Journal of Experimental Psychology: Human Perception and Performance 2002, Vol. 28, No. 5, 1019 –1038
Copyright 2002 by the American Psychological Association, Inc. 0096-1523/02/$5.00 DOI: 10.1037//0096-15184.108.40.2069
Why Do Non-Color Words Interfere With Color Naming? Jennifer S. Burt University of Queensland In the non-color–word Stroop task, university students’ response latencies were longer for low-frequency than for higher frequency target words. Visual identity primes facilitated color naming in groups reading the prime silently or processing it semantically (Experiment 1) but did not when participants generated a rhyme of the prime (Experiment 3). With auditory identity primes, generating an associate or a rhyme of the prime produced interference (Experiments 2 and 3). Color-naming latencies were longer for nonwords than for words (Experiment 4). There was a small long-term repetition benefit in color naming for low-frequency words that had been presented in the lexical decision task (Experiment 5). Facilitation of word recognition speeds color naming except when phonological activation of the base word increases response competition.
recognition. In Warren’s task, participants name the color of a base word having no relation to color or color names. Typically, their response latency is only slightly slower than when they name the color of a meaningless letter string (e.g., XXXXX; Klein, 1964). The small interference observed in the non-color–word Stroop task has been attributed to competition between activated baseword and color-name responses, just as response competition traditionally has played a significant role in explanations of the standard color–word Stroop task (see the review by C. M. MacLeod, 1991; Morton & Chambers, 1973). On the basis of Morton’s analysis of color naming, Warren (1972, 1974) suggested that the word and color attributes of the stimulus are processed in parallel until a response is prepared for output in a limited capacity response buffer that can accommodate only one of the two responses. On trials in which the word response becomes available before the color, there is a delay while the inappropriate response is cleared from the buffer. This is a horse race model, in that the magnitude of interference is related to the probability that the word becomes available before the color name. Manipulations that increase this probability by facilitating recognition of the base word can be expected to increase the magnitude of color-naming interference. Consistent with this expectation, Warren found that when colored base words (e.g., king in blue letters) were targets in an associative priming arrangement, and they were preceded a second or so earlier by associatively related primes (e.g., queen) or unrelated primes (e.g., water), associative primes increased the latency to name the color of the target words. This result was obtained regardless of whether the primes were presented visually in achromatic lettering (Warren, 1974) or auditorily (Warren, 1972, 1974) and has been interpreted as showing that semantic primes facilitate word recognition. Horse race accounts of color-naming interference were popular for the standard Stroop task (see the review by C. M. MacLeod, 1991), but they have been challenged by findings during the past decade or so. For example, Dunbar and MacLeod (1984) showed that manipulations of the relative speed of reading and color processing (e.g., by geometrical transformations of print) did not abolish the interference as predicted by the horse race model, and manipulations of color–word stimulus onset asynchrony (SOA) do
Since the early work of Stroop (1935), it has been well established that the concurrent processing of word and color dimensions of an incongruent Stroop stimulus—for example, red written in blue letters—interferes substantially with rapid naming of the letter color. Decades of research on this standard Stroop test (see the review by C. M. MacLeod, 1991) have confirmed what is apparent to anyone who has performed the task: If the task format requires attention to the location of the word in the display, it is difficult to avoid reading the word, despite instructions to ignore it. This means neither that word reading is automatic nor that interference by word distractors is all or none. Stroop interference is affected by the spatial integration of the color and word attributes of the Stroop stimulus (e.g., Kahneman & Henik, 1981) and by the nature of the context provided by other trials (Logan & Zbrodoff, 1979; Lowe & Mitterer, 1982). However, among highly literate adults, and in the traditional, spatially integrated colored-word display used in the present studies, reading is an overlearned response that is difficult to avoid. Because word reading in the conventional Stroop display is highly probable but incidental to the primary task, this selective attention task is a useful tool for answering questions about the operation and control of the visual word-processing system. In comparison with the commonly used lexical decision task (LDT), the color-naming task at least potentially allows researchers to make inferences about word recognition that are less likely to reflect task-specific strategic influences on word processing. The present studies had university students as participants and focused on Warren’s (1972, 1974) non-color–word version of the Stroop task, which has been used by researchers investigating visual word
These experiments were supported by a University of Queensland Special Projects grant and a small grant from the Australian Research Council. Experiments 1 and 3 were reported at the Australian Experimental Psychology Conference, University of Tasmania, Hobart, Australia (April 1998). I thank Karin Humphreys, Ruth Bouma, Andrew Ede, and Toni Cannon for assistance with data collection. Correspondence concerning this article should be addressed to Jennifer S. Burt, School of Psychology, University of Queensland, Queensland 4072, Australia. E-mail: [email protected]
not produce the results predicted by the horse race model (M. O. Glaser & Glaser, 1982; W. R. Glaser & Dungelhoff, 1984). Thus, although response competition remains an important component of current models of Stroop interference, the amount of interference is thought to be determined more by the relative strength of activation in color-name and color-processing pathways (Cohen, Dunbar, & McClelland, 1990; Cohen & Huston, 1994); or by the compatibility of word reading and word naming (Phaf, van der Heijden, & Hudson, 1990), than by the relative speed of processing in the two pathways. In an analogous manner, for the non-color–word Stroop task, an alternative view is that color-naming interference may be explained in terms of competition for output between activated word responses, as suggested by Warren, but that such response competition does not depend on the relative processing speed of word and color attributes. As reviewed below, evidence against the horse race model as an account of the Warren task comes from recent color-naming research on the effects of manipulations designed to facilitate base-word processing and, hence, increase recognition speed. There is no doubt as to the replicability of Warren’s finding that associative priming in the non-color–word Stroop task produces color-naming interference. The result has been reported frequently (Burt, 1994; Henik, Friedrich, & Kellogg, 1983; C. MacLeod & Rutherford, 1992; Mathews & MacLeod, 1985; McClain, 1983; Merrill, Sperber, & McCauley, 1981; Oden & Spira, 1983; Parkin, 1979; Whitney, 1986; Whitney, McKay, Kellas, & Emerson, 1985). However, the associative priming results from the noncolor–word Stroop task differ in important respects from those reported in traditional word recognition tasks. Associative primes produce interference in color naming only when participants make a meaning-based response to primes; no priming effect has been reported when participants silently read primes or make a letter search or syllable count (Burt, 1994, 1999; Henik et al., 1983; Parkin, 1979). A somewhat similar effect of prime response requirement, in that some semantic processing of the prime enhances or is necessary for priming, has been found in word-naming tasks (e.g., Parkin, 1984), the LDT (e.g., Smith, Theodor, & Franklin, 1983), and perceptual identification tasks (Burt, Walker, Humphreys, & Tehan, 1993), but in these tasks priming also is observed when the prime is read silently (see the review by Neely, 1991). Furthermore, the salience of the prime–target relation is critical for priming interference in color naming. Temporal grouping of prime–target pairs to enhance the salience of their relationship appears to be necessary for associative priming in color naming (Burt, 1999) but not in the LDT (Fischler, 1977; Shelton & Martin, 1992). It is arguable that the preceding results can be explained in terms of lower sensitivity of the color-naming task, as compared with traditional word recognition tasks, to variations in word-processing speed. More difficult to reconcile with a relative processing speed model of interference in the non-color–word Stroop task are inconsistent results within the color-naming task for a number of word-processing manipulations. For example, Burt (1994, 1999) has reported that base-word frequency effects in the non-color– word Stroop task run counter to the direction predicted on the basis of word-processing speed (see also Kachelski, 1997). That is, the color of high-frequency words is named faster than is the color of low-frequency words. Also, Burt (1994) found that identity prim-
ing, in which the target is the same word as the prime, produced color-naming facilitation in the non-color–word Stroop task, although theoretical presuppositions and available evidence (e.g., Evett & Humphreys, 1981) suggest that identity primes more strongly facilitate word recognition than do associative primes. Finally, Burt (1999) found that associative primes can facilitate color naming at a short prime–target SOA. Observations of facilitation rather than interference in color naming as a result of priming manipulations are not unique to the present laboratory. Facilitation of color naming also has been reported in sentence priming studies in which participants covertly generated the target word on sentence presentation (Dosher & Corbett, 1982; Whitney, 1986). In the literature on processing biases in anxiety disorders and phobias, facilitation of color naming has been observed (Kyrios & Iob, 1998; C. MacLeod & Rutherford, 1992), a result that contrasts with the typical finding of retardation of color naming to threat words (e.g., Mathews & MacLeod, 1985; Mogg, Mathews, & Weinman, 1989; Watts, McKenna, Sharrock, & Trezise, 1986). Reconciling the evidence about the effects of variation in lexical processing speed in the non-color–word Stroop task is possible if it is acknowledged that associative priming, like other manipulations used in word recognition tasks, may have effects on color naming that do not arise from the facilitation by a prime of base-word recognition. The present view (Burt, 1999) is that characteristics of the base words and primes may foster strategies that in traditional word recognition tasks facilitate responding to a target word, but in the color-naming task disrupt preparation of the color-naming response. In the following discussion, the base-word component of the colored Stroop stimulus is referred to as the target, reflecting its role in the associative and identity priming paradigms. In the typical associative priming paradigm used in the colornaming task, there is a high proportion of associated prime–target pairs, and the SOA is relatively long (500 ms or more). Under these conditions, participants may generate expectancies about the identity of the target when processing the prime (Neely, 1991). In the related-prime condition, the target may be one of the words generated during prime processing. Thus, relative to unrelated primes, there may be increased elaborative processing of the target, perhaps including retrospective meaning-relatedness judgments about the prime and target (Neely, Keefe, & Ross, 1989) or the target and the generated associates, or reactivation of the related prime when associative relationships are bidirectional. As discussed later, this elaborative processing may conflict directly with preparation of a color-naming response with related targets or indirectly, by increasing the activation of the phonology or articulatory program of the target word. In the context of theories of attention, the conflict may result from bottlenecks in response selection (Pashler & Johnston, 1998) or more general processing resource limitations if resources are diverted from the colornaming task to the word stimuli (Kahneman, 1973). Finally, it may reflect difficulties in maintaining the appropriate set or task orientation (cf. Monsell, 1996). Regardless of the precise mechanism of interference in word priming paradigms in the non-color–word Stroop task, it is proposed here that contrary to horse race models of the non-color–word Stroop task, the color-naming interference often observed in the associative priming paradigm does not arise because primes facil-
PHONOLOGY AND COLOR NAMING
itate target word recognition. Facilitation of word reading arguably reduces the overall concurrent processing load and consequently facilitates color naming— or, more accurately, reduces colornaming interference—relative to an unrelated prime condition. The present studies used the identity priming arrangement, wherein the prime and target are identical or unrelated, to further delineate the conditions that produce facilitation or interference in color naming in the non-color–word Stroop task and to provide a basis for explaining these effects. It was predicted that, except in instances in which conditions foster a failure of selective attention as outlined earlier, identity priming and word repetition would facilitate color naming, as demonstrated previously (Burt, 1994) and as shown in the LDT, naming, and perceptual identification tasks. In contrast, in the identity priming paradigm, manipulations devised to activate the pronunciation of the prime or induce effortful semantic or elaborative processing of the prime were expected to produce greater interference on identity trials, in which the target is primed by itself, than on unrelated prime trials. A major focus was to investigate the basis of the color-naming interference observed in situations in which priming manipulations are expected to facilitate base-word recognition. The effects of base-word frequency were assessed because this variable is agreed to affect word recognition speed (e.g., Monsell, 1991).
Experiment 1 In a previous study (Burt, 1994), facilitation was produced by identity primes in color naming when participants were instructed to remember the prime and recall it after each of their colornaming responses. This outcome contrasts with the results under prime recall in associative priming, wherein associative primes have produced interference in color naming relative to unrelated primes (Burt, 1994; Warren, 1972). In the identity priming case, substantial facilitation of target processing—and, thus, facilitation of color naming—was expected on the basis of previously described research from the present laboratory. However, it is unclear why, in the previous study, identity prime recall did not engage working memory processes that interfered with color naming. Rehearsal of the identity prime and, thus, the to-be-ignored target word might more strongly interfere than rehearsal of an unrelated word. Indeed, interference from identity primes was reported by McClain (1983) in a study in which prime recall was required. In McClain’s study, color-related base words were present in the materials and participants responded under considerable time pressure, factors that may have promoted interference in the identity prime condition. One explanation is that, in the earlier study (Burt, 1994), the prime recall task may have produced less interference in the identity prime condition because the prime recall task was less demanding than in the unrelated condition. Participants need only say the target at the recall prompt on identity trials, and thus they do not need to recall the prime. In the associative priming paradigm, the demands of the prime recall task may have been comparable for the two prime conditions. An additional factor differentiating associative from identity priming is that elaborative and semantic processing of the prime and its associates is more likely in the associative than the identity case. Interference in color naming across a range of prime processing instructions in the associative priming paradigm may be a
direct result of semantic processing of the prime and target. Alternatively, semantic processing may increase color-naming interference only indirectly. A possibility that is a focus of interest here is that semantically based strategies such as expectancy generation may foster activation of the phonology of the target word during preparation of the color word, and it is this phonological activation that is the primary source of the color-naming interference (Burt, 1999). One way to assess the contribution of elaborative semantic processing to color-naming interference is to investigate identity priming with a semantically based prime task that is likely to promote elaborative associative processing of the prime (and hence the target in the identity condition), and perhaps other words, but is unlikely to promote strong activation of prime and target pronunciations. In Experiment 1, the prime task involved semantic or deep processing of the prime (Craik & Tulving, 1975) but did not require that a phonological representation of the prime be held in memory during preparation of the color-naming response, nor was it likely to promote phonological activation of the target after the onset of the target display. Participants in a deep prime processing condition (hereafter termed the associative group) were asked to silently generate an associate to the prime, and participants in the comparison group (the read group) were instructed to read primes silently (cf. Burt et al., 1993). There was no overt response requirement in case the response itself interfered with preparation for the color-naming task. Indirect evidence that participants in the associative group were following instructions was obtained from an episodic recognition test of primes conducted immediately after the priming session. Target words were selected to represent three levels of natural language frequency.
