- Email: [email protected]
Accessing previous mental sleep experience in REM and NREM sleep
Biofogicaf Psychoiogy 29 (1989) 27-38 North-Holland
ACCESSING AND NREM IGINO
CARLO
PREVIOLJS SLEEP *
27
MENTAL
SLEEP
EXPERIENCE
IN REM
FAGIOLI
CIPOLLI
**
instirute of Human Physiology, Uniuemity of Modem, Modem, rta& GIOVANNI Depnrtment Accepted
TUOZZI
of Psychalogv, Universify of f3aiogna. Bologna, Italy for pubiicatinn
27 February
1989
This study inv~tigated the processes by which contents previously stored in memory are retrieved and inserted into mental sleep experience (MSE). MSE reports were collected from six subjects awakened three times on each of eight nights in two alternate sequences of awakenings (NREM-REM-NR~M; REM-REM-REM}. The occurrences of interrelations between contents of report pairs were scored using Clark’s (1970) feature matching model. These were greater for same night pairs than for different night pairs, and did not differ with respect to sequence of awakenings or order of report pairs (first-second, second-third, first-third}. Contents of previous MSEs, therefore, seem to be accessible in both sleep types for insertion into current MSE. The interrelated units were more frequently lexical than propositional, with more paradigmatic than syntagmatic relationships in report pairs from both sequences of awakenings. Thus, the re-etaboradon of contents of previous MSEs seems to occur mainly at the level of single contents in both types of sleep, with similar modalities of processing.
Since the earliest psychophy~iolo~cal studies of sleep and dreaming, it has been claimed that the capacity for information processing differs in the two main types of sleep, namely REM and NREM sleep (Emmons & Simon, 1956). This view seems to have been supported on the one hand by studies of * This research was carried out within the Department of Psychology, University of Bologna, Italy, and was supported by a grant from C.N.R. (Italy) No, X7.0161334. The authors are indebted to Guy Aston for comments and assistance in preparation of the manuscript. ** Requests for reprints should be sent to Carlo Cipotli, Institute of Human Physiology, University of Modena, Via Campi 287, 41100 Modena, Italy.
0301-0511/89/$3.50
0 1989, Usevier
Science Publishers
B.V. (North-Holland)
28
I. Fagdr
et al. / Accessing previous mentul sleep experience
retention of stimuli externally delivered before (e.g., Fowler, Sullivan, & Ekstrand, 1973) or during sleep (Shimizu, Takehashi. Sumitsuji, Tanaka. Yoshida, & Kaneko, 1977). and, on the other hand, by studies of the production of mental sleep experience (MSE). The latter have shown that the recall of MSE is more frequent after awakening in REM than NREM sleep and that REM reports have a more organized structure than NREM reports (Foulkes, 1962; Foulkes & Schmidt, 1983). The overall difference in capacity for information processing in REM and NREM sleep might be accounted for by certain or all of the processes operating in the retention of external stimuli and in MSE production. The first possibility seems to be more parsimonious and more easily controlled in an experimental situation. The well-known phenomenon of the thematic continuity of MSEs provides a potentially useful context for such experimentation. There is substantial evidence that contents of MSEs recalled after multiple awakenings are more frequently interrelated (i.e, share some or all of their features) in paired reports of the same night than in those paired across different nights. This thematic continuity is not an artefact of the experimental procedure of multiple awakenings and reportings, but steadily characterizes MSE production. In recent studies, in which we believe some of the methodological biases inherent in previous studies (Dement & Wolpert, 1958; Rechtschaffen, Vogel, & Shaikun, 1963; Trosman, Rechtschaffen, Offenkrantz, & Wolpert, 1960) have been overcome, we have found such continuity to be present both for sequences of reports given after awakenings in REM sleep (Cipolli, Baroncini, Fagioli, Fumai, & Salzarulo, 1987) and for sequences of NREM and REM reports (Cipolli, Fagioli, Baroncini, Fumai, Marchio’, & Sancini, 1988). The phenomenon of thematic continuity immediately poses the question of the organization of the processes involved in MSE production. Since MSE production implies several processes operating at various levels, such as accessing information stored in memory (the so-called mnestic sources), MSE planning, and insertion of single contents into the frame of the ongoing MSE, these processes have to be sketched in multilevel systems (for one such system, see Foulkes, 1982). We might posit different systems guiding those processes for each sleep type, or a single system operating in more or less distinct manners in the two sleep types. Some support for the latter hypothesis is provided by our two previous studies on sequences of REM and of NREM and REM report pairs. Stronger support might be provided by comparing sequences of REM and of NREM and REM report pairs collected from the same subjects. If (a) MSE contents previously elaborated in NREM and REM sleep are both potentially accessible subsequently, and (b) interrelated units and relationships between contents of paired reports have similar characteristics in both types of sleep, this would show that elaboration of the contents of previous MSE for insertion into
ongoing MSE takes place at the same level in both sleep types. This level is the local one of single contents insertion rather than the more global one of narrative structure (Foulkcs, 1982). The present experiment was designed to gather further insight as to the processes by which the contents of previous MSE are accessed and daborated for insertion into current MSE in REM and NREM sleep on a more frequently than chance basis (as estimated by comparing report pairs across the same night with those acxoss different nights}. We aimed to assess whether the extent of thematic ~ntinuity differs in the two main types of sleep, by comparing report pairs collected in nights where awakenings were provoked only in REM sleep with those cdlected after alternate REM and NREM awakenings, testing rhe hy~oth~s~~ that access to contents of previous MSE should be different in the two main types of sleep. This hypothesis seems plausible inasmuch as it has been ascertained that capacity for processing i~forrnatio~~ is different For externally delivered stimuli such as flashes (varying from I to 30) and names (Shim&u et al., 19?“?), being lower in each case in NREM than in REM sleep. In the light of these fi~~dings, it might be expected that accessibility will afsa differ according to sleep type for internally delivered stimuli. Besides day residues, abstract knowledge, and recent or remote waking events, such stimuli inctude contents of previous MSEs, whose long-term storage has been ascertained in several studies for both sleep types fCipalli, Calasso, Maccohni, Pani, & Salzarulo, 1984; Sdzarulo & Cipdli, 15379). Were the assailability of previous IVISE contents found not to differ, we would have to consider the hypothesis that processing during sleep differs between internally and externally delivered information, with memory functioning being similar in NREM and REM sleep for the former, In the terms of our experiment greater thematic continuity across reports from REM sleep awakenings than across REM and NREM reparts would thus support the ~~otbesis that access to internally delivered i~f~r~~~tion differs according to sleep type, whereas a similar degree of continuity in both sequence-types wouId suggest that memory functioning is similar in both sieep types at least for this kind of material.
