What We Know Currently about Mirror Neurons - The Most [Over] Hyped Topic in Neuroscience

If you ask Christian Jarrett, Ph.D, editor of the British Psychological Society's Research Digest blog and staff writer on their magazine The Psychologist, mirror neurons are the most (over) hyped topic in neuroscience.

Back in December of 2012 he posted an entry at his Psychology Today blog entitled, "Mirror Neurons: The Most Hyped Concept in Neuroscience?" A year later, he authored an article for WIRED on the study presented below. Be sure top check out his overview of the article if the actual article proves too long or geeky for you - he offers an excellent assessment.

For your entertainment, here is Dr. Jarrett's original piece, followed by the new article published (open access) in Current Biology.

Mirror Neurons: The Most Hyped Concept in Neuroscience?

Mirror neurons are fascinating but they aren’t the answer to what makes us human

Published on December 10, 2012 by Christian Jarrett, Ph.D in Brain Myths


Back in the 1990s neuroscientists at the University of Parma identified cells in the premotor cortex of monkeys that had an unusual response pattern. They were activated when the monkeys performed a given action and, mirror-like, when they saw another individual perform that same movement. Since then, the precise function and influence of these neurons has become perhaps the most hyped topic in neuroscience.

The hype

In 2000, Vilayanur Ramachandran, the charismatic neuroscientist, made a bold prediction: “mirror neurons will do for psychology what DNA did for biology.” He's at the forefront of a frenzy of excitement that has followed these cells ever since their discovery. For many, they have come to represent all that makes us human.

Perhaps, in those early heady years, Ramachandran was just getting a little carried away? Not at all. For his 2011 book, The Tell-Tale Brain, Ramachandran took his claims further. In the chapter “The neurons that shaped civilisation”, he argues that mirror neurons underlie empathy, allow us to imitate other people, that they accelerated the evolution of the brain, that they help explain the origin of language, and most impressively of all, that they prompted the great leap forward in human culture that happened about 60,000 years ago.

“We could say mirror neurons served the same role in early hominin evolution as the Internet, Wikipedia, and blogging do today,” he concludes. “Once the cascade was set in motion, there was no turning back from the path to humanity.”

Ramachandran is not alone. Writing for The Times (London) in 2009 about our interest in the lives of celebrities, the eminent philosopher AC Grayling traced it all back to those mirror neurons. “We have a great gift for empathy,” he wrote. “This is a biologically evolved capacity, as shown by the function of ‘mirror neurons’.” In the same newspaper this year, Eva Simpson wrote on why people were so moved when Tennis champ Andy Murray broke down in tears. “Crying is like yawning,” she said, “blame mirror neurons, brain cells that make us react in the same way as someone we’re watching (emphasis added)”. In a New York Times article in 2007, about one man’s heroic actions to save another, those cells featured again: “people have ‘mirror neurons,’” Cara Buckley wrote, “which make them feel what someone else is experiencing (emphasis added)”.

If mirror neurons grant us the ability to empathise with others, it follows that attention should be drawn to these cells in attempts to explain why certain people struggle to take the perspective of others – such as can happen in autism. Lo and behold the “broken mirror hypothesis” of autism.

The reality

The ubiquitous idea that mirror neurons “cause” us to feel other people’s emotions can be traced back to the original context in which they were discovered – the motor cells in the monkey brain that responded to the sight of another person performing an action. This led to the suggestion that mirror neurons play a causal role in allowing us to understand the goals behind other people’s actions. By representing other people’s actions in the movement-pathways of our own brain, so the reasoning goes, these cells provide us with an instant simulation of their intentions – a highly effective foundation for empathy.

It’s a simple and seductive idea. What the newspaper reporters (and over-enthusiastic neuroscientists) don’t tell you is just how controversial it is. The biggest and most obvious problem for anyone advocating the idea that mirror neurons play a central role in our ability to understand other people’s actions, is that we are quite clearly capable of understanding actions that we are unable to perform.

A non-player tennis fan who’s never held a racket doesn’t sit baffled as Roger Federer swings his way to another victory. They understand fully what his aims are, even though they can’t simulate his actions with their own racket-swinging motor cells. Similarly, we understand flying, slithering, coiling and any number of other creaturely movements, even if we don’t have the necessary motor cells to simulate them. From the medical literature there are also numerous examples of comprehension surviving after damage to motor networks – people who can understand speech, though they can’t produce it; others who recognise facial expressions, though their own facial movements are compromised. Perhaps most awkward of all, there’s evidence that mirror neuron activity is greater when we view actions that are less familiar – such as a meaningless gesture – as compared with gestures that are imbued with cultural meaning, such as the victory sign.

Mirror neuron fans generally accept that action understanding is possible without corresponding mirror neuron activity, but they say mirror neurons bring an extra depth to understanding. In a journal debate published this year in Perspectives in Psychological Science, Marco Iacoboni insists mirror neurons are important for action understanding, and he quotes others saying how they allow “an understanding from within”. Critics in the field believe otherwise. Gregory Hickok at the University of California Irvine thinks the function of mirror neurons is not about understanding others’ actions per se, but about using others actions’ in the process of making our own choice of how to act. Seen this way, mirror neuron activity is just as likely a consequence of action understanding, as a cause.

What about the grand claims that mirror neurons played a central role in accelerating human social and cultural evolution by making us empathise with each other? Troublesome findings here include the fact that mirror neurons appear to acquire their properties through experience. Research by Cecelia Heyes and others has shown that learning experiences can reverse, nullify or exaggerate mirror-like properties in motor cells. It can’t reasonably be claimed that mirror neurons made us imitate and empathise with each other, if the way we choose to behave instead dictates the way our mirror neurons work. On their role in cultural evolution, Heyes says mirror neurons are affected by cultural practices, such as dancing and music, just as much they influenced them.

Finally, what about the suggestion that mirror neurons play a role in autism? It’s here that the hype is probably the least justified. There are numerous findings showing that people with autism have no problem understanding other people’s actions (contrary to the broken mirror hypothesis) and that they show normal imitation abilities and reflexes. For a new review paper, Antonia Hamilton assessed the results from 25 relevant studies, concluding: “there is little evidence for a global dysfunction of the mirror system in autism.”

