakan and Alperson (1967) investigated bothmeaningfulness and pronounceability and claimed thatboth were important factors directly related to theamount of interference that word stimuli generated forcolor naming. Their results were unambiguous for themeaningfulness dimension. but due to confounding ofmeaningfulness with pronounceability, the role ofpronounceability is somewhat in question.

Ellison and Lambert (1968) used the Stroopphenomenon to study semantic or verbal satiation(Kanungo. 1967) in which word meaning is decreased oreven eliminated by continuous repetition or viewing of aword. Ellison and Lambert hypothesized that if thecolor names were subjected to the satiation procedure,they would produce less interference on a subsequentadministration of the Stroop test. They found a verysmall reduction of interference following satiation ofcolor names relative to a condition with satiation ofother words. The effect was stronger for the first half ofthe card of colored color names. This suggests that astudy of this type using the individual stimulusprocedure might more adequately assess the effects ofsemantic satiation on color naming. In fact. semanticsatiation may be one possible interpretation of thereduced interference in the study in which words werepreexposed in a neutral color prior to coloration (Dyer.197Ic).

Interlingual vs Intralingual Interferencein the Stroop Test

Dyer (1971 b) showed that English-speakingrnonolinguals showed interference to a form of theStroop test using translation equivalents of incongruentcolor names as interfering stimuli with the amount ofinterference directly related to the similarity of theseinterfering words to their English equivalents. Prestonand Lambert (1969) and Dyer (1971 b) both showedthat for bilinguals. the condition where the interferingwords were in one of the bilingual's languages and thecolor naming was to be done with the other languageproduced a great amount of interference. althoughsignificantly less than when color naming and interferingwords were both in the same language. The highi n t e r l i n g u a l interference levels Sllpp,\rt

THE STROOP PHENOMENON AND ITS USE 113

Dalrymple-Alford's (1968) position that bilinguals donot "turn off' or even greatly attenuate one languagewhile speaking in the other. However, the significantlysmaller interference in the interlingual conditioncompared to the intralingual condition does appear toargue for some separation between languages forbilinguals other than the differences between individualwords which exist as well for different words of thesame language.

Assessment of theContents and Structure of Memory

A very interesting application of the Stroop task wasrecently made by Warren (1972). He determined thestrength of word representations in memory bymeasuring the amount of interference such wordsgenerated to color naming when they were used as colorpatches. He used words that were unrelated to color andintroduced the word to memory at various periods priorto its use in the color-naming task. Presentation of thewords was auditory and was found to greatly increasethe word's interference to color naming when it waslater written in colored ink. Such introduction alsoincreased the interference potential of the categoryname to which the word belonged. Warren also madeimportant initial steps to assess changes over time in thestrength of these word representations in memory.Although time and trials were confounded, the strengthof the word in memory was shown to decline linearlywith 1-31 sec of storage prior to its use as a Stroopstimulus.

Subliminal Perception

A great deal of controversy has been generatedregarding whether word stimuli can be responded toalthough they are exposed so briet1y that they cannot berecognized. Following the format of the study (Dyer &Severance. in press) that produced interference to colornaming with black words followed by colors, Severanceand Dyer (1972) presented black words for durationsthat were too short for recognition and followed thiswith 50 msec supraliminal colored Xs. Times to namethe colors did not differ as a function of whether thesubliminal word stimulus was an incongruent colorname. congruent color name. or a series of Vs. The studywith supraliminal presentations of the words hadproduced over 40 msec of delay between theincongruent and control conditions, and facilitation of10-20 msec occurred for color naming in the congruentcondition. This was in marked contrast to the subliminalstudy where a nearly total absence of differencesbetween any important conditions occurred.

Hemispheric Processing Differences

vlilne r (I l)71) has discussed processing advantages

114 DYER

that accrue to particular dimensions when they arelocalized to one hemisphere instead of the other. Notonly does there exist the well-known superiority of thedominant hemisphere for processing verbal materials,but the nondominant hemisphere processes certainnonverbal dimensions better than the dominanthemisphere. If color were one such dimension that wasbetter handled by the nondominant hemisphere, onewould predict that stimulating this hemisphere alonewith a Stroop stimulus would reduce the interference.Presumably. the color would be processed more rapidlythan the word by this hemisphere and a strongrepresentation of the color would be transmitted back tothe dominant hemisphere for naming,alongwith a weakor delayed representation of the word. On the otherhand. when the Stroop stimulus was sent only to thedominant hemisphere, the strong representation of theirrelevant word would be expected to interfere greatlywith the color representation which might requiretransmission across the corpus callosum to "its"hemisphere and back. Even if color was not a dimensionfavored by the nondominant hemisphere, one mightpredict that the more "robust" representation of theirrelevant word when the Stroop stimulus wastransmitted to the dominant hemisphere would result inmore interference for this condition.