Six male subjects, Italian native-speakers, aged 20 to 26 years, were recorded at weekly intervals for eight nights and awakened each night three times af3er 5 min of uninterrupted sleep in the stage where the awakening was planned, The sequence and the order of awakenings were counterbalanced across subjects and nights, combining the criteria of (a) the sequence of awakenings
30 Order of the provoked awakeninEs 1st
REM sequence
/
REM
2nd
1-1
REM
3rd
I_
REH
/
NREM - REM - NREM
Fig. 1. Schedule
of awakenings.
(alternately REM and NREM-REM), and (b) the four possible combinations of stages in NREM sleep (II and III) awakenings before and after REM sleep. The schedule of the awakenings is represented in fig. 1. Subjects had previously experienced awakening and the spontaneous reporting in an adaptation night. Immediately after each awakening subjects were asked “What was going through your mind before awakening?“. At the end of the report they were allowed to go back to sleep. 2.2. Apparatus Electroencephalogram (EEG: six channels), electrooculogram (EOG: vertical and horizontal leads), chin electromyogram (EMG) were recorded on a Battaglia-Rangoni polygraph (paper speed 15 mm/s, gain 50 EL-V/cm, time constant 0.3 s (EEG and EOG), 0.1 s (EMG)). Placement of electrodes and scoring of sleep stages followed Rechtschaffen and Kales’s (1968) criteria. 2.3. Scoring Contentful reports (i.e., reports composed of at least one sentence describing the contents of MSE: Cohen, 1972) were provided after 140 of the 144 awakenings performed (97.22%). To analyze the interrelations between reports, we matched the contents of all pairings of contentful reports for each type of awakening sequence both of the same night and of different nights. For the latter match, reports were paired across the four nights of each type of awakening sequence according to a forward-oriented procedure (see fig. 2). Two trained judges scored the reports, unaware of their order and night of origin. Interrelated contents in pairs of reports were identified: only where there was agreement between judges were then retained. These relationships
I. Fagioli et al. / Accessrngprevious
mental sleep experience
31
(n = 264; 108 between report pairs of the same night for the sequence REM-REM-REM and 83 for that NREM-REM-NREM; 43 between report pairs of different nights for the sequence REM-REM-REM and 30 for that NREM-REM-NREM) were then classified: (a) according to the type of linguistic unit involved, as lexical (single words or nominal/verbal groups denoting an object, character or action) or propositional (clauses or sentences). Where the interrelated contents occurred in different types of unit in the two reports, the interrelation was scored as propositional; (b) according to the type of relationship, as paradigmatic (with substitution of linguistic units of the same grammatical class) or syntugmutic (combination of linguistic units of different grammatical classes), using a slightly modified version of Clark’s (1970) model: for a complete description, see Cipolli, Fagioli, Maccolini, and Salzarulo (1983) and, for some examples, see table 1. Interjudge agreement in scoring units as lexical/propositional was 97.38%, and in classifying interrelations as paradigmatic/syntagmatic was 95.29% for report pairs of the same nights and respectively 95.89% and 94.52% for report pairs of different nights. Cases of disagreement were resolved through discussion between the scorers. The proportion of interrelated report pairs per subject and the frequency of interrelated units per pair of reports were computed with respect to the total number of pairs of contentful reports (i.e., 97.22%). These scoring and classification operations yielded two indicators, namely the proportion of pairs with interrelated reports and the frequency of interrelated units per report pair. The factors considered for the first indicator were: (a) the night for report pairings (same vs different night); (b) the sequence of awakenings (NREM-REM-NREM vs REM-REM-REM);
32
I. Fagioh et al. / Accessing previous mrntrrl sleep experience
Table 1 Examples
of relationships
between
Clark’s rules
contents
of reports
pairs Examples
Paradigmatic
Minimal
contrast
(a) identity (b) antonymy (c) synonymy
train-train east-west my cat-Felix
Marking
dog-bitch
Feature Addition/Deletion
vegetable-cabbage grandmother-old mountain-fountain
Category
supply-replenish
Maintenance
woman
Syntagmatic
Selectional Feature-Realization
film-script
Idiom-Completion
garter-knight
(c)
the ordering of the reports paired (lst-2nd vs 2nd-3rd addition to those above, two further factors considered
indicator were: (d) the type of interrelated units (lexical/propositional); (e) the type of relationships (paradigmatic/syntagmatic)
vs lst-3rd). In for the second
between
these units.
The proportion of report pairs with interrelated reports was transformed in arc-sine and the frequency of interrelations per report pair was transformed in log (x + 1) to reduce within group variance before computing ANOVAs (Myers, 1979). The transformed data were then tested for homogeneity: distribution was not significantly different from normal. Confirmation of our prediction of a similar accessibility
of previous
MSE
contents in NREM and REM sleep implied acceptance of the null hypothesis. Given the prejudice against accepting null hypothesis, for each nonsignificant relevant finding a power test was planned to estimate the probability (1-p) of rejecting of false null hypothesis (Myers, 1979). To compare occurrences of interrelations and frequencies of interrelated units the alternative hypothesis was defined as any difference with respect to the null one, since we had no criteria to establish the minimal relevant difference.