--

Motor cells that respond to the sight of other people moving are intriguing, there’s no doubt. It’s likely they play a role in important social cognitions. But to claim that they make us empathic, and to raise them up as neuroscience’s holy grail, as the ultimate brain-based root of humanity, is ridiculous. The evidence I’ve mentioned is admittedly somewhat biased, designed to counteract the hype and show just how much debate and doubt persists. In fact, the very existence of mirror neurons in the human brain is still disputed by some. That’s where we’re at with the study of these cells. We’re still trying to find out whether they exist in humans, where they are, and what exactly it is they do. Mirror neurons are fascinating but they aren’t the answer to what makes us human.

Update December 2013: A new review paper has provided a sober assessment of what we currently know about mirror neurons. [This is the overview I referred to above.]

And now, your featured article.

Full Citation:

Kilner, JM, Lemon, RN. (2013, Dec 2). What We Know Currently about Mirror Neurons. Current Biology; 23(23), pR1057–R1062. DOI: http://dx.doi.org/10.1016/j.cub.2013.10.051

What We Know Currently about Mirror Neurons

J.M. Kilner, R.N. Lemon
Sobell Department of Motor Neuroscience and Movement Disorders, UCL Institute of Neurology, London, UK, WC1N 3BG

Summary

Mirror neurons were discovered over twenty years ago in the ventral premotor region F5 of the macaque monkey. Since their discovery much has been written about these neurons, both in the scientific literature and in the popular press. They have been proposed to be the neuronal substrate underlying a vast array of different functions. Indeed so much has been written about mirror neurons that last year they were referred to, rightly or wrongly, as “The most hyped concept in neuroscience”. Here we try to cut through some of this hyperbole and review what is currently known (and not known) about mirror neurons.

Introduction

Mirror neurons are a class of neuron that modulate their activity both when an individual executes a specific motor act and when they observe the same or similar act performed by another individual. They were first reported in the macaque monkey ventral premotor area F5 [1] and were named mirror neurons in a subsequent publication from the same group [2]. Ever since their discovery, there has been great interest in mirror neurons and much speculation about their possible functional role with a particular focus on their proposed role in social cognition. As Heyes [3] wrote “[mirror neurons] intrigue both specialists and non-specialists, celebrated as a ‘revolution’ in understanding social behaviour … and ‘the driving force’ behind ‘the great leap forward’ in human evolution…”. Indeed so much has been written in both peer-review literature and elsewhere about mirror neurons and their proposed functional role(s) that they have recently been given the moniker “The most hyped concept in neuroscience” [4].

For us, the discovery of mirror neurons was exciting because it has led to a new way of thinking about how we generate our own actions and how we monitor and interpret the actions of others. This discovery prompted the notion that, from a functional viewpoint, action execution and observation are closely-related processes, and indeed that our ability to interpret the actions of others requires the involvement of our own motor system.

The aim of this article is not to add to this literature on the putative functional role(s) of mirror neurons, but instead to provide a review of the studies that have directly recorded mirror neuron activity. To date, there have been over 800 published papers on mirror neurons (from a PubMed search using: “mirror neuron” OR “mirror neurons”). Here, we restrict our attention to only the primary literature on mirror neurons. Mirror neurons were originally defined as neurons which “discharged both during monkey’s active movements and when the monkey observed meaningful hand movements made by the experimenter” [2]. Thus, the key characteristics of mirror neurons are that their activity is modulated both by action execution and action observation, and that this activity shows a degree of action specificity. This distinguishes mirror neurons from other ‘motor’ or ‘sensory’ neurons whose discharge is associated with either execution or observation, but not both. It also distinguishes mirror neuron responses from other types of response to vision of objects or other non-action stimuli. As the activity of mirror neurons cannot yet be unambiguously detected using neuroimaging techniques, we have excluded human and non-human primate imaging studies from this review. We therefore focus on the 25 papers [1, 2, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27] that have reported quantitative results of recording mirror neurons or mirror-like neurons in macaque monkeys since 1992 (Table 1).

Table 1 

Proportion of neurons recorded in macaque premotor cortex (area F5) and posterior parietal cortex that showed mirror neuron properties.

Reference	Recording area	No. neurons	No. mirror	% mirror	1Action specificity	Observed effector
Bonini et al. [5]	F5	154	36	23.4%	y	Hand
Caggiano et al. [6]	F5	299	149	49.8%	n	Hand
Caggiano et al. [8]	F5	219	105	48%	n	Hand
Caggiano et al. [7]	F5	224	123	54.9%	n	Hand (video)
Caggiano et al. [9]	F5	785	247	31.5%	n	Hand (video)
Ferrari et al. [11]	F5	485	130	26.8%	y	Mouth
Ferrari et al. [12]	F5	209	52	24.9%	y	Hand
Gallese et al. [2]	F5	532	92	17.3%	y	Hand
Kohler et al. [16]2	F5	497	63	12.7%	y	Auditory
Kraskov et al. [17]	F5	64	31	48.4%	y	Hand (PTNs)
di Pellegrino et al. [1]	F5	184	18	9.8%	y	Hand
Rizzolatti et al. [18]	F5	300	60	20%	y	Hand
Rochat et al. [19]	F5	282	92	32.6%	y	Hand
Umilta et al. [23]	F5	220	103	46.8%	y	Hand
Bonini et al. [5]	IPL	120	28	23.3%	y	Hand
Fogassi et al. [13]	IPL	165	41	24.8%	y	Hand
Rozzi et al. [20]	IPL	423	51	12%	y	Hand
Shepherd et al. [21]	LIP	153	30	19.6%	n	Eye-gaze
Dushanova and Donoghue [10]	M1	303	105	34.6%	y	Reaching
Tkach et al. [22]	M1	829	581	70.1%	y	Tracking arm
Vigneswaran et al. [24]	M1	132	77	58.3%	n	Hand (PTNs)
Tkach et al. [22]	PMd	128	77	60.1%	y	Tracking arm
Ishida et al. [14]	VIP	541	48	8.9%	y	Bimodal tactile/visual
Fujii et al. [27]	PM3	148	_	3–14%4	n	Hand
	IPS5	148	-	10–42%4	n

1This column indicates if mirror neurons were tested for any form of action specificity.

