No mechanism known to explain positive observed effect of 2.45 GHz Wi‐Fi exposure on sleep‐dependent memory consolidation

Effects of 2.45 GHz Wi‐Fi exposure on sleep‐dependent memory consolidation. Ana Bueno‐Lopez  Torsten Eggert  Hans Dorn  Gernot Schmid  Rene Hirtl  Heidi Danker‐Hopfe. Journal of Sleep Research, November 9 2020. https://doi.org/10.1111/jsr.13224

Rolf Degen's take: https://twitter.com/DegenRolf/status/1326395159362342913

Abstract: Studies have reported that exposure to radiofrequency electromagnetic fields (RF‐EMF) emitted by mobile telephony might affect specific sleep features. Possible effects of RF‐EMF emitted by Wi‐Fi networks on sleep‐dependent memory consolidation processes have not been investigated so far. The present study explored the impact of an all‐night Wi‐Fi (2.45 GHz) exposure on sleep‐dependent memory consolidation and its associated physiological correlates. Thirty young males (mean ± standard deviation [SD]: 24.1 ± 2.9 years) participated in this double‐blind, randomized, sham‐controlled crossover study. Participants spent five nights in the laboratory. The first night was an adaptation/screening night. The second and fourth nights were baseline nights, each followed consecutively by an experimental night with either Wi‐Fi (maximum: psSAR10g = <25 mW/kg; 6 min average: <6.4 mW/kg) or sham exposure. Declarative, emotional and procedural memory performances were measured using a word pair, a sequential finger tapping and a face recognition task, respectively. Furthermore, learning‐associated brain activity parameters (power spectra for slow oscillations and in the spindle frequency range) were analysed. Although emotional and procedural memory were not affected by RF‐EMF exposure, overnight improvement in the declarative task was significantly better in the Wi‐Fi condition. However, none of the post‐learning sleep‐specific parameters was affected by exposure. Thus, the significant effect of Wi‐Fi exposure on declarative memory observed at the behavioural level was not supported by results at the physiological level. Due to these inconsistencies, this result could also be a random finding.

4 DISCUSSION

The present provocation study, which can only address acute effects, analysed whether a Wi‐Fi exposure during TIB (8 h) might affect sleep‐dependent memory consolidation processes (declarative, procedural and emotional memory) and their learning‐associated brain activity during sleep in young healthy male volunteers.

4.1 Sleep‐dependent memory consolidation: Behavioural level

Results show that although Wi‐Fi did not affect retention in the procedural and emotional memory tasks, the data reveal that retention in the declarative memory was increased after Wi‐Fi as compared to sham exposure.

In the WPT, overnight performance gain was higher after Wi‐Fi exposure compared to sham (see Figure 3a.i), with an effect size of 0.40. According to Cohen (1988) this is a small effect, which, however, has a large uncertainty (95% CI [0.11; 0.70]). This observed difference in overnight retention of correctly recalled word pairs between sham and Wi‐Fi exposure conditions represented moderate evidence for the alternative hypothesis when evaluated based on the corresponding Bayes factor (BF01 = 0.254) (see Table 2). However, the interaction of several factors needs to be taken into account in order to interpret this result accurately. Small differences in the number of correctly recalled word pairs during immediate recall might have affected performance gains in the WPT. That is, the number of correctly recalled word pairs in the evening was slightly, but not significantly, higher in the sham nights as compared to Wi‐Fi, whereas the opposite was observed in the morning (see Table 2). The lower “reference level” in the evening preceding the Wi‐Fi condition might explain why overnight change was significantly higher under Wi‐Fi compared to sham. On the other hand, as both versions of the WPT had the same level of difficulty, it is unlikely that encoding difficulties could explain this finding. Regardless of the exposure condition, the performance on the evening of the two experimental nights did not differ, which supports the absence of a learning effect between experimental nights (see Table S5). Moreover, the data did not reflect the presence of floor or ceiling effects.

Wi‐Fi exposure did not affect performance in the FRT. Overnight retention was similar between Wi‐Fi and sham exposure. Bayes factors showed that overnight retention in all categories presented moderate evidence for the absence of a decline or improvement after exposure (all faces: BF01 = 3.931; neutral faces: BF01 = 4.538; positive faces: BF01 = 3.155; negative faces: BF01 = 5.527) with effect sizes (Cohen's d) that vary from no (negative faces) to small effects (all, neutral and positive faces; see Table 2). Thus, recognition memory in the emotional task did not differ between exposure conditions.