33
I. Fagioli et al. / Accessrng preurous mental sleep experience
3. Results 3.1. Proportion
of report pairs with interrelations
The proportions of report pairs with interrelated contents are shown in table 2. A three-way ANOVA with repeated measures showed that the proportions of report pairs with interrelated contents were significantly higher for same night pairs than for different night pairs (F(1,5) = 89.115, p < O.OOl), whereas they did not differ significantly for the two sequences of awakenings (see fig. 1) (F(1,5) < 1; 1-p > .95), nor for the three types of report pairs (F(2,lO) = 2.252, n.s., 1-p > .99). No possible interaction between two or three of the run factors reached statistical significance: “Night of report pairings” X “sequence of awakenings” (F(1,5) < 1; 1-p > .99), “night of report pairings”~ “order of reports paired” (F(2,lO) < 1; 1-p > .75), “sequence of awakenings” X “order of reports paired” (F(2,lO) = 1.334, n.s.; 1-p > .25), “night of report pairings” x “sequence of awakenings” X “order of reports paired” (F(2,lO) < 1; 1-p > .99). 3.2. Frequency
of interrelations
per report pair
The frequencies of interrelated units per report pair are shown in table 3. A five-way ANOVA showed that interrelated contents were significantly more frequent in report pairs from the same night than in pairs from different nights (F(1,5) = 73.130; p < .Ol), whereas they were not significantly different in the two sequences of awakenings (F(1,5) < 1, n.s.; 1-p > .99), nor in differently ordered report pairs (F(2,lO) = 3.537; n.s.; 1-p > .99). Interrelated lexical units were significantly more frequent than propositional ones (F(1,5) = 83.450, p < .OOl) and paradigmatic relationships were significantly more frequent than syntagmatic ones (F(1,5) = 29.790, p < .Ol). The interaction between the factors “night of report pairing” and “type of interrelated unit” was statistically significant (F(1,5) = 10.510, p < .05); no other possible inter-
Table 2 Proportions
of interrelations
between
contents
of report
pairs
Rl-R2
R2-R3
RlLR3
Same night NREM-REM-NREM REM-REM-REM
0.87 f 0.21 0.79 f 0.25
0.79 + 0.19 0.92kO.13
0.58 k 0.41 0.75 k 0.42
Different nights NREM-REM-NREM REM-REM-REM
0.46kO.17 0.46 f 0.25
0.35 + 0.16 0.54 i_ 0.29
0.29 k 0.25 0.38 k 0.30
I. Fagid
34 Table
et al. / Accessing prevrous menfal sleep experwnce
3
Frequencies
of interrelated
units
Lexcial Par
Propositional Synt
Par
Synt
NREM-REM-NREM Same night Rl-R2
0.65 k 0.30
0.49 + 0.26
0.18f0.14
0.17kO.26
R2-R3
0.87 & 0.77
0.29 + 0.33
0.22 + 0.11
0.04 * 0.10
Rl-R3
0.33 & 0.44
0.31 kO.19
0.29 k 0.40
0
Different
nights
Rl-R2
0.35 + 0.25
0.20 f 0.11
0.04+0.06
0.02 * 0.05
R2-R3
0.26kO.14
0.08 i 0.10
0.07 + 0.08
0
Rl-R3
0.15 + 0.27
0.09 i 0.07
0.07 f 0.11
0
REM-REM-REM Same night Rl-R2
0.50 f 0.35
0.29 k 0.37
0.29 + 0.45
0.08 + 0.20
R2-R3
1.00+0.61
0.42 + 0.41
0.42 k 0.41
0.13+0.21
RI-R3
0.67 c 0.41
0.33 * 0.26
0.29 f 0.19
0
Rl-R2
0.21 t 0.24
0.33 + 0.20
0
0
R2-R3
0.33 & 0.20
0.13*0.13
0.25 + 0.27
0.08 f 0.13
RI-R3
0.29 k 0.33
0.08+0.13
0.04+0.10
0.04~0.10
Different
nights
action between two or more of the run factors reached (with a power value ranging from > .30 to > .99). 3.3. Conservative
measures
statistical
significance
of interrelations
These results were then controlled for a possible parasitic source of variance. Interrelated contents in pairs of 1st NREM-2nd NREM reports may also occur in the intervening REM report, thus forming a chain of three interrelated reports. Interrelations between the NREM reports in such a chain may, therefore, be an artefact of successive interrelations. A post-hoc analysis was carried out excluding such chainings; the proportions and frequencies of interrelations between contents of report pairs are shown in table 4. Proportions of interrelated report pairs were again significantly more frequent in the same night than in different nights (F(1,5) = 67.974; p < .OOl), whereas they were not significantly different with respect to the factor “sequence of awakenings” (F(1,5) < 1, n.s.: 1-p > .95), or to “order of report pairs” (F(2,lO) = 2.076, n.s., 1-p > .99). No possible interaction between two or three of the run factors reached significance (with a power value ranging from > .20 to > .99). A three-way ANOVA showed that frequencies of interrelated units across reports were significantly higher in pairs of the same night than in those of
I. Fagioli et al. / Accessing preuious mental sleep experience
35
Table 4 Proportions artefacts
of interrelations
PROPORTIONS Same night NREM-REM-NREM REM-REM-REM Different nights NREM-REM-NREM REM-REM-REM FREQUENCIES Same nights NREM-REM-NREM REM-REM-REM Different nights NREM-REM-NREM REM-REM-REM
and
frequencies
of interrelated
units
excluding
cases of possible
Rl-R2
W-R3
RlLR3
0.87 f 0.21 0.79 * 0.25
0.78 k 0.19 0.86kO.15
0.53 + 0.36 0.75 + 0.42
0.44*0.15 0.46 rt 0.25
0.35 i 0.16 0.54 * 0.29
0.29 + 0.25 0.38 k 0.30
0.34 f 0.09 0.27 k 0.14
0.33 kO.17 0.45 f0.17
0.20+0.15 0.28 + 0.15
0.15 + 0.07 0.13 f 0.07
0.11 kO.05 0.20+0.11
0.08 k 0.07 0.11~0.11
different nights (F(1,5) = 75.758; awakenings (F(1,5) = 1.826, n.s., pairs (F(2,lO) = 3.684, n.s.; 1-p > three of the run factors reached ranging from > .40 to > .99).
p < .OOl), but not in different sequences of 1-p > .99), or in differently ordered report .90). No possible interaction between two or statistical significance (with power values
4. Discussion The main results of the present experiment seem to be more compatible with the hypothesis of a similar availability of the contents of previous MSEs in both types of sleep than with one of differing availability in REM and NREM sleep. In fact, proportions of interrelations in report pairs were as expected higher in report pairs from the same night than in those from different nights, and overall were quite similar in both sequences of awakenings: although the power test did not reach a high value, the occurrences in NREM-REM pairs were more frequent than those in NREM-NREM pairs, against the expected direction of the difference between these values: see table 2. The two indicators of modalities of information recruitment and elaboration, namely the type of unit and the type of associative relationships (Cipolli et al., 1987), also had similar values in both cases. The first indicates that this information is in each case mainly at the lexical level (single words or phrases). The latter, which showed a prevalence of paradigmatic relationships between the contents of paired reports, suggests that information from previous MSE(s)
36
I. Fagioli et 01. / Accessing prevmxs mental sleep experrence
is more likely to be inserted because of its coherence with the ongoing MSE than as a means of guiding thematic progression through its relationships with other information nodes in the narrative. This means that the matching mechanism by which contents of previous MSEs are elaborated for insertion as elements of the current MSE can be described in terms of the procedure operating to instantiate single contents which Foulkes (1982) has called “lexical look-up”. These data also speak in favour of the hypothesis that memory functioning is similar in the two main types of sleep, at any rate for the lower levels of the system of MSE production. In evaluating the findings provided by our experiment, the possibility that access to MSE contents is partly due to a contingent waking activation of those linguistic structures used by the subject in describing the contents of MSEs of a particular night cannot be completely excluded, given that any analysis of MSE contents presupposes verbal reporting. However, a contingent use of certain words and clauses in reporting would not modify the amount of interrelations scored, given that these words and clauses should be denotative of the MSE contents and, thus, synonymous with others that might be employed: interrelations between reports of the same night would be in any case scored. Moreover, a contingently greater accessibility of certain words or clauses might be expected to influence the proportions of interrelated contents encoded in lexical and propositional units respectively, and of the proportions of types of associative relationships in a given night. The only interaction found. that between “night of report pairings” and “type of unit”, seems too weak to support this hypothesis. In fact, it is additive with respect to the two main effects shown (namely, the superiority of the pairs of reports in the same night with respect to different ones, and of lexical with respect to propositional units). How are we to reconcile the evidence of similar functioning in accessing and elaborating contents of previous MSE with that provided by the studies on the processing of externally-delivered stimuli cited in the introduction? It seems useful to distinguish, first of all, evidence involving stimuli delivered during sleep from evidence involving stimuli delivered before sleep. It is mainly for the former that higher processing capacity in REM than NREM sleep has been shown (Shimizu et al., 1977). For the latter, although some studies support the superiority of REM sleep in processing externally delivered information, others do not show significant differences between REM and NREM for simple types of information (such as single words and simple sentences: Cipolli & Salzarulo, 1979; Empson & Clarke, 1970). Given our finding that the interrelations in paired reports more frequently concern the simpler (lexical) than the more complex (propositional) units in both REM and NREM-REM paired reports it is thus less striking that these contents should be re-processed in a similar way in both sleep types.
