Participants were recruited either from previous studies or by advertisement via different media about lucid dreaming, including a German web page (http://klartraum.de), or from lucid dream induction studies in which specific techniques were applied in order to induce lucidity (e.g., MILD, LaBerge, 1980). Table 1 depicts the participants who successfully finished one of the three experimental protocols (the walking and gymnastic tasks included not only lucid dreamers but also sports students who participated in a lucid dream induction study. The average lucid dream frequency in these groups was thus somewhat lower). Informed consent was obtained from the participants and participation was paid.

The task descriptions for the three conditions:

In all studies, polysomnography was conducted to register the sleep stages. Sleep was recorded by means of the following standard procedures: electroencephalogram (EEG; C3 and C4 for counting and walking; F3, F4, C3, C4, O1, and O2 for gymnastic), EOG, submental electromyogram (EMG) and electrocardiogram (ECG). The data was recorded during the entire night (or during afternoon nap for one participant) by a standard recording device (XLTEK Trex Longtime EEG recorder or Schwarzer ComLab 32). Sleep stages for the counting and walking conditions were scored according to Rechtschaffen and Kales (1968) while those for the gymnastic condition were scored in accordance to the Manual of the American Academy of Sleep Medicine (2008).

The participants spent one to three non-consecutive nights in a sleep laboratory. One participant was recorded twice during an afternoon nap at about 3 pm.

Before sleep, participants received task instructions (see above) in written and oral forms. Afterwards, participants were instructed about left-right-left-right (LRLR) eye signals to mark task events in a lucid dream. The first signal was always to mark the onset of lucidity. In the counting and walking task participants had to mark the beginning of each task sequence as well as the end of the task (five signals for each successful dream). As an example, the exact protocol for the counting task is depicted in Figure 1. In the gymnastic routine, apart from the first signal for the onset of lucidity, only the beginning and the end of the task had to be marked (three signals for each successful dream).

After the participants were familiar with the task and eye signaling, they carried out the task five times in wakefulness (including eye signals). In order to determine the duration of the task in wakefulness, in the counting and walking task the participants measured the times by themselves using a stopwatch—starting after the first eye signal and stopping with the onset of the second one. Because in the gymnastic routine it was not practical for the participants to handle the stopwatch, the experimenter started and stopped the times, according to a verbal signal from the participant, which was given immediately after and before the respective eye signal. For lucid dreams the time intervals were defined as the interval from the end of one LRLR eye signal to the beginning of the next LRLR and so on (see Figure 2).

During the night the experimenter monitored the recordings and woke participants up when recordings showed any of the following criteria: (1) A false awakening, i.e., the recording showing LRLRLRLR eye movements (signal for being awake, see below) but the EEG and EMG channel still showing characteristics for REM sleep. (2) Loss of lucidity, i.e., the recording showing five correct LRLR eye movements in the EOG channel, but no further eye signals occurring 30 s after the previous signal. These criteria were set in order to keep participants from sleeping on and forgetting specific parts of their lucid dreams (Erlacher and Schredl, 2008a). After accomplishing the task successfully in one lucid dream, the participants were to wake themselves up by the technique of focusing on a fixed spot in the lucid dream as described by Tholey (1983). In two cases the experimenter had to wake up participants after false awakening; in all other cases the participants woke up by themselves after finishing the lucid dream task (no cases of loss of lucidity).

The awakening had to be signaled by left-right-left-right-left-right-left-right eye movements (LRLRLRLR). After each lucid dream, participants wrote down a complete and precise dream report. Also they were asked whether they had been lucid and the task had been performed correctly by using a protocol which checked for each element of the task (e.g., eye signals). Any deviations from the protocol were highlighted (e.g., “only a single LR eye movement instead of a pair”) and evaluated to determine whether the data should be excluded. The complete set of dream reports used for data analysis can be found in Supplement 2.

Out of n = 37 recorded lucid dreams n = 16 cases (counting: n = 2; walking: n = 4; gymnastic routine: n = 10) could not be used for the analysis. The criteria for inclusion of a data set were strict, in order to ensure that only lucid dreams conforming exactly to the protocol were used. A data set was excluded for one or more of the following reasons:

To illustrate the last category, the two dream reports will be presented in detail (Original dream reports were in German, translations were done by the authors):

Dream example 1 Slow motion in the dream (gymnastic routine)

[longer dream sequence before] Then I did the LRLR and then I was here, the water was gone, but the floor was dark. I also felt that after this eye signal suddenly it was blurry again. I waited until it got better and then I walked around, wanting to find a brighter spot where I could see better and have more space. I went to a garden where it was bright and I thought, “Okay, I am doing the experiment now.” I gave a LRLR and I jumped and I felt immediately that jumping was very different compared to wakefulness. Just a different perception of the body, also slower. I continued and I did the forward roll—which lasted almost eternally. When I finished the task I gave a LRLR again”

The task duration was indeed 163% longer than in wakefulness (14.8 vs. 5.6 s).