2These data were further analysed by Keysers et al. [15].

3Including area F5.

4See text.

5Included anterior bank of the intraparietal sulcus (IPS). PTN, pyramidal tract neuron.

Mirror neurons were first described in the rostral division of the ventral premotor cortex (area F5) of the macaque brain, and have subsequently been reported in the inferior parietal lobule, including the lateral and ventral intraparietal areas, and in the dorsal premotor and primary motor cortex. But despite the large array of areas in which mirror neurons have been reported, the majority of mirror neuron research has studied the activity of mirror neurons in area F5 (15/25 papers; Figure 1A).

Figure 1. Number of mirror neurons recorded in areas F5 and in the IPL.

(A) The percentage of mirror neurons as a function of publication year for studies reporting mirror neurons in F5 when observing hand actions. The black line shows the line of best fit. (B) The percentage of mirror neurons in premotor area F5 and in the inferior parietal lobule (IPL). The average percentage of mirror neurons for each region is shown in black and the percentage of total mirror neurons is shown in grey with the total number of mirror neurons and neurons recorded given above.

Mirror Neurons in Ventral Premotor Region F5

Of the 15 papers reporting mirror neuron activity in area F5, 11 provide details of the number of mirror neurons recorded when observing the experimenter (not a video) reaching and grasping objects with their hand. On average, 33.6% of neurons recorded in F5 have been described as mirror neurons when the monkey observed hand actions performed by a human experimenter in front of them (ranging from 9.8–49.8%; Figure 1A,B). It is of note that the percentage of mirror neurons reported appears to increase as a function of time. This most likely reflects a sampling bias during data collection.

The first three papers [1, 2, 18] described the basic properties of mirror neurons, and their percentages are low compared with later studies. The more recent papers, in general, have investigated modulations of mirror neuron activity with some form of task manipulation. The methodological approach of these later papers is to first select neurons based on their motor properties (for example, selectivity for grasping) and then investigate the responses of this neuronal population to observed actions. This subtle change in the experimental strategy might explain the apparent increase in the percentage of mirror neurons in F5 as a function of time. Some investigators have avoided the sampling bias based on mirror properties by studying identified pyramidal tract neurons in area F5, selected on the basis of their antidromic response and not for their properties during action execution or observation [18]. A large proportion of pyramidal tract neurons in F5 and in M1 appear to show mirror-like responses (Table 1).

The three early papers [1, 2, 18] provided details about the relative selectivity of mirror neuron discharge during action execution and observation. On average, 48.9% of mirror neurons were classified as broadly congruent. Some mirror neurons discharged for only one action type, such as grasping, during both execution and observation, but showed no specificity for the type of grasp, for example precision grip or whole hand prehension. Others discharged for more than one type of observed action, for example grasping and holding. One of the three papers [2] describes a further category of mirror neurons, strictly congruent mirror neurons; these are defined as mirror neurons that respond selectively to one action type, such as precision grip, during both action execution and observation, and are reported as constituting 31.5% of mirror neurons recorded. Two of the three papers [2, 18] report a further category of neuron in F5 that discharged during action observation but not during action execution; on average these neurons, which would not be included as mirror neurons, have been reported as making up 5.1% of the neurons in F5.

Further neuroanatomical studies of area F5 have revealed three interconnected sub-divisions [28]. The sub-division in which mirror neurons are located is suggested to be on the convexity of the precentral gyrus, adjacent to the inferior limb of the arcuate sulcus, and referred to as area F5c. This is distinguished from area F5p (posterior), which is reciprocally connected both with posterior parietal area AIP and primary motor cortex M1, and from area F5a (anterior) in the depth of the sulcus, which has prefrontal connections [29].

Two studies [7, 9] have been reported that have shown that F5 mirror neurons discharged both to the observation of an action performed in front of the monkey by the experimenter and to videos of the same action. On average 26.9% of F5 neurons discharged when the monkey observed a video of a grasping action. One of the two studies [7] reported the relative number of mirror neurons that discharged to real and to videoed actions: 46.4% of neurons in F5 that responded to an executed action also responded when observing a real action, whereas only 22.3% responded when observing a videoed action. Although fewer mirror neurons responded when the monkey was observing the video of an action, for those mirror neurons that did discharge, there was no significant difference in the pattern or rate of mirror neuron discharge between real and videoed actions.

Two of the early papers [2, 18] on mirror neurons reported that they could not find any neurons that discharged when monkeys observed an object being grasped with a tool. Subsequently, two studies [12, 19] showed that mirror neurons did respond to such a tool-based action. In both these latter cases, however, the monkeys had received a high exposure to tool use during the training period prior to the recordings. One study [12] reported that 20% of F5 neurons were tool-responding mirror neurons, whereas the other reported the much higher percentage of 66.6% [20]. This high percentage most likely reflects a combination of a small sample size (n = 27) and strict inclusion criteria.

Two papers [15, 16] have reported that neurons in F5 responded to the sound of an action: so-called auditory mirror neurons. On average, 17% of F5 neurons have been reported to have auditory properties (12.7% and 21.3%, respectively, in the two papers). Four papers [6, 7, 8, 23] have reported that mirror neurons not only discharged during action observation but that their firing is further modulated by different factors: occlusion [23], relative distance of observed action [8], reward value [6] and the view point of the observed action [7]. Umilta et al. [23] showed that 19/37 mirror neurons discharged even when the observed action was occluded or hidden from the observer, demonstrating that direct vision of the action was not necessary to elicit mirror neuron discharge. Caggiano et al. [7] showed that 149/201 mirror neurons discharged preferentially for one or more of three different views of the same action (at 0, 90 and 180 degrees). Sixty of these neurons showed a preference for only one view point.