Method Participants. A total of 48 introductory psychology students, all of whom spoke English as their first language, participated for course credit. There were 24 participants assigned to each of the read prime and associative prime processing groups. Materials. The targets were 120 three- to six-letter words, mainly nouns and verbs, divided into three frequency sets of 40 words each. High-frequency words had a mean frequency of 512 in the Kucera and Francis (1967) corpus of approximately one million words (median: 440; range: 200 –1,360). The mean was 53 (median: 52; range: 21– 83) for the medium-frequency words, and the mean was 1.9 (median: 2; range: 1–3) for the low-frequency words. The mean word length for each category was 4.5 letters. An additional 120 words of similar characteristics served as unrelated primes. These words had a frequency range of 1–760 (Kucera & Francis, 1967) and were allocated to targets such that the mean length of primes (4.5 letters) and their mean frequency (88) were comparable over target frequency sets. For one counterbalanced list, the targets in the first half of each frequency set were primed by themselves, and those in the second half had the unrelated prime words. The converse allocation was made for the second list. Each target in each condition was rotated through the four colors red, blue, green, and purple to produce eight sets. Items were counterbalanced separately within session halves in such a way that each condition was represented equally often in each session half and each target appeared equally often in each condition in each session half. Each target appeared once in each condition over the eight sets, but in only two of the four colors in a session half. The sequence of trials was randomized with the constraint that no color appeared more than three times in succession. Four unrelated practice trials were added at the beginning of each set.
For the recognition test, 10 of the unrelated primes of each frequency set were tested with a set of 30 unrelated primes not presented during the priming session (i.e., unrelated primes allocated to targets presented in the identity condition). The same test of randomly sequenced primes was presented to all participants, but, as a result of counterbalancing, the set of items that was old for half of the participants was new for the remaining participants, and vice versa. The words were typed one per line on a sheet of paper, and participants were asked to circle words that they saw in white letters during the priming session. Procedure. Participants were tested individually on a BBC microcomputer. They wore a headset microphone that was connected to the microcomputer by means of a voice-operated relay. Words were presented in the center of a color monitor in double height (5 to 9 mm) lowercase letters, primes in white and targets in red, blue, green, or purple (BBC magenta). On each trial, the prime was displayed for 1,500 ms, the screen was clear for 1,000 ms, and the target was displayed until terminated by the participant’s vocal response and a keypress by the experimenter to indicate whether the response was correct or incorrect or the microphone failed. It was discovered in pilot testing that participants required a long prime– target SOA to complete the associate generation task. A 4-s interval was interpolated before the next trial. Participants were instructed to name the color of the target word as quickly as possible, without making an excessive number (more than 1 in 20) of errors. Participants in the read group read the primes silently, whereas those in the associative processing group silently generated an associate of the prime (the first word to come to mind). Participants in the latter group were given examples of typical associates exemplifying various semantic relationships and were instructed to terminate any associate generation as soon as the target appeared. Before the test session, participants practiced on 25 unrelated pairs.
Results Mean latencies for the two prime processing groups during the priming phase are shown in Figure 1. The error rate was negligible (M ⫽ 0.2%), with no errors made in the identity condition and a maximum rate of 0.6% for unrelated low-frequency targets in the read group. No analyses were conducted on these data. Before analysis of latency data, errors, microphone failures, and times more than three standard deviations away from a participant’s correct mean within the identity and unrelated prime conditions were removed, resulting in the loss of 4.8% of trials for the read group (3.6% microphone failures and 1.2% extreme latencies) and
Figure 1. Mean correct color-naming latencies for the group reading (read group) versus the group generating associates to (associative group) the visual primes in Experiment 1.
3.2% of trials for the associative group (2.2% microphone failures and 1.0% extreme latencies). In this and all subsequent studies, statistical hypotheses were tested against an alpha value of .05. A Counterbalance Set (two levels) ⫻ Prime Processing Group ⫻ Frequency ⫻ Prime Type analysis of variance (ANOVA) was conducted on mean color-naming latencies (throughout F1 refers to analyses with participants as the random effect, and F2 refers to analyses by items). There were main effects of group, with the associative group slower than the read group, F1(1, 44) ⫽ 11.51, MSE ⫽ 88,118.8, and F2(1, 114) ⫽ 1,480.67, MSE ⫽ 1,154.7; target frequency, with slower responses to low-frequency targets, F1(2, 88) ⫽ 8.81, MSE ⫽ 630.4, and F2(2, 114) ⫽ 6.85, MSE ⫽ 1,558.0; and prime type, with the identity condition faster than the unrelated condition, F1(1, 44) ⫽ 28.81, MSE ⫽ 1,060.7, and F2(1, 114) ⫽ 31.90, MSE ⫽ 1,521.2. The latter two effects were qualified by a significant Frequency ⫻ Prime Type interaction, with priming greater for lower frequency targets, F1(2, 88) ⫽ 5.74, MSE ⫽ 819.5, and F2(2, 114) ⫽ 3.94, MSE ⫽ 1,521.2. No other effects were significant. Simple main effects analyses of the Prime Type ⫻ Frequency interaction revealed that the identity priming facilitation was marginally significant for high-frequency targets, F1(1, 44) ⫽ 5.14, p ⬍ .05, MSE ⫽ 749.3, and F2(1, 38) ⫽ 3.41, p ⫽ .07, MSE ⫽ 1,610.8, and reliable for medium-frequency targets, F1(1, 44) ⫽ 4.20, MSE ⫽ 875.5, and F2(1, 38) ⫽ 5.66, MSE ⫽ 1,460.9, and low-frequency targets (Fs ⬎ 30). In the test of episodic recognition of primes, performance was scored as hits minus false alarms. The associative prime processing group had a mean of 55% and the read group a significantly lower mean of 18%, F1(1, 46) ⫽ 34.80, p ⬍ .01, MSE ⫽ 34.8.
Discussion In the recognition memory test of unrelated primes, the associative group showed a substantial superiority, indicating that participants were engaged in elaborative or semantic processing of the primes that increased memory for these items. Therefore, it is inferred that participants probably were complying with the instructions concerning the primes. In color naming, there was a significant disadvantage in latencies for low-frequency targets that replicated previous findings in the present laboratory (Burt, 1994, 1999). This finding was consistent with the hypothesis that color naming is facilitated by factors that increase the speed or ease of base-word recognition. One qualification is that, in this and previous studies, the frequency effect has been demonstrated in the context of a priming manipulation, and thus expectancies about targets may have affected latencies, perhaps differentially so for rare and common words. Consequently, a replication of the base-word frequency effect was attempted in Experiment 4, in which there were no prime words and color-naming responses were made to individually presented colored words. The associate generation task had a severe effect on colornaming latencies, with responses on unrelated trials 120 ms slower in the associative group than in the read group, even though participants were instructed to stop processing primes when the target appeared. There was no significant effect of depth of prime processing manipulation on magnitude of priming, in that there was no Group ⫻ Prime Type interaction. The associative group showed substantial identity priming facilitation averaged over tar-
PHONOLOGY AND COLOR NAMING
get frequency and exhibited no evidence of priming interference. Consequently, together with the previous findings (Burt, 1994), the priming results confirm that identity priming facilitation with visual primes is robust over variations in prime response requirement. The predictability of identity targets may have contributed to the identity priming facilitation in color naming. Predicted targets may be easier to ignore during preparation of the color-naming response (cf. W. A. Johnston & Hawley, 1994). The significant variation in the magnitude of priming over levels of target frequency was foreshadowed in the pattern of results observed in the prime recall group in the previous study (Burt, 1994), wherein priming was observed at all frequency levels but there was a (nonsignificant) tendency for magnitude of priming to increase as target frequency decreased. Reduction in identity priming at high levels of target frequency has been observed in the LDT and the perceptual identification task (e.g., Forster & Davis, 1984; Jacoby & Dallas, 1981) and was termed by Forster and Davis (1984) the frequency attenuation effect. Forster and Davis observed frequency attenuation in the LDT when participants were required to respond to primes but not when primes were presented briefly with masks to prevent participants from remembering the prime when responding to the target. In the present study, frequency attenuation was more evident for the deep than the shallow prime processing groups, although not significantly so. Thus, the present results are consistent with the inference drawn by Forster and Davis that the frequency attenuation effect is unlikely to arise early in lexical processing. A contributing factor to the frequency attenuation effect with visual prime presentation in color naming may be a perceptual fluency effect (Jacoby & Dallas, 1981) that supports a classification of a target as a repeated stimulus that, once identified, does not require additional processing. Such an effect may be greater for lowfrequency targets because they are processed with more difficulty. The fact that identity priming resulted in color-naming facilitation when elaborative processing of primes was required goes some way toward ruling out a semantically based account of the interference observed previously in the associative priming paradigm (Burt, 1994). If semantic processing of the prime and target is directly implicated in color-naming interference, then it is difficult to explain the strong facilitation observed for identity primes in the present experiment. It is true that the prime–target SOA was almost twice as long in the present study as the SOA of 1,300 ms used in Burt’s previous study of associative priming with prime recall. However, as noted earlier, the substantial increase in response latencies for the associate generation group over the read prime group in the present experiment indicates that the prime processing task may have continued for a duration sufficiently long to affect response preparation. The associate generation task can be expected to have a different impact on attention to the color-naming task in the identity arrangement case than in the associative priming case investigated previously. With associated primes, an associate (e.g., queen) of the prime (king) that is activated as a result of elaborative prime processing may be identical to the target (queen). When this activated associate is the same as the target word, its phonological activation may magnify response conflict arising during processing of the target word. In the identity priming case, a primary difference is that a generated associate (queen) is unlikely to match the identical target (king) or unrelated target (e.g., house). There-
fore, less activation of the target word phonology is likely. The activated phonology of the associate can be expected to cause a similar amount of interference over prime conditions. Thus, in the absence of a source of differential interference for identity relative to unrelated primes, the facilitation of lexical processing caused by the identity prime (Forster & Davis, 1984) may determine the consequences for color naming; that is, more efficient target word processing leads to faster color naming. In summary, the tentative conclusion at this point is that colornaming interference in associative priming in Warren’s non-color– word Stroop task arises from processes supporting activation of the phonological representation of the target and perhaps also the prime. Thus, although the priming effect appears to depend on participants’ making a meaning-based response to the prime (Burt, 1999), it is not semantic processing per se of the prime or target that is the basis of the interference effect. In Experiments 2 and 3, the phonological activation hypothesis was tested more directly in the identity priming paradigm with auditory prime presentation.
Experiment 2 It was suggested earlier that interference in the non-color–word paradigm is fostered by activation of the target word phonology just before or during the preparation of the color-naming response. If this is the case, then manipulations devised to activate the phonology of target words can be expected to produce colornaming interference in the Warren priming paradigm. In Experiment 2, phonological activation of the target was attempted in the identity priming paradigm by means of auditory prime presentation. Greater interference was expected in the identity than in the unrelated condition, because in contrast to the unrelated case, the identity prime should promote activation of the phonology of the target word, thus increasing conflict between the target word and the color name. Auditory prime presentation may have contributed to associative priming interference in color naming in a number of previous studies (Conrad, 1974; Oden & Spira, 1983; Warren, 1972, 1974; Whitney et al., 1985). Auditory presentation of a word may make its phonological code available in working memory and thus potentially a source of response conflict in the color-naming paradigm. A considerable body of research on the unattended speech effect indicates that auditorily presented words disrupt rehearsal of visually presented verbal material in working memory tasks (Jones, 1993; Salame´ & Baddeley, 1982). In Experiment 2, in parallel with Experiment 1, one group listened to the prime and a second group silently generated an associate of the prime. The materials from Experiment 1 were used to facilitate comparison with the visual priming case. As discussed previously, the associate generation task requires attention to the prime and perhaps phonological activation of the prime word and its associate. Therefore, priming interference was predicted for this group. In the listen-to-prime group, participants can be expected to ignore the prime as far as possible, as a way of minimizing interference in the color-naming task. Thus, in this group, the effect of the auditory prime presentation might dissipate over the 2.5-s prime–target SOA, and interference was predicted to be weaker than in the associative group or absent. In terms of the effects of auditory identity primes on lexical processing of the target, previous research indicates that there is some cross-modal facilitation of word recognition when the prime
and target are identical but differ in presentation modality. For word repetitions over intervening trials or study–test blocks in the LDT and perceptual identification, facilitation has been found for auditory first presentations and visual tests (e.g., Clarke & Morton, 1983; Kirsner, Milech, & Standen, 1983; Kirsner & Smith, 1974). Kirsner, Dunn, and Standen (1987) concluded that on average, aggregated over studies, the repetition benefit is reduced by 50% to 60% by a change in stimulus modality between first and second presentations. There have been few investigations of modality effects in an identity priming paradigm, wherein the interval between first and second presentations is typically less than 3 s. Available evidence (McKone & Dennis, 2000; Whatmough & Arguin, 1998) shows a substantial identity priming benefit in the modality-change condition in the LDT. In Experiment 2, an assessment of the effect of auditory identity primes on responding to target words was obtained from groups tested with the same materials and procedures, except that they were instructed to name the target word and ignore its color. For the word-naming groups, substantial identity priming facilitation was predicted, with the associative group perhaps showing larger facilitation than the group only listening to primes (Parkin, 1984). For the color-naming groups, as indicated, interference was predicted, with the associative group predicted to show the larger effect.
