Conceived and designed the experiments: JL DFD KK. Performed the experiments: JL. Analyzed the data: JL CYT MM. Contributed reagents/materials/analysis tools: RE MM PSL. Wrote the paper: JL.
The authors have declared that no competing interests exist.
Prolonged demands on the attention system can cause a decay in performance over time known as the time-on-task effect. The inter-subject differences in the rate of this decline are large, and recent efforts have been made to understand the biological bases of these individual differences. In this study, we investigate the genetic correlates of the time-on-task effect, as well as its accompanying changes in subjective fatigue and mood. N = 332 subjects performed a 20-minute test of sustained attention (the Psychomotor Vigilance Test) and rated their subjective states before and after the test. We observed substantial time-on-task effects on average, and large inter-individual differences in the rate of these declines. The 10-repeat allele of the variable number of tandem repeats marker (VNTR) in the dopamine transporter gene and the Met allele of the catechol-o-methyl transferase (COMT) Val158Met polymorphism were associated with greater vulnerability to time-on-task. Separately, the exon III DRD4 48 bp VNTR of the dopamine receptor gene DRD4 was associated with subjective decreases in energy. No polymorphisms were associated with task-induced changes in mood. We posit that the dopamine transporter and COMT genes exert their effects by increasing dopaminergic tone, which may induce long-term changes in the prefrontal cortex, an important mediator of sustained attention. Thus, these alleles may affect performance particularly when sustained dopamine release is necessary.
Sustaining attention for a length of time is a prerequisite for performing many cognitive tasks. Failure to do so typically results in a vigilance decrement, or time-on-task (TOT) effect
Recently, investigators have observed that there is substantial inter-individual variability in the rate of TOT decline, with the most resilient subjects showing almost no decrement over a 20–30 minute period
The existence of such biomarkers supports the hypothesis that time-on-task effects arise because resources in the brain are finite, and must be economically deployed during times of high attentional demand. If this is the case, individual differences in TOT vulnerability may be determined either by resource availability, or the ability of the neural system to tap into this reservoir. It has been proposed that dopamine (DA) may be one of the resources in question
In populations with ADHD, all of the alleles listed above (with the exception of the DRD2 TaqI polymorphism) have been linked to cognitive outcomes, in particular measures of executive attention such as response inhibition
Aside from the paucity of data on healthy individuals, studies of the effects of dopamine on cognition have focused largely on global variables, and have not examined the degradation of performance over time. As slope variables may measure a different facet of brain function than global variables, two competing hypotheses present themselves. Maintaining a steady level of functioning may depend on adequate dopamine signaling during the period of task performance in much the same way this benefits global outcomes. If so, we would expect individuals with alleles promoting higher levels of dopamine to show reduced TOT effects. Alternatively, TOT declines may be more greatly affected by efficient baseline levels of functioning, as suggested by previous fMRI data
This study was approved by the Institutional Review Board of the National University of Singapore, and all subjects provided written informed consent before participating in the study.
A sample of N = 350 undergraduates, graduate students and staff members from the National University of Singapore were recruited through online advertising and word-of mouth. Subjects were pre-screened via a short telephone interview to ensure that they met all inclusion criteria. To qualify for the study, participants needed to be between 18 and 35 years of age, and ethnically Han Chinese. This latter criterion was chosen to decrease the possibility of artifacts owing to ethnic stratification, as Singapore is a racially diverse immigrant society. We excluded subjects who admitted to chronic physical or mental illness, had been diagnosed with a sleep disorder, or who were taking long-term medication. Subjects were also instructed to obtain a full night (>7 hours) of sleep for the 2 nights prior to the study, and to refrain from caffeine and alcohol for 6 hours prior to coming into the lab. All testing took place in the Cognitive Science Laboratory of Temasek Laboratories, Singapore.
