Resting EEG in alpha and beta bands predicts individual differences in attentional blink magnitude

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). There are reliable individual differences in the magnitude of the AB. Recent evidence has shown that the attentional approach that an individual typically adopts during a task or in anticipation of a task, as indicated by various measures, predicts individual differences in the AB deficit. It has yet to be observed whether indices of attentional approach when not engaged in a goal-directed task are also relevant to individual differences in the AB. The current studies investigated individual differences in the AB by examining their relationship with attention at rest using quantitative measures of EEG. Greater levels of alpha at rest were associated with larger AB magnitudes, where greater levels of beta at rest were associated with smaller AB magnitudes. Furthermore, individuals with more beta than alpha demonstrated a smaller AB effect than individuals with more alpha than beta. Our results suggest that greater attentional engagement at rest, when not engaged in a goal-directed task, is associated with smaller AB magnitudes.

Introduction

Attention allows for the selective processing of information. This selectivity can help to ensure that relevant information is available to influence behavior. However, attention is a finite resource, and there are limits on how much information one can attend to at any one time. When the relevant information available exceeds the available attentional resources, the probability that relevant information will be missed increases. The attentional blink (AB) is a deficit in performance attributed to such attentional limitations (Raymond, Shapiro, & Arnell, 1992).

When the second of two to-be-attended targets is presented within half a second of the first target (T1) in a rapid serial visual presentation (RSVP) stream, report accuracy for the second target (T2) is impaired compared to when T2 is presented after a longer interval following T1 (Broadbent and Broadbent, 1987, Raymond et al., 1992). This T1–T2 lag dependant dual-task effect is referred to as the attentional blink (AB; Raymond et al., 1992).

Generally, the AB has been attributed to a lack of sufficient attentional resources necessary for multiple targets to be appropriately processed into awareness. For some authors, this has been instantiated in terms of bottleneck information processing models. For example, bottleneck models of the AB (e.g., Chun and Potter, 1995, Jolicoeur, 1998) propose that encoding the temporary representation of T1 into working memory consumes resources to the extent that any subsequent encoding of relevant information, such as T2, would need to be suspended until encoding of T1 was complete. Suspending the encoding of the fragile, temporary representation of T2 when its processing overlaps with that of T1 is thought to reduce the fidelity of the T2 representation leading to decreased report accuracy for T2 at short lags.

For other authors, a shortage of attentional resources has been instantiated in terms of limitations on cognitive control functions. For example, in the temporary loss of control model (Di Lollo, Kawahara, Ghorashi, & Enns, 2005) top-down input filters control the efficient selection of task-relevant targets. However, because encoding T1 into working memory requires limited attentional resources, these resources are not available to exert top-down control over the input filter, and the filter falls under bottom-up control while T1 is being encoded. If T2 is presented before attention is again free to control this filter, then T2 selection is impaired reducing the likelihood of accurate T2 report at short lags.

In the context of these accounts it appears that the AB is an unavoidable consequence of the way in which selective information processing is necessarily carried out. However, the AB is not observed in some individuals, so called “non-blinkers” (Martens, Munneke, Smid, & Johnson, 2006), and individuals differ reliably in the magnitude of their AB (i.e. the slope of the lag dependant effect on T2 performance; McLaughlin, Shore, & Klein, 2001). This suggests that individuals may differ in their speed and efficiency of information processing and/or their approach to selective information processing (i.e. their attentional approach), which then influences the magnitude of their AB.

Measures of fluid intelligence (Arnell et al., 2010, Colzato et al., 2007), information processing speed (Arnell, Howe, Joanisse, & Klein, 2006), and working memory capacity (Arnell and Stubitz, 2010, Arnell et al., 2010) do not appear to predict individual differences in the AB. This suggests that it is not the quality or amount of cognitive resources that determines the magnitude of an individual’s AB (Arnell et al., 2006). There is, however, evidence that the attentional approach that an individual adopts, for example whether attention is generally diffused or focused (Dale & Arnell, 2010), or whether attention tends to be more or less invested in irrelevant information (Arnell and Stubitz, 2010, Dux and Marois, 2008, Martens and Valchev, 2009), predicts individual differences in AB magnitude. These results show that individuals who focus less on irrelevant information and have a diffuse, global processing style produce smaller ABs. Dispositional factors like trait affect (MacLean, Arnell, & Busseri, 2010) and personality (MacLean & Arnell, 2010) also predict AB magnitude. Individuals high in traits such as positive affect, openness to experience and extraversion (traits that have been linked to a diffuse processing style – e.g., Fredrickson & Branigan, 2005) show smaller ABs, while individuals high in negative affect and neuroticism (traits linked to a focused attentional style – e.g., Kramer, Buckhout, & Eugenio, 1990) show larger ABs. So, it appears that how an individual tends to employ their cognitive resources is relevant to the magnitude of their AB, rather than the amount of resources they possess.

