- Email: [email protected]
Using interstimulus interval to maximise sensitivity of the Psychomotor Vigilance Test to fatigue
G Model
ARTICLE IN PRESS
AAP-3941; No. of Pages 5
Accident Analysis and Prevention xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
Using interstimulus interval to maximise sensitivity of the Psychomotor Vigilance Test to fatigue Raymond W. Matthews a,b,∗ , Sally A. Ferguson a , Charli Sargent a , Xuan Zhou a , Anastasi Kosmadopoulos a , Gregory D. Roach a a b
Appleton Institute for Behavioural Science, Central Queensland University, PO Box 42, Goodwood, South Australia 5032, Australia Centre for Sleep Research, University of South Australia, GPO Box 2471, Adelaide, South Australia 5001, Australia
a r t i c l e
i n f o
Article history: Received 28 June 2015 Received in revised form 4 October 2015 Accepted 13 October 2015 Available online xxx Keywords: Psychomotor Vigilance Task Fatigue Sleep Neurocognitive performance Preparedness Forced desynchrony
a b s t r a c t There is some evidence that short interstimulus intervals (ISIs) on the Psychomotor Vigilance Test (PVT) are associated with longer and more varied reaction times (RTs). Preparation processes may impede RT following short ISIs, resulting in additional unexplained variance. The aims of this study were to investigate whether there is an effect of ISI on RT and errors within the PVT, and whether such an effect changes with three elements of fatigue: time of day, prior wake and time on task. Twelve male participants completed 49 PVTs across 7× 28 h periods of forced desynchrony. For analysis, RTs, reciprocal reaction times (1/RT), false starts and lapse responses within each 10 min session were assigned to a 1-s ISI group, a 2-min time of task group, a 2.5-h PW level and a 60◦ phase of the circadian rhythm of core body temperature (as a measure of time of day). Responses following short ISIs (2–5 s) were significantly slower and more varied than responses following longer ISIs (5–10 s). The likelihood of a lapse was also higher for short ISIs, while the probability of a false start increased as a function of ISI. These effects were independent of the influences of time of day, prior wake and time on task. Hence, mixed model ANOVAs comprising only long ISIs (5–10 s) contained stronger effect sizes for fatigue than a model of all ISIs (2–10 s). Including an ISI variable in a model improved the model fit and explained more variance associated with fatigue. Short ISIs resulted in long RTs both in the presence and absence of fatigue, possibly due to preparation processes or ISI conditioning. Hence, omitting short ISI trials from RT means or including an ISI variable in analysis can reduce unwanted variance in PVT data, improving the sensitivity of the PVT to fatigue. © 2015 Elsevier Ltd. All rights reserved.
Over the last three decades the Psychomotor Vigilance Task (PVT) has become the ubiquitous behavioural assay to measure the effects of fatigue and sleepiness (Dorrian et al., 2005). The PVT is a sustained attention, stimulus-response task developed by Dinges and Powell (1985). It was developed as an evolution of Wilkinson’s earlier simple visual reaction time (VRT) task (Glenville et al., 1978; Wilkinson and Houghton, 1982), which was itself based on the previous auditory reaction time (ART) task by Lisper and Kjellberg (1972). The standard form of the PVT is a 10-min test containing approximately 90 response trials. Each trial consists of a stimulus
∗ Corresponding author at: Appleton Institute for Behavioural Science, Central Queensland University, PO Box 42, Goodwood, South Australia 5032, Australia. E-mail addresses: [email protected] (R.W. Matthews), [email protected] (S.A. Ferguson), [email protected] (C. Sargent), [email protected] (X. Zhou), [email protected] (A. Kosmadopoulos), [email protected] (G.D. Roach).
presentation, a response, 1 s of feedback and a randomly varied 2–10 s interstimulus interval (ISI) before the next stimulus (see Fig. 1). The PVT had been shown to be both valid and reliable (Dorrian et al., 2005) and thus has been employed extensively in the study of sleep loss (Belenky et al., 2003; Van Dongen et al., 2003) and the interaction between circadian and homeostatic sleep–wake processes on performance (Dinges et al., 1994; Matthews et al., 2010; Van Dongen et al., 2003; Zhou et al., 2010). It has been presented on different devices with various durations and ISI ranges, with and without feedback (Lamond et al., 2005, 2008; Roach et al., 2006, 2016). Recently, Basner and Dinges experimented with the sensitivity of the PVT using variable length tests (Basner and Dinges, 2011, 2012; Basner et al., 2011). Due to its wide spread use, the PVT has become the modern benchmark against which other measures of performance deficits are compared. This is due to its ‘low order’ nature reflecting basic frontal cortex dysfunction (Horne, 1993). As a component of more complex operations, the value of the PVT lies
http://dx.doi.org/10.1016/j.aap.2015.10.013 0001-4575/© 2015 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Matthews, R.W., et al., Using interstimulus interval to maximise sensitivity of the Psychomotor Vigilance Test to fatigue. Accid. Anal. Prev. (2015), http://dx.doi.org/10.1016/j.aap.2015.10.013
G Model AAP-3941; No. of Pages 5 2
ARTICLE IN PRESS R.W. Matthews et al. / Accident Analysis and Prevention xxx (2015) xxx–xxx
Fig. 1. Timeline of the PVT response trial shown above the theorised response components of Inspection Time (Vickers et al., 1972), Responding Time (Kirby and McConaghy, 1986), Response Initiation Time, and Checking and Preparation Time (Kirby and Nettelbeck, 1991). While research has shown that the Reaction Time component is influenced by fatigue it is unknown whether Checking and Preparation Time is also affected by fatigue or has a fixed relationship with the ISI range.
