Exploring in a climbing task during a learning protocol: a complex sense-making process

In perceptual-motor learning, constant and variable practice conditions have been found to have differential effects on learners’ exploratory activity and their ability to transfer their skills to novel environments. However, how learners make sense of these practice conditions during practice remains unclear. This study aimed to analyse learners’ experiences of different practice conditions during a climbing learning protocol and to examine how these experiences might further inform learners’ exploratory activity. Twelve participants assigned to either ‘Constant practice’, ‘Imposed Novelty’, or ‘Chosen novelty’ groups climbed a ‘Control route’ (i.e. a route common to all groups) and a ‘transfer route’ (i.e. a new route) before and after a ten session learning protocol. Descriptions of learners’ experiences during previews and climbs were collected using self-confrontation interviews. After identifying general dimensions via a thematic analysis, a hierarchical cluster analysis on these general dimensions allowed the identification of phenomenological clusters (PhC). The distribution of these PhCs was compared between the first and last learning sessions, the control and transfer routes, and the practice condition groups. We identified seven PhCs reflecting learners’ meaningful exploratory activity during the previews and climbs. Significant differences in the distribution of these PhCs were found between (i) the first and the last session, (ii) the control and the transfer route and (iii) the Chosen-novelty group and the other two practice groups. These results suggest that exploration is part of a complex sense-making process linked to practice conditions, which can be described by a joint analysis of the intentions, perceptions and actions.


Introduction
In radical embodied cognitive science (Newen et al., 2018;Seifert et al., 2020), learning new sporting skills is characterised as learners changing their interactions with their performance environment, thereby affecting their ability to deal with novel performance environments (Hacques et al., 2021a). These changes are referred to as an exploratory activity and are mostly investigated using a behavioural approach to examine learner-environment interactions from the experimenters' perspective (Button et al., 2020). In what follows, we present the main findings of motor learning studies that have investigated exploration from behavioural observations, and then describe how a phenomenological account can enrich our understanding of the processes at stake when learning. We then present an investigation of how learners make sense of their performance environment over the course of practice and show how different practice conditions can affect this sense-making process.

Studying exploration in motor learning: the major contribution of the ecological dynamics approach
Ecological psychology places interactions between an individual and his/her environment at the centre of action/perception/cognition through the concept of exploration (Gibson, 1988;Gibson, 1966). From this viewpoint, exploration is defined as the active process of revealing and picking up information to adaptively move in the environment.
In line with this conceptualisation of exploration, a series of studies in climbing has analysed how the exploratory activity of climbers changes with practice or expertise in terms of modality and temporal organisation and how these changes support skill transfer to new climbing routes (Hacques et al., 2021a;Orth et al., 2018b;Seifert et al., 2015Seifert et al., , 2018. A study by Orth et al. (2018b) showed that less experienced climbers always present intense exploratory activity by touching the handholds as they search to determine how best to grasp and/or use them, and at the hip level to regulate their body posture. In contrast, expert climbers explore only at the hip level, suggesting that visual information is sufficient for them to determine which actions need to be performed without the need for haptic information. In line with these results, Hacques et al. (2021b) showed that haptic exploration (i.e. touching the holds) and visual exploration (i.e. number of visual fixations) decreased with learning, and the entropy of the learners' gaze path decreased. According to these works, these changes do not imply that perceptualmotor exploration disappears with skill acquisition, but that exploration is reorganised to better specify the opportunities for action offered by the environment (i.e. the hold graspability and use-ability in climbing, Seifert et al., 2018).
The reorganisation of exploratory activity also affects how performers transfer their skills to a new performance environment. For example, the study of Orth et al. (2018b) suggested that, with practice, less experienced climbers were able to transfer their skills to new routes when they developed the ability to explore the handholds haptically while maintaining their balance. This exploratory behaviour enabled them to achieve a performance score (i.e. measure of climbing fluency) on a transfer test similar to that of the skilled climbers, although the skilled climbers displayed different behaviour with fewer exploratory touches of the handholds (Orth et al., 2018b). Thus, improving skill transfer does not necessarily require the development of a specific form of exploratory activity, but it instead requires practice conditions that foster learners' attunement to the affordances of their performance environment (Hacques et al., 2021a;Seifert et al., 2018).
Indeed, practice conditions have been shown to affect perceptual-motor learning and transfer, notably when they induce variability in the task constraints. Variability in practice conditions is generally induced by an experimenter (or a coach, or physical trainer, if outside of a laboratory) imposing a new task constraint to encourage learner exploration during practice (Ranganathan & Newell, 2013). The aim is to support the search of the perceptual-motor workspace and discover a wider range of solutions to achieve the task at hand (Button et al., 2020). However, recent studies have shown that learning benefits from an optimal ratio between exploration (i.e. trying a new solution) and exploitation (i.e. repeating a solution already performed) (Komar et al., 2019;Sidarta et al., 2022), suggesting that changing task constraints too frequently would be detrimental to learning. Furthermore, given that individual learning dynamics may differ importantly (Chow et al., 2008;Orth et al., 2018a), the rate at which task constraints are changed should also differ between individuals. Liu et al. (2012) proposed to give one group of learners the possibility to choose when to change a task constraint during practice (i.e. the initial speed of the ball in a rollerball task). In contrast, another group practised the same task, but the change in the task constraint was imposed by a success criterion set by the experimenters (i.e. the speed of the ball was reduced after each series of ten trials with eight task successes). At the end of training, the imposed variability group showed a lower improvement rate in the task than the chosen variability group. In addition, the success rate of the chosen variability group was maintained at 50% throughout practice, whereas the success rate of the imposed variability group was initially high and decreased by the end of the training. These results show that learners are able to optimally control the rate at which task constraints change, whereas imposed variability appears less effective when participants are not ready to engage with a new condition. A recent study examined the effect of self-controlled practice on visual-motor exploration during a climbing learning protocol (Hacques et al., 2022). In this study, the self-controlled group could choose when to practice on a new climbing route, while the rate of route change was imposed on another group. The imposed rate of change led the participants to develop a gaze behaviour adapted to the confrontation with a new climbing route. Self-controlled practice induced two distinct adaptations: some participants may have used their choice to challenge themselves during practice and changed their gaze behaviour similarly to the imposed variability group, while others may have remained in a comfort zone, as the change in their gaze behaviour was more similar to the change observed in a constant practice group who always trained on the same route. In any case, the way in which learners make sense of exploration under these practice conditions remains unresolved.
Overall, the research that has analysed perceptual-motor exploration from behavioural variables has made a major contribution to the literature by highlighting the role of behavioural (e.g. in locomotor tasks, visual-haptic behaviour) variability to characterise exploration in the learning process. The enrichment of this work has been pursued by studies to identify the meanings that actors construct as they explore the environment with which they are interacting. This line of research seems to be heuristic regarding the work done from the phenomenological approach for skill acquisition (Dreyfus, 2002;Ravn & Höffding, 2022;Høffding & Schiavio, 2021), including the enactive framework (Maturana & Varela, 1987;Varela et al., 1991).