Recently, MacLean and Arnell (2011) operationalized attentional resource deployment in the AB task by measuring the amount of anticipatory attentional investment prior to each RSVP trial using alpha event-related desynchronization (ERD). Alpha ERD refers to a decrease in alpha power from baseline following an event; in the case of MacLean and Arnell the event was a cue that the RSVP was to begin shortly. Greater alpha ERD immediately before an RSVP trial began was shown to be beneficial for T1 performance and T2 performance at long T1–T2 lags. In contrast, alpha ERD was greater preceding short T1–T2 lag trials where T2 performance was subsequently incorrect (an AB trial) than on those trials where T2 performance was subsequently correct (a no-AB trial), providing evidence that greater investment of attentional resources in advance of the RSVP trial, is associated with the AB.

In summary, measuring attentional approach tendencies between individuals using cognitive task performance and self-report questionnaires, and using EEG to measure trial-to-trial changes in state attentional investment within individuals has provided evidence that individual differences in attentional approach and attentional investment are relevant to understanding the AB and individual differences in the AB.

Personality and trait affect measures index individual differences in general tendencies over relatively long epochs of time, and alpha ERD was used to measure intra-individual states of anticipatory attentional investment several seconds before the first target appears. Thus, even before an RSVP stream begins, individuals appear to approach the trial in characteristic ways that influence performance outcomes. Therefore, it is possible that individual differences in attentional investment at rest, when not engaged in the primary goal-directed task of interest (i.e. the AB task), may predict individual differences in the AB. Enduring dispositions in attentional approach could be reflected in individual differences in attentional investment at rest. Individual differences in attentional investment at rest could be indicative of the level of investment that is likely to occur during the AB task. In the current study we examined electrophysiological measures of individual differences in attentional investment at rest for the purpose of predicting individual differences in the AB.

Oscillatory activity in the alpha range of frequencies (8–12 Hz) is characteristically observed in the resting state. Specifically, alpha range oscillatory activity is increased during periods of rest with eyes closed compared to periods of rest with eyes open, an effect which is attributed to the desynchronizing effect of visual stimulation on the cortex (for a review of alpha oscillations see Niedermeyer, 1997). The presence of alpha in the waking state is thought to denote an “idling” or unoccupied cortex (Pfurtscheller, Stancák, & Neuper, 1996), or alternatively, the presence of cortical inhibition (Klimesch et al., 2006, Klimesch et al., 2007). In support of these conceptualizations, investigations of alpha oscillations at rest with concurrent fMRI imaging have observed a negative correlation between alpha and metabolic activity such that when alpha is present metabolic activity is reduced in the cortex (Goldman et al., 2002, Laufs et al., 2003a, Laufs et al., 2003b, Laufs et al., 2006, Moosmann et al., 2003). Alpha is also seen to decrease in response to a signal that an attention-demanding event is soon to occur (Brunia & van Boxtel, 2001), a phenomenon referred to as alpha event-related desynchronization (ERD) previously discussed here in relation to attentional investment and the AB. The desynchronization of alpha in response to a warning stimulus is thought to represent an engagement of attention (Brunia & van Boxtel, 2001) and/or a release of inhibition (Klimesch et al., 2006).

However, there is evidence to suggest that alpha desynchronization is not sufficient to indicate an increase in attentional investment at rest. Laufs et al., 2003b, Laufs et al., 2006 suggest that the presence of alpha, more specifically alpha at rest, is indicative of an idling cortex and represents a baseline state. However, reduction of alpha at rest can accompany two different kinds of fluctuations in attention. Alpha reductions may accompany an increase in vigilance, which increases their attention-demanding cognitive processes – a state that is accompanied by increased beta oscillations and decreased theta oscillations. Alpha reductions may also accompany an increase in drowsiness and a decrease in vigilance – a pattern associated with increased theta oscillations and decreased beta oscillations. Indeed, when alpha reduction was observed to correlate with increases in metabolic activity in frontal–parietal cortical areas thought to compose the attention network, faster oscillations in the beta range were increased while slower oscillations in the theta range were reduced (Laufs et al., 2006). This pattern, which Laufs et al. call a state of high vigilance, appears to represent an increase in attentional investment at rest where the cortex is engaged in information processing and various mental activities could be occurring. However, when alpha reduction was seen to correlate with increases in metabolic activity in occipital and parietal areas of the cortex, beta was reduced and theta increased, indicating a state of drowsiness, or low vigilance.