in its simple nature. This is because without the ability to sustain attention on a given task, goal directed action becomes impossible (Dorrian et al., 2005). While a core element of the task is simple RT, the PVT is not exclusively a measure of RT like its historical counterparts. The PVT runs for a designated duration (traditionally 10 min), with the specific ISI range of 2–10 s. This allows researchers to sample across many responses from a short period of time. The relatively high stimulus rate (or signal load) is essential to avoid boredom and task fatigue effects. These distinct test characteristics are tailored towards capturing and exploring behavioural responses related to ‘vigilance’ (Basner and Dinges, 2012). The other aspect that makes PVT more than simple RT is the test outcome metrics used to study this concentrated data. RTs are valid if they are ≥100 ms and ≤500 ms. RTs less than 100 ms are termed ‘false starts’ and represent ‘errors of commission’. These premature responses are an attempt by participants to anticipate when the stimulus will appear and represent increased compensatory effort consistent with the state instability hypothesis (Dorrian et al., 2005). Reaction times greater than 500 ms are deemed ‘lapses’. These represent transient moments of increased sleepiness (Dorrian et al., 2005) which are particularly sensitive to sleep loss and are the most common outcome metrics published in papers (Basner and Dinges, 2011). A reciprocal transformation of reaction time (1/RT) is also commonly analysed as this metric limits the influence of outliers and the skewed nature of RT distributions (Dorrian et al., 2005). While the performance effects of many dimensions of the PVT have been studied, such as practice effects, test duration and time on task effects (Dorrian et al., 2005; Lamond et al., 2005, 2008; Roach et al., 2006), little has been published on the relationship between ISI and PVT performance. This is likely due to early research on the effect of absolute and relative durations of ISI on participants’ preparation and expectation states (Nickerson, 1967, 1968). These effects had been explored prior to Lisper and Kjellberg’s (1972) ART task, and led to the ISI range of 2–10 s being chosen for the precursor of the PVT. A fixed ISI range will condition
participants’ ‘state of preparedness’, influencing what is perceived to be a relatively long or short ISI (Los et al., 2001). Hence, if research was conducted on other ISI ranges the result would be a task measuring different dimensions of vigilance and fatigue, incomparable to the standard version of the PVT. The current authors have observed participants having apparent difficulty responding to stimuli following short ISIs in comparison to stimuli following long ISIs on the PVT. This effect has only been anecdotally reported in published papers. In one example, while investigating the RTs of lead-exposed workers it was reported that 2-s ISIs were followed by long RT’s for both lead exposed and nonexposed workers (Balbus et al., 1998). The effect of ISI on RT was so great that the influence of neurotoxic damage caused by lead exposure (the effect of interest) was only observed with ISIs longer than 2 s. In separating the cognitive processes involved in responding to a stimulus, cognitive psychology may have provided an explanation for the effect of ISI on RT. Vickers et al. (1972) used the visual perception theory that ‘visual information is not continuous but is sampled’ to define a stage of input processing. This stage described the time required for a stimulus to be perceived and discriminated, called Inspection Time (IT) shown in Fig. 1. Elaborating on this, Kirby and McConaghy (1986) found that immediately after a response was made, input processing was inhibited. The idea of a stimulus inhibiting the perception of a second stimulus is not new (Telford, 1931), but Kirby and McConaghy (1986) definitively showed that time is needed for a participant to process the outcome of their response and ready themselves for the next stimulus. This inhibitory period was defined by Kirby and Nettelbeck (1991) as checking and preparation time but is also referred to as a psychological or response refractory period (Nickerson, 1967), or preparation state (Los et al., 2001). It is possible that checking and preparation processes may make responding to short ISIs particularly difficult, resulting in unexplained (within-group) variance such as in the lead exposure example. If so, this would mean that responses following short ISIs on the PVT are less sensitive to the effects of fatigue. A second
Please cite this article in press as: Matthews, R.W., et al., Using interstimulus interval to maximise sensitivity of the Psychomotor Vigilance Test to fatigue. Accid. Anal. Prev. (2015), http://dx.doi.org/10.1016/j.aap.2015.10.013
G Model AAP-3941; No. of Pages 5
ARTICLE IN PRESS R.W. Matthews et al. / Accident Analysis and Prevention xxx (2015) xxx–xxx
possibility is that preparatory processes are affected by fatigue in a similar way to response processes. It has been well documented that RTs become longer and more varied under the effects of fatigue induced by sleep homeostatic and endogenous circadian influences (Belenky et al., 2003; Dinges et al., 1994; Matthews et al., 2010; Van Dongen et al., 2003; Zhou et al., 2010, 2011). However, it is unknown whether preparatory processing (such as ‘checking and preparation time’) is also affected by fatigue in a similar manner. If it is affected by fatigue, this would lead to longer RTs, as well as an inability to respond to quick successive stimuli, with a higher probability of lapsing when fatigued. Hence, the current study aims to investigate whether there is an effect of ISI length on task vigilance, and if the proposed effect is sensitive to changes in time of day, prior wake and time on task. 1. Methods 1.1. Participants The study sample consisted of 12 male participants (M 22.42 ± SD 2.31 years) with a healthy body mass index (BMI; M 21.99 ± SD 2.18 kg/m2 ). Participants were non-smoking, non-shift working, and low coffee drinkers, having no sleep disorders or recent international flights. Following an expression of interest, participants underwent a three-stage screening process consisting of a general health questionnaire, an interview and one week of activity monitoring. Ethics approval for the study was granted by the University of South Australia Human Research Ethics Committee using guidelines established by the National Health and Medical Research Council of Australia. 1.2. Apparatus and measures Vigilance throughout the study was assessed using a 10-min PVT with ISIs randomly and continuously distributed from 2 to 10 s, across the session. The PVT was presented on a handheld PVT-192 (Ambulatory Monitoring Inc., Ardsley, NY, USA). The device, measuring 21 cm × 11 cm × 6 cm, combines a four digit red LED screen, an alpha numeric information display screen and two response push buttons measuring 1.1 cm2 (Dinges and Powell, 1985). Vigilance was assessed by RTs, the standard deviation of RTs, 1/RT, lapses and false starts. Lapses were defined as RT > 500 ms and false starts were defined as responses that occurred before the stimulus was presented or with a RT < 100 ms. RTs from the first and last trials within each task were also excluded (Dinges and Powell, 1985). Time of day was determined from participants’ circadian cycle of core body temperature. Continuous measurements of core body temperature were recorded using a self-administered indwelling rectal thermistor (Steri-probe 491B, Cincinatti Sub-Zero Products, Cincinnati, OH) connected to a Mini-Mitter data logger (Bend, OR, USA) worn throughout the study by participants in a waist pack. 1.3. Protocol In order to assess the independent effects of prior wake and circadian phase, a 28 h forced desynchrony protocol was used. The protocol, described in greater detail elsewhere (Darwent et al., 2010; Matthews et al., 2011, 2012a,b), was conducted over 12 days consisting of 2 training days and a baseline day, followed by seven 28-h forced desynchrony “days” comprising 18.67 h of wake followed by 9.33 h of time in bed. On each forced desynchrony day, the 12 participants completed 7 PVTs at 2.5-h intervals, 2 h after waking, giving a total of approximately 53,000 individual response
3
Table 1 Empirical results from each mixed models ANOVA of RRT. df
F
p
7, 78,676.07 5, 78,676.16 6, 78,676.05 4, 78,676.05 30, 78,676.23 35, 78,676.06 42, 78,676.06 28, 78,676.06
966.33 322.61 169.64 256.13 27.28 1.23 1.14 1.37
<.001 <.001 <.001 <.001 <.001 .168 .247 .090
Model 2a: fatigue model without ISI variable AICa = −533,362.29 5, 82,551.51 Time of day 6, 82,551.38 Prior wake 4, 82,551.38 Time on task Prior wake × time of day 30, 82,551.58
274.35 142.82 216.57 24.12
<.001 <.001 <.001 <.001
Model 2b: fatigue model with ISI variable AICa = -539,403.67 Time of day Prior wake Time on task Prior wake × time of day ISI
5, 78,874.07 6, 78,873.95 4, 78,873.94 30, 78,874.13 7, 78,873.97
324.74 170.47 257.13 27.32 981.29
<.001 <.001 <.001 <.001 <.001
Model 3a: 2–10 ISI model Time of day Prior wake Prior wake × time of day
Cohen’s d 0.37 0.54 1.07
5, 667.27 6, 667.25 30, 667.28
22.77 11.59 2.16
<.001 <.001 <.001
Model 3b: 5–10 ISI model Time of day Prior wake Prior wake × time of day
Cohen’s d 0.44 0.75 2.39
5, 705.60 6, 705.58 30, 705.61
32.48 19.15 2.76
<.001 <.001 <.001
Model 1: full model ISI Time of day Prior wake Time on task Prior wake × time of day ISI × time of day ISI × prior wake ISI × time on task
a AIC: Akaike information criterion value where a smaller number (more negative) denotes a better model fit.