1. Fagioli et al. / Accessing previous
mental sleep experience
31
The contrast between our finding and those of studies on the processing of stimuli delivered externally during sleep may be tentatively explained in terms of the temporal characteristics of their processing. As shown by Shimizu et al. (1977) the effectiveness of REM sleep processing of during-sleep stimuli is higher than that of NREM sleep until 15 min after the stimulus presentation, after which they are similar. Given the intervals between awakenings in the sequences scheduled in our experiment, we cannot suppose that such an eventual contingent superiority of REM sleep has been operative. It seems, therefore, more economical to suppose that the availability of internally-delivered during sleep stimuli (such as contents of previous MSEs) parallels that of pre-sleep stimuli with similar characteristics of organization (single words and simple sentences). The results of our experiment may also be considered from the perspective of the state-dependency of the processes of MSE production and recall. The essence of this view (see Koukkou & Lehmann, 1983) is that the availability of information varies with respect to the type of sleep in which MSE is either produced (dependency in production) or retrieved (dependency in recall). Our findings go against the predictions of this view as far as concerns dependency in recall, at least for one specific type of mnestic source, namely the contents of previous MSE. They do not imply, however, that there are no differences across sleep stages in the functioning of any of the processes involved in MSE production. Comparison of REM reports collected after the first awakening with following ones suggests higher-level differences in MSE planning (Cipolli & Poli, submitted). An asymmetrical functioning of some higher-level processes (that is, stage-specific constraints) in MSE production would be a plausible source of variability in recall frequency and in length and organization of MSE reports (not only between reports from different stages of sleep, but also within the same type or stage of sleep).
References Cipolli, C., Baroncini, P., Fagioli, I., Fumai, A., & Salzarulo, P. (1987). The thematic continuity of mental sleep experience in the same night. Sleep, 10, 473-479. Cipolli, C., Calasso, E., Maccolini, S. Pani R., & Salzarulo, P. (1984). Memory processes in morning recall after multiple night awakenings. Perceptual and Motor Skills, 59, 435-446. Cipolli, C., Fagioli, I., Baroncini, P., Fumai, A., Marchio’, B., & Sancini, M. (1988). The thematic continuity of mental experiences in’REM and NREM sleep. International Journal of Psychophysiology, 6, 307-313. Cipolli, C., Fagioli, I., Maccolini, S., & Salzarulo, P. (1983). Associative relationships between pre-sleep sentence stimuli and reports of mental sleep experience. Perceptual and Motor Skills, 56. 223-234. Cipolli, C., & Poli, D. (submitted). Story structure in verbal reports of mental sleep experience after awakening in REM sleep.
38
I. Fagwli ef al. / Acce.wing preuious
mental sleep experrence
Cipolli, C., & Salzarulo, P. (1979). Sentence memory and sleep: A pilot study. Sleep, 2, 1933198. Clark, H.H. (1970). Word associations and linguistic theory. In J. Lyons (Ed.), New horizons m iinguistics (pp. 271-286). Harmondsworth: Penguin Books. Cohen, D.B. (1972). Failure to recall dream content: Contentless versus dreamless reports. Perceptual and Motor Ski& 34, lMK)- 1002. Dement. W.C., & Wolpert, E. (1958). Relationships in the manifest content of dreams occurring on the same night. Journal of Nervous and Menrrrl Diseuse, 126, 568-578. Emmons, W.H., & Simon, C.W. (1956). The non-recall of material presented during sleep. American Journal oi Psychology, 69, 76-81. Empson, J., & Clarke, P. (1970). Rapid eye movements and remembering. Nurure, 227, 287-288. Foulkes, D. (1962). Dream report from different stages of sleep. Journal of Abnormai and Socrui Psychology, 65, 14-25. Foulkes, D. (1982). A cognitive-psychological model of REM dream production. Sleep, 5, 1699187. Foulkes, D., & Schmidt. M. (1983). Temporal sequence and unit composition in dream reports from different stages of sleep. Sleep, 6, 265-280. Fowler, M.J.. Sullivan. M.J., & Ekstrand, B.R. (1973). Sleep and memory. Science, 179, 3022304. Koukkou, M., & Lehmann, D. (1983). Dreaming: The functional state-shift hypothesis. 3~ztIsh Journul of Psychic&y, 142. 221-231. Myers, J.L. (1979). Fundamentals of experimentul desrgn. Boston: Allyn & Bacon. Rechtschaffen, A., & Kales, A. (1968). A manual of stundurdized termmolo~~, techniques and scoring system for sleep stages of human subjects. Los Angeles: Brain Information Service. Rechtschaffen, A., Vogel, G., & Shaikun, G. (1963). Interrelatedness of mental activity during sleep. Archives of General Psychiatry, 9. 536-547. Salzarulo, P.. & Cipolti, C. (1979). Linguistic organization and cognitive implications in REM and NREM sleep related reports. Perceptual and Motor Skills, 49, 767-777. Shimizu, A., Takehashi, H., SumitsuJi, N., Tanaka, M., Yoshida, I.. & Kaneko, Z. (1977). Memory retention of stimulations during REM and NREM stages of sleep. Elecrroencephalography and Chnical Neurophysiology, 43, 658-665. Trosman, H., Rechtschaffen, A., Offenkrantz, W., & Wolpert, E. (1960). Studies in psychophysiology of dreams: IV. Relations among dreams in sequence. Archiues of General Psychiatv, 3, 602-607.
