Reaching and Grasping With a Sensory Substitution Glove: Variability of Practice Effects

Journal of Motor Learning and Development, 2025, 13, 450–472 https://doi.org/10.1123/jmld.2024-0060 ORIGINAL RESEARCH Reaching and Grasping With a Sensory Substitution Glove: Variability of Practice Effects Saroosh Bilal, 1 Raoul M. Bongers, 2 and David M. Jacobs 1 1 Universidad Auto´noma de Madrid, Madrid, Spain; 2 University of Groningen, University Medical Center Groningen, Groningen, The Netherlands The usability of sensory substitution devices can be enhanced with practice. However, practice conditions impact the effectiveness of skill acquisition. This study investigates the effect of constant and variable practice conditions on performing a reach and grasp task with a sensory substitution glove. The glove provides vibrotactile information on the index finger or thumb whenever the index finger or thumb points toward the to-be-grasped object. We recruited 44 participants and divided them into two groups of 22 participants each. Both groups performed similar pretests and posttests. One group practiced with constant task conditions in which the object always had the same size and position. The other group encountered variations in object size and position over practice trials. The variable group showed improvements in the constant and variable posttests. The constant group, in contrast, showed improved performance only in the constant posttest. Both practice types led to increased exploration and information detection. Different strategies to perform the task were identified, stressing the importance of individual differences. Taken together, the study emphasizes the importance of incorporating variability of practice to enhance skill acquisition with sensory substitution devices, as well as suggesting that practice leads to improved exploration and information detection. Keywords:exploration, individual differences, information detection, practice conditions, skill acquisition, visual impairment © 2025 The Authors. Published by Human Kinetics, Inc. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, CC BY-NC-ND 4.0, which permits the copy and redistribution in any medium or format, provided it is not used for commercial purposes, no modifications are made, appropriate credit is given, and a link to the license is provided. See http://creativecommons.org/licenses/by-nc-nd/4.0. This license does not cover any third-party material that may appear with permission in the article. For commercial use, permission should be requested from Human Kinetics, Inc., through the Copyright Clearance Center (http://www.copyright.com). Bilal https://orcid.org/0009-0000-0193-2449 Bongers https://orcid.org/0000-0002-3518-7464 Jacobs ([email protected]) is corresponding author, https://orcid.org/0000-0003-2954-5293 450 Unauthenticated | Downloaded 11/13/25 03:28 PM UTC According to the World Health Organization, the number of individuals worldwide with moderate or severe blindness is greater than 250 million (Bourne et al., 2017). This large number of individuals with visual impairments implies that effective technological aids may impact the lives of numerous individuals (Hoffmann et al., 2018). Many invasive and noninvasive aids exist (Elli et al., 2014). One group of noninvasive devices is known as sensory substitution devices (SSDs). These devices provide either tactile information (Bach-y-Rita et al., 1969)oracoustic information (Abboud et al., 2014)about the surroundingsof the user. Bach-y-Rita and colleagues created the first widely-known SSD that substitutes visual information through the tactile modality. In contrast to the earlier SSDs, more recent SSDs mostly aim to help individuals to perform specific perceptual motor tasks, rather than substituting vision as a whole (examples of recent SSDs are the Enactive Torch [Froese & Ortiz-Garin, 2020]andtheTSIGHT[Cancar et al., 2013]). One of the many tasks that may be challenging under conditions of visual impairment is the task of locating nearby objects in order to reach to and grasp the objects. With this challenge in mind, de Paz et al. (2023) developed a sensory substitution glove. In the here-reported experiment, we used a glove with the same design as the one of de Paz et al. The functioning of the glove is based on the distance to the first-encountered object in the pointing direction of the index finger and thumb at a particular moment. When one of the fingers points in the direction of an object, the distance to the object is transformed in an on-line manner into a vibration that is applied to the tip of the respective finger. When no object is encountered in the pointing direction, no vibration is applied. Moving the fingers while wearing the glove produces changes in the vibration, and these changes depend on the distance and orientation of the fingers with respect to the object (cf. Lenay et al., 1997). De Paz et al. reported that participants in their experiments were able to grasp a high percentage of the objects with the glove. In the grasping task, therefore, and in many other tasks, laboratory research indicates that SSDs allow relatively high levels of performance (de Paz et al., 2023; Kilian et al., 2022;Lobo et al., 2018,2019). SSDs thus have the potential to be useful for visually impaired individuals. Even so, SSDs are not frequently used in situations other than scientific laboratories. Among other reasons, Elli et al. (2014) have argued that the limited use of SSDs by individuals with a visual impairment may partly be related to the difficulty of learning to interact with the environment when using an SSD. Kristjánsson et al. (2016) have likewise argued that the use of SSDs outside the laboratory requires appropriate training, and that the topic of learning to use SSDs has received insufficient attention in research. It is therefore crucial that the field of research on SSDs concentrates more on the issue of skill acquisition in order to enhance the usability of SSDs. Key Points •Variable practice has benefits over constant practice in the acquisition of reaching and grasping skills with a sensory substitution glove. •Practice with a