Method Participants. A total of 80 introductory psychology students who spoke English as their first language participated for course credit. They were randomly assigned to associate generation and listen color-naming groups (24 participants in each) and to associate generation and listen word-naming groups (16 participants in each). Materials. The item set from Experiment 1 was used, counterbalanced as described in that experiment. Procedure. Participants sat in front of the BBC microcomputer, wearing a headset microphone whose earpieces had little effect on participants’ hearing. The prime words were recorded on audiotape by the author, an Australian speaker. A National Panasonic tape recorder connected to the microcomputer played the prime word through a speaker above and approximately 1 m away from the participant’s head, at a volume of 65–70 dB. The recorder was activated by a microswitch on the microcomputer before each trial and stopped 1,500 ms after target presentation. The target was displayed in colored lettering as described previously, 2,500 ms after the onset of the prime was detected through a voice-operated relay connected to a second microswitch on the microcomputer. Participants in the listen groups were instructed to listen to the prime word, and participants in the associative groups were instructed to silently generate an associate, as described in Experiment 1. Color-naming participants named the color of the target, and word-naming participants were told to disregard the color of the target and to name the word as quickly as possible. Participants completed 25 practice trials with unrelated primes and targets before beginning the test block. Immediately after the priming phase, participants completed a recognition memory test on a subset of 60 unrelated primes (half seen and half unseen in the priming phase), as described in Experiment 1.
dition are shown for the word-naming and color-naming groups in Figures 2 and 3, respectively. Word naming. Error rates were negligible (0.1%), recorded at a maximum of 0.6% for low-frequency unrelated targets in each group, and were not analyzed. Two percent of trials in the listen group and 1.3% in the associative group were lost because of extreme latencies, and an additional 2.6% (listen) and 2.3% (associative) of trials were lost through microphone failure. A Counterbalance Set (two levels) ⫻ Prime Processing Group ⫻ Target Frequency (high vs. medium vs. low) ⫻ Prime Type (identity vs. unrelated) ANOVA was conducted on mean naming latencies. There were main effects of group, with latencies longer for the associative group, F1(1, 28) ⫽ 6.61, MSE ⫽ 67,869.5, and F2(1, 114) ⫽ 606.84, MSE ⫽ 1,910.9; target frequency, with high frequency fastest and low frequency slowest, F1(2, 56) ⫽ 58.02, MSE ⫽ 717.4, and F2(2, 114) ⫽ 28.52, MSE ⫽ 3,618.2; and prime type, with identity faster than control, F1(1, 28) ⫽ 42.43, MSE ⫽ 2,450.3, and F2(1, 114) ⫽ 145.33, MSE ⫽ 1,920.1. The two-way interactions between group and prime type and between frequency and prime type were reliable in the item but not the participant analysis. There was a three-way interaction of group, frequency, and prime type, F1(2, 56) ⫽ 4.86, MSE ⫽ 670.3, and F2(2, 114) ⫽ 3.93, MSE ⫽ 2,823.1, reflecting a larger priming effect for medium- and low-frequency items than for high-frequency items in the associative but not the listen group. The three-way interaction was decomposed through separate analyses for each group. These analyses revealed that the Frequency ⫻ Prime Type interaction was reliable in the associative group, F1(2, 28) ⫽ 6.53, MSE ⫽ 834.3, and F2(2, 114) ⫽ 6.97, MSE ⫽ 2,985.4, but not in the listen group, F1(2, 28) ⫽ 1.44, p ⬎ .05, MSE ⫽ 506.4, and F2 ⬍ 1. Episodic recognition of primes. The recognition memory data, scored as hits minus false alarms, revealed better memory of primes in the associative processing group (53% correct) than in the listen group (31% correct), F(1, 30) ⫽ 8.04, MSE ⫽ 45.4. Color naming. Error rates varied from 0.2% to 1.3% over the six conditions for the two groups, with means of 0.6% and 0.4% for the listen and associative groups, respectively, and 0.6% and 0.5% for the identity and unrelated conditions, respectively. A
Results Latency data were preprocessed as described previously, with errors, microphone failures, and latencies in excess of three standard deviations outside a participant’s mean within prime type conditions removed before analysis. Mean latencies in each con-
Figure 2. Mean correct word-naming latencies for the group listening (listen group) versus the group generating associates to (associative group) the auditory primes in Experiment 2.
PHONOLOGY AND COLOR NAMING
Figure 3. Mean correct color-naming latencies for the group listening (listen group) versus the group generating associates to (associative group) the auditory primes in Experiment 2.
Prime Processing Group ⫻ Target Frequency ⫻ Prime Type ANOVA revealed no significant effects. A total of 3.5% and 2.1% of trials were lost as extreme times for the listen and associative groups, respectively, with microphone failures causing losses of 2.3% (listen) and 1.1% (associative). A Counterbalance Set ⫻ Prime Processing Group ⫻ Target Frequency ⫻ Prime Type ANOVA on mean color-naming latencies revealed a significant main effect of target frequency, with color-naming latencies faster for higher frequency targets, F1(2, 88) ⫽ 10.09, MSE ⫽ 1,458.7, and F2(2, 114) ⫽ 16.38, MSE ⫽ 2,991.1, and a main effect of prime type such that, averaged over groups, there was faster color naming for the identity prime condition, F1(1, 44) ⫽ 4.71, MSE ⫽ 2,043.8, and F2(1, 114) ⫽ 6.01, MSE ⫽ 2,511.4. However, the prime type effect was qualified by a Group ⫻ Prime Type interaction reflecting priming interference for the associative group and no priming effect for the listen group, F1(1, 44) ⫽ 4.55, MSE ⫽ 2,043.6, and F2(1, 114) ⫽ 11.62, MSE ⫽ 1,352.9. In relation to the priming effect, simple effects analyses within each group confirmed the interference observed with identity primes in the associative group, F1(1, 22) ⫽ 9.49, MSE ⫽ 1,994.1, and F2(1, 114) ⫽ 16.29, MSE ⫽ 1,891.9; there was no evidence of priming in the listen group (Fs ⬍ 1). Episodic recognition of primes. The recognition data of 1 participant in the associative group were lost. Among the remaining participants, the percentages of hits minus false alarms were 36% and 15% for the associative and listen groups, respectively, F(1, 45) ⫽ 14.52, MSE ⫽ 30.7.
Discussion The target word frequency effect, with a marked disadvantage for low-frequency relative to medium- and high-frequency words, is consistent with the color-naming results of Experiment 1 and the word-naming literature (Forster & Chambers, 1973). As predicted, the word-naming groups showed identity priming facilitation regardless of prime processing requirement. In addition to crossmodal facilitation of word recognition, activation of the phonological representation of the target word in the identity prime
condition may have facilitated the articulatory preparation of the naming response and thus contributed to the substantial identity priming effect. Also, the long prime–target SOA and the fact that 50% of targets had identity primes presumably led participants to make predictions about the target identity from the prime. The magnitude of frequency and priming effects in the wordnaming task tended to be greater (although not significantly so) for the associative processing group than for the listen group. In addition, there was a frequency attenuation effect in the associative group, with priming effects increasing with decreasing target frequency, and no such effect in the listen group. In the LDT with a study–test design, it has been found that auditory repetition priming effects (Joyce, Paller, Schwartz, & Kutas, 1999) are significantly greater for low-frequency than high- or medium-frequency words (C. A. Joyce, personal communication, May 12, 2000); a similar result has been reported for perceptual identification (Kirsner, Dunn, & Standen, 1989). The fact that the associative group showed a frequency attenuation effect together with a tendency to exhibit a larger priming effect than the listen group may reflect nothing more than the baseline differences between the groups, with the data of the slower associative processing group providing a greater opportunity for priming and frequency effects to be observed. Alternatively, it is possible that the attention to the prime required by the associative prime task caused identity targets to be better predicted. Prediction in the word-naming tasks may have curtailed or facilitated postlexical access procedures that are more time consuming for low-frequency words and thus produced the observed frequency attenuation effect. These procedures may include orthographic verification of the target word identity and preparation of the word’s articulatory program. The color-naming groups showed target frequency effects in the same direction as the word-naming groups. In contrast, and broadly consistent with predictions, their priming results displayed an entirely different pattern, with the listen group showing no priming facilitation and the associative processing group showing interference from identity auditory primes. The interference effect of auditory identity primes, unlike the facilitation from visual identity primes of Experiment 1, did not interact with target frequency. It appears that the interference does not depend on characteristics of words that are thought to covary with normative frequency, such as the time required for lexical processing. It is noteworthy that auditory prime presentation presumably eliminates any facilitatory influence of perceptual fluency. Given that the possibility of predicting identity targets with 50% accuracy did not allow participants to avoid the interference effect, prediction strategies may not be useful in a situation in which interference arises as a result of priming. In view of the absence of a main effect of prime processing group, it appears that the associate generation task was no more disruptive in terms of color naming than was listening to primes. The color-naming latencies of the auditory prime groups were intermediate between those of the two visual prime groups of Experiment 1, with the listen group of Experiment 2 tending to respond more slowly than the read group of Experiment 1, F1(1, 44) ⫽ 4.05, MSE ⫽ 78,758.9, p ⫽ .05, and F2(1, 114) ⫽ 486.37, MSE ⫽ 1,139.7, p ⬍ .01. The fact that only the associative processing group showed color-naming interference indicates that, at the relatively long
prime–target SOA used here, merely listening to the primes made a short-lived and perhaps weaker contribution to phonological activation of the target. Thus, the results are broadly consistent with the present view that activities promoting activation of the target word phonology exacerbate response conflict in the noncolor–word Stroop task. An alternative hypothesis entertained previously is that the color-naming interference observed here and in studies of associative priming depends on the amount of elaborative or semantic processing of the prime words. However, Experiments 1 and 2 showed that silent generation of associates to the prime, a semantic task, resulted in interference with target word-color naming for auditory but not visually presented identity primes. Therefore, the semantic prime processing task used here may be important because it involves internal speech about the prime and associates or other activities that promote phonological activation of the target in the identity condition, not because it involves semantic or elaborative processing per se. This inference appears to be at odds with an earlier finding (Davelaar & Besner, 1988) that imageability is the factor that accounts for content words showing more interference than function words in color naming. However, it is possible that imagery also promotes implicit naming of the colored words or associated concepts. Taken together, Experiments 1 and 2 indicate that neither the associate generation task nor auditory prime presentation was sufficient to produce color-naming interference with identity relative to unrelated primes. In the associative group of the present experiment, reading the target in the identity condition presumably acted in concert with the prime processing task to cause sufficient activation of the target phonology to interfere with preparation of the color-naming response. However, it must be noted that auditory prime presentation for the listen group may have interfered with naming the target color had the prime–target SOA been short. Interference in the standard color–word Stroop task has been reported when the base word was presented auditorily while participants named the color of a color patch, but only when the auditory distractor was presented simultaneously with or very shortly after the visual stimulus (Elliott, Cowan, & Valle Inclan, 1998; Shimada, 1990).
Experiment 3 If the concurrent availability of the phonological representation of the target word disrupts preparation or execution of the colornaming response, then a prime task requiring the production of a phonological representation of the target should produce a substantial interference effect in color naming for identical targets. The additional prime task used in Experiment 3 was generating words that rhymed with the prime. There is abundant evidence that a related task, making rhyme judgments, requires phonological coding of to-be-judged items. For example, concurrent articulation has been found to impair rhyme judgments about visually presented items (Besner, Davies, & Daniels, 1981; Brown, 1987; R. S. Johnston & McDermott, 1986; Wilding & White, 1985), as has shadowing an auditory digit list (Kleiman, 1975). In Experiment 3, identity priming in color naming was assessed in three prime processing tasks with auditory primes: listen only, silently generate an associate of the prime, and silently generate a rhyme of the prime. Because rhyme generation was expected to
activate the pronunciation of the prime more strongly than associate generation, it was of interest whether rhyme generation would produce color-naming interference with visual primes. Consequently, the effect of rhyme generation with visual primes was assessed in a fourth condition. As in Experiments 1 and 2, a recognition memory test was conducted after the priming phase to provide an assessment of the effects of the prime processing tasks. For the identity priming effects in the auditory prime groups, the amount of color-naming interference for identity relative to unrelated primes was predicted to increase in order over the listen, associative, and rhyme tasks. For episodic recognition of primes, levels-of-processing research indicates that if a standard recognition test is used (Morris, Bransford, & Franks, 1977), associative generation, a semantic task, produces better recognition accuracy than rhyme generation (Craik & Tulving, 1975; Schnur, 1977); listening-only tasks are likely to produce the poorest recognition performance. Therefore, a partial dissociation of priming effects on color naming and recognition memory for primes was predicted, in that episodic recognition accuracy was expected to be higher for the associative group than the rhyme group, whereas priming effects were expected to be larger for the rhyme than the associative group.
Method Participants. A total of 96 participants were selected from the pool used previously. The data of an additional 5 participants were discarded because more than 25% of trials were lost through microphone failures. Materials and design. Twenty unrelated prime–target pairs were selected from each of the high-frequency and low-frequency target sets of Experiment 2 on the basis that there were rhymes to the primes and targets. An additional 20 high- and 20 low-frequency targets with unrelated primes were taken on the same basis from the corpus of Kucera and Francis (1967). The Kucera and Francis (1967) corpus was searched for one- and two-syllable rhymes with stress on the final syllable. Mean numbers of rhymes were 11.0 for high-frequency targets and 11.2 for low-frequency targets. The mean length of targets and primes was 4.5 letters, and the mean and median frequencies were 576 and 429, respectively (range: 201–2,216) for the high-frequency targets and 2.9 and 2, respectively (range: 1– 6), for the low-frequency targets. The unrelated primes had a mean frequency of 136 (range: 1– 897), and they were matched on mean length, mean frequency, and mean number of rhymes (10.9) over target frequency sets. Targets were counterbalanced over prime type and color in eight lists, as described previously. Twenty-four unrelated primes were chosen from each target frequency set for the recognition test of primes. For the first counterbalance set, the first 12 primes from each subset were presented during the priming phase and the remainder were not (vice versa for the second counterbalance set). The primes were presented in a random sequence on a response sheet as in Experiment 2. Procedure. For the auditory groups, the primes were recorded by a female Australian speaker in stereo, at 16 bits and a 22050-Hz sampling rate, on an IBM-compatible PC with a Soundblaster sound card and software. They were played from the PC at a volume of 65–70 dB through two small speakers placed at chest height in front of and at either side of the participant. Participants were instructed to listen to the prime or silently generate an associate or rhyme, according to condition. Examples were given to participants in the rhyme and associate groups, and participants were instructed to abandon the prime task as soon as the target word appeared. For the visual rhyme group, primes were presented in uppercase white letters 11 mm high. Targets were presented in lowercase letters 8 –11 mm high, in red, green, blue, or yellow. The interval from prime onset to target onset was 2.5 s, as in Experiment 2. Twelve practice trials preceded the 80 test trials, which were presented in two blocks of 40.
