Effect of infant sleeping position on sleep spindles


: Dr Rosemary S. C Horne, Department of Paediatrics, Monash Medical Centre, 246 Clayton Road, Clayton, Victoria, Australia 3168. Tel.: +61 39594 4504; fax: +61 39594 6259; e-mail: rosemary.horne@med.monash.edu.au


Sleep spindles play an active role in inducing and maintaining sleep and may affect arousal by blocking the transmission of external stimuli through the thalamus to the cortex. Previously we have demonstrated that sleeping in the prone position impairs arousal in infants at 2–3 months of age, but not at 5–6 months. We aimed to examine if sleeping position and postnatal age affected duration and/or density of sleep spindles. Twenty-one healthy term infants were studied using daytime polysomnography at 2–3 months and 16 were again studied at 5–6 months. Infants slept both prone and supine at each study. The mean duration of non-rapid eye movement (NREM) sleep was not different between the two studies in either position. At 2–3 months both spindle density (P < 0.001) and proportion of NREM sleep (P < 0.025) with spindles were significantly greater in the supine than in the prone position. At 5–6 months spindle duration was longer in the supine than in the prone position (P < 0.03). Spindle density in the supine position was not different between the two studies, however, when infants slept prone, it was significantly increased at 5–6 months compared with 2–3 months (P < 0.001). Arousal threshold was not correlated with either spindle density or percentage of NREM sleep with spindles in either position at either study. In this study spindle density and the percentage time spent with spindles were not well correlated with infant arousability, and hence may not be able to be used as markers of depressed arousal responses in infants.


It has been proposed that sleep spindles play an active role in inducing and maintaining sleep (Yamadori 1971), and may serve to block transmission of external stimuli through the thalamus to the cortex (Steriade and Amzica 1998). In animal studies, it has been demonstrated that spontaneous arousal and localized stimulation of the brainstem arousal centres are associated with sleep-spindle termination (Munk et al. 1996; Steriade et al. 1996; Steriade and Amzica 1998). Additionally, in infants, sighs and startles are also associated with spindle suppression (Wulbrand et al. 1998). The association between arousal defects and sudden infant death syndrome (SIDS) has led to the hypothesis that a failure to arouse from sleep may be the final pathway for SIDS (Hunt 1992). Sleep spindles emerge between 4 and 9 weeks postnatally, and spindle activity is maximal between 3 and 6 months of age (Ellingson 1982; Hen-Shin et al. 1980; Hughes 1996; Louis et al. 1992; Metcalf 1969; Shibagaki et al. 1982; Tanguay et al. 1975), which corresponds with the peak incidence for SIDS (Hoffman and Hillman 1992). It has been postulated that spindles may serve as a marker of central nervous system maturation (Schultz et al. 1968), which may be impaired in SIDS infants (Hunt 1992).

The prone sleeping position has been identified as one of the major risk factor for SIDS in numerous epidemiological studies carried out in western countries (Brooke et al. 1997; Mitchell 1993; Mitchell et al. 1997; Oyen et al. 1997; Ponsonby and Dwyer 1995; Taylor et al. 1996). Several studies have now demonstrated that arousability of infants, to a variety of external stimuli, is depressed when they are placed prone to sleep in rapid eye movement sleep (REM sleep) (Franco et al. 1998; Galland et al. 1998) and in both REM and non-rapid eye movement sleep (NREM sleep) (Horne et al. 2001). Our previous study demonstrated that this depression was most marked at 2–3 months of age and sleeping position did not affect arousability at 5–6 months (Horne et al. 2001). However, the mechanisms by which arousability could be altered are unknown. The aim of this study was therefore to examine the duration and density (frequency of occurrence) of sleep spindles in infants sleeping prone and supine. We hypothesized that sleep spindles would be increased in the prone position and this would be most marked at 2–3 months of age.



