Word recall correlates with sleep cycles in elderly subjects


: Giuliana Mazzoni, Dipartimento di Scienze, dell’ Educazione, Box 18B, Arcavacata di Rende, CS, Italy. Tel.:+39 055 631783.


Morning recall of words presented before sleep was studied in relation to intervening night sleep measures in elderly subjects. Night sleep of 30 elderly subjects aged 61–75 years was recorded. Before sleep, subjects were presented with a list of paired non-related words and cued recall was asked immediately after the morning awakening. Recall positively correlated with average duration of NREM/REM cycles, and with the proportion of time spent in cycles (TCT) over total sleep time (TST). No significant correlations were found with other sleep or wake measures. These results suggest the importance of sleep structure for sleep-related memory processes in elderly adults.


Age-related changes in sleep have been well documented in recent decades. Several investigations reported that some sleep parameters undergo important modifications with age ( Lairy et al. 1962; Kahn and Fisher 1969 ); night sleep is shorter and less consolidated than in younger adults with an increase in wake after sleep onset and sleep fragmentation ( Feinberg 1974; Feinberg et al. 1967; Webb 1982 ). The number of NREM/REM cycles and the proportion of TCT over TST decreases in older adults ( Lombardo et al. 1998 ); stage shifts are frequent and REM sleep is often interrupted ( Webb and Dreblow 1982). Changes in slow wave sleep (SWS) are a debated issue. Depending on the method of analysis, the percentage of SWS is decreased in the elderly or is similar to that observed in young adults ( Kahn and Fisher 1969; Webb and Dreblow 1982; Blois et al. 1983; Feinberg et al. 1983; Vitiello and Prinz 1994; Bonnet 1989 ).

Studies on young adults have shown that sleep can have an important role in the retention of material for learning and recall ( Empson and Clarke 1970; Fishbein et al. 1974; Salzarulo and Cipolli 1979; Smith 1985; Tilley and Warren 1984 ); some research has provided evidence for a specific facilitating effect of sleep on memory ( Jenkins and Dallenbach 1942; for a recent replication of the sleep effect on perceptual learning, see Karni et al. 1994 ). The role of single sleep states on memory has been investigated with unclear results. Several studies have stressed the positive role of REM sleep in both humans ( Karni et al. 1994; De Konink et al. 1989; Rotenberg 1992; Smith and Lapp 1991 ) and animals ( Smith and Kelly 1988; Hennevin et al. 1995; Smith 1985 ; for a review on research in both humans and animals, see Dujardin et al. 1990 ). Yet, other studies have questioned the specific role of both NREM and REM sleep for memory processes ( Ekstrand et al. 1977; Cipolli and Salzarulo 1979; Cipolli 1995 ).

Given the potential role of sleep on retention in memory and learning tasks, the varying degrees of spontaneous disruption of sleep in elderly subjects can provide an interesting condition to study the effect of sleep disturbances on memory processes. If sleep is relevant to intra-night memory processes, disrupted sleep might be detrimental to elderly subjects ( Salzarulo 1995).

In this study we assessed whether morning recall of material presented before sleep in elderly subjects could be related to sleep disorganization.

We hypothesized specifically that sleep organization (NREM/REM cycle) more than sleep states perse, is relevant to memory processes. Previous research has pointed out the possible positive role of NREM/REM cycles in anabolic processes and protein synthesis ( Fagioli et al. 1981; Salzarulo and Fagioli 1995 ). On the other hand, protein synthesis is important for long-term potentiation (LTP) and hence for memory ( Bliss and Collinridge 1993; Mattheis 1982; Barondes 1983). Specifically, NREM/REM cycle interruption or disorganization could affect protein synthesis, with consequent impairment of the processes involved in the consolidation of to-be-remembered material, resulting in poor recall.



Thirty elderly volunteers (13 females and 17 males; aged 68 ±5 years) took part in the experiment. Subjects were living at home and were recruited from the general population. All subjects were in good physical health, as verified by an extensive screening process that included the subject's medical history and a clinical examination. The subjects reported no symptoms of sleep apnoea or other respiratory disorders, no current physical illness requiring treatment, no evidence of current or past psychiatric illness, no cardiovascular problems, no history of drug or alcohol abuse, and no complaints of insomnia or daytime sleepiness. Average education (i.e. number of years spent in school) was 8±3 years; average Wechsler adult intelligence scale (WAIS) verbal score was 97.71 (81–111); average WAIS performance score was 106.08 (88–132), with an average index of deterioration (calculated according to the procedure reported in the WAIS manual) of 1.13.


