Circadian- and wake-dependent influences on face-name memory in healthy men and women over three weeks of chronic sleep restriction

Abstract

Objectives:

Facial recognition is one of the key functions of the human brain, and linking a face to a name is critical in many social and occupational settings. This study assessed circadian- and wake-dependent effects on face-name recognition in healthy adults.

Methods:

13 healthy adults (20–70yrs; 7F) were studied in a 39-day inpatient protocol that included three weeks of 28h forced desynchrony with sleep restriction (6.5:21.5h sleep:wake). Starting 3h after scheduled wake, 6 novel face-name pairs were presented every 4 waking hours; recognition was tested 2h later. Performance data were averaged across ~4h circadian phase or time-awake bins.

Results:

Face-name recognition deteriorated with increased time awake (p<0.0001) and exhibited significant circadian variation (p<0.0001), with worst performance shortly after the core temperature nadir. There was a significant interaction between sex and circadian phase (p=0.0177), with women performing significantly better than men at all circadian phases except 60° and 120°. Women exhibited a significantly higher amplitude than men during the third week of forced desynchrony (p<0.01).

Conclusions:

Like many other aspects of neurobehavioral performance, recalling face-name associations is impacted by both duration of time awake and circadian phase. These results have implications for face recognition testing in medical contexts, such as in testing for dementia, because performance may be impacted by sleep deficiency and the time of testing.

Introduction

Remembering someone’s face and name is a social function unique to humans. Losing this ability can lead to difficulty building and maintaining social connections(1), which are strongly correlated with mental and physical health outcomes(2, 3). Indeed, the recent COVID-19 pandemic and subsequent widespread usage of face masks highlight how usual social interactions in various settings are impacted when face recognition is impaired(4). For example, patients’ difficulty in differentiating among treating clinicians has been shown to hinder the establishment of positive patient-doctor relationships(5), and surveys of the general population reveal reduced community involvement associated with mask usage(6).

Neurobehavioral performance on many tasks is susceptible to accumulated sleep pressure, deteriorating with increasing time awake(7–11). Many of these tasks, such as the psychomotor vigilance task, digit-symbol substitution test, and addition test, also exhibit circadian variation(7–9), suggesting that both a sleep-wake homeostatic process and the circadian system influence human cognitive performance. However, fewer studies have assessed these effects with ethologically relevant tasks. Here, we used a forced-desynchrony protocol with chronic sleep restriction to explore whether the ability to remember recently learned face-name pairs is impacted by time awake and/or circadian phase in healthy adults.

Participants and Methods

Participants.

Data were collected from 13 young (n=4; mean±sd: 24.8±1.5yrs; 3F) and older (n=9; 60.1±5.1yrs; 4F) healthy adults who participated in a 39-day study designed to test the impact of chronic sleep restriction on metabolism, sleep, and performance(12, 13). All participants gave written informed consent, and the study was approved by the Partners Health Care Human Research Committee. Detailed recruitment/screening procedures have been published elsewhere(12, 13); briefly, participants were screened to exclude medical, psychological, and sleep disorders; acute and chronic illnesses; regular medication usage (except birth control); recent history of shiftwork; and recent trans-meridian travel.

Protocol.

Participants maintained a regular sleep-wake schedule at home with 10h time-in-bed (TIB) for at least 3 weeks immediately before admission. The inpatient study began with three sleep extension days (12h overnight TIB plus 4h naps) and three baseline days (10h TIB), followed by three weeks on a 28h forced desynchrony (FD) protocol, during which time participants were scheduled to 5.6h/24h TIB (Figure 1; details previously published(12, 13)). This protocol distributes scheduled sleep and wake across the circadian cycle, allowing wake- and circadian- dependent effects to be separated and quantified(14).

Figure 1. Study schematic.