Dream example 2: Running in the dream (walking)

[longer dream sequence before] We talked for about 5 min about the dream I had and that I often have nightmares. Suddenly, I was back at the party and saw the lights again but this time I realized that I was dreaming and did the LRLR. Afterwards I did the protocol but I was running instead of walking the steps. First 10, then 20 and then 30 steps. Finally I woke up”

The task duration was indeed significantly shorter than in wakefulness and therefore the data set was excluded for the statistical comparison of absolute times between wakefulness vs. lucid dream state (3.2). However, the data set is of special interest for the relative time and therefore it was included in the comparison of the relative timing analysis (3.1).

Due to the small sample sizes, individual data are presented and analysis focuses mainly on a descriptive level. Furthermore, for the comparison of times between wakefulness and lucid dreaming, no predictions were made and, therefore, two-tailed statistical t-tests (dependent samples) as well as Wilcoxon tests were applied. For the comparison of task complexity, time differences between wakefulness and lucid dreaming for walking 10 steps and the gymnastic routine were calculated and two-tailed statistical t-test (independent samples) as well as Mann-Whitney-test applied. For all statistical tests a significance level of alpha = 0.05 was used. SPSS Statistics 20 software was used for the statistical analysis. For differences in times between wakefulness and lucid dreaming effect sizes d (Cohen, 1988) were calculated by the open-source software G^∗Power V 3.1.3 (Faul et al., 2007). Cohen (1988) differentiated between small (d = 0.2), medium (d = 0.5), and large (d = 0.8) effect sizes.

Figure 2 shows a sample of a correctly signaled lucid dream for the counting task with five LRLR eye signals. The participant reported the following dream after awakening:

Dream example 3. Correctly signaled lucid dream (counting)

“I was awake and tried WILD [WILD stands for Wake-Initated Lucid Dream which is a technique to induce lucid dreams] which did not induce lucidity immediately. There was a long dream sequence where I had barbecue with some friend. Then I was in a basement with some cupboards and I played with some kids and adults. I knew that I was dreaming and I started to do the protocol: 1. LRLR for “I'm lucid,” 2. LRLR for counting from 1 to 10, 3. LRLR for counting from 1 to 20, 4. LRLR for counting from 1 to 30. After finishing the protocol I waited for a couple of seconds and the dream started to dissolve.”

The interval between two LRLR eye signals corresponds to the counting interval (gray area). Figure 3 depicts the absolute times for the counting task during wakefulness and lucid dreaming. In three cases (P2m32, P3m23, P4f24) the absolute time was longer during lucid dreaming than in wakefulness. Figure 4 depicts the relative times for the counting task during wakefulness and lucid dreaming, e.g., the total time for the whole task equals 100%. Because the ratio for the three parts are 1/6 the expected relative time for counting from 1 to 10 is 16.7%, for counting from 1 to 20 is 33.3% and for counting from 1 to 30 is 50% (marked with the red lines in Figure 4). The differences between the expected percentage and the relative time structure of the counting task in wakefulness are M = 1.1% (SD = 0.6%), M = 0.5% (SD = 0.7%) and M = −1.5% (SD = 0.6%) and in lucid dreaming are M = 0.6% (SD = 0.3%), M = 1.6% (SD = 1.5%) and M = −2.2% (SD = 1.6%) (for counting to 10, 20, and 30, respectively).

Figure 5 depicts the absolute times for the walking task during wakefulness and lucid dreaming. In five cases (P2m32, P3m23, P4f24, P5m34, P7f22) the absolute time was longer during lucid dreaming than in wakefulness. P8m24 exhibits significantly shorter time; however, the participant in this experiment experienced his first lucid dream and reported he was running instead of walking in the steps. Figure 6 depicts the relative times for the walking task during wakefulness and lucid dreaming, e.g., the total time for the whole task equals 100%. Again, the ratio for the three parts are 1/6 and the expected relative time for walking 10 steps is 16.7%, walking 20 steps is 33.3% and walking 30 steps is 50% (marked with the red lines in Figure 6). The differences between the expected percentage and the relative time structure of the walking task in wakefulness are M = 1.2% (SD = 0.8%), M = −0.2% (SD = 0.6%) and M = −1.0% (SD = 0.5%) and in lucid dreaming are M = 1.8% (SD = 2.7%), M = −1.5% (SD = 1.9%) and M = −0.3% (SD = 3.2%) (for walking 10, 20, and 30 steps, respectively).