trials. Sound, temperature (22 ± 1 ◦ C) and light (10–15 lux) were controlled, and participants were temporally and social isolated throughout the experimental protocol. 1.4. Statistical analysis The RTs, lapses, and false starts from response trials within the 10-min PVTs were assessed separately and assigned to a 1-sec ISI group (from 2 to 10 s), a 2.5 h prior wake group (from 2 to 17 h), one of six 60◦ phase bins of the circadian rhythm of core body temperature (as a measure of time of day) and a 2-min time on task group (from 1 to 10 min). The 1-s ISI groups were needed as the PVT randomly distributes ISIs continuously, rather than at discrete 1-s intervals. A mixed models analysis of variance (ANOVA) was used to analyse the reciprocal of RT (1/RT) with the four independent variables (ISI, time of day, prior wake and time on task) entered as non-continuous fixed effects, as well as a prior wake × time of day interaction (Model 1: full model, in Table 1). Furthermore, ISI interaction terms with each variable were also included in this model to explore the relationship between ISI and the effects of the contributors of fatigue. The variable of ‘Participant ID’ was entered as a random effect into the analysis to account for within-subject variability. To explore the effect of ISI further, a mixed models ANOVA (Model 2a, in Table 1) was run containing only the fatigue variables: time of day, prior wake, time on task, and the prior wake × time of day interaction. This model was then rerun with the addition of the ISI variable (Model 2b). The fit of each model with the data was compared using Akaike’s Information Criterion (AIC), where a smaller score depicts a better model fit. These analyses focus on individual response trials whereas conventionally the mean RT (or 1/RT) from each 10 min test is
Please cite this article in press as: Matthews, R.W., et al., Using interstimulus interval to maximise sensitivity of the Psychomotor Vigilance Test to fatigue. Accid. Anal. Prev. (2015), http://dx.doi.org/10.1016/j.aap.2015.10.013
G Model AAP-3941; No. of Pages 5
ARTICLE IN PRESS R.W. Matthews et al. / Accident Analysis and Prevention xxx (2015) xxx–xxx
4
calculated and groups of means are analysed (Dorrian et al., 2005). To investigate the effect of ISI in analyses of this form, two additional mixed models ANOVAs were compared. Model 3a contains mean reciprocal reaction times from each PVT with the standard 2–10 s ISI range, while in Model 3b, response trials of short ISI groups (2–5 s) were removed prior to calculating the mean reciprocal reaction times. These two models were compared by the effects sizes of prior wake, time of day and the prior wake × time of day interaction. Time on task was not included as these models contained mean RT from the whole 10 min PVT rather than individual response trials. The frequency of lapses and false starts within each ISI group were compared with a chi-squared test. The frequencies of false starts were progressively under-represented as ISI increased. This was corrected by dividing the false start frequency by the probability that a stimulus had not been presented. This was the cumulative number of observed stimuli over the total number of stimuli, inversed, multiplied by 100.
2. Results The effect of ISI length on RT and 1/RT is represented in Panel A and C of Fig. 2. A mixed models ANOVA indicated that there was a significant effect of ISI on 1/RT (see Table 1). Pairwise comparisons showed that the three shortest ISI groups of 2–3, 3–4 and 4–5 s were significantly different from each other (p < .001) and contained significantly slower RTs and 1/RT than all other groups (Panel A and C, Fig. 2). The 5–6 s ISI group was significantly different from the three shorter groups and the longest ISI group (9–10 s). The other long ISI groups were not significantly different from each other but were different from the three shortest ISI groups. Time of day, prior wake and time on task had a significant main effect on 1/RT (Model 1, Table 1). However, the main effect of ISI on performance was greater than these other influences, with a change in RT of 50 ms between short and long ISIs. Mean RT at a long duration of prior wake (17 h), the poorest time of day (circadian nadir), and long time on task (10 min) was 26 ms, 24 ms, and 18 ms slower, respectively, than best performance at a short duration of prior wake (2 h), the best time of day (circadian acrophase), and short time on task (1 min). Correspondingly, the Cohen’s d effect size of ISI on 1/RT was large (.96) while the effect sizes of prior wake, time of day, and time on task were small to moderate (.37, .31 and .30 respectively). The main effect of ISI on 1/RT was also independent of the effects of time of day, prior wake and time on task. This is shown by the non-significant interactions between ISI and the time of day, prior wake and time on task factors in Model 1 (Table 1). In addition to producing longer responses, short ISIs were also associated with more varied response times as depicted in Panel B, Fig. 2. Short ISIs consisted of significantly more lapses than longer ISIs (2(7) = 250.48, p < .001) (Panel D, Fig. 2). Finally, there was a significant positive association between errors of commission and ISI (2(8) = 393.55, p < .001) suggesting that the probability of an false start response increased as the ISI length increased (Panel E, Fig. 2). In accordance with these findings, when investigating the effects of fatigue (by time of day, prior wake, time on task and prior wake × time of day interaction) on 1/RT, a model that captured the variance associated with ISI (Model 2b) provided a better model fit than one without an ISI factor (Model 2a), shown by a smaller Akaike information criterion (Table 1). Furthermore, when short ISIs were removed from PVT mean 1/RT scores, distributions containing long ISIs of 5–10 s (Model 3b) had larger effect sizes for prior wake, time of day and interaction, than means of distributions containing all ISIs (Model 3a).
Fig. 2. Descriptive statistics of the effect of interstimulus interval group on mean reaction time for response trials (Panel A), the standard deviation of reaction time (Panel B), and mean reciprocal reaction time (1/RT) (Panel C), after lapse and false start responses were removed, shown in milliseconds with standard error of the means. Panel D shows the frequency of lapses as a percentage of each interstimulus interval group. The frequencies of false starts are shown as percentages, separated by the elapses time since the previous stimulus in 1-s groups (Panel E).
Please cite this article in press as: Matthews, R.W., et al., Using interstimulus interval to maximise sensitivity of the Psychomotor Vigilance Test to fatigue. Accid. Anal. Prev. (2015), http://dx.doi.org/10.1016/j.aap.2015.10.013
G Model AAP-3941; No. of Pages 5
ARTICLE IN PRESS R.W. Matthews et al. / Accident Analysis and Prevention xxx (2015) xxx–xxx
3. Discussion Recently, rigorous work has been conducted to investigate the sensitivity of PVT outcome measures (Basner and Dinges, 2011) to partial and total sleep deprivation. The current study considered the main outcome measures across individual response trials as a function of ISI. Evidence was found that responses that followed short ISIs (2–5 s) on the PVT were accompanied by slower and more varied reaction times and more lapses, independent to the effects of fatigue. This effect of ISI on RT was at least twice as large as that of 17 h of prior wake and the influence of the circadian nadir—masking the variance of interest. Including an ISI factor or covariate can capture this variance, improving the model fit and the sensitivity of the test to the effects of the homeostatic and circadian influences. The finding that short ISIs are associated with slower RTs is consistent with the theory that after a response is made, time is needed to be ready for the next stimulus. It is likely that after short ISIs, checking and preparation processes may be inhibited leading to longer RTs and more lapses (errors of omission) (Kirby and McConaghy, 1986). While processing inhibition may prevent responses (correct or error), false starts increased as a function of the time participants waited for the stimulus to appear. This could also be explained by participant expectation and preparation state (Los et al., 2001). Participants were aware that the longer they waited the more likely it was that a stimulus would appear. Thus, the longer they maintained a high level of motor readiness, the more likely they made anticipatory responses (Nickerson, 1967, 1968). These longer ISIs may represent a response period where the participant is most primed and ready to respond. Thus, any deficit in vigilance due to sleep–wake processes is more likely to be captured by these ISIs. This study may have identified the upper limit of stimulus frequency, that is, the shortest ISIs that participants can maximally respond to before a sharp speed-frequency tradeoff. Alternatively, it is possible that this effect of ISI is proportionate to the absolute ISI range. For example, were a longer ISI range of 5–15 s to be implemented, the 5 s ISIs may produce slow responses because they are short relative to the conditioned ISI range. The 2–10 s ISI range is an important component of the PVT as it dictates the signal load of the test. It is not suggested that the ISI range of the PVT should be altered; rather, the findings suggest that if the short ISIs were removed from a response dataset the mean RT (or 1/RT) would contain less variability and hence would be more sensitive to the effects of fatigue. Identifying which response domains are more sensitive to the homeostatic and circadian influences may prove useful for other recent PVT developments. For example, Basner and Dinges (2012) have developed an ‘adaptive-duration’ PVT that uses an algorithm to predict whether a participant is a high, medium or low performer based on an initial set of responses. The task stops once sufficient responses have been sampled to classify the participant. If ISIs from 5 to 10 s represents an ‘ideal’ response range, sampling from this range may improve the algorithms’ predicative ability, leading to better predictions with less data. Acknowledgment This study was financially supported by the Australian Research Council. References Balbus, J.M., Stewart, W., Bolla, K.I., Schwartz, B.S., 1998. Simple visual reaction time in organolead manufacturing workers: influence of the interstimulus interval. Arch. Environ. Health 53 (4), 264–270. Basner, M., Dinges, D.F., 2011. Maximizing sensitivity of the psychomotor vigilance test (PVT) to sleep loss. Sleep 34 (5), 581.
