EXPERIMENT 1: E FFECTS OF STIMULUS ENCODING ON SLEEP - RELATED EXPLICIT MEMORY

Experiment 1 investigated the effects of sleep in face learning using implicit and explicit measures of memory. Experiment 1 also tested whether the effect of sleep on face learning was affected by the level of encoding of the faces prior to sleep. We tested two groups of participants across two sessions and different encoding and testing tasks performed on photographs of cropped faces. The two sessions were separated by a 12-hour period, with intervening sleep for one group of participants and without sleep for the other group.

4.2.1. Methods Participants

Twenty-eight healthy volunteers (12 males, mean age 26.5 ± 4.1 years) gave informed consent to participate in this study approved by the Geneva University Hospitals Ethics Committee. They were randomly assigned to two distinct experimental groups (14 participants each); there was no significant age or gender difference between the groups.

None of the volunteers had any history of neurological or psychiatric disorders, of drug or alcohol abuse. None suffered from sleep disturbances as assessed both by clinical interview and by a questionnaire adapted from The St Mary’s Hospital Questionnaire (Ellis et al., 1981)

and the Pittsburgh Sleep Quality Index (Buysse et al., 1989). They were not under the influence of hypnotic, anti-allergic, sedative or anti-depressant medications at the time of the study.

Stimuli

The stimuli consisted of black and white frontal photographs of 146 unfamiliar individuals (73 males) with neutral expressions. The stimuli were selected from three different databases:

the Karolinska Directed Emotional Faces (Lundqvist et al., 1998), the AR face database (Martinez & Benavente, 1998), and some pictures from George et al. (George et al., 2001) were used with permission. Each photograph underwent four processing steps: (1) faces with distinctive features (glasses, moustache or beard, etc.) were excluded; skin spots and scars were removed using Photoshop (program ref); (2) the size of the face was adjusted to align the eyes, nose, and mouth from all faces at the same vertical positions; (3) the distribution of grey levels was adjusted to equalize all the photographs for contrast and luminance using a homemade Matlab script (www.mathworks.com); (4) the faces were cropped by placing a black mask covering the hair, ears, neck and face outlines. Figure 1 shows two exemplars of the resulting face stimuli. Faces were presented at the centre of a 17 inches CRT monitor (refresh rate: 60 Hz; resolution: 1024x768) covering 15 degrees of visual angle. Two stimuli were used as fillers at the beginning of each run to allow for adaptation to the task; they were removed from subsequent analyses. The remaining 144 stimuli were randomly divided into 6 sets of 24 items (12 males) each, as explained below in the experimental protocol. Different lists of stimuli were created and counterbalanced across participants. The experiments were programmed using the E-Prime software (www.pstnet.com, Psychology Software Tools) allowing for stimulus presentation and response recording with precise timing.

Figure 1. Two examples of face stimuli used in the experiments. Pictures of faces were carefully controlled for mean luminosity and contrast, as well as for size and vertical position of eyes. Stimuli did not include faces with beards, glasses, make-up, scars or other skin imperfection.

Experimental protocol

Fourteen different participants forming the day-group were first tested at 8 a.m. and then retested at 8 p.m. Fourteen participants belonging to the night-group underwent the first session at 8 p.m. and the second one at 8 a.m. Between the sessions, the subjects were asked to follow their usual activities, but were not allowed any nap, any exhausting sportive exercise, nor any stressful mental performance (such as university exams or intensive preparation for such exams). Participants from the night-group were also asked to fill a sleep questionnaire in the morning to judge the quality of their sleep during the experimental night.

All the participants in this group reported good subjective sleep quality. Both groups of participants performed three study-test pairs sequentially distributed across both experimental sessions, with no delay between study and test for the first and third pairs, but with a 12h delay for the second pair (with or without sleep between study and test; see Figure 2). A surprise recognition test was administrated at the end of the second session. Three completely different sets of stimuli were presented during each study-test pair. Each study phase consisted of 24 stimuli and each test phase of 48 stimuli (24 old and 24 new).

