Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2018 Dec 1.
Published in final edited form as: Chronobiol Int. 2018 Jan 17;35(6):746–749. doi: 10.1080/07420528.2017.1411361

Total Sleep Deprivation Does Not Significantly Degrade Semantic Encoding

KA Honn 1,2,*, DA Grant 1,2,*, JM Hinson 1,3, P Whitney 1,3,†, HPA Van Dongen 1,2
PMCID: PMC6126540  NIHMSID: NIHMS1502372  PMID: 29341785

Abstract

Sleep deprivation impairs performance on cognitive tasks, but it is unclear which cognitive processes it degrades. We administered a semantic matching task with variable stimulus onset asynchrony and both speeded and self-paced trial blocks. The task was administered at baseline and 24 hours later after 30.8 hours of total sleep deprivation or matching well-rested control. After sleep deprivation, the 20% slowest response times were significantly increased. However, the semantic encoding time component of the response times remained at baseline level. Thus, the performance impairment induced by sleep deprivation on this task occurred in cognitive processes downstream of semantic encoding.

Keywords: cognitive performance, dissociated components of cognition, vigilant attention

Introduction

Sleep deprivation has considerable negative impact on performance across a wide range of cognitive tasks (Killgore, 2010). Sleep deprivation-induced cognitive impairments have been shown to be task-dependent and individual-specific (Van Dongen et al., 2004). Targeted countermeasure development would be facilitated by a deeper understanding of the underlying causes of such impairments, including which components of cognition are most susceptible to sleep deprivation. One aspect of cognition that has been extensively studied in this context is vigilant attention. A hallmark effect of sleep deprivation on vigilant attention is the emergence of performance variability, giving rise to attentional lapses and response time (RT) distributions with a long right tail (Dinges & Kribbs, 1991). This performance variability increases with longer time awake and is modulated by time of day (Doran et al., 2001).

While vigilant attention has consistently been shown to be sensitive to sleep deprivation, studies of other cognitive functions have yielded mixed results (Killgore, 2010). This may be due to use of performance tasks that incorporate multiple functions and fail to differentiate them (the “task impurity problem”; Whitney & Hinson, 2010). Tasks that can be deconstructed to isolate particular cognitive functions – such as stimulus encoding, working memory scanning, suppression of interference, and motor response preparation – provide greater insight into the specific effects of sleep deprivation on cognition (Jackson et al., 2013). One important, outstanding question is whether sleep deprivation impairs cognitive performance because it degrades the encoding of information. Here, we used a variable stimulus onset asynchrony (SOA) semantic matching task to investigate the specific effect of total sleep deprivation (TSD) on semantic encoding.

Methods

N=69 carefully screened, healthy subjects (33 males, 36 females), ages 22–40 (mean ± SD: 26.7 ± 4.8), completed a 4-day/3-night in-laboratory study. The study was approved by the Institutional Review Board of Washington State University. Subjects gave written informed consent and were paid for their time.

Subjects were in the laboratory continuously for 4 days (3 nights). They were randomized to either a TSD condition (n=38) or a control condition (n=31). Subjects in the TSD condition had 10-hour sleep opportunities (22:00–08:00) on the first and third nights, with a 38-hour period of acute TSD from 08:00 on day 2 through 22:00 on day 3. Subjects in the control condition had 10-hour sleep opportunities (22:00–08:00) on each of the three nights. Subjects remained in the laboratory throughout the study and were continuously monitored by trained staff.

The variable SOA semantic matching task was administered twice: at 14:50 on day 2 and day 3. That is, test sessions took place at 6.8 hours awake on day 2 and at 30.8 hours awake (TSD condition) or 6.8 hours awake (control condition) on day 3. There were two equivalent versions of the task, which were administered in randomized order. Subjects were instructed that in each trial they would see a word and then, either immediately or after a brief delay (the SOA), a second word. They were to determine whether or not the two words belong in the same semantic category (e.g., “corn” and “peas” are both vegetables). They were asked to respond as fast as possible, while still being accurate, with either 1 (yes, the words belong in the same category) or 2 (no, the words do not belong in the same category). Category terms were drawn from a normative database (the American young adult norms; Yoon et al., 2004).