PHONOLOGY AND COLOR NAMING
Results Mean latencies for each prime processing group and condition are shown in Figure 4. Error rates were low (0.7% overall), with no consistent effect of prime type or target frequency. A mixed-factor Group ⫻ Target Frequency (within participants) ⫻ Prime Type (within participants) ANOVA for the four groups showed only marginally significant effects of frequency, F1(1, 92) ⫽ 4.53, p ⬍ .05, MSE ⫽ 2.4, and F2(1, 78) ⫽ 3.18, p ⫽ .08, MSE ⫽ 1.2, and the interaction of group with frequency, F1(3, 92) ⫽ 2.96, p ⬍ .05, MSE ⫽ 2.4, and F2(3, 234) ⫽ 2.57, p ⫽ .06, MSE ⫽ 1.2, reflecting a higher error rate for high-frequency words that was evident only in the auditory rhyme and listen groups. The means over the latter two groups were 1.1% for high-frequency targets and 0.4% for low-frequency targets. Latency data were processed as described previously, with a loss in addition to errors of 4.8% of trials because of microphone failure and 1.2% of trials because of extreme latencies (range: 0.8% to 1.4% over groups). The three auditory prime groups were compared in Counterbalance Set (two levels) ⫻ Prime Processing ⫻ Target Frequency ⫻ Prime Type ANOVAs, with three planned comparisons (at ␣ ⫽ .05) assessing pairwise group differences in identity priming. The visual prime–rhyme group was compared with the auditory prime–rhyme group in separate Counterbalance ⫻ Prime Modality ⫻ Target Frequency ⫻ Prime Type ANOVAs. An initial set of analyses was conducted with the additional factor item set (old items from Experiment 2 vs. new items). Because there was no significant main or interactive effect of this variable, items were pooled for the analyses reported below. Analysis of the data for the auditory prime groups revealed significant main effects of target frequency, with shorter latencies for high-frequency targets, F1(1, 66) ⫽ 77.51, p ⬍ .01, MSE ⫽ 985.2, and F2(1, 76) ⫽ 38.45, p ⬍ .01, MSE ⫽ 3,910.0, and prime type, with interference by identity primes, F1(1, 66) ⫽ 19.30, p ⬍ .01, MSE ⫽ 2,378.8, and F2(1, 76) ⫽ 20.19, p ⬍ .01, MSE ⫽ 2,947.2. The main effect of group was reliable in the analysis by items and marginal in the analysis by participants, F1(2, 66) ⫽ 2.62, p ⫽ .08, MSE ⫽ 37,118.7, and F2(2, 152) ⫽ 74.73, p ⬍ .01, MSE ⫽ 1,819.9. The Group ⫻ Prime Type interaction approached significance in the analysis by participants and was reliable in the
Figure 4. Mean correct color-naming latencies for the three auditory prime processing groups (listen, associative, and rhyme) and the visual rhyme group in Experiment 3.
analysis by items, F1(2, 66) ⫽ 2.83, p ⫽ .07, MSE ⫽ 2,378.6, and F2(2, 152) ⫽ 4.57, p ⬍ .05, MSE ⫽ 1,653.4. There was a Counterbalance ⫻ Prime Type interaction reflecting a larger interference effect in one set than the other, F1(1, 66) ⫽ 4.14, p ⬍ .05, MSE ⫽ 2,378.5, and F2(1, 76) ⫽ 17.24, p ⬍ .01, MSE ⫽ 1,653.4. Figure 4 shows that, consistent with predictions, the magnitude of the interference produced by identity primes was smallest for the listen group (12 ms), intermediate for the associative group (20 ms), and largest for the rhyme group (44 ms). Planned comparisons on the priming effect revealed that the rhyme group displayed greater priming interference than the listen group, F1(1, 44) ⫽ 4.93, p ⬍ .05, MSE ⫽ 2,490.4, and F2(1, 76) ⫽ 8.24, p ⬍ .01, MSE ⫽ 1,616.5, but that the associative and listen groups did not differ on magnitude of priming effect (F1 and F2 ⬍ 1); the difference between the associative and rhyme groups in priming was reliable in the item but not the participant analysis, F1(1, 44) ⫽ 2.40, p ⫽ .13, MSE ⫽ 3,040.6, and F2(1, 76) ⫽ 4.66, p ⬍ .05, MSE ⫽ 1,915.0. In contrast to the auditory rhyme group, the visual rhyme group showed only a small priming interference effect of 10 ms. The Prime Modality ⫻ Target Frequency ⫻ Prime Type ANOVAs on the mean latencies for the rhyme groups revealed main effects of target frequency, F1(1, 44) ⫽ 27.33, p ⬍ .01, MSE ⫽ 1,243.6, and F2(1, 76) ⫽ 16.64, p ⬍ .01, MSE ⫽ 3,629.7, and prime type, F1(1, 44) ⫽ 14.04, p ⬍ .01, MSE ⫽ 2,511.5, and F2(1, 76) ⫽ 27.79, p ⬍ .01, MSE ⫽ 2,029.5, that paralleled the results for the auditory group analysis. The group main effect (slower latencies for the auditory group) was reliable only in the item analysis, F1 ⬍ 1, MSE ⫽ 51,230.7, and F2(1, 76) ⫽ 8.41, p ⬍ .01, MSE ⫽ 2,457.1. The two main effects were qualified by a pair of two-way interactions: a Modality ⫻ Prime Type interaction reflecting greater priming for the auditory group, F1(1, 44) ⫽ 5.54, p ⬍ .05, MSE ⫽ 2,511.5, and F2(1, 76) ⫽ 7.20, p ⬍ .01, MSE ⫽ 2,195.2, and a Target Frequency ⫻ Prime Type interaction indicative of greater interference for high- than low-frequency targets, F1(1, 44) ⫽ 9.24, p ⬍ .01, MSE ⫽ 1,101.8, and F2(1, 76) ⫽ 7.17, p ⬍ .01, MSE ⫽ 2,029.5. From the means displayed in Figure 4, it can be seen that although the three-way interaction of modality, frequency, and prime type was not significant, the primary source of the two-way interactions just described was the inconsistency of priming effects over target frequency for the visual but not the auditory prime group. The interactions were examined in separate ANOVAs for each group. In the auditory group, the main effects of frequency and prime type were reliable (all ps ⬍ .01), and there was no Prime Type ⫻ Frequency Interaction. In the visual group, there was a main effect of frequency, F1(1, 22) ⫽ 8.67, p ⬍ .01, MSE ⫽ 856.3, and F2(1, 76) ⫽ 4.46, p ⬍ .05, MSE ⫽ 2,655.6; no main effect of prime type; and a significant disordinal Frequency ⫻ Prime Type interaction, F1(1, 22) ⫽ 7.07, p ⬍ .05, MSE ⫽ 1,550.5, and F2(1, 76) ⫽ 7.03, p ⬍ .05, MSE ⫽ 1,874.0, with interference for high-frequency targets and facilitation for low-frequency targets. A simple main effects analysis of the latter Frequency ⫻ Prime Type interaction showed that the interference for high-frequency targets was reliable, F1(1, 22) ⫽ 6.58, p ⬍ .05, MSE ⫽ 1,804.1, and F2(1, 38) ⫽ 7.21, p ⬍ .05, MSE ⫽ 2,439.4, whereas the facilitation for low-frequency targets was not, F1(1, 22) ⫽ 1.82, p ⬎ .05, MSE ⫽ 843.6, and F2 ⬍ 1. In summary, there was a consistent interference effect over target frequency for the auditory
prime group, but interference was found only for high-frequency targets in the visual group. In the episodic recognition test of primes, participants in the four groups made an average of 16.5% false alarms in the recognition test of primes and new items. Consequently, in scoring, the number of false alarms made by each participant was subtracted from the number of hits. The mean percentage of hits minus false alarms for the visual rhyme group was 34%, and the percentages for the auditory prime groups were 48% (associative), 37% (rhyme), and 21% (listen). The two rhyme groups did not differ (F1 ⬍ 1), and their data were pooled in subsequent analyses. Planned comparisons confirmed that the combined rhyme groups and the associative group each had episodic recognition superior to that of the listen group, F1(1, 70) ⫽ 10.91, p ⬍ .01, MSE ⫽ 311.9, F1(1, 46) ⫽ 22.13, p ⬍ .01, MSE ⫽ 397.7, respectively. The associative group showed superior recognition to the rhyme groups, F1(1, 70) ⫽ 6.19, p ⬍ .05, MSE ⫽ 403.6.
Discussion The results of Experiment 3 were broadly consistent with expectations, in that substantial interference was produced by auditory identity primes when participants were asked to generate a rhyme of the prime word. As in Experiment 2, the interference effect produced by auditory identity primes did not interact with target frequency. In addition, although the Group ⫻ Prime Type interaction for the auditory prime groups fell short of reliability in the participant analysis, the ordering of interference in the auditory groups was as predicted, with most interference for the rhyme group and least for the listen group. Over the visual prime and the three auditory prime groups, episodic recognition of the primes varied in the predicted manner, with best memory for primes in the associative processing group and worst for the listen group. There was partial confirmation of the prediction that visual identity primes would interfere with target color naming when participants performed a prime task requiring activation of a phonological representation of the prime. Rhyme generation to visual primes produced identity priming interference, but only for highfrequency targets. This result contrasted with that for the auditory rhyme group, in which interference was observed over the two levels of target frequency. Two major results of Experiment 2 for the associative and listen groups were replicated in Experiment 3: the effect of target frequency, with faster color naming for high- than low-frequency targets, and the superiority of the associative group in terms of memory for the primes. The difference in the magnitude of interference observed in the associative and listen groups was smaller in Experiment 3 than in Experiment 2, and the Group ⫻ Prime Type interaction was not reliable in the present study. The present experiment was less sensitive in having only two levels of target frequency, whereas Experiment 2 had three levels. Analysis of the results for high- and low-frequency targets of Experiment 2 revealed that the Group (associative vs. listen) ⫻ Prime Type interaction fell short of reliability in the participant analysis. Experiment 3 provided additional information about the effects of rhyme generation. The auditory and visual rhyme generation groups had comparable episodic recognition of primes, and they had poorer memory for primes than did the associative processing group. Nevertheless, rhyme generation, but not associate genera-
tion, produced greater interference in color naming than did listening to the primes. The results for the auditory and visual prime groups of Experiments 1, 2, and 3 converge on the conclusion that, relative to a semantic prime task, rhyme generation increases interference or reduces facilitation from identity primes in color naming. That is, for visual primes with high-frequency targets, associative processing of the prime produced a (nonsignificant) facilitation of 11 ms in color naming (Experiment 1), whereas rhyme generation to visual primes produced interference of 32 ms (Experiment 3). The facilitation by visual identity primes for low-frequency targets was reduced from 45 ms under associate generation to primes (Experiment 1) to 11 ms under rhyme generation to primes (Experiment 3). With auditory prime presentation, averaged over high and low target frequency, associate generation produced interference rates of 28 ms and 20 ms in Experiments 2 and 3, respectively, whereas rhyme generation produced an interference effect of 45 ms (Experiment 3). The tendency of rhyme generation to promote identity priming interference in color naming, yet poorer memory for primes than does associative prime processing, points to a specific effect beyond general attention to primes or degree of elaborative processing of primes. Thus, the results of Experiment 3 are consistent with the present view that interference arises from the joint contribution of rhyme generation and target word reading to the activation of a phonological or articulatory code of the target word, concurrent with preparation of the vocal color-naming response. One somewhat puzzling result is the failure of rhyme generation in response to visual primes to produce interference with lowfrequency targets when interference was observed in the same group for high-frequency targets. It is possible that the difficulty of the rhyme generation task was confounded with target frequency and prime type, given that the frequency of unrelated primes was matched over the two target frequency sets and therefore was lower relative to high-frequency identity primes and higher relative to low-frequency identity primes. However, militating against such item-based confounds are that primes were matched on the number of rhymes over conditions and that, in the same item set, identity priming interference was observed in the auditory rhyme group for both high- and low-frequency targets. The explanation favored here is based on an assumption that the magnitude of priming effects for the visual prime group is determined jointly by identity priming facilitation of word recognition and interference with color naming by the phonological activation of the target. It is clear from Experiment 1 that regardless of the prime task, the facilitation effects of visual identity primes are substantial and frequency dependent, with larger facilitation for low-frequency targets. It was suggested that a differential contribution of a perceptual fluency effect for low-frequency targets was responsible for this frequency dependency. That is, participants may be more easily able to ignore a target that is judged to be the same as a prime, and perceptual fluency (Jacoby & Dallas, 1981) may facilitate this “same” judgment, particularly for lowfrequency targets. It is possible, then, that the differential results for high- and low-frequency targets reflect the additive combination of an interference effect that is constant over frequency and a visual prime facilitation effect that is greater for low-frequency targets. The present results do not rule out more complex accounts involving the interactive combination of word-processing facilita-
PHONOLOGY AND COLOR NAMING
tion and response conflict or the differential activation of phonology for high- relative to low-frequency words.
Experiment 4 Experiments 1–3 focused on the conditions under which priming manipulations produce interference in target color naming in the Warren task. The final two experiments reported here addressed facilitation of color naming by factors thought to affect word recognition speed. It was argued in the introduction that, in the absence of factors promoting response conflict or processing of word stimuli at the expense of the color-naming task, facilitation of target word recognition can be expected to facilitate color naming. Thus, in contrast to the traditional interpretation of the priming interference effects demonstrated by Warren and others, it is proposed here that facilitation of color naming is the more valid index of facilitation of word recognition. Some evidence to support this contention has been reported previously as well as in the present experiments, namely the identity priming facilitation with visual primes and the consistent target word frequency effect, with faster color naming for high- than low-frequency targets. Although the results for word frequency and identity priming neatly parallel findings in commonly used word recognition tasks such as the LDT and naming, it is not yet established that the basis of the facilitation is the same in the color-naming task and other tasks. A complication is that some effects in word recognition undoubtedly are task specific and may reflect strategies or characteristics of materials that have little to do with visual word recognition. For example, relevant to attentional processing in the color-naming task is the mismatch theory of W. A. Johnston and Hawley (1994), who proposed that responses to expected stimuli are guided by conceptual top-down processes that dampen detailed perceptual analysis of these stimuli, with the consequence that resources are available for prompt perceptual analysis of unexpected inputs. In light of this fact, a number of authors (e.g., Andrews, 1992) have suggested that converging evidence across word recognition tasks is the most useful way to diagnose which effects are related to word recognition. In the spirit of this aim, Experiments 4 and 5 were designed to provide additional evidence that factors affecting word recognition produce color-naming results that converge with those from more conventional word recognition tasks. Experiment 4 addressed the word frequency effect and compared color-naming latencies to high-frequency and lowfrequency words in a single-word presentation format so that the frequency effect could be assessed in the absence of the effects of prime words and prime tasks.