During the three study phases, the participants were asked to judge whether the face presented was pleasant or unpleasant by pressing two different keys using their dominant hand. Each trial began with a fixation cross (1 s), followed by the stimulus presentation (3 s), and then the word ‘response’ appeared. The participants were instructed to respond when the response screen appeared. Thus, encoding time was kept constant for all faces. During the three test phases, the participants were presented with the faces from the last study phase intermingled with new ones and had to indicate on each trial whether the face displayed after the fixation cross (1 s) was a male or a female. They were required to respond as quickly as possible after the face appeared on the screen. The next trial started 2.5 s after the response. These test phases allowed us to measure any priming effects (faster reaction times) for the previously studied faces. At no time during the study or test phases participants were informed that memory was investigated, in order to avoid individual strategy of memorization. During the recognition phase, we tested explicit memory for the faces that had been presented in the three study-test pairs. Each trial started with a fixation cross (1 s), followed by the presentation of one of the 144 faces. The participants had to tell whether they had already seen the face presented. Like in the test phases, the face remained on the screen until the response was recorded (time limit 3 s). Key presses and reaction times (RTs) were collected during all phases.

Figure 2. Experimental protocols for Experiment 1. Effects of sleep versus wake periods on memory for faces. In Experiment 1, participants performed pleasantness judgments and gender discrimination judgments, repeatedly on three different sets of faces, followed by a surprise recognition task that was administered at the very end of the protocol. Two groups of participants were tested across 12-hour delays including either daytime wakefulness or one night of normal sleep. In this protocol, items from the first study-test pair were presented during the first session and recognition performance was assessed after 12 hours during the second session, thus providing measures of long-term explicit memory. Items from sets 2 and 3 were presented at least once during the second session, i.e., a few minutes before the recognition task, which therefore reflected measures of short-term explicit memory.

4.2.2. Results

We first report the analyses on RTs for the test phases, for recognition phase, and then the results from the recognition performance (number of hits). All the data were analyzed using repeated-measure analysis of variance (ANOVAs), with additional post-hoc comparisons whenever relevant to the main hypotheses tested by the experiment (significance level set to p<.05). Please note that since the participants performed the pleasantness task without any stringent time constraint, RTs from the study phases may not convey meaningful information about memory processing and thus were not considered for further analysis. Hit rate during test phases was not analyzed because gender discrimination performance was at ceiling for all the subjects across all sessions (>90% correct).

Reaction times

In each test phase, mean RTs for correct responses were calculated separately for faces that had already been presented in the corresponding study phase (‘old’) and for faces that were presented for the first time (‘new’). RTs from the test phases were analyzed using a repeated-measure ANOVA with one between-subject factor (Group: ‘night’, ‘day’) and two within-subject factors (Test Phase: test 1, test 2, test 3; Repetition: ‘old’, ‘new’). This analysis revealed a main effect of Test Phase [F(2,52) = 8.55 , p < .001 ], with faster response in phase 2 [F(1, 26) = 6.55, p = .02] and in phase 3 [F(1, 26) = 11.44, p < .01] than in phase 1 [RTs in

T1: M = 724 ms, SE = 36 ms; T2: M = 671 ms, SE = 19 ms; T3: 652 ms, SE = 19 ms]. There was a trend for RTs to be faster in phase 3 than in phase 2 [F(1, 26) = 3.76, p = .06].

However, there was no effect of Group [F(1, 26) = 1.02, p = .32; RTs for ‘day’ group: M = 704 ms, SE = 37 ms; ‘night’ group: M = 661 ms, SE = 22 ms], no Repetition effect [F(1, 26)

< 1, p = .43; RTs for ‘old’: M = 679 ms, SE = 21 ms; ‘new’: M= 686 ms, SE = 22 ms] and no interaction (all p > .73). Thus, a general motor learning process occurred through phases but no immediate or delayed repetition priming effects, and no sleep-related modulation, during the test phases were observed (gender discrimination task).

Mean RTs from the recognition phase were analyzed using an ANOVA with one between-subject factor (Group) and two within-between-subject factors (Study-Test Pair in which the faces were shown: 1, 2, 3; Repetition: once, twice). There was a significant Repetition effect, with faster responses for repeated items than for those presented only once [F(1, 26) = 5.23, p <

.05]. However, there was no effect of Group [F(1, 26) = 1.40, p = .25], no effect of Study-Test Pair [F(2, 52) = 1.27, p = .29], and no interaction (all p > .14).

Explicit recognition

Experiment 1 was designed to investigate the effect of sleep on implicit memory for faces through a repetition-priming paradigm. A recognition test was added at the very end of the experiment as a preliminary indication of a possible effect of sleep on explicit memory for faces (this issue will be addressed more thoroughly in Experiment 2). Due to a very large number of items (144) and time constraints, we did not add any novel faces during the recognition phase, thus precluding any d-prime calculation.