ACCESSING AND NREM IGINO
CARLO
PREVIOLJS SLEEP *
27
MENTAL
SLEEP
EXPERIENCE
IN REM
FAGIOLI
CIPOLLI
**
instirute of Human Physiology, Uniuemity of Modem, Modem, rta& GIOVANNI Depnrtment Accepted
TUOZZI
of Psychalogv, Universify of f3aiogna. Bologna, Italy for pubiicatinn
27 February
1989
This study inv~tigated the processes by which contents previously stored in memory are retrieved and inserted into mental sleep experience (MSE). MSE reports were collected from six subjects awakened three times on each of eight nights in two alternate sequences of awakenings (NREM-REM-NR~M; REM-REM-REM}. The occurrences of interrelations between contents of report pairs were scored using Clark’s (1970) feature matching model. These were greater for same night pairs than for different night pairs, and did not differ with respect to sequence of awakenings or order of report pairs (first-second, second-third, first-third}. Contents of previous MSEs, therefore, seem to be accessible in both sleep types for insertion into current MSE. The interrelated units were more frequently lexical than propositional, with more paradigmatic than syntagmatic relationships in report pairs from both sequences of awakenings. Thus, the re-etaboradon of contents of previous MSEs seems to occur mainly at the level of single contents in both types of sleep, with similar modalities of processing.
Since the earliest psychophy~iolo~cal studies of sleep and dreaming, it has been claimed that the capacity for information processing differs in the two main types of sleep, namely REM and NREM sleep (Emmons & Simon, 1956). This view seems to have been supported on the one hand by studies of * This research was carried out within the Department of Psychology, University of Bologna, Italy, and was supported by a grant from C.N.R. (Italy) No, X7.0161334. The authors are indebted to Guy Aston for comments and assistance in preparation of the manuscript. ** Requests for reprints should be sent to Carlo Cipotli, Institute of Human Physiology, University of Modena, Via Campi 287, 41100 Modena, Italy.
0301-0511/89/$3.50
0 1989, Usevier
Science Publishers
B.V. (North-Holland)
28
I. Fagdr
et al. / Accessing previous mentul sleep experience
retention of stimuli externally delivered before (e.g., Fowler, Sullivan, & Ekstrand, 1973) or during sleep (Shimizu, Takehashi. Sumitsuji, Tanaka. Yoshida, & Kaneko, 1977). and, on the other hand, by studies of the production of mental sleep experience (MSE). The latter have shown that the recall of MSE is more frequent after awakening in REM than NREM sleep and that REM reports have a more organized structure than NREM reports (Foulkes, 1962; Foulkes & Schmidt, 1983). The overall difference in capacity for information processing in REM and NREM sleep might be accounted for by certain or all of the processes operating in the retention of external stimuli and in MSE production. The first possibility seems to be more parsimonious and more easily controlled in an experimental situation. The well-known phenomenon of the thematic continuity of MSEs provides a potentially useful context for such experimentation. There is substantial evidence that contents of MSEs recalled after multiple awakenings are more frequently interrelated (i.e, share some or all of their features) in paired reports of the same night than in those paired across different nights. This thematic continuity is not an artefact of the experimental procedure of multiple awakenings and reportings, but steadily characterizes MSE production. In recent studies, in which we believe some of the methodological biases inherent in previous studies (Dement & Wolpert, 1958; Rechtschaffen, Vogel, & Shaikun, 1963; Trosman, Rechtschaffen, Offenkrantz, & Wolpert, 1960) have been overcome, we have found such continuity to be present both for sequences of reports given after awakenings in REM sleep (Cipolli, Baroncini, Fagioli, Fumai, & Salzarulo, 1987) and for sequences of NREM and REM reports (Cipolli, Fagioli, Baroncini, Fumai, Marchio’, & Sancini, 1988). The phenomenon of thematic continuity immediately poses the question of the organization of the processes involved in MSE production. Since MSE production implies several processes operating at various levels, such as accessing information stored in memory (the so-called mnestic sources), MSE planning, and insertion of single contents into the frame of the ongoing MSE, these processes have to be sketched in multilevel systems (for one such system, see Foulkes, 1982). We might posit different systems guiding those processes for each sleep type, or a single system operating in more or less distinct manners in the two sleep types. Some support for the latter hypothesis is provided by our two previous studies on sequences of REM and of NREM and REM report pairs. Stronger support might be provided by comparing sequences of REM and of NREM and REM report pairs collected from the same subjects. If (a) MSE contents previously elaborated in NREM and REM sleep are both potentially accessible subsequently, and (b) interrelated units and relationships between contents of paired reports have similar characteristics in both types of sleep, this would show that elaboration of the contents of previous MSE for insertion into
ongoing MSE takes place at the same level in both sleep types. This level is the local one of single contents insertion rather than the more global one of narrative structure (Foulkcs, 1982). The present experiment was designed to gather further insight as to the processes by which the contents of previous MSE are accessed and daborated for insertion into current MSE in REM and NREM sleep on a more frequently than chance basis (as estimated by comparing report pairs across the same night with those acxoss different nights}. We aimed to assess whether the extent of thematic ~ntinuity differs in the two main types of sleep, by comparing report pairs collected in nights where awakenings were provoked only in REM sleep with those cdlected after alternate REM and NREM awakenings, testing rhe hy~oth~s~~ that access to contents of previous MSE should be different in the two main types of sleep. This hypothesis seems plausible inasmuch as it has been ascertained that capacity for processing i~forrnatio~~ is different For externally delivered stimuli such as flashes (varying from I to 30) and names (Shim&u et al., 19?“?), being lower in each case in NREM than in REM sleep. In the light of these fi~~dings, it might be expected that accessibility will afsa differ according to sleep type for internally delivered stimuli. Besides day residues, abstract knowledge, and recent or remote waking events, such stimuli inctude contents of previous MSEs, whose long-term storage has been ascertained in several studies for both sleep types fCipalli, Calasso, Maccohni, Pani, & Salzarulo, 1984; Sdzarulo & Cipdli, 15379). Were the assailability of previous IVISE contents found not to differ, we would have to consider the hypothesis that processing during sleep differs between internally and externally delivered information, with memory functioning being similar in NREM and REM sleep for the former, In the terms of our experiment greater thematic continuity across reports from REM sleep awakenings than across REM and NREM reparts would thus support the ~~otbesis that access to internally delivered i~f~r~~~tion differs according to sleep type, whereas a similar degree of continuity in both sequence-types wouId suggest that memory functioning is similar in both sieep types at least for this kind of material.
