Greater attentional blink magnitude is associated with higher levels of anticipatory attention as measured by alpha event-related desynchronization (ERD)

Abstract

Accuracy for a second target (T2) is reduced when it is presented within 500 ms of a first target (T1) in a rapid serial visual presentation (RSVP)—an attentional blink (AB). Reducing the amount of attentional investment with an additional task or instructing the use of a more relaxed cognitive approach has been found to reduce the magnitude of the AB. As well, personality and affective traits, as well as affective states, associated with a more diffused or flexible cognitive approach have been found to predict smaller AB magnitudes. In the current study, event-related desynchronization in the alpha range was used to investigate whether the degree of attentional investment in anticipation of a RSVP trial was related to the behavioral outcome of that trial. As hypothesized, greater alpha ERD before the RSVP trial, indicating greater anticipatory attentional investment, was observed on short lag trials where an AB was present (inaccurate T2 performance) compared to short lag trials where an AB did not occur. However, on trials where T2 was presented after a longer period relative to T1, greater alpha ERD before the RSVP trial was found on trials with accurate T2 performance. Results support models of the AB that propose that greater attentional investment underlies the AB, and furthermore that this attentional investment is prepared in anticipation before each RSVP trial.

Introduction

When two to-be-attended targets are presented in a rapid serial visual presentation (RSVP) stream, accuracy for the second target (T2) is reduced when it is presented within 500 ms of the first target (T1), relative to longer T1–T2 separations—a phenomenon known as the attentional blink (AB; Raymond et al., 1992). The AB has been interpreted as reflecting attentional limitations where attentional processing of T1 interferes with and/or delays the allocation of attention to T2 if T2 is presented before T1 processing has been completed (Shapiro et al., 1997).

Traditional models of the AB tend to characterize the AB in terms of bottlenecks on information processing (e.g., Chun and Potter, 1995, Jolicoeur, 1998). For example, in the two-stage model of the AB (Chun and Potter, 1995), it is proposed that there are two stages to target processing. At the first stage, multiple stimuli can be processed in parallel and temporary fragile representations of the stimuli are created. In the second stage of processing, the fragile and temporary representations are encoding into more durable working memory representations that can be used for later report. Stage two processing is time and attention demanding such that a bottleneck is created at stage two processing if T2 is presented while T1 is still undergoing stage two processing, or if RSVP distractors are currently competing for stage two processing resources. Until that bottleneck is resolved, the encoding of any subsequent targets is delayed leaving their perceptual representations vulnerable to decay and reducing the probability that they will be accurately reported. Thus, any unnecessary investment of stage 2 processing resources in T1 would be expected to exacerbate the AB.

More recently, there have been models of the AB suggesting that some feature of cognitive control is responsible for the pattern of attentional investment that results in the failure to accurately report T2 at short target separations. For example, in the Temporary Loss of Control model (TLC; Di Lollo et al., 2005), it is suggested that cognitive control initially optimizes an input filter in favor of T1. When attention is needed to process the T1 stimulus, less attention is available to control the input filter and the filter falls under bottom-up control. If T2 is presented before cognitive control of the input filter is restored, this loss of cognitive control impairs selection of T2, resulting in the AB. Therefore, the TLC model implies that a lack of top-down cognitive control following T1 is responsible for the AB.

In the Boost-and-Bounce model (Olivers and Meeter, 2008), it is proposed that the T1 item elicits an excitatory “boost” that lasts long enough to also boost the distracter item that immediately follows T1 into working memory. Cognitive control then responds to the presence of this distracter with an inhibitory “bounce” that prevents subsequent items, including T2, from entering working memory. According to this model, poor cognitive control over the “bounce” response (i.e., an inability to prevent the “bounce”) seems to initiate the context necessary for an AB.

The Threaded Cognition model (Taatgen et al., 2009) also suggests that a memory function initiated by T1 prevents the further detection of targets. Taatgen et al. (2009) characterize this memory function as an overexertion of control, and suggest that when this control function is not engaged, the probability of accurate T2 performance is increased.

In their Overinvestment Hypothesis, Olivers and Nieuwenhuis, 2005, Olivers and Nieuwenhuis, 2006 propose that the AB results from the unrestrained investment of attentional resources extending to all RSVP items such that distractors become effective competitors for entrance into working memory. When T2 appears soon after T1, it is particularly vulnerable to this interference given the additional attention required for encoding T1, resulting in the AB. However, Olivers and Nieuwenhuis, 2005, Olivers and Nieuwenhuis, 2006 suggest that if investment of attention was reduced to a level just sufficient to encode the targets, then interference would be reduced and the probability of accurate T2 performance would increase, particularly at short target separations.

In all of the above models, limited attentional resources and inappropriate application of attention underlie the AB. Cognitive control models further suggest that this is a result of maladaptive management of attentional resources by top-down cognitive control. If more or less adaptive cognitive control and the resultant investment of attentional resources could influence the magnitude of the AB, then that would imply that the AB does not reflect a fundamental attentional processing limitation. Instead, the AB would be conceptualized as resulting from a particular attentional style, where its magnitude is influenced by the kind of cognitive control or attentional investment of attentional resources with which an individual approaches the RSVP task.