Studying exploration in motor learning: the enactive approach
The enactive approach is grounded in phenomenological philosophy and the cognitive sciences Poizat et al., 2022;Stewart et al., 2010;Varela et al., 1991). This framework is designed to examine agents' sense-making as it emerges from their structural coupling with the environment (Di . In the enactive framework, exploration can be defined as "a fundamental way of making sense of oneself as coupled with the world" (Høffding & Schiavio, 2021, p. 811). More precisely, exploration is conceived as an autonomous active regulation that ensures the enactment of meaningful interactions with the world during the dynamics of activity (e.g. a learning activity) Di Paolo et al., 2017). Thus, in the enactive perspective, exploration is understood as a meaningful activity-involving behavioural and phenomenological dimensions-that emerges from an actor coupled with an environment as highlighted by Høffding and Schiavio's definition: "Exploration is a condition of possibility of meaning-making. In the minimal sense of the action of investigating an unknown area, explorative behaviours allow agents to put forward their concerned 'point of view'" (Høffding & Schiavio, 2021, p. 814).
Research from within this framework has investigated exploration in motor learning (Bermejo et al., 2020). For example, Bermejo et al. (2020) studied exploration in a sound sources localisation task under different conditions (i.e. passive/dynamic, normal/reversed sounds). They characterised exploration strategies from behavioural data (head movements) and experiential data (reports of participants' experiences). The results showed a transformation of exploration strategies for both behavioural and experience reports analysis. With learning, exploration strategies became more goal-directed (i.e. fewer head saccades and more performance-directed analytical strategies). They also found the emergence of individual exploration strategies in learning pathways.
Another study in a learning context focussed on sensemaking in the search for fluency in climbing, from the identification of a learner's intentions, perceptions and actions (Rochat et al., 2020). By articulating phenomenological accounts with performance indicators, this case study provided additional insights to the existing literature on the effect of practice conditions and practice duration on the emergence of exploratory activity. Indeed, the results showed that the learner's confrontation with a new route fostered an exploratory engagement aimed at constructing action sequences in order to be effective. Furthermore, this study also showed that with the learning and stabilisation of a functional plan of action (i.e. functional hand/foot chainings), the learner sought to challenge himself to improve his climbing performance, which contributed to the redefinition of what the authors called the 'phenomenological synthesis'. Exploration in this sense is lived as a new way of interacting with the environment, becoming sensitive to new sensations that are progressively refined in the course of these interactions.
Overall, the studies by Bermejo et al. (2020) and Rochat et al. (2020) highlighted the continuous nature of exploration during motor learning, as learners' lived experiences are constantly redefined through their actions, meaning that we never stop of exploring (Høffding & Schiavio, 2021). In addition, the study of Rochat et al. (2020) also highlighted the importance of considering the subjectivity of learners when learning a new skill. The authors described learning as 'enacting significations ' (p. 19).

The current study
Based on studies that have investigated the effects of practice conditions on perceptual-motor exploration, our study rooted in the enactive approach aimed to enrich the understanding of exploration. Our objective was to characterise the meanings made by the actor during his/her exploratory interactions with the learning environment. More specifically, we sought to characterise exploration from a phenomenological point of view in order to extract phenomenological syntheses, respecting the holistic dimension of activity (including lived experience) as defended by the enactive approach, that can reveal exploration macro-formats. The earlier work that investigated the learning dynamics at a phenomenological level showed that by determining learners' intentions, perceptions and actions, we are able to gain insight into how and why these learners transform their interactions with their practice environment (Rochat et al., 2020). In this vein, the present study further investigates the effect of different practice schedules (constant, variable and self-regulated) on how actors make sense when exploring their learning environment.

Participants
Twelve undergraduate students (9 males and 3 females; mean age = 19.6 ± 1.2 years) volunteered to participate in this study. All were enrolled in the Sports Sciences School of Rouen Normandy University and had no prior experience in climbing. They were thus considered to have the lowest grade skill level according to the International Rock-Climbing Research Association scale (Draper et al., 2015). The protocol was approved by the French National Agency of Research (ID: ANR-17-CE38-0006 DynACEV).