Six male subjects, Italian native-speakers, aged 20 to 26 years, were recorded at weekly intervals for eight nights and awakened each night three times af3er 5 min of uninterrupted sleep in the stage where the awakening was planned, The sequence and the order of awakenings were counterbalanced across subjects and nights, combining the criteria of (a) the sequence of awakenings
30 Order of the provoked awakeninEs 1st
REM sequence
/
REM
2nd
1-1
REM
3rd
I_
REH
/
NREM - REM - NREM
Fig. 1. Schedule
of awakenings.
(alternately REM and NREM-REM), and (b) the four possible combinations of stages in NREM sleep (II and III) awakenings before and after REM sleep. The schedule of the awakenings is represented in fig. 1. Subjects had previously experienced awakening and the spontaneous reporting in an adaptation night. Immediately after each awakening subjects were asked “What was going through your mind before awakening?“. At the end of the report they were allowed to go back to sleep. 2.2. Apparatus Electroencephalogram (EEG: six channels), electrooculogram (EOG: vertical and horizontal leads), chin electromyogram (EMG) were recorded on a Battaglia-Rangoni polygraph (paper speed 15 mm/s, gain 50 EL-V/cm, time constant 0.3 s (EEG and EOG), 0.1 s (EMG)). Placement of electrodes and scoring of sleep stages followed Rechtschaffen and Kales’s (1968) criteria. 2.3. Scoring Contentful reports (i.e., reports composed of at least one sentence describing the contents of MSE: Cohen, 1972) were provided after 140 of the 144 awakenings performed (97.22%). To analyze the interrelations between reports, we matched the contents of all pairings of contentful reports for each type of awakening sequence both of the same night and of different nights. For the latter match, reports were paired across the four nights of each type of awakening sequence according to a forward-oriented procedure (see fig. 2). Two trained judges scored the reports, unaware of their order and night of origin. Interrelated contents in pairs of reports were identified: only where there was agreement between judges were then retained. These relationships
I. Fagioli et al. / Accessrngprevious
mental sleep experience
31
(n = 264; 108 between report pairs of the same night for the sequence REM-REM-REM and 83 for that NREM-REM-NREM; 43 between report pairs of different nights for the sequence REM-REM-REM and 30 for that NREM-REM-NREM) were then classified: (a) according to the type of linguistic unit involved, as lexical (single words or nominal/verbal groups denoting an object, character or action) or propositional (clauses or sentences). Where the interrelated contents occurred in different types of unit in the two reports, the interrelation was scored as propositional; (b) according to the type of relationship, as paradigmatic (with substitution of linguistic units of the same grammatical class) or syntugmutic (combination of linguistic units of different grammatical classes), using a slightly modified version of Clark’s (1970) model: for a complete description, see Cipolli, Fagioli, Maccolini, and Salzarulo (1983) and, for some examples, see table 1. Interjudge agreement in scoring units as lexical/propositional was 97.38%, and in classifying interrelations as paradigmatic/syntagmatic was 95.29% for report pairs of the same nights and respectively 95.89% and 94.52% for report pairs of different nights. Cases of disagreement were resolved through discussion between the scorers. The proportion of interrelated report pairs per subject and the frequency of interrelated units per pair of reports were computed with respect to the total number of pairs of contentful reports (i.e., 97.22%). These scoring and classification operations yielded two indicators, namely the proportion of pairs with interrelated reports and the frequency of interrelated units per report pair. The factors considered for the first indicator were: (a) the night for report pairings (same vs different night); (b) the sequence of awakenings (NREM-REM-NREM vs REM-REM-REM);
32
I. Fagioh et al. / Accessing previous mrntrrl sleep experience
Table 1 Examples
of relationships
between
Clark’s rules
contents
of reports
pairs Examples
Paradigmatic
Minimal
contrast
(a) identity (b) antonymy (c) synonymy
train-train east-west my cat-Felix
Marking
dog-bitch
Feature Addition/Deletion
vegetable-cabbage grandmother-old mountain-fountain
Category
supply-replenish
Maintenance
woman
Syntagmatic
Selectional Feature-Realization
film-script
Idiom-Completion
garter-knight
(c)
the ordering of the reports paired (lst-2nd vs 2nd-3rd addition to those above, two further factors considered
indicator were: (d) the type of interrelated units (lexical/propositional); (e) the type of relationships (paradigmatic/syntagmatic)
vs lst-3rd). In for the second
between
these units.
The proportion of report pairs with interrelated reports was transformed in arc-sine and the frequency of interrelations per report pair was transformed in log (x + 1) to reduce within group variance before computing ANOVAs (Myers, 1979). The transformed data were then tested for homogeneity: distribution was not significantly different from normal. Confirmation of our prediction of a similar accessibility
of previous
MSE
contents in NREM and REM sleep implied acceptance of the null hypothesis. Given the prejudice against accepting null hypothesis, for each nonsignificant relevant finding a power test was planned to estimate the probability (1-p) of rejecting of false null hypothesis (Myers, 1979). To compare occurrences of interrelations and frequencies of interrelated units the alternative hypothesis was defined as any difference with respect to the null one, since we had no criteria to establish the minimal relevant difference.
