Force Adaptation across Ages: Investigation of Internal Models in Early Childhood and Adulthood

doi:10.21203/rs.3.rs-4862796/v1

Download PDF

Article

Force Adaptation across Ages: Investigation of Internal Models in Early Childhood and Adulthood

https://doi.org/10.21203/rs.3.rs-4862796/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Adapting movements to rapidly changing conditions is fundamental for interacting with our dynamic environment. This adaptability relies on internal models that predict and evaluate sensory outcomes to adjust motor commands. Even infants anticipate object properties for efficient grasping, suggesting the use of internal models. However, how internal models are adapted in early childhood remains largely unexplored. This study investigated a naturalistic force adaptation task in 1.5-, 3-year-olds, and young adults. Participants opened a drawer with temporarily increased resistance, creating sensory prediction errors between predicted and actual drawer dynamics. After perturbation, all age groups showed lower peak speed, longer movement time, and more movement units with trial-wise changes analyzed as adaptation process. Results revealed no age differences in adapting peak speed and movement units, but 1.5- and 3-year-olds exhibited higher trial-to-trial variability and were slower in adapting their movement time, although they also adapted their movement time more strongly. Upon removal of perturbation, we found significant aftereffects across all age groups, indicating effective internal model adaptation. These results suggest that even 1.5-year-olds form internal models of force parameters and adapt them to reduce sensory prediction errors, possibly through more exploration and with more variable movement dynamics compared to adults.

motor adaptation

motor learning

development

cognition

toddlers

Around the first year of life, children use internal models to guide their grasping behavior ^⁠^1–5. Internal models are representations of the external world used to predict and simulate behavior based on sensory information and convert them into motor actions^⁠6,7. Previous research indicated that infants can simulate their behavior and select appropriate actions dependent on force requirements when using both hands to lift heavy objects and a single hand for light objects^⁠1. Further, they lift light objects higher than heavy objects^⁠5 and selectively scale their forces even when no visual information is provided^⁠4. Moreover, they exhibit different neural responses when processing objects of varying weights^⁠2. However, in our dynamic environment, conditions often change unpredictably, for example when opening doors or drawers with unexpected resistances. Then, the applied force of the chosen action needs to be spontaneously adapted to allow an efficient interaction with current object’s properties. Research has shown that children aged four and older can adapt their forces^⁠8–10. However, it remains unclear whether and how force adaptation unfolds in even younger children. This study examines to what extent children at 1.5- and 3-years of age can adapt their forces while grasping and pulling an object whose force requirements change unpredictably. This can provide insights to which extent young children are able to adapt their internal models when experiencing mismatches between their predicted and actual sensory outcomes.

Force adaptation, a subcategory of motor adaptation, refers to a neural process within motor control whereby internal models are adapted in response to external force perturbations, such as changes in resistance when moving an object or handle^11–13. Participants initially perform a baseline block to measure their individual motor performance. Subsequently, an adaptation block introduces an external perturbation to the resistance of the target object, resulting in sensory prediction errors, i.e. mismatches between predicted and actual sensory consequences of movements. During the adaptation block, these sensory prediction errors reduce gradually through internal model adaptation. Once the perturbation is removed, aftereffects emerge as errors in the opposite direction to the baseline, resulting from the continued use of adapted movements despite the absence of the perturbation. During a following deadaptation (or washout) block, aftereffects (over- or undershoot of movements) are gradually reduced as the internal model deadapts to the original task conditions^11–16.

Although force adaptation in adults has been widely explored in laboratory settings^13,17–20, and more naturalistic environments²¹, literature on the development of force adaptation in children is relatively limited. Previous studies investigated force adaptation in children aged 4–11 years^⁠8–10 in laboratory settings. In these studies, children performed goal-directed forearm extension and flexion movements on a robotic manipulandum that applied velocity-dependent forces altering the movement direction^⁠8,9, or they moved a robot handle that applied force-fields to a specific target¹⁰. The results demonstrated that movements were initially perturbed in all participants, regardless of age. Trial-to-trial adaptation showed age-differences only in some parameters. For instance, younger children exhibited prolonged adaptation times^⁠8,10 and deadaptation times^⁠9,10, as well as greater movement variability compared to adults. The authors concluded that motor adaptation functions similarly across this age range, although the internal model parameters appear to be less fine-tuned and more affected by neuromotor noise in younger participants^⁠8–10. These findings on force adaptation align with motor adaptation studies using visuomotor adaptation tasks^16,22,23 and gait adaptation^24–26. However, it remains unclear when this ability starts to function and to what extent children below 4 years can adapt their forces to unpredictable changes.