On entering the lab, subjects first provided self-reports of their sleep history and alcohol/medication use over the previous 48 hours. They then provided self-ratings of several subjective states (fatigue, stress, anxiety, depression, sleepiness, motivation and energy) on a 9-point Likert-type scale. All questions took the form “How fatigued/sleepy/anxious etc. do you feel?” Following this, subjects were asked to remove their wristwatches and turn off their cell phones. They were then given instructions for performing the Psychomotor Vigilance Test (PVT)
To check on the test-retest reliability of the PVT, a subset (N = 56) of participants returned to the lab one week after their initial session to complete the PVT a second time. Sleep history, caffeine and substance use were similarly controlled for this session.
Subjects were reimbursed $15 for their participation in the experiment ($25 for subjects who visited the lab twice).
The Psychomotor Vigilance Test (PVT) is a test of simple reaction time that is mentally demanding because of its high stimulus-load. The standard version of the test is 10 minutes in length; however, in order to elicit greater TOT effects, a 20-minute version was administered in this experiment. The Windows PennPVT (Pulsar Informatics, Philadelphia, PA) was used for stimulus presentation. During the test, subjects are required to monitor a small box subtending approximately 4.1 (width)×1.2 degrees (height) of visual angle for the appearance of a millisecond counter, whereupon they respond with a button press on the keyboard (space bar) as quickly as they can. The inter-stimulus interval of this counter lasts from 2 to 10 seconds (mean = 6 s), and RTs are uniformly distributed across this range of ISIs. Subjects are given 1 s after the counter stops to read their reaction time. Further details of the PVT and its psychometric properties can be found in references
DNA was extracted from saliva samples, which were collected with Oragene DNA OG-500 tubes (DNA Genotek Inc., Ontario Canada) according to the manufacturer's protocol. Single nucleotide polymorphisms (SNPs) were analyzed by Sequenome MassArray genotyping; this analysis was performed at the Analytical Genetics Technology Centre, Princess Margaret Hospital, Toronto, Canada. The variable number of tandem repeats (VNTR) marker in the DAT gene was analyzed by PCR with ReddyMix™ PCR Master Mix (Thermo Scientific, Waltham, MA, USA). Primer sequences were: forward
Data analysis was performed using SPSS for Windows, Version 17.0 and MATLAB R2011B. Bivariate associations between objective and subjective data were assessed using Pearson's correlations. To reduce the possibility of Type I error in analyzing the genetic data, we restricted our search to six dopaminergic alleles, consisting of two VNTRs and four SNPs. We created two groups from each gene as follows: dopamine transporter gene (DAT1) VNTR (10/10 vs. 10/9, 9/9, 11/10 and other rare variants), DRD4 VNTR (2/2 and 2/4 vs. 4/4 and other rare variants), dopamine receptor genes DRD4 -521C/T (T/T vs. C/T and C/C), DRD2 TaqIA (any A1 allele vs. A2/A2), DBH Taq I (A/A vs. A/G and G/G) and COMT Val/Met (Val/Val vs. Val/Met and Met/Met). Overall univariate analysis of covariance (ANCOVA) was performed using, in turn, the three TOT measures (RRT slope, T1 and T2) as dependent variables, allele subgroups as between-subjects fixed factors, and subjective change in energy as a covariate. A separate ANCOVA was conducted using subjective change in energy as the dependent variable, allele subgroups as between-subjects fixed factors, and RRT slope as a covariate.
Data from 18 participants were excluded for the following reasons: not complying with instructions to obtain sufficient sleep (n = 11) and suspected non-compliance to task instructions (lapses and/or false alarms >20; n = 7). Examination of the PSQI revealed that a substantial number of individuals reported habitual poor sleep (PSQI>5; n = 72) as defined by the standards of Buysse et al.