Method Participants. A total of 48 introductory psychology students who spoke English as their first language participated for course credit. Materials. Eighty words 4 –5 letters long (mean length: 4.5 letters) were selected from the Kucera and Francis (1967) corpus. Forty had frequencies below 10 (mean: 4), and the remaining 40 were of high frequency (range: 185– 895; mean: 398). Most of the words had not appeared in the stimulus sets for the previous experiments. The words were rotated through four colors (red, blue, green, and yellow) to generate four lists, with the trial sequence randomized except that no more than two successive trials had the same letter color. Procedure. Participants were tested individually on an IBM-compatible computer. Words were presented in lowercase letters 8 –11 mm
high on a color monitor with a dark background. On each trial, a ready signal (⫹⫹⫹) was displayed in the center of the screen for 500 ms, the screen was dark for 250 ms, and the word was presented in red, blue, green, or yellow lettering until the participant’s vocal response occurred and the experimenter pressed a key to indicate response accuracy or microphone failure. The intertrial interval was 2,500 ms. Twelve practice trials preceded the test list, which was presented in two blocks.
Results After preprocessing, 7.5% of word trials were lost, of which 5.8% involved microphone failures and 1.3% involved extreme latencies with respect to a participant’s overall correct mean latency. Errors were rare, with means of 0.2% for high-frequency words and 0.5% for low-frequency words. Analysis of the error data indicated that the higher error rate for low- than highfrequency words was reliable by participants, F1(1, 47) ⫽ 6.01, MSE ⫽ 0.5, but only marginally so by items, F2(1, 78) ⫽ 3.89, p ⫽ .05, MSE ⫽ 0.7. In the latency data, a one-way ANOVA revealed a significant advantage for high- over low-frequency words, with means of 633 and 645 ms, respectively, F1(1, 47) ⫽ 17.54, MSE ⫽ 186.5, and F2(1, 78) ⫽ 9.83, MSE ⫽ 323.4.
Discussion The significant latency advantage for high-frequency words was consistent with predictions and supported previous evidence indicating that when word frequency effects occur in the non-color– word Stroop task, they are in the same direction as in the LDT, naming, and perceptual identification tasks. The observation by Klein (1964) that high-frequency words support slower color naming than low-frequency words is in direct contrast to the present results and those reported previously (Burt, 1994, 1999). Klein’s word frequency effects were replicated in a study reported (in German) by Effler (1977). Potentially critical differences between Klein’s (1964) study and the present study are the composition of the stimulus lists and whether a single-trial or a list reading format was used. In the present study, there were no color names or color-related words, and each letter string was seen only once by each participant during the color-naming task. In the Klein study, color names and words with a color association, such as lemon, constituted half of the stimuli, and each letter string was repeated 20 times during the experiment. Furthermore, interference was calculated from times taken to complete a block of trials for a single condition presented on a card. Therefore, participants had the opportunity to adjust their strategies to trial types. Finally, a salient difference between Klein’s and the present study is that at least three of the four low-frequency words used by Klein (sol, belot, eft, and abjure) are so rare in modern English usage that university students may not be familiar with them. The rare words in the present study were selected as likely to be familiar to university students. The distinctiveness of the rare words would presumably be greatly reduced by repetition, and when participants process words largely outside their vocabulary there may be weaker activation of phonological or semantic information in comparison with highfrequency words.
Experiment 5 In the LDT, as indicated previously, a well-replicated finding is that visually presented words repeated during the session are
responded to more quickly and perhaps more accurately than unrepeated words (Balota & Spieler, 1999; Forster & Davis, 1984; McKone, 1995; Monsell, 1985; Scarborough, Cortese, & Scarborough, 1977), an effect also observed in perceptual identification (Feustel, Shiffrin, & Salasoo, 1983) and naming (Balota & Spieler, 1999). The benefit occurs even when many items intervene between repetitions. As discussed previously, in cases in which a response is required to the word on its first presentation, the repetition effect often is larger for low-frequency than highfrequency words (Forster & Davis, 1984; Jacoby & Dallas, 1981; Norris, 1984; Ostergaard, 1998; Rajaram & Neely, 1992; Scarborough et al., 1977). Because the interaction with frequency does not occur in the identity priming paradigm when primes are masked and cannot be reported by participants (Bodner & Masson, 1997; Forster & Davis, 1984), it has been argued that it does not arise within the processes that must be executed for word identification. Given that there has been little support for an account offered by Forster and Davis in terms of episodic distinctiveness on the recollection of low-frequency words (Duchek & Neely, 1989; Kinoshita, 1995), more viable accounts are that influences of familiarity, task-specific and item-specific practice effects, or the duration of orthographic checks later in lexical processing are greater for low- than high-frequency words. Regardless of whether the frequency attenuation effect is informative about lexical access per se, this effect (together with the repetition benefit effect) might be expected to occur in the noncolor–word Stroop task to the extent that it is sensitive to the factors affecting performance in traditional word recognition tasks. The primary purpose of the final experiment was to assess longer term repetition effects on color naming. Repeated items were classified in the LDT before the color-naming phase of the experiment. Requiring a lexical decision to words on their first presentation was expected to increase participants’ attention to words and, hence, increase the power of the experiment to show repetition effects. The color-naming task is relatively insensitive to factors that produce large effects in the LDT and small to moderate effects in naming, as is evident in that the frequency effect in Experiment 4 was only 12 ms. In a similar manner, identity priming effects at relatively short prime–target SOAs are very small in the color-naming task in comparison with the effects observed in the LDT (Feustel et al., 1983; Forster & Chambers, 1973; Kirsner et al., 1987; Monsell, 1985) and word-naming tasks (Burt, Mardle, & Humphreys, 1996; Durso & Johnson, 1979). Therefore, it is likely to be difficult to demonstrate longer term repetition effects in color naming. However, given the common observation of frequency attenuation, with a substantial benefit observed for low-frequency words, there is a possibility that a measurable color-naming facilitation will be produced at least for low-frequency words. An additional feature of Experiment 5 was the presentation of nonword stimuli, some of which had been classified in the LDT. The prediction that follows from findings in the LDT and the naming task is straightforward; that is, color-naming responses to nonwords will be slower than those to words, and nonwords will show weak repetition effects relative to words (den Heyer, Goring, Gorgichuk, Richards, & Landry, 1988). In contrast, Klein’s (1964) work with non-color words showed that color-naming responses were faster to nonwords than to words. However, Klein used unpronounceable and thus unwordlike letter strings, whereas the
nonword stimuli in the current experiment were orthographically acceptable in English. Bakan and Alperson (1967) found that the interference (assessed against color-patch naming) produced by nonsense syllables increased monotonically with their pronounceability but that color-naming interference was less for nonwords than for unisyllabic words. In the standard Stroop task in which color words are part of the stimulus list, color-naming latencies for nonwords typically do not differ from those for non-color, unrelated words (Hintzman et al., 1972; Redding & Gerjets, 1977). Color-naming latencies were recorded to nonwords and high-, medium-, and low-frequency words, half of which had previously served as items in the LDT. It was predicted that long-term repetition benefits would occur for low-frequency words, with perhaps smaller effects observable for other words.
Method Participants. A total of 48 introductory psychology students who spoke English as their first language participated for course credit. Materials. One hundred twenty words and 120 orthographically legal nonwords 4 –5 letters long (mean length: 4.2 letters) were the stimuli. Within the word set, there were 40 high-frequency words (mean and median frequencies: 263 and 217, respectively; range: 123– 897; Kucera & Francis, 1967), 40 medium- to high-frequency words (mean frequency: 62; median: 61; range: 41–90), and 40 low-frequency words (mean frequency: 1.5; median: 1; range: 1–3). Most of the words had not appeared in the stimulus sets for Experiments 1– 4. Within each frequency category, words were divided into two subsets of 20 words each, matched approximately on length and frequency. The nonwords were divided into two sets of 60 items, of which 20 were allocated to the repetition test and 40 were fillers. For the initial lexical decision phase, one stimulus list was constructed from the first of the three subsets from each frequency category and the first of the nonword subsets, and a second list was made from the second subset from each of these item sets. Thus, each list for the LDT consisted of 60 words and 60 nonwords. For the second, color-naming phase, the total set of 120 words was presented with the 40 test nonwords. Each word in the stimulus list was rotated through the four colors red, blue, green, and purple to make four sets, and each set was presented with the trial sequence randomized except that a color did not occur more than twice in succession. Twelve participants were allocated to each of the four color-naming sets. Within each group of 12, 6 participants had the first lexical decision list, and 6 had the second. Thus, for the 120 words and the 40 test nonwords, each item was old for half of the participants and new for the remaining participants. Procedure. Participants were tested individually on the apparatus used in Experiments 1 and 2. In the first, lexical decision phase, a plus sign was displayed in the center of the screen for 500 ms, the screen was blank for 800 ms, and the letter string was displayed in the center of the screen in double height lowercase white letters. Participants rested their right and left index fingers on the right and left buttons of a response box and were asked to press the right button for words and the left for nonwords (vice versa for left-handed individuals). They were instructed to respond as quickly as possible while maintaining high accuracy. A button press cleared the screen and started a 3-s intertrial interval. There was an initial practice block of eight words and eight nonwords, randomly sequenced. In the second, color naming phase, participants wore a headset microphone. A warning signal (⫹) was displayed in the same format and temporal arrangement as in the LDT, and participants named the color of words presented singly in double height, lowercase, colored lettering, as described previously. After the participants responded and the experimenter pressed a key to score the response, the letter string was removed, and a 3.5-s interval was interpolated before the next trial. A practice block of 15 trials was given before the test block.
PHONOLOGY AND COLOR NAMING
Results Before analysis of latency data for the LDT and color naming, errors were removed, as were latencies in excess of three standard deviations from a participant’s correct mean for words and nonwords separately (lexical decision) and for words and nonwords within each repetition condition separately (color naming). In addition, for color naming, microphone failures were excluded. Lexical decision. Pooled over word sets but excluding fillers, mean lexical decision latencies were 704 ms for words and 820 ms for nonwords. There was a substantial frequency effect for the word stimuli, mainly evident as slower latencies for low- than for high- or medium-frequency words, with means of 658 ms (high), 672 ms (medium), and 781 ms (low), F1(2, 94) ⫽ 138.49, MSE ⫽ 1578.6, and F2(2, 117) ⫽ 44.27, MSE ⫽ 4,916.3. Color naming. Error rates were low, at less than 0.2% overall, and were not analyzed. An additional 1.2% and 1.0% of trials were lost as a result of microphone failures and extreme times, respectively. A one-way lexical-status ANOVA revealed a significant latency advantage for words over nonwords, F1(1, 47) ⫽ 18.95, MSE ⫽ 288.7, and F2(1, 158) ⫽ 14.57, MSE ⫽ 475.3. Mean latencies were 631 ms for words and 646 ms for nonwords. Mean latencies for old and new words are shown in Table 1. Repetition effects were analyzed for words and nonwords separately. For nonwords, there was no repetition effect (F1 and F2 ⬍ 1), with latencies for old and new items of 645 ms and 648 ms, respectively. For words, a Counterbalance Set (two levels) ⫻ Frequency (high, medium, or low) ⫻ Repetition (old vs. new) ANOVA revealed a main effect of frequency, with latencies fastest for high-frequency words and slowest for low-frequency words, F1(2, 92) ⫽ 6.87, MSE ⫽ 612.1, and F2(2, 114) ⫽ 3.51, MSE ⫽ 974.5. There was no main effect of repetition, F1(1, 46) ⫽ 1.13, MSE ⫽ 642.3, and F2 ⬍ 1, but the Frequency ⫻ Repetition interaction was reliable, F1(2, 92) ⫽ 3.93, MSE ⫽ 670.4, and F2(2, 114) ⫽ 4.01, MSE ⫽ 551.1. Simple effects analysis confirmed the predicted repetition benefit for low-frequency words, with old items having shorter response latencies than new items, F1(1, 46) ⫽ 11.94, MSE ⫽ 441.0, and F2(1, 38) ⫽ 8.16, MSE ⫽ 502.4. As is evident from Table 1, there was no repetition benefit for high- or medium-frequency words.
Discussion The results of Experiment 5 confirm that the non-color–word Stroop task displays effects of lexicality and word frequency similar to, albeit smaller than, those typically observed in the LDT. That is, averaged over frequency, words are responded to more
Table 1 Experiment 5: Mean Color-Naming Latencies (in Milliseconds; With Percentages of Error) for Old and New Words as a Function of Item Frequency Frequency Repetition status
624 (0.3) 624 (0.0)
636 (0.1) 631 (0.0)
629 (0.4) 644 (0.2)
quickly than nonwords, and high- and medium-frequency words are responded to more quickly than rare words. Although there was no main effect of repetition for words, there was a repetition benefit for low-frequency words. The finding that nonwords produced slower color-naming latencies than words is inconsistent with the studies of Klein (1964) and Bakan and Alperson (1967). In the Klein study, each letter string was presented 20 times, as noted previously; the same was true in the Bakan and Alperson study. This repetition may allow participants to process nonwords as familiar but meaningless stimuli. A similar explanation was offered in Experiment 4 for Klein’s finding that very rare words produced less color-naming interference than common words. Also, in the studies of Klein and Bakan and Alperson, strategic factors may have been introduced by the test procedure, wherein conditions were blocked by lists in a list reading format. Conversely, their list presentation method is susceptible to interference from preceding stimuli, and such interference may be greater in word lists by virtue of their semantic content. The fact that the repetition benefit was confined to lowfrequency words does not indicate that word repetition has fundamentally different effects in color naming and other word recognition tasks. The small magnitude of the repetition effect in color naming was anticipated and is not surprising in that, in the colornaming paradigm, word reading is a secondary task. Given that repetition effects typically are larger from low-frequency words, it is likely that the present results reflect merely the low sensitivity of the color-naming task to the characteristics of non-color words. The similarity of the effects of word frequency, lexicality, and repetition over color naming and other word recognition tasks lends some credence to arguments that the underlying mechanisms are similar. It was suggested previously that perceptual fluency for a repeated word may play a role in the frequency attenuation effect in identity priming. In the longer term repetition paradigm of the present experiment, perceptual fluency may have incremented the perceived familiarity of a word, especially a low-frequency word, and thereby reduced the salience of the identified base word. This frequency-dependent, perceptual-fluency-based increment in familiarity may likewise be a cause of the frequency attenuation effect in other word recognition tasks, particularly the commonly used LDT, wherein familiarity appears to affect the decision component of processing (Balota & Chumbley, 1984).