Protocol
The participants were randomly assigned to one of three groups. The protocol included ten sessions divided into two weekly sessions for 5 weeks. In the first group, called 'constant group (C)', participants systematically climbed the same 'control route', which was 525 cm high with 20 holds (i.e. 13 handholds and 7 footholds, Fig. 1). The second group was called 'imposed novelty (IN)'. In each session, these participants climbed three times on the control route and three (in sessions 1 and 10) to six (in sessions 2-9) times on variant routes prescribed by the experimenter. In sessions 2-9, they were confronted each time with a novel variant route. Nine variant routes-which differed from the control route-were designed on a 480 cm high wall. Each variant route was composed of 20 holds and differed from the others such that only the spatial layout of the handholds was manipulated without affecting the difficulty of the routes. The size and shape of the hand-and footholds were the same for all the routes. The last group, called 'chosen novelty (CN)', followed the same practice schedule as the IN group except that the participants had to choose at the end of each session whether they wanted to keep the same variant routes or have a novel variant route for their next session.
For all groups, the very first and very last trials of the protocol were performed on a 'transfer' route. The transfer route was 495 cm high and also had 20 holds (Fig. 2). Like the variants, the transfer route was designed by changing the spatial layout of the handholds without manipulating the difficulty of the route. Before each climbing session, participants had a 10-min warm-up on easy boulder routes in the bouldering area of the climbing gym. Before each climb, they had a 30-s preview that could be used completely or not. Then, the experimenter gave the following instructions: Use all the handholds in a bottom-up order, do not to use holds with both hands or both feet at the same time. In addition, find a way to climb the route as fluently as possible: that is, avoiding pauses and jerky movements.

Data collection
This research mobilised the methodological framework of the course-of-action research program (Poizat et al., 2022;Theureau, 2003) rooted in the enactive approach. In keeping with the pre-reflective consciousness hypothesis and semiotic hypothesis (Poizat et al., 2022), two types of data were collected: (1) audiovisual recordings of the participants while they were previewing and climbing each route and (2) phenomenological data concerning the lived experience of the participants during their preview and climb.

Audiovisual data collection
Audiovisual data were collected for each climb by recording the activity of each participant. Ascents were filmed using a wide-angle standing camera to see the participant during the preview and the climb.

Phenomenological data collection
Phenomenological data were collected immediately after each session with self-confrontation interviews (Rochat et al., 2020;Theureau, 2003) based on the video recordings of each preview and climb. These video recordings were used as past-activity traces to help the climbers re-enact their experience (Theureau, 2003). This means that they were invited to chronologically relive their meaningful experience throughout the climbing trials. Based on these video traces, the interviews consisted of asking them to comment on their activity during each preview and climb while avoiding retrospective judgments and generalisations. The interview prompts were especially aimed to document (1) the climbers' intentions ('What are you trying to do?'), (2) their actions ('What are you doing?'), and (3) their perceptions ('What is drawing your attention?'; 'What are you feeling?'). The interviews were conducted by two trained researchers who were experienced in conducting self-confrontation interviews with athletes from different sports. For all participants and all sessions and trials, a total of 120 self-confrontations were performed. Each interview lasted 45 min on average.

Data processing
Given the study objectives, the analysis focussed particularly on the trials common to the three learning groups for sessions 1 and 10 to detect potential transformations in the emergence of the phenomenological dimensions of exploration between the beginning and the end of the learning protocol (i.e. trials on the 'control route' and the 'transfer route'). Thus, eight trials per participant were analysed (i.e. four in session 1 and four in session 10; 96 trials for the 12 participants).
The data were processed in three steps. The first two steps followed the methodology of Rochat et al. (2020). First, the participants' courses of experience were reconstructed by identifying their actions, perceptions, and intentions. Second, a thematic analysis of the course of experience identified the general dimensions of the participants' experience. The third step consisted in identifying phenomenological clusters of exploration by applying a hierarchical cluster analysis to the general dimensions across trials. This last step aims at revealing macro-level exploration formats that could help practitioners' intervention.

Reconstruction of the courses of experience
The first step was to restore the course of experience of each participant and each climb. From the verbalisations in the self-confrontation interviews and the climbing activity recorded on the videos, this consisted of a semiotic labelling of (1) intentions, reported in response to the question: 'What is expected?'; (2) perceptions, reported in response to the question: 'What is meaningful?'; and (3) actions, reported in response to the question: 'What am I doing?' (e.g. as already done in Rochat et al., 2020, andSeifert et al., 2017a).

Identification of general dimensions in the courses of experience
The second step consisted of a thematic analysis to identify the general dimensions of intentions, perceptions, and actions (Rochat et al., 2020). We conducted this analysis to inductively find similarities in the climbers' intentions, actions, and perceptions (Braun & Clarke, 2006;Braun et al., 2016;Vaismoradi et al., 2013) in order to characterise the general dimensions that made up their courses of experience. The raw data of the intentions, actions, and perceptions were examined in detail and the detection of similarities among them helped identify the first-order themes, which were merged into second-order themes and then into general dimensions (third-order themes).
For example, 'aligned holds', 'snake shape' and 'zigzags' as first-order themes from the reconstruction of the learners' courses of experience were grouped into a secondorder theme called 'spatial arrangement of holds', which in turn was grouped into the general dimension of 'physical characteristics'.
These steps were jointly performed by the first, second and last authors, who were all trained in performing this type of phenomenological data processing. In cases of disagreement, the researchers re-watched the video recordings of the self-confrontation interviews and the audiovisual recordings and discussed what they had seen and heard until a consensus was found.