Figure 7 depicts the absolute times for the gymnastic task during wakefulness and lucid dreaming. In six cases (P9f25, P11m25, P12f24, P13f20, P14f25, P16f24) the absolute time was longer during lucid dreaming than in wakefulness. In the other two cases (P10m24, P15f35) the duration of the gymnastic routine was slightly shorter in the lucid dream state than in wakefulness.

Table 2 summarizes the absolute times required for the counting, walking and the gymnastic task during wakefulness and lucid dreaming. For the counting and walking tasks, the total time is calculated by sum of counting to 10, 20, and 30 or walking 10, 20, and 30 steps. P8m24 was excluded for this statistical analysis because he was running instead of walking the 10, 20, and 30 steps. Statistically significant differences were found for the two tasks with motor activity, walking (p = 0.03) and gymnastics (p = 0.03) but not for the counting task (p = 0.10) (for statistical details see Table 2). In the lucid dream condition, the durations for counting were 27.2%, for walking 52.5% and for the gymnastic routine 23.2% longer than in wakefulness. The effect sizes for all three conditions were quite high (between 0.94 and 1.06), but for the counting task the statistical power was low (0.54).

Figure 8 depicts means and standard deviations for the walking 10 steps and the gymnastic routine during wakefulness and lucid dreaming. In wakefulness the gymnastic routine lasted M = 6.6 s (SD = 0.1) and therefore matched the time for walking 10 steps (M = 6.7 s, SD = 0.3). Comparing the two tasks with motor activity but different complexity, no statistically significant effects were found, t₍₁₃₎ = 1.6, p = 0.14, d = 0.78, power = 0.42; Mann-Whitney-U: Z = 1.04, p = 0.30. Moreover, the more complex gymnastic routine required less time (8.1 s) than walking 10 steps (10.6 s) during lucid dreaming. Again, for this statistical analysis P8m24 was excluded because he was running instead of walking the 10, 20, and 30 steps.

Two different task modalities were used in the present study: those involving motor activity (walking and gymnastic conditions) and those not involving motor activity (counting condition). While increased durations in lucid dreams were observed for both modalities, only tasks with motor activity resulted in significant increases in time (with the caution of possible type II error for counting). These findings are in accordance with Erlacher and Schredl (2004) who also demonstrated that a task involving motor activity (performing squats) yielded an increased duration in lucid dreams. In contrast, tasks which did not involve motor activity (counting) led to negligible differences between wakefulness and lucid dreaming (3.5 and 9.6%). Also no differences were found in study by LaBerge (1985). However, in the present study the difference for counting was considerably higher (27.2%) and it is possible that only the small sample size did not allow it to reach statistical significance. Thus, while prolonged times are quite consistent across the range of different tasks involving motor activity (walking, gymnastics, performing squats), the findings regarding tasks without motor activity (counting) are still inconclusive.

It is important to note that all our conditions actually involved counting. Thus it is possible that the counting itself had an influence on the duration of the motor tasks. Therefore motor tasks which do not involve counting should be investigated in future studies in order to find out if the prolonged durations can still be found and if the extent of a probable increase is smaller or higher than when counting is involved.

Taking a closer look, there was also motor activity in the counting condition because participants were asked to count aloud. Even though the motor activation of the muscles involved during counting seems negligible in contrast to the gross motor activation during walking or the gymnastic routine, future studies should explore the difference for counting aloud and silent.

In two conditions (counting and walking), in addition to the absolute task time, also interim task times (after counting to 10 and to 20; and after walking 10 and 20 steps) have been measured. The analysis showed that relative times for both conditions did not differ between wakefulness and the lucid dream state. This was also true for one participant who accidentally ran the 10, 20, and 30 steps in his dream. Therefore it appears that extended durations in lucid dreams are not dependent on the task length or, in other words, there is not a disproportional time effect when accomplishing longer tasks.

