Edited by:
Reviewed by:
*Correspondence:
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
When attention has to be maintained over prolonged periods performance slowly fluctuates and errors can occur. It has been shown that lapses of attention are correlated with BOLD signals in frontal and parietal cortex. This raises the question how attentional fluctuations are linked to the fronto-parietal default network. Because the attentional state fluctuates slowly we expect that potential links between attentional fluctuations and brain activity should be observable on longer time scales and importantly also before the execution of the task. In the present study we used fMRI to identify brain activity that is correlated with vigilance, defined as fluctuations of reaction times (RT) during a sustained attention task. We found that brain activity in visual cortex, parietal lobe (PL), inferior and superior frontal gyrus, and supplementary motor area (SMA) was higher when the subject had a relatively long RT. In contrast to our expectations, activity in the default network (DN) was higher when subjects had a relatively short RT, that means when the performance was improved. This modulation in the DN was present already several seconds before the task execution, thus pointing to activity in the DN as a potential cause of performance increases in simple repetitive tasks.
Many tasks in our daily life require that we focus our attention and remain alert over prolonged periods of time. This sustained attention is also referred to as vigilance (
Typical laboratory vigilance tests employ repetitive tasks across long periods, for example the MacWorth’s classic clock test (
Such vigilance tasks are not well suited for fMRI experiments because they provide not enough events for the analysis. The psychomotor vigilance task (PVT) is similar to the classical vigilance tasks. Here subjects also have to respond to the presentation of a simple stimulus, however, these events are more frequent.
The continuous performance task (CPT) is slightly more complex and requires simple perceptual decision making. While performing the task subjects are frequently presented with non-target stimuli (for example a set of consonants) and not so frequently with target stimuli (for example a set of vowels). Subjects are asked to respond to the target stimuli only, and to withhold the response in other cases. In the sustained attention to respond task (SART) the response pattern is inverted. Here subjects are asked to withhold the response to target stimuli but respond to the frequent non-target stimuli.
Both PVT, CPT, and SART have been mostly used to investigate vigilance with fMRI by contrasting blocks of or responses to PVT/CPT/SART performance with blocks of or responses to control tasks that required a lower level of sustained attention (for example repetitively pressing a button).
The described vigilance tasks have been used to reveal that mostly right lateralized frontal cortex, parietal cortex, thalamus and the brain-stem are involved in tasks that require a high level of sustained attention vs. tasks that require a low level of sustained attention (
However, if the activity of a brain region is associated with tasks that require high levels of sustained attention compared to low level of attention this could simply reflect the different difficulty levels or the workload of the two tasks. In order to confirm that activity in a brain region is really associated with fluctuations of the attentional state the activation has to meet the following two criteria:
(1) The activity has to reflect
(2) The activity has to
Here we used a CPT with interspersed prospective memory events in order to identify brain regions that show such a response profile. The prospective memory events were not analyzed here and will be subject of another paper. The vigilance state is reflected in time-varying reaction times (RT). For example it has been shown that sleep deprived (and thus less vigilant) subjects show longer RTs (
Twenty-two participants between the ages of 19 and 34 (mean age 25.83, 11 female) took part in the fMRI study. They had normal or corrected-to-normal vision. No participant reported a history of neurological or psychiatric disorders. One participant was excluded due to strong motion (more than 5 mm) during the scanning session. All participants gave informed consent and were compensated with €10 for every hour they participated in the experiment. The study was approved by the ethics committee of the Faculty of Psychology at the Humboldt-Universität zu Berlin (Antrag 2013-34). All subjects gave written informed consent in accordance with the Declaration of Helsinki.
Subjects were asked to perform a sustained attention task that has previously been used in prospective memory rather than vigilance research and that contains a continued load requiring high attention (
Illustration of three trials of the paradigm. A 4 × 4 grid with a triangle and another random polygon was presented every 3 s for 500 ms. In most trials (90%), the ongoing trials (OG), subjects indicated by button-press whether the non-triangle-shape was positioned left or right to the triangle (adapted from
A new configuration of a triangle and polygon was presented every 3 s for a duration of 500 ms with a 2500 ms inter-stimulus-interval. Subjects were asked to respond to the stimuli as soon they were visible and responses were considered valid until the onset of the next stimulus. During each run of the experiment, 204 trials were presented. Twenty trials (9.8%) of the trials involved the PM task. These were randomly distributed across the run with a minimum of 4 OG trials between two PM trials. The first four trials of reach run were always OG trials.