The number of correctly recognized faces (or hits) during the recognition phase was analyzed separately for items presented twice (during both study and test: T1old, T2old, T3old) and once (only during test: T1new, T2new, T3new) during the study-test pairs 1, 2, or 3.

Items from study-test pair 1 clearly allowed us to test sleep-related consolidation effect. Items for study-test pair 3 were presented just before the recognition test, without sleep or wake delay, and were used as circadian controls. However, items from study-test pair 2, given their presentation across two sessions, did not allow us to clearly assess a sleep-related consolidation effect and were consequently excluded from this analysis.

Recognition hits for items from study-test pairs 1 and 3 were analyzed using a repeated-measure ANOVA with one between-subject factor (Group: ‘night’, ‘day’) and two within-subject factors (Study-test pair: 1 and 3; Repetition: ‘Told’, ‘Tnew’). The ANOVA showed a significant effect of Repetition, with a better recognition for the ‘Told’ (repeated stimuli) compared to the ‘Tnew’ ones that were presented only once prior to the recognition phase [F(1, 26) = 81.29, p < .0001], but no effect of Group [F(1, 26) < 1, p < .47] or Study-Test Pair

[F(1, 26) < 1, p < .40]. There was significant a triple Group X Study-Test Pair X Repetition interaction [F(1, 26) = 4.65, p < .05]. Post-hoc Scheffé test confirmed that recognition for

‘Told’ items from pair 1 was higher for the night group than for the day group (p < .05), whereas recognition for ‘Tnew’ items from this same pair did not differ between the two groups (p = .99). No difference was found between the two groups either for old or new items from pair 3, excluding any circadian effect (Figure 3).

Figure 3. Explicit recognition. Performance was selectively enhanced after one night of sleep for stimuli seen twice across a pleasantness judgment and a gender discrimination tasks. ** p < .01); ***

p < .0001.

4.2.3. Discussion

Here, we found no evidence for immediate repetition priming effect during the test phases (gender discrimination task). Although we used similar stimuli, our result departs from those reported by Goshen-Gottstein & Ganel (Goshen-Gottstein & Ganel, 2000). Several differences between the studies may explain this discrepancy such as the use of a pleasantness judgment task versus an intelligence judgment task at encoding, as well as shorter stimulus-presentation duration and many more face-stimuli in our study. This might have led to the formation of comparatively weaker or less differentiated traces for the face-stimuli in our study, limiting subsequent priming effect when the participants performed the gender discrimination task.

By contrast, repetition facilitation effect was found during recognition for both ‘day’ and

‘night’ groups, with faster RTs for ‘old’ (presented both in study and test) as compared to

‘new’ (presented only in test) items. While we consider the recognition task here primarily as an explicit measure of memory performance (number of hits), reaction times collected during

a recognition task might be used to assess implicit aspects of memory processing (Jacoby, 1991), as in the study by Wagner et al. (2003). However, unlike Wagner et al., (2003), the

“priming” effect found in our study did not appear to be strictly sleep-dependent since both day and night groups showed similar RT decreases. This discrepancy between the studies might possibly be due to differences in the stimuli (cropped versus whole faces), time delays (3 hours in Wagner et al.’s study versus 12 hours in ours) and to the fact that Wagner et al.

(2003) compared RT from learned faces and really old ones whereas we used faces with different numbers of expositions (once or twice during the learning). The recognition task from this experiment does not allow the measurement of standard repetition priming, however, it demonstrates that long-term response facilitation to repeated face stimuli can occur after both one day of wakefulness or one night of sleep to the same extend. In agreement with Goshen-Gottstein & Ganel’s results (2000), these “repetition priming” effects were observed in our experiment when configural information was extracted from the faces.

No external features were presented and the task focused on identity.

An important result from Experiment 1 is that enhanced recognition of individual faces occurred specifically in the night group for faces from study-test pair 1 (i.e. stimuli exclusively shown before sleep). As both groups displayed similar levels of performance for faces encoded in the study-test pair 3 (despite different time schedules between groups), it is unlikely that group differences for faces from phase 1 would be due to circadian effects. This result is a strong indicator of real consolidation effects of sleep. However, we are aware that the lack of novel items in the recognition test weakens this finding. Therefore, the recognition memory issue was tested more thoroughly in Experiment 2.