To test these hypotheses, both predicting moreinterference from presentations of Stroop and controlstimuli to the right visual field than to the left withright-handed Ss, Dyer and Harker' used verticallywritten stimuli presented briefly on one or the other sideof the visual field. Stimuli were written vertically tocontrol for peripheral location, and the brief exposurewas to prevent eye movements from centering them.Although considerable interference occurred in each oftwo studies which differed in the peripheral location ofthe stimuli (45 min vs 2 deg), no difference in responsetimes or in interference appeared between the differenthemiretinal presentations in either study.

Some promise for the existence of a Stroop paradigmthat will differentiate hemispheric function was shownby the presentation procedure described earlier (Dyer,1972b) in which separate words in black and coloredrectangles were presented bilaterally with the side of thecolor (and word) varied randomly from trial to trial.This procedure showed considerable interference, butfor the color red this interference was much greaterwhen the word was on the right and the color on theleft. To some extent, this was also true for the colorblue, but it was not true for green in the first experimentwith three colors or for green and yellow in the secondexperiment with four colors. The result for red appearedwith considerable strength in each of the experiments.Further research is under way to determine if thisparadigm can reliably indicate functional differencesbetween hemispheres.

PROPOSED MECHANISMS AND LOCIFOR THE STROOP PHENOMENON

The Stroop phenomenon has been "explained" bymany different investigators over the past years. Thiscontinuing tendency to provide newexplanations and torepeat old ones suggests that the various explanationshave been somewhat less than adequate. Most. of theseexplanations have considered the phenomenon in termsof response competition with a stronger reading responseto the irrelevant word aspect of the stimulus dominatingand delaying the color-naming response. However. thesesimilar explanations often differ in their accounts of thediffering strengths of reading and naming responses.Pro b ab ly the most notable exception to theresponse-conflict explanation was that of Hock andEgeth (1970), who claimed the interference resultedfrom different color encoding for incongruentlycolored color words than for control color stimuli.Dalrymple-Alford and Azkoul (1972) have recentlyshown the inappropriateness of this explanation. Theirarguments will be reiterated in this section, and furtherdiscussion of the inappropriate conclusions of the Hockand Egeth (1970) work will be presented. Followingthis, the various versions of the response-conflictexplanation will be discussed, including one (Treisman &Fearnley, 1969) which the authors deny fits thiscategory. Finally, a new explanation will be presentedinvolving both response competition and a failure ofselective attention. Prior to these discussions. however. aseries of studies utilizing colored incongruent colornames as stimuli are described. These studies generallyinvolve responses to the color aspect of the stimuli otherthan naming or categorizing on the basis of names andshow little or no interference from the irrelevant wordaspect. These studies provide strong evidence for theimportance of response conflict in the Stroopphenomenon.

TasksWhere Irrelevant Wordsare Actually Irrelevant

Derks and Calder (1969) required Ss to count thenumber of times a particular color appeared on a card ofcolor patches. No differences in counting times appearedbetween a card of color patches that were Xs and a cardwhere the patches were incongruent color names. Egeth,Blecker. and Kamlet (1969) required their Ss to indicatewhether the colors of a pair of color patches were thesame or different and found no difference betweenconditions where the pairs of patches were embossedwith white Xs or with white incongruent color names.Pritchatt (1968) found that keypress responses toincongruently colored color names showed very littleinterference from the irrelevant words when the keyswere labeledwith colors. A similar unpublished study by

Azkoul, reported by Dalrymple-Alford and Azkoul(1972), showed that with repeated practice, Pritchatt'sSs would probably have eliminated the small amount ofinterference found with his keypress task. TreismanandFearnley (1969) required sorting of cards with two colorpatches on the basis of whether the two colors were thesame or different and found that this wasextremely fastdespite the fact that one color patch was a Stroopstimulus.

None of the above tasks required overt color naming,but under certain conditions, even a task that requiresnaming will not be interfered with by irrelevant words.Derks and Calder(1969), in a second study. required theS to name a single target color as well as to count it andshowed this too produced no difference in response timebetween XS and Stroop stimuli. Uleman and Reeves(1971) found that times to scan for a single color on acard of Stroop stimuli, both checking and naming itupon detection. did not differ from times for suchscanning. checking, and naming with control colorstimuli. It appears in all these studies that the S isattending to "the rapidly formed but nondurablestimulus representations formed by the sensing stage ofperceptual processing [Hock & Egeth, 1970, p. 300}."Even when naming was involved, the single namingresponse for the single target stimulus eliminated anyneed for selectionof such a response and could probablybe produced automatically upon reception of a"sensing-stage" cue.