General Discussion The five experiments reported here focused on the interpretation of color-naming interference and facilitation observed in Warren’s (1972, 1974) non-color–word Stroop task, which has served as a vehicle for studying lexical processing. Three experiments examined the conditions producing interference and facilitation under identity priming at a long prime–target SOA. Taken together, these studies showed that, over silent reading and silent associate- and rhyme-generation prime tasks, visually presented primes that were lexically identical to their targets facilitated naming the color of the target lettering, except that interference was observed when rhyme generation involved primes of high-frequency targets. Facilitation was greater for low- than high-frequency targets. When primes were presented auditorily, identity primes had no effect on color naming unless participants were required to silently generate
associates or rhymes to the primes, in which case color-naming interference was observed at a magnitude that was similar over levels of target frequency. The final two experiments assessed whether the effects of target frequency, item lexicality, and item repetition reported in the word recognition literature generalized to the color-naming task. The answers to these questions generally were affirmative. High-frequency words showed faster color naming than low-frequency words in the single-word trial arrangement of Experiment 4, replicating the results averaged over prime type in the three identity priming experiments. In the final experiment, color naming was faster to words than nonwords, and a repetition benefit was observed for low-frequency words previously responded to in the LDT. A primary aim of the present research was to discover under what conditions, and why, factors expected to facilitate recognition of the base word of a Stroop stimulus produce interference in color naming. The traditional horse race account endorsed by Warren (1972, 1974) and many subsequent researchers cannot explain the pattern of facilitation and interference effects that has resulted from priming manipulations directed at the base word. That is, in the non-color–word Stroop task, facilitation of word recognition may result in facilitation of color naming (Burt, 1994, 1999; Dosher & Corbett, 1982; C. M. MacLeod, 1996; Whitney, 1986). Therefore, the interference that results from manipulations thought to facilitate word recognition is unlikely to be the product of faster word processing per se. The approach taken in the present and recent work (Burt, 1999) is that these manipulations may concomitantly enhance response conflict, or otherwise impair selective attention, and thus produce interference. Taken together with previous research, the present findings support the generalization that priming interference in color naming in the non-color–word Stroop task is not observed unless priming manipulations involve semantic processing of primes, a meaning-based response to the prime (Burt, 1994, 1999; Henik et al., 1983; Parkin, 1979), or auditory prime presentation (Conrad, 1974; Oden & Spira, 1983; Warren, 1972, 1974; Whitney et al., 1985).
Explaining the Priming-Based Interference Effect A number of possible explanations of the color-naming interference sometimes produced by auditory identity primes were canvassed in the introduction of this article. An account based on diversion of resources from color naming through differential elaborative or semantic processing of identity primes (and hence targets) can be rejected. Because Experiment 1 of the present series showed that identity primes produced color-naming facilitation rather than interference when associative prime processing was required, it is unlikely that semantic or elaborative processing is sufficient for the interference effect. Furthermore, because Experiment 3 showed that rhyme generation produced substantial interference under some conditions, even though rhyme generation produced poorer episodic memory for primes than that produced by semantic processing, it is unlikely that a high degree of semantic or elaborative processing is necessary for interference. In summary, semantic elaboration of prime words appears to be neither necessary nor sufficient for the interference effect. It appears that semantic processing exerts its effects on color naming through other effects on processing of the target word.
Another alternative raised previously was that difficulties associated with switching between prime tasks and color naming may be responsible for the interference effects observed in the noncolor–word Stroop task. Monsell (1996) suggested that the major source of color-naming interference in the non-color–word Stroop task is that with colored words or wordlike letter strings, the participant’s task set becomes word reading rather than color naming. This provides a useful heuristic for understanding why the stimulus lists in the present experiments revealed effects of lexicality, word frequency, and word repetition similar to those reported in more conventional word recognition tasks, whereas in previous studies (e.g., Klein, 1964) very different mixtures of stimuli, containing many unpronounceable letter strings and frequently repeated letter strings, did not. However, with respect to the priming results, an experiment-wide set for word reading cannot explain the effects of prime task and modality observed in Experiments 1–3. Although it is possible that switching from prime task to color naming was a factor in the interference, it is difficult to provide other than a post hoc account of the difference in interference produced for rhyme and associate generation prime tasks. There is no principled basis for assuming that there is a greater switching cost from rhyme generation to color naming than from associate generation to color naming. Phonology and articulatory influences. A simple and viable explanation of the increase in interference produced by priming is that it depends on phonological activation of the base word. On this account, the nature of the prime task is implicated in colornaming interference to the extent that the task directly or indirectly fosters activation of the target (base-word) phonology. Also, as discussed in relation to Experiment 2, auditory presentation is assumed to promote phonological activation. A number of factors not examined here may modulate the effect of the prime task on phonological activation of the target word. These factors include prime–target SOA, whether an overt response to the prime is required, and features of the prime task other than semantic versus rhyme processing. According to current theories of speech production, there is a hierarchical structure of utterance programming that proceeds from abstract plans through phonological activation of words and phonological segments, such as syllables, phonemes, and phonological features, to motor program assembly and execution (Dell & O’Seaghdha, 1992; Mackay, 1987). In principle, the color-naming interference observed here may occur anywhere along a phonological continuum from activation of abstract phonological representations to covert pronunciation of an interfering word. Evidence from natural speech errors and experimental priming manipulations in listening and speech production indicates that speech slips usually arise at the level of phonological representations rather than during the assembly of the motor program (e.g., Dell & Repka, 1992; Fay & Cutler, 1977; Stemberger, 1985). There is support for a close link between listening and speaking, in that classification of speech sounds may suffer interference from similar phonemes in already-prepared utterances (see Meyer & Gordon, 1985), and the inner voice that is heard during reading and mental planning is associated with an articulatory component measurable as electromyographic activity in speech muscles (e.g., Sokolov, 1972). Furthermore, inner speech resembles overt speech in terms of the nature of slips reported when participants are asked to hear tongue twisters in their minds (Dell & Repka, 1992).
PHONOLOGY AND COLOR NAMING
From the preceding considerations, it is likely that the interference observed in color naming in the present research can be explained in terms of the activation of phonological representations without any assumption that motor programs for interfering words have been prepared or are implicated in the interference. Thus, in the present experiments phonological activation of an interfering word may have disrupted the phonological coding of the color-name response, or the availability of multiple phonological representations may have impeded selection of the color name for assembly and execution of the articulatory program. However, there is no evidence in the present studies to rule out the possibility that the delay arises later in response preparation, when assembled articulatory programs for components of the competing responses are selected and sequenced for execution. A factor that is likely to be important in the color-naming interference observed in the non-color–word Stroop task is the vocal response requirement. Color-naming interference in the standard color–word task typically is reduced (although not eliminated) when the vocal response is replaced by a manual response (see the review by C. M. MacLeod, 1991). VanVoorhis and Dark (1995) suggested that difficulty experienced by participants in making a vocal response in the LDT may result from a tendency to pronounce the word or letter string when they were preparing to respond. They supposed, as is assumed here, that participants have a tendency to implicitly pronounce letter strings that they read, with consequent difficulty when a different vocal response is required. This supposition is supported by evidence from the standard color–word color-naming task (e.g., Dennis & Newstead, 1981). A more speculative suggestion is that a vocal response requirement contributes to interference by priming the speech system to pronounce concurrently processed words. This latter idea is broadly compatible with the spread-of-activation metaphor that is central to some current models of speech production (e.g., Dell & O’Seaghdha, 1992), and it also fits naturally with work on attentional set in perceptual–motor tasks (e.g., Monsell, 1996). However, there appears to be no evidence on this possibility. Prelexical versus postlexical phonological activation. An enduring debate in visual word recognition research concerns the role of phonological activation in word identification. Two focal questions are whether activation or generation of a word’s phonology is obligatory and whether phonological activation is a precursor of lexical access and perhaps a causal agent in word identification (see Frost, 1998; Jared, Levy, & Rayner, 1999; Perfetti & Bell, 1991). It is arguable that the present manipulations producing activation of prime phonology exerted their effects on color naming by increasing the probability of prelexical activation of the target word phonology. If so, it might be inferred that such prelexical activation is not obligatory. However, it is difficult to obtain clear evidence of an influence of prelexical phonology without special arrangements to preclude access of phonology postlexically (e.g., using nonwords as phonological primes and masking them to prevent awareness). Also, reading an identity prime at a prime– target SOA as short as 700 ms (Burt, 1994) does not interfere with color naming, although it would be reasonable to suppose, on the basis of the phonological priming literature (e.g., Grainger & Ferrand, 1996; Lee, Binder, Kim, & Pollatsek, 1999; Lukatela,
Frost, & Turvey, 1998; Rouibah, Tiberghien, & Lupker, 1999), that an identity prime preactivates the phonology of the target. Prelexical activation of phonology in skilled adult readers presumably occurs early and outside awareness. Interference in color naming is assumed here to arise at the stage of response selection or preparation, with a delay in color naming depending on the availability of the word as a potential response. Therefore, it is more likely that the influence of phonological activation on color naming is exerted after word identification (and hence lexical access) is essentially complete. If this supposition is correct, then the present findings can shed light only on the obligatoriness of phonological activation occurring after lexical access. On this issue, the present results clearly indicate that phonological activation after word identification is not obligatory in the sense that it is not all or none. It appears that task requirements and perhaps strategic factors modulate the magnitude and duration of phonological activation of a visually presented word.
Facilitation of Color Naming The interaction of target frequency with facilitatory but not interfering effects of primes supports the proposal that the mechanisms underlying the facilitation and interference effects observed here are different. The effects of identity primes and variation in base-word frequency and lexicality are in line with expectations based on the present assumption that, in the absence of factors promoting interference, facilitation of base-word processing facilitates color naming. That is, facilitation of target word identification reduces the duration or the magnitude of the concurrent processing demands associated with identifying the base word and preparing to name the color of its letters. The only problematical result is that no long-term repetition effects were observed for high- and medium-frequency words, whereas these effects are reliably observed in the LDT. It was argued that finding facilitation in Experiment 5 only for lowfrequency words—that is, for the frequency range in which repetition effects typically are larger—is to be expected given the low sensitivity of the color-naming task. An alternative explanation of the color-naming facilitation effects observed in the present experiments is that they are entirely caused by attentional factors that are not relevant to performance in the LDT and other word recognition tasks. For example, longterm repetition and word frequency effects may depend on higher salience or distinctiveness attributed to low-frequency words (as observed by Greene & Thapar, 1994), and the identity priming facilitation reported here and previously may reflect the ability of participants to ignore base words that are predicted by the prime (cf. W. A. Johnston & Hawley, 1994; Kamin, 1969; Wagner, 1976). This kind of attentional explanation may apply to the increased color-naming interference caused by other word characteristics, such as imageability (Davelaar & Besner, 1988). A more radical proposal is that the facilitation effects reflect reductions in phonological activation and thus may be subsumed under the account of the interference effects. This latter account is rejected on the grounds that it is implausible that phonological activation is lower for high-frequency words than low-frequency words or lower for words than nonwords. The preceding task-specific account of frequency and priming effects is open to a number of challenges. First, it is difficult to
define salience or expectedness in a manner that precisely maps onto priming, frequency, and lexicality effects. Second, Burt (1999) previously found that, at a short SOA unlikely to support expectancy generation, associative primes facilitated base-word color naming. This result indicates that color naming is speeded by facilitation of target word identification. Third, item analyses aggregated over the present experiments indicated a small but marginally reliable correlation (r ⫽ ⫺.11, p ⫽ .05) between word frequency (range: 1–9) and color-naming latencies for lowfrequency words. One measure of distinctiveness was obtained for the same words by counting the number of their word-body neighbors in the Kucera and Francis corpus (1967), with a low count indicating high orthographic distinctiveness. These ratings did not correlate with color-naming latencies (r ⫽ .01). Finally, as discussed in Experiment 5 in relation to item familiarity, some factors that are separable from word identification per se may affect color naming and the LDT in a similar manner. In view of the convergence of the facilitation effects with the phenomena observed in traditional word recognition tasks, the tentative conclusion is that these effects in the color-naming and other tasks reflect common mechanisms. That is, it is proposed that (a) participants do read the base word in the Stroop stimulus on most trials, as is assumed in the literature on the standard Stroop task (see the review by C. M. MacLeod, 1991), and (b) the facilitation effects reflect at least in part the efficiency or speed of lexical processing. It is important to note that these effects may not apply in color-naming studies in which unwordlike letter strings and color-related words form a substantial proportion of the stimulus lists. An attentional set for word reading (Monsell, 1996) appears to be engendered in the present paradigm, a set that may depend on the nature of the stimulus lists used here. It is noteworthy that in keeping with the present and earlier results (Burt, 1994), a recent investigation of color naming in the non-color–word Stroop task reported a consistent, albeit not always reliable, tendency for color naming to be faster for high- than low-frequency words (Monsell, Taylor, & Murphy, 2001). The frequency manipulation was less extreme than in the present studies, with high-frequency words defined as above 30 per million in two cases and above 50 per million in another. The primary result of the Monsell et al. studies was that color-naming latencies were slowest for words and wordlike nonwords, fastest for nonalphanumeric characters, and intermediate for unpronounceable nonwords. They interpreted this result in terms of participants’ adopting a whole-task set for reading, such that wordlike stimuli engage a set for word reading that must be suppressed in favor of a set for color naming. The marginal frequency effect they attributed to an occasional breakthrough of a word to lexical access and response activation. It was suggested that activation of a word’s name and interference with response selection would occur and be resolved more quickly for high- than low-frequency words. The present view is that lexical access of words almost always occurs in the color-naming task (consistent with the long-term repetition results of Experiment 5) and that lexical access occurs in parallel with, rather than after, the processing and response preparation associated with the color dimension of the Stroop stimulus (C. M. MacLeod, 1991). If the argument of Monsell et al. were correct, then breakthrough lexical access should cause color naming to be slower for low-frequency words than for pronounceable nonwords, a result they did not obtain.
