Intervening response events between identification targets do not always turn repetition benefits into repetition costs

Hilchey, Matthew D.; Rajsic, Jason; Huffman, Greg; Pratt, Jay

doi:10.3758/s13414-016-1262-9

Intervening response events between identification targets do not always turn repetition benefits into repetition costs

Published: 06 January 2017

Volume 79, pages 807–819, (2017)
Cite this article

Download PDF

Attention, Perception, & Psychophysics Aims and scope Submit manuscript

Intervening response events between identification targets do not always turn repetition benefits into repetition costs

Download PDF

Matthew D. Hilchey¹,
Jason Rajsic¹,
Greg Huffman¹ &
…
Jay Pratt¹

1184 Accesses
17 Citations
Explore all metrics

Abstract

When there is a relatively long interval between two successive stimuli that must be detected or localized, there are robust processing costs when the stimuli appear at the same location. However, when two successive visual stimuli that must be identified appear at the same location, there are robust same location costs only when the two stimuli differ in their responses; otherwise same location benefits are observed. Two separate frameworks that inhibited attentional orienting and episodic integration, respectively, have been proposed to account for these patterns. Recent findings hint at a possible reconciliation between these frameworks—requiring a response to an event in between two successive visual stimuli may unmask same stimulus and same location costs that are otherwise obscured by episodic integration benefits in identification tasks. We tested this hybrid account by integrating an intervening response event with an identification task that would otherwise generate the boundary between same location benefits and costs. Our results showed that the intervening event did not alter the boundary between location repetition benefits and costs nor did it reliably or unambiguously reverse the common stimulus-response repetition benefit. The findings delimit the usefulness of an intervening event for disrupting episodic integration, suggesting that effects from intervening response events are tenuous. The divide between attention and feature integration accounts is delineated in the context of methodological and empirical considerations.

When do response-related episodic retrieval effects co-occur with inhibition of return?

Article 27 April 2020

Investigating the contribution of task and response repetitions to the sequential modulations of attentional cueing effects

Article Open access 29 November 2017

Responding, fast and slow: Visual detection and localization performance is unaffected by retrieval

Article Open access 20 November 2023

A complex network of attentional and memory processes causes the visuomotor system to be biased toward or away from select sources of sensory input. Such biases often are observed in the reaction time (RT) and accuracy data obtained from visual cueing procedures (Posner, 1980; Posner, Rafal, Choate, & Vaughan, 1985). In broad strokes, cueing procedures are used to evaluate how features (location, color, shape, response, etc.) associated with an initial stimulus (the “cue”^{Footnote 1}) bias responding to a later stimulus (the “target”), which may or may not share features with the cue. Importantly, the responses required by cues and targets (e.g., the task) heavily influence the type of bias observed. On the one hand, when simple detection or localization responses are required to targets and/or cues, responses, over time, typically become slower when the target appears at the same compared with a different location than the cue. We will refer to such effects as spatial repetition costs (SRCs). Conversely, when identity discrimination responses are required to cues and targets, SRCs are typically observed over time only when the cue and target require different responses. Otherwise, responses are faster when cues and targets share a location. We will refer to the latter effects as spatial repetition benefits (SRBs). We refer to the response-mediated division between SRCs and SRBs in identification tasks as the SRC-SRB boundary. The present study examined the hypothesis that an intervening response event between a cue and target abolishes or alters the established SRC-SRB boundary in stimulus identification tasks.

Before we begin our investigation into the role played by intervening response events, it is necessary to set the stage for the comparison between SRCs and SRBs. One of the best known examples of SRCs comes from the canonical attentional cueing paradigm (Posner & Cohen, 1984; Maylor & Hockey, 1985). When the interval between the cue and target onset exceeds approximately 300 ms, responding is slowest to targets appearing near a prior cue’s location, even though there is no correlation between the cue and target locations. This bias, sometimes referred to as inhibition of return (IOR), is long-lasting and particularly robust in tasks that require simple detection or localization responses (Samuel & Kat, 2003; Taylor & Klein, 2000). Indeed, this SRC is common even when all cue features, including the response made to the cue, repeat as the target (Rabbitt & Rogers, 1965; Kirby, 1972, 1976; Maylor & Hockey, 1985; Soetens et al., 1985; Tanaka & Shimojo, 1996; Soetens, 1998; Taylor & Donnelly, 2002; Coward, Poliakoff, O’Boyle, & Lowe, 2004); that goes for detection and localization tasks. In the attention literature, SRCs of this sort are commonly attributed to mechanisms that bias orienting against locations or objects to which attention was involuntarily captured (Klein, 2000).

