Let’s do the time warp again – embodied learning of the concept of time in an applied school setting

ABSTRACT Embodied Cognition approaches suggest that movements influence the understanding of abstract concepts such as time. It follows that moving the arms as watch hands should boost children’s learning to read the clock. In a school setting, we compared three learning conditions: an embodied (movement) condition, an interactive App condition, and a text condition. Age, self-reported enjoyment, and group size were controlled. In a clock-time-test, the embodied condition resulted in better performances than the mean of the other conditions in small, but not in large groups. This innovative, theory-informed approach may advance learning of abstract concepts in children.


Introduction
From an embodied cognition perspective, our ability to build conceptual knowledge of the world is based on the fact that (and how) we move with our body and its perceptual system in and interact with the world (Shapiro, 2011). One of the basic tenets of embodied accounts of cognition therefore is that a concept arises by associating perceptual, sensorimotor, and mental processes in a coherent and meaningful manner. For instance, the spatial concept "front" emerges from perceiving, for example, the front door, by moving to the front of a line, or by cognitively anticipating how a ball is being kicked to the front. This information from perceptual, sensorimotor, and mental processes is tied to the concept "front" and it is argued that the stronger this network, the more efficient is the reactivation of the learned information at retrieval (Barsalou et al., 2003).
Empirical research aiming at testing these theoretical ideas in education, thereby eventually sparking novel teaching methods, is scarce. A recent exception is a study by Kontra et al. (2015) in which the authors examined whether embodying a physical concept facilitates learning of the concept. Children who physically experienced the forces associated with angular momentum by tilting a set of wheels showed significantly better performances in a subsequent quiz about angular momentum than a control group. Further analyses confirmed that enhanced performance was related to the activation of sensorimotor brain regions when students later reasoned about angular momentum. Next to the evidence for advantages of embodied learning of abstract physical concepts (Kontra et al., 2015), there is also evidence for advantages of children's embodied learning of foreign language vocabulary (Toumpaniari et al., 2015), embodied learning of force-tracing behavior (Han & Black, 2011), and embodied learning of geography (Mavilidi et al., 2016). In parallel to research on the benefits of embodied learning, research on virtual learning methods such as using mobile tablets received increasing attention over the last years (e.g. Hung et al., 2015;Lindgren & Johnson-Glenberg, 2013). However, whether virtual teaching methods like mobile tablets facilitate or are detrimental to the learning process is still under debate (e.g. Rossing et al., 2012;Wang, 2017).
In the present study, we examined in an applied school setting to what extent different learning conditions ("moving the arms as watch hands" = embodied condition, "learning with an App" = App condition, "learning by reading a text on paper" = text condition) improve children's performance in a subsequent clock-time-test (see Appendix).
Based on the Conceptual Metaphor Theory (Lakoff & Johnson, 1999) we postulate that the emergence of the abstract concept of time is grounded in more concrete, spatial concepts. This groundedness of time is among other things reflected in our gestures: When we talk about something that is repeated various times, we possibly make a movement like a clock (e.g. arms going round and round). Based on Embodied Cognition Approaches and the Conceptual Metaphor Theory, embodying an abstract concept like time should hence facilitate the learning process of this concept. We therefore hypothesized that embodying time would benefit children's learning to read the clock in their second language more than learning with an App or reading on paper.

Method
In a within-subject design, we compared the impact of three different learning conditions with regard to children's understanding of time.

Participants
An a-priori power analysis revealed that a minimum of 22 children was required. We tested 37 children (two classes), of which 30 completed all three learning conditions (15 male, M age = 8.7 years, SD age = .73; 15 female, M age = 8.8 years, SD age = .41). After completion of the study, children received sweets for their participation. The experiment was approved by the ethical committee of the local institution. All parents provided written consent for their children's participation in the research. All children were free to withdraw from testing at any time.