Caggiano et al. [8] also found that F5 mirror neurons have a preference for whether an observed action occurred in peripersonal or extrapersonal space: 27/105 mirror neurons discharged preferentially when the observed action occurred in the monkeys extra-personal space, whereas 28/105 mirror neurons discharged preferentially when the observed action occurred in the monkey's peri-personal space. The remaining 50 mirror neurons showed no preference. Caggiano et al. [6] reported that mirror neuron discharge is modulated by the value of the reward associated with the action: they showed that 40/87 mirror neurons responded more when a rewarded object was grasped, while 11/87 responded more when observing an action to a non-rewarded action. The remaining mirror neurons showed no preference.

One study [17] recorded from 64 neurons in F5 that were identified as pyramidal tract neurons. Thirty-one of these neurons were classified as mirror neurons, with 14/31 mirror neurons showing the ‘classic’ facilitation response during the action observation condition. Compared with baseline, the activity of the remaining 17 mirror neurons was significantly suppressed during action observation. The inclusion of these ‘suppression mirror neurons’ [8, 17, 24, 25] clearly changes the overall proportion of neurons responsive during action observation.

In a recent study, Maranesi et al. [30] compared multiunit activity responses in areas F5, F4 (premotor regions) and F1 (primary motor cortex, M1). They reported a higher proportion of recording sites showing mirror type responses in area F5 (particularly in area F5c), compared with area F4 (caudal part of the ventral premotor cortex) and with F1. In addition, they reported that in penetration sites where they identified mirror responses, they were rarely able to evoke movement using intracortical microsimulation and argued that this might be due to presence of suppression mirror neurons, as first identified by Kraskov et al. [17].

One interesting study [27] looked at activity in premotor and parietal cortex neurons of the left hemisphere of a Japanese macaque monkey, either while it observed another monkey sitting opposite making reach-to-grasp movements for food rewards, or when it performed similar actions itself. Many neurons in both cortical areas were active during the other monkey’s movements, with the proportion varying across different actions (Table 1). Premotor cortex neurons showed a distinct preference for movements involving the observed monkey’s right arm and hand, and showed a similar preference for the monkey’s own right-sided actions.

Mirror Neurons in the Inferior Parietal Lobule

Four papers [5, 13, 20, 25] have reported neuronal activity recorded in the inferior parietal lobule that the authors have described as that of mirror neurons (Figure 1B). None of these papers explicitly specifies the percentage of neurons that were classified as mirror neurons; for three of these papers, however, we were able to estimate from the numbers in the papers that the average percentage of sampled neurons that were mirror neurons was 20% (41/165 Fogassi et al. [13]; 28/120 Bonnini et al. [5]; 51/423 Rozzi et al. [20]).

Two papers [5, 13] describe the modulation of mirror neuron activity in the inferior parietal lobule by the overall goal of the observed action. Here monkeys observed an experimenter reaching for and grasping an object and either placing it in the mouth (eating) or placing it in a container (placing). On average 53% of mirror neurons had a significantly greater firing rate when the monkey observed the ‘eating’ compared with the ‘placing’ condition, 17% had a significantly greater firing rate for ‘placing’ compared with ‘eating’. The remaining 30% showed no difference between the two conditions. Yamazaki et al. [25] reported examples of mirror neuron activity in macaque area inferior parietal lobe; these neurons responded to the same action carried out in rather different contexts, suggesting that they are involved in encoding the ‘semantic equivalence’ of actions carried out by different agents in different contexts.

Rozzi et al. [20] investigated the properties of mirror neurons in the IPL. They reported that 58% of mirror neurons were responsive to only one type of hand action, for example grasping, and 25% were responsive to two different hand actions. The remaining 17% were responsive to either observed mouth actions or mouth and hand actions. Furthermore, they reported that 29% of IPL mirror neurons were strictly congruent and 54% were broadly congruent.

Mirror Neurons in the Primary Motor Cortex

The first few papers [2, 18] that described mirror neurons in area F5 also reported that the authors found no evidence of mirror activity in M1. Indeed, Gallese et al. [2] argued that, because most neurons in M1 show activity during self-movement, the absence of detectable mirror activity in M1 was evidence against the idea that this activity might actually represent monkey’s making small, covert movements while they watched the experimenter. Similarly, a recent multiunit recording study [29] found only a low level of mirror activity within primary motor cortex. However, three papers [10, 22, 24] have reported mirror neuron-like responses in M1.

Tkach et al. [22] reported that when monkeys either performed a visuomotor tracking task themselves, or watched the same target and cursor being operated by an experimenter, 70% (581/829) of recorded neurons in M1 showed stable preferred direction tuning during both execution and observation. These authors also reported that 60% (77/128) of neurons in dorsal premotor cortex were modulated in the same way.

Dushanova and Donoghue [10] recorded from neurons in M1 whilst the monkey either performed a point-to-point arm-reaching task or observed a human experimenter performing the same action. This study reported that 34.7% (105/303) of the neurons recorded in M1 were directionally tuned during both action execution and action observation. The mean firing rate during the observation condition was on average 46% of that during the execution condition. In addition, 38% of neurons retained the same directional tuning during both execution and observation conditions. It should be noted that these studies differ from those previously described that recorded from F5 and IPL.

All the studies on mirror neurons in F5 and IPL have employed tasks where the macaque monkey observed either a video or the experimenter performing simple reach and grasp actions. The two studies [10, 22] described above on mirror-like responses in M1 differed in that they used tasks in which the monkey had been extensively trained on the motor execution task. It is unclear whether the relatively high percentage of these mirror-like responses, compared with those in F5 and IPL, reflects differences between the task or real differences in the number of mirror neurons.

The final paper [24] on M1 mirror neurons recorded from 132 neurons that were identified as pyramidal tract neurons; 58% of these neurons (77/132) were classified as mirror neurons. As in F5, these authors found that these pyramidal tract neurons were either facilitation mirror neurons (58.5%) or suppression mirror neurons (41.5%) during the action observation condition. In contrast to F5, facilitation mirror neurons in M1 fired at significantly lower rates during action observation vs execution, with the former reported as “less than half of that when the monkey performed the grip”. It is noteworthy that these authors made simultaneous EMG recordings from up to 11 different arm, hand and digit muscles and confirmed complete absence of activity during action observation.
 