33
I. Fagioli et al. / Accessrng preurous mental sleep experience
3. Results 3.1. Proportion
of report pairs with interrelations
The proportions of report pairs with interrelated contents are shown in table 2. A three-way ANOVA with repeated measures showed that the proportions of report pairs with interrelated contents were significantly higher for same night pairs than for different night pairs (F(1,5) = 89.115, p < O.OOl), whereas they did not differ significantly for the two sequences of awakenings (see fig. 1) (F(1,5) < 1; 1-p > .95), nor for the three types of report pairs (F(2,lO) = 2.252, n.s., 1-p > .99). No possible interaction between two or three of the run factors reached statistical significance: “Night of report pairings” X “sequence of awakenings” (F(1,5) < 1; 1-p > .99), “night of report pairings”~ “order of reports paired” (F(2,lO) < 1; 1-p > .75), “sequence of awakenings” X “order of reports paired” (F(2,lO) = 1.334, n.s.; 1-p > .25), “night of report pairings” x “sequence of awakenings” X “order of reports paired” (F(2,lO) < 1; 1-p > .99). 3.2. Frequency
of interrelations
per report pair
The frequencies of interrelated units per report pair are shown in table 3. A five-way ANOVA showed that interrelated contents were significantly more frequent in report pairs from the same night than in pairs from different nights (F(1,5) = 73.130; p < .Ol), whereas they were not significantly different in the two sequences of awakenings (F(1,5) < 1, n.s.; 1-p > .99), nor in differently ordered report pairs (F(2,lO) = 3.537; n.s.; 1-p > .99). Interrelated lexical units were significantly more frequent than propositional ones (F(1,5) = 83.450, p < .OOl) and paradigmatic relationships were significantly more frequent than syntagmatic ones (F(1,5) = 29.790, p < .Ol). The interaction between the factors “night of report pairing” and “type of interrelated unit” was statistically significant (F(1,5) = 10.510, p < .05); no other possible inter-
Table 2 Proportions
of interrelations
between
contents
of report
pairs
Rl-R2
R2-R3
RlLR3
Same night NREM-REM-NREM REM-REM-REM
0.87 f 0.21 0.79 f 0.25
0.79 + 0.19 0.92kO.13
0.58 k 0.41 0.75 k 0.42
Different nights NREM-REM-NREM REM-REM-REM
0.46kO.17 0.46 f 0.25
0.35 + 0.16 0.54 i_ 0.29
0.29 k 0.25 0.38 k 0.30
I. Fagid
34 Table
et al. / Accessing prevrous menfal sleep experwnce
3
Frequencies
of interrelated
units
Lexcial Par
Propositional Synt
Par
Synt
NREM-REM-NREM Same night Rl-R2
0.65 k 0.30
0.49 + 0.26
0.18f0.14
0.17kO.26
R2-R3
0.87 & 0.77
0.29 + 0.33
0.22 + 0.11
0.04 * 0.10
Rl-R3
0.33 & 0.44
0.31 kO.19
0.29 k 0.40
0
Different
nights
Rl-R2
0.35 + 0.25
0.20 f 0.11
0.04+0.06
0.02 * 0.05
R2-R3
0.26kO.14
0.08 i 0.10
0.07 + 0.08
0
Rl-R3
0.15 + 0.27
0.09 i 0.07
0.07 f 0.11
0
REM-REM-REM Same night Rl-R2
0.50 f 0.35
0.29 k 0.37
0.29 + 0.45
0.08 + 0.20
R2-R3
1.00+0.61
0.42 + 0.41
0.42 k 0.41
0.13+0.21
RI-R3
0.67 c 0.41
0.33 * 0.26
0.29 f 0.19
0
Rl-R2
0.21 t 0.24
0.33 + 0.20
0
0
R2-R3
0.33 & 0.20
0.13*0.13
0.25 + 0.27
0.08 f 0.13
RI-R3
0.29 k 0.33
0.08+0.13
0.04+0.10
0.04~0.10
Different
nights
action between two or more of the run factors reached (with a power value ranging from > .30 to > .99). 3.3. Conservative
measures
statistical
significance
of interrelations
These results were then controlled for a possible parasitic source of variance. Interrelated contents in pairs of 1st NREM-2nd NREM reports may also occur in the intervening REM report, thus forming a chain of three interrelated reports. Interrelations between the NREM reports in such a chain may, therefore, be an artefact of successive interrelations. A post-hoc analysis was carried out excluding such chainings; the proportions and frequencies of interrelations between contents of report pairs are shown in table 4. Proportions of interrelated report pairs were again significantly more frequent in the same night than in different nights (F(1,5) = 67.974; p < .OOl), whereas they were not significantly different with respect to the factor “sequence of awakenings” (F(1,5) < 1, n.s.: 1-p > .95), or to “order of report pairs” (F(2,lO) = 2.076, n.s., 1-p > .99). No possible interaction between two or three of the run factors reached significance (with a power value ranging from > .20 to > .99). A three-way ANOVA showed that frequencies of interrelated units across reports were significantly higher in pairs of the same night than in those of
I. Fagioli et al. / Accessing preuious mental sleep experience
35
Table 4 Proportions artefacts
of interrelations
PROPORTIONS Same night NREM-REM-NREM REM-REM-REM Different nights NREM-REM-NREM REM-REM-REM FREQUENCIES Same nights NREM-REM-NREM REM-REM-REM Different nights NREM-REM-NREM REM-REM-REM
and
frequencies
of interrelated
units
excluding
cases of possible
Rl-R2
W-R3
RlLR3
0.87 f 0.21 0.79 * 0.25
0.78 k 0.19 0.86kO.15
0.53 + 0.36 0.75 + 0.42
0.44*0.15 0.46 rt 0.25
0.35 i 0.16 0.54 * 0.29
0.29 + 0.25 0.38 k 0.30
0.34 f 0.09 0.27 k 0.14
0.33 kO.17 0.45 f0.17
0.20+0.15 0.28 + 0.15
0.15 + 0.07 0.13 f 0.07
0.11 kO.05 0.20+0.11
0.08 k 0.07 0.11~0.11
different nights (F(1,5) = 75.758; awakenings (F(1,5) = 1.826, n.s., pairs (F(2,lO) = 3.684, n.s.; 1-p > three of the run factors reached ranging from > .40 to > .99).
p < .OOl), but not in different sequences of 1-p > .99), or in differently ordered report .90). No possible interaction between two or statistical significance (with power values
4. Discussion The main results of the present experiment seem to be more compatible with the hypothesis of a similar availability of the contents of previous MSEs in both types of sleep than with one of differing availability in REM and NREM sleep. In fact, proportions of interrelations in report pairs were as expected higher in report pairs from the same night than in those from different nights, and overall were quite similar in both sequences of awakenings: although the power test did not reach a high value, the occurrences in NREM-REM pairs were more frequent than those in NREM-NREM pairs, against the expected direction of the difference between these values: see table 2. The two indicators of modalities of information recruitment and elaboration, namely the type of unit and the type of associative relationships (Cipolli et al., 1987), also had similar values in both cases. The first indicates that this information is in each case mainly at the lexical level (single words or phrases). The latter, which showed a prevalence of paradigmatic relationships between the contents of paired reports, suggests that information from previous MSE(s)
36
I. Fagioli et 01. / Accessing prevmxs mental sleep experrence
is more likely to be inserted because of its coherence with the ongoing MSE than as a means of guiding thematic progression through its relationships with other information nodes in the narrative. This means that the matching mechanism by which contents of previous MSEs are elaborated for insertion as elements of the current MSE can be described in terms of the procedure operating to instantiate single contents which Foulkes (1982) has called “lexical look-up”. These data also speak in favour of the hypothesis that memory functioning is similar in the two main types of sleep, at any rate for the lower levels of the system of MSE production. In evaluating the findings provided by our experiment, the possibility that access to MSE contents is partly due to a contingent waking activation of those linguistic structures used by the subject in describing the contents of MSEs of a particular night cannot be completely excluded, given that any analysis of MSE contents presupposes verbal reporting. However, a contingent use of certain words and clauses in reporting would not modify the amount of interrelations scored, given that these words and clauses should be denotative of the MSE contents and, thus, synonymous with others that might be employed: interrelations between reports of the same night would be in any case scored. Moreover, a contingently greater accessibility of certain words or clauses might be expected to influence the proportions of interrelated contents encoded in lexical and propositional units respectively, and of the proportions of types of associative relationships in a given night. The only interaction found. that between “night of report pairings” and “type of unit”, seems too weak to support this hypothesis. In fact, it is additive with respect to the two main effects shown (namely, the superiority of the pairs of reports in the same night with respect to different ones, and of lexical with respect to propositional units). How are we to reconcile the evidence of similar functioning in accessing and elaborating contents of previous MSE with that provided by the studies on the processing of externally-delivered stimuli cited in the introduction? It seems useful to distinguish, first of all, evidence involving stimuli delivered during sleep from evidence involving stimuli delivered before sleep. It is mainly for the former that higher processing capacity in REM than NREM sleep has been shown (Shimizu et al., 1977). For the latter, although some studies support the superiority of REM sleep in processing externally delivered information, others do not show significant differences between REM and NREM for simple types of information (such as single words and simple sentences: Cipolli & Salzarulo, 1979; Empson & Clarke, 1970). Given our finding that the interrelations in paired reports more frequently concern the simpler (lexical) than the more complex (propositional) units in both REM and NREM-REM paired reports it is thus less striking that these contents should be re-processed in a similar way in both sleep types.