Whereas SRCs are robust in simple stimulus detection and localization tasks, there is a firmly established boundary on SRCs in stimulus identification tasks at cue-target onset asynchronies (CTOAs) greater than 300 ms (Terry, Valdes, & Neill, 1994; Tanaka & Shimojo, 1996; Pratt & Castel, 2001; Taylor & Donnelly, 2002; Taylor & Ivanoff, 2005). This boundary is a demarcation line, or flashpoint, between inhibited attentional orienting and episodic integration accounts. Most crucially for present purposes, when identity discrimination responses are required to cues and targets, SRCs are observed only when the cues and targets require different responses (Rajsic, Bi, & Wilson, 2014; Notebaert & Soetens, 2003). If the cues and targets require the same response, regardless of cue-target identity (e.g., same or different colors), SRBs are observed (Fig. 1). SRBs are thought to occur because the memory trace of the cue’s integrated location-response representation is efficiently re-enacted when the target’s location-response representation matches it. In the episodic integration literature, this SRC-SRB boundary can be accounted for by mechanisms that generate costs and/or benefits when the target’s representation partially mismatches and perfectly matches/mismatches, respectively, the cue’s integrated representation (Hommel, Proctor, & Vu, 2004; Rajsic et al., 2014; discussed in more detail in the General Discussion). As such, there is rarely a need to appeal to inhibited orienting in identification tasks in which stimuli appear in sequence, because SRCs are conspicuously absent unless there are partial mismatches, on some level, between cue and target events (but see Christie & Klein, 2001). On the other hand, appeal to episodic integration falls short in detection and localization tasks because SRCs arise even when there is perfect feature (and response) overlap between the cue and target, barring their asynchrony (see Lupiáñez, 2010, for consideration of timing).

Independent of the SRC-SRB boundary, there also are RT benefits when a target matches, compared with mismatches, a previously executed cue response in sequential identification tasks (Bertelsen, 1965; Rabbitt, 1968; Kornblum, 1969). In some cases, this repetition benefit depends on the repetition of the cue stimulus as the target stimulus whereas in other cases response repetition is sufficient (Pashler & Baylis, 1991; Campbell & Proctor, 1993; see Tretyak & Yarrow, 2014, for recent review and extension). In general, stimulus-response repetition benefits in identification tasks are most robust when the interval between the cue response and target onset is ≤1 s (Bertelson & Renkin, 1966; Hale, 1967), although these repetition benefits can persist unabated well beyond 1 s when the stimulus-response mappings are many and/or arbitrary (Schvaneveldt & Chase, 1969). It is a classic stimulus-response repetition benefit that Spadaro, He, and Milliken (2012; also Spadaro & Milliken, 2013) recently reversed by simply requiring responses to events in between the cue and target. In identification tasks, a reversal of a repetition benefit by a variable other than time in the context of otherwise full feature overlap is remarkably rare; as such, any variable capable of doing so is of significant theoretical interest. The present experimental work was designed to develop a deeper understanding of the nature and robustness of this peculiar event-induced reversal in sequential identification tasks.

The basic conditions for reversing repetition benefits seem simple enough. At fixation, Spadaro et al. randomly presented one of two to-be-discriminated stimuli—each mapping onto a different response and separated by an intervening response event—as a cue and target on a given trial. In each experiment, the two stimuli belonged to a superordinate dimension (e.g., color, line, or word) and stimulus, response or stimulus-response repetition costs (all indistinguishable) were found whenever stimuli were separated by an intervening response event, but not otherwise. Taking color as a presently relevant example (Spadaro et al., 2012, Experiment 1A), a blue rectangle required one manual response and a yellow rectangle another. In one condition, the target appeared 1,500 or 2,500 ms after the cue response. In this case, responding was faster when the cue repeated as the target. In another condition, an intervening event (e.g., a red circle), which was orthogonal to the identification task, appeared between the cue and target. Whenever a response was made to this intervening event—regardless of its modality (e.g., visual or auditory), response (e.g., pressing both identification response keys simultaneously or verbal), the interval between the cue and target, or whether the intervening event occurred mixed within or separated between blocks—responding was slowest when the cue repeated as the target (Fig. 2). These findings contrasted markedly with Spadaro et al.’s observations of no effect or stimulus-response repetition benefits without intervening response events. To account for the boundary between the stimulus-response repetition benefit and the stimulus-response repetition cost, Spadaro et al. advocated for a two parallel process model. One process, episodic integration, relates to core principles in the aforementioned memory and feature integration literature, in that the retrieval of a response code for a recently encountered stimulus is efficiently reenacted in the absence of updating costs (partial feature mismatches). Critically, Spadaro et al. proposed that responding to an intervening event disrupts this process. The other process is an extension of space-based inhibition of return into the nonspatial domain (Law, Pratt, & Abrams, 1995; Fox & de Fockert, 2001; Francis & Milliken, 2003) with the core idea being that the brain habituates to repeat presentations of the same stimulus (similar to Dukewich & Boehnke, 2008; Dukewich, 2009). Thus, it is hypothesized that the intervening event knocks out the repetition benefits caused by episodic integration, leaving only repetition costs due to habituation (see also Spadaro & Milliken, 2013).