Before the scanning sessions subjects completed three runs of the experiment outside the fMRI scanner in order to familiarize themselves with the task and to avoid training effects during the scanning session. During the fMRI session subjects completed six runs of the experiment. Presentation was controlled and responses were recorded using the Cogent toolbox
Gradient-echo EPI functional MRI volumes were acquired with a Siemens TRIO 3 T scanner with standard head coil (33 slices, TR = 2000 ms, echo time TE = 30 ms, resolution 3 mm × 3 mm × 3 mm with 0.75 mm gap, FOV 192 mm × 192 mm). In each run 309 images were acquired for each participant. The first three images were discarded to allow for magnetic saturation effects. For every subject six runs of functional MRI were acquired. We also acquired structural MRI data (T1-weighted MPRAGE: 192 sagittal slices, TR = 1900 ms, TE = 2.52 ms, flip angle = 9°, FOV = 256 mm × 256 mm).
Data were preprocessed using SPM8.
First we used a GLM based fMRI analysis to investigate which individual voxels were activated during the task and furthermore modulated by RT during the trial execution. We assumed a time lag of the BOLD signal by using a canonical haemodynamic response function (HRF). In this analysis we applied a univariate GLM (
The second analysis (cross-correlation) extended the first analysis in a way that did not assume any fixed time lag between RT and BOLD signal, therefore we investigated the link between the BOLD-response and RT on longer timescales
Behavioral results. Participants responded faster to the OG compared to the prospective memory (PM) trials. The accuracy on the OG trials was higher compared to the PM trials. Error bars show the standard error of the mean (∗∗∗
We also analyzed the RT of correct OG trials directly preceding PM trials. There was no difference in RT of OG trials preceding correct vs. incorrect PM trials [OGbefore_correct_PM: mean = 805.42 ms; SEM = 30.39; OGbefore_incorrect_PM: mean = 806.14 ms; SEM = 33.02;
We conducted an exploratory analysis to investigate whether the performance on the OG trials depends on the similarity to the PM trials. Therefore, we considered the Euclidian distance between the two presented shapes. We calculated the correlation between the absolute difference between the Euclidian distance of the 8 possible OG configurations and the PM configuration and the RT (OG-PM). The correlation was not significantly different from 0 (
The first analysis aimed to reflect previous studies of vigilance and focused on whether canonical BOLD responses elicited by ongoing trials reflected performance fluctuations. Thus, this analysis did not look across longer time scales. Brain responses in the insula, inferior temporal gyrus/V5, middle frontal gyrus, inferior frontal gyrus, supplementary motor area (SMA), postcentral gyrus, inferior parietal lobe, precentral gyrus, early visual cortex, thalamus, and cerebellum were positively modulated by the participants’ response times (
Results of the parametric general linear model. Positively modulated regions (red) are more active when participants respond relatively slow. Negatively modulated regions (blue) are more active when participants respond relatively fast. Brain regions that are deactivated by the task (green) highly overlap with regions that are positive modulated (yellow) (
Illustration of the cross-correlation analysis.
HRF model; positive modulation
Anatomical area | L/R | |||||
---|---|---|---|---|---|---|
Insula | R | 11.15 | Inf | 33 | 20 | 7 |
L | 11.98 | Inf | –30 | 17 | 7 | |
V5; inferior temporal | R | 7.52 | 6.95 | 48 | –58 | –8 |
gyrus | L | 9.14 | Inf | –39 | –61 | –5 |
Middle frontal gyrus | R | 6.59 | 6.19 | 39 | 35 | 16 |
L | 7.96 | 7.29 | –39 | 32 | 28 | |
Inferior frontal gyrus | R | 13.29 | Inf | 48 | 8 | 25 |
L | 12.85 | Inf | –48 | 5 | 28 | |
Supplementary motor area | 10.48 | Inf | 6 | 8 | 49 | |
Postcentral gyrus | R | 13.00 | Inf | 45 | –37 | 49 |
(Inf parietal lobe) | L | 14.90 | Inf | –48 | –34 | 46 |
Inferior parietal lobe | R | 12.96 | Inf | 30 | –49 | 46 |
L | 14.94 | Inf | –33 | –43 | 43 | |
Precentral gyrus; Area 6 | R | 13.65 | Inf | 27 | –4 | 52 |
L | 14.35 | Inf | –24 | –7 | 52 | |
Visual cortex; Area 17 | R | 5.53 | 5.28 | 21 | –61 | 4 |
L | 5.12 | 4.92 | –18 | –67 | 7 | |
Thalamus | R | 5.86 | 5.57 | 12 | –16 | –2 |
L | 5.68 | 5.41 | –12 | –19 | 1 | |
Cerebellum; Lobule VI | R | 5.81 | 5.53 | 33 | –46 | –26 |
Cortical responses in several regions overlapping with the default network, superior medial frontal lobe/ACC, bilateral temporal parietal junction (TPJ), and precuneus, were negatively modulated by response times (
HRF model; negative modulation
Anatomical area | L/R | Z-value | ||||
---|---|---|---|---|---|---|
Superior medial frontal lobe; ACC | 9.52 | Inf | –3 | 56 | 4 | |
Angular gyrus; TPJ | R | 9.06 | Inf | 54 | –64 | 37 |
L | 8.39 | 7.62 | –51 | –70 | 34 | |
Precuneus | 7.89 | 7.24 | –3 | –46 | 34 | |
Middle temporal gyrus | R | 6.40 | 6.03 | 63 | –16 | –14 |
Middle frontal gyrus | L | 4.99 | 4.80 | –39 | 14 | 58 |
Cerebellum; Lobule VIIa Crus I | R | 6.60 | 6.20 | 33 | –82 | –35 |
Cerebellum; Lobule VIIa Crus I | L | 6.29 | 5.93 | –30 | –79 | –32 |
A number of brain regions showed task dependent deactivation (
The results of the whole brain cross correlation analysis are shown in