sensory substitution glove leads to an increase in the amount of exploration and in the effectiveness of exploration. •Different strategies are used to grasp objects with a sensory substitution glove, stressing the importance of individual differences. JMLD Vol. 13, No. 2, 2025 VARIABILTY OF PRACTICE AND SENSORY SUBSTITUTION 451 Unauthenticated | Downloaded 11/13/25 03:28 PM UTC From research on skill acquisition in other contexts, it is well known that the conditions that are encountered during practice are crucial for the resulting performance. One of the parameters that has been varied is whether practice conditions are constant or variable. Since Schmidt (1975) introduced the concept of variability of practice, a substantial number of studies have reviewed or reported empirical evidence related to it. Although some authors are cautious about the empirical support for the positive effects of variability (van Rossum, 1990), others provide more encouraging views (e.g., Raviv et al., 2022;Schöllhorn et al., 2012; Shea & Wulf, 2005; cf. Table 8.1 in Gray, 2021). In the specific case of comparing constant practice with variable practice, several studies have shown a stronger overall improvement after variable practice than after constant practice (North et al., 2019;Oliveira et al., 2024). Schoenfelt et al. (2002) studied basketball freethrow shooting and showed that although the constant practice group practiced under conditions that were similar to the retention test, their performance at the retention test did not differ from the variable practice groups. Yao et al. (2009) studied wheelchair propulsion and observed a superior effect of variable practice for most of the experimental conditions. Questions about how to implement variable practice are generally discussed in the literature about contextual interference (Goode & Magill, 1986;Shea & Morgan, 1979), demonstrating superior effects of random practice over blocked practice. Overall, we believe that it is fair to conclude from this body of work that variable practice has advantages over constant practice, most particularly because the transfer of learning to novel task conditions is generally better after variable practice. If one accepts this conclusion, and one believes that learning with an SSD is in essential ways similar to other forms of perception-action learning, then it is a small step to see the potential of variability of practice for the acquisition of skills with an SSD. Testing this notion is the main focus of the present study. Different explanations have been provided about why variability in practice conditions may have beneficial effects on skill acquisition. Huet et al. (2011), for example, argued that variability helps individuals to transition toward reliance on more useful informational variables, because the variability reduces the usefulness of initially used variables and therefore helps people to abandon these variables. Consequently, better ones may be discovered (Smeeton et al., 2013). Another explanation is that variable practice may help each individual to discover his or her own specific perceptual-motor solutions, going against the idea that what is learned is one optimal technique that should be the same on every execution of the action. Using the formulation by Gray (2021, p. 55): Moving skillfully involves coming up with new solutions to new problems, not just repeating the same old solution. How do we learn to solve problems? By practicing a lot of different ones. In other words, by having variability in our practice conditions. In this view, successful task accomplishment is linked to an efficient search for functional solutions (Hacques et al., 2021;Orth et al., 2019). This search process is affected by the individuals’intrinsic dynamics, and it is therefore important to consider individual behavioral patterns in studies on skill acquisition. A stronger focus on individual differences has been defended by several authors in the JMLD Vol. 13, No. 2, 2025 452 BILAL, BONGERS, AND JACOBS Unauthenticated | Downloaded 11/13/25 03:28 PM UTC research fields of perception-action and learning (Golenia et al., 2014;Vidal & Lacquaniti, 2021). Relatedly, individual differences have been argued to lead to different learning routes (Kostrubiec et al., 2012; cf. Pacheco et al., 2019), which might motivate training regimens inspired by individual characteristics. The present study, then, addresses the effects of practice on reaching and grasping objects with an SSD. We applied a pretest–practice–posttest design, where two groups of individuals practiced under constant or variable task conditions. Our analysis focused on the differences between the training conditions, on exploration, on information usage, and on individual differences in how the task is performed. The purpose of the study was (a) to demonstrate that practice with the SSD-based reach-and-grasp task leads to improvements in performance, (b) to test the differences in the effectivity of constant and variable practice, (c) to examine if improvements in performance with practice are accompanied with increases in exploration and in active information detection, and (d) to determine individual differences in exploration and information detection. Methods Participants We recruited 44 right-handed participants (mean age: 19.6 years). This number is in line with previous experiments with a sensory substitution glove (de Paz et al., 2023, used between 32 and 48 participants in their experiments). Of our participants, 37 were females and seven were males. All of them were students at the Faculty of Psychology of the Autonomous University of Madrid. They received course credits for their participation. All participants had normal or corrected-tonormal vision and had no self-reported movement problems. The local ethics committee approved the experiment (UAM-CEI-125-2563). Written informed consent was obtained from all participants. Materials The SSD and the experimental setup are schematically illustrated in Figure 1. The setup included a wooden table that was covered with a black tablecloth to avoid reflections of the infrared light used by the movement-registration