In the present study, we investigated force adaptation in 1.5-year-olds who have recently acquired critical motor milestones^27–29 and in 3-year-olds, who display nearly adult-like kinematics^⁠8,29–32. A control group of young adults was also tested. We employed an experimental, more naturalistic task than the previously reported. The approach was chosen to align with the motor and cognitive abilities of our younger age groups and to ensure that all participants had prior experience and corresponding neural representations (i.e., internal models) for the given task. In the task, participants were asked to open a drawer whose resistance was temporarily increased. Thus, participants experienced spontaneous sensory prediction errors between intended and actual drawer dynamics. Rewards were provided after each trial regardless of task performance. Consequently, sensory errors between internal models and action effects were the main drivers of the adaptation and deadaptation processes we investigated. Figure 1 presents our hypothetical adaptation and deadaptation processes of the drawer opening peak speed, movement time and number of movement units. This is based on (de)adaptation curves demonstrated in previously presented studies with adults. Concerning the age comparison, we hypothesized that even 1.5- and 3-year-olds would show force adaptation and deadaptation due to their developed ability to predict and interact with objects after the first year of life^{⁠3–5,11,28,33–36}. However, based on force-adaptation studies with older children (4 years and above), we hypothesized that children would show lower adaptation and deadaptation performance, represented by a slower return of peak speed, movement time, and movement units to baseline average, along with higher residual errors at the end of both experimental blocks.

Participants

Thirty-seven 1.5-year-olds, twenty-five 3-year-olds, and twenty-four young adults, who served as a control group, participated in this study. The 1.5- and 3-year-olds were recruited by obtaining their birth records from local municipal councils and neighboring communities, and contacting their parents by mail or e-mail. Adult participants were undergraduate students at the Justus Liebig University Giessen. Data of eighteen 1.5-year-olds were excluded from the analysis due to lack of cooperation (n = 7), technical problems (n = 3), or early termination of the experiment (n = 8). Additionally, six 3-year-olds (lack of cooperation: n = 2, technical problems n = 4) and four adults (technical problems: n = 1; left-handedness (n = 3) were excluded. The final sample comprised nineteen 1.5-year-olds, nineteen 3-year-olds, and twenty young adults with a middle to high socio-economic status. Anthropometric data for the final sample are outlined in Table 1. All included adults were right-handed or ambidextrous (n = 2), as measured by the Edinburgh Handedness Inventory³⁷. The 1.5- and 3-year-olds received a gift and a certificate for participation, while adults were compensated with course credits or a 5 euro voucher. This study was conducted in accordance with the German Psychological Society (DGPs) research ethics guidelines. The Office of Research Ethics at the University of Giessen approved the experimental procedure and the informed consent protocol. Written informed consent was obtained from the parents of the children or the adult participants themselves before study participation.

Table 1

Anthropometric data of participants
		3-year-olds	Young adults
N	19	19	20
Age	17.4 ± 0.6 months	40.0 ± 2.2 months	22.7 ± 2.4 years
Gender (f/m/d)	7/12/0	13/6/0	18/2/0
Height (cm)	83.3 ± 2.4	100.5 ± 4.6	169 ± 8.6
Weight (kg)	11.6 ± 1.2	16.1 ± 2.1	61.3 ± 15.5
Arm length (cm)	31.4 ± 2.4	39.9 ± 2.9	66.7 ± 13.0

Apparatus

The experimental setup consisted of two small tables (height: 45 cm), a drawer (dimensions: 42 x 28.5 x 14 cm), an age-appropriate chair, 36 cubes (2 x 2 cm) of different colors, a box for collecting the cubes, and a curtain above the drawer (Fig. 2a). The drawer was equipped with two green knobs (scope: 10 cm, diameter: 3 cm), positioned 9 cm from each edge of the drawer. The drawer lacked roller rails, so the movement would stop immediately when no force was applied. Participants were instructed to use only the left knob. At the back of the drawer, there was a cutout through which small cubes could be inserted into the drawer. Additionally, the drawer had an aperture (15 cm) at the upper part, requiring participants to fully open the drawer before accessing cubes through the back slot (4.5 cm). Drawer resistance manipulation was adjusted for the various age groups and varied for each age group: a weight of 150 g for 1.5-year-olds, 250 g for 3-year-olds, and 500 g for adults was used. This resistance was achieved by attaching weights via a small string that passed over a pulley located at the back of the drawer (Fig. 2b). The chosen age-related drawer resistances were piloted beforehand to ensure that the change of drawer resistance was perceived comparably across the age groups (see Supplementary Fig. S1 online).