Overall, subjects performed the PVT well, as indicated by the low number of lapses (mean [SD] = 2.59 (3.28)) and false alarms (mean [SD] = 1.59 (2.17)). The average median response time over the entire sample was 264.3 ms (SD = 26.2). Inspection of the RRT slope showed that there was a decline in performance on average over the 20-minute period (mean [SD] = −0.026 (0.020)), as well as a substantial amount of interindividual variation around that mean (min = 0.0312; max = −0.089). More intuitively, raw reaction times were, on average, 14.2% slower during the last compared to the first four minutes of the task, with the worst performer showing a 101.4% increase, and the best a 12.9% decrease. Interestingly, and in agreement with previous studies of this phenomenon, 28 out of 332, or 8.4% of participants showed either no vigilance decrement, or a slight improvement in performance over time.
We measured the stability of the time-on-task effect by inviting a subset of participants (n = 56) to undergo an identical second session of testing. Intra-class correlation coefficients were computed on these results to assess the test-retest reliability of TOT vulnerability. Between session (trait-like) reliability was significant, but moderate (ICC1,1 = .54,
Time-on-task slope is significantly correlated across two test sessions spaced one week apart (ICC1,1 = .54,
In addition to the RRT slope, we obtained parameters T1 and T2 using the exponential fit in equation 1. While T2 was normally distributed, T1 showed a bimodal distribution, thus necessitating the use of bootstrapping in order to compare between-group means. We created a bootstrapped distribution using 10,000 draws from our original dataset, and calculated the difference of the adjusted means (accounting for subjective energy change) of T1 between each allele group to obtain an estimated p-value for each of these genes. Comparison of these p-values with those obtained from the straightforward ANCOVA revealed that they are extremely similar; we thus report results from the ANCOVA for simplicity.
In our main sample, performing the PVT resulted in significant changes in all subjective variables, with the largest differences in fatigue, sleepiness and energy (
Subjective variable | Pre-test average [SD] | Post-test average [SD] | Pre-post task correlation ( |
|
Fatigue | 3.39 (2.07) | 5.54 (2.13) | −18.82 |
.561 |
Stress | 3.54 (1.76) | 3.87 (1.71) | −4.26 | .684 |
Anxiety | 2.97 (1.60) | 3.26 (1.69) | −3.97 | .674 |
Sleepiness | 3.28 (1.51) | 4.99 (1.80) | −19.78 | .564 |
Depression | 2.18 (1.29) | 2.52 (1.51) | −6.40 | .778 |
Motivation | 4.95 (1.40) | 4.39 (1.34) | 8.25 | .624 |
Energy | 5.28 (1.17) | 4.17 (1.34) | 17.94 | .611 |
degrees of freedom = 311 due to missing data.
All t and r values significant at
Variable | Factor 1 (subjective energy) | Factor 2 (mood) |
Fatigue |
|
.112 |
Stress | .092 |
|
Anxiety | −.033 |
|
Sleepiness |
|
.136 |
Depression | .191 |
|
Motivation |
|
.016 |
Energy |
|
−.101 |
Bold values indicate factor loaded on to by each variable.