Mirror Neurons in Other Regions

Above, we have described the results of studies reporting mirror neurons in ventral premotor cortex, dorsal premotor cortex, primary motor cortex and inferior parietal lobule. Three further papers [14, 21, 26] have reported mirror neuron-like responses in two further areas. The first [14] recorded visuotactile bimodal neurons in the ventral intraparietal area (VIP). These are neurons that exhibit tactile receptive fields for a particular body part (for example, face or head) and also exhibit visual receptive fields in the congruent location. This study demonstrated that 48/541 bimodal neurons also exhibited visual receptive fields when observing the congruent area being touched on the experimenter. These neurons were not called mirror neurons but ‘body-matching bimodal neurons’.

Shepherd et al. [21] reported mirror neuron-like responses in the lateral intraparietal (LIP) area. These authors reported that 30/153 neurons in LIP responded not only when monkeys oriented attention towards the receptive field of those neurons, but also when they observed other monkeys orienting in the same direction.

Yoshida et al. [26] recently recorded from neurons in the medial frontal cortex, some of which selectively responded to self or observed actions within a social context. The neurons were recorded in one of two monkeys who, on alternate trials, chose a movement in order to earn a reward. Correct (or incorrect) choices rewarded (or punished: no reward) both monkeys. ‘Partner-type’ neurons were selectively responsive to the choices made by the other monkey, signalling the correct or incorrect choice made; interestingly around 19% of these ‘partner neurons’ showed decreased activity during self-movement.

Relating Human Neuroimaging Data to Mirror Neuron Activity

Of the over 800 papers returned when searching PubMed for ‘mirror neurons’ or ‘mirror neuron’, the vast majority report the results of experiments on human subjects. Of these, the results of human neuroimaging experiments, specifically fMRI [31], confirm a broad overlap between cortical areas active in humans during action observation and areas where mirror neurons have been reported in macaque monkeys (see above). Thus, changes in the BOLD signal during action observation seem to be consistent with the existence of a mirror neuron system in humans, but they cannot yet furnish conclusive proof. There has, however, also been a report of single neuron activity recorded from human neurosurgical patients that has demonstrated mirror neuron activity [32]. Recordings were focused on medial frontal cortex and temporal lobe structures, and show the extensive nature of the mirror neuron system. Unfortunately, neither of the premotor or posterior parietal areas so heavily investigated in monkeys were available for study in these patients.

Central to being able to interpret human fMRI studies of the mirror neuron system is understanding the relationship between the BOLD signal in human and mirror neuron activity in macaque monkey. To this end, monkey fMRI studies have now demonstrated significant activity during action observation in regions where mirror neurons have been previously reported [33, 34]. These monkey imaging studies have taken advantage of enhancing the neurovascular responses with an iron-based (MION) contrast agent. As with the vast majority of human fMRI studies, however, there is difficulty in relating these results to mirror neurons, in that they only employ an action observation condition and have no action execution condition. This makes it difficult to calibrate the activity changes in observation to those in execution, and also raises the possibility that sensory responses other than mirror responses contribute to the neurovascular changes (see Introduction).

One possible way of attributing the fMRI response to a single neuronal population, such as mirror neurons, is to use fMRI adaptation, or repetition suppression. This is a neuroimaging tool that has been adopted to identify neural populations that encode a particular stimulus feature [35]. The logic behind fMRI adaptation is that neurons decrease their firing rate with repeated presentations of the stimulus feature that those neurons encode. By extension it has been argued that the BOLD signal will also decrease with repeated presentations. It has been argued that areas of the cortex that contain mirror neurons should show fMRI adaptation both when an action is executed and subsequently observed, and when an action is observed and subsequently executed. This is because the stimulus feature encoded in mirror neurons is repeated irrespective of whether the action is observed or executed [36].

The results of such studies have produced mixed results. Of the five studies using this technique published to date [36, 37, 38, 39, 40], only three have demonstrated significant fMRI adaptation consistent with the presence of mirror neurons in the human brain [38, 39, 40]. One possible explanation for the mixed results is that humans do have mirror neurons, but that they do not alter their pattern of activation when stimuli that evoke their response are repeated. Indeed a recent study [9] has shown some evidence that mirror neurons may not alter their firing rate during repetitions of the same action; however, in this work the neuronal activity represented in the local field potential (LFP) did modulate with repetition. Further work is clearly required to determine why the BOLD signal in humans and the LFP in monkeys do adapt with repetition, while the evidence to date suggests that mirror neurons may not.

Great care must be taken when comparing the results from human and monkey studies. Specifically, readers must pay careful attention to the difference in the level of inference between the different modalities. The majority of human neuroimaging studies report significant results at the population level where the variance is estimated across subjects. This is in contrast to the studies reporting mirror neurons in macaque monkeys, where the aim is to test whether individual neurons show a consistent modulation of firing rate during periods of action observation and execution. Here the inference is closer to the analysis of fMRI at the single subject level. Therefore, when it is reported that 30% of neurons in any region were significantly modulated during both action observation and execution this does not mean that the remaining 70% do not modulate at all. Rather, it means there was not sufficient statistical evidence that these neurons displayed mirror activity. Indeed it is quite possible that when tested at the population level, the neurons that are non-significant at the single neuron level could be significantly modulated when observing an action.

The point here is that care must be taken when arguing that ‘only’ X% of neurons in any brain region are mirror neurons. The ‘only’ implies that the remaining neurons are not significantly modulated in any way during action observation. This is not a valid inference as to do so would be to accept the null hypothesis. This may be particularly problematic for cortical regions where responses in individual mirror neurons are relatively weak, such as in M1.

It is often assumed that mirror neuron activity during action observation is driven, bottom-up, by the visual (or auditory) input. The review of mirror neuron discharge presented here provides evidence that this is, at best, an incomplete description of mirror neuron firing. We now know that mirror neuron firing rates are modulated by view point [7], value [6] and that they even discharge in the absence of any visual input [23]. This suggests that mirror neurons can be driven or modulated top-down by backward connections from other neuronal populations. Indeed, the requirement for such top-down input to regions containing mirror neurons was realized by Jacob and Jeannerod [41], who argued that it was impossible for a mirror neuron system driven uniquely by the visual input to correctly infer an intention from an observed action if two or more different intentions would generate the same action. The fact that mirror neurons can be driven by backward connections is consistent with recent predictive coding models of mirror neuron function [42, 43, 44]. Within this framework, mirror neurons discharge during action observation not because they are driven by the visual input but because they are part of a generative model that is predicting the sensory input. This framework provides a theoretical account of mirror neuron activity that resolves the one-to-many mapping problem described by Jacob and Jeannerod [41] and is consistent with top-down modulation of mirror neuron firing rates.