A child-friendly curtain in front of the participants covered the experimenter and any weight adjustments or inserted cubes of the drawer. Participants sat on an age-appropriate chair at the left side of the drawer. In front of the participants, a box was positioned with an animal (monkey, sheep, or elephant) with a cutout (2.5 x 2.5 cm) in the cover.

Drawer kinematics were recorded using a Vicon Nexus 2.2.3 motion capture system (Vicon, Oxford, England). Six infrared cameras (Bonita) tracked the motion of two small reflective markers (6 mm in diameter) located on the front right and lateral lower edge of the drawer, with a sampling frequency of 100 Hz.

Experimental procedure

Participants (and their parents) were welcomed at the department entrance and familiarized with the laboratory environment. Two experimenters were present during the testing for the children. One communicated with families and administered the technical equipment, while the other supervised and interacted with the child. The parents of the children and the participating students were informed about the procedure and were given consent forms to sign. Meanwhile, one experimenter engaged with the child about the animals depicted on the curtain. Once consent forms were signed, the experimental setup was explained. The experimenter opened the drawer, took the inserted cube, and handed it to the child. The participant then placed the cube into the animal box nearby. All participants were instructed to open the drawer several times to collect little cubes to “feed” the animal depicted on the box (i.e. yellow cubes as bananas for the monkey, blue cubes as water for the elephant and green cubes as grass for the sheep). To do that, participants were asked to open the drawer with their right hand only and use only the left knob on the drawer. Once the drawer was fully opened, the cube was collected and placed through the box’s cutout. The animal box was replaced with another one after every sixth trial to maintain participant motivation. The order of appearance of the three animals on the boxes was randomized across blocks and participants. After two warm-up trials, the experiment began with its three blocks: the baseline block (trials 3–12) with normal drawer resistance (around 50 N), the adaptation block (trials 13–24) with increased drawer resistance (adults: 500 g, 3-year-olds: 250 g, 1.5-year-olds: 150 g), and the deadaptation block (trials 25–36) where drawer resistance returned to baseline level by removing the weights. Participants were not informed about the change of drawer resistance before the experiment.

Data preprocessing

Vicon Nexus 2.2.3 (Vicon, Oxford, England) was used to preprocess the kinematic data captured from the drawer. First, we reconstructed and labeled the data. Missing markers were interpolated using the pattern fill algorithm, as both markers consistently maintained the same distance and moved together in the same direction. When both markers were occluded simultaneously spline fill interpolation was applied with a maximum gap length of 500 samples³⁸. A Butterworth low-pass filter with a cutoff frequency of 8 Hz was applied to the data. All following data analyses were conducted using MATLAB R2022a (MathWorks, Natick, MA, USA).

Data analysis

We obtained the three-dimensional position of the infrared markers on the drawer for each trial and calculated drawer opening speed by numerically differentiating positional data. The movement onset was defined as the first point at which the drawer movement speed exceeded 30 mm/s. Movement offset was determined when drawer speed fell below 30 mm/s and was less than 3.5 cm away from the drawer’s maximum pullout position.

We calculated peak speed of the drawer opening as the maximum speed detected between movement onset and offset for each trial. Movement time defined the duration between movement onset and offset. Movement units indicated the number of local maxima in drawer speed and served as a measurement of motor planning. Well-planned adult movements are typically bell-shaped with exactly one movement unit, whereas infants and children often exhibit multiple movement units^30,39,40.

Outliers were corrected first at the individual level. Trials deviating more than the mean ± 2.5 standard deviations from the individual’s remaining trials for peak speed [0.88%], movement time [2.70%], and movement units [1.91%] were excluded. The first adaptation and deadaptation trials were exempt from individual’s outlier correction due to their expected divergence from the remaining trials. At the group level, data (including first (de)adaptation trials) deviating more than the mean ± 2.5 standard deviations from the group mean for peak speed [0.02%] movement time [0.03%], and movement units [0.02%] were excluded.