1. Fagioli et al. / Accessing previous
mental sleep experience
31
The contrast between our finding and those of studies on the processing of stimuli delivered externally during sleep may be tentatively explained in terms of the temporal characteristics of their processing. As shown by Shimizu et al. (1977) the effectiveness of REM sleep processing of during-sleep stimuli is higher than that of NREM sleep until 15 min after the stimulus presentation, after which they are similar. Given the intervals between awakenings in the sequences scheduled in our experiment, we cannot suppose that such an eventual contingent superiority of REM sleep has been operative. It seems, therefore, more economical to suppose that the availability of internally-delivered during sleep stimuli (such as contents of previous MSEs) parallels that of pre-sleep stimuli with similar characteristics of organization (single words and simple sentences). The results of our experiment may also be considered from the perspective of the state-dependency of the processes of MSE production and recall. The essence of this view (see Koukkou & Lehmann, 1983) is that the availability of information varies with respect to the type of sleep in which MSE is either produced (dependency in production) or retrieved (dependency in recall). Our findings go against the predictions of this view as far as concerns dependency in recall, at least for one specific type of mnestic source, namely the contents of previous MSE. They do not imply, however, that there are no differences across sleep stages in the functioning of any of the processes involved in MSE production. Comparison of REM reports collected after the first awakening with following ones suggests higher-level differences in MSE planning (Cipolli & Poli, submitted). An asymmetrical functioning of some higher-level processes (that is, stage-specific constraints) in MSE production would be a plausible source of variability in recall frequency and in length and organization of MSE reports (not only between reports from different stages of sleep, but also within the same type or stage of sleep).
References Cipolli, C., Baroncini, P., Fagioli, I., Fumai, A., & Salzarulo, P. (1987). The thematic continuity of mental sleep experience in the same night. Sleep, 10, 473-479. Cipolli, C., Calasso, E., Maccolini, S. Pani R., & Salzarulo, P. (1984). Memory processes in morning recall after multiple night awakenings. Perceptual and Motor Skills, 59, 435-446. Cipolli, C., Fagioli, I., Baroncini, P., Fumai, A., Marchio’, B., & Sancini, M. (1988). The thematic continuity of mental experiences in’REM and NREM sleep. International Journal of Psychophysiology, 6, 307-313. Cipolli, C., Fagioli, I., Maccolini, S., & Salzarulo, P. (1983). Associative relationships between pre-sleep sentence stimuli and reports of mental sleep experience. Perceptual and Motor Skills, 56. 223-234. Cipolli, C., & Poli, D. (submitted). Story structure in verbal reports of mental sleep experience after awakening in REM sleep.
38
I. Fagwli ef al. / Acce.wing preuious
mental sleep experrence
Cipolli, C., & Salzarulo, P. (1979). Sentence memory and sleep: A pilot study. Sleep, 2, 1933198. Clark, H.H. (1970). Word associations and linguistic theory. In J. Lyons (Ed.), New horizons m iinguistics (pp. 271-286). Harmondsworth: Penguin Books. Cohen, D.B. (1972). Failure to recall dream content: Contentless versus dreamless reports. Perceptual and Motor Ski& 34, lMK)- 1002. Dement. W.C., & Wolpert, E. (1958). Relationships in the manifest content of dreams occurring on the same night. Journal of Nervous and Menrrrl Diseuse, 126, 568-578. Emmons, W.H., & Simon, C.W. (1956). The non-recall of material presented during sleep. American Journal oi Psychology, 69, 76-81. Empson, J., & Clarke, P. (1970). Rapid eye movements and remembering. Nurure, 227, 287-288. Foulkes, D. (1962). Dream report from different stages of sleep. Journal of Abnormai and Socrui Psychology, 65, 14-25. Foulkes, D. (1982). A cognitive-psychological model of REM dream production. Sleep, 5, 1699187. Foulkes, D., & Schmidt. M. (1983). Temporal sequence and unit composition in dream reports from different stages of sleep. Sleep, 6, 265-280. Fowler, M.J.. Sullivan. M.J., & Ekstrand, B.R. (1973). Sleep and memory. Science, 179, 3022304. Koukkou, M., & Lehmann, D. (1983). Dreaming: The functional state-shift hypothesis. 3~ztIsh Journul of Psychic&y, 142. 221-231. Myers, J.L. (1979). Fundamentals of experimentul desrgn. Boston: Allyn & Bacon. Rechtschaffen, A., & Kales, A. (1968). A manual of stundurdized termmolo~~, techniques and scoring system for sleep stages of human subjects. Los Angeles: Brain Information Service. Rechtschaffen, A., Vogel, G., & Shaikun, G. (1963). Interrelatedness of mental activity during sleep. Archives of General Psychiatry, 9. 536-547. Salzarulo, P.. & Cipolti, C. (1979). Linguistic organization and cognitive implications in REM and NREM sleep related reports. Perceptual and Motor Skills, 49, 767-777. Shimizu, A., Takehashi, H., SumitsuJi, N., Tanaka, M., Yoshida, I.. & Kaneko, Z. (1977). Memory retention of stimulations during REM and NREM stages of sleep. Elecrroencephalography and Chnical Neurophysiology, 43, 658-665. Trosman, H., Rechtschaffen, A., Offenkrantz, W., & Wolpert, E. (1960). Studies in psychophysiology of dreams: IV. Relations among dreams in sequence. Archiues of General Psychiatv, 3, 602-607.