It is worth mentioning that we did not randomize the order of lengths (e.g., P1: 10, 20, 30; P2: 30, 20, 10; etc.). This might confound the results with respect to order effects, however, one might speculate that possible order effects should have distorted the relative times in a systematic proportional way, but this was not the case.

Two different tasks with motor activity were included in the present study: a simple motor task (walking) and a complex motor task (gymnastic routine). While both motor tasks resulted in increased durations in lucid dreams, greater complexity of the task was not associated with greater increases in time. In fact, the trend was even in the opposite direction: Highest increases were observed for the most simple task, walking (52.5%), followed by somewhat more complex task from a previous study, performing squats (39.9%; Erlacher and Schredl, 2004), and finishing with the lowest increases for the most complex task, gymnastic routine (23.3%). While it is not clear if these differences just occurred by chance or there is indeed some inverse relationship between the task complexity and prolonged durations in lucid dreams, from the present data we conclude that more complex actions do not lead to longer durations.

However, it is important to acknowledge, that it is nearly impossible to provide an exact definition of “complexity” (Wulf and Shea, 2002) and the concept has been used in various ways. For example, Guillot and Collet (2005) use this notion in the sense of highly automatic movements (simple) in comparison to cyclical closed movements (complex). The gymnastic routine task, which has been employed in the present study, can be termed complex in several ways: it consisted of a sequence of different elements and was therefore a discrete as opposed to a continuous (walking) motor task. Also the various elements required higher levels of motor coordination and balance. It is still to be investigated whether and to what extent motor tasks which are complex in other ways than the gymnastic routine (e.g., regarding attention, task difficulty) affect dream state durations.

Since the difference in duration between wakefulness and the dream state was observed only for the tasks which involved motor activity, it is worth taking a look into studies which investigated the durations of motor tasks which were mentally simulated by participants while awake. Both in mental simulations and in the dream state motor activity is performed only in one's mind, without moving the physical body. Some mental simulation studies indeed found prolonged durations for mental simulations of walking tasks (Decety et al., 1989; Decety and Jeannerod, 1995) as well as in golf, swimming and weight lifting (for overview see Guillot and Collet, 2005). The difficulty of task, perceived force and skill complexity seem to be time-enhancing factors (Guillot and Collet, 2005). However, the findings from mental simulation studies are ambiguous: Some authors report equivalence of time (e.g., Munzert, 2002), others found shortened durations (Calmels and Fournier, 2001).

One possible explanation from mental stimulation studies for the prolonged durations might be centrally encoded force (Jeannerod, 1994). In the experiment by Decety et al. (1989) the participants who mentally simulated a walking task with an actual 25-kg weight on their back had increased mental simulation durations by about 30%. Jeannerod (1994) suggests that somehow the programmed increased level of force—as a reaction to the actual weight perceived—could not be used to overcome physical resistance and was thus misread by participants as a longer duration. Physically perceived force thus led to the program “increased effort required.” In dreams the perceived force, in the sense of gravity or resistance, might not correspond to the ordinary gravity force in wakefulness, because no real gravity force exists in the dream simulation and muscular feedback is lacking due to REM sleep atonia. Therefore the movements may also be programmed with “increased effort” to compensate for the lack of muscular feedback.

Another possible explanation might be related to neural specifics of REM sleep. Louie and Wilson (2001) found that when rats were trained in a behavioral task their hippocampal activity during the task in wakefulness was replayed in REM sleep but with a somewhat different temporal scaling factor. Most scaling factors were bigger than 1.0 (i.e., there was a slower corresponding activity during REM sleep) and the average was 1.4 ± 0.6. This average duration increase by 40% in REM sleep are in line with our findings on increased duration of motor tasks in lucid dreams (gymnastic: 23.3%; squats: 39.9%; walking: 52.5%). However, it is not clear if the observed replayed neural patterns are indeed linked to (dreamed) motor activity or if they rather represent learning procedures regarding temporal-spatial orientation. The task for the rats involved motor activity and therefore it is possible that the observed neural activity during REM sleep was connected to motor learning, although it is impossible to say if the rats actually dreamed of accomplishing the task. Louie and Wilson (2001) also found that the theta EEG rhythm during REM sleep was about 1.2 times slower compared to the practice in wakefulness and therefore provides two possible explanations. Firstly, this might reflect a globally slower neural processing during sleep due to lower brain temperature. Further, the theta rhythm itself might serve as a pacing mechanism to coordinate interactions during information processing across multiple brain regions.