We performed bivariate correlations to determine the relationship between objective performance declines and subjective mental states. RRT slope was correlated with the weighted index of subjective changes in energy (
Allele frequencies are presented in
Genetic polymorphism | Genotype | Allele frequencies | Hardy-Weinberg equilibrium | |||||
DAT1 VNTR | 9/9 | 9/10 | 9/11 | 10-repeat | 9-repeat | 11-repeat | Others | Yes ( |
1 (0.3%) | 40 (12.1%) | 0 (0.0%) | 631 (91.2%) | 42 (6.1%) | 14 (2.0%) | 5 (0.7%) | ||
10/10 | 10/11 | 11/11 | ||||||
287 (87.0%) | 12 (3.6%) | 1 (0.3%) | ||||||
DRD4 VNTR | 2/2 | 2/4 | 2/5 | 4-repeat | 2-repeat | 5-repeat | Others | Yes ( |
12 (3.8%) | 89 (27.9%) | 5 (1.6%) | 519 (74.8%) | 140 (20.2%) | 14 (2.0%) | 21 (3.0%) | ||
4/4 | 4/5 | 5/5 | ||||||
206 (64.6%) | 6 (1.8%) | 1 (0.3%) | ||||||
DRD4 -521C/T | T/T | T/C | C/C | T allele | C allele | Yes ( |
||
150 (45.2%) | 131 (39.5%) | 48 (14.5%) | 431 (67.0%) | 227 (33.0%) | ||||
DRD2 TaqIA | T/T | T/C | C/C | T allele | C allele | Yes ( |
||
112 (33.7%) | 158 (47.6%) | 55 (16.6%) | 382 (58.8%) | 268 (41.2%) | ||||
COMT Val/Met | G/G | G/A | A/A | G allele | A allele | Yes ( |
||
183 (55.1%) | 121 (36.4%) | 20 (6.0%) | 487 (75.2%) | 161 (24.8%) | ||||
DBH TaqI | A/A | A/G | G/G | A allele | G allele | Yes ( |
||
247 (74.4%) | 70 (21.1%) | 3 (0.9%) | 566 (88.2%) | 76 (11.8%) |
We first examined the genetic data using a univariate approach to determine which of our key outcome variables differed depending on genotype group. These outcome variables were RRT slope and the indices for subjective change in energy and mood. None of the covariate by group interactions was significant at the
Means and standard errors for reciprocal reaction time slope in each allele group. * represents
Curves were plotted using predicted values calculated from average parameters A, T1 and T2 for each allele group. The thinner dotted curves represent the mean ± one standard error. Note that the curvature of the trends is only slight despite the use of an exponential equation; this is due to the bimodality of T1, for which large values cause the curve to approach linearity. Panel A: COMT allele groups. Panel B: DAT1 allele groups.
Source | Type III SS | df | Mean square | F | Significance | Partial η2 |
Full model | 0.013 | 7 | 0.002 | 4.93 | <.001 | .103 |
Intercept | 0.029 | 1 | 8×10−6 | 76.67 | <.001 | .203 |
|
||||||
DRD4 VNTR | 0.001 | 1 | 0.001 | 3.87 | .051 | .013 |
DAT1 VNTR | 0.002 | 1 | 0.002 |
|
.033 | .015 |
DRD4 -521 C/T | 0.001 | 1 | 0.001 | 1.64 | .201 | .005 |
DRD2 Taq1A | 0.000 | 1 | 0.000 | 1.07 | .302 | .004 |
COMT Val/Met | 0.002 | 1 | 0.002 |
|
.026 | .016 |
DBH TaqI | 8.5×10−5 | 1 | 8.5×10−5 | 0.22 | .684 | .001 |
|
||||||
Subjective change in energy | 0.008 | 1 | 0.008 | 21.5 | <.001 | .067 |
Values in bold text are significant at
T1 | T2 | |||||||||||
Source | Type III SS | df | Mean square | F | Significance | Partial η2 | Type III SS | df | Mean square | F | Significance | Partial η2 |
Full model | 2990.1 | 7 | 427.2 | 1.32 | .211 | .028 | 0.086 | 7 | 0.012 | 3.31 | .001 | .077 |
Intercept | 22722.4 | 1 | 22722.4 | 66.61 | <.001 | .181 | 0.159 | 1 | 0.159 | 46.28 | <.001 | .133 |
|
||||||||||||
DRD4 VNTR | 365.2 | 1 | 365.2 | 1.07 | .302 | .004 | 0.007 | 1 | 0.007 | 2.27 | .151 | .007 |
DAT1 VNTR | 70.1 | 1 | 70.1 | 0.11 | .651 | .001 | 0.020 | 1 | 0.020 |
|
.017 | .019 |
DRD4 -521 C/T | 59.6 | 1 | 59.6 | 0.28 | .676 | .001 | 0.006 | 1 | 0.006 | 2.18 | .185 | .006 |
DRD2 Taq1A | 38.1 | 1 | 38.1 | 0.06 | .739 | .000 | 0.002 | 1 | 0.002 | 0.78 | .400 | .002 |
COMT Val/Met | 704.5 | 1 | 704.5 | 2.15 | .152 | .007 | 0.018 | 1 | 0.018 |
|
.024 | .017 |
DBH TaqI | 0.1 | 1 | 0.1 | 0.01 | .984 | .000 | 0.008 | 1 | 0.008 | 2.36 | .131 | .008 |
|
||||||||||||
Subjective change in energy | 1401.2 | 1 | 1401.2 | 4.78 | .044 | .013 | 0.031 | 1 | 0.031 | 9.82 | .002 | .029 |
Values in bold text are significant at