Identification of phenomenological clusters of exploration: a hierarchical clustering analysis (HCA)
The third step, a hierarchical cluster analysis (HCA), was used to classify all the analysed trials. The objectives were to identify association patterns between the general dimensions (i.e. intentions, perceptions, actions). The methodology of the HCA had four steps. First, trials were classified as follows: a value of 1 was indicated when the general dimension was present during the trial and a value of 0 when the general dimension was not present. Second, the dissimilarity matrix was obtained with the computation of binary distances between trials. Third, the Ward method was applied to aggregate trials into clusters by minimising the total withincluster variance (Ward, 1963). Fourth, the optimal number of clusters was assessed based on the dendrogram representation of the clustering. Specifically, the dendrogram was inspected to find the partitions that maximised the distance between nodes (i.e. we looked for a large vertical section of the dendrogram without nodes) (Landau et al., 2011). We then calculated the average within-cluster and betweencluster sum of squares of the identified partitions. The final partition was the one that minimised the ratio between these two measures.
Last, an analysis of the distribution (in percentage) of the general dimension of intentions, perceptions and actions for all clusters obtained in preview and climb was applied (see Tables 1 and 2). This analysis was used to label and define the phenomenological clusters. All treatments were performed with the NbClust 3.0 package under RStudio (Rstudio © 2021.09.1 build 372 that used R version 4.1.2., Boston, MA, USA).

Statistical analysis
Chi-square tests were performed to compare the distributions of the phenomenological clusters from previews and climbs between (1) the three groups of practice, (2) the first session of learning (i.e. session 1) and the last session of practice (i.e. session 10), and (3) the control route and the transfer route. The level of statistical significance was set at p < 0.05. The chi-square tests were performed on JASP software (JASP© 0.13.1, Amsterdam, The Netherlands).

Identification of the general dimensions of exploration
This section of the results describes the general dimensions in preview and climb obtained from the thematic analysis carried out for the three constituent dimensions of the participants' experience (for more details see online resource in the online resource supplementary materials). The general dimensions are illustrated by a few verbatims extracted from the participants' courses of experience.

Preview
The thematic analysis of the climbers' intentions the preview revealed four general dimensions: (1) designing sequences of actions, (2) 'reproducing sequences of actions', (3) 'climbing fluently', and (4) 'respecting the instructions'. The general dimensions are detailed in Table 3 of the online resource supplementary materials.
The general dimension of 'designing sequences of actions' refers to the participants' intentions to organise optimal discrete actions such as handhold or foothold positioning. As an illustration, they tried to avoid crossing their hands to ensure the design of an optimal handhold position: I try to do the preview from the bottom up so I can find a sequence that will avoid ending up with poor placement for this group of holds; I want to finish with my right hand here and then chain with the left hand so that I can grab this hold with the right hand without crossing hands (Participant from CR group, Session 1, Trial 1, Transfer route).
This general dimension was also related to the design of action sequences like same-hand grabbing of the next handhold or alternating movements. An example would be when they tried to avoid movement chainings like hand crossings in favour of grabbing two consecutive holds with the same hand or when they designed hand-foot coordination sequences: 'I have to find a way to put one foot off-centred to the right so I can take the left hand off-centred' (Participant from CR group, Session 1, Trial 1, Control route).
The general dimension of 'reproducing sequences of actions' refers to the participants' intentions to reproduce actions performed during previous climbs that they considered to be efficient. In this case, they tried to reproduce hand and foot sequences that felt fluid to them on either part of the route or the whole route: 'I plan to start again in the same way as on my first try and thus show that my earlier sequence was valid' (Participant from the CN group, Session 1, Trial 2, Control route); 'I'll do the same start as before with a left hand raise and then a cross, which worked well' (Participant from the CN group, Session 1, Trial 2, Control route).
The general dimension of "climbing fluently" refers to the participants' intentions to improve the fluency of their climbs. For example, to be more fluent, they tried to limit stops and avoid losing balance, rushing or being dynamic on the holds: Now I'm trying to climb like a cat, so I tell myself to go slowly, to be much more delicate on the foot holds, the rhythm is not going to be the same − it's a good technique to improve fluidity (Participant from the IN group, Session 10, Trial 2, Control route).
The general dimension of 'respecting the instructions' refers to the participants' intentions to remain in line with the requirements that defined the task: 'I have to find a sequence on this route that takes the instructions into account, especially the one about starting with both hands on the same handhold' (Participant from the CR group, Session 1, Trial 1, Transfer route).
The thematic analysis of the climbers' perceptions during the preview revealed three general dimensions: (1) 'physical characteristics', (2) 'functional characteristics', and (3) 'previous climbs'. The general dimensions are detailed in Table 4 of the online resource supplementary materials.
The general dimension of 'physical characteristics' reflects the participants' perceptions of the environment, particularly the physical characteristics of the route. For example, they were sensitive to the distance between holds, the layout of the holds, and even the shape evoked by the spatial arrangement of the holds: 'There's a certain pattern that is repeated on the route. There are small ladders of three holds that go up to the left, three that go up to the right, and so on' (Participant from the CR group, Session 1, Trial 1, Transfer route).
The general dimension of 'functional characteristics' reflects the participants' perceptions of the possibilities for action sequences offered by the spatial arrangement of the holds. For example, the learners sometimes recognised the spatial arrangement of the holds as helpful for their future climbs and thus had more positive perceptions of the climb.
On the contrary, they could perceive the route negatively, assuming that the spatial arrangement of the holds would make for particular difficulties: When I see the route, I think I'll have to do a lot of hand-crossings, but for me it's problematic, especially since there aren't many footholds and the holds are generally spaced out (Participant from the CN group, Session 1, Trial 1, Transfer route).
The general dimension of 'previous climbs' refers to the participants' memories of previous ascents related to the physical characteristics of the route or the sequences of actions. To illustrate, this general dimension was at times related to the perception of previous unpleasant stops 'I remember perfectly the three places on the route where I stopped; each time it was at the moment of relaunching to the top' (Participant from the CR group, Session 10, Trial 1, Control route).
The thematic analysis of the climbers' actions during the preview revealed four general dimensions: (1) 'miming one part of the route', (2) 'miming the complete ascent', (3) 'focusing on one part of the route', and (4) 'scanning the route'. The general dimensions are detailed in Table 5 of the online resource supplementary materials.
The general dimension of 'miming one part of the route' refers to the participants' simulating themselves grasping sequences of handholds. These actions were mainly directed at the beginning or end of the route: 'I mime the start sequence by including the places where I think I will make crossings' (Participant from the IN group, Session 1, Trial 1, Transfer route).
The general dimension of 'miming the entire ascent' refers to the participants' actions as they simulated hands sequences from the bottom to the top of the route to test the possibilities of sequences: I start directly by miming with my hands to try to find the easiest sequence and see if it's feasible or not; So, I keep miming the rest of the route and try to do a preview up to the top (Participant from the CR group, Session 1, Trial 1, Transfer route).
The general dimension of 'focusing on one part of the route' refers to the participants' actions as they concentrated on sections of the route considered difficult, especially handor footholds: 'This time I do my preview focussing on one of the two centre holds and I think of several possibilities at that spot, but I don't choose one' (Participant from the IN group, Session 1, Trial 2, Control route).
The general dimension of 'scanning the route' refers to the participants' actions when they looked at the whole route from top to bottom and from bottom to top, once and several times: 'I looked at the route from bottom to top and scanned it several times' (Participant from the IN group, Session 1, Trial 1, Transfer route).

