In line with our hypotheses, we could show with the fourth study that olfactory TMR of therapy success after in vivo exposure-based group therapy in spider phobic patients, compared with odorless vehicle presentation, resulted in an increase in spindle power in response to the reminder cues associated with therapy success. However, unexpectedly, targeting the consolidation process during sleep did not result in a benefit in extinction memory as measured by fear response.
As above described and already replicated with study one, the active system consolidation hypothesis assumes an interaction between slow oscillations, sharp wave-ripples, and spindles in the process of sleep-dependent memory consolidation. Thus, not surprisingly, an effect of TMR was also visible in enhanced spindle power in the fourth study.
The active system consolidation hypothesis assumes the above described mechanisms especially during the consolidation of declarative memory. Exposure therapy consists of learning processes encompassing both, declarative and non-declarative memory, although the main proportion is declarative memory. Non-declarative elements are contained especially in the habituation to stimuli associated with the feared object, such as pictures, or the object itself. The larger part of the extinction learning process consists of consciously acquiring new memory traces associated with information about the irrationality of the person’s fear, the establishing of clinical behavior due to avoiding the phobic object, and possible alternative behaviors towards the phobic object. Even though this information is acquired in a very emotional context, a clear declarative component is given.
The increases in EEG slow and fast spindle power can be interpreted in line with the active system consolidation hypothesis. In the active system consolidation hypothesis, a key role of sleep in the consolidation of memory by a reorganization of hippocampal information to the neocortex is assumed. The transfer is enabled by an interaction between slow oscillations, spindles, and sharp wave-ripples (Mölle & Born, 2011). As shown in the first study, this mechanism is reflected in specific oscillatory changes in the human EEG in response to TMR, namely in increases in delta activity, slow oscillation slopes, and fast
spindle activity. In study four, there was only evidence for increase in spindle power in response to TMR of therapy success. However, this increase in spindle power might reflect higher efficacy of spindles, for example established by a stronger coupling to sharp wave-ripples, and thus a bottom–up bias of the process of a hippocampal-neocortical transfer.
Importantly, the increase in spindle power was only revealed by an interaction with the amount of activity over the different hemispheres. A possible explanation for this lateralization effect is the strong emotional nature of the exposure therapy, where patients have to face objects they are extremely afraid of. Lateralization of EEG activity is a well-observed phenomenon during learning of emotional tasks (Davidson & Fox, 1982; Ekman &
Davidson, 1993).
Contrary to the findings of the first study, a change in the delta power spectrum is not identified by TMR reactivations of extinction memory. Therefore, we did not investigate changes in slow oscillation slopes as done in the first study. A possible speculation about this non-finding is that we used a 1.5 hr-afternoon nap in the fourth study, whereas the sleep period in the first study contained a whole night of sleep. Therefore, we focused on deeper N2 sleep stages to re-present the extinction memory-associated odor, which is also considered as NREM sleep, but with different characteristic electrophysiological oscillations. Of course, we continued the re-presentation of the TMR cues if sleep became deeper and subjects entered SWS, but in the end, the analyzed EEG segments contained N2 and SWS, and not SWS alone, as in the first study.
Interestingly, sleep spindles are the hallmark of N2 sleep, even though they also occur in SWS, whereas slow oscillations are only visible shortly before SWS when N2 sleep gets very deep. The absence of an influence of our TMR on the slow oscillation band could possibly be due to only a small amount of slow oscillations in N2 sleep, whereas spindles are the main oscillatory event of N2 sleep. However, spindles are also included as an important parameter in the active system consolidation paradigm, and a hippocampal bottom-up transfer could possibly be reflected in different oscillatory events based on different NREM sleep stages – slow and fast spindle power in N2 sleep, and delta and fast spindle power in SWS. It is important to underline that this is only a vague speculation and needs to be experimentally verified. A way to do this would be to use the same paradigm of study four with the sleep period extended to a whole night, with TMR only during SWS.
Moreover, the active system consolidation hypothesis, as the sequential hypothesis, speculates about a possible role of REM sleep in system consolidation (Mölle & Born, 2011);
that is a consolidation at a cortical synaptic level after reorganization of the information from
the hippocampus to the neocortex. Not surprisingly, only few subjects had REM sleep during their afternoon nap. To analyze an interaction of NREM sleep and REM sleep, more subjects would need to undergo REM sleep during their sleep, which could also be investigated when using a whole night-design.