Ethical approval for this project was granted by the Monash Medical Centre Human Ethics Committee. All infants were recruited from the maternity wards and Jessie MacPherson Private Hospital, Monash Medical Centre, Melbourne, Victoria, Australia. No monetary incentive was provided to mothers participating in the study, and none of the mothers used illegal drugs. Written informed consent was obtained from parent(s) prior to commencement of the study.

Infants in this study were recruited as part of a previously reported larger study to assess the effects of sleeping position on arousal from sleep (Horne et al. 2001). Twenty-one infants (14F/7M) of the original 24 had electroencephalograph (EEG) signals suitable for spindle analysis at 2–3 months of age and 16 infants (F/M) were studied again at 5–6 months of age. All infants were born at term (range 38–42 week gestation), with normal birth weights mean 3590 ± 77 g (range 2940–4080 g) and Apgar scores averaged 9 and 9 at 1 and 5 min, respectively. The mean age of the infants at the 2–3 month-study was 74 ± 1 days (range 63–83 days) and at the 5–6 month 181 ± 4 days (range 155–213 days).

Recording methods

All infants were studied using polysomnography between 10:00 and 16:00 h, at the Melbourne Children's Sleep Unit, Monash Medical Centre. Electrodes for recording of physiological variables were attached to the baby while it fed and, when drowsy, the infant was placed in a bassinet under dim lighting and constant room temperature (22–23°C). The study did not begin until the infant was in a stable sleep state. Infants generally had both a morning and afternoon sleep interrupted by a midday feed, when sleep position was changed. Each infant slept in both prone and supine sleeping positions at both studies. The initial sleep position was randomized for the morning of the first study, and the opposite position used for the morning of the second study to minimize any time of day effects. None of the infants studied routinely slept prone.

Recordings were made using a polygraph recorder (Grass Model 78 A, Grass Instrument Company, Quincey, MA, USA) of EEG, electrooculogram (EOG), submental electromyogram (EMG), electrocardiogram (ECG), instantaneous heart rate, thoracic and abdominal breathing movements (Resp-ez Piezo-electric sensor; EPM Systems, Midlothian, VA, USA) and blood oxygen saturation (Biox 3700e Pulse Oximeter, Ohmeda, Louisville, CO, USA). Sleep-state was assessed as either NREM or REM sleep using EEG, behavioural, heart rate and breathing pattern criteria (Guilleminault & Souquet 1979). Two EEG signals were recorded at a sampling rate of 500 Hz, one with a high-pass filter (HPF) set at 0.3 Hz and low-pass filter (LPF) set at 30 Hz, and the second with a HPF set at 10 Hz and a LPF set at 30 Hz. The filtering of the second EEG channel aided identification and measurement of spindle duration by removing slow high-amplitude background waves which are a feature of the infant EEG. Signals were also recorded onto a computerized sleep analysis system (S-series Sleep-System V5.2; Compumedics, Melbourne, Victoria, Australia). Spindle frequency and duration were measured visually on the computer screen at 20 s per page. Spindles were defined as having a frequency of between 11 and 15 Hz, but most commonly were 12–14 Hz and duration of ≥0.5 s.

Stimulus and arousal criteria

Infant arousability was determined using an intermittent randomly presented pulsatile air-jet stimulus (frequency 3 Hz for 5 s) delivered to the nostrils of the infant. Arousal thresholds were calculated as described previously (Horne et al. 2001). Briefly, the stimulus was presented alternately to the left and right nostrils; if the infant failed to arouse, the air-jet pressure was increased when the stimulus was again presented to that nostril. Whenever an arousal response occurred the pressure was then decreased. Arousal threshold was calculated as the mean value between each arousing and non-arousing stimulus. Subcortical arousal was defined as a change in at least three of the following criteria: a change in ventilation pattern of more than two breaths, an observed behavioural response, a heart rate acceleration of greater than 10% above baseline, and an increase in submental EMG activity (Horne et al. 2001). For each physiological variable the 10 s of recording immediately preceding the stimulus presentation provided the baseline level used to assess the change in that variable.