Performance improvements in the SFTT after sleep were not affected by Wi‐Fi exposure. The results for the overnight retention in this memory task did not differ between exposure conditions. Moreover, retention in this task showed moderate evidence for the null hypothesis (BF01 = 4.539), which is supported by a small effect size (see Table 2). In contrast, Lustenberger et al. (2013) reported a reduction of the performance improvement, measured as the variance of the reaction time, in a similar SFTT under RF‐EMF exposure compared to sham (with an effect size of |d| = 0.57 representing a medium effect; effect size calculated from data presented in Lustenberger et al., 2013). This effect could not be confirmed by our results. The variance in reaction time performance in the present study did not differ significantly between the exposure conditions, the effect size indicates no effect (|d| = 0.13) and the Bayes factor indicates moderate evidence for the null hypotheses (BF01 = 6.355) (see Table S4, and Figure S2). However, beside different signal characteristics, Lustenberger et al. (2013) used substantially higher intensities of RF‐EMF exposure, whereas in the present study the applied RF‐EMF intensities represent realistic worst‐case exposure from real Wi‐Fi installations.

Irrespective of exposure, the present results confirmed the beneficial role of sleep for memory consolidation. Performance in the three memory tasks improved after a night of sleep, reflecting small (FRT, 0.014) to medium effect sizes (WPT, 0.069; SFTT, 0.116) as indicated by generalized η2 values. Sleep‐dependent improvements in memory consolidation have been extensively discussed using different declarative and non‐declarative memory tasks showing that post‐sleep memory retention is better than retention after a wake period (Rasch & Born, 2013). This sleep‐specific beneficial effect is assumed to be reflected in the present results. In particular, in the WPT, declarative memory enhancements after a night of sleep under both experimental conditions are in line with multiple other studies (for reviews, see Diekelmann et al., 2009; Rasch & Born, 2013). Regarding the FRT, recognition memory performance for all faces, regardless of their emotional valence, improved after a night of sleep, which is in agreement with previous findings (Solomonova et al., 2017; Wagner et al., 2007). Additionally, memory performance was better after sleep for neutral and positive facial expressions. These findings are consistent with the results of a recent meta‐analysis (Schäfer et al., 2020), which revealed an enhancement of recognition memory for both emotional and neutral stimuli. In contrast, recognition for negative stimuli did not improve after sleep in the present study. In this respect, only the neutral faces were recognized during the evening recall phase more effectively on the second experimental night when compared with the first night, regardless of the exposure condition (see Table S5). Finally, results of the SFTT are in line with the evidence of the contribution of sleep to procedural memory consolidation (for review, see King et al., 2017).

4.2 Sleep‐specific features related to memory consolidation: Physiological level

There is compelling evidence that depending on the type of memory, certain sleep stages and sleep EEG characteristics are related to the previously mentioned memory consolidation processes. With regard to the macrostructure of sleep, overnight improvements in declarative memory have been related to slow‐wave sleep (N3) (e.g., Diekelmann et al., 2012), whereas overnight improvements in procedural memory have been proposed to be related to time spent in stage N2 sleep (e.g., Walker et al., 2002). Additionally, REM sleep has been associated with both procedural and declarative memory consolidation (Fogel et al., 2007). Finally, the consolidation of emotional memory has been proposed to be dependent on both REM sleep and NREM sleep (Tempesta et al., 2018).

The present analysis revealed that Wi‐Fi exposure had no effect on time spent in sleep stages N2, N3 (slow‐wave sleep), NREM or REM sleep. Bayes factors for N2 and N3 sleep supported this interpretation by providing moderate evidence for the absence of an exposure effect on these two sleep stages (N2, BF01 = 6.672; N3, BF01 = 5.379). The corresponding Cohens' d values indicated also no effect. However, Bayes factors for NREM and REM sleep indicated only anecdotal evidence for the H0 (NREM, BF01 = 2.414; REM, BF01 = 2.266), with Cohens' d values representing small effects (see Table 3). In other words, these results pointed out that N2 and N3 sleep were rather unlikely to be affected by Wi‐Fi exposure, but that an exposure effect on NREM and REM sleep cannot be excluded. It could be speculated that the evaluation of these two effects, whether they are supportive of the null or alternative hypothesis, would have been more convincing if the sample size had been larger. Then, if this supported the tendency observed in NREM sleep at the descriptive level under Wi‐Fi exposure compared to sham (see Table 3), this possible change in NREM could explain at least partially the improvement of declarative memory consolidation.