Climb
The thematic analysis of the climbers' intentions during the climb revealed four general dimensions: (1) 'ensuring the correct execution of action chaining', (2) 'improving the timing of the climb', (3) 'maintaining balance while climbing', and (4) 'respecting the instructions'. The general dimensions are detailed in Table 6 of the online resource supplementary materials.
The general dimension of 'ensuring the correct execution of action chaining' refers to the participants' intentions to reproduce the hand and/or foot sequences planned during the preview. For example, they focussed on avoiding mistakes and performing a known sequence of hand actions: 'Now I'm doing what I'd planned, right then left, then right hand. It goes well' (Participant from the CR group, Session 1, Trial 3, Control route).
The general dimension 'improving the timing of climbing' refers to the participants' intentions to be efficient in their climbing. For example, they tried to keep their hands or feet on the holds for as short a time as possible or to avoid jerks or stops during the climb: I want to have my pelvis moving all the time, I want my arms and legs to work together, to be able to pull with my arms and push with my legs at the same time to shorten the downtime (Participant from the CR group, Session 10, Trial 1, Control route).
The general dimension of 'maintaining balance while climbing' refers to the participants' intentions to maintain a stable, well-oriented body while climbing to perform well. They used their feet to remain stable and focussed on maintaining the orientation of the body along a vertical axis: I try to keep my pelvis from moving to the sides as much as possible even though the track goes to the right and to the left; I try to keep it in the centre when I climb even if it goes to the right at the beginning (Participant from the IN group, Session 10, Trial 1, Transfer route).
As in the preview, the general dimension of 'respecting the instructions' refers to the participants' intentions to meet the task requirements, such as to be efficient and avoid putting both hands on the same holds: 'When I arrive to take this hold, I tell myself that I shouldn't take it with both hands because it's prohibited in the instructions' (Participant from the IN group, Session 10, Trial 1, Transfer route).
The thematic analysis of the climbers' perceptions during the climb revealed four general dimensions: (1) 'sensation 1 3 of being unbalanced', (2) 'sensation of being balanced', (3) 'sensation of efficient timing', and (4) 'sensation of perturbed timing'. The general dimensions are detailed in Table 7 of the online resource supplementary materials.
The general dimension of 'sensation of being unbalanced' refers to the participants' negative perceptions of, for example, a fall, discomfort, or a loss of body stability. They were also sensitive to hand-crossings or to unreliable footholds or discomfort: From the start I feel an imbalance towards the back and a feeling of discomfort in my feet [...]. Then, when I have to let go of my right hand in the middle of the track, I feel very unbalanced and uncomfortable on my feet (Participant from the CN group, Session 1, Trial 1, Transfer route).
The general dimension of 'sensation of being balanced' refers to their positive perceptions during the climbs. For example, these positive sensations could be felt in a specific way through the reliability of hand or foot support or in a global way in the sensation of a well-aligned body: 'It's the best trial of the session! My hips didn't move too much towards either side, except a little to the right at the beginning, I had the impression that the red light pretty much stayed in the middle' (Participant from the IN group, Session 10, Trial 1, Transfer route).
The general dimension of 'sensation of efficient timing' refers to body markers experienced as positive during the climbs. These markers were perceived either in the movement of the whole body or in certain sequences of hand-foot actions such as the feeling of power between the hands and feet: 'I feel that I put the tip of my feet on the holds and that helps me go fast […] My arms and legs are working at the same time to help me move quickly' (Participant from the CR group, Session 10, Trial 2, Control route).
The general dimension of 'sensation of perturbed timing' refers to the negative perceptions of the participants regarding efficiency during climbing. These feelings of inefficiency were perceived, for example, through the sensations of stops or jerky movements or through the selected holds that impeded the synchronisation of the hand-foot actions: I really have the feeling that I'm not moving forward, that my feet are preventing me from moving forward, that in fact it's as if I were climbing only with my hands; I have the impression that everything is complicated and it's taking forever (Participant from the CR group, Session 10, Trial 1, Transfer route).
The general dimensions are detailed in Table 8 of the online resource supplementary materials.
The general dimension of 'making errors' refers to unplanned actions experienced by the participants as errors. These significant errors concerned both footholds and handholds: 'I'm making a mistake here because I put a right hand on the handhold that is completely to the left and it makes me unbalanced' (Participant from the CR group, Session 10, Trial 1, Transfer route).
The general dimension of 'carrying out the planned actions' refers to the actions of the participants corresponding to the movements they had planned during the preview. These actions concerned the chains of both hand actions and hand-foot actions: I start as planned and continue kind of automatically until I finish without any problems. In fact, I do the same sequence as in the previous trial because it was good; so once again I put my knee against the wall and think that all I need to do is repeat what I did before to improve the fluidity (Participant from the CR group, Session 1, Trial 3, Control route).
The general dimension of 'improvising actions' refers to actions not planned by the participants during their preview. This dimension therefore refers to participants' actions that emerged during the climb without having been considered beforehand. These actions concerned either the hand or the foot chaining: 'Well, here, I'm climbing as best as best can because I don't have my feet! So, I make a jump to get my feet up and then I just finish however I can, it's crazy what I did there' (Participant from the CR group, Session 10, Trial 1, Transfer route).
The general dimension of 'modifying the planned actions' refers to the modifications in hand and foot actions that the participants made during the climbs. These modifications were intended to improve or disrupt their climbing fluency. For example, they used a hold with the other hand: I don't do what I had planned because I wanted to put my right hand to the right but instead, I put it to the left and got unbalanced, so I swing a bit like a pendulum to restart on the next hold (Participant from the CR group, Session 1, Trial 1, Transfer route).
These modifications could also disrupt their climbing fluency: I don't feel comfortable enough to start up again how I had planned, so I modified it to do a cross, except that I find myself in a very uncomfortable position... nevertheless I think that I didn't do so bad (Participant from the IN group, Session 1, Trial 1, control route).