5
Basner, M., Dinges, D.F., 2012. An adaptive-duration version of the PVT accurately tracks changes in psychomotor vigilance induced by sleep restriction. Sleep 35 (2), 193. Basner, M., Mollicone, D., Dinges, D.F., 2011. Validity and sensitivity of a brief psychomotor vigilance test (PVT-B) to total and partial sleep deprivation. Acta Astronaut. 69 (11), 949–959. Belenky, G., Wesesten, N.J., Thorne, D.R., Thomas, M.L., Sing, H., Redmond, D.P., Balkin, T.J., 2003. Patterns of performance degradation and restoration during sleep restriction and subsequent recovery: a sleep dose–response study. J. Sleep Res. 12, 1–12. Darwent, D., Ferguson, S.A., Sargent, C., Paech, G.M., Williams, L., Zhou, X., Dawson, D., 2010. Contribution of core body temperature, prior wake time and sleep stages to cognitive throughput performance during forced desynchrony. Chronobiol. Int. 27 (5), 898–910. Dinges, D., Pack, F., Gillen, K., Powell, J., Kribbs, N., Ott, G., Pack, A., 1994. Cumulative effects of acute partial sleep deprivation on PVT lapses and SSS ratings. J. Sleep Res. 23, 408. Dinges, D.F., Powell, J.W., 1985. Microcomputer analyses of performance on a portable, simple visual RT task during sustained operations. Behav. Res. Methods Instrum. Comput. 17 (6), 652–655. Dorrian, J., Rogers, N.L., Dinges, D.F., 2005. Psychomotor vigilance performance: neurocognitive assay sensitive to sleep loss. In: Kushida, C.A. (Ed.), Sleep Deprivation: Clinical Issues, Pharmacology and Sleep Loss Effects, vol. 39–70. Marcel Dekker Inc., New York. Glenville, M., Broughton, R., Wing, A.M., Wilkinson, R.T., 1978. Effects of sleep deprivation on short duration performance measures compared to the Wilkinson auditory vigilance task. Sleep 1, 169–176. Horne, J.A., 1993. Human sleep, sleep loss and behaviour. Implications for the prefrontal cortex and psychiatric disorder. Br. J. Psychiatry 162, 413–419. Kirby, N., McConaghy, J., 1986. Responding time: a response equivalent to inspection time. Pers. Individ. Diff. 7 (5), 687–694. Kirby, N., Nettelbeck, T., 1991. Speed of information processing and age. Pers. Individ. Diff. 12 (2), 183–188. Lamond, N., Dawson, D., Roach, G.D., 2005. Fatigue assessment in the field: validation of a hand-held electronic psychomotor vigilance task. Aviat. Space Environ. Med. 76 (5), 486–489. Lamond, N., Jay, S.M., Dorrian, J., Ferguson, S.A., Roach, G.D., Dawson, D., 2008. The sensitivity of a palm-based psychomotor vigilance task to severe sleep loss. Behav. Res. Methods 40 (1), 347–352. Lisper, H.-O., Kjellberg, A., 1972. Effects of 24-hour sleep deprivation on rate of decrement in a 10-minute auditory reaction time task. J. Exp. Psychol. 96, 287–290. Los, S.A., Knol, D.L., Boers, R.M., 2001. The foreperiod effect revisited: conditioning as a basis for nonspecific preparation. Acta Psychol. (Amst.) 106 (1), 121–145. Matthews, R., Ferguson, S., Zhou, X., Roach, G., 2011. A forced desynchrony correction: accounting for longitudinal confounders. In: Little Clock, Big Clock: Molecular to Physiological Clocks. Australasian Chronobiology Society, Melbourne, Australia., pp. 43–47. Matthews, R., Ferguson, S.A., Zhou, X., Darwent, D., Sargent, C., Kennaway, D.J., Roach, G.D., 2010. The interactive influence of the internal body clock and prior wake on task vigilance. Ergon. IJE&HF 32 (1), 37–48. Matthews, R.W., Ferguson, S.A., Zhou, X., Kosmadopoulos, A., Kennaway, D.J., Roach, G.D., 2012a. Simulated driving under the influence of extended wake, time of day and sleep restriction. Accid. Anal. Prev. 45 (Suppl.), 55–61. Matthews, R.W., Ferguson, S.A., Zhou, X., Sargent, C., Darwent, D., Kennaway, D.J., Roach, G.D., 2012b. Time-of-day mediates the influences of extended wake and sleep restriction on simulated driving. Chronobiol. Int. 29 (5), 572–579. Nickerson, R.S., 1967. Psychological refractory phase and the functional significance of signals. J. Exp. Psychol. 73 (2), 303. Nickerson, R.S., 1968. Response time to the second of two signals following varied vs constant intersignal intervals. Percept. Psychophys. 4 (2), 78–80. Roach, G.D., Dawson, D., Lamond, N., 2006. Can a shorter psychomotor vigilance task be used as a reasonable substitute for the ten-minute psychomotor vigilance task? Chronobiol. Int. 23 (6), 1379–1387. Roach, G.D., Lamond, N., Dawson, D., 2016. Feedback has a positive effect on cognitive function during total sleep deprivation if there is sufficient time for it to be effectively processed. Appl. Ergon. 52, 285–290. Telford, C.W., 1931. The refractory phase of voluntary and associative responses. J. Exp. Psychol. 14 (1), 1. Van Dongen, H.P.A., Maislin, M.A., Mullington, J.M., Dinges, D.F., 2003. The cumulative cost of additional wakefulness: dose–response effects on neurobehavioural functions and sleep physiology from chronic sleep restriction and total sleep deprivation. Sleep 26 (2), 117–126. Vickers, D., Nettelbeck, T., Willson, R., 1972. Perceptual indices of performance: the measurement of ‘inspection time’ and ‘noise’ in the visual system. Perception 1 (3), 263–295. Wilkinson, R.T., Houghton, D., 1982. Field test of arousal: a portable reaction timer with data storage. Hum. Fact. 24 (4), 487–493. Zhou, X., Ferguson, S.A., Matthews, R., Sargent, C., Darwent, D., Kennaway, D.J., Roach, G.D., 2011. Dynamics of neurobehavioral performance variability under forced desynchrony: evidence of state instability. Sleep 34 (1), 57–63. Zhou, X., Ferguson, S.A., Matthews, R.W., Sargent, C., Darwent, D., Kennaway, D.J., Roach, G.D., 2010. Interindividual differences in neurobehavioural performance in response to increasing homeostatic sleep pressure. Chronobiol. Int. 27 (5), 922–933.