Each test session on the variable SOA semantic matching task had four alternating blocks of trials. Two blocks were computer-paced, with the next trial starting automatically after the previous trial; and two blocks were subject-paced, with the subjects controlling the task pace by pressing the space bar to begin the next trial. Each test session included 224 trials, with 14 positive (yes, 1) and 14 negative (no, 2) trials at each of four SOAs (0, 250, 500, and 750 ms) and two levels of pacing (computer-paced and subject-paced). The two types of pacing allowed assessment of the impact of vigilant attention deficits on performance. The variability in the SOA allowed separation of the time needed for semantic encoding from overall RT.

Results

Response accuracy on the variable SOA semantic matching task was greater than 90% even during TSD, with no significant difference in accuracy between subject-paced and computer-paced blocks (mixed-effects ANOVA by pacing: F1,31000=0.83, p=0.36). RT analyses were based on correct responses only. RT was defined as the time between the presentation of the second word and the subject’s response. For each subject, test session, and pacing (computer-paced/subject-paced), RTs were divided into quintiles, ranging from the fastest 20% (Q1) to the slowest 20% (Q5) of responses. Quintile-based results are depicted in Fig. 1 (top).

Figure 1.

Figure 1.

Open in a new tab

Overall response times and encoding times on the variable SOA semantic matching task. Top: mean RTs (± SE) per quintile in session 1 (top left) and session 2 (top right) for the control condition (gray) and TSD condition (black) for each SOA (0/250/500/750 ms). Quintile 1 (Q1) represents the fastest 20% of each subject’s responses; quintile 5 (Q5) represents the slowest 20%. The subject-paced blocks are shown with solid lines and circle markers and the computer-paced blocks are shown with dashed lines and triangle markers. Bottom: Mean encoding time (± SE) in session 1 (bottom left) and session 2 (bottom right) for the control condition (gray) and TSD condition (black) for the subject-paced trial blocks (solid bars) and computer-paced trial blocks (dashed bars).

RT encompassed the time used for stimulus encoding, decision-making, and motor response processes. At 0 ms SOA, when the two words were presented simultaneously, RT included the semantic encoding time of both words. In contrast, at 750 ms SOA, which should have been more than enough of a delay to fully encode the first word before the second word was presented, RT included the semantic encoding time of the second word only. The encoding time of the first word could thus be assessed as the difference between the mean RTs for the 0 ms and 750 ms SOAs (Whitney et al., 2001) – this was used as our measure of semantic encoding. Semantic encoding results are depicted in Fig. 1 (bottom).

Response times

First, task performance was analyzed by subjecting the RTs (regardless of quintiles) to mixed-effects ANOVA by condition (TSD/control), test session (1/2), pacing (computer-paced/subject-paced), and SOA (0/250/500/750 ms). There was a significant main effect of SOA (F3,29000=465.48, p<0.001). At 0 ms SOA (when the two words were presented simultaneously), RTs were slowest because they included the full semantic encoding time of both words. As expected, greater SOAs were associated with faster RTs (Fig. 1, top).

There were also significant effects of condition (F1,29000=16.57, p<0.001), session (F1,29000=843.00, p<0.001), condition by session interaction (F1,29000=1517.04, p<0.001), and condition by SOA interaction (F3,29000=5.12, p=0.002). As anticipated, subjects in the TSD condition responded overall more slowly during session 2, when they had been awake for 30.8 hours (Fig. 1, top right panel, black curves). Additionally, there was a significant session by pacing interaction (F1,29000=13.01, p<0.001) and there were trends for interactions of condition by pacing (F1,29000=3.43, p=0.064) and condition by session by pacing (F1,29000=3.33, p=0.068). The computer-paced trial blocks tended to have slower mean RTs during TSD than the subject-paced trial blocks (Fig. 1, top right panel, black dashed curves).

Quintile RTs were analyzed with mixed-effects ANOVA by condition, session, pacing, SOA, and quintile (Q1–Q5). There was a significant 3-way interaction of condition by session by quintile (F4,29000=609.83, p<0.001), such that the slowest 20% of RTs (Q5) were particularly impaired in session 2 for the TSD condition (Fig. 1, top right panel, black curves, Q5). There was also a significant 4-way interaction of condition by session by pacing by quintile (F4,29000=4.21, p=0.002), such that the slowest quintile in session 2 for the TSD condition was most impacted in the computer-paced trial blocks (Fig. 1, top right panel, black dashed curve Q5).