Statistical analyses

Statistical analyses were performed using JASP (Version 0.18.3, JASP Team 2024) and RStudio (Version 2023.12.0, RStudio Team 2023). All tests were conducted at a significance level of α < .05, with post-hoc tests Bonferroni-corrected. Errorbars and variances in statistical tests are presented as standard errors (s.e.m). For analyses in the adaptation block, repeated measures ANOVAs were conducted to test for differences in peak speed, movement time, and movement units between the mean of baseline block and the first adaptation trial (initial error), as well as between the mean of baseline and the mean of the last three adaptation trials (residual error adaptation block), with age as a between-subjects factor. To examine the course of adaptation, linear mixed model analyses were performed in Rstudio using the lme4 package⁴¹. Fixed effects included trials and age, with subjects treated as random effect. The significance was calculated using the lmerTest package⁴², which applies the Kenward-Rogers method for estimating degrees of freedom and generating p-values for mixed models. A priori defined customized contrasts were conducted when a significant fixed effect of age or the interaction of age and trials were indicated. Additionally, the trend function was used to investigate post-hoc slope differences. The model specification for the adaptation block was as follows: dependent variable ~ 1 + Trial * Age + (1|subject). Model fits and extended outputs are outlined in the Supplementary Table S1-S6 online. Similar analyses were conducted for the deadaptation process with regard to peak speed, movement time, and movement units. A repeated measures ANOVA was used to investigate the aftereffect as described by the comparison of those measures between the first deadaptation trial to the baseline average, while the residual error in the deadaptation block compared the average of the last three deadaptation trials with the baseline average. The model specification for the linear mixed model in the deadaptation block was as follows: dependent variable ~ 1 + Trial * Age + (1|subject).

(I) Adaptation

Figure 3 shows that the speed profiles of the drawer opening movement were perturbed by the spontaneous increase in drawer resistance, but adapted in subsequent trials.

Initial Error between Baseline and First Adaptation Trial

The 3 (age) x 2 (initial error: baseline vs. first adaptation trial) repeated measures ANOVAs for each dependent variable (peak speed, movement time, number of movement units) revealed significant initial errors of all variables across all age groups.

In detail, peak speed showed a main effect of initial error, F(1, 54) = 65.99, p < .001, η²_p = 0.55, with a significant decrease from baseline (M = 354.57 ± 10.19 mm/s) to the first adaptation trial (M = 266.12 ± 10.28 mm/s). No main effect of age, F(2, 54) = 1.17, p = .317, η²_p = 0.04, or interaction between age and initial error, F(2, 54) = 1.46, p = .242, η²_p = 0.05, was found (Fig. 4a).

Movement time revealed a significant main effect of initial error, F(1, 52) = 30.14, p < .001, η²_p = 0.37, with an increase from baseline (M = 990.89 ± 35.38 ms) to the first adaptation trial (M = 1338.91 ± 77.10 ms). A main effect of age was detected, F(2, 52) = 6.08, p = .004, η²_p = 0.19, with shorter movement time in adults (M = 945.28 ± 37.82 ms) compared to 3-year-olds (M = 1242.21 ± 85.66 ms), p_bonf = .035, d = ˗0.71, and compared to 1.5-year-olds (M = 1319.41 ± 129.74 ms), p_bonf = .005, d = ˗0.89. There was no significant interaction between age and initial error, F(2, 52) = 1.06, p = .355, η²_p = 0.04 (Fig. 4b).

Movement units displayed a significant main effect of initial error, F(1, 54) = 18.02, p < .001, η²_p = 0.25 with an increase from baseline (M = 1.33 ± 0.04) to the first adaptation trial (M = 1.65 ± 0.08). A main effect of age was identified, F(2, 54) = 18.49, p < .001, η²_p = 0.41. Post-hoc tests revealed fewer movement units in adults (M = 1.13 ± 0.10) compared to 3-year-olds (M = 1.65 ± 0.08), p_bonf < .001, d = ˗1.23, and compared to 1.5-year-olds (M = 1.69 ± 0.11), p_bonf < .001, d = ˗1.33. No significant interaction between age and initial error was determined, F(2, 55) = 0.23, p = .799, η²_p = 0.01 (Fig. 4c). A histogram for the number of movement units for all age groups is attached in the Supplementary Figure S2 online.

Residual Errors in Adaptation Block

The 3 (age) x 2 (residual error: baseline vs. mean last three adaptation trial) repeated measures ANOVAs revealed non-significant residual errors of all variables across all age groups.

Hence, peak speed displayed no main effect of residual error, F(1, 55) = 3.70, p = .059, η²_p = 0.06), indicating no significant differences between end of adaptation (M = 341.34 ± 11.36 mm/s) and baseline (M = 360.20 ± 11.49 mm/s). A main effect of age was identified, F(2, 55) = 7.10, p = .002, η²_p = 0.21), with higher peak speed in adults (M = 360.47 ± 11.36 mm/s) compared to 3-year-olds (M = 303.10 ± 15.15 mm/s), p_bonf = .044, d = 0.71, and in 3-year-olds compared to 1.5-year-olds (M = 388.22 ± 25.75 mm/s), p_bonf = .002, d = ˗1.06. The interaction between age and residual error was not significant, F(2, 55) = 1.43, p = .249, η²_p = 0.05 (Fig. 4a).