Clock-time-test
To measure understanding of time in an encompassing way, a clock-time-test with six different types of tasks (e.g. "Draw the correct time", "Write the correct time, for detailed information", see Appendix) was applied. Children had eight minutes to work on the clock-time-test. A learning rate was calculated as the difference between the clock-time-tests completed before and after the respective condition and served as dependent variable. All children completed the clock-time-test four times (parallel versions). To create groups that were equally good in reading the time in English, the subsequent assignment to the learning groups was based on their score in the first clock-time-test: We assigned the child with the highest pre-test score to Group 1, the child with the second highest pretest score to Group 2, and so on, therefore all groups (small and large) were matched with respect to their prior knowledge of reading the clock in English. In the following sessions the groups rotated (Latin square randomized).

Learning conditions
The learning conditions (embodied condition, App condition, text condition) represented the independent variable. In all conditions, children learned to read the time in English. Four to five days passed between the learning conditions. In all three conditions, a poster with a clock (and no watch hands) was attached to the wall. To concentrate on learning effects of the learning conditions of interest, no prior training condition was applied.
In the embodied condition, one child received either an analog or written clock time on a card (randomized) and was asked to show this clock time to his/her peers by embodying it with the whole body. When the correct time was named, the next child proceeded. As we kept the learning time equal for all three learning conditions ( = 20 min), some children repeated the embodied learning procedure more often than others in the given time.
In the App condition, each child got a tablet, on which he/she played the App "Learning to tell Time", which was developed to teach children how to read the clock.
In the text condition, children read a text with explanations about how the time is expressed in English. The text also included pictures of clocks and the time written in digitals or letters beside it. In case a child got distracted, the experimenter asked him/her to pay attention to the text until the learning time was over.

Control variables
As we had three different learning conditions, both classes were divided into three groups (= six groups in total). Due to practical reasons the group sizes differed. Small groups consisted of three to four children (n = 3, 4, 4), large groups consisted of six to seven children (n = 6, 6, 7). Most studies have reported that groups with small size tend to perform better than larger groups (Kooloos et al., 2011). Group size might impact in particular the embodied condition, because the group scenario in the embodied condition implied a higher intensity (e.g. more repetitions of moving the arms as watch hands) of the manipulation. Group size is unlikely to have had an impact on the text condition and the App condition because each child got his/her own text and tablet. To control for possible modulations of learning effects due to group size, we included group size as control variable. In addition, age and self-reported enjoyment during the learning conditions were included as control variables, as both are reported to potentially affect learning outcome (Birdsong, 1999). Children indicated their enjoyment after each learning condition on a Visual Analogue Scale (i.e.: "How much fun did you have when learning the clock with the App?"). The Visual Analogue Scale is a psychometric measuring instrument. It consists of 10-cm lines anchored at the ends by words that define the bounds of the question of interest (i.e. "No fun at all", "A lot of fun").

Experimental design
A linear mixed model analysis was computed, with a random intercept for participants and a stepwise integration of fixed effects (condition, enjoyment, age, group size, condition*enjoyment, condition*age, condition*group size). The models were compared using Likelihood ratio tests. Post hoc tests were calculated by comparing each mean with the overall mean in the small/large groups (p-value adjustment: fdr = false discovery rate method, Benjamini & Yekutieli, 2001). Visual inspection of residual plots did not reveal any deviations from normality.

Results
Results did not reveal a main effect of condition, but a significant interaction between condition and group size χ²(1) = 16.6, p = .002, r 2 = .18. Post hoc tests revealed that in small groups, participants had significantly more correct items in the clock-time-test after the embodied condition (M Embodied, small group = 4.8) compared to the other conditions (M App, small group = 1.7, M Text, small group = -.9), t.ratio(28) = 3.10, p = .03, estimate = 3.24, Cohen's d = .87, whereas in large groups there were no differences between conditions (see Figure 1).
Including age did not improve the model. There were no other significant differences between conditions. The self-reported enjoyment (based on the Visual Analogue Scale) was higher in the App condition (M enjoyment = 9.26, SD enjoyment = 1.61) than in the other conditions (Embodied: M enjoyment = 8.40, SD enjoyment = 1.23; Text: M enjoyment = 7.04, SD enjoyment = 2.23). However, including self-reported enjoyment did not improve the model.