system (Qualisys AB). The tablecloth was attached firmly to the table to avoid it (and the objects on it) being displaced due to the hand and arm movements. A chinrest was used to make sure that all participants sat at the same position with respect to the table, and to restrict head and trunk movements during the experiment. The main part of the SSD was a right-hand glove, made of soft and stretchable material. Two vibrotactile motors were attached to the glove, one on the back of the index finger and one on the back of the thumb. There was a wired connection of the motors to a printed circuit board, which was further connected to a digital/analogue conversion card and a PC (Intel Core i7, 3.07 GHz). This system allowed the intensity of vibration to be controlled by MATLAB (MathWorks, Inc., Version R2022b). The SSD did not include actual sensors to determine the distance to the nearest object. Instead, it used the movement-registration system, which consisted of seven cameras. The glove had six infrared-reflecting markers attached to it. These JMLD Vol. 13, No. 2, 2025 VARIABILTY OF PRACTICE AND SENSORY SUBSTITUTION 453 Unauthenticated | Downloaded 11/13/25 03:28 PM UTC markers were placed on the tips, middle joints, and lower joints of the index finger and thumb. The movement-registration system detected the locations of the markers at 100 Hz. These locations were used to analyze the movement trajectories and to determine the distance between the markers on the fingertips and the nearest surface in the pointing direction of the index finger and thumb. That distance, in turn, was used to control the vibration of the motors on the glove. In addition to the six reflective markers on the glove, markers were placed on the right wrist, right elbow, right shoulder, and left shoulder. These four markers Figure 1 —(A) Sensory substitution glove. The wired blue circles circles represent the vibrotactile motors and the white dots the reflective markers. The dashed red lines indicate the directions of sensitivity, which were determined with the two markers closest to the tip of each finger. In the figure, the index finger points to the target, leading to a vibration that depends on the distance indicated by D. (B) Experimental setup. The red (right) and blue (left) rectangles represent the initial and target objects, respectively, and the ×symbols represent possible positions of the objects in the experiment. Also shown are the seven movement-registration cameras around the table. The used coordinate system is indicated in the lower left corner. See online article for color version of the figure. JMLD Vol. 13, No. 2, 2025 454 BILAL, BONGERS, AND JACOBS Unauthenticated | Downloaded 11/13/25 03:28 PM UTC were attached to a jacket with Velcro. Before the actual experiment, three individuals were asked to perform reach and grasp movements similar to the ones in the experiment. With their data, we generated an Automatic Identification of Markers model with the Qualisys software. This model allowed automatic labeling of trajectories that was applicable also to the trajectories of future participants. This means that the Qualisys software was able to identify the markers before their locations were sent to MATLAB. As to-be-grasped objects, two sets of three rectangular Lego objects were used. One set of objects was referred to as the initial objects. These objects were held by the participants before the trial. They were removed by the experimenter just before the trial initiation. This procedure allowed precise control over the initial position of the hand and fingers. The other set of objects were the targets. The goal of the participants was to reach to and grasp the target objects during the trial. All objects had a height of 7.0 cm and a depth of 3.2 cm. According to the width of the objects (3.2, 4.8, or 6.4 cm), we refer to them as small, medium, or large. The objects were registered by the Qualisys system as rigid bodies. We used two unique marker configurations for the initial and target objects, so that Qualisys could distinguish them and send their position information online to MATLAB. Functioning of the Glove As mentioned above, the vibration intensity of the motors was computed in an online manner. Customized MATLAB routines first used the markers on the fingers to trace imaginary lines in the pointing direction of the two fingers in the horizontal plane. More specifically, these lines were based on the top two markers on each finger. The programs then determined if these lines crossed any of the sides of the rectangular target (also considered in the horizontal plane). If one of the lines crossed the target, then the distance between the fingertip and the point where the line reached the target was computed. The vibration intensity was computed independently for each finger. If the line in the pointing direction of a finger did not cross the target, or the distance between the marker on the fingertip and the object was larger than the maximum detectable distance (D max ; set at 45 cm), then no vibration was applied to the finger. If the line crossed the target and the distance was less than D max , then the vibration intensity increased from a minimum intensity (I min ; set at 4) to a maximum intensity (I max ; set at 10) when the distance was reduced from D max to 0. Said more precisely, the vibration intensity was calculated with the following equation: I=Imin +ðImax −IminÞ×Dmax −D Dmax 3 : In this equation, Dis the finger-target distance, Iis the vibration intensity, 1 and the other parameters are as specified above. The minimum intensity value of 4 was chosen because it is slightly above the minimum value needed to activate the vibrotactile motors. The maximum intensity value of 10 corresponds to the maximum admitted intensity for the motors. Note that we used a cubic equation to map the hand-target distance to the intensity of the vibration. We did pilot work with ourselves as participants to ensure that this is convenient. JMLD Vol. 13, No. 2, 2025 VARIABILTY OF PRACTICE AND SENSORY SUBSTITUTION 455 Unauthenticated | Downloaded 11/13/25 03:28 PM UTC During a trial, participants moved