In summary, it is proposed that the basic mechanism of facilitation of color naming is a reduction in the concurrent processing load, or the duration of the processing overlap, for word reading and preparation of the color-naming response. Other strategies adopted by participants may enhance their ability to ignore target words once they have been identified. A profitable direction for future research on the color-naming task would be to explore the relationship between color-naming and other word recognition tasks at short prime–target SOAs. From the perspective of reading research, a prediction that follows from the faster responding to well-learned, frequent words than rarer words is that analogous effects on color-naming speed might be observed among participants differing in reading skill. That is, within populations for whom reading is sufficiently well learned that word reading reliably occurs in the Stroop task, highly skilled readers are expected to exhibit less interference for non-color words relative to nonlexical control stimuli than poor readers. This prediction is counterintuitive in light of the claims about the importance of the automaticity of word reading to interference in the standard Stroop task (see the review by C. M. MacLeod, 1991) and highlights the complexity of the relationships between the standard and noncolor–word tasks.1
Applicability of Current Models of the Standard Stroop Task The priming interference effects demonstrated here in the noncolor–word Stroop task have been interpreted in terms of phonological activation of words and color names and resultant response competition. As indicated earlier, response competition is an important source of interference in recent connectionist models of interference in the standard color–word task, and the models might be applied to the recent results if processing units for non-color words were incorporated. Two key differences between the standard and non-color versions of the Stroop task limit the applicability of the models. First, when base words are color names, there is potential for semantic and output-level similarity-based interference between base words and color responses that does not arise with non-color words. Second, base words that are color names constitute a source of priming of alternative responses that is absent in the non-color–word task. The computational models produce interference primarily through stimulus–response compatibility differences for word naming and color naming (W. R. Glaser & Glaser, 1989; Phaf et al., 1990; Trainham, Lindsay, & Jacoby, 1997; Zhang, Zhang, & Kornblum, 1999) or through effects of learning on the relative strengths of activation in color and word pathways (Cohen et al., 1990; Cohen & Huston, 1994). Attentional input (e.g., as task demand nodes in the Cohen et al. models) ensures that the correct (color-name) response is produced. The simplest way to incorporate the priming interference effects demonstrated in the present experiments would be to increase the activity in the output unit for the non-color base word. Cohen et al. (1990) used this adjustment to model response priming effects. A comprehensive model of the present priming results would be rather cumbersome in that there 1 I thank Derek Besner for this suggestion, made in his review of the article.
PHONOLOGY AND COLOR NAMING
are many potential responses, and the model would require augmentation to deal with prime task and modality effects and trialto-trial changes in response activation depending on prime type. Models of the standard Stroop task offer little explanatory utility for the facilitation effects of identity primes on color naming in the non-color–word task or for the variations in color-naming speed associated with base-word frequency and lexicality. The facilitation effects are not comparable to the effects of congruent color names on color naming. Indeed, the present priming facilitation effects may be more aptly described as reductions in the interference normally produced by base words relative to a color-patch– only condition. An increase in the weights on the word reading pathway or its compatibility with the word-naming response would facilitate word processing in the models, but the result for color naming is likely to be increased interference. The problem is that strength of word processing and the capacity of the word pathway to interfere with color naming are conflated within models of the standard Stroop task. A distinction between the speed or ease of word recognition and the attentional consequences of word recognition has not proved necessary to account for interference and facilitation in the color–word Stroop task. The present results indicate that it is necessary to make this distinction in accounts of interference and facilitation in the non-color–word Stroop task.
Implications for Clinical Applications of the Non-Color– Word Stroop Task Sufferers of anxiety disorders typically show longer colornaming latencies in the non-color–word Stroop to a word related to threat than to an affectively neutral control word (see the review by Williams, Mathews, & MacLeod, 1996). The attentional bias toward threat that has been imputed to anxious individuals on the basis of the color-naming task and other tasks (e.g., the dot probe task; Broadbent & Broadbent, 1988) may influence color naming through emotional reactions to and rumination about the content of threat words (cf. Dawkins & Furnham, 1989). If so, given that these reactions are likely to involve implicit pronunciation of the threat word, the emotional Stroop effect can be explained in terms of competition with the color-name response arising from activation of the phonology of the threat word. It is predicted that manipulations of the kind used here to activate target word phonology would attenuate the difference in color-naming latencies between threat and neutral words. However, it has been suggested by a number of clinical researchers that, in addition to any distraction from color naming attributable to focusing on the recognized threat word, there is a preconscious attentional bias the effects of which can be observed when briefly presented, pattern-masked threat words produce color-naming interference. Although there are several reports of this subliminal emotional Stroop effect (Harvey, Bryant, & Rapee, 1996; Lundh, Wikstrom, Westerlund, & Ost, 1999; C. MacLeod & Rutherford, 1992; Mogg, Bradley, Williams, & Mathews, 1993), the interference typically is small, lack of awareness of words has not been established convincingly for individual participants, and there have been a number of failures to obtain the effect (Kyrios & Iob, 1998; Sackville, Schotte, Touyz, Griffiths, & Beumont, 1998; Thorpe & Salkovskis, 1997). On the present account that the availability in working memory of a phonologically activated word is the source of color-naming interference, it is difficult to imagine
how the interference can arise outside the respondent’s awareness of the word. Transient and perhaps preconscious phonological activation of the base word occurring as a by-product of word recognition is likely to have only a small effect that would not be differential for threat relative to control words. It is arguable that masked threat words require less processing time than matched control words, a factor that, on the present view, should produce facilitation for threat words; however, the evidence for a significant difference in speed of processing for the two word types is lacking at present.
Conclusion: The Non-Color–Word Task as a Vehicle to Investigate Word Recognition The pattern of interference and facilitation effects demonstrated here and previously (Burt, 1999) suggests that there may be considerable flexibility in word encoding according to the structure of the task and specific response requirements. There is a substantial literature on the effects of task requirements, materials, and contextual variables on the way that words are processed, for example, in semantic priming (e.g., Henik et al., 1983; Smith et al., 1983), episodic memory (e.g., perceptual vs. conceptual processing; Roediger, Weldon, & Challis, 1989), and color naming (Lowe & Mitterer, 1982). The previously discussed work of Monsell (1996) on attentional set and task switching raises issues that have been the focus of interest in a number of recent articles on lexical processing in color-naming tasks (e.g., Besner & Stolz, 1999). It is arguable that progress in word recognition research currently is limited by task factors impairing the ability of performance data to discriminate increasingly precise theoretical predictions. Given the importance of participants’ strategies for dealing with task requirements, progress in the elucidation of the mechanisms of visual word recognition may be well served by a more sophisticated understanding of strategic responses to experimental situations, together with the deployment of a variety of tasks providing convergent evidence on theoretical issues. The noncolor–word Stroop task has the potential to illuminate phonological processes in word naming and verbal working memory through their impact on response conflict in color naming. The utility of the task was demonstrated in a recent study in which initial phoneme overlap for the base word and color name reduced interference (Coltheart, Woollams, Kinoshita, & Perry, 1999). Furthermore, the color-naming task may be a useful vehicle to investigate factors affecting the ease or rapidity of word identification, through the impact of these factors on the efficiency and thus the latency of color naming. For example, masked priming effects provide a rich but complex source of information in word recognition (Forster, 1999), and additional evidence from the non-color–word task may allow validation of critical findings in a situation free from the influence of requirements to make lexical classifications or pronounce words.
References Andrews, S. (1992). Frequency and neighborhood effects on lexical access: Lexical similarity or orthographic redundancy? Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 234 –254. Bakan, P., & Alperson, B. (1967). Pronounceability, attensity, and interference in the color–word test. American Journal of Psychology, 80, 416 – 420.
Balota, D. A., & Chumbley, J. I. (1984). Are lexical decisions a good measure of lexical access? The role of word frequency in the neglected decision stage. Journal of Experimental Psychology: Human Perception and Performance, 10, 340 –357. Balota, D. A., & Spieler, D. H. (1999). Word frequency, repetition, and lexicality effects in word recognition tasks: Beyond measures of central tendency. Journal of Experimental Psychology: General, 128, 32–55. Besner, D., Davies, J., & Daniels, S. (1981). Reading for meaning: The effects of concurrent articulation. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 33(A), 415– 438. Besner, D., & Stolz, J. A. (1999). Unconsciously controlled processing: The Stroop effect reconsidered. Psychonomic Bulletin & Review, 6, 449 – 455. Bodner, G. E., & Masson, M. E. J. (1997). Masked repetition priming of words and nonwords: Evidence for a nonlexical basis for priming. Journal of Memory and Language, 37, 268 –293. Broadbent, D. E., & Broadbent, M. (1988). Anxiety and attentional bias: State and trait. Cognition and Emotion, 2, 165–183. Brown, G. D. (1987). Phonological coding in rhyming and homophony judgement. Acta Psychologica, 65, 247–262. Burt, J. S. (1994). Identity primes produce facilitation in a colour naming task. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 47(A), 957–1000. Burt, J. S. (1999). Associative priming in color naming: Interference and facilitation. Memory & Cognition, 27, 454 – 464. Burt, J. S., Mardle, L., & Humphreys, M. S. (1996). Expectancy-based associative and identity priming in pronunciation. Australian Journal of Psychology, 48, 64 –74. Burt, J. S., Walker, M., Humphreys, M. S., & Tehan, G. (1993). Associative priming in perceptual identification: Effects of prime-processing requirements. Memory & Cognition, 21, 125–137. Clarke, R., & Morton, J. (1983). Cross modality facilitation in tachistoscopic word recognition. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 35(A), 79 –96. Cohen, J. D., Dunbar, K., & McClelland, J. L. (1990). On the control of automatic processes: A parallel distributed processing account of the Stroop effect. Psychological Review, 97, 332–361. Cohen, J., & Huston, T. (1994). Progress in the use of interactive models for understanding attention and performance. In C. Umilta & M. Moscovitch (Eds.), Attention and performance XV: Conscious and nonconscious information processing (pp. 453– 476). Cambridge, MA: MIT Press. Coltheart, M., Woollams, A., Kinoshita, S., & Perry, C. (1999). A positionsensitive Stroop effect: Further evidence for a left–right component in print-to-speech conversion. Psychonomic Bulletin & Review, 6, 456 – 463. Conrad, C. (1974). Context effects in sentence comprehension: A study of the subjective lexicon. Memory & Cognition, 2, 130 –138. Craik, F. I., & Tulving, E. (1975). Depth of processing and the retention of words in episodic memory. Journal of Experimental Psychology: General, 104, 268 –294. Davelaar, E., & Besner, D. (1988). Word identification, imageability, semantics, and the content–functor distinction. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 40(A), 789 –799. Dawkins, K., & Furnham, A. (1989). The colour naming of emotional words. British Journal of Psychology, 80, 383–389. Dell, G. S., & O’Seaghdha, P. G. (1992). Stages of lexical access in speech production. Cognition, 42, 287–314. Dell, G. S., & Repka, R. J. (1992). Errors in inner speech. In B. J. Baars (Ed.), Experimental slips and human error: Exploring the architecture of volition (pp. 237–262). New York: Plenum Press. den Heyer, K., Goring, A., Gorgichuk, S., Richards, L., & Landry, M. (1988). Are lexical decisions a good measure of lexical access? Repe-