Please cite this article in press as: Matthews, R.W., et al., Using interstimulus interval to maximise sensitivity of the Psychomotor Vigilance Test to fatigue. Accid. Anal. Prev. (2015), http://dx.doi.org/10.1016/j.aap.2015.10.013
ARTICLE IN PRESS
AAP-3941; No. of Pages 5
Accident Analysis and Prevention xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
Using interstimulus interval to maximise sensitivity of the Psychomotor Vigilance Test to fatigue Raymond W. Matthews a,b,∗ , Sally A. Ferguson a , Charli Sargent a , Xuan Zhou a , Anastasi Kosmadopoulos a , Gregory D. Roach a a b
Appleton Institute for Behavioural Science, Central Queensland University, PO Box 42, Goodwood, South Australia 5032, Australia Centre for Sleep Research, University of South Australia, GPO Box 2471, Adelaide, South Australia 5001, Australia
a r t i c l e
i n f o
Article history: Received 28 June 2015 Received in revised form 4 October 2015 Accepted 13 October 2015 Available online xxx Keywords: Psychomotor Vigilance Task Fatigue Sleep Neurocognitive performance Preparedness Forced desynchrony
a b s t r a c t There is some evidence that short interstimulus intervals (ISIs) on the Psychomotor Vigilance Test (PVT) are associated with longer and more varied reaction times (RTs). Preparation processes may impede RT following short ISIs, resulting in additional unexplained variance. The aims of this study were to investigate whether there is an effect of ISI on RT and errors within the PVT, and whether such an effect changes with three elements of fatigue: time of day, prior wake and time on task. Twelve male participants completed 49 PVTs across 7× 28 h periods of forced desynchrony. For analysis, RTs, reciprocal reaction times (1/RT), false starts and lapse responses within each 10 min session were assigned to a 1-s ISI group, a 2-min time of task group, a 2.5-h PW level and a 60◦ phase of the circadian rhythm of core body temperature (as a measure of time of day). Responses following short ISIs (2–5 s) were significantly slower and more varied than responses following longer ISIs (5–10 s). The likelihood of a lapse was also higher for short ISIs, while the probability of a false start increased as a function of ISI. These effects were independent of the influences of time of day, prior wake and time on task. Hence, mixed model ANOVAs comprising only long ISIs (5–10 s) contained stronger effect sizes for fatigue than a model of all ISIs (2–10 s). Including an ISI variable in a model improved the model fit and explained more variance associated with fatigue. Short ISIs resulted in long RTs both in the presence and absence of fatigue, possibly due to preparation processes or ISI conditioning. Hence, omitting short ISI trials from RT means or including an ISI variable in analysis can reduce unwanted variance in PVT data, improving the sensitivity of the PVT to fatigue. © 2015 Elsevier Ltd. All rights reserved.
Over the last three decades the Psychomotor Vigilance Task (PVT) has become the ubiquitous behavioural assay to measure the effects of fatigue and sleepiness (Dorrian et al., 2005). The PVT is a sustained attention, stimulus-response task developed by Dinges and Powell (1985). It was developed as an evolution of Wilkinson’s earlier simple visual reaction time (VRT) task (Glenville et al., 1978; Wilkinson and Houghton, 1982), which was itself based on the previous auditory reaction time (ART) task by Lisper and Kjellberg (1972). The standard form of the PVT is a 10-min test containing approximately 90 response trials. Each trial consists of a stimulus
∗ Corresponding author at: Appleton Institute for Behavioural Science, Central Queensland University, PO Box 42, Goodwood, South Australia 5032, Australia. E-mail addresses: [email protected] (R.W. Matthews), [email protected] (S.A. Ferguson), [email protected] (C. Sargent), [email protected] (X. Zhou), [email protected] (A. Kosmadopoulos), [email protected] (G.D. Roach).
presentation, a response, 1 s of feedback and a randomly varied 2–10 s interstimulus interval (ISI) before the next stimulus (see Fig. 1). The PVT had been shown to be both valid and reliable (Dorrian et al., 2005) and thus has been employed extensively in the study of sleep loss (Belenky et al., 2003; Van Dongen et al., 2003) and the interaction between circadian and homeostatic sleep–wake processes on performance (Dinges et al., 1994; Matthews et al., 2010; Van Dongen et al., 2003; Zhou et al., 2010). It has been presented on different devices with various durations and ISI ranges, with and without feedback (Lamond et al., 2005, 2008; Roach et al., 2006, 2016). Recently, Basner and Dinges experimented with the sensitivity of the PVT using variable length tests (Basner and Dinges, 2011, 2012; Basner et al., 2011). Due to its wide spread use, the PVT has become the modern benchmark against which other measures of performance deficits are compared. This is due to its ‘low order’ nature reflecting basic frontal cortex dysfunction (Horne, 1993). As a component of more complex operations, the value of the PVT lies
http://dx.doi.org/10.1016/j.aap.2015.10.013 0001-4575/© 2015 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Matthews, R.W., et al., Using interstimulus interval to maximise sensitivity of the Psychomotor Vigilance Test to fatigue. Accid. Anal. Prev. (2015), http://dx.doi.org/10.1016/j.aap.2015.10.013
G Model AAP-3941; No. of Pages 5 2
ARTICLE IN PRESS R.W. Matthews et al. / Accident Analysis and Prevention xxx (2015) xxx–xxx
Fig. 1. Timeline of the PVT response trial shown above the theorised response components of Inspection Time (Vickers et al., 1972), Responding Time (Kirby and McConaghy, 1986), Response Initiation Time, and Checking and Preparation Time (Kirby and Nettelbeck, 1991). While research has shown that the Reaction Time component is influenced by fatigue it is unknown whether Checking and Preparation Time is also affected by fatigue or has a fixed relationship with the ISI range.