their hand from the initial object (red rectangle in Figure 1B) to the target (blue rectangle in Figure 1B) with the goal of grasping the latter. The small black crosses in the figure show the possible object positions. As indicated in the figure, we defined the xcoordinate as the coordinate that indicates the lateral direction (parallel to the frontal plane) and the ycoordinate as the one that indicates the anterior–posterior direction (parallel to the sagittal plane). The distance between the initial and target objects in the xdirection was 40 cm on every trial. In contrast, the objects could both have one of three ycoordinates, which we referred to as near (y=5), middle (y=10), and far (y=15). Remember that both objects could also have one of three sizes. Thus, each trial was specified by choosing four parameters: the positions and sizes of the initial and target objects. Instructions to Participants The participants were given oral and written instructions. The written instructions were as follows: “You will perform a grasping task while being blindfolded and wearing a sensory substitution glove. You will also wear a jacket that carries markers for the right upper limb and left shoulder. The glove has one vibrating motor on the index finger and another one on the thumb. It is comfortable and safe to use. The task will be performed in a seated position while your chin will be resting on a chinrest. Before each trial, the experimenter will guide you to hold an initial object. When the initial object is removed you will be requested to maintain the positioning of your hand. When the experimenter says, ‘you can begin now’, you may start the trial. In most cases, you must start moving the hand and fingers to generate the vibration in the glove that will guide you to the object. If one of the fingers points in the appropriate direction (i.e., in the direction of the target), there will be vibration of the coin motor of that finger. Once you have grasped the target, the trial will be finished, and the motors will stop vibrating. There are no time limits to complete the task.”With the verbal instructions that followed the written ones, the experimenter made sure that the participant had understood the procedure and functioning of the glove. Experimental Design Participants were divided into a constant group and a variable group (n=22 for each group). The groups differed in the practice conditions that they encountered in the training phase: constant or variable. A block of trials with constant conditions repeated the same trial throughout the block. This trial was always the one with the medium-sized initial and target objects placed in the middle positions (i.e., the example shown in Figure 1B). A block of trials with variable conditions used a different combination of initial and target positions and sizes on each trial. The experiment was divided into four phases. These phases were the familiarization phase (three trials), the pretest (12 trials under constant conditions and 12 trials under variable conditions), the training (96 trials under either constant or variable conditions, depending on the group), and the posttest (12 trials under constant conditions and 12 trials under variable conditions). The phases are summarized in Table 1. An additional phase of 12 trials was performed under constant conditions immediately after the experiment, but the analyses on the data of this phase will not be presented in the current paper. Except for the training JMLD Vol. 13, No. 2, 2025 456 BILAL, BONGERS, AND JACOBS Unauthenticated | Downloaded 11/13/25 03:28 PM UTC phase, all phases were the same for the constant and variable groups. We next provide a more detailed description of each phase. Familiarization The familiarization consisted of three trials performed under constant task conditions. The familiarization phase was the only phase that was performed with vision in addition to the vibrotactile information. Pretest and Posttest The pretest consisted of two blocks, referred to as the constant and variable pretests. During the constant pretest, constant conditions were used, and during the variable pretest, variable conditions were used. In the variable pretest, the near and far positions were used for both objects. All three sizes were used for the target, but only the medium size for the initial object. This led to 2 (initial positions) ×1 (initial size) ×2 (target positions) ×3 (target sizes) =12 combinations. The order of the combinations was randomized. The order of the constant and variable pretests was counterbalanced among the participants. The posttest was in all aspects identical to the pretest. Practice The third phase was the practice phase, consisting of four blocks of 24 trials each. The constant group practiced with constant conditions and the variable group practiced with variable conditions. During the variable training, the small and large initial objects were used at either the near or far positions. For the target, all three sizes were used at either the near or far positions. This led to 2 (initial positions) × 2 (initial sizes) ×2 (target positions) ×3 (target sizes) =24 combinations. Each Table 1 Positions and Sizes of Initial and Target Objects in Different Experimental Phases Initial object Target object Phase (no. of trials) Position Size Position Size Familiarization (3) M M M M Pretest (24) Constant (12) M M M M Variable (12) N, F M N, F S, M, L Training (4 ×24) Constant (4 ×24) M M M M Variable (4 ×24) N, F S, L N, F S, M, L Posttest (24) Constant (12) M M M M Variable (12) N, F M N, F S, M, L Note. For position: N =near, M =middle, F =far; for size: S =small, M =medium, L =large. See text for explanation. JMLD Vol. 13, No. 2, 2025 VARIABILTY OF PRACTICE AND SENSORY SUBSTITUTION 457 Unauthenticated | Downloaded 11/13/25 03:28 PM UTC combination was used once per block of trials, in a randomized order. One may note that the variable training did not include the middle position for either of the objects or the middle size for the initial object. This was done with the aim to slightly increase the difference between the trials that were used in the variable training