Preview
Visual inspection of the dendrogram suggested partitioning the data into two or three clusters (Online resource supplementary materials, Figure 2). The lowest ratio of within-cluster and between-cluster sums of squares was obtained for the three-cluster partition (values of the ratio for two and three clusters were 0.64 and 0.62, respectively). Based on the distribution of the general dimensions in the three phenomenological clusters (Table 1), we respectively named the clusters 'Construction-directed exploration' (CE), 'Reconstruction by scanning the entire route' (RCS) and 'Reconstruction focussed on parts of the route' (RCP).
The CE cluster was defined as activities aimed at defining the orientation of actions based on the perception of the physical characteristics of the environment. CE was mainly characterised by the intention of 'designing sequences of actions' (86.8%), the perception of 'physical characteristics' (78.9%), and the action of 'scanning the route' (71%).
The RCS cluster was defined as activities to 're-construct' past actions by scanning the environment in order to perform well. RE was mainly characterised by the intention of 'performing' (80.0%), the intention of 'reproducing sequences of actions' (63.3%), the perception of 'previous climbs' (100%), and actions of 'looking at part of the route' (51.7%) or 'scanning the route' (93.3%).
The RCP cluster was defined as activities to 're-construct' past actions by focussing on parts of the climbing route to perform well. RE was mainly characterised by the intention of 'performing' (92.9%), the perception of 'previous climbs' (100%), and actions of 'looking at part of the route' (78.6%).

Climb
Visual inspection of the dendrogram suggested partitioning the data into two or four clusters (Online resource supplementary material, Figure 3). The lowest ratio of within-cluster and between-cluster sums of squares was obtained for the four-cluster partition (values of the ratio for two and four clusters were 0.77 and 0.70, respectively). Based on the distribution of the general dimensions in the four phenomenological clusters (Table 2), we respectively named the clusters 'Refined exploration' (RE) and 'Performance-directed exploration' (PE), 'Timing-focussed exploration' (TE) and 'Balance-focussed exploration' (BE).
The action 'carrying out the planned actions' was frequent in the four clusters (ranging from 63.6% to 95.8%).   Therefore, this general dimension has not been included in the following description of the cluster characteristics. The RE cluster included activities aimed at confirming the known direction of some of the actions and defining, during action, the direction of other potentially disruptive actions. It was mainly characterised by the intention of 'ensuring the correct execution of action chainings' (97.3%), the perception of 'sensation of being unbalanced' (62.2%), the perception of 'sensation of perturbed timing' (89.2%), and the action of 'improvising actions' (100%).
The PE cluster included activities that confirmed the direction of actions known to be effective. PE was mainly characterised by the intention of 'ensuring the correct execution of actions chainings' (88.5%) and the perception of 'sensation of efficient timing' (92.3%).
The TE cluster included activities that focussed on the timing of the actions. TE was mainly characterised by the intention of 'improving the timing of the climb' (95.5%), the perception of 'sensation of efficient timing' (72.7%), and the perception of 'sensation of perturbed timing' (63.6%).
The BE cluster included activities that focussed on balance during climbing. BE was mainly characterised by the intention of 'ensuring the correct execution of action chainings' (90.9%), the intention of 'maintaining balance while climbing' (100%), the perception of 'sensation of being balanced' (81.8%), and the action of 'improvising actions' (81.8%).

Distribution of the phenomenological clusters for the preview and the climb Preview
Overall, the CE, RCS, and RCP phenomenological clusters represented 36.6%, 31.2% and 29.2% of the trials, respectively. A marginal effect of learning groups on the distribution of the clusters was observed [X 2 (4, N = 96) = 8.839, p = 0.065]. We could notice that the distribution of the three clusters was homogenous in the Constant practice (CR) group (CE: 37.5%; RCS: 31.3%; RCP: 31.3%). In the Imposed-Novelty group (IN), the RCS cluster was less represented than the two others (CE: 37.5%; RCS: 18.8%; RCP: 43.8%). Finally, in the Chosen-Novelty group (CN), the RCP cluster was less represented than the two others (CE: 43.8%; RCS: 43.8%; RCP: 12.5%).
A significant difference between session 1 and session 10 in the distribution of the phenomenological clusters was observed [X 2 (2, N = 96) = 19.777, p = < 0.001], which indicated an effect of practice. The representation of the CE cluster dropped from 54.2% in session 1 to 25.0% in session 10, while the representation of the RCS cluster increased from 10.4% to 52.1% in session 10. The representation of the RCP cluster decreased from 35.4% in session 1 to 22.9% in session 10.
The three phenomenological clusters were also distributed differently in the trials performed on the control route and the transfer route, [X 2 (2, N = 96) = 42.423, p < 0.001]. This result indicates an effect of the route. Almost all the trials performed on the transfer route belonged to the CE cluster (95.8%), whereas on the control route, 20.8% of the trials belonged to this cluster. The remaining trials performed on the transfer route (representing the remaining 4.2%) belonged to the RCS cluster. On the control route, RCS and RCP clusters represented 40.3% and 38.9%, respectively.