in its simple nature. This is because without the ability to sustain attention on a given task, goal directed action becomes impossible (Dorrian et al., 2005). While a core element of the task is simple RT, the PVT is not exclusively a measure of RT like its historical counterparts. The PVT runs for a designated duration (traditionally 10 min), with the specific ISI range of 2–10 s. This allows researchers to sample across many responses from a short period of time. The relatively high stimulus rate (or signal load) is essential to avoid boredom and task fatigue effects. These distinct test characteristics are tailored towards capturing and exploring behavioural responses related to ‘vigilance’ (Basner and Dinges, 2012). The other aspect that makes PVT more than simple RT is the test outcome metrics used to study this concentrated data. RTs are valid if they are ≥100 ms and ≤500 ms. RTs less than 100 ms are termed ‘false starts’ and represent ‘errors of commission’. These premature responses are an attempt by participants to anticipate when the stimulus will appear and represent increased compensatory effort consistent with the state instability hypothesis (Dorrian et al., 2005). Reaction times greater than 500 ms are deemed ‘lapses’. These represent transient moments of increased sleepiness (Dorrian et al., 2005) which are particularly sensitive to sleep loss and are the most common outcome metrics published in papers (Basner and Dinges, 2011). A reciprocal transformation of reaction time (1/RT) is also commonly analysed as this metric limits the influence of outliers and the skewed nature of RT distributions (Dorrian et al., 2005). While the performance effects of many dimensions of the PVT have been studied, such as practice effects, test duration and time on task effects (Dorrian et al., 2005; Lamond et al., 2005, 2008; Roach et al., 2006), little has been published on the relationship between ISI and PVT performance. This is likely due to early research on the effect of absolute and relative durations of ISI on participants’ preparation and expectation states (Nickerson, 1967, 1968). These effects had been explored prior to Lisper and Kjellberg’s (1972) ART task, and led to the ISI range of 2–10 s being chosen for the precursor of the PVT. A fixed ISI range will condition
participants’ ‘state of preparedness’, influencing what is perceived to be a relatively long or short ISI (Los et al., 2001). Hence, if research was conducted on other ISI ranges the result would be a task measuring different dimensions of vigilance and fatigue, incomparable to the standard version of the PVT. The current authors have observed participants having apparent difficulty responding to stimuli following short ISIs in comparison to stimuli following long ISIs on the PVT. This effect has only been anecdotally reported in published papers. In one example, while investigating the RTs of lead-exposed workers it was reported that 2-s ISIs were followed by long RT’s for both lead exposed and nonexposed workers (Balbus et al., 1998). The effect of ISI on RT was so great that the influence of neurotoxic damage caused by lead exposure (the effect of interest) was only observed with ISIs longer than 2 s. In separating the cognitive processes involved in responding to a stimulus, cognitive psychology may have provided an explanation for the effect of ISI on RT. Vickers et al. (1972) used the visual perception theory that ‘visual information is not continuous but is sampled’ to define a stage of input processing. This stage described the time required for a stimulus to be perceived and discriminated, called Inspection Time (IT) shown in Fig. 1. Elaborating on this, Kirby and McConaghy (1986) found that immediately after a response was made, input processing was inhibited. The idea of a stimulus inhibiting the perception of a second stimulus is not new (Telford, 1931), but Kirby and McConaghy (1986) definitively showed that time is needed for a participant to process the outcome of their response and ready themselves for the next stimulus. This inhibitory period was defined by Kirby and Nettelbeck (1991) as checking and preparation time but is also referred to as a psychological or response refractory period (Nickerson, 1967), or preparation state (Los et al., 2001). It is possible that checking and preparation processes may make responding to short ISIs particularly difficult, resulting in unexplained (within-group) variance such as in the lead exposure example. If so, this would mean that responses following short ISIs on the PVT are less sensitive to the effects of fatigue. A second
Please cite this article in press as: Matthews, R.W., et al., Using interstimulus interval to maximise sensitivity of the Psychomotor Vigilance Test to fatigue. Accid. Anal. Prev. (2015), http://dx.doi.org/10.1016/j.aap.2015.10.013
G Model AAP-3941; No. of Pages 5
ARTICLE IN PRESS R.W. Matthews et al. / Accident Analysis and Prevention xxx (2015) xxx–xxx
possibility is that preparatory processes are affected by fatigue in a similar way to response processes. It has been well documented that RTs become longer and more varied under the effects of fatigue induced by sleep homeostatic and endogenous circadian influences (Belenky et al., 2003; Dinges et al., 1994; Matthews et al., 2010; Van Dongen et al., 2003; Zhou et al., 2010, 2011). However, it is unknown whether preparatory processing (such as ‘checking and preparation time’) is also affected by fatigue in a similar manner. If it is affected by fatigue, this would lead to longer RTs, as well as an inability to respond to quick successive stimuli, with a higher probability of lapsing when fatigued. Hence, the current study aims to investigate whether there is an effect of ISI length on task vigilance, and if the proposed effect is sensitive to changes in time of day, prior wake and time on task. 1. Methods 1.1. Participants The study sample consisted of 12 male participants (M 22.42 ± SD 2.31 years) with a healthy body mass index (BMI; M 21.99 ± SD 2.18 kg/m2 ). Participants were non-smoking, non-shift working, and low coffee drinkers, having no sleep disorders or recent international flights. Following an expression of interest, participants underwent a three-stage screening process consisting of a general health questionnaire, an interview and one week of activity monitoring. Ethics approval for the study was granted by the University of South Australia Human Research Ethics Committee using guidelines established by the National Health and Medical Research Council of Australia. 1.2. Apparatus and measures Vigilance throughout the study was assessed using a 10-min PVT with ISIs randomly and continuously distributed from 2 to 10 s, across the session. The PVT was presented on a handheld PVT-192 (Ambulatory Monitoring Inc., Ardsley, NY, USA). The device, measuring 21 cm × 11 cm × 6 cm, combines a four digit red LED screen, an alpha numeric information display screen and two response push buttons measuring 1.1 cm2 (Dinges and Powell, 1985). Vigilance was assessed by RTs, the standard deviation of RTs, 1/RT, lapses and false starts. Lapses were defined as RT > 500 ms and false starts were defined as responses that occurred before the stimulus was presented or with a RT < 100 ms. RTs from the first and last trials within each task were also excluded (Dinges and Powell, 1985). Time of day was determined from participants’ circadian cycle of core body temperature. Continuous measurements of core body temperature were recorded using a self-administered indwelling rectal thermistor (Steri-probe 491B, Cincinatti Sub-Zero Products, Cincinnati, OH) connected to a Mini-Mitter data logger (Bend, OR, USA) worn throughout the study by participants in a waist pack. 1.3. Protocol In order to assess the independent effects of prior wake and circadian phase, a 28 h forced desynchrony protocol was used. The protocol, described in greater detail elsewhere (Darwent et al., 2010; Matthews et al., 2011, 2012a,b), was conducted over 12 days consisting of 2 training days and a baseline day, followed by seven 28-h forced desynchrony “days” comprising 18.67 h of wake followed by 9.33 h of time in bed. On each forced desynchrony day, the 12 participants completed 7 PVTs at 2.5-h intervals, 2 h after waking, giving a total of approximately 53,000 individual response
3
Table 1 Empirical results from each mixed models ANOVA of RRT. df
F
p
7, 78,676.07 5, 78,676.16 6, 78,676.05 4, 78,676.05 30, 78,676.23 35, 78,676.06 42, 78,676.06 28, 78,676.06
966.33 322.61 169.64 256.13 27.28 1.23 1.14 1.37
<.001 <.001 <.001 <.001 <.001 .168 .247 .090
Model 2a: fatigue model without ISI variable AICa = −533,362.29 5, 82,551.51 Time of day 6, 82,551.38 Prior wake 4, 82,551.38 Time on task Prior wake × time of day 30, 82,551.58
274.35 142.82 216.57 24.12
<.001 <.001 <.001 <.001
Model 2b: fatigue model with ISI variable AICa = -539,403.67 Time of day Prior wake Time on task Prior wake × time of day ISI
5, 78,874.07 6, 78,873.95 4, 78,873.94 30, 78,874.13 7, 78,873.97
324.74 170.47 257.13 27.32 981.29
<.001 <.001 <.001 <.001 <.001
Model 3a: 2–10 ISI model Time of day Prior wake Prior wake × time of day
Cohen’s d 0.37 0.54 1.07
5, 667.27 6, 667.25 30, 667.28
22.77 11.59 2.16
<.001 <.001 <.001
Model 3b: 5–10 ISI model Time of day Prior wake Prior wake × time of day
Cohen’s d 0.44 0.75 2.39
5, 705.60 6, 705.58 30, 705.61
32.48 19.15 2.76
<.001 <.001 <.001
Model 1: full model ISI Time of day Prior wake Time on task Prior wake × time of day ISI × time of day ISI × prior wake ISI × time on task
a AIC: Akaike information criterion value where a smaller number (more negative) denotes a better model fit.