In summary, levels of alpha are related to attentional investment such that when alpha is high, attention is not engaged and the cortex is “idling” (Brunia and van Boxtel, 2001, Pfurtscheller et al., 1996) or inhibited (Klimesch et al., 2007). The relationship between reduced alpha and attention at rest depends on the cortical areas that are activated during alpha desynchronization, which is indicated by levels of oscillatory activity in frequency bands that neighbor alpha (Laufs et al., 2006). Attentional investment at rest is accompanied by high beta and low alpha, while low alpha and high theta is indicative of drowsiness not attentional investment. Thus, by measuring the relative contribution of EEG oscillations in alpha and beta bands, one can estimate the degree of attentional investment of an individual at rest. In the current study we aimed to use individual differences in levels of oscillatory activity in the alpha and beta ranges of EEG frequencies to measure attentional investment at rest for the purpose of predicting individual differences in the AB.¹

Previously we discussed evidence from various investigations indicating that increased attentional investment is associated with a larger AB. This might lead one to hypothesize that greater attentional investment at rest, operationalized in this study as more beta, less theta and less alpha power, should correlate positively with AB magnitude. However, there is additional evidence to suggest that the attentional investment at rest is not necessarily positively correlated with attentional investment during a goal-directed task as measured by alpha frequency oscillations.

As noted above, within participants, greater alpha ERD during the RSVP foreperiod predicted poor T2 performance at lags during the AB interval but better T2 performance at lags outside of the AB interval as well as better T1 performance (MacLean & Arnell, 2011). MacLean and Arnell (2011) also reported non-significant trend (r = .27) where individuals with greater overall alpha ERD showed greater ABs.

Alpha power has been shown to predict cognitive performance on other tasks in specific ways. Several studies have shown that good performance on memory tasks is related to higher resting alpha power (Klimesch, Vogt, & Doppelmayr, 1999), and larger alpha ERD is associated with better performance on a task requiring semantic search (Doppelmayr, Klimesch, Hodlmoser, Sauseng, & Gruber, 2005). For perceptual tasks, good performance is related to low alpha immediately preceding the stimulus (Ergenoglu et al., 2004, Hanslmayr et al., 2005) – a pattern that matches the results observed by MacLean and Arnell (2011) outside the AB interval (i.e., where greater alpha ERD/lower pre-stimulus alpha was associated with better T1 performance and better T2 performance at longer lags). Klimesch et al. (2007) suggest that this is because performance on perceptual tasks is enhanced if the cortex is already activated. The MacLean and Arnell (2011) findings suggest that the alpha levels that benefit T1 performance and T2 performance at longer lags, which also apparently benefit perceptual performance more generally (Ergenoglu et al., 2004), are costly for T2 performance at shorter lags during the AB interval.

Finally, several studies have shown a positive relationship between alpha ERD and resting alpha power (tonic), in that greater desynchronization is seen in response to greater alpha power (see Doppelmayr et al., 1998, Klimesch, 1999 for reviews).

Therefore, based on the within participant findings of MacLean and Arnell (2011), that greater alpha ERD is associated with poor T2 performance at short lags but better T2 performance at long lags and better T1 performance, and the positive relationship between resting alpha and alpha ERD (e.g., Klimesch, 1999), we hypothesized that individual differences in resting alpha power would predict AB magnitude in that lower resting alpha would be associated with better T2 accuracy at short lags, and therefore smaller AB magnitudes. Low resting alpha power would be expected to be accompanied by either high resting beta or high resting theta. We hypothesized that high resting beta, if accompanied by low resting alpha (indicating a state of attentional investment), would predict improved AB performance. However, high resting theta, if accompanied by low resting alpha (reflecting greater drowsiness), should not correlate with AB performance (Laufs et al., 2006). We examined these hypotheses in two independent investigations.