Given the effects of DAT1 and COMT on TOT, we wished to further probe the effect of these alleles on other common PVT parameters. Accordingly, we extracted the number of lapses (responses>500 ms) and false alarms, and computed a metric of attentional stability
When confronted with a vigilance task, it is possible that subjects may withhold effort at the beginning of the trial so as to conserve resources, and thus better maintain performance. As part of the study setup, we took steps to mitigate this issue by instructing participants to exert maximum effort at all times, and by having an experimenter observe the subject throughout the study. We checked on this potential confound by correlating RRT slope with mean reaction time from the first minute of the PVT, and found a small but significant positive correlation between them (r = .13,
For the subjective data, the overall effect of the six dopamine polymorphisms was also significant (F7,308 = 3.43, p = .002); however, only the DRD4 VNTR (F1,325 = 13.71,
Means and standard errors for average subjective energy change in each allele group. * represents
Source | Type III SS | df | Mean square | F | Significance | Partial η2 |
Full model | 11.9 | 9 | 1.70 | 3.43 | .002 | .074 |
Intercept | 48.2 | 1 | 48.2 | 97.37 | <.001 | .244 |
|
||||||
DRD4 VNTR | 6.78 | 1 | 6.78 |
|
<.001 | .044 |
DRD1 VNTR | 0.12 | 1 | 0.12 | 0.24 | .624 | .001 |
DRD4 -521 C/T | 0.01 | 1 | 0.01 | 0.01 | .905 | .000 |
DRD2 Taq1A | 1.19 | 1 | 1.19 | 2.40 | .122 | .008 |
COMT Val/Met | 0.17 | 1 | 0.17 | 0.33 | .564 | .001 |
DBH TaqI | 0.15 | 1 | 0.15 | 0.30 | .585 | .001 |
|
||||||
RRT slope | 4.40 | 1 | 4.40 | 8.90 | <.003 | .029 |
Values in bold text are significant at
The results of this study demonstrate links between several functional dopaminergic alleles and the propensity to both decline in performance and feel mentally fatigued. Specifically, our data suggest that the dopamine transporter, DAT1, as well as COMT, may have an impact on the rate at which the vigilance decrement occurs, and that the dopamine receptor DRD4 may be related to subjective declines in mental energy. This is one of the first demonstrations that these polymorphisms play such a role in attention in non-clinical populations.
For the two genes that were associated with TOT, alleles that typically confer risk of poorer cognitive performance (i.e. the Val/Val allele
The direction of the dopamine effect in this study may be best explained by the tonic-phasic model of dopamine regulation. In a seminal set of papers, Grace and colleagues
In the long-term, the tonic and phasic DA systems are not independent. For instance, stimulants such as cocaine block reuptake of DA from the synaptic cleft
As mentioned previously, alleles associated with greater DA availability are usually found to be of benefit in a range of cognitive tasks. However, our data suggest that slope variables may be affected much more greatly by DA tone than phasic DA release, thus explaining the inferior performance of individuals with DA-promoting alleles. Differences between DAT1 and COMT genotype groups in this sample were found only for TOT variables, and not lapses, false alarms or reaction time variability, despite the fact that these variables are highly inter-correlated. We also found evidence that some subjects were able to ameliorate their level of TOT decline by withholding effort (possibly unconsciously), as mean first-minute reaction times were positively correlated with RRT slope. Nevertheless, DAT1 and COMT genotype groups did not differentiate performers on this variable either. We therefore speculate that individuals with chronically high DA availability may more quickly exhaust the benefits of greater phasic release when longer-term attentional effort is required, due to the long-term plastic changes in DA neurons in the striatum and PFC described above. These effects may be especially prominent when elicited by the PVT, which is a task with consistently high signal load and attentional demand.
