This preliminary report examined the influence of sleep state on the neural mechanisms underlying the anticipatory biasing of spatial attention. Previous reports have shown that sleep deprivation impairs performance and alters brain activity when subjects perform tasks targeting attention, verbal learning, and working memory (Chee and Choo, 2004
; Drummond et al., 2000
; Drummond et al., 2001
; Thomas et al., 2000
; Wu et al., 1991
). The current results suggest that sleep deprivation may also impair the anticipatory allocation of attention in response to spatially predictive cues and alter the underlying neural correlates.
Following sleep deprivation, subjects performed less accurately and made significantly more errors of omission than when rested, which has been seen in other studies of sleep deprivation (Williams et al., 1959
). Although there was no main effect of sleep state on reaction time, an interaction with the type of cue did affect subjects’ responses such that valid trials were significantly faster than both neutral and invalid trials when subjects were rested but not when they were sleep-deprived. Sleep-deprived subjects also had significantly fewer trials conferring a cue benefit, despite the lack of differences in means and variances of reaction times when cue types were compared individually between sleep states. These results suggest that sleep deprivation may lead to an impairment in the anticipatory allocation of spatial attention through interacting effects on both spatial and non-spatial attentional components (de Gonzaga Gawryszewski et al., 1987
; Jongen et al., 2006
; Mesulam et al., 2001
; Small et al., 2003
In the rested state, the presence of a cue benefit was associated with activity within the PCC, whereas following sleep deprivation cue benefits were associated with activity within the IPS. Previous reports have also shown increased PCC activity when attention is shifted in response to spatially predictive cues (Hopfinger et al., 2001
; Mesulam et al., 2001
; Small et al., 2003
). This relationship between performance and brain activity was demonstrated to be independent of reaction time, per se (Small et al., 2003
). Instead, it was suggested that PCC activity was associated with the generation of a motivational bias for attending to a focal location in space (Small et al., 2003
In contrast, studies have shown that IPS activity may display the opposite relationship to spatial cues, by demonstrating decreased activity when attention is allocated predictively to a location in space (Constantinidis and Steinmetz, 2001
; Robinson et al., 1995
; Small et al., 2003
). These data have been taken to suggest IPS suppression may be necessary in order limit attentional receptivity to stimuli at unexpected locations. Consistent with these reports, the current study showed no change in IPS activity to spatial cues when subjects were rested, . However, in the SD state, IPS activity was increased for trials showing a cue benefit. Thus, when sleep-deprived, subjects appear to preferentially recruit the IPS when spatially orienting attention.
It is unclear why the relationship between cue benefit and brain activation is altered by sleep deprivation. One possibility is that PCC recruitment is impaired by sleep deprivation. This notion is supported by studies demonstrating reduced activity within the PCC following SD as compared to the rested state when subjects performed serial addition/subtraction and complex navigation tasks (Strangman et al., 2005
; Thomas et al., 2000
). Furthermore, positron emission tomography (PET) imaging of wake and non-rapid eye movement sleep (NREM) states have shown that activity in posterior cingulate is significantly reduced in NREM sleep, while medial parietal regions remain as active as when awake (Nofzinger et al., 2002
). It is often argued that in a sleep-deprived state, errors of omission predominantly represent a brief transition to NREM sleep (Dinges and Kribbs, 1991
; Williams et al., 1959
). Taken together, these data suggest that it is possible that PCC recruitment is impaired in the sleep-deprived state due to the intermittent suppression of PCC activity when subjects generate errors of omission.
Because the recruitment of the PCC in the sleep-deprived state is impaired, cue benefits may depend on a strategy other than generating a motivational bias for attending to a focal location in space. We have shown that this strategy is associated with IPS activity. Greater IPS activity in the sleep-deprived state may reflect a strategy that relies on increasing receptivity to stimuli at unexpected locations. Thus, sleep-deprived individuals may shift from a focal endogenous orienting strategy to a more global exogenous orienting strategy.