Recent evidence where researchers have manipulated or measured the level of cognitive control and/or attentional investment supports this conceptualization of the AB—specifically the possibility that overly stringent cognitive control and inappropriate attentional investment contribute to the AB. For example, when participants engaged in concurrent task such as detecting yells in music or performing a match to sample task, Olivers and Nieuwenhuis, 2005, Olivers and Nieuwenhuis, 2006¹ found that the AB was reduced relative to control conditions where participants performed only the AB task. Similarly, the AB has been reduced when task instructions emphasized a more passive target search strategy where you let the targets jump out at you (Olivers and Nieuwenhuis, 2005), and when AB task instructions emphasized reporting the two targets as a combination or pair (Ferlazzo et al., 2007). Olivers and Nieuwenhuis (2006) also observed a reduced AB when participants were exposed to positive affective pictures, relative to negative or neutral pictures. This result has implications for models of the AB given that positive affect is associated with an open and flexible cognitive processing style and diffused attention (e.g., Fredrickson, 2001) while negative affect is associated with heightened focusing of attention (e.g., Kramer et al., 1990).

Individual differences in trait affect (MacLean et al., 2010) and state affect (MacLean and Arnell, 2010) have been shown to predict AB magnitude where greater positive affect is associated with reduced AB magnitudes and greater negative affect is associated with increased AB magnitudes. Personality dimensions related to attentional investment and focus have also been shown to predict the magnitude of the AB where higher scores on extraversion and openness to experience predicted smaller AB magnitudes, and higher scores on neuroticism predicted larger AB magnitudes (MacLean and Arnell, 2010). Individual differences in the degree of global versus local processing also predict AB magnitude, where an individual's tendency to focus on the local information as opposed to seeing the global overall picture was positively associated with larger AB magnitudes (Dale and Arnell, 2010). Individual differences in the ability to effectively inhibit or ignore RSVP distractors have been shown to relate to the AB where greater inhibition of irrelevant RSVP distractors was associated with smaller AB magnitudes (Dux and Marois, 2008). Similarly, individual's ability to ignore irrelevant visual material presented beside the RSVP stream (Martens and Valchev, 2009), or in a separate visual working memory task (Arnell and Stubitz, 2010), was negatively related to AB magnitude, where greater ability to ignore the irrelevant material predicted smaller AB magnitudes. Notice that in each of these studies inappropriate allocation of attentional resources is associated with larger ABs.

While the above evidence supports models of the AB discussed previously, which propose that maladaptive applications of cognitive control and inappropriate attentional investment contribute to the AB, it also suggests that cognitive control and attentional investment are not determined once the RSVP stream starts or within the 500 ms following T1. Instead, these results suggest that even before the RSVP task begins the degree of cognitive control or attentional investment with which an individual approaches the trial can influence the AB. This suggests that there may be a relationship between readiness to invest attention before the RSVP trial, and T2 performance on that trial.

Attentional investment during the RSVP stream has often been measured using event-related brain potentials (ERPs). Electrophysiological investigations of the role of the AB have focused on attentional investment only during the RSVP task, either relative to a target (Martens et al., 2006a, Martens et al., 2006b, Sessa et al., 2007, Vogel and Luck, 2002, Vogel et al., 1998) or distracters (Martens, et al., 2006b). Vogel et al., 1998, Sessa et al., 2007 both demonstrated that the P3 component, (an ERP component thought by many to be related to updating working memory; e.g., Donchin, 1981), is absent following T2 at shorter target separations, suggesting that T2 fails to enter working memory when presented at shorter T1–T2 intervals. Martens et al. (2006a) showed that a low probability T1 resulted in a larger P3 component and a larger AB compared to a high probability T1 target, a finding they attributed to the greater attentional investment required for improbable T1s. Martens et al. (2006b) were also able to show that non-blinkers (individuals who reliably show no AB) had more discrete and significantly earlier P3's to T1 and T2, indicating that their attentional investment in targets differed from individuals who demonstrate an AB. Non-blinkers also had significantly reduced attentional investment in distracters, as measured by activation during distracter-only RSVP trials, suggesting that they were also investing less attention in distracter items compared to individuals who demonstrate an AB (Martens et al., 2006b).

The electrophysiological evidence reviewed above supports models of the AB that propose that an inappropriate investment of attentional resources underlies the AB in that greater ABs were associated with greater P3s to T1 and greater activation to RSVP distracters. However, the electrophysiological measures used (the P3 and distracter activation) were confined to measuring the attentional investment that occurs during the RSVP trial. The goal of this study is to investigate whether the degree of anticipatory attentional investment with which an individual approaches an AB trial influences the behavioral outcome on that trial.