The literature shows that RF‐EMF effects on sleep architecture are quite heterogeneous. Although some studies found effects in the discussed sleep parameters, others did not (for detailed overview, see Danker‐Hopfe et al., 2016). Therefore, the present results can be assigned to the group of studies that reported null findings with regard to effects of exposure on sleep macrostructure. The same applies to the study by Danker‐Hopfe et al. (2020), which examined the impact of Wi‐Fi exposure on a large number of objective sleep parameters in addition to some subjective sleep variables. This previous study, however, considered sleep data from all 34 recruited participants and disregarded deliberately some of the sleep‐specific variables that are thought to be associated with memory consolidation processes. Thus, the present study fills this gap and complements this previous publication, but with results restricted to a subsample of 30 subjects for whom behavioural data were available.

With regard to sleep microstructure, sleep spindle frequency ranges, as well as slow‐wave activity (0.1–3.5 Hz), have been associated with both declarative and procedural memory improvements (Fogel et al., 2007; Holz et al., 2012). However, other studies did not find a clear association between performance improvements and related sleep stages or EEG power in declarative (Gais et al., 2002) or procedural memory (Rångtell et al., 2017). Sleep spindle density has been proposed to be involved in declarative (e.g., Gais et al., 2002) and in procedural (e.g., Barakat et al., 2011) memory consolidation. Additionally, emotional memory has been positively correlated with fast spindle densities (13–16 Hz) and negatively with slow spindle (10–13 Hz) densities (Solomonova et al., 2017).

The present results did not reveal any Wi‐Fi exposure effect on the EEG power in the ranges of slow oscillations (0.5–0.1 Hz) and narrow (12–14 Hz) and wide (12–16 Hz) sleep spindles. Nor was the sleep spindle density in stages N2 and N3 sleep affected by exposure (see Table 4). This is supported by Cohen's d values, which indicate small or no effects (see Table S2). Bayes factors revealed moderate evidence for the absence of a Wi‐Fi effect on the narrow sleep spindle frequency range at all regions in N2 and N3. Similarly, Bayes factors indicated moderate evidence for the absence of a Wi‐Fi effect on the EEG power in the wide spindle frequency range and in the range of slow oscillations in all cortical regions in both sleep stages, except for the occipital region in N2 and N3. In these cases, Bayes factors revealed only anecdotal evidence for the absence of Wi‐Fi effects. As mentioned above, a larger sample size could have provided stronger evidence for the presence or absence of the reduced EEG power under Wi‐Fi exposure that can be observed at the descriptive level (see Table S2). Furthermore, Bayes factors revealed moderate evidence for an absence of an exposure effect on sleep spindle densities in both sleep stages, with Cohen's d values indicating no effects (see Table S3).

In this respect, Lustenberger et al. (2013) reported that pulsed RF‐EMF induced an increase of slow‐wave activity at the end of the sleep period, whereas spindle activity remained unchanged and sleep‐dependent procedural memory gains were downscaled. Similarly, other RF‐EMF studies did not report effects on the EEG in the spindle frequency range (Fritzer et al., 2007; Hinrichs et al., 2005; Nakatani‐Enomoto et al., 2013; Wagner et al., 1998, 2000) or for spindle density (Lustenberger et al., 2015), in line with the present results. However, as pointed out previously, RF‐EMF effects on the sleep EEG power show mixed results.

In summary, the results at the physiological level did not reveal an impact of Wi‐Fi exposure on any of the sleep parameters that are generally associated with sleep‐dependent memory consolidation processing, such as NREM sleep, specifically slow‐wave sleep, as well as EEG power values in the SO and spindle frequency ranges, and sleep spindle densities. Accordingly, the positive effects that Wi‐Fi exposure had on memory retention in the declarative task were not supported by physiological changes associated with memory consolidation processes during sleep. Thus, the present behavioural and neurophysiological findings did not provide evidence that night‐time Wi‐Fi exposure affects sleep‐dependent memory consolidation, so the positive exposure effect on declarative memory should be classified as inconclusive.