Concluding Remarks

The discovery of mirror neurons has had a profound effect on the field of social cognition. Here we have reviewed what is currently known about mirror neurons in the different cortical areas in which they have been described. There is now evidence that mirror neurons are present throughout the motor system, including ventral and dorsal premotor cortices and primary motor cortex, as well as being present in different regions of the parietal cortex. The functional role(s) of mirror neurons and whether mirror neurons arise as a result of a functional adaptation and/or of associative learning during development are important questions that still remain to be solved. In answering these questions we will need to know more about the connectivity of mirror neurons and their comparative biology across different species.

Acknowledgements

J.K. and R.N.L. were both funded by the Wellcome Trust, London, UK. We would like to thank Alexander Kraskov for helpful comments on an earlier version.


References are available at the Current Biology page.

From Frontiers in Psychology: Consciousness Research, this recent article looks at the nature of time in task performance in lucid dreams vs. waking space. They found that any motor task such as squats (previous study), walking, or gymnastics takes more time in lucid dreams than in the waking world.

Among the questions raised by this study:

Longer durations in lucid dreams might be related to the lack of muscular feedback or slower neural processing during REM sleep. Future studies should explore factors that might be associated with prolonged durations.

I would wager that a significant portion of the difference in time is a result of the lack of muscular feedback in performing the motor task.

Full Citation:
Erlacher D, Schädlich M, Stumbrys T andSchredl M. (2014, Jan 16). Time for actions in lucid dreams: Effects of task modality, length, and complexity. Frontiers in Psychology: Consciousness Research; 4:1013. doi: 10.3389/fpsyg.2013.0101

Time for actions in lucid dreams: effects of task modality, length, and complexity

Daniel Erlacher [1], Melanie Schädlich [2], Tadas Stumbrys [2], and Michael Schredl [3]

1. Institute of Sport Science, University of Bern, Bern, Switzerland
2. Institute of Sports and Sports Sciences, Heidelberg University, Heidelberg, Germany
3. Medical Faculty Mannheim, Central Institute of Mental Health, Heidelberg University, Mannheim, Germany

The relationship between time in dreams and real time has intrigued scientists for centuries. The question if actions in dreams take the same time as in wakefulness can be tested by using lucid dreams where the dreamer is able to mark time intervals with prearranged eye movements that can be objectively identified in EOG recordings. Previous research showed an equivalence of time for counting in lucid dreams and in wakefulness (LaBerge, 1985; Erlacher and Schredl, 2004), but Erlacher and Schredl (2004) found that performing squats required about 40% more time in lucid dreams than in the waking state. To find out if the task modality, the task length, or the task complexity results in prolonged times in lucid dreams, an experiment with three different conditions was conducted. In the first condition, five proficient lucid dreamers spent one to three non-consecutive nights in the sleep laboratory. Participants counted to 10, 20, and 30 in wakefulness and in their lucid dreams. Lucidity and task intervals were time stamped with left-right-left-right eye movements. The same procedure was used for the second condition where eight lucid dreamers had to walk 10, 20, or 30 steps. In the third condition, eight lucid dreamers performed a gymnastics routine, which in the waking state lasted the same time as walking 10 steps. Again, we found that performing a motor task in a lucid dream requires more time than in wakefulness. Longer durations in the dream state were present for all three tasks, but significant differences were found only for the tasks with motor activity (walking and gymnastics). However, no difference was found for relative times (no disproportional time effects) and a more complex motor task did not result in more prolonged times. Longer durations in lucid dreams might be related to the lack of muscular feedback or slower neural processing during REM sleep. Future studies should explore factors that might be associated with prolonged durations.

Introduction

The question of time in dreams is frequently debated in science, philosophy and recently also by Hollywood film makers. For instance, in the movie Inception (Nolan and Thomas, 2010), dream time runs much slower than real time, 5 min of real time equaling 1 h of dream time. The idea, which inspired Christopher Nolan, the director of Inception, that time is scaled down during dreams, can be traced back a century and a half to the work of the French scholar Alfred Maury (1861), who was convinced that dreams are created at the moment of waking up. He based this assumption on a subjectively long-lasting dream about the French Revolution, at the end of which the dreaming Maury was to be beheaded under the guillotine. When he was roughly awoken by a piece of his bed (la flèche de mon lit) which had fallen on his neck, Maury assumed that the whole dream had been created at that very moment, leading up to the guillotine scene.

Maury's dream explanation led to the so-called Goblot hypothesis. In 1896 the French logician Edmond Goblot (1896) proposed that remembered dreams occur during the process of awakening and that a difference exists, therefore, between the time experienced in a dream and the time which actually passes while the dream is taking place. Hall (1981) tried to find evidence to support the Goblot hypothesis by showing that stimuli of a sleeper's surrounding as well as internal stimuli, such as hunger, were represented in the dreams of his subject who had recorded his dreams for two years. While such correspondence was found to some extent, Hall admitted himself that this does not prove that these dreams are generated during awakening, as external and internal stimuli “… are or may be present while we are asleep or before we go to sleep” (Hall, 1981, p. 245). In this approach the assumptions concerning time in dreams were indirect implications of a hypothesis on the origin of dreams in general. The idea that dreams are instantaneous memory insertions experienced at the moment of awakening also plays a major role in philosophical debates, for example in Dennett's cassette-theory of dreaming (Dennett, 1976).