The hypothesis that a response to an intervening event is sufficient for knocking out the episodic integration process, of the sort that is also presumed responsible for the SRC-SRB boundary in stimulus identification tasks, has tremendous potential for unifying divergent findings and perspectives derived from simple detection/localization and identification tasks. If, as suggested, a response to a stimulus in between the cue and target reliably disrupts or knocks out the integration process between the cue and the target, it could be very useful for unmasking otherwise overshadowed repetition costs or inhibited orienting effects. In pursuit of this possibility, the present work seeks to evaluate whether Spadaro et al.’s (2012) intervening stimulus-response method (which may be similar to a “cue-back” in the inhibition of return literature; Prime, Visser, & Ward, 2006) can be used to alter, if not abolish, the SRC-SRB boundary. For this, Notebaert and Soetens’ cueing procedure (2003; Experiment 1) is the perfect place to start, in part because its SRC-SRB boundary was recently replicated and extended by Rajsic, Bi, and Wilson (2014) with fixed CTOAs (1,500, 10,500, and 13,500 ms). Thus, there is strong reason to believe that the SRC-SRB boundary is stable in this task in the absence of intervening response events at fixation in between the cue and target. In Experiment 1, we integrated Notebaert and Soetens’ (2003) cueing procedure with Spadaro et al.’s procedure primarily to determine whether intervening response events reverse the standard location-response repetition benefit (i.e., unmask spatial inhibition of return). By design, we also have an opportunity to evaluate whether the stimulus-response repetition benefit (presumed same-stimulus habituation or non-spatial inhibition of return) reverses in this more demanding task.

Experiment 1

In this experiment, the cues and targets randomly took on one of four colors, as in Notebaert and Soetens (2003, Experiment 1), instead of only two, as in Spadaro et al.’s procedure. With this kind of approach, two colors map onto a unique response finger so that the effect of color repetition can be distinguished from the effect of response repetition (Bertelsen, 1965), which is not possible using Spadaro et al.’s two color identification tasks. As in Notebaert and Soetens’, the cue and target appeared randomly to the left or right of a central fixation stimulus, but as in Spadaro et al., our cues and targets also appeared randomly at fixation. We used response-stimulus intervals (RSIs) instead of fixed CTOAs, more closely approximating Spadaro et al.’s procedure. Comparisons between Notebaert and Soetens’ (2003) and Rajsic et al.’s (2014) findings suggest that similar patterns are obtained (e.g., the SRC-SRB boundary is present) regardless of whether the cue-target interval is fixed or contingent on the response to the cue (Fig. 1), so we have little reason to be concerned that the SRC-SRB boundary hinges on this design choice. Thus in Experiment 1, the cue and target identities, locations, and responses randomly matched or mismatched, allowing for analyses of matching costs and/or benefits.

Spadaro et al.’s intervening response event was integrated into this design as follows: The response to the cue triggered a response-stimulus interval (RSI, 500 ms) followed by a centrally-presented intervening response event, the response to which triggered another 500 ms RSI. Conceptually, this approach yields all cue-target relationships in Spadaro et al. and Notebaert and Soetens experiments, whilst ensuring an intervening response event. We also add two new conditions (i.e., a peripheral cue followed by a central target and a central cue followed by a peripheral target), mainly to maintain a balanced design, as appropriately suggested by others (Christie & Klein, 2001). We ask simply: Will the response to the intervening event, in this context, disrupt or alter the integration process, thereby revealing non-spatial and/or spatial repetition costs instead of benefits? An affirmative to this question would help reconcile inhibited orienting and episodic integration accounts. A negative would delimit Spadaro et al.’s (2012) intervening stimulus-response method for unmasking repetition costs while giving rise to important methodological and theoretical considerations.