trials. Sound, temperature (22 ± 1 ◦ C) and light (10–15 lux) were controlled, and participants were temporally and social isolated throughout the experimental protocol. 1.4. Statistical analysis The RTs, lapses, and false starts from response trials within the 10-min PVTs were assessed separately and assigned to a 1-sec ISI group (from 2 to 10 s), a 2.5 h prior wake group (from 2 to 17 h), one of six 60◦ phase bins of the circadian rhythm of core body temperature (as a measure of time of day) and a 2-min time on task group (from 1 to 10 min). The 1-s ISI groups were needed as the PVT randomly distributes ISIs continuously, rather than at discrete 1-s intervals. A mixed models analysis of variance (ANOVA) was used to analyse the reciprocal of RT (1/RT) with the four independent variables (ISI, time of day, prior wake and time on task) entered as non-continuous fixed effects, as well as a prior wake × time of day interaction (Model 1: full model, in Table 1). Furthermore, ISI interaction terms with each variable were also included in this model to explore the relationship between ISI and the effects of the contributors of fatigue. The variable of ‘Participant ID’ was entered as a random effect into the analysis to account for within-subject variability. To explore the effect of ISI further, a mixed models ANOVA (Model 2a, in Table 1) was run containing only the fatigue variables: time of day, prior wake, time on task, and the prior wake × time of day interaction. This model was then rerun with the addition of the ISI variable (Model 2b). The fit of each model with the data was compared using Akaike’s Information Criterion (AIC), where a smaller score depicts a better model fit. These analyses focus on individual response trials whereas conventionally the mean RT (or 1/RT) from each 10 min test is
Please cite this article in press as: Matthews, R.W., et al., Using interstimulus interval to maximise sensitivity of the Psychomotor Vigilance Test to fatigue. Accid. Anal. Prev. (2015), http://dx.doi.org/10.1016/j.aap.2015.10.013
G Model AAP-3941; No. of Pages 5
ARTICLE IN PRESS R.W. Matthews et al. / Accident Analysis and Prevention xxx (2015) xxx–xxx
4
calculated and groups of means are analysed (Dorrian et al., 2005). To investigate the effect of ISI in analyses of this form, two additional mixed models ANOVAs were compared. Model 3a contains mean reciprocal reaction times from each PVT with the standard 2–10 s ISI range, while in Model 3b, response trials of short ISI groups (2–5 s) were removed prior to calculating the mean reciprocal reaction times. These two models were compared by the effects sizes of prior wake, time of day and the prior wake × time of day interaction. Time on task was not included as these models contained mean RT from the whole 10 min PVT rather than individual response trials. The frequency of lapses and false starts within each ISI group were compared with a chi-squared test. The frequencies of false starts were progressively under-represented as ISI increased. This was corrected by dividing the false start frequency by the probability that a stimulus had not been presented. This was the cumulative number of observed stimuli over the total number of stimuli, inversed, multiplied by 100.
2. Results The effect of ISI length on RT and 1/RT is represented in Panel A and C of Fig. 2. A mixed models ANOVA indicated that there was a significant effect of ISI on 1/RT (see Table 1). Pairwise comparisons showed that the three shortest ISI groups of 2–3, 3–4 and 4–5 s were significantly different from each other (p < .001) and contained significantly slower RTs and 1/RT than all other groups (Panel A and C, Fig. 2). The 5–6 s ISI group was significantly different from the three shorter groups and the longest ISI group (9–10 s). The other long ISI groups were not significantly different from each other but were different from the three shortest ISI groups. Time of day, prior wake and time on task had a significant main effect on 1/RT (Model 1, Table 1). However, the main effect of ISI on performance was greater than these other influences, with a change in RT of 50 ms between short and long ISIs. Mean RT at a long duration of prior wake (17 h), the poorest time of day (circadian nadir), and long time on task (10 min) was 26 ms, 24 ms, and 18 ms slower, respectively, than best performance at a short duration of prior wake (2 h), the best time of day (circadian acrophase), and short time on task (1 min). Correspondingly, the Cohen’s d effect size of ISI on 1/RT was large (.96) while the effect sizes of prior wake, time of day, and time on task were small to moderate (.37, .31 and .30 respectively). The main effect of ISI on 1/RT was also independent of the effects of time of day, prior wake and time on task. This is shown by the non-significant interactions between ISI and the time of day, prior wake and time on task factors in Model 1 (Table 1). In addition to producing longer responses, short ISIs were also associated with more varied response times as depicted in Panel B, Fig. 2. Short ISIs consisted of significantly more lapses than longer ISIs (2(7) = 250.48, p < .001) (Panel D, Fig. 2). Finally, there was a significant positive association between errors of commission and ISI (2(8) = 393.55, p < .001) suggesting that the probability of an false start response increased as the ISI length increased (Panel E, Fig. 2). In accordance with these findings, when investigating the effects of fatigue (by time of day, prior wake, time on task and prior wake × time of day interaction) on 1/RT, a model that captured the variance associated with ISI (Model 2b) provided a better model fit than one without an ISI factor (Model 2a), shown by a smaller Akaike information criterion (Table 1). Furthermore, when short ISIs were removed from PVT mean 1/RT scores, distributions containing long ISIs of 5–10 s (Model 3b) had larger effect sizes for prior wake, time of day and interaction, than means of distributions containing all ISIs (Model 3a).
Fig. 2. Descriptive statistics of the effect of interstimulus interval group on mean reaction time for response trials (Panel A), the standard deviation of reaction time (Panel B), and mean reciprocal reaction time (1/RT) (Panel C), after lapse and false start responses were removed, shown in milliseconds with standard error of the means. Panel D shows the frequency of lapses as a percentage of each interstimulus interval group. The frequencies of false starts are shown as percentages, separated by the elapses time since the previous stimulus in 1-s groups (Panel E).
Please cite this article in press as: Matthews, R.W., et al., Using interstimulus interval to maximise sensitivity of the Psychomotor Vigilance Test to fatigue. Accid. Anal. Prev. (2015), http://dx.doi.org/10.1016/j.aap.2015.10.013
G Model AAP-3941; No. of Pages 5
ARTICLE IN PRESS R.W. Matthews et al. / Accident Analysis and Prevention xxx (2015) xxx–xxx
3. Discussion Recently, rigorous work has been conducted to investigate the sensitivity of PVT outcome measures (Basner and Dinges, 2011) to partial and total sleep deprivation. The current study considered the main outcome measures across individual response trials as a function of ISI. Evidence was found that responses that followed short ISIs (2–5 s) on the PVT were accompanied by slower and more varied reaction times and more lapses, independent to the effects of fatigue. This effect of ISI on RT was at least twice as large as that of 17 h of prior wake and the influence of the circadian nadir—masking the variance of interest. Including an ISI factor or covariate can capture this variance, improving the model fit and the sensitivity of the test to the effects of the homeostatic and circadian influences. The finding that short ISIs are associated with slower RTs is consistent with the theory that after a response is made, time is needed to be ready for the next stimulus. It is likely that after short ISIs, checking and preparation processes may be inhibited leading to longer RTs and more lapses (errors of omission) (Kirby and McConaghy, 1986). While processing inhibition may prevent responses (correct or error), false starts increased as a function of the time participants waited for the stimulus to appear. This could also be explained by participant expectation and preparation state (Los et al., 2001). Participants were aware that the longer they waited the more likely it was that a stimulus would appear. Thus, the longer they maintained a high level of motor readiness, the more likely they made anticipatory responses (Nickerson, 1967, 1968). These longer ISIs may represent a response period where the participant is most primed and ready to respond. Thus, any deficit in vigilance due to sleep–wake processes is more likely to be captured by these ISIs. This study may have identified the upper limit of stimulus frequency, that is, the shortest ISIs that participants can maximally respond to before a sharp speed-frequency tradeoff. Alternatively, it is possible that this effect of ISI is proportionate to the absolute ISI range. For example, were a longer ISI range of 5–15 s to be implemented, the 5 s ISIs may produce slow responses because they are short relative to the conditioned ISI range. The 2–10 s ISI range is an important component of the PVT as it dictates the signal load of the test. It is not suggested that the ISI range of the PVT should be altered; rather, the findings suggest that if the short ISIs were removed from a response dataset the mean RT (or 1/RT) would contain less variability and hence would be more sensitive to the effects of fatigue. Identifying which response domains are more sensitive to the homeostatic and circadian influences may prove useful for other recent PVT developments. For example, Basner and Dinges (2012) have developed an ‘adaptive-duration’ PVT that uses an algorithm to predict whether a participant is a high, medium or low performer based on an initial set of responses. The task stops once sufficient responses have been sampled to classify the participant. If ISIs from 5 to 10 s represents an ‘ideal’ response range, sampling from this range may improve the algorithms’ predicative ability, leading to better predictions with less data. Acknowledgment This study was financially supported by the Australian Research Council. References Balbus, J.M., Stewart, W., Bolla, K.I., Schwartz, B.S., 1998. Simple visual reaction time in organolead manufacturing workers: influence of the interstimulus interval. Arch. Environ. Health 53 (4), 264–270. Basner, M., Dinges, D.F., 2011. Maximizing sensitivity of the psychomotor vigilance test (PVT) to sleep loss. Sleep 34 (5), 581.