Results of the cross correlation analysis between the (time shifted) fMRI signal and the response times. The Fisher-Z normalized correlation maps with positive lags (3 and 8 s) are very similar to the GLM results (
We then further investigated the temporal development of signals in the regions of interest (ROI) obtained by analysis I. We created two masks that contained all positive or all negative modulated voxels (
Results of the cross correlation analysis between the (time shifted) fMRI signal and the response times for selected regions. Upper row shows Fisher-Z normalized and averaged correlation coefficients and standard error of the mean for different time lags. Lower row shows –log(p) of the
In the present study we implemented a prospective memory task as a type of a CPT to investigate neural correlates of vigilance, as defined by fluctuations of the trial-by-trial performance (RT). We identified two large networks that were modulated by RT, one positively and one negatively. Importantly, RT modulated activity of the default network negatively
The two networks that were modulated by RT had different temporal profiles.
A large number of brain regions, such as the insula, inferior temporal gyrus, middle frontal gyrus, inferior frontal gyrus, supplementary motor area, visual cortex and parietal cortex, showed a positive modulation with RT. This means, activity in these regions was higher when subjects responded relatively slower. Importantly, the activity in these regions was modulated by RT mostly
Brain regions that showed a task dependent deactivation to a large degree overlapped with regions that were also positively modulated by RT. This finding supports the previous interpretation, that regions showing a positive modulation are task specific. In contrast regions that were negatively modulated by RT are candidates for a task independent vigilance modulation.
More interestingly, the default network (DN) (medial frontal lobe/ACC, precuneus, angular gyrus/TPJ) showed a negative modulation with RT. This means, in these regions activity was higher when subjects responded relatively faster. Importantly, activity in these regions was modulated by RT not only after but also
This result might seem surprising because the DN has been defined as a network that showed increased activity in rest conditions compared to different task conditions (
In this study we focus on the performance in the OG task. There is the possibility that performance in the OG task is anti-correlated with performance in the PM task, because attention is directed to task demands specific to the OG task or vice versa. RT of OG trials directly preceding PM trials were not different between correct and incorrect PM trials. This result suggests that in our data a pre-error fastening in the PM task (
Increased activity during episodes of rest or during episodes of decreased performance could be linked to processes such as daydreaming or mind wandering. Mind wandering during the execution of a task could also explain poor performance in the form of increased error rates or increased RTs. Indeed, DN activity has been linked to mind wandering in several studies (
Taken these results together this shows that DN activity seems not exclusively be linked to poor performance. Furthermore,
Furthermore, DN activity has been identified to be linked with good performance in previous sustained attention studies.
It has been shown that the strength of the anti-correlation between the DN and the task-positive network mediates behavioral variability in a flanker task (
Finally, the important role of prefrontal cortex during a vigilance task was directly demonstrated with a transcranial direct current stimulation study (tDCS) (
We have introduced a new version of a CPT here. Brain regions that modulate vigilance should, in theory, show similar responses for different tasks. Therefore, we think that a variety of different tasks that all require sustained attention are useful in order to identify brain regions that modulate
More research is needed to investigate the circumstances under which DN activity is associated with poor or with good performance. Based on the present findings in combination with previous research we conclude that DN activity is related to good performance if participants are engaged in a demanding sustained attention task during which it is required to monitor the external environment in order to decide which task to perform. Furthermore, during such task requirements DN activity predicts task performance even before the execution of the task. Therefore, it is likely that DN activity reflects the attentional state during certain vigilance tasks.
CB planned the study, collected and analyzed the data and wrote the manuscript. AV programmed the experiment, collected data. PZ collected data, analyzed the experiment. J-DH planned the study, wrote the manuscript.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
This work was funded by the Bernstein Computational Neuroscience Program of the German Federal Ministry of Education and Research (BMBF grant 01GQ0411), the Bernstein Focus Neurotechnology (BMBF grant 01GQ0851), the German Research Foundation (DFG FK:JA945/3-1), the Excellence Initiative of the German Federal Ministry of Education and Research (DFG grant GSC86/1-2009), and the European Regional Development Fund of the European Union (10153458 and 10153460).