Episodic memory was tested with two verbal memory tasks. The first task measured baseline memory ability and the second task tested sleep-related memory. The memory tasks each utilized a list of 20 pairs of unrelated words. Words were chosen from a large pool and were controlled for frequency (between 100 and 500 occurrences over 500 000 words) ( Bortolini et al. 1972 ) and imagery values (5–6 on a 7 point scale) ( Cornoldi 1975). The two lists of words were similar in difficulty.

The subjects were randomly assigned to two groups. Lists were counterbalanced across conditions (baseline and sleep-related memory tasks), so that the first group of subjects had List 1 for the baseline memory task and List 2 for the sleep-related memory task. For the second group of subjects the order was reversed.


On the morning preceeding the sleep recording night, subjects received the baseline memory task. Given that episodic memory is commonly impaired in the elderly ( Guttentag 1985; Hartley 1989; Light 1991), the measure of baseline recall allowed us to control for initial differences in memory ability in our subjects. The experimenter presented the list of word pairs. Each pair was printed on a separate card and was presented for 2 s. Instructions informed the subject to read and study each pair of words for a subsequent cued-recall test. The experimenter read each pair of words aloud, and the subject read them silently. A 10-min interference task followed the list presentation, to reduce any recency effect on recall. The task required the subject to orally answer a set of questions. Cued recall was oral: the experimenter read the first word of each pair and the subject tried to recall the second word of the pair.

The list for the sleep-related memory task was presented in the evening of the same day, when subjects were in bed and ready for sleep recording. The experimenter presented the list of word pairs not used for the baseline memory task. The procedure was the same as for the baseline memory task, except for the absence of an interfering task and the timing of cued recall. Instructions asked subjects to study the words and informed them of the memory test to be held in the morning, immediately after awakening. Word pairs were presented at a 2-s rate, and were read aloud by the experimenter and silently by the subject. The oral cued recall test took place in the morning immediately after awakening. Subjects were presented with the first word of each pair and were asked to try to recall the target word. Time to recall each word was paced by the subject. No precaution was used to prevent rehearsal before falling asleep.

Sleep was recorded following the standard procedures on a polygraph Era 14–21. Records were visually scored every 20 s according to the Rechtschaffen & Kales rules. However, the minimum length of each sleep state and stage was 2 min, following a procedure previously used by our group ( Bes et al. 1991 ). Similarly, a period of wake was counted as an awakening when it exceeded 2 min. Slow wave sleep was scored by eliminating the amplitude criterion, as suggested by Webb and Dreblow (1982).

A NREM/REM cycle was defined as a sequence of NREM and REM sleep not interrupted by a waking period longer than 2 min. To define a NREM/REM sequence, both NREM and REM periods had to be longer than 2 min. REM epochs shorter than 2 min were included in the previous sleep state. Therefore only sequences of NREM sleep followed by a REM period longer than 2 min were computed as NREM/REM cycles. Similarly, a sequence of NREM stages interrupted by a period of wake longer than 2 min was not considered part of a NREM/REM cycle.


Mean recall in the baseline memory task was 5.97 items (sd=2.91); mean recall of the sleep-related memory task was 2.63 (sd=3.24).

Average length of NREM/REM cycles was 110 min. This and other sleep parameters are reported in Table 1. The relationship between sleep and cognitive performances was examined by means of multiple parametric correlations (Pearson's).

Table 1.  Descriptive statistics of sleep parameters and their correlation with morning recall Thumbnail image of

Baseline memory was considered a measure of the subject's basic memory ability, and, as such, was partialled out from the correlations between sleep-related memory and sleep parameters. Partialization was automatically performed by the multiple correlation program.

As shown in Table 1, sleep-related memory performance was significantly and positively correlated with two parameters of sleep, i.e. average duration of NREM/REM cycles and proportion of total TCT over TST. No significant correlations were found between memory and either the number of awakenings, amount of wake, percentage of REM sleep or other sleep parameters, including sleep latency.