The Case for a PerceptualEncoding Explanation

Hock and Egeth (1970) used the Sternberg (1969)paradigm in which reaction time is determined forclassification of stimuli as membersor nonmembers of apreviously learned target set to study the encoding andclassification of colors presented as verbs. incongruentcolor names. or a series of Xs. The memory sets wereone, two. or three colors defined by name prior topresentation of the series of color stimuli. Sternberg(1969) has shown that variables which affect theencoding of the stimuli lead to differences in reactiontime that are constant for different memory set sizes. Onthe other hand, variables which affect the more centralcomparisons with the memory set produce curves ofdifferent slopes when reaction time data is plottedagainst memory set size. As Dalrymple-Alford andAzkoul (1972) point out, since the memory set isprobably color names. the encoding of the stimuli formaking the memory match would involve covertgeneration of the names of these color stimuli. Becauseof this, response competition effects from irrelevantwords which delay this encodingwould not be expectedto have any different effect depending on memory setsize and no interaction of Type of Verbal Material bySize of Target Set would be expected. Dalrymple-Alfordand Azkoul's arguments indicate that it is inappropriatefor Hock and Egeth (1970) to conclude that differences

THE STROOP PHENOMENON AND ITSUSE 115

between word types represent differences in speed ofperceptual encoding of the color component for thethree types of stimuli unless they extend this encodingto include covert naming of the color stimuli. Hock andEgeth do distinguish between low-level color encodingwhich is not delayed by irrelevant words combinedwiththe stimuli and "high-level" encoding which is sodelayed. They thus appear to have provided a valid anduseful distinction but do not admit that thecharacteristic of high-level encoding is that covert orovert word responses are generated, as Dalrymple-Alfordand Azkoul have convincingly argued.

In their conclusion that type of verbal material doesnot interact with size of the memory set, Hock andEgeth (1970) also fall into an error that Sternberg(1971) has warned against-using statistical testsdesigned to reject the hypothesis of no interaction as abasis for accepting the hypothesis that there is nointeraction. The error is to assert a lack of interactionbetween variables when the insignificance may merelyreflect imprecision in the experiment that preventedachievement of significance. Their Type of VerbalMaterial by Size of the Target Set interaction producedan F of 1.58, which, with df =4 and 60, would occuronly one in five times by chance. Their probableinappropriate rejection of the existence of thisinteraction does not bear on the verbal conflict thataccounts for longer response latencies withincongruently colored color words, but it does maskwhat may be an even more interesting finding. Thememory set may be qualitatively different when it is ofSize I than when it is larger. It is also possible that thememory set is qualitatively different at different times inthe sequence of memory matches. The absence ofin t erference from irrelevant words in scanning,matching, and counting of colors implies that a sensoryrepresentation can be held at least as long as it takes tofind successive stimuli of the target color. Even if it isnot possible to generate such a sensory memory from aname, following presentation of a stimulus from thepositive set. it would be expected that this would beavailable for at least the next comparison. The fastertimes found by Hock and Egeth (1970) with the Size 1memory set than with larger set sizes, for incongruent incomparison to control stimuli. could well reflect savingsresulting from memory matches of such a sensory naturefor Size 1 memory sets.

The failure of the attempt by Hock and Egeth toshow that sensory or perceptual encoding of colors canbe interfered with by written words does not necessarilyindicate that such an effect is not possible. In fact. astudy by Tecce and Dimartino (1965) might be taken asevidence that words can affect perception of colors.Their data appear to indicate that words spoken at thetime that brief color flashes occurred facilitatedrecognition of the color when they were the same as thecolor and delayed their recognition when they wereincongruent to the color. Recentlv. Dyer and Behar"

116 DYER

conducted a study in which color names and control Vswere exposed in black for 50 msec and then one of threecolors was flashed in the black area for O. 2. 5. 10, or20 msec. The S was required to identify the color or toguess it if unable to recognize it. The Ornsec exposureswere included to obtain a measure of guessing rate forthe different word conditions. In the "0" condition. itwas found that the control Vs produced a hit rate of33%. exactly in line with what would be expected forthree colors. When word stimuli were used in the "0"condition, Ss guessed the color corresponding to theword 427c of the time. This meant that only 58% of thetime did they guess a color different from that denotedby the word. Approximately the same proportion ofcolor responses congruent to the word were found forthe very brief 2-msec exposures of the color. It wasprobably this influence of the word on guessing thataccounted for what might instead be interpreted as afacilitation of recognition with congruent combinationsof words and colors and a delay of recognition withincongruent combinations. Longer color flashes whichdefinitely did allow some recognition showed noconsistent differential improvement or delay as a resultof their combination with congruent and incongruentwords. In light of this result, the claim of Tecce andDimartino (1965) for an effect of spoken words onrecognition can be largely discounted.