and the trial that was used in the constant pretest and posttest. To make it easier for participants to maintain a high level of concentration, the experiment was divided into two sessions that were performed on different days. The first session comprised the familiarization, the pretest, and the first two blocks of training. The second session comprised the last two blocks of training, the posttest, and the additional phase (of which the data are not reported). Each session took approximately 1.5 hr. The two sessions were in most cases performed within 1 week. Previous research has shown that changes in performance often occur in a largely continuous manner over blocks of trials despite a division in sessions (see, e.g., Figure 4 of Jacobs et al., 2000). Data Processing and Statistical Analysis The analysis that is reported in this article concerns the horizontal components of two markers: the ones on the tips of the index finger and thumb. The time series for these markers and dimensions were gap-filled and smoothed with the MATLAB function “interp1.”The movement initiation was determined as the moment at which the speed of the marker on either the index finger or the thumb surpassed 5 m/s (after filtering the speed data). The movement trajectories were analyzed until the moment at which the analyzed markers had both reached a position that was 2 cm removed, in the xdirection, from the target. For the majority of the trials, this was just before the target was touched with the fingertips. We refer to this moment as the end of the trial. Trials were automatically removed from the analyses if (a) there were gaps in the data of more than 0.5 s in the interval between the movement initiation and the end of the trial, (b) the movement initiation did not occur during the first 5 s after the start of the data registration, or (c) at the moment that was identified as the movement initiation the hand had already moved more than 10 cm in the x direction. The latter situation could occasionally happen if the participant began the movement before the instruction of the experimenter to do so, meaning that the hand had already moved substantially when the experimenter started the registration. Apart from the automatic removal of the trials, the first author inspected all movement trajectories visually, comparing the raw trajectories and the gap-filled and smoothed trajectories. On the basis of the visual inspection, additional trials were removed if (a) the gap filling (for gaps smaller than 0.5 s) did not work in a satisfactory way, (b) the smoothed data differed substantially from the registered raw data, or (c) there were other problems with the registered data such as peaks in the marker locations that may have been due to unanticipated reflections. In total, after the automatic and manual inspections, 5.1% of the trials were removed for the test-phase data and 3.1% for the practice data. Of the trials that were included in the analysis for the test phases, 90.1% had no missing values, meaning that all markers were properly registered at all frames. Furthermore, for 67.7% of the test-phase trials that had missing values and were included in the analysis, the maximum length of the interval for which data were missing waslessthan50ms.Thesenumbersweresimilar for the practice data. When missing JMLD Vol. 13, No. 2, 2025 458 BILAL, BONGERS, AND JACOBS Unauthenticated | Downloaded 11/13/25 03:28 PM UTC Figure 6 —Trial that illustrates abrupt periods of exploration. (A) Anterior–posterior (y) components of markers on tips of index finger and thumb. (B) Vibration applied to index finger and thumb. (C) Distance between markers on tips of index finger and thumb. Figure 5 —Trial that illustrates hand-opening strategy of exploration. (A) Anterior– posterior (y) components of markers on tips of index finger and thumb. (B) Vibration applied to index finger and thumb. (C) Distance between markers on tips of index finger and thumb. JMLD Vol. 13, No. 2, 2025 465 Unauthenticated | Downloaded 11/13/25 03:28 PM UTC first two periods of exploration with the hand opening did not lead to vibration on the fingers. This indicates that the participant failed to locate the object until he or she explored with a wider hand opening in the third attempt. The pattern of exploration in Panel C contrasts to the exploration that was observed in many other trials in the sense that it showed clearly identifiable and rather abrupt periods in which the exploration took place. In this case, the abrupt exploration was performed with the hand opening. Cases with abrupt exploration with the hand position were also seen. Case 4: Exploration With One Finger Panel B of Figure 7shows that, in this case, vibratory information was obtained only with the thumb. This means that the individual moved toward the target with the thumb pointing in the direction of the object, apparently without useful exploratory movements of the index finger. Although exploration with both fingers was more common, several other examples were seen of a clear dominance of either the index finger or thumb. Discussion In the present study, we addressed the effects of practice on performance with a sensory substitution glove. Forty-four blindfolded participants performed a reach and grasp task with the glove. A pretest–training–posttest design was applied. Participants were divided into two groups: the constant and variable groups. The constant group practiced the task with constant conditions, which means that the object always had the same position and size. For the variable group, the object positions and sizes differed per trial. Practicing the task led to substantial improvements in performance. The overall proportion of correct grasps increased from 0.79 in the pretest to 0.89 in the posttest, and, even though participants were not asked to perform the task as quickly as possible, the average time it took to complete the trials decreased from 4.9 s in the pretest to 3.1 s in the posttest. In other words, performance became more accurate and faster. These results constitute an additional contribution to the findings of de Paz et al. (2023), who studied the use of a