Black bars indicate scheduled sleep episodes in the laboratory. Participants underwent 3 weeks of regular 10h time-in-bed at home (dark gray bars) before admission. The inpatient study began with 3 sleep extension days, each with a 12h nighttime sleep opportunity and a 4h nap opportunity, followed by 3 baseline days with 10h overnight sleep opportunities. They then underwent 3 weeks of forced desynchrony consisting of 28h “days” with 6.5h sleep opportunities (equivalent to 5.6h/24h) and 21.5h wake episodes. Core body temperature was collected every minute throughout the forced desynchrony segment to estimate circadian period and phase. The Face Name Task was administered every 4 hours starting 3 hours after wake; each test session (white circles) occurred 2 hours after the corresponding encoding session (not shown).

Face-Name Task (FNT).

Beginning three hours after scheduled wake, each participant was prompted to take a computerized FNT every four waking hours. FNT details are published elsewhere(15). Briefly, each encoding session consisted of six novel face-name pairs, presented one at a time on the computer screen for five seconds each; each pair was presented twice within a session. Memory was assessed two hours after encoding, during which time face-name pairs were presented one at a time in random order. Each face was presented twice, once with the correct name and once with an incorrect name, and the participant indicated whether the name was correct (yes/no). Neither faces nor names were repeated across encoding sessions. Photos of male and female faces of multiple ethnicities and ages with neutral expressions were presented throughout the study; ethnicity and age group were selected randomly in proportion to the number of faces with that characteristic in the photo database. Each session contained faces of the same sex and age group.

Data analysis.

The first test session was excluded to minimize learning effects. The FNT was not administered on the third baseline day and only once on the second; therefore, only data from the first baseline day and three weeks of FD were included in analysis. Data from each wake episode were binned into 4h time-awake bins to examine the homeostatic effect of time awake on FNT performance. The intrinsic circadian period of each participant was estimated from continuously recorded core body temperature (CBT) data using non-orthogonal spectral analysis(14, 16), allowing a circadian phase from 0 to 359° to be assigned to each minute of the study, with 0° corresponding to the fitted CBT minimum. To examine the effect of circadian phase on FNT performance, data were also binned into 60° (~4h) circadian phase bins.

Statistics.

Statistical analyses were conducted with SAS version 9.4 (SAS Institute, Cary, NC). We used generalized linear mixed models with a binomial distribution and logit link function to evaluate FNT performance, with time awake, circadian phase, and forced desynchrony week as fixed effects and participant as a random effect. We tested sex assigned at birth (male, female) and age as covariates, along with all two-way interactions. Significant covariates and interactions were retained in the final models. For interactions with sex as a factor, post-hoc analyses were conducted by testing the main effect of sex at each level of time awake, circadian phase, or FD week. The critical significance level was set to α=0.05. Bonferroni corrections were applied to all post-hoc comparisons for the linear mixed models, and only significant results are reported. Finally, we fit a cosinor model (~24h period) with sex, FD week, phase, time-awake, and 2-way interactions with sex included as fixed effects and participant as a random effect to confirm circadian rhythmicity.

Results

We found no significant effect of age on face-name recognition, so data from young and older participants were combined for all analysis. During the baseline, we found no effect of time awake but a significant effect of sex such that women outperformed men (p<0.001).

During FD, we found a significant main effect of time awake, such that face-name performance declined significantly between 5h and 9h of wakefulness (unadjusted p<0.0001; adjusted p<0.01) and was significantly worse at 17h compared to 5h (unadjusted p<0.0001; adjusted p<0.001). We also found significant main effects of FD week, sex, and circadian phase, as well as significant interaction effects between FD week and sex as well as phase and sex. Post-hoc analyses revealed that women improved significantly between the first and third week of FD (unadjusted p<0.0001; adjusted p<0.001), whereas men did not. Women and men did not perform significantly differently in the first week of FD, but women performed significantly better than men during the second (unadjusted p<0.001; adjusted p<0.01) and third (unadjusted p<0.001; adjusted p=0.0135) FD weeks. Finally, women significantly outperformed men at circadian phases 0° (unadjusted p<0.01; adjusted p=0.033), 180° (unadjusted p<0.001; adjusted p=0.0105), 240° (unadjusted p<0.001; adjusted p<0.01), and 300° (unadjusted p<0.01; adjusted p=0.018).