The 50·msec word exposure prior to color flash in theDyer and Behar study" would probably not besufficiently ahead of the color to generate activitycorresponding to the word that could influence encodingof color stimuli. This is because such color encodingappears to be faster at sensory stages than is theencoding of words (e.g., Uleman & Reeves, 1971). The50-msec word exposure was selected since 40 and60 msec of preexposure of black words had maximizedinterference in the Dyer (1971c) study of the effect ofblack word preexposure on color naming. The Dyer andBehar study" thus does not prove that words cannotaffect color encoding as determined by recognition, onlythat it does not do it in normal presentations of Stroopstimuli and even when the word is presented slightlybefore the color. This factor thus cannot account forany part of the delay in color naming that occurs whenwords are presented simultaneously with colors orslightly before.

Response Conflict Explanations

It is almost ironic that the counting, scanning, andmatching tasks which were described earlier in thissection are faster for a target color than for a targetword (e.g., U1eman & Reeves, J 971), whereas a namingresponse occurs much more quickly to a written wordthan to a color patch. This faster reading than namingapplies to other dimensions than color and is classicallyillustrated in Fraisse's (1969) finding that reading an"0" when it was presented as one of four possible letters

was 166 msec faster than naming the identical stimulusas a "circle" when it was presented as one of fourpossible geometric forms. Fraisse (1969) concluded aftera series of experiments that neither differential practiceon reading compared to naming nor stimulusdiscriminability could account for the basically fasterreading than naming. Another aspect of this basicdifference is shown in the work of Morin, Konick,Troxell. and McPherson (1965) and of Gholson andHohle (1968), who showed that letter and word readingwere relatively independent of the size of the set ofletters or words to be read, whereas the time to namecolors, faces, animals, and geometric forms all increasedgreatly as the size of the stimulus set increased.

Almost every investigator of the color-wordphenomenon since Stroop (1935) has viewed the furtherincrease in color-naming time when the patches arewords as a direct result of this faster assignment ofspoken words to written word stimuli than to colors andthe resulting conflict between this faster response to theirrelevant word aspect of the stimulus and the responseto the relevant color aspect of it. Some have speculatedfurther about the difference between naming andreading, and a few have been concerned with the natureof the response conflict and its resolution. At least oneinvestigator (Treisman. 1969) has discussed the questionof why selective attention, which can gate the irrelevantword when the task is one of counting. scanning. andmatching colors. fails to gate such irrelevant inputsduring color naming.

Stroop (1935) suggested that the faster naming ofwords than of colors resulted from the fact that a varietyof responses are learned to a specific color besides itsname, whereas only the name response is learned to thewritten color name. Such an explanation does not seemparticularly satisfactory. and to the author's knowledgeno one has shown that reading responses are delayed as aresult of learning other nonreading responses to a wordstimulus. This would seem to be an appropriate test ofthis hypothesis. Gholson and Hohle (1968). however.found this explanation to be the only one availablewhich could provide even a somewhat satisfactoryaccount for their findings of differential increases intime for naming of colors and forms with increasedstimulus set size relative to naming of words. Schiller(1966) rejected the differential practice explanation ofreading/naming differences. but his alternativeexplanation-that the individual letters, word length.etc.. constituted a much more redundant stimulus than acolor patch that differed only on the dimension ofhue-does not appear to square with the fasterprocessing of these color patches when the task iscounting, scanning, or matching.

Although Treisman and Fearnley (1969) deny thattheir explanation of the Stroop phenomenon is aresponse-conflict explanation. it would appear that theyhave reemphasized and provided a partial explanation ofthe fact that word responses are produced more rapidly

in response to written word stimuli than to color stimuli.Their experiment showed that judgments of same anddifferent (in a card-sorting task) were much faster forpairs of stimuli that were two words or two colors thanfor pairs of stimuli that were a color and a color name.Large further delays in "cross-attribute" matchingresulted when one of the stimuli was combined with anirrelevant stimulus from the other dimension. Thesedifferent matching tasks produced sorting-timedifferences that closely parallel the differences foundbetween word reading, control color naming, and colornaming with interference from irrelevant words. On thebasis of this, they reach the conclusion that readingwords is similar to judging whether two-word stimuli arethe same or different and that naming a color is likedetermining the same thing with a word stimulus and acolor stimulus. This implies that stimuli and responsesvary in their degree of similarity to each other and,specifically, that a written word is more similar than anobject to the spoken word that names both. A possiblebasis for greater similarity between written words andtheir spoken names than between objects and theirspoken names could be that the perceptual eventassociated with viewing a word includes an auditorycomponent. while perception of the object does notinclude this auditory activity. Such auditory activitycould provide a stronger connection between the wordand naming response. since it would be similar to thesensory feedback produced by the naming response.Greenwald (1970) has recently provided considerableevidence for a direct relation between the degree ofsimilarity of a stimulus to the feedback from theresponse and speed of responding. For example. he hasshown that a word can be spoken faster when it is aresponse to the same auditorily' presented word thanwhen it is read. The auditory stimulus is much moresimilar than the printed word to the sensory feedbackfrom the naming response. A parallel closer relationshipof the stimulus to the "feedback from the namingresponse" for the written word than for the object mayalso exist and account for faster reading than naming.