Another possibility is that sleep-deprived subjects rely more on eye-movements to perform the task than when rested. Eye movements were not monitored in the present experiment, thus this possibility cannot be definitively addressed. However, we suggest that if this was the case we would have expected to see greater activity in the frontal eye fields and lateral IPS, which was not found (Corbetta et al., 1998
; Nobre et al., 2000
). In order to examine further the linkage between the generation of cue benefits and subject performance a cue benefit score calculated as previously described (Small et al., 2003
) and regressed against BOLD responses. Results were similar to those seen in the categorical V+
analysis further supporting the possibility that PCC and IPS activations were associated with cue benefits (see Table S1
One limitation of the current study is the small number of subjects, which may potentially affect both the generalizability of the findings and the assumptions underlying the parametric statistics used to analyze the fMRI data. Random effects statistics were used to address the issue of population inference, and activations were found in the PCC and IPS. Similar sites of activation were also seen previously in other studies examining the anticipatory allocation of spatial attention (Mesulam, 2003
; Small et al., 2003
In order to address our use of cluster level parametric statistics in the setting of low degrees of freedom, we also calculated the mean FWHM values for each of the clusters, and compared these values with the average value used by SPM. In all cases, the FWHM in the cluster was smaller than the value used by SPM, suggesting that clusters identified as significant were unlikely to be false positives.
In the setting of low degrees of freedom, non-parametric statistics could have been used to analyze the fMRI data (Hayasaka et al., 2004
). However, the power of non-parametric statistics may also be reduced by the small number of subjects (n=6), which would have only allowed a limited number of resamplings (26
= 64). The constrained number of resamplings limits the lowest possible p-value to 1/64 = 0.0156, thereby reducing the power of the technique (S. Hayasaka, personal communication). In the face of limitations to both parametric and non-parametric techniques, we chose to utilize standard parametric statistics, while attempting to minimize the chance of a Type I error. Nevertheless, replication and extension of these findings in a larger study will be important.
The standard reaction time pattern (Valid RT < Neutral RT < Invalid RT) was not replicated in the rested condition. Instead, neutral RT was slower than both valid and invalid trials in the rested state. We do not have a definitive explanation for this phenomenon. This pattern of reaction times with neutral trials being slower than invalid trials has been previously reported, albeit infrequently (Amir et al., 2003
; Perchet et al., 2001
; Posner et al., 1987
). In some of these reports “neutral” trials were uncued, which was not the case in the present study. Posner (1987)
attributed this reaction time pattern to diminished transient arousal due to the lack of an alerting signal (Posner et al., 1987
). Nevertheless, we still believe that the neutral trials served as a valid comparison in the current study. The neutral trials were used to calculate V+
trials as a way of normalizing subjects’ response times, i.e., to provide a baseline across subjects for dividing valid trials into those with (V+
) and without (V−
) cueing effects. Since subjects showed similar patterns of responses across runs for each sleep condition, we were reassured by the consistency of subjects’ responses. We therefore suggest that the neutral trials represent a reasonable choice for a baseline both within and across sleep conditions. We also note that standard errors were similar in both sleep states for neutral trials, making it unlikely that changes in variance affected the determination of V+
To address further concerns about the neutral trials the analysis was also performed by using invalid trial reaction times as a baseline to categorize the valid trials Results were similar to those reported with neutral trials (see Table S2
). This result provides further support that the current analysis demonstrates an effect of sleep deprivation on the anticipatory spatial biasing of attention.
In conclusion, sleep deprivation impairs the ability to utilize a predictive cue to shift attention towards relevant locations in space. This impairment is reflected in a lack of PCC activation, which has been implicated in the generation of an anticipatory bias for target location. Instead, it appears that SD subjects recruit the IPS when allocating spatial attention. This alternate strategy may depend on enhancing receptivity to stimuli in unexpected locations, thus shifting to more global, exogenous, attentional mechanisms that would rely more on IPS recruitment. Nevertheless, this strategy appears to be less effective overall, as there was no benefit of informative cues on reaction time (comparison of valid vs. neutral cues) in the SD state. These data suggest that sleep loss may affect performance by interfering with the ability to predictively allocate attention and to suppress distractibility to irrelevant spatial events. The consequence of this is that sleep-deprived individuals may miss predictive environmental cues and react impulsively to behaviorally irrelevant stimuli. Both responses are likely to increase errors and result in accidents even while individuals appear to be awake and responding.