A few years after the discovery of rapid-eye movement (REM) sleep and its initial association with dreaming (Aserinsky and Kleitman, 1953), Dement and Kleitman (1957) explored more precisely the relationship between REM sleep and dream activity. In one of their experiments, they wanted to demonstrate the relation between the lengths of periods of rapid eye movements and the subjects' estimations of how long they had been dreaming. In their study, participants were awakened randomly, either 5 or 15 min after REM onset, and were then asked if they had dreamed 5 or 15 min. In 92 out of 111 awakenings (83%) the participants judged correctly. The authors also found a correlation between the elapsed amount of time and length of dream reports (r = 0.40 to r = 0.71). These results were replicated by other researchers (e.g., Glaubman and Lewin, 1977; Hobson and Stickgold, 1995) and nowadays it is a widely accepted hypothesis that subjectively experienced time in dreams corresponds with the actual time. Yet, a study conducted by Moiseeva (1975) found that in dreams with a complex and bizarre structure or in very emotional dreams, time can be perceived as flowing much faster, exceeding the absolute time span of a dream by 2–10, 25–50 or even 100 times.

While in regular dream studies, this correspondence can only be explored on a correlational basis and retrospectively, a completely different approach opens when conducting studies with lucid dreamers. A lucid dream is defined as a dream during which dreamers, while dreaming, are aware they are dreaming (LaBerge, 1985). Lucid dreams are considered to be mainly REM sleep phenomena (LaBerge, 1990). Lucid dreamers can consciously influence the dream content and are thus able to carry out prearranged tasks while dreaming (e.g., Fenwick et al., 1984; Erlacher and Schredl, 2008a, 2010). In order to mark events or actions in a lucid dream, lucid dreamers can produce a specific pattern of eye movements (e.g., left-right-left-right) that can be objectively identified on an electrooculogram (EOG) recording (cf. Erlacher et al., 2003). Lucid dreams are especially useful for studying time intervals in the dream state because the beginning and end of a certain action can be marked with eye signals while the sleep is recorded using standard polysomnography.

In general, lucid dream studies conducted in sleep laboratories demonstrated that a certain time is needed during the recorded REM period. However, only two studies explored time in lucid dreams explicitly. In a pilot study, LaBerge (1985) demonstrated that the time interval for counting from one to ten in a lucid dream is about the same compared to that of wakefulness. Erlacher and Schredl (2004) investigated the duration of a sequence of squats (deep knee bends) compared to what would have been necessary in wakefulness. Five participants performed the following task both in wakefulness and while dreaming lucidly: Counting five seconds, performing ten squats and counting five seconds again. By means of eye signals, the durations of each counting or squat sequence could be determined and compared to the duration of waking performances. While there was no significant difference between wakefulness and dream state for the counting intervals, participants required about 40% more time for performing squats in lucid dreams than in the waking state. This finding contradicts the results of prior studies which supported equivalence of dream time and physical time.

Different explanations can be used to explain why more time was required for performing squats in the dream state. Firstly, there might be a difference between the task modalities. For example, tasks that involve an activation of the body concept in the dream could require more time due to a more complex simulation of this body schema. Secondly, there might be a difference due to the task duration: In the study described above by Erlacher and Schredl (2004), the motor task (M = 17.84 s, SD = 6.8) lasted almost three times as long as the counting task (first counting: M = 6.26 s, SD = 1.7; second counting: M = 6.48 s, SD = 1.0), when measured in wakefulness. Therefore it might be possible that longer tasks generally lead to increased durations in the dream state. Further, if there is indeed a need for more complex simulation to take more time in the dreaming state, then more complex actions in the dream should also lead to longer durations.

In the present study we conducted further experiments to explore the effects of task modality (involving motor activity vs. not involving motor activity), length (intervals of 10, 20, or 30 s/steps), and complexity (simple motor task vs. complex motor task) on task durations in lucid dreams. The durations of three different tasks were compared in wakefulness and in lucid dreams: counting, walking and a gymnastic routine.

Materials and Methods 

Participants

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.

TABLE 1


Table 1. Participants characteristics.

Experimental Conditions

The task descriptions for the three conditions:

Counting

For the counting task, participants had to count from 1 to 10, from 1 to 20, and from 1 to 30 at their own regular pace. During counting, participants were asked not to move (see Figure 1 as an example).

FIGURE 1


Figure 1. Experimental protocol for the lucid dream task (counting).

Walking

For the walking task, participants had to walk 10 steps, 20 steps, and 30 steps at their own regular pace.

Gymnastic routine

The gymnastic routine consisted of four consecutive elements starting in an upright position with feet together. Participants were instructed to count along while performing the elements (see Supplement 1):

Count 1, 2: Straight jump, landing with feet apart to the left and right, straight jump, putting feet together again

Count 3, 4: Straight jump, landing with feet apart to front and back, straight jump, putting feet together again

Count 5, 6, 7, 8: roll forward, standing up

Count 9, 10: Straight jump with half turn (180°)

For the counting and walking task, participants performed the task at their own regular pace. The gymnastic routine was developed to match the walking 10 steps condition regarding the task duration in wakefulness. The task was presented by the experimenter and the participants were asked to perform the task at the same speed and pace.

Sleep Recordings

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).

Procedure

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).

FIGURE 2


Figure 2. A sample of one correctly signaled lucid dream for the counting task. Five LRLR eye signals are depicted. The interval between two LRLR eye signals corresponds to the counting interval (gray area).

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.

Excluded Data

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:

● One or more LR eye signals were not detectable in the recording (counting: n = 2; walking: n = 3; gymnastic routine: n = 5)
● An element of the task was skipped or the participant was unsure about having performed one or more of the elements (gymnastic routine: n = 1)
● The participant stated in the dream report that he or she had imagined the performance rather than carried it out “physically” (gymnastic routine: n = 2)
● The dream report showed that there was a delay between eye signal and task performance, e.g., one participant stated in the protocol that she had hesitated for a moment between the second eye signal and the start of the motor routine to recall the exact sequence of the task (gymnastic routine: n = 1)
● The dream content directly influenced the time of the task performance (walking: n = 1; gymnastic routine: n = 1).

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).

Statistical Analysis

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.

Results 

Absolute and Relative Times for Counting, Walking and Gymnastics

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 3


Figure 3. Absolute durations for the counting task during wakefulness and lucid dreaming (Labels: e.g., P1m28 = Participant 1, male, 28 years. ∗Participants of the counting task also completed the walking task). 