Climb
Overall, the RE, PE, TE and BE phenomenological clusters represented 38.5%, 27.1%, 22.9%, and 11.5% of the trials, respectively. A significant difference in the distribution of the phenomenological clusters was observed between the three learning groups [X 2 (6 A significant difference between session 1 and session 10 in the distribution of the phenomenological clusters was observed [X 2 (3, N = 96) = 28.419, p < 0.001], indicating an effect of practice. Notably, the representation of the RE cluster dropped from 56.3% in session 1 to 20.8% in session 10 whereas the representation of the TE cluster increased from 2.1% to 43.8% in session 10. Proportion of the PE cluster increased from 25.0% in session 1 to 29.2% in session 10 and the proportion of the BE cluster decreased from 16.7% to 6.3% in session 10.
A significant difference in the distribution of the phenomenological clusters between the control route and the transfer route was observed [X 2 (3, N = 96) = 27.317, p < 0.001]. The main difference appeared to be the rare to non-existent representation of the PE and TE clusters on the transfer route with 4.2% and 0% of the trials against 34.7% and 30.6% on the control route. Conversely, the RE and BE clusters were more represented on the transfer route (70.8% and 25% respectively) than on the control route (27.8% and 6.9% respectively).

Discussion
The twofold aim of this study was to (a) characterise the meaningful exploratory activity of novice climbers at the beginning and the end of a 10-week learning protocol and (b) to investigate how different practice conditions (constant, variable-imposed or variable-chosen) would affect learner's exploratory activity on control and transfer routes. The general dimensions identified inductively from the learners' courses of experiences were brought together to identify phenomenological clusters that synthesised the learners' exploratory activity during the previews and climbs. Overall, our results revealed significant effects of the practice groups and routes on the distribution of the phenomenological clusters, in both the previews and climbs. In what follows we address the following points: (a) the effect of practice and route familiarity on exploration, (b) the effect of practice conditions on exploration, (c) the enrichment of the enactive definition of exploration, and finally (d) the methodological contributions to the enactive approach.

The effect of practice and route familiarity on exploration
Our results suggest that exploration during preview and climbing evolves with practice. At the beginning of training (i.e. session 1), the preview was characterised by two main types of exploration activity: one that aimed at defining a sequence of actions in relation to the characteristics of the environment (the CE cluster) and another aimed at reconstructing sections of the route (the RCP cluster). In addition, the CE cluster was systematically observed in the transfer route, suggesting that this mode of exploration seems to be characteristic of the discovery of a new climbing route. In contrast, relying on past experience to reconstruct parts of the route (the RCP cluster) or the entire route (the RCS cluster) is associated with a familiar route (i.e. the control route). However, these two types of exploration are differentially affected by practice, as RCS increased with practice while RCP decreased. In sum, previews were used to construct an action sequence based on the characteristics of the environment at the beginning of practice. Learners then relied on previous experiences to reconstruct action sequences: first locally on sections of the route, and finally globally by scanning the entire route. These changes in exploration during preview support that climbing experience promotes the development of a functional perception of the climbing environment (i.e. the perception of possible actions) (Boschker et al., 2002;Pezzulo et al., 2010). As shown by Boschker et al. (2002), our results confirm that novices rely on the physical characteristics of the climbing environment to discover the route. However, our results add that this changes rapidly as the learners gain experience with the route, which facilitates the reconstruction of action sequences.
At the beginning of training (i.e. session 1) and during the discovery of a new route (i.e. the transfer route), the climbs were characterised by two main types of exploration activity: one was related to improvisation of actions (the RE cluster) and the other related to the search for balance while climbing (the BE cluster). Rochat et al. (2020) reported that when the climber was confronted with a new route, he made errors in his hand/foot action sequences, which negatively affected his sensations of climbing fluency. Similarly, the type of exploration based on action improvisation was related to sensations of poor fluency and sensations of being unbalanced, confirming that initial perceptions in a novel condition are negative for novice participants. In contrast, the end of practice and the performances on a familiar route (the control route) were associated with exploration focussed on movement timing (the TE cluster) and repetition of a sequence of actions (the PE cluster). These findings confirm the observations of Rochat et al. (2020) that a focus on the timing of actions and ensuring the correct execution of a chain of actions characterised the lived experience of their participants.

