P300-based Stroop study with low probability and target Stroop oddballs: The evidence still favors the response selection hypothesis

Abstract

This paper addresses the issue of the locus of action in cognitive processing during Stroop effects. It uses the P300 latency to assess stimulus processing effects, but, for the first time, under conditions in which Stroop stimuli are rare and target stimuli. The study was also concerned with demonstrating that apparent P300s during verbal responding are in fact uninterpretable due to contamination of EEG by speech-related artifact. Three studies were presented. In Study 1, there were 3 blocks, each containing 1 of 3 types of rare Stroop stimuli (p = .15), congruent, neutral, and incongruent. There were also 3 response modes: button press (BUTTON), speaking aloud (VERBAL), and speaking to self (SILENT). Three sessions were used, each for a different response style. The only task was to name the color on each trial. In the 2 non-verbal blocks, Reaction Time (RT) varied by stimulus type; congruent < neutral < incongruent. P300 latency was the same across blocks in these non-verbal conditions in which one saw the classic Pz > Cz > Fz distribution. The much larger, speech artifact-contaminated “P300s” in the VERBAL blocks did suggest a Stroop effect, especially at Fz and Cz, where “P300s” were larger than at Pz. In Study 2, there were 2 response modes, VERBAL and SILENT, and only two rare Stroop stimuli; neutral and incongruent, 1 per block. In each of these blocks, one word–color combination was a designated target requiring a unique response. The subject was to name the color followed by a yes or no to categorize the target or non-target. Again the RT for incongruents was greater than RT for neutrals, without a parallel effect in P300 latency. Again, the rostral ERPs appeared artifactual in the VERBAL condition. Study 3 was a replication of the second study, except that motivated subjects, versus Psychology pool recruits, were used. The latency–RT correlation still failed to obtain. Thus, using classic P300-eliciting antecedents–rare and target (Stroop) stimuli–this study supports the view that the locus of Stroop interference is in response processing.

Introduction

The Stroop effect (Stroop, 1935) is one of the most robust and best known in psychology. There are actually two Stroop phenomena; Facilitation and Interference. The former effect is demonstrated by comparison of behavioral reaction times (RTs) in response to congruent stimuli (names of colors, e.g., the word, “RED” presented in congruent colors, e.g., red color), in comparison to RTs to neutral stimuli, which are non-color words (e.g., the word, “HAT”) presented in some color. The subject's task is to name the color of the displayed word. With congruent stimuli, the facilitated RT is typically less than the neutral RT. On the other hand, in response to incongruent stimuli, such as the word “RED” presented in the color blue, the RT is typically greater than the neutral RT. The typically more robust effect is interference, and the response style yielding the most robust Stroop effects is verbal as opposed to manual responding (MacLeod, 1991).

As discussed by MacLeod (1991) and others, a major theoretical question about the effect has centered on its locus in the cognitive processing sequence. The question is whether Stroop effects are active during the stimulus evaluation phase or the response selection phase or both. Although recent studies have used functional magnetic resonance imaging (fMRI) and other methods so as to find alternative ways to conceptualize the Stroop effect in terms of cognitive control, conflict monitoring and task switching (Kerns et al., 2004, MacDonald et al., 2000), Duncan-Johnson and Kopell (1981) were the first to approach the Stroop mechanism question with a well-conceived psychophysiological study utilizing the latency of the P300 ERP. P300 is an electrically positive-going wave classically elicited during the presentation of a Bernoulli series of rare (e.g., p < .3) target stimuli requiring unique behavioral responses and frequent (e.g., p > .7) meaningless stimuli; the former but not the latter elicit P300. Since Kutas et al. (1977), Duncan-Johnson (1981), and others had shown that the latency of P300 was a potentially useful index of the stimulus evaluation time, Duncan-Johnson and Kopell (1981) reasoned that if the behavioral Stroop effects were correlated with simultaneously collected P300 latencies, this would support the stimulus evaluation hypothesis of the locus of the Stroop effects. Dissociation of P300 latency and RTs, on the other hand, would support the response selection hypothesis, and indeed, this is what was reported.