Interference was found by Treisman and Fearnley(1969) in the cross-attribute matching task regardless ofwhether the irrelevant aspect was a word or color. Whenthe relevant color is combined with an irrelevant word.this would appear to be interference from covert readingof the word to the covert naming of the relevant color(this is required for the judgment). Except for the factthat the naming is not overt. this is similar to theresponse-conflict processes that investigators fromStroop (1935) to Dalrymple-Alford and Azkoul (1972)have called on to explain the Stroop phenomenon itself.When the relevant word is combined with an irrelevantcolor. on the other hand. it could be transformation ofthe relevant word to a color code that is disrupted bythe presence of the irrelevan t color. II' this were true.then the Stroop phenomenon might be thought of as aspecific case of a general interference with stimulus

THE STROOP PHENOMENON AND ITS USE 117

transformation that occurs when the stimulus to betransformed is in the presence of an irrelevant stimulusthat is already ciose to the form to which the relevantstimulus will be transformed. This might apply only totransformations between words and things or might beeven more general.

An extension of the Treisman and Fearnley (I969)experiment by Dyer (1971 d) analyzed latencies ofindividual same and different responseswhen the pair ofstimuli was a word and a color. Same responses wereequally delayed when the irrelevant stimulus was a wordor a color. Different judgments were slower when theirrelevant stimulus was a word than when it was a color.Largely for this reason, it was concluded that for thedifferent responses. a word code was used for the centralcomparison. When the irrelevant stimulus was a color, itwould not be in the form of the central comparison andhence would not particularly interfere with the matchbetween the relevant word and the transformeduncombined relevant color. Morton and Chambers- haveproduced further data indicating more delay for cardsorting in the "cross-attribute" matching task when theirrelevant value is a word than when it - is a color.although it appears that they interpret this resultsomewhat differently. It may be that the form of thecentral comparison depends on the form of theuncombined stimulus and also the equivalence ornonequivalence of the word and the color. i.e.. whetherthe response is same or different.

Perhaps the most elaborate response-conflict model ofinterference in Stroop-like tasks has been provided byMorton (1969). Interference to card sorting onenumeration tasks was seen to occur from an irrelevantsymbol which readily generates a naming response thatoccupies a serial response buffer. This prevents theproper naming response used for sorting from occupyingthe buffer as soon as it would without the irrelevantstimulus present. It is a very general model that allowsfor auditory as well as visual interference to bothenumeration of visual objects and enumeration ofauditory events. However. Morton's general conclusionsfrom a series of experiments are much the same as thosepresented here. The interference is not seen to occurprior to availability of the symbol name. and theinterference is a form of response competition. It is ofinterest that auditory interference to color naming andvoice naming was not found by Thackray and Jones(1971) and Dyer and Mosko.s This suggests thatMorton's model may not apply to the Stroopphenomenon as well as it does to his enumeration tasks.Morton's model will be returned to in a subsequentanalysis of response competition processes.

The Role of Selective Attentionin the Stroop Phenomenon

Treisman and Feamley's (1969) study of within- andcross-attribute matching illustrated that it was not a

118 DYER

differential ability to focus on words and gate colorscompared to focusing on colors and gating of words thatproduced the Stroop phenomenon. since color-colormatches were as fast as word-word matches and were notdisrupted by the presence of irrelevant words. However.largely on the basis of the Stroop phenomenon,Treisman (1969) concluded that it was difficult. andperhaps impossible, to focus on either the word or coloranalyzer with gating of the input from the otheranalyzer. It is thus possible to conceive of the responsecompetition in the Stroop task as occurring because ofthis failure of selective attention to focus on colors andgate word input. The fact that fast counting, matching,and scanning for colors do not result in interferencefrom irrelevant words may only illustrate that attentionmust be directed to both aspects of the stimulus longenough for this parallel processing of relevant andirrelevant aspects to proceed to a level whereinterference will occur. However. it may be more thantime that is involved; attention may have to be directedto the level (perhaps auditory) at which name responsesto the sensory representations emerge.

The generality of Treisman's (1969) conclusion thatfocusing on one analyzer is difficult or impossible iscalled into question by the failure of many dimensionsto show much naming interference when combinedwithirrelevant words (White, 1969; Dyer. 1971a, 1972a;Shor. 1970. 1971). It could be that the color analyzer isa special type on which attention cannot be focused.whereas achromatic shades, movement direction, spatialposition. etc.. do allow at least some focusing withconsequent attenuation of the irrelevant words.