Finally, it is important to underline that in each condition two participants also produced quite similar time or even slightly shorter times compared to wakefulness. Unfortunately, from our data it is not possible to conclude why those participants performed differently. For example, P1m28 was a highly frequent lucid dreamer and he showed very exact times in his lucid dreams. On the other side, P6f24, who also showed quite exact yet slightly shorter time in the walking condition during her lucid dream, was a very infrequent lucid dreamer.

The relative timing of motor skills plays an important role in motor control theories. Schmidt (1975), for example, proposed in the motor schema theory that the relative time (e.g., the temporal structure of a motor skill) is an invariant component of a so-called generalized motor program and that parameters could scale this structure proportionately in time. For example, throwing a ball can be done fast or slow, however, the relative timing of the involved force impulse need to be proportional in order to speak of the same motor skill. If the relative time structure is not rigidly structured within a certain motor skill then this action is just something else but not the motor skill at hand (e.g., throwing a ball is no longer throwing but something else). The present findings of this study demonstrate that despite the longer absolute durations for tasks involving motor activity, the relative durations remain the same. This finding has important implications for lucid dream applications, such as using lucid dreams for motor skill practice: Athletes practicing long movement sequences seem to practice the same movement sequences as in wakefulness because the temporal structure is still given in their lucid dreams. With respect to relative time issues, it seems that lucid dreaming can be successfully applied for motor skill learning in sports (cf. Erlacher, 2007).

Practice in lucid dreams is similar to mental rehearsal in wakefulness: Movements are rehearsed with an imagined body on a cognitive level. Mental rehearsal is a well-established and widely used technique in sports science and practice. Meta-analyses (Feltz and Landers, 1983; Driskell et al., 1994) demonstrated that it has a positive and significant effect on performance. The evidence suggests that imagined and executed actions to some extent seem to share the same central neural structures. Decety (1996) presented three lines of evidence in support of this correspondence hypothesis: measurement of central nervous activity, autonomic responses, and mental chronometry. Similar correspondence can be demonstrated between dreamed actions in REM sleep and executed actions in wakefulness (Fenwick et al., 1984; LaBerge, 1990; Erlacher and Schredl, 2008b). The present study provides further evidence about the correspondence of mental chronometry (albeit with some scaling factor).

Previous studies with lucid dreamers demonstrated that complex sports skills, such as skiing or gymnastics, can indeed be successfully practiced in lucid dreams (Tholey, 1981). Also in this study, the participants were able to memorize a gymnastic routine and to recall and perform it within a lucid dream. It seems that athletes indeed are able to perform their sports in lucid dreams (Erlacher et al., 2011–2012) and that practice in lucid dreams can increase performance in wakefulness (Erlacher and Schredl, 2010).

The present findings should be replicated in future studies by using bigger sample sizes. It might be possible that not only experienced lucid dreamers can be involved, but also novices, supported by a lucid dream induction technique. In the third condition of the present study, some participants were not experienced lucid dreamers but sport students who took part in a lucid dream induction study. Nevertheless some of them were able to have their first lucid dream and successfully accomplished the requested task in it. A plethora of different methods have been suggested for lucid dream induction and some of them do look promising (see Stumbrys et al., 2012).

Future studies should explore the discrepancies found in the counting condition, as well a possible negative relation between the task complexity and prolonged times, i.e., that more simple motor tasks for some reason lead to longer durations. Further, measures of perceived effort (e.g., Borg, 1982) could be included to explore the relationship between prolonged durations and perceived effort when accomplishing a motor task. Also it might be worth investigating other possible influencing factors that were found to have an effect on durations in mental simulations, e.g., the level of expertise and task familiarity (Guillot and Collet, 2005). Concerning the features of the tasks used in our own studies, it might also be worth exploring possible differences between continuous (walking, squats) and discrete (gymnastic routine) motor tasks.

One of the difficulties with chronometric lucid dream research is that it mainly relies on subjective time perception. Therefore it would be interesting to approach this problem with another way of measuring the durations of dreamed actions by incorporating physical time intervals into lucid dreams with external auditory signals. In a recent study Strelen (2006) showed that in a lucid dream the dreamer can hear and distinguish an externally provided acoustic stimulus. These audio cues could serve as a start and stop signal of an interval, during which lucid dreamers, for example, could count numbers or count their steps while walking. The problem of subjectivity within dreams of course could not be avoided, however, this would allow another comparison of physical clock time with subjective time experience in lucid dreams.