In contrast to the confirmation of our hypothesis concerning oscillatory changes during TMR, we could not find an effect of TMR on fear behavior after sleep. We explain this finding by a ceiling effect of extinction learning, since exposure therapy is considered very effective in the treatment of anxiety disorders, and very successful even after one session (Öst, Brandberg, & Alm, 1997; Öst, Ferebee, & Furmark, 1997; Öst, 1996). Further, it is important to underline that our patients participated on a self-initiated, voluntary basis and were thus highly motivated and highly committed. Therefore, future studies using less effective paradigms are suggested in order to investigate if TMR can influence fear behavior, such as in sensu exposure by watching videos with spiders or only psychoeducation without exposure. Furthermore, since our results indicate also a maximal treatment outcome in the wake group, therapies for other specific phobias that are often treatment-resistant, for example social anxiety disorder (Blanco et al., 2003; Cuthbert, 2002; Heimberg, 2002), could serve as extinction memory learning. In fact, a main reason for the decision to test spider phobic patients in this first clinical application of TMR was the easy access to this patient group because of the high prevalence, usually in addition to little comorbidity. Since our finding constitutes the first attempt to apply basic research to a clinical setting, it would of course be interesting and necessary to replicate the study in a clinical population with another fear disorder.
A possible speculation about memory consolidation during sleep is that post-learning sleep only initiates a process, and thus effects on behavior are not visible immediately, but only after a time. In case of system consolidation, an effective consolidation of the information by a transfer from the hippocampus to existing neocortical networks could probably only be initiated immediately after therapy, and will be more pronounced some time later. In line with this assumption, a previous study found reduced fear after exposure therapy in post-treatment sleep compared with wake patients not immediately after sleep, but one week later (Kleim et al., 2013). To investigate an increase in extinction memory after several time, we designed our study with two sessions separated by one week, and we compared the fear values between sleep groups after this week. But also then, no differences in behavior were found between sleep with reactivation, sleep without reactivation, and wake groups,
which could also be explained by a ceiling effect. Also three months later, subjects of the different groups did not show differences in questionnaires assessing fear of spiders.
Alternatively, the missing effect of TMR could be explained by an ineffectiveness of the reactivation to induce behavioral changes when delta power was not affected by TMR.
This could possibly be revealed by the absence of changes in slow wave activity after TMR, reflecting an “incomplete” attempt of memory transfer which could not be finished because of missing time-locked slow oscillations. As a consequence, triggering previously learned declarative information that is still hippocampus-dependent is possible by TMR and entrains spindles, but does not result in preferential consolidation of this information because of the asynchrony of the third key player in this mechanism, slow oscillations.
Since the specific investigation of oscillatory changes was neglected in most human TMR studies, no other studies hint to a causal role of changes in EEG delta power and behavioral changes. This assumption of course needs to be verified by future studies. Also for this research question, a full night of post-extinction learning sleep with TMR could provide answers.
In case future whole night-designs or studies in other fear disorders will uncover an effect of olfactory TMR on extinction memory, it would also be interesting to examine if sounds are also potent as extinction memory cues. This would be important with regard to an application in an everyday clinical setting, since presenting sounds during sleep does not require much equipment, whereas presenting olfactory cues during sleep depends on an olfactometer.
Moreover, given the case that refined paradigms would succeed to find behavioral effects of extinction memory consolidation together with a replication of our EEG findings, a lot of follow-up experiments would be interesting. For example, since an increase in spindles after memory cue presentation seems to reflect the memory consolidation mechanism, spindles during post-learning sleep could also be induced by sinus tones with spindle frequencies, which could lead to an entrainment of the brain in the spindle frequency.
To sum up, the changes in spindle power in response to memory cues seem to reflect targeted extinction memory reactivation on an electrophysiological level. Alternatively, they also could reflect sleep stage-specific effects of TMR during N2 sleep, hallmarked by spindles. Further studies using sleep over a whole night instead of 1.5 hours are needed to investigate these speculations. Moreover, a whole-night paradigm would also be able to account for the role of REM sleep in the active system level consolidation during sleep, a question we did not investigate in study one. Furthermore, our paradigm could be refined by
using patients with different, more treatment-resistant anxiety disorders or less intense treatments for anxiety disorders to prevent ceiling effects, which opens the possibility to enhance the extinction memory.