Discussion
The aim of the study was to examine whether embodying an abstract concept (i.e. time) benefits the learning process of that particular concept more than interacting with an app or reading a text on paper. The main result was that this was true for small, but not for larger learning groups. Further, despite children's self-reported enjoyment indicating that they enjoyed the App condition most, the learning benefits were largest in the embodied condition. Given the limited number of studies in applied school settings and the exploratory nature of our study caution is demanded when interpreting this finding. However, with respect to the transfer of theoretical embodied cognition assumptions to a realistic implementation at school, this result may motivate researchers as well as teachers to use embodied methods while taking group size as a potential moderator into consideration. Another factor coming along with a smaller group size is the number of movement repetitions. In small groups, children showed the time by moving their arms as watch hands more often than in large groups. Embodied learning research is often conducted without specific assumptions about the necessity of minimum number of movements (repetitions) required to show an effect. As a consequence, the reported embodied learning effects across different studies may be difficult to compare. The present study might be considered as an initial step towards a reflected analysis of the number of movement repetitions required to increase the learning process in embodied research settings as well as in applied educational settings.
Furthermore, as alluded to by Nobel laureate Richard Feynman in his commencement address 1974 at Caltech, the core of the scientific method is to rigorously identify all the relevant variables that help to explain the phenomena it tries to unravel. Yet, as is often the case in psychological experiments including ours, there is a risk of oversimplification by pre-selecting only a few (perhaps the most obvious) variables at the expense of missing out on other relevant variables that may also explain part of the variance in the observed behavior. Having said this, we cannot rule out that other variables may have influenced our findings, such as the voices and behavior of the teachers. We therefore recommend that future research needs to carefully take other potentially influencing variables into account. In addition, the present study might contribute to the debate on the cautious implementation of electronic devices in schools. Although our results showed an advantage of embodied learning over learning with a tablet, learning with electronic devices could also include embodied tasks (see considerations about computers and "syntonic learning" by Papert, 1993).
There are some limitations in the present study coming along with the fact that we aimed to realize a standardized, within-subject design within an applied school setting. First, although we conducted an a-priori power analysis, measuring more participants is necessary to confirm the robustness of the effect. Second, we cannot disentangle if the reason for the increased learning rate in the embodied condition was based on perceptual (= observing other children embodying the time) or motor (= embodying time oneself) or a combination of both processes. Nevertheless, the fact that the effect was only observable in small groups speaks in favor of movement processes causing the effects, because children in both groups observed the same amount of children embodying time.
To conclude, although future research is necessary to prove our findings robust, the integration of embodied learning methods in educational settings seems to be a promising approach to enhance learning outcomes in children. Further research may focus on qualitative data to explain the changes identified and quantifying the learning effects of embodying abstract concepts such as time, by for example systematically varying the number of movement repetitions.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Notes on contributors
Dr. Markus Raab (1968) is Full Professor and Director of the Department of Performance Psychology, Head of the Institute of Psychology at the German Sport University as well as Research Professor at the London South Bank Uni-versity, UK. He is president of the current Managing Council of FEPSAC and served in other associations such as asp and EFPA. He was awarded by the DOSB, asp, FEPSAC, DGPs, ECSS, Springer Publisher, NCA, NRW state and the University of Heidelberg. He serves (or has served) as an editor (in-chief, associate, section) for ZfS, PSE, IJSEP, SEPP, Frontiers in Psychology, Latin American Journal on Sport and Exercise Psychology. His main areas of interest are exercise psychology, embodied cognition, motor learning and motor control. After predicting 2017 research in sport psychology until 2050 he currently tries to self-fulfil that prediction.

Rouwen Cañal-Bruland
Research Interests: • Motor learning and motor control • Implicit and explicit motor learning