There nevertheless were reservations about the strength of support provided by this study due to methodological concerns: Most importantly, the ERP data were collected in a paradigm in which subjects made verbal responses on each trial. It has been long known that speech produces large artifacts in an EEG record (Szirtes and Vaughan, 1977) which precede as well as follow verbalizations by about 1 s. This fact is what initially prompted us (Rosenfeld and Woodley, 1994) to utilize silent (mental) responding, as we will do more systematically, here. Both previously and presently, we observe clear evidence of speech artifact during verbalization trials at all sites. More recently, others have shared this concern and utilized alternative response modes in ERP Stroop studies, e.g., Ilan and Polich (1999), West and Alain (1999), Liotti et al. (2000), who considered all three response modalities which we use here in our Experiment 1, and Atkinson et al. (2003).

Duncan-Johnson and Kopell (1981) moreover presented data from Pz only. The two problem issues associated with this fact are that 1) speech artifact is usually most evident at more anterior sites (as we will also show again here), so that it is not as obvious at Pz; thus noisy Pz data during speech can be mistaken for clean P300 activity; 2) it is impossible without Pz, Fz and Cz data to satisfy the usual scalp distribution attribute that helps identify P300, namely, that Pz > Cz > Fz (Fabiani et al., 1987). If one does not have a definite, artifact-free P300, one cannot confidently utilize the P300 metric of stimulus evaluation in Stroop studies. This latter issue becomes especially important when the classical antecedents for P300 elicitation–rare, target stimuli–are absent, as they were in Duncan-Johnson and Kopell (1981) as well as in the other ERP-Stroop papers cited above. In all these studies, incongruent, neutral, and congruent stimuli were equally probable within a block, and the only task was color naming; i.e., there was no unique target response executed.

The importance of the present study is implied by the preceding two sentences. While it is likely that the late positive components in at least some of the aforementioned, ERP-enhanced Stroop studies based on P300 were or contained elements of the classic P300 or “P3b” components identified by Donchin and Coles (1988) and Fabiani et al. (1987), it is also the case that lacking two of the classic antecedent conditions for P300 elicitation–rare, target stimuli–it could be argued that the late positive components measured were likely to contain other positive components which overlap classic P300 (see Spencer et al., 2001). For example, researchers studying late positive components related to memory attributes usually wish to eliminate confounded probability effects and so use equally probable targets and non-targets. Thus studies using equally probable Stroop stimuli may be looking at memory-related EEG events other than the context updating mechanism theorized by Donchin and Coles (1988) to underlie classic “P3b.” As has been emphasized by Donchin and colleagues many times (though not universally accepted), strict constraints are required for component identification (e.g., Donchin et al., 1978). This is what we have tried to do here–for the first time–that is, utilize the Duncan-Johnson and Kopell (1981) P300 latency–RT correlation approach to investigating the locus of Stroop effects but with particular attention to P300 identification in terms of antecedent conditions emphasized by Donchin and colleagues.

The present set of three experiments was intended to remedy the situation by utilizing rare Stroop stimuli in the first study, and rare, target Stroop stimuli in the latter two studies. Three different response modes are utilized, and ERP data from Pz, Cz, and Fz during speech are compared with those data collected during silent and button-press response blocks. Correlation of P300 Pz latency with behavioral RT, as initially proposed by Duncan-Johnson and Kopell (1981), is the key observation in all studies.

The point of using three modalities here was as follows: The VERBAL modality, in which subjects name colors of stimuli aloud, was used so as to compare our RT and P300 latency data (elicited using classic P300 antecedent conditions) with those of Duncan-Johnson and Kopell (1981), and with special attention to waveforms at anterior sites near oro-facial structures. The manual (button press) and silent responding modalities were used so as to explore results with classic P300-eliciting antecedent conditions; such data did not previously exist.