Of possible relevance to the question of selectiveattention and the Stroop phenomenon is the work ofHouston and 1. Jones (1967) and Houston (1969).Although Thackray and K. Jones (1971) found thatspoken names did not affect either control color namingor interference color naming, these other studies havedemonstrated a very interesting interaction relating tothe effect on color namingof continuous sounds such astrains. gibberish, and electronic music. These soundsdelay color naming in the condition without interferencebut actually speed color naming in the condition withinterference from irrelevant words. Unlike spoken colornames, the Houston sounds may reliably elicit nameresponses and as such may serve to interfere with colornaming much as the irrelevant written words do. Thefact that such stimuli are continually present and namingresponses stimulated by them are continually beingrejected apparently aids the S in rejecting the irrelevantwords on the interference card as well. This is basicallythe explanation that Houston and 1. Jones put forth fortheir data. This result appears to argue for someperceptual strengthening of color or inhibition of otherinputs.

WHAT IS THE BASISFOR RESPONSE COMPETITION?

The most elaborate response-conflict explanation was

provided by Klein (1964), who discussed a need forrestimulation by the color aspect of the colored colorword to overcome the irrelevant response to the wordand to allow generation of the proper response to thecolor. In Klein's position, the need for restimulationoccurred because it was considered easier to attend tothe word and ignore the color than the converse, anattentional hypothesis that Treisman and Fearnley(1969) have well discounted. Still Klein's restimulationposition without the emphasis on attention may havevalidity. Klein found that reading the word prior tocolor naming reduced the S's interference effects greatly,and. in fact, the double response required only a littlemore time than naming colors alone on the interferencecard. Similarly, black preexposures of the word prior tocoloration (Dyer. 1971 c) greatly reduced interferencefrom the irrelevant word when the preexposures werelonger than 100 msec. Thus, some relevance of Klein'srestimulation hypothesis may exist such that, withtraditional simultaneous presentation of both the wordand color. a person waits for the word activity to "diedown" and then processes the color to naming.Attractive as this "explanation" of the delay is, therestimulation by the color must be possible from anonerasable iconic image of the stimulus because Dyerand Kuehne (1972) showed that very brief presentationsof single Stroop stimuli as short as 25 msec produced"normal" interference to color naming in comparison tosimilar short presentations of control stimuli. This wastrue both with and without an erasure stimulus thatfollowed the brief Stroop stimulus. Without this erasurestimulus, interference and response times were almostidentical for presentations of 25, 50, 100, 200. and500 msec. With the mask, only when it immediatelyfollowed the 25-msec exposure was there a change ininterference(a nonsignificant reduction).

Perhaps the best present explanation for responsecompetition includesboth Treisman's(1969) claim of aninability to focus on either the color or word analyzersand Morton's (1969) assumption of a single responsechannel. The faster reading response tends to occupy thechannel before the color-naming response can do so. Theword response activity, however, has a very short timecourse as shown by the Dyer (1971c) black preexposurestudy, and this frees the response channel for theappropriate color-naming response. This correct responsehas already been selected or else can be generated fromvery brief "masked" stimuli (Dyer & Kuehne, 1972).Associative closeness of the irrelevant words to theresponse words [Morton's (1969) priming hypothesis]may account for the differential ability of irrelevantwords from different classes to occupy the singleresponse channel.

Of course, many things are absent in this explanation,including how the identity of the correct aspect and/orincorrect aspect for response is maintained. A possiblesolution could be derived from the Dalrymple-Alfordand Budayr (1966) study of sequential effects with thefirst available (reading) response being suppressed and

the second response being amplified. However, thefacilitation of color naming by congruent combinationswhich occurs with immediate feedback of response times(Dyer, 1971c; Dyer & Kuehne, 1972) indicates that thetagging of responses to stimuli must be different than atemporal process of suppression followed byamplification.

Of all the gaps in our explanation of the Stroopphenomenon, the most critical unanswered question isprobably that asked by Fraisse (1969): "Why is the timefor reading faster than the time for naming?" The searchfor this answer and for the answers to other questionsabout the Stroop phenomenon appears to be highlyworthwhile. The answers. when obtained, will apply tomany other important questions about the attentional.reading, and naming processes which are not specific tothis phenomenon but which constitute the basic fabricof cognitive behavior.

REFERENCES

Bakan. P.. & Alperson. B. Pronounceability. attensity, andinterference in the color-word test. American Journal ofPsychology. 1967.80,416-420.

Cramer, P. The Stroop effect in preschool aged children: Apreliminary study. Journal of Genetic Psychology, 1967. Ill,9-12.

Dalrymple-Alford. E. C. Interlingual interference in acolor-naming task. Psychonomic Science, 1968, 10. 215-216.

Dalrymple-Alford, E. C. Associative facilitation and interferencein the Stroop color-word task. Perception & Psychophysics.1972a.ll.274-276.

Dalrymple-Alford, E. C. Sound similarity and color-wordinterference in the Stroop task. Psychonomic Science. 1972b.28.209-210.

Dalrymple-Alford. E. Coo & Azkoul, J. The locus of interferencein the Stroop and related tasks. Perception & Psychophysics.1972. 11. 385-388.