Data analysis

Mean values for spindle duration, spindle density (the mean number of spindles per minute of NREM sleep), per cent NREM sleep time in which spindles occurred (total duration of NREM sleep for each infant divided by total length of time that spindles occurred) and arousal threshold were calculated for each infant in each position and both ages. Data were firstly checked for normality and then were compared between sleeping positions at each age with paired Student's t-test. The effects of postnatal age were compared with two-way analysis of variance for repeated measures (n=16). Individual infant arousal thresholds were compared with individual infant spindle density and duration and per cent of NREM sleep with spindles at each study with regression analysis. All values are expressed as mean ± SEM and a P-value of <0.05 was considered significant.


The mean duration of NREM analysed was not different between the two studies in either position (Table 1). Similar amounts of NREM sleep were recorded in the morning compared with the afternoon at both studies (2–3 months: morning 557 min; afternoon 800 min; 5–6 months: morning 429 min; afternoon 409 min). The mean number of spindles in the supine position was also not different between the two studies, however, there were significantly more spindles at 5–6 months than at 2–3 months when infants slept prone (P < 0.001) (Table 1).

Table 1.  Summary of the effects of sleeping position on sleep spindle characteristics and arousal threshold at 2–3 months and 5–6 months. Values are mean ± SEM with n=number of infants studied
2–3 months (n=21)
 Total NREM sleep duration (min)623734 
 Total number of spindles analysed1361899 
 Mean NREM sleep duration (s)1780 ± 2032097 ± 173NS
 Mean spindle length (s)1.63 ± 0.11.71 ± 0.1NS
 Mean spindle density (spindles min−1 NREM sleep)2.40 ± 0.31.37 ± 0.3P < 0.001
 NREM sleep with spindles (%)6.50 ± 1.14.81 ± 1.11P < 0.025
 Arousal threshold (cm H2O)274 ± 46430 ± 45P < 0.01
5–6 months (n=16)
 Total NREM sleep duration (min)357481 
 Total number of spindles analysed10071443 
 Mean NREM sleep duration (s)1338 ± 811805 ± 178P < 0.01
 Mean spindle length (s)2.10 ± 0.141.95 ± 0.12P < 0.035
 Mean spindle density (spindles min−1 NREM sleep)2.91 ± 0.362.98 ± 0.33NS
 NREM sleep with spindles (%)10.93 ± 1.6910.59 ± 1.52NS
 Arousal threshold (cm H2O)509 ± 79592 ± 62NS

Effects of sleeping position

At 2–3 months, there was no difference in the duration of NREM sleep between the two sleeping positions. Spindle density was significantly higher when infants slept supine (P < 0.001) compared with prone, and likewise the percentage NREM sleep time in which spindles occurred was longer (P < 0.025) (Table 1).

At 5–6 months, infants spent more time in NREM sleep when they slept prone than supine (P < 0.01). Spindle duration was shorter in the prone than in the supine position (P < 0.03). However, there was no difference between sleeping positions in spindle density or the percentage of NREM sleep with spindles occurring (Table 1).

Effects of postnatal age

When data for the 16 infants who completed both studies were compared, spindle length was not different in either position. In the supine position mean spindle density was not different between studies, however, in the prone position, spindle density increased at 5–6 months (P < 0.001). In both supine (P < 0.02) and prone (P < 0.005) positions the percentage of NREM sleep with spindles was greater at 5–6 months than at 2–3 months (Fig. 1).

Figure 1.

Effects of postnatal age in individual infants (n=16) on percentage of non-rapid eye movement sleep with spindles in the prone and supine sleeping positions.

Correlation with arousal threshold

Arousal thresholds were significantly higher in the prone position at 2–3 months of age (P < 0.01), however, at 5–6 months there was no difference (Table 1). Arousal thresholds were also significantly elevated at 5–6 months compared with 2–3 months in both the supine (P < 0.01) and prone positions (P < 0.05). When mean spindle density and length and the percent of NREM sleep with spindles were compared with mean arousal threshold in individual infants there was not significant correlation in either sleep position at either age.