5
Basner, M., Dinges, D.F., 2012. An adaptive-duration version of the PVT accurately tracks changes in psychomotor vigilance induced by sleep restriction. Sleep 35 (2), 193. Basner, M., Mollicone, D., Dinges, D.F., 2011. Validity and sensitivity of a brief psychomotor vigilance test (PVT-B) to total and partial sleep deprivation. Acta Astronaut. 69 (11), 949–959. Belenky, G., Wesesten, N.J., Thorne, D.R., Thomas, M.L., Sing, H., Redmond, D.P., Balkin, T.J., 2003. Patterns of performance degradation and restoration during sleep restriction and subsequent recovery: a sleep dose–response study. J. Sleep Res. 12, 1–12. Darwent, D., Ferguson, S.A., Sargent, C., Paech, G.M., Williams, L., Zhou, X., Dawson, D., 2010. Contribution of core body temperature, prior wake time and sleep stages to cognitive throughput performance during forced desynchrony. Chronobiol. Int. 27 (5), 898–910. Dinges, D., Pack, F., Gillen, K., Powell, J., Kribbs, N., Ott, G., Pack, A., 1994. Cumulative effects of acute partial sleep deprivation on PVT lapses and SSS ratings. J. Sleep Res. 23, 408. Dinges, D.F., Powell, J.W., 1985. Microcomputer analyses of performance on a portable, simple visual RT task during sustained operations. Behav. Res. Methods Instrum. Comput. 17 (6), 652–655. Dorrian, J., Rogers, N.L., Dinges, D.F., 2005. Psychomotor vigilance performance: neurocognitive assay sensitive to sleep loss. In: Kushida, C.A. (Ed.), Sleep Deprivation: Clinical Issues, Pharmacology and Sleep Loss Effects, vol. 39–70. Marcel Dekker Inc., New York. Glenville, M., Broughton, R., Wing, A.M., Wilkinson, R.T., 1978. Effects of sleep deprivation on short duration performance measures compared to the Wilkinson auditory vigilance task. Sleep 1, 169–176. Horne, J.A., 1993. Human sleep, sleep loss and behaviour. Implications for the prefrontal cortex and psychiatric disorder. Br. J. Psychiatry 162, 413–419. Kirby, N., McConaghy, J., 1986. Responding time: a response equivalent to inspection time. Pers. Individ. Diff. 7 (5), 687–694. Kirby, N., Nettelbeck, T., 1991. Speed of information processing and age. Pers. Individ. Diff. 12 (2), 183–188. Lamond, N., Dawson, D., Roach, G.D., 2005. Fatigue assessment in the field: validation of a hand-held electronic psychomotor vigilance task. Aviat. Space Environ. Med. 76 (5), 486–489. Lamond, N., Jay, S.M., Dorrian, J., Ferguson, S.A., Roach, G.D., Dawson, D., 2008. The sensitivity of a palm-based psychomotor vigilance task to severe sleep loss. Behav. Res. Methods 40 (1), 347–352. Lisper, H.-O., Kjellberg, A., 1972. Effects of 24-hour sleep deprivation on rate of decrement in a 10-minute auditory reaction time task. J. Exp. Psychol. 96, 287–290. Los, S.A., Knol, D.L., Boers, R.M., 2001. The foreperiod effect revisited: conditioning as a basis for nonspecific preparation. Acta Psychol. (Amst.) 106 (1), 121–145. Matthews, R., Ferguson, S., Zhou, X., Roach, G., 2011. A forced desynchrony correction: accounting for longitudinal confounders. In: Little Clock, Big Clock: Molecular to Physiological Clocks. Australasian Chronobiology Society, Melbourne, Australia., pp. 43–47. Matthews, R., Ferguson, S.A., Zhou, X., Darwent, D., Sargent, C., Kennaway, D.J., Roach, G.D., 2010. The interactive influence of the internal body clock and prior wake on task vigilance. Ergon. IJE&HF 32 (1), 37–48. Matthews, R.W., Ferguson, S.A., Zhou, X., Kosmadopoulos, A., Kennaway, D.J., Roach, G.D., 2012a. Simulated driving under the influence of extended wake, time of day and sleep restriction. Accid. Anal. Prev. 45 (Suppl.), 55–61. Matthews, R.W., Ferguson, S.A., Zhou, X., Sargent, C., Darwent, D., Kennaway, D.J., Roach, G.D., 2012b. Time-of-day mediates the influences of extended wake and sleep restriction on simulated driving. Chronobiol. Int. 29 (5), 572–579. Nickerson, R.S., 1967. Psychological refractory phase and the functional significance of signals. J. Exp. Psychol. 73 (2), 303. Nickerson, R.S., 1968. Response time to the second of two signals following varied vs constant intersignal intervals. Percept. Psychophys. 4 (2), 78–80. Roach, G.D., Dawson, D., Lamond, N., 2006. Can a shorter psychomotor vigilance task be used as a reasonable substitute for the ten-minute psychomotor vigilance task? Chronobiol. Int. 23 (6), 1379–1387. Roach, G.D., Lamond, N., Dawson, D., 2016. Feedback has a positive effect on cognitive function during total sleep deprivation if there is sufficient time for it to be effectively processed. Appl. Ergon. 52, 285–290. Telford, C.W., 1931. The refractory phase of voluntary and associative responses. J. Exp. Psychol. 14 (1), 1. Van Dongen, H.P.A., Maislin, M.A., Mullington, J.M., Dinges, D.F., 2003. The cumulative cost of additional wakefulness: dose–response effects on neurobehavioural functions and sleep physiology from chronic sleep restriction and total sleep deprivation. Sleep 26 (2), 117–126. Vickers, D., Nettelbeck, T., Willson, R., 1972. Perceptual indices of performance: the measurement of ‘inspection time’ and ‘noise’ in the visual system. Perception 1 (3), 263–295. Wilkinson, R.T., Houghton, D., 1982. Field test of arousal: a portable reaction timer with data storage. Hum. Fact. 24 (4), 487–493. Zhou, X., Ferguson, S.A., Matthews, R., Sargent, C., Darwent, D., Kennaway, D.J., Roach, G.D., 2011. Dynamics of neurobehavioral performance variability under forced desynchrony: evidence of state instability. Sleep 34 (1), 57–63. Zhou, X., Ferguson, S.A., Matthews, R.W., Sargent, C., Darwent, D., Kennaway, D.J., Roach, G.D., 2010. Interindividual differences in neurobehavioural performance in response to increasing homeostatic sleep pressure. Chronobiol. Int. 27 (5), 922–933.
Please cite this article in press as: Matthews, R.W., et al., Using interstimulus interval to maximise sensitivity of the Psychomotor Vigilance Test to fatigue. Accid. Anal. Prev. (2015), http://dx.doi.org/10.1016/j.aap.2015.10.013