In addition, performance at the two memory tests was significantly and positively correlated (r=0.77, P< 0.01) and sleep-related memory performance was significantly correlated with WAIS verbal (r=0.39, P< 0.05).


In this experiment we examined the relationship between sleep parameters and performance in an episodic memory task in elderly adults.

Firstly, we have a methodological remark. The percentage of SWS obtained in this study is higher than the percentage found in other studies with elderly subjects. However, such a percentage could be due to both the criterion of analysis (see Method) or to the low amount of TST observed in our subjects. If we transform in minutes the percentage of SWS obtained in other studies ( Feinberg et al. 1967 ), we obtain a quite similar amount of SWS: an average of 91 min in Feinberg et al. vs. an average of 94 min in our study.

We did not find any significant correlation between recall and number of awakenings in our group of elderly subjects. This result could suggest that continuity of sleep processes does not seem to be an indispensable condition for retention of words in elderly subjects. This result is in agreement with previous findings by Bonnet (1989), despite differences between the two procedures (experimental manipulation vs. spontaneous awakenings; young vs. old subjects).

The absence of a significant correlation between percentages of REM and NREM sleep and after-sleep recall is in agreement with previous results obtained in younger adults after either sleep deprivation ( Empson and Clarke 1970), or when memory performance was compared across sleep periods with different quantities of REM and NREM ( Smith 1985).

The correlation between SWS and morning recall is close enough to be significant and negative. This may suggest that SWS can have a detrimental role on memory, as already discussed in previous studies on sleep functions ( Cai 1995) and on dream recall ( Feinberg 1974; Foulkes et al. 1993 ).

Our data show that a better sleep-related recall is significantly associated with longer NREM/REM cycles and with a greater proportion of TCT over TST. This result suggests that the functional NREM/REM unit might have an important role in sleep-related memory processes.

The observed relationship between memory and NREM/REM cycles has two possible explanations. One explanation is that the relationship between memory and cycles reflects the role of an iterative mechanism ( Cipolli 1995; Cipolli et al. 1987 ), where during each sleep cycle information is accessed and processed.

A second explanation goes beyond the role of single sleep states, and takes into account the role of sleep organization as reflected by NREM/REM cycles. The influence of sleep organization on memory is in agreement with the hypothesis that connects sleep cycles, protein synthesis and consolidation of material in memory during sleep.

The two hypotheses, however, are not mutually exclusive. On the contrary, they can be usefully integrated, so that during each cycle the presence of both NREM and REM sleep facilitates protein synthesis, and at the same time consolidation in memory increases with the amount of time spent in NREM/REM cycles over TST.

However, there is an additional aspect to the results that deserves to be mentioned. Since according to our definition of NREM/REM cycle there was no REM period outside of cycles, the significant correlation between sleep-related recall and time spent in cycles could stand also for the time spent in REM sleep. In this case, the correlation would suggest that REM per se is important for retention during sleep. This would be in agreement with previous data supporting the role of REM for memory consolidation ( De Konink et al. 1989; Rotenberg 1992; Smith and Lapp 1991; Hennevin et al. 1995 ). Our data, however, showed no correlation between REM percentage and recall, while we found a significant correlation between recall and TCT over TST. This result underscores the role of the NREM/REM functional unit on intrasleep memory processes, and provides support to previous suggestions about the importance of good sleep organization (as expressed by NREM/REM cycles), in addition to REM, for memory consolidation in humans and animals ( Buszaki 1989; Giuditta et al. 1984; Hars and Hennevin 1987; Karni 1995; Lairy and Salzarulo 1975; Salzarulo 1995; Smith and Lapp 1991; Wilson and McNaughton 1994 ).


In conclusion, this study showed in elderly adults the importance of sleep structure for intrasleep memory processes. The effect is currently limited to episodic memory of verbal material. The possibility to extend this result to other types of material or to other memory systems (e.g. semantic memory) requires additional study.


This research was supported by an Agenzia Spaziale Italiana (ASI) Grant No. 94–RS–162 and by the Comitato Nazionale per la Ricerca (CNR).

Accepted in revised form 27 January 1999; received 8 June 1998