This study has demonstrated that, although spindle density and duration are altered by sleeping position in young infants, this was not well correlated with the depressed arousability observed previously in 2–3-month-old infants sleeping prone. We had hypothesized that sleep spindle density would be increased when infants slept prone at 2–3 months. However, our data showed that at 2–3 months, spindle density and the percentage of NREM sleep with spindles was significantly greater when infants slept supine than when they slept prone. If spindles do indeed play an active role in maintaining sleep, then the increased spindle density in the supine position may act to promote sleep in this position. It is known that infants have more spontaneous arousals from sleep in this position (Kahn et al. 1993) as well as being more arousable to external stimuli (Horne et al. 2001). We had previously demonstrated that at 5–6 months, there was no difference in arousability between sleeping positions (Horne et al. 2001). In this study, we also showed that sleeping position had no effect of on spindle density or percentage of NREM sleep with spindles at 5–6 months.

Previous studies have reported that spindle duration and density are maximal at between 3 and 6 months of age and decrease or remain unchanged with increasing postnatal age (Hughes 1996; Louis et al. 1992; Metcalf 1969), however, sleeping position was not reported in these studies. The mean duration and density of spindles in our study are similar to values reported by others at 3 and 6 months (Louis et al. 1992; Tanguay et al. 1975). The only other longitudinal study in infants (Louis et al. 1992) showed that spindle density increased from 1.5 to 3 months of age but remained unchanged until 6 months. Similarly, the longest spindles were observed at 1.5 and 3 months of age, however, there was no statistically significant difference across the four ages studied. Hughes (1996) reported a much longer spindle duration of 6 s at 3 months; this study, however, was not carried out longitudinally and the data were reported for one infant only. As reported previously there is a wide individual variation between infants in both density and duration of spindles (Louis et al. 1992; Yamadori 1971) and this was also found in our study.

There are several limitations to our study. During the study, repeated arousal stimuli were presented to the infants, which may have affected our results (Horne et al. 2001). However, the number of stimuli presented was not different between the two sleep positions. In addition, it has been demonstrated that presentation of episodic stimuli had no effect on either spindle length or spindle density in infants (Tanguay et al. 1975; Yamadori 1971). Our studies were carried out during the day and this may have had an effect on spindles. It had previously been reported that the periodicity of sleep spindles was genetically determined (Hori et al. 1989), and that although there is a large variation between subjects, there was little variation between the same individual studied on different nights (Shirakawa et al. 1980; Yamadori 1971). However, recent data in adults has shown that low-frequency (12.25–13.0 Hz) spindle activity is altered by circadian phase (Dijk 1999). Our studies were all carried out during the same part of the day with each infant following its normal feeding and sleeping routine and thus this variation may not have had a significant effect on results. Sleep spindles have a distinctive frequency of 10–16 Hz, but most frequently are 12–14 Hz, and are the only EEG pattern, which remains at the same frequency throughout life (Hughes 1996; Shibagaki et al. 1982; Tanguay et al. 1975). It has been proposed that they represent a marker of neural maturation (Louis et al. 1992). Our study group were all healthy infants born at term and were studied at similar postnatal ages to minimize maturational differences between infants.

Despite the limitations of this study, we have demonstrated that both spindle duration and the percentage of time spent with spindles are not well correlated in this study with arousability. Other factors, such as autonomic control of cardiorespiratory function, may be involved in the depressed arousal observed in infants sleeping in the prone position.


We wish to thank the parents and infants who gave up their time to assist us with this project. Also, we wish to acknowledge the assistance of Professor Richard Harding, Department of Physiology, Monash University, Melbourne and Dr Lilia Curzi-Dascalova INSERM, Service de Physiologie, Hôpital Robert-Debré, Paris for commenting on earlier drafts of this manuscript. This project was supported by grants from SIDSaustralia, Sudden Infant Death Research Foundation (South Australia) and SIDassist.