tition blocking suggests the affirmative. Canadian Journal of Psychology, 42, 274 –296. Dennis, I., & Newstead, S. E. (1981). Is phonological recoding under strategic control? Memory & Cognition, 9, 472– 477. Dosher, B. A., & Corbett, A. T. (1982). Instrument inferences and verb schemata. Memory & Cognition, 10, 531–539. Duchek, J. M., & Neely, J. H. (1989). A dissociative word-frequency ⫻ levels-of-processing interaction in episodic recognition and lexical decision tasks. Memory & Cognition, 17, 148 –162. Dunbar, K., & MacLeod, C. M. (1984). A horse race of a different color: Stroop interference patterns with transformed words. Journal of Experimental Psychology: Human Perception and Performance, 10, 622– 639. Durso, F. T., & Johnson, M. K. (1979). Facilitation in naming and categorizing repeated pictures and words. Journal of Experimental Psychology: Human Learning and Memory, 5, 449 – 459. Effler, M. (1977). Experimental contributions toward an analysis of the interference phenomenon observed with the Stroop test. Zeitschrift fuer Experimentelle und Angewandte Psychologie, 24, 244 –281. Elliott, E. M., Cowan, N., & Valle Inclan, F. (1998). The nature of cross-modal color–word interference effects. Perception & Psychophysics, 60, 761–767. Evett, L. J., & Humphreys, G. W. (1981). The use of abstract graphemic information in lexical access. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 33(A), 325–350. Fay, D., & Cutler, A. (1977). Malapropisms and the structure of the mental lexicon. Linguistic Inquiry, 3, 505–520. Feustel, T. C., Shiffrin, R. M., & Salasoo, A. (1983). Episodic and lexical contributions to the repetition effect in word identification. Journal of Experimental Psychology: General, 112, 309 –346. Fischler, I. (1977). Associative facilitation without expectancy in a lexical decision task. Journal of Experimental Psychology: Human Perception and Performance, 3, 18 –26. Forster, K. I. (1999). The microgenesis of priming effects in lexical access. Brain and Language, 68, 5–15. Forster, K. I., & Chambers, S. M. (1973). Lexical access and naming time. Journal of Verbal Learning and Verbal Behavior, 12, 627– 635. Forster, K., & Davis, C. (1984). Repetition priming and frequency attenuation in lexical access. Journal of Experimental Psychology: Learning, Memory, and Cognition, 10, 680 – 698. Frost, R. (1998). Toward a strong phonological theory of visual word recognition. Psychological Bulletin, 123, 71–99. Glaser, M. O., & Glaser, W. R. (1982). Time course analysis of the Stroop phenomenon. Journal of Experimental Psychology: Human Perception and Performance, 8, 875– 894. Glaser, W. R., & Dungelhoff, F. J. (1984). Time course of picture–word interference. Journal of Experimental Psychology: Human Perception and Performance, 10, 640 – 654. Glaser, W. R., & Glaser, M. O. (1989). Context effects in Stroop-like word and picture processing. Journal of Experimental Psychology: General, 118, 13– 42. Grainger, J., & Ferrand, L. (1996). Masked orthographic and phonological priming in visual word recognition. Journal of Memory and Language, 35, 623– 647. Greene, R. L., & Thapar, A. (1994). Mirror effect in frequency discrimination. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 946 –952. Harvey, A. G., Bryant, R. A., & Rapee, R. M. (1996). Preconscious processing of threat in posttraumatic stress disorder. Cognitive Therapy and Research, 20, 613– 623. Henik, A., Friedrich, F. J., & Kellogg, W. A. (1983). The dependence of semantic relatedness effects upon prime processing. Memory & Cognition, 11, 366 –373. Hintzman, D. L., Carre, F. A., Eskridge, V. L., Owens, A. M., Shaff, S. S.,
PHONOLOGY AND COLOR NAMING & Sparks, M. E. (1972). “Stroop” effect: Input or output phenomenon? Journal of Experimental Psychology, 95, 458 – 459. Jacoby, L. L., & Dallas, M. (1981). On the relationship between autobiographical memory and perceptual learning. Journal of Experimental Psychology: General, 110, 306 –340. Jared, D., Levy, B. A., & Rayner, K. (1999). The role of phonology in the activation of word meanings during reading: Evidence from proofreading and eye movements. Journal of Experimental Psychology: General, 128, 219 –264. Johnston, R. S., & McDermott, E. A. (1986). Suppression effects in rhyme judgment tasks. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 38(A), 111–124. Johnston, W. A., & Hawley, K. J. (1994). Perceptual inhibition of expected inputs: The key that opens closed minds. Psychonomic Bulletin & Review, 1, 56 –72. Jones, D. (1993). Objects, streams, and threads of auditory attention. In A. D. Baddeley & L. Weiskrantz (Eds.), Attention: Selection, awareness and control. A tribute to Donald Broadbent (pp. 87–104). Oxford, England: Oxford University Press. Joyce, C. A., Paller, K. A., Schwartz, T. J., & Kutas, M. (1999). An electrophysiological analysis of modality-specific aspects of word repetition. Psychophysiology, 36, 655– 665. Kachelski, R. A. (1997). An investigation of dual-task interference using a modified Stroop color naming task. Dissertation Abstracts International, 58B, 3332. Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NJ: Prentice Hall. Kahneman, D., & Henik, A. (1981). Perceptual organization and attention. Hillsdale, NJ: Erlbaum. Kamin, L. J. (1969). Predictability, surprise, attention and conditioning. New York: Appleton-Century-Crofts. Kinoshita, S. (1995). The word frequency effect in recognition memory vs. repetition priming. Memory & Cognition, 23, 569 –580. Kirsner, K., Dunn, J., & Standen, P. (1987). Record-based word recognition. Hove, England: Erlbaum. Kirsner, K., Dunn, J. C., & Standen, P. (1989). Domain-specific resources in word recognition. In S. Lewandowsky, J. C. Dunn, & K. Kirsner (Eds.), Implicit memory: Theoretical issues (pp. 99 –122). Hillsdale, NJ: Erlbaum. Kirsner, K., Milech, D., & Standen, P. (1983). Common and modalityspecific processes in the mental lexicon. Memory & Cognition, 11, 621– 630. Kirsner, K., & Smith, M. C. (1974). Modality effects in word identification. Memory & Cognition, 2, 637– 640. Kleiman, G. M. (1975). Speech recoding in reading. Journal of Verbal Learning and Verbal Behavior, 14, 323–339. Klein, G. S. (1964). Semantic power measured through the interference of words with color naming. American Journal of Psychology, 77, 576 – 588. Kucera, H., & Francis, W. N. (1967). Computational analysis of presentday American English. Providence, RI: Brown University Press. Kyrios, M., & Iob, M. A. (1998). Automatic and strategic processing in obsessive– compulsive disorder: Attentional bias, cognitive avoidance or more complex phenomena? Journal of Anxiety Disorders, 12, 271–292. Lee, Y. A., Binder, K. S., Kim, J. O., & Pollatsek, A. (1999). Activation of phonological codes during eye fixations in reading. Journal of Experimental Psychology: Human Perception and Performance, 25, 948 –964. Logan, G. D., & Zbrodoff, N. J. (1979). When it helps to be misled: Facilitative effects of increasing the frequency of conflicting stimuli in a Stroop-like task. Memory & Cognition, 7, 166 –174. Lowe, D. G., & Mitterer, J. O. (1982). Selective and divided attention in a Stroop task. Canadian Journal of Psychology, 36, 684 –700. Lukatela, G., Frost, S. J., & Turvey, M. T. (1998). Phonological priming by
masked nonword primes in the lexical decision task. Journal of Memory and Language, 39, 666 – 683. Lundh, L., Wikstrom, J., Westerlund, J., & Ost, L. (1999). Preattentive bias for emotional information in panic disorder with agoraphobia. Journal of Abnormal Psychology, 108, 222–232. Mackay, D. G. (1987). The organisation of perception and action: A theory for language and other cognitive skills. New York: Springer. MacLeod, C. M. (1991). Half a century of research on the Stroop effect: An integrative review. Psychological Bulletin, 109, 161–203. MacLeod, C. M. (1996). How priming affects two speeded implicit tests of remembering: Naming colors versus reading words. Consciousness and Cognition: An International Journal, 5, 73–90. MacLeod, C., & Rutherford, E. M. (1992). Anxiety and the selective processing of emotional information: Mediating roles of awareness, trait and state variables, and personal relevance of stimulus materials. Behaviour Research and Therapy, 30, 479 – 491. Mathews, A., & MacLeod, C. (1985). Selective processing of threat cues in anxiety states. Behaviour Research and Therapy, 23, 563–569. McClain, L. (1983). Color priming affects Stroop interference. Perceptual and Motor Skills, 56, 643– 651. McKone, E. (1995). Short-term implicit memory for words and nonwords. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 1108 –1126. McKone, E., & Dennis, C. (2000). Short-term implicit memory: Visual, auditory, and cross modality priming. Psychonomic Bulletin & Review, 7, 341–346. Merrill, E. C., Sperber, R. D., & McCauley, C. (1981). Differences in semantic encoding as a function of reading comprehension skill. Memory & Cognition, 9, 618 – 624. Meyer, D. E., & Gordon, P. C. (1985). Speech production: Motor programming of phonetic features. Journal of Memory and Language, 24, 3–26. Mogg, K., Bradley, B. P., Williams, R., & Mathews, A. (1993). Subliminal processing of emotional information in anxiety and depression. Journal of Abnormal Psychology, 102, 304 –311. Mogg, K., Mathews, A., & Weinman, J. (1989). Selective processing of threat cues in anxiety states: A replication. Behaviour Research and Therapy, 27, 317–323. Monsell, S. (1985). Repetition and the lexicon. Hillsdale, NJ: Erlbaum. Monsell, S. (1991). The nature and locus of word frequency effects in reading. In D. Besner & G. W. Humphreys (Eds.), Basic processes in reading: Visual word recognition (pp. 148 –197). Hillsdale, NJ: Erlbaum. Monsell, S. (1996). Control of mental processes. In V. Bruce (Ed.), Unsolved mysteries of the mind (pp. 93–143). Hove, England: Erlbaum. Monsell, S., Taylor, T. J., & Murphy, K. (2001). Naming the color of a word: Is it responses or task sets that compete? Memory & Cognition, 29, 137–151. Morris, C. D., Bransford, J. D., & Franks, J. J. (1977). Levels of processing versus transfer appropriate processing. Journal of Verbal Learning and Verbal Behavior, 16, 519 –533. Morton, J., & Chambers, S. M. (1973). Selective attention to words and colours. Quarterly Journal of Experimental Psychology, 25, 387–397. Neely, J. H. (1991). Semantic priming effects in visual word recognition: A selective review of current findings and theories. In D. Besner & G. W. Humphreys (Eds.), Basic processes in reading: Visual word recognition (pp. 264 –336). Hillsdale, NJ: Erlbaum. Neely, J. H., Keefe, D. E., & Ross, K. L. (1989). Semantic priming in the lexical decision task: Roles of prospective prime-generated expectancies and retrospective semantic matching. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15, 1003–1019. Norris, D. (1984). The effects of frequency, repetition, and stimulus quality in visual word recognition. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 36(A), 507–518.
Oden, G. C., & Spira, J. L. (1983). Influence of context on the activation and selection of ambiguous word senses. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 35(A), 51– 64. Ostergaard, A. L. (1998). The effects on priming of word frequency, number of repetitions, and delay depend on the magnitude of priming. Memory & Cognition, 26, 40 – 60. Parkin, A. J. (1979). Specifying levels of processing. Quarterly Journal of Experimental Psychology, 31, 175–195. Parkin, A. J. (1984). Levels of processing, context, and the facilitation of pronunciation. Acta Psychologica, 55, 19 –29. Pashler, H., & Johnston, J. C. (1998). Attentional limitations in dual-task performance. In H. Pashler (Ed.), Attention (pp. 155–189). Hove, England: Psychology Press/Erlbaum/Taylor & Francis. Perfetti, C. A., & Bell, L. (1991). Phonemic activation during the first 40 ms of word identification: Evidence from backward masking and priming. Journal of Memory and Language, 30, 473– 485. Phaf, R. H., van der Heijden, A. H., & Hudson, P. T. (1990). A connectionist model for attention in visual selection tasks. Cognitive Psychology, 22, 273–341. Rajaram, S., & Neely, J. H. (1992). Dissociative masked repetition priming and word frequency effects in lexical decision and episodic recognition tasks. Journal of Memory and Language, 31, 152–182. Redding, G. M., & Gerjets, D. A. (1977). Stroop effect: Interference and facilitation with verbal and manual responses. Perceptual and Motor Skills, 45, 11–17. Roediger, H. L. I., Weldon, M. S., & Challis, B. H. (1989). Explaining dissociations between implicit and explicit measures of retention: A processing account. In H. L. Roediger & F. I. M. Craik (Eds.), Varieties of consciousness: Essays in honor of Endel Tulving (pp. 3– 41). Hillsdale, NJ: Erlbaum. Rouibah, A., Tiberghien, G., & Lupker, S. J. (1999). Phonological and semantic priming: Evidence for task-independent effects. Memory & Cognition, 27, 422– 437. Sackville, T., Schotte, D. E., Touyz, S. W., Griffiths, R., & Beumont, P. J. V. (1998). Conscious and preconscious processing of food, body weight and shape, and emotion-related words in women with anorexia nervosa. International Journal of Eating Disorders, 23, 77– 82. Salame´ , P., & Baddeley, A. (1982). Disruption of short-term memory by unattended speech: Implications for the structure of working memory. Journal of Verbal Learning and Verbal Behavior, 21, 150 –164. Scarborough, D. L., Cortese, C., & Scarborough, H. S. (1977). Frequency and repetition effects in lexical memory. Journal of Experimental Psychology: Human Perception and Performance, 3, 1–17. Schnur, P. (1977). Testing the encoding elaboration hypothesis: The effects of exemplar ranking on recognition and recall. Memory & Cognition, 5, 666 – 672. Shelton, J. R., & Martin, R. C. (1992). How semantic is automatic semantic priming? Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 1191–1210. Shimada, H. (1990). Effect of auditory presentation of words on color naming: The intermodal Stroop effect. Perceptual and Motor Skills, 70, 1155–1161.
Smith, M. C., Theodor, L., & Franklin, P. E. (1983). The relationship between contextual facilitation and depth of processing. Journal of Experimental Psychology: Learning, Memory, and Cognition, 9, 697– 712. Sokolov, A. N. (1972). Inner speech and thought. New York: Plenum Press. Stemberger, J. P. (1985). An interactive activation model of speech production. In A. Ellis (Ed.), Progress in the psychology of language (Vol. 1, pp. 143–186). London: Erlbaum. Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18, 643– 662. Thorpe, S. J., & Salkovskis, P. M. (1997). Information processing in spider phobics: The Stroop colour naming task may indicate strategic but not automatic attentional bias. Behaviour Research and Therapy, 35, 131– 144. Trainham, T. N., Lindsay, D. S., & Jacoby, L. L. (1997). Stroop process dissociations: Reply to Hillstrom and Logan (1997). Journal of Experimental Psychology: Human Perception and Performance, 23, 1579 – 1587. VanVoorhis, B. A., & Dark, V. J. (1995). Semantic matching, response mode, and response mapping as contributors to retroactive and proactive priming. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 913–932. Wagner, A. R. (1976). Priming in S.T.M.: An information processing mechanism for self-generated and retrieval-generated depression in performance. Hillsdale, NJ: Erlbaum. Warren, R. E. (1972). Stimulus encoding and memory. Journal of Experimental Psychology, 94, 90 –100. Warren, R. E. (1974). Association, directionality, and stimulus encoding. Journal of Experimental Psychology, 102, 151–158. Watts, F. N., McKenna, F. P., Sharrock, R., & Trezise, L. (1986). Colour naming of phobia related words. British Journal of Psychology, 77, 97–108. Whatmough, C., & Arguin, M. (1998). Semantic mediation of auditory priming in dyslexia. Brain and Cognition, 37, 86 – 88. Whitney, P. (1986). Processing category terms in context: Instantiations as inferences. Memory & Cognition, 14, 39 – 48. Whitney, P., McKay, T., Kellas, G., & Emerson, W. A. J. (1985). Semantic activation of noun concepts in context. Journal of Experimental Psychology: Learning, Memory, and Cognition, 11, 126 –135. Wilding, J., & White, W. (1985). Impairment of rhyme judgments by silent and overt articulatory suppression. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 37(A), 95–107. Williams, M., Mathews, A., & MacLeod, C. (1996). The emotional Stroop task and psychopathology. Psychological Bulletin, 120, 3–24. Zhang, H., Zhang, J., & Kornblum, S. (1999). A parallel distributed processing model of stimulus–stimulus and stimulus–response compatibility. Cognitive Psychology, 38, 386 – 432.
Received November 29, 2000 Revision received January 11, 2002 Accepted February 5, 2002 䡲