Methods

Participants

Twenty-four undergraduate students (mean age = 18.9 years; 18 females) from the University of Toronto consented to participate for course credit. All were naive to the purposes of the experiment and reported normal or corrected-to-normal vision.

Stimuli and apparatus

All stimuli were displayed on a 17” CRT monitor connected to a Dell computer running custom Python software. Head position was stabilized by a chin rest 44 cm from the monitor. Responses were made using the “z” and “/” keys on a standard QWRTY keyboard. All stimuli were displayed on a black background (0, 0, 0). The three cue and target locations—one left of center (7.5°), one at center, and one right of center (7.5°)—were marked by gray (RGB: 128, 128, 128) outline (1 pixel) placeholder boxes (2° × 2°). The cue and target colors were red (RGB: 128, 0, 0), green (RGB: 0, 128, 0), blue (RGB: 0, 0, 128) and purple (RGB: 128, 0, 128). The cues and targets filled the placeholder boxes. Two different colors mapped onto each index finger; the color-response mappings with completely counterbalanced across participants. Each trial began with a small (0.3° × 0.3°) white (RGB: 255, 255, 255) fixation cross centered in the middle placeholder. The intervening response event was a small white circle (radius = 0.3°) in the center placeholder box, to which both keys had to be pressed simultaneously. Text at the end of a trial appeared in the middle of the display in white font.

Procedure

Each person participated in one block of 432 trials, except one (396 trials). This trial count corresponds with three randomly intermixed runs of all factorial combinations of cue color (red, green, blue, or purple), target color (red, green, blue, or purple), cue location (left, center, or right) and target location (left, center, or right). Each participant sat down with the experimenter before testing and was fully instructed on the task. The experimenter emphasized the stimulus-response mappings, that eye gaze should remain on the center box, and that there was no correlation between the cue and target colors and locations.

The sequence of stimulus events appears in Fig. 3. Each trial began with the fixation cross and placeholder boxes; the cross was extinguished after 500 ms, which signaled the beginning of the trial. The color cue appeared 1 s later. The response to the cue extinguished the cue and triggered a 500 ms RSI, after which the intervening (bimanual) response event, as in Spadaro et al., appeared at center. The response to the intervening event extinguished the intervening event and triggered another 500 ms RSI, after which the color target appeared. If all trial responses were correct, the target response triggered a black screen and a 2 s intertrial interval (ITI). If a response error occurred at some point during the trial, a feedback screen appeared (“A response error was made…”), which also included all of the response mappings as reminders (e.g., “respond to red and green squares with the ‘z’ key…”). Feedback was acknowledged by a response key and triggered a 2 s ITI.

Results

Central and peripheral targets were analyzed separately. Peripheral target data were analyzed with 3 × 3 repeated measures ANOVAs. One factor was Cueing, which refers to the spatial relation between the cue and target [cued, uncued peripheral (cue), or uncued central (cue)]. The other factor, which we call Transition Type (following Notebaert & Soetens, 2003), represented the relationship between the cue and target’s color and response features [(Color Repetition (CR), Color Alternation (CA) with response repetition, or Response Alternation (RA)]. Central target data were analyzed with 2 × 3 repeated measures ANOVAs. In this case, Cueing had only two levels [cued, or uncued peripheral (cue)].

Peripheral targets

Most trials (89.9%) were completed without error, leaving ample data for RT analysis on error-free trials. Z scores for correct target RTs were computed for each participant at each level (9 levels) of the ANOVA; trials (3.1%) containing z scores greater than 2.5 were excluded as outliers. Next, we examined the data for exceptionally long CTOAs. We considered intervals greater than 5 s between the cue and target onset to be unreasonably long and thus excluded all such trials (<1%). Descriptively, we note that M RTs to the cue and intervening events were 741 ms and 327 ms, respectively. Accordingly, the mean CTOA was 2,068 ms.