The effect of practice conditions on exploration
Our results tended to show that the two variable practice conditions induced different types of exploration activity in the preview. Imposed confrontation with new training routes promoted a partial reconstruction of action sequences, whereas the choice of confrontation with a new training route led learners to reconstruct the action sequence on the entire route. Contrary to our expectations, our results did not support that variable practice conditions promote types of exploration during preview that facilitate adaptation to new environments, as all the participants defined an action sequence in relation to the characteristics of the transfer route, but they did not use their previous experience to construct an action sequence on this route.
During climbing, the types of exploration activity also differed in the two variable practice conditions. Imposing novelty led learners to improvise their actions in half of the trials (the RE cluster) and to focus less on balance (the BE cluster), whereas providing learners with the possibility to choose environmental novelty promoted the repetition of the action sequence (the PE cluster), and at the same time reduced improvisation and focus on timing (the TE cluster). Moreover, learners' engagement in the CN group was more related to the repetition of known action sequences (46.9%) than in the control group (18.8%), although we expected the opposite phenomenon as the control group performed all the trials of the climbing sessions on the same route. We expected learners in the CN group to regulate their exploration activity more optimally, as the literature has shown that learners are able to challenge themselves optimally when they can control their practice schedule (Liu et al., 2012). Taken together, our results rather suggest that learners from the CN group were more likely to repeat known action sequences, thus, maintaining a 'comfort zone' rather than challenging themselves. This consideration was also suggested by Hacques et al. (2022), who also highlighted that some participants in their self-controlled variability group showed poor adaptations in their gaze behaviour. Similarly, Liu et al. (2022) reported that the one of their participants who systematically chose not to change the task constraints during practice did not improve his/her performance.
In contrast to the CN group, participants from the IN group mainly improvised their actions when climbing, supporting that imposing the rate of change in task constraints may not be optimal to stabilise the effect of practice (Liu et al., 2012). This result contrasts with those of Hacques et al. (2022) as this study showed that participants of the IN group adapted their gaze pattern with practice in a way that seemed to match performance on the transfer route, suggesting a positive learning effect of imposed variability. Therefore, these contrasting conclusions between studies invite a joint analysis of learners' lived experience and behaviour to further investigate the effects of imposed and chosen variability protocols. They also invite closer examination of the individual learning trajectories of participants in protocols that give them control over learning conditions, to capture what drives them to maintain a practice condition or to challenge themselves with a new task constraint.

Enrich an enactive definition of exploration in motor learning
In our introduction, we started from a minimal enactive definition of exploration. As a reminder, Høffding and Schiavio (2021) defined exploration as "a condition of possibility of meaning-making. In the minimal sense of the action of investigating an unknown area, explorative behaviours allow agents to put forward their concerned 'point of view' (p. 814). Our results enriched this definition on three points.
First, our work highlighted the presence of different forms of exploration from the seven phenomenological clusters. These forms reflected singular articulations between intentions, perceptions and actions. From these articulations, emerged a process of sense-making from the learner's interactions with the environment. As an illustration, exploration in the preview at the beginning of learning could be related to the construction-directed exploration cluster which was defined as scanning the route (action) from the physical characteristics of the environment (perceptions), to construct hand and foot action sequences (intentions).
Second, our work highlights that exploration is linked to the conditions of practice. As an illustration, for the preview the IN and CN groups actualized (from their actions) the same form of exploration about the re-construction of sequences of actions but, in a different way. For the IN group, they attempted to reconstruct action sequences on parts of the route whereas the CN group reconstructed action sequences on the entire route.
Finally, our work has highlighted that exploration is a continuous process of sense-making/actualization of this sense-making. Indeed, there is an effect of practice on the emergence of forms of exploration. As an illustration, in the preview, exploration evolved from the construction of action sequences in relation to the physical characteristics of the environment at the beginning of practice to the reconstruction of action sequences by scanning the entire route at the end of practice.

Methodological contributions to the enactive approach
First, our results converge with those obtained by Rochat et al. (2020) in the general dimensions of intentions, perceptions and actions. These similarities between the two studies suggest the reliability of our common methodology: thematic analysis (Braun & Clarke, 2006), which was mobilised to identify the general dimensions of intentions, perceptions, and actions. This methodology highlights the stability in the results obtained for the general dimensions, supporting its reliability.
Moreover, our methodology goes beyond the methodology proposed by Rochat et al. (2020). We aimed to characterise a phenomenological synthesis of exploration from the general dimensions of intentions, perceptions and actions, without distinguishing these general dimensions from each other, as was done by Rochat et al. (2020). Since the reference to holism is an important aspect of the enactive approach, it seems that the way of presenting the results must preserve the holistic character of the phenomenon under study. Thus, for us the Hierarchical Clustering Analysis (HCA) seems to be a promising approach that goes in this direction. On the assumption that doing so might limit the impact of the results, we used hierarchical cluster analysis to connect the general phenomenological dimensions. However, although our clustering-based method was an opportunity to identify macro-level exploration formats, it did not allow us to distinguish variations within clusters. Indeed, a cluster remains a representation of a reality that hides a diversity of realities at a finer level. Yet, this level of representation might be useful to professionals as a way to help them better identify generic forms of exploration in the learner.
Second, our study seems to be distinct from those usually conducted within an enactive phenomenological framework in the sense that it brought together many participants (12 participants), a variety of practice conditions (3 practice conditions), and a large quantity of practice (86 trials). These characteristics enabled us to conduct a quantitative analysis using the hierarchical clustering method, based on current works studying motor learning from behavioural analysis (Komar et al., 2019;Seifert et al., 2017b). This step follows a first methodological step traditionally used in research involving the enactive framework and mobilising reduced samples data (Girard et al., 2021;Gottsmann et al., 2021;Jourand et al., 2018;Rochat et al., 2019Rochat et al., , 2020. This opportunity offers a new way of extracting new meaning from phenomenological data, especially by preserving the experiential globality (by combining the general phenomenological dimensions). Thus, tendencies in the distribution of forms of enacted exploration can be identified between learning sessions, groups, and conditions of practice.

Conclusion
Our study, conducted from an enactive and phenomenological approach, has provided new findings into how learners explore a learning environment from their lived experiences. First, the results showed that exploration is a process linked to practice conditions. These findings open up new research directions for intervention to better understand how we learn to explore. Second, this study has enriched the enactive definition of exploration through the identification of meanings from a joint analysis of the intentions, perceptions and actions. This work was possible thanks to the importation into the course-of-action research program of a data mining tool and methods.
Although our study allowed us to identify the distribution of the general dimensions of intentions, perceptions and actions between the beginning and the end of the learning process, it did not allow us to identify the dynamics of their transformation over the 10 sessions of the learning protocol. In the future, our goal will be to track the dynamics of emergence and changes in the intentions, perceptions and actions during exploration over 10 sessions of a learning protocol in order to identify general tendencies in the changes from the general dimensions of enacted exploration.