Dalrymple-Alford, E. Coo & Budayr. B. Examination of someaspects of the Stroop color-word test. Perceptual & MotorSkills. 1966. 23. 1211-1214.

Danielv J. Performance in an interference test and some changesin the vegetative functions. Studia Psychologica, 1969. 11.267-271.

Derks. P. L.. & Calder. E. S. Information processing and verballabels: The Stroop color-word test. Paper presented at themel' t ing 0 I' the E a stern Psychological Association.Philadelphia. 1969.

Dyer. F. N. Word reading. color naming and Stroop interferenceas a function of background luminance. U.S. Army MedicalResearch Laboratory Report No. 889. August 20. 1970.

Dyer. F. N. A comparison of chromatic and achromatic versionsof the Stroop color-word test. Psychonomic Science. 1971a.22,235-237.

Dyer. F. N. Color naming interference in monolinguals andbilinguals. Journal of Verbal Learning & Verbal Behavior.1971b. 10.297-302.

Dyer, F. N. The duration of word meaning responses: Stroopinterference for different preexposures of the word.Psychonornic Science. 1971c. 25. 229-231.

Dyer. F. N. Latencies for matching of word-color pairs with"irrelevant" words or colors. U.S. Army Medical ResearchLaboratory Report No. 920. February 26. 1971d: Journal ofExperimental Psychology. in press.

Dyer. F. N. Latencies for movement naming with congruent and

THE STROOPPHENOMENON AND ITS USE 119

incongruent word stimuli. Perception & Psychophysics,1972a, 11. 377-280.

Dyer, F. N. Interference and facilitation for color naming withseparate bilateral presentations of the word and color. U.S.Army Medical Research Laboratory Report No. 982, May 23.1972b; Journal of Experimental Psychology, in press.

Dyer, F. N., & Kuehne, T. E. Color naming latencies with briefexposures of individual Stroop and control stimuli. U.S. ArmyMedical Research Laboratory Report No. 970, March 13.1972.

Dyer. F. N., & Severance, L. 1. Effects of irrelevant colors onreading of color names: A controlled replication of the"reversed Stroop" effect. Psychonomic Science. 1972. 28,336-338.

Dyer, F. N,; & Severance, L. J. Interference with successivepresentations of separate incongruent words and colors.Journal of Experimental Psychology, in press.

Egeth, H. E., Blecker. D. L., & Kamlet, A. S. Verbal interferencein a perceptual comparison task. Perception & Psychophysics.1969,6,355-356.

Ellison, A. E., & Lambert, W. E. Reduction of responseinterference through verbal repetition. British Journal ofPsychology, 1968,59, 147-155.

Fox, L. A., Shor, R. E., & Steinman, R. J. Semantic gradientsand interference in naming color, spatial direction, andnumerosity. Journal of Experimental Psychology. 1971. 91,59-65.

Fraisse, P. Why is naming longer than reading? ActaPsychologica, 1969.30,96-103.

Gholson, B., & Hohle. R. H. Verbal reaction times to hues vs huenames and forms vs form names. Perception & Psychophysics.1968.3.191-196.

Grand. S. Color-word interference: II. An investigation of therole of vocal conflict and hunger in associative priming.Journal of Experimental Psychology. 1968,77,31-40.

Greenwald. A. G. Sensory feedback mechanisms in performancecontrol: With special reference to the ideo-motor mechanism.Psychological Review, 1970,77,73-99.

Gumenik. W. E., & Glass. R. Effects of reducing the readabilityof the words in the Stroop color-word test. PsychonornicScience. 1970,20.247-248.

Hamers. J. Foo & Lambert. W. E. Bilingual interdependencies inauditory perception. Journal of Verbal Learning & VerbalBehavior, 1972, 11. 303-310.

Hochman, S. H. The effects of stress on Stroop color-wordperformance. Psychonomic Science. 1967.9.475-476.

Hochman. S. H. Stress and response competition in children'scolor-word performance. Perceptual & Motor Skills. 1969.28.115-118.

Hock. H. S; & Egeth, H. E. Verbal interference with encoding ina perceptual classification task. Journal of ExperimentalPsychology, 1970, 83.299-303.

Houston. B. K. Noise. task difficulty, and Stroop color-wordperformance. Journal of Experimental Psychology. 1969. 82.403-404.

Houston. B. K. & Jones. T. M. Distraction and Stroopcolor-word performance. Journal of Experimental Psychology.1967.74.54-56.

Jensen. A. Roo & Rohwer. \\'. D; Jr. The Stroop color-word test:A review. Acta Psychologica. 1966. 25. 36-93.

Kamlet, A. S; & Egeth, H. E. Note on construction ofStroop-type stimuli. Perceptual & Motor Skills, 1969.29.914.

Kanungo. R. N. Semantic satiation and verbal learning. Journalof Special Education. 1967.2.45-59.