similar glove without addressing the issue of skill acquisition. The observed improvements with practice are consistent with other studies in the field of sensory substitution. For example, a study by Kilian et al. (2022) revealed that structured training helps individuals with visual impairments in navigation and wayfinding with a sensory substitution glove. With respect to the effect of the different practice conditions, the variable group showed a notable increase in the proportion of successful grasps for the variable and constant test conditions. In contrast, the constant group improved in the proportion of correct grasps only for the constant test conditions. Such findings are in line with other studies that have reported benefits of variable practice conditions (Raviv et al., 2022). Our results demonstrate that variability of practice techniques are useful also in the field of sensory substitution. Given that the development of training methods is a crucial issue for SSD users (Elli et al., 2014;Kristjánsson et al., 2016), we believe that this is a promising contribution. JMLD Vol. 13, No. 2, 2025 466 BILAL, BONGERS, AND JACOBS Unauthenticated | Downloaded 11/13/25 03:28 PM UTC Figure 7 —Trial that illustrates exploration with one finger. (A) Anterior–posterior (y) components of markers on tips of index finger and thumb. (B) Vibration applied to index finger and thumb. (C) Distance between markers on tips of index finger and thumb. JMLD Vol. 13, No. 2, 2025 467 Unauthenticated | Downloaded 11/13/25 03:28 PM UTC To understand what type of change underlies the observed improvements in performance, we examined how practice affected the exploration that occurred during the trials. A significant increase was observed in the SD of the hand-opening from the pretest to the posttest (one may want to remember that the exploration measures were computed during the first 80% of each of the test trials). In addition, a marginally significant increase was observed for the anterior–posterior movements of the hand (in that period). In our interpretation, this increase in variability occurred because individuals came to explore more with the sensory substitution glove. In their review, Hacques et al. (2021) included a section on the question whether exploration decreases with practice (p. 1372). In that section, studies are described that indeed indicate that less exploration is observed after practice. This is not what we observed in our study. More importantly, however, Hacques and colleagues also argued that one should distinguish the quantity of exploration from the effectiveness of exploration. Experts may seem to explore less because their exploration is more effective. This means that analyses cannot be limited to measures of the amount of exploration. Measures of the effectiveness of exploration are also needed. In the present task, exploration is effective if it provides information about the edges of the to-be-grasped object. This is so because the object is usually grasped at the edges. Crucial information about the edges is detected whenever the vibration switches from off to on or vice versa, which occurs whenever the pointing direction of the finger sweeps over the edges. We quantified the effectiveness of the exploration, or the detection of useful information, as the number of bouts of vibration (off-on-off periods). The number of bouts that occurred per second increased with practice. In sum, in our experiment, both the quantity of exploration and the effectiveness of exploration increased with practice. Even though the increase in the quantity of exploration may not be in agreement with the studies that were reviewed by Hacques et al. (2021), the overall pattern of an increasing exploration is consistent with studies that claim that active exploration is a key component of successful performance with SSDs (de Paz et al., 2023;Díaz et al., 2012). Let us note at this point that our study is not without limitations. First, we did not include transfer tests. This is unfortunate in part because the benefits of variable practice are often associated with the generalization to novel task conditions. Related to this issue, there was an asymmetry between our practice conditions. For the variable group, the trial that was used in the constant posttest fell within the range of conditions that was encountered during practice. In contrast, for the constant group (who practiced with a single trial), the variable posttest obviously included trials that fell beyond the range of conditions with which they had practiced. Second, we did not use a retention test to measure if the observed improvements in performance were stable over time. Even so, the curves that indicate the improvements over practice blocks (Figure 2) demonstrate the durability of the improvements. This is so because there was a break of up to 1 week between the second and third practice blocks, and this break did not lead to a substantial reduction in performance. Third, our glove did not include real distance sensors and therefore requires a voluminous (and expensive) movement registration system. This means that further technical developments are needed before the glove can be of practical use for individuals with visual impairments. JMLD Vol. 13, No. 2, 2025 468 BILAL, BONGERS, AND JACOBS Unauthenticated | Downloaded 11/13/25 03:28 PM UTC Let us also mention a limitation regarding the obtained results. One could have expected that training in a variable way would have stimulated exploration and the search for information-movement couplings more than training under constant conditions. Nevertheless, we observed largely similar increases in exploration and information usage for the different training regimens. Possibly related to this limitation, we observed a large variability between participants in how they performed the task. This high variability may have obscured more subtle differences between constant and variable practice. Future studies may therefore aim to reduce the variability or use a larger sample size to increase statistical power. To ensure a higher statistical power, sample size should be based on desired power