Cosinor analyses confirmed that FNT performance had a circadian rhythm with a peak at 240° (phase: p<0.001). Amplitude was not significant for men in any of the FD weeks, although it approached significance in the first FD week (p=0.059). In contrast, amplitude was significant for women in FD week 1 (p=0.04), week 2 (p<0.01), and week 3 (p<0.001). We next tested for differences in amplitude between men and women in each FD week and found a significantly higher amplitude in women compared to men in FD week 3 (p<0.01).

Discussion

We found that correctly remembering a newly-learned face-name association exhibits both circadian- and wake-dependent effects in healthy adults. Performance was at its worst at circadian phase 60°, corresponding to usual wake time under entrained conditions. Performance reached a peak at circadian phase 240°, corresponding with the wake maintenance zone, a 2–3h interval of reduced sleep propensity in the biological evening(17). We also found that performance deteriorated over 17h of wakefulness, suggesting that these two influences may oppose each other to allow stable face-name memory across a typical entrained day, as has been observed for other types of neurobehavioral performance(18).

Consistent with previous literature(19), we found that women performed significantly better on this task than men overall. We also identified a significant interaction between sex and phase. Sex differences in the circadian regulation of other cognitive domains such as verbal working memory and vigilant attention have previously been reported, with women exhibiting more pronounced circadian rhythms compared to men(20). These sex differences have typically been attributed to sex hormones, which have been shown to impact cognitive performance(21). Here we found that circadian modulation of FNT performance appeared to increase in women but decrease in men over the three weeks of the protocol. By the third FD week, women exhibited a significantly greater amplitude than men, with an average increase of ~12.1% from the trough to peak of their performance compared to ~3.3% in men. Unexpectedly, we also found that women demonstrated improved performance from week 1 to week 3, suggesting either a potential sex difference in the response to chronic sleep loss, or possible sex differences in the strategy employed during the face-name recognition task. Indeed, some evidence suggests that the female advantage in face processing could be driven by differences in visual scanning behaviors during initial learning(22), and that women show more neural activity and faster structural face encoding than men(23). It may be the case that these sex differences in the initial encoding session are differentially impacted by chronic sleep loss, resulting in the different patterns of performance in men and women over the 3-week study.

Limitations.

Age effects have been previously reported on face-name recognition tasks(24, 25), but we found no significant effect of age in this study. However, our sample did not include middle-aged adults, and male participants were primarily older whereas female participants included a fairly even distribution of both young and older adults. Thus, it is possible that some of the interactions we detected between sex and phase could be influenced by the uneven distribution of ages between the two groups. Menstrual phase has been shown to interact with vulnerability to sleep loss(26). However, we were unable to control for this given the duration of each experiment and that four of the women were post-menopausal. Finally, although we chose to include sex in our statistical model due to the considerable evidence in the literature for sex-based differences in performance, the original study was not powered to investigate sex differences; thus, further appropriately powered studies would be prudent to confirm our findings.

Figure 2. Circadian- and wake- dependent effects on Face Name Task (FNT) performance in men and women across 3 weeks of forced desynchrony (FD) with restricted sleep.

Panels A-C: FNT performance decreased with longer time awake in each of the three weeks of FD. Panels D-F: Women (red diamonds) exhibited significant circadian modulation of FNT performance in all three FD weeks, whereas men’s (blue squares) performance did not exhibit a significant circadian rhythm in FD weeks 2 or 3, and only a trend in FD week 1. Fitted cosine waves are shown for women (orange) and men (blue) in D-F, along with average values. Open symbols and error bars depict means±SEM.

Public health relevance.

As highly social primates, humans rely heavily on face-name recognition to develop interpersonal connections and navigate everyday interactions. Numerous studies have demonstrated that the strength of these social connections is associated with both physical and mental health, making it important to better understand the factors influencing this critical cognitive ability.