Klein. G. S. Semantic power measured through the interferenceof words with color-naming. American Journal of Psychology.1964.77.576-588.

Liberman. A.. :\1.. Cooper. F. s.. Harris. K. Soo :\Iac:\,'ilagc. P. roo& Studdert-Kennedv. :\1. Some observations on a model forspeech perception. In \\. \Llthen-Dunn tLd.). .110,/,,15 tor file

120 DYER

perception of speech and visual form. Cambridge. "lass: \I.I.T.Press. 1967.

Lund. F. H. The role of practice in the speed of association.Journal of Experimental Psychology. 1927. 10.424-433.

Milner, B. Interhemispheric differences in the localization ofpsychological processes in man. British Medical Bulletin,1971. 27. 272-277.

Morin. R. E., Konick, A., Troxell. N., & Mcl'herson, S.Information and reaction time for "naming" responses.Journal of Experimental Psychology, 1965,70,309-314.

Morton, 1. Categories of interference: Verbal mediation andconflict in card sorting. British Journal of Psychology, 1969,60. 329-346.

Preston. M. S., & Lambert. W. E. 1nterlingual interference in abilingual version of· the Stroop color-word task. Journal ofVerbal Learning & Verbal Behavior. 1969,8,295-301.

Pritchatt, D. An investigation in some of the underlyingassociative verbal processes of the Stroop colour effect.Quarterly Journal of Experimental Psychology. 1968, 20,351-359.

Scheibe. K. E.. Shaver, P. R., & Carrier, S. C. Color associationvalues and response interference on variants of the Strooptest. Acta Psychologica, 1967,26,286-295.

Schiller. P. H. Developmental study of color-word interference.Journal of Experimental Psychology, 1966,72,105-108.

Severance, 1. J., & Dyer, F. N. Failure of subliminal wordpresentations to generate interference to color-naming. U.S.Army Medical Research Laboratory Report No. 1.005,September 27, 1972.

Shor, R. E. The processing of conceptual information on spatialdirections from pictorial and linguistic symbols. ActaPsychologica, 1970.32, 346-365.

Shor. R. E. Symbol processing speed differences and symbolinterference effects in a variety of concept domains. Journalof General Psychology, 1971,85.187-205.

Sichel. J. 1., & Chandler, K. A. The color-word interference test:The effects of varied color-word combinations upon verbalresponse latency. Journal of Psychology, 1969, 72. 219-231.

Smith, B. J. W., & Borg, G. W. U. The problem of retesting in theserial colour-word test. Psychological Research Bulletin IV:6,Lund University, Sweden. 1964.

Sternberg. S. The discovery of processing stages: Extensions ofDonders' method. Acta Psychologica, 1969, 30, 276-315.

Sternberg, S. Decomposing mental processes with reaction-timedata. Invited address, Midwestern Psychological Association,Detroit, May 1971.

Stroop, J. R. Studies of interference in serial verbal reactions.Journal of Experimental Psychology, 1935, 18,643-662.

Tecce, 1. 1.. & Dimartino, M. Effects of heightened drive (shock)on performance in a tachistoscopic color-word interferencetask. Psychological Reports. 1965. 16, 93-94.

Tecce, J. J., & Happ. S. J. Effects of shock-arousal on acard-sorting test of color-word interference. Perceptual &Motor Skills, 1964. 19.905-906.

Thackray, R. I.. & Jones. K. N. Level of arousal during Stroopperformance: Effects of speed stress and "distraction."Psychonornic Science, 1971.23. 133-135.

Treisman, A. M. Strategies and models of selective attentiun.Psychological Review.-1969. 76. 282-299.

Treisman, A. M., & Feamley, S. The Stroop test: Selectiveattention to colours and words. Nature, 1969, 222, 437-439.

Ulernan, J. S., & Reeves. J. A reversal of the Stroop interferenceeffect, through scanning. Perception & Psychophysics, 1971,9,293-295.

Warren, R. E. Stimulus encoding and memory. Journal ofExperimental Psychology, 1972,94,90-100.

White, B. W. Interference in identifying attributes and attributenames. Perception & Psychophysics, 1969,6, 166-168.

Windes, J. D. Reaction time for numerical coding and naming ofnumerals. Journal of Experimental Psychology, 1968. 78,318-322.

NOTES

1. Dyer, F. N., & Severance. L. J. Failure of spokenincongruent names to delay color naming. In preparation.

2. Dyer, F. N.. & Mosko. J. D. Failure of spoken incongruentspeaker names to interfere with voice naming. In preparation.

3. Dyer, F. N., & Harker. G. S. Stroop interference withstimulus presentations to the right and left hemispheres. Inpreparation.

4. Dyer, F. N.. & Behar. I. Absence of effects on colorrecognition of congruent and incongruent color name stimuli. Inpreparation.

5. Morton, 1. Personal communication.

(Received for publication October 22. 1972;accepted November 13,1972.)