The assertion that tonic, resting levels of DA may influence TOT changes is consistent with data from a recent study in which Lim et al.
While both DAT1 and COMT were found to have an effect on PVT performance in this study, presumably via the modulation of levels of extracellular DA, we note that these molecules likely influence this phenotype via different sites in the brain. COMT acts to degrade DA primarily in PFC, with a ∼40% difference in enzyme activity between the Val and Met alleles
By decomposing individual subject curves into two components using a set of exponential functions, we found that DAT1 and COMT polymorphisms were associated with differences in the slow, but not the fast decaying component. Effect sizes of genotype on the slow component were, in fact, slightly larger than for the simple linear fit. These data suggest that dopamine exerts its effects on longer-term TOT decay, which lends credence to the hypothesis that the direction of the effects observed in this experiment (i.e. COMTMet<COMTVal; DAT110-repeat absent<DAT110-repeat present) may only be observable when the dependent measure is a slow-evolving process.
As expected, performing the PVT for 20-minutes resulted not only in robust TOT decrements, but also significant declines in energy and mood. This is consistent with findings from previous studies which suggest that vigilance tests are a resource-demanding form of mental work
Interestingly, subjective changes in energy were strongly associated with the DRD4 polymorphism; subjects with at least one copy of the 2-repeat allele tended to show a greater subjective change in energy over time. To our knowledge, this is the first association of the allele with this effect, although previous work has implicated DAT1, COMT and DRD2 in mental fatigue
Finally, significant increases in anxiety, stress and depression represent a separate problem that is putatively caused by high mental workload. Our data suggested that changes in mood are not directly associated with TOT; nevertheless, they represent an undesirable side effect that may lead to other negative consequences. Further research is needed to characterize how these changes are instantiated in the brain, and how they might affect behavior and performance.
The current study has a small number of limitations. First, subjects were asked to abstain from caffeine, which may have caused some regular users to experience a withdrawal effect. However, data collected from other studies by the first author indicate that caffeine usage in Singapore students is low, and we thus suggest that any such effects were relatively minor. Second, and more importantly, we note that our sample size was moderate for a study of this nature, and that the effects observed were in the small to medium range, exposing us to the possibility of Type I error. Nevertheless, we count two points in our favor. First, we had strong
We note that the reliability of the TOT effect, as is typical with change scores, is only in the moderate range (.54), as are the intra-class correlations of the subjective measures used in this paradigm. This is not unexpected, as difference or slope scores by their nature have lower reliability than measures of central tendency
An increasing amount of evidence points to TOT vulnerability as being the result of the cortical attention system drawing on a brain-limiting resource that is determined by resting levels of neuronal activity. Our genetic data strengthen the case that one of the resources in question may be dopamine, and that these individual differences may be more highly related to tonic brain functioning than phasic task-related activity. We further demonstrate that TOT vulnerability is associated with changes in energy but not mood, and that DRD4 is robustly related to the feelings of energy depletion. The results of this study have implications for the emerging field of neuroergonomics
(DOCX)
We acknowledge the assistance of Gabriel Chew, Sheralyn Tan, Michael Loh, He Haitao, Tania Kong and Tan Ying Ying in data collection, analysis and literature review, and Xiong Gaogao and Zhu Qingdi in DNA extraction and analysis.