instead of on comparisons with previously published studies. Studies with a higher power may provide a more reliable picture of the differences between the training regimens. Alternatively, future studies may focus specifically on the different exploratory strategies, fomenting their occurrence and aiming to reveal whether, and if so how, they relate to performance. As a final limitation of our study, let us mention a shortcoming that is relevant to the field of research on variability of practice in general. We used variability in the anterior–posterior coordinate and the size of the initial and target objects. However, the type and amount of variability that we used could have been chosen in different ways (e.g., in the left–right coordinates and/or the shape of the trajectories, etc.). Which ways to introduce variability of practice would be most beneficial for skill acquisition? The shortcoming that we want to mention here is that we do not have an answer to this question. We believe that this is true for the field of research on variability of practice as a whole. In general, predictions concerning the amounts and types of variability that are most beneficial are difficult to make. The positive message of the present study is that the benefits of variable practice and the increases in the quantity and effectiveness of exploration were so robust that the effects showed up in spite of the limitations that have been mentioned in this discussion. To conclude, therefore, we would like to emphasize the importance of the incorporation of variability in practice conditions as well as the possibility for active exploration in order to allow users of SSDs to gain proficiency. Notes 1. The used measure of intensity is proportionally related to the driving voltage of the motors. The driving voltage, in turn, is related to the frequency as well as the amplitude of the vibrations. We refer the reader to the appendix of Díaz et al. (2012) for a more precise specification of these relations. 2. For completeness, let us report the main differences between the ANOVAs with and without outliers. To do so, we analyzed all changes among levels of significance, which is to say, changes in pvalue that crossed the levels of .10, .05, .01, or .001. All effects that are reported to be significant in the “Result”section had the same level of significance in the ANOVAs with outliers. For the effects that are not reported to be significant, the following changes occurred. First, two interactions that are reported to be nonsignificant were significant in the ANOVAs with the outliers, both in the ANOVA on trial duration. More specifically, for the interaction Test phase ×Test condition the pvalue changed from p=.096 to p=.029 and for the interaction JMLD Vol. 13, No. 2, 2025 VARIABILTY OF PRACTICE AND SENSORY SUBSTITUTION 469 Unauthenticated | Downloaded 11/13/25 03:28 PM UTC Test phase ×Practice condition the pvalue changed from p=.160 to p=.036. Furthermore, two effects that are reported to be marginally significant were not so in the ANOVAs with the outliers, both in the ANOVA on the hand positioning variable. For this ANOVA, the pvalue for the main effect of test phase changed from p=.082 to p=.126 and the pvalue for the interaction Test condition ×Practice condition changed from p=.083 to .139. Acknowledgments Funding Source: This project has received funding from the European Union’s Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement No. 956003. References Abboud, S., Hanassy, S., Levy-Tzedek, S., Maidenbaum, S., & Amedi, A. (2014). EyeMusic: Introducing a “visual”colorful experience for the blind using auditory sensory substitution. Restorative Neurology and Neuroscience, 32(2), 247–257. https://doi.org/ 10.3233/RNN-130338 Bach-Y-Rita, P., Collins, C.C., Saunders, F.A., White, B., & Scadden, L. (1969). Vision substitution by tactile image projection. Nature, 221(5184), 963–964. https://doi.org/ 10.1038/221963a0 Bourne, R.R., Flaxman, S.R., Braithwaite, T., Cicinelli, M.V., Das, A., Jonas, J.B., Keeffe, J., Kempen, J.H., Leasher, J., Limburg, H., Naidoo, K., Pesudovs, K., Resnikoff, S., Silvester, A., Stevens, G.A., Tahhan, N., Wong, T.Y., & Taylor, H.R., & Vision Loss Expert Group. (2017). Magnitude, temporal trends, and projections of the global prevalence of blindness and distance and near vision impairment: A systematic review and meta-analysis. The Lancet Global Health, 5(9), e888–e897. https://doi.org/10. 1016/S2214-109X(17)30293-0 Cancar, L., Díaz, A., Barrientos, A., Travieso, D., & Jacobs, D.M. (2013). Tactile-Sight: A sensory substitution device based on distance-related vibrotactile flow. International Journal of Advanced Robotic Systems, 10(6), 272. https://doi.org/10.5772/56235 de Paz, C., Ibá˜nez-Gijo´n, J., Travieso, D., & Jacobs, D.M. (2023). Grasping objects with a sensory substitution glove. International Journal of Human-Computer Studies, 170, Article 102963. https://doi.org/10.1016/j.ijhcs.2022.102963 Díaz, A., Barrientos, A., Jacobs, D.M., & Travieso, D. (2012). Action-contingent vibrotactile flow facilitates the detection of ground level obstacles with a partly virtual sensory substitution device. Human Movement Science, 31(6), 1571–1584. https://doi. org/10.1016/j.humov.2012.05.006 Elli, G.V., Benetti, S., & Collignon, O. (2014). Is there a future for sensory substitution outside academic laboratories? Multisensory Research, 27(5–6), 271–291. https://doi. org/10.1163/22134808-00002460 Froese, T., & Ortiz-Garin, G.U. (2020). Where is the action in perception? An exploratory study with a haptic sensory substitution device. Frontiers in Psychology, 11, 528286. https://doi.org/10.3389/fpsyg.2020.00809 Golenia, L., Schoemaker, M.M., Mouton, L.J., & Bongers, R.M. (2014). Individual differences in learning a novel discrete motor task. PLoS One, 9(11), Article e112806. https://doi.org/10.1371/journal.pone.0112806 Goode, S., & Magill, R.A. (1986). Contextual interference effects in learning three badminton serves. Research Quarterly for Exercise and Sport, 57(4), 308–314. https:// doi.org/10.1080/02701367.1986.10608091 JMLD Vol. 13, No. 2, 2025 470 BILAL, BONGERS, AND JACOBS Unauthenticated | Downloaded 11/13/25 03:28 PM UTC Gray, R. (2021). How we learn to move: A revolution in the way we coach and practice sports skills. Perception Action Consulting & Education LLC. Hacques, G., Komar, J., Dicks, M., & Seifert, L. (2021). Exploring to learn and learning to explore. Psychological Research, 85(4), 1367–1379. https://doi.org/10.1007/s00426020-01352-x Hoffmann, R., Spagnol, S., Kristjánsson, Á., & Unnthorsson, R. (2018). Evaluation of an audio-haptic sensory substitution device for enhancing spatial awareness for the visually impaired. Optometry and Vision Science, 95(9), 757–765. https://doi.org/ 10.1097/OPX.0000000000001284 Huet, M., Jacobs, D.M., Camachon, C., Missenard, O., Gray, R., & Montagne, G. (2011). The education of attention as explanation of variability of practice effects: Learning the final approach phase in a flight simulator. Journal of Experimental Psychology: Human Perception and Performance, 37(6), 1841–1854. https://doi.org/10.1037/a0024386 Jacobs, D.M., Michaels, C.F., & Runeson, S. (2000). Learning to perceive the relative mass of colliding balls: The effects of ratio scaling and feedback. Perception & Psychophysics, 62(7), 1332–1340. https://doi.org/10.3758/BF03212135 Kilian, J., Neugebauer, A., Scherffig, L., & Wahl, S. (2022). The unfolding space glove: A wearable spatio-visual to haptic sensory substitution device for blind people. Sensors, 22(5), 1859. https://doi.org/10.3390/s22051859 Kostrubiec, V., Zanone, P.G., Fuchs, A., & Kelso, J.S. (2012). Beyond the blank slate: Routes to learning new coordination patterns depend on the intrinsic dynamics of the learner—Experimental evidence and theoretical model. Frontiers in Human Neuroscience, 6, 222. https://doi.org/10.3389/fnhum.2012.00222 Kristjánsson, Á., Moldoveanu, A., Jo´hannesson, Ó.I., Balan, O., Spagnol, S., Valgeirsdo´ttir, V.V., & Unnthorsson, R. (2016). Designing sensory-substitution devices: Principles, pitfalls and potential. Restorative Neurology and Neuroscience, 34(5), 769–787. https://doi.org/10.3233/RNN-160647 Lenay, C., Canu, S., & Villon, P. (1997, August 25–28). Technology and perception: The contribution of sensory substitution systems [Conference session]. Proceedings of the Second International Conference on Cognitive Technology, pp. 44–53, AizuWakamatsu, Japan. https://doi.org/10.1109/CT.1997.617681 Lobo, L., Nordbeck, P.C., Raja, V., Chemero, A., Riley, M.A., Jacobs, D.M., & Travieso, D. (2019). Route selection and obstacle avoidance with a short-range haptic sensory substitution device. International Journal of Human Computer Studies, 132,25–33. https://doi.org/10.1016/j.ijhcs.2019.03.004 Lobo, L., Travieso, D., Jacobs, D.M., Rodger, M., & Craig, C.M. (2018). Sensory substitution: Using a vibrotactile device to orient and walk to targets. Journal of Experimental Psychology: Applied, 24, 108–124. https://doi.org/10.1037/xap0000154 North, J.S., Bezodis, N.E., Murphy, C.P., Runswick, O.R., Pocock, C., & Roca, A. (2019). The effect of consistent and varied follow-through practice schedules on learning a table tennis backhand. Journal of Sports Sciences, 37(6), 613–620. https://doi.org/10. 1080/02640414.2018.1522683 Oliveira, V., Menayo, R., & Fuentes-García, J.P. (2024). Training tennis through induced variability and specific practice: Effects on performance in the forehand approach shot. Applied Sciences, 14(8), 3287. https://doi.org/10.3390/app14083287 Orth, D., McDonic, L., Ashbrook, C., & van der Kamp, J. (2019). Efficient search under constraints and not working memory resources supports creative action emergence in a convergent motor task. Human Movement Science, 67, Article 102505. https://doi.org/ 10.1016/j.humov.2019.102505 Pacheco, M.M., Lafe, C.W., & Newell, K.M. (2019). Search strategies in the perceptualmotor workspace and the acquisition of coordination, control, and skill. Frontiers in Psychology, 10, 1874. https://doi.org/10.3389/fpsyg.2019.01874 JMLD Vol. 13, No. 2, 2025 VARIABILTY OF PRACTICE AND SENSORY SUBSTITUTION 471 Unauthenticated | Downloaded 11/13/25 03:28 PM UTC Raviv, L., Lupyan, G., & Green, S.C. (2022). How variability shapes learning and generalization. Trends in Cognitive Sciences, 26(6), 462–483. https://doi.org/10. 1016/j.tics.2022.03.007 Schmidt, R.A. (1975). A schema theory of discrete motor skill learning. Psychological Review, 82(4), 225–260. https://doi.org/10.1037/h0076770 Schoenfelt, E.L., Snyder, L.A., Maue, A.E., McDowell, C.P., & Woolard, C.D. (2002). Comparison of constant and variable practice conditions on free-throw shooting. Perceptual and Motor Skills, 94, 1113–1123. https://doi.org/10.2466/pms.2002.94.3c.1113 Schöllhorn, W., Hegen, P., & Davids, K. (2012). The nonlinear nature of learning: A differential learning approach. The Open Sport Sciences Journal, 7, 100–1012. https:// doi.org/10.2174/1875399X01205010100 Shea, C.H., & Wulf, G. (2005). Schema theory: A critical appraisal and reevaluation. Journal of Motor Behavior, 37(2), 85–102. https://doi.org/10.3200/JMBR.37.2.85-102 Shea, J.B., & Morgan, R.L. (1979). Contextual interference effects on the acquisition, retention, and transfer of a motor skill. Journal of Experimental Psychology: Human Learning and Memory, 5(2), 179–187. https://doi.org/10.1037/0278-7393.5.2.179 Smeeton, N.J., Huys, R., & Jacobs, D.M. (2013). When less is more: Reduced usefulness training for the learning of anticipation skill in tennis. PLoS One, 8(11), Article e79811. https://doi.org/10.1371/journal.pone.0079811 van Rossum, J.H. (1990). Schmidt’s schema theory: The empirical base of the variability of practice hypothesis: A critical analysis. Human Movement Science, 9(3–5), 387–435. https://doi.org/10.1016/0167-9457(90)90010-B Vidal, P.P., & Lacquaniti, F. (2021). Perceptual-motor styles. Experimental Brain Research, 239(5), 1359–1380. https://doi.org/10.1007/s00221-021-06049-0 Yao, W.X., DeSola, W., & Bi, Z.C. (2009). Variable practice versus constant practice in the acquisition of wheelchair propulsive speeds. Perceptual and Motor Skills, 109(1), 133–139. https://doi.org/10.2466/pms.109.1.133-139 JMLD Vol. 13, No. 2, 2025 472 BILAL, BONGERS, AND JACOBS Unauthenticated | Downloaded 11/13/25 03:28 PM UTC