Anticipatory attention has been captured by examining event-related changes in alpha frequency (~ 8–12 Hz) oscillations present in cortical electrophysiological activity (Bastiaansen and Brunia, 2001, Bastiaansen et al., 2002, Bastiaansen et al., 2001, Capotosto et al., 2009, Onoda et al., 2007, Yamagishi et al., 2005). Alpha frequency oscillations are frequently observed over widespread cortical areas and are thought to be generated by thalamo-cortical connections as well as cortico-cortical communication (Lopes da Silva, 1991, Steriade et al., 1990). Specifically, during a particular thalamic state, afferent stimulus information is prevented from proceeding to the cortex and this thalamic state results in synchronized cortical activity in the alpha range and an increase in alpha power. When afferent information is then allowed to reach the cortex, synchronization is disrupted, in other words, a desynchronization in the alpha range and a reduction in alpha power results (Lopes da Silva, 1991, Steriade et al., 1990). Lopes da Silva (1991) suggest that the different thalamic states could represent a gating function very similar to an early attentional filter which controls the flow of specific information to the cortex as well as between cortical areas, and that the function of this gating system is reflected in changes to alpha frequency oscillations of the cortex. Brunia and van Boxtel (2001) suggest that anticipatory attention is initiated by top-down influences generated by cortical areas that control attention via the thalamus, such that the flow of information to the cortex is regulated in order to facilitate the processing of an upcoming relevant stimulus. These authors propose that the top-down influence is directed at the reticular nucleus which is thought to control the thalamic states which result in the synchronization and desynchronization of alpha frequency oscillations of the cortex (Lopes da Silva, 1991, Steriade et al., 1990). Therefore, event-related desynchronization (ERD) in alpha frequency oscillations could be considered an index of anticipatory attention facilitating, in a top-down manner, the flow of information from the thalamus to the cortex. If this assumption is true, then alpha ERD should be observed prior to a to-be-attended stimulus when that stimulus can be anticipated and greater alpha ERD should reflect greater anticipation.

In support of the idea of alpha ERD as a measure of anticipatory attentional investment, alpha ERD has been observed following a cue to shift attention toward the location of an upcoming target, and found to persist until the target appeared (Yamagishi et al., 2005). Alpha ERD has also been observed prior to a visual stimulus providing feedback on performance of a time-estimation task (Bastiaansen and Brunia, 2001, Bastiaansen et al., 2002, Bastiaansen et al., 2001). Furthermore, interrupting alpha ERD prior to a relevant stimulus, using trans-cranial magnetic stimulation on fronto-parietal areas of the attention network, impaired target performance in a spatial attention task (Capotosto et al., 2009). Alpha ERD has also been shown to be larger following a cue indicating that the upcoming affective stimulus was negatively-valenced, compared to when the cue indicated it was positively-valenced (Onoda et al., 2007). These authors interpreted this as evidence that cortical sensory-perceptual areas were activated via top-down control in anticipation of a relevant stimulus, in order to facilitate the processing of that stimulus.

It has been suggested that the assignment of limited attentional resources to T1 and distracters (Chun and Potter, 1995), an overexertion of cognitive control (Olivers and Meeter, 2008, Taatgen et al., 2009), and/or a general overinvestment of attention (Olivers and Nieuwenhuis, 2006) underlies the AB. These models have been supported in that: (1) manipulations meant to reduce the degree of attentional investment have been shown to reduce the AB (e.g., Olivers and Nieuwenhuis, 2005, Olivers and Nieuwenhuis, 2006), (2) affective and personality traits associated with an open and flexible cognitive control and/or reduced attentional investment predict smaller ABs (MacLean and Arnell, 2010, MacLean et al., 2010), and (3) an individual inability to avoid processing irrelevant information predicts larger ABs (Arnell and Stubitz, 2010, Dux and Marois, 2008, Martens and Valchev, 2009).

The goal of this study is to test the hypothesis that an inappropriate investment of attention just prior to the onset of the RSVP stream contributes to the production of the AB. This was done by investigating whether the degree of anticipatory attentional investment with which an individual approaches an AB trial influences the behavioral outcome on that trial. To answer this question, we measured the level of anticipatory attentional investment prior to the RSVP stream, using alpha ERD as a more direct measure of anticipatory attention. If greater levels of anticipatory attentional investment are associated with a reduction in the probability of accurate T2 performance at shorter T1–T2 intervals, then alpha ERD should be greater prior to the RSVP stream on trials when T2 performance was incorrect than when T2 performance was correct at shorter T1–T2 intervals. We would not expect, however, that greater alpha ERD prior to the RSVP stream should be associated similarly with poor T2 performance at longer T1–T2 intervals. At shorter T1–T2 intervals, overinvestment of anticipatory attention would be expected to increase attention to T1, and this would leave less attention for a closely trailing T2. However, at longer T1–T2 intervals, T1 processing would likely be complete by the time T2 was presented, and thus overinvestment would not have a detrimental effect on T2 processing.

When T2 is presented at longer T1–T2 intervals (more than 500 ms after T1), T2 performance resembles single target performance (Raymond et al., 1992). Greater attentional investment would generally be expected to improve target performance, except in the particular circumstances in which an AB is found to occur as discussed above. So, when those circumstances are absent, as is the case for T1 processing or when T2 is presented following longer T1–T2 intervals, greater levels of anticipatory attentional investment as indicated by greater alpha ERD should be associated with better target performance. This hypothesis assumes that attentional investment beyond some minimum threshold required for accurate identification of a target stimulus could still increase target performance.