Correct participant M RTs were submitted to a 3 × 3 repeated measures ANOVA. The main effects of Cueing [F(2, 46) = 6.36, MSE = 1107, p < 0.01] and Transition Type [F(2, 46) = 35.53, MSE = 5410, p < 0.01] were significant and these effects were qualified by the two-way interaction [F(4, 92) = 10.74, MSE = 1159, p < 0.01]. Replicating Notebaert and Soetens (2003), and Rajsic et al. (2014) (Fig. 4), SRBs were observed when the cue and target shared a response, whereas an SRC was observed when they did not. Also consistent with earlier reports and as can be gleaned from the parallel lines between CR and CA in Fig. 4, the alternating color feature did not interact with Cueing (F < 1; i.e., no stimulus-location repetition benefits or costs). The corresponding analysis on error rates to the target on otherwise correct trials revealed similar patterns [Cueing x Transition Type: F(4, 92) = 9.20, MSE = 26.74, p < 0.01; Cueing: F < 1; Transition Type: F(2, 46) = 31.33, MSE = 65.20, p < 0.01], although the alternating color feature appeared to interact with Cueing when the cue and target shared a response [F(2, 46) = 3.25, MSE = 20.0, p < 0.05]; however, this may be because accuracy was near ceiling when color repeated. Regardless, clearly typical patterns were observed and not reversed by the response to the intervening event.

Central targets

Most trials (90.7%) were completed without error. As before, z-scores were computed and trials containing z-scores greater than 2.5 were excluded (3.3%). Trials containing unreasonably long CTOAs (>5 s) were likewise removed (<1%). M RTs to the cue and intervening event were 731 ms and 327 ms, respectively. Accordingly, the mean CTOA was 2,058 ms.

Correct participant M RTs were submitted to a 2 × 3 repeated measures ANOVA (Fig. 5). The main effects of Cueing and the interaction between Cueing and Transition Type were not significant (Fs < 1). There was a main effect of Transition Type [F(2, 46) = 31.47, MSE = 6248, p < 0.01]. Simply, RTs were fastest on CR trials (481 ms), whereas there was no statistically meaningful difference between CA (600 ms) and RA (583 ms) trials [t(23) = 1.15, SE = 14.94, p > 0.10]. Clearly, the intervening response event did not reverse the usual benefit on CR trials (Fig. 5; i.e., there was no evidence for a stimulus-response repetition cost). Although Cueing and Transition Type did not interact in the RT analysis, they did in the error rate analysis [F(2, 46) = 9.66, MSE = 25.74, p < 0.01; also, no main effect of Cueing: F < 1 but a main effect of Transition Type: F(2, 46) = 18.73, MSE = 45.20, p < 0.05]. Simply, there was an SRB-SRC boundary in error rate but not, as noted, in RT. Apparently, as documented by Notebaert and Soetens (2003) in a different context (their Experiment 2), alternation biases are more prominent in error rate than RT when a cue differs from a target feature when the target appears at fixation.

Discussion

Typical SRB and SRC patterns were obtained despite intervening response events. Moreover, comparison between CR and RA transition types revealed robust stimulus-response repetition benefits, not costs, which occurred regardless of cueing (cf, CR-RA transition types for cued and uncued trials). Although beyond our purview, we note further that these differences imply a level of independence between stimulus-response and location-response representations. In other words, location, stimulus and response features were not completely integrated into a single representation. This is not to say that complete feature integration cannot occur (Hommel et al., 2004) nor does this diminish the hypothetical role of matching and mismatching operations as they apply to latent, binary feature-response representations (Hommel, 2007). These patterns, as they bear on episodic integration and inhibited orienting theories, will be elaborated on in the General Discussion.

Experiment 2

At this point, it seems that intervening response events do not reverse spatial or nonspatial repetition benefits when more complex coding operations are involved. In broad strokes, we are defining increased complexity by the two major differences between our experiment and those of Spadaro et al. (2012): four instead of two colors (2:1 response mappings), and three locations instead of a single target location at fixation. One or both of these variables likely mattered. However, because a major theoretical interest in the present report concerns the SRC-SRB boundary, it is not sensible to reduce the number of target locations to one. Thus, for present purposes, we are left contemplating whether our failure to abolish episodic integration processes is due to the increased stimulus-response mappings. That is, paradoxically, maybe our information reduction approach for studying the relative contributions of stimulus and response repetition to the stimulus-response repetition cost fundamentally abolished it. After all, and as noted in the introduction, repetition benefits tend to be persistent and long-lasting in tasks involving arbitrary and/or many-to-one response mappings. Experiment 2 therefore assessed whether the intervening response event reverses repetition benefits in the context of only two stimulus-response mappings (1:1 color-response mappings).