Semantic encoding times

Mean encoding times were analyzed with mixed-effects ANOVA by condition, session, and pacing. There were no significant main effects or interactions (F1,201<1.06, p>0.30). Mean encoding times were around 250 ms regardless of condition, session, or pacing (Fig. 1, bottom). Mean encoding times restricted to Q5 responses for subjects in the TSD condition, subjected to mixed-effects ANOVA by session and pacing, also showed no significant main effects or interactions (F1,111<0.77, p>0.38). Thus, acute TSD did not significantly increase encoding time, even within the slowest 20% of RTs (where the TSD effect was most pronounced).

Discussion

Exposure to 30.8 hours of TSD had a substantial, adverse effect on performance on the variable SOA semantic matching task. This effect was particularly profound in the slowest 20% of responses (Q5), that is, the right tail of the RT distribution. The effect was further amplified in the trial blocks that were computer-paced, which emphasizes the vigilance aspect of task performance (Broadbent, 1953). These observations strongly suggest that performance impairment on the task reflected TSD-induced deficits in vigilant attention (Dinges & Kribbs, 1991; Doran et al., 2001).

Importantly, the design of the variable SOA semantic matching task allowed us to dissociate deficits in stimulus encoding from deficits in other components of cognition involved in performance (Whitney et al., 2001). There was, however, no significant effect of TSD on semantic encoding times. Thus, semantic encoding was resilient to TSD even though overall cognitive performance was substantially degraded. This result implies that the effect of TSD on overall performance – whether caused by vigilant attention deficits or otherwise – must occur downstream of the encoding of information. This finding contributes to efforts seeking to identify which cognitive processes are most sensitivity to TSD (e.g., Tucker et al., 2010; Ratcliff et al., 2011) and helps to narrow down the options for targeted countermeasure development.

Footnotes

Declaration of interest

The authors have no conflicts of interest related to this paper. This research was supported by NIH grant R21CA167691 to John Hinson.

References

  1. Broadbent DE. (1953). Noise, paced performance and vigilance tasks. Br J Psychol. 44(4):295–303. [DOI] [PubMed] [Google Scholar]
  2. Dinges DF, Kribbs NB. (1991). Performing while sleepy: effects of experimentally-induced sleepiness In Monk TH, ed. Sleep, Sleepiness, and performance. New York: Wiley, pp. 97–128. [Google Scholar]
  3. Doran SM, Van Dongen HPA, Dinges DF. (2001). Sustained attention performance during sleep deprivation: evidence of state instability. Arch Italian Biol. 139:253–67. [PubMed] [Google Scholar]
  4. Jackson ML, Gunzelmann G, Whitney P, Hinson JM, Belenky G, Rabat A, Van Dongen HPA. (2013). Deconstructing and reconstructing cognitive performance in sleep deprivation. Sleep Med Rev. 17:215–225. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Killgore WDS. (2010). Effects of sleep deprivation on cognition. Prog Brain Res. 185:105–29. [DOI] [PubMed] [Google Scholar]
  6. Tucker AM, Whitney P, Belenky G, Hinson JM, Van Dongen HPA. (2010). Effects of sleep deprivation on dissociated components of executive functioning. Sleep, 33(1):47–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Van Dongen HPA, Baynard MD, Maislin G, Dinges DF. (2004). Systematic interindividual differences in neurobehavioral impairment from sleep loss: evidence of trait-like differential vulnerability. Sleep, 27(30):423–33. [PubMed] [Google Scholar]
  8. Whitney P, Arnett PA, Driver A, Budd D. (2001). Measuring central executive functioning: what’s in a reading span? Brain Cogn. 45:1–14. [DOI] [PubMed] [Google Scholar]
  9. Whitney P, Hinson J. (2010). Measurement of cognition in studies of sleep deprivation. Prog Brain Res. 185:37–48. [DOI] [PubMed] [Google Scholar]
  10. Yoon C, Feinberg F, Hu P, Gutchess AH, Hedden T, Chen H-Y M, Jing Q, Cui Y, Park DC. (2004). Category norms as a function of culture and age: comparisons of item responses to 105 categories by American and Chinese adults. Psychol Aging. 19(3): 379–93. [DOI] [PubMed] [Google Scholar]

RESOURCES