FIGURE 4


Figure 4. Relative durations for the counting task during wakefulness and lucid dreaming (Labels: e.g., P1m28 = Participant 1, male, 28 years. ∗Participants of the counting task also completed the walking task).

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 5

Figure 5. Absolute durations for the counting task during wakefulness and lucid dreaming (Labels: e.g., P1m28 = Participant 1, male, 28 years. ∗Participants of the counting task also completed the walking task). 

FIGURE 6

Figure 6. Relative durations for the counting task during wakefulness and lucid dreaming (Labels: e.g., P1m28 = Participant 1, male, 28 years. ∗Participants of the counting task also completed the walking task).

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.

FIGURE 7

Figure 7. Absolute durations of the gymnastic routing during wakefulness and lucid dreaming (Labels: e.g., P1f25 = Participant 1, female, 25 years). 

Comparison of Absolute Times between Wakefulness vs. Lucid Dream State

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).

TABLE 2


Table 2. Comparisons of times in wakefulness and lucid dreaming. 

Comparison of Walking 10 Steps vs. Gymnastic Routine

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.

FIGURE 8


Figure 8. Means and standard deviations for walking 10 steps and the gymnastic routine during wakefulness and lucid dreaming.

Discussion

In this study, longer durations were observed for all types of tasks in lucid dreams as compared to those when awake. The greatest increase in time was for walking (52.5%) while the lowest increase was for gymnastics (23.2%). The increase for counting was 27%, but did not reach statistical significance. The differences in time, however, were observed only for the absolute durations of the task, but not for the relative durations.

Before discussing the results, some limitations of the present study should be acknowledged. One of the biggest limitations is the small sample sizes. Small sample size is always related to statistical drawbacks because it is hard to determine if the data meet all prerequisites for parametrical testing (e.g., normality). In order to account for such statistical problems we, firstly, concentrated on presenting sufficient descriptive statistics and, secondly, ran additional non-parametric tests (Wilcoxon test). The obvious advantage of t-tests is that effect size (Cohens d) and test power can be calculated and therefore those results are presented in Table 2. Because in this study effect sizes are large (d > 0.8) and test power ranges from 0.5 to 0.8 the probability for type II error is high (as in the case of counting).

Increasing sample size in lucid dream studies is not easy because the enrolment of proficient participants is always complicated. In a representative survey by Schredl and Erlacher (2011) it was shown that about 50% of the population experienced at least one lucid dream, however only 1.2% have lucid dreams on a very frequent basis (e.g., several times a week) which is necessary for sleep laboratory studies. Further, in addition to becoming lucid, participants also need to remember the task, accomplish it, and produce unambiguous eye signals. A recent survey of lucid dreamers (Stumbrys et al., in press) showed that lucid dreamers are able to remember their waking intentions in lucid dreams in only about half of the occasions and only less than half of those remembered intentions can be successfully accomplished in lucid dreams (failures most often occur due to awakening or hindrances within the dream environment). This seems to be borne out by our own study: Recall that half of the data sets had to be excluded because dreamers failed to carry out the task.

Next, the sleep recordings for the present study were conducted over the period of several years and the electrode montage has slightly changed over the time. The first two conditions (counting and walking) were recorded in accordance with the guidelines by Rechtschaffen and Kales (1968), while the third condition (gymnastic) was recorded in accordance with the American Academy of Sleep Medicine (2008) guidelines.

It should also be mentioned that in the present study lucid dreams were used to explore a special feature of a motor routine and that the results and conclusion should not be generalized to “the dream state” as a matter of course. Dreams in general—referring to REM dreams—also include non-lucid dreams. An EEG study by Voss et al. (2009) indicated that there might be a difference between lucid and non-lucid REM sleep concerning frontal lobe activation. These findings are supported by Dresler et al. (2012) who demonstrated in an EEG/fMRI study that during the lucid dream state a network of different brain areas appear to be reactivated which are normally deactivated during REM sleep (including prefrontal, occipito-temporal cortices, precuneus, cuneus, parietal lobules). These studies do not indicate differences between lucid and non-lucid dreams concerning motor activity per se. However, we cannot simply exclude such a difference a priori. Future studies using EEG/fMRI recordings should also investigate motor activation during non-lucid dreams, based upon the correlation of activation patterns and reported motor activity.

It is also worth mentioning that in our study the counting and walking task was performed at the participants' own regular pace, e.g., counting to 10 did not match 10 s of physical clock time (see also Table 2). LaBerge (1985) for example explicitly trained his participants to estimate a specific interval of time as accurately as possible, namely 10 s by counting “One thousand and one, one thousand and two, … one thousand and ten” at a rate attempting to match 10 s of physical clock time. In our study for the counting and walking condition, we did not intend to match the lucid time durations exactly to physical clock time (e.g., 10 s). This allows participants to do the task at their own pace and has the advantage that they don't have to pay attention to this additional demand of concentrating to match a certain time interval. However, for the gymnastic routine the participants were trained to match the walking 10 steps condition regarding the task duration in wakefulness.

Effects of Task Modality

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.

Effects of Task Length

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.

Effects of Task Complexity

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.

Explaining Extended 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.

Implications for Sports Science

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).

Future Directions

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.

Conclusion

In summary, the present study confirms the findings of Erlacher and Schredl (2004) that motor actions lead to prolonged durations in lucid dreams. The findings for the durations of cognitive actions (without motor activity) are as yet inconclusive. The relative time structure of motor tasks that last longer in the dream state than in wakefulness do not result in disproportional task durations in the dream state. Lucid dreams, therefore, can be successfully applied for motor skill practice in sports, music and other areas. Prolonged durations might be related to the lack of muscular feedback or slower neural processing during REM sleep. Future studies should explore factors that might be associated with prolonged durations (e.g., level of perceived effort, continuous vs. discrete tasks, motor task with counting vs. without counting) and try to incorporate physical time intervals within the dream by external auditory signals (e.g., implementing audio cues as start and stop signals).

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Acknowledgments

This project was funded by the BIAL Foundation, Portugal (Grant 72/06).

Supplementary Material

The Supplementary Material for this article can be found online at: http://www.frontiersin.org/journal/10.3389/fpsyg.2013.01013/abstract

References are available at the Frontiers site.