Movement time obtained no main effect of residual error, F(1, 55) = 1.21, p = .276, η²_p = 0.02, suggesting that movement time did not differ significantly between baseline (M = 989.64 ± 34.02 ms) and end of adaptation (M = 1031.58 ± 44.42 ms). A main effect of age was detected, F(2, 55) = 10.91, p < .001, η²_p = 0.28, with shorter movement time in adults (M = 831.72 ± 25.09 ms) compared to 3-year-olds (M = 1152.61 ± 60.24 ms), p_bonf < .001, d = ˗1.19, and compared to 1.5-year-olds (M = 1052.34 ± 84.54 ms), p_bonf = .008, d = ˗0.82). No significant interaction between age and residual error was observed, F(2, 55) = 2.32, p = .108, η²_p = 0.08 (Fig. 4b).

Movement units revealed no main effect of residual error, F(1, 55) = 0.09, p = .763, η²_p < 0.01. Thus, the number of movement units did not differ significantly between baseline (M = 1.33 ± 0.04) and end of adaptation (M = 1.32 ± 0.05). We found a main effect of age, F(2, 55) = 34.55, p < .001, η²_p = 0.56, with adults (M = 1.00 ± 0.00), having fewer movement units compared to 3-year-olds (M = 1.53 ± 0.07), p_bonf < .001, d = ˗1.99, and compared to 1.5-year-olds (M = 1.46 ± 0.08), p_bonf < .001, d = ˗1.72. We found a significant interaction between age and residual error, F(2, 55) = 4.83, p = .012, η²_p = 0.15. Compared to the baseline, adults had equal movement units at the end of adaptation (M_diff = 0.00 ± 0.00), 3-year-olds had more movement units at the end of adaptation (M_diff = 0.14 ± 0.01), and 1.5-year-olds had fewer movement units at the end of adaptation (M_diff = ˗0.17 ± 0.04) (Fig. 4c).

Linear Mixed Model Trial-wise Adaptation Analysis

Linear mixed models were applied to each dependent variable, with trials (13–24) and age as fixed effects, and subject as a random factor, to analyze the trial-to-trial adaptation process between age groups (cf., statistical analyses). Model fits and estimates are presented in the Supplementary Tables S4-6 online.

A significant main effect of trials on peak speed was found, F(1, 618.29) = 29.72, p < .001, η_p² = 0.05. This supports the adaptation process (i.e. an increase) of peak speed with subsequent trials. Peak speed was not significantly different between age groups across all trials, F(2, 504.80) = 0.20, p = .822, η_p² < 0.01. There was no significant interaction effect between age and trials, F(2, 618.29) = 1.77, p = .171, η_p² < 0.01 (Fig. 5a).

We found a significant main effect of trials on movement time, F(1, 602.46) = 33.72, p < .001, η_p² = 0.05. Thus, movement time adapted (i.e. increased) with subsequent trials. Further, a main effect of age was identified, F(2, 570.34) = 7.66, p < .001, η_p² = 0.03. Customized contrasts showed that adults had shorter movement time compared to 3-year-olds (p < .001), and compared to 1.5-year-olds (p = .008). We found a significant interaction between age and trials, F(2, 602.46) = 3.30, p = .038, η_p² = 0.03. Post-hoc analysis with customized contrasts and trend function revealed that adults exhibited smaller reduction in movement time across trials with an estimated slope of ˗10.1 ms per trial. In contrast, 3-year-olds showed a significant reduction in movement time of ˗21.2 ms per trial, while 1.5-year-olds had the greatest reduction in movement time with ˗33.6 ms per trial. Statistically adults reduced their movement time significantly less than 1.5-year-olds (p = .028). Descriptively, adults reduced their movement time to baseline level already in the second trial, while children needed more trials to adapt their movement time and had a higher trial-to-trial variability (see Fig. 5b). To check the trial-by-trial variability statistically we conducted a 1 x 3(age) ANOVA on the participants individual standard deviation of the movement time divided by the individual mean of the movement time¹⁰. A significant main effect of age on trial-to-trial variability was found, F(2, 55) = 15.48, p < .001, η_p² = 0.36. Post-hoc tests revealed a significantly lower trial-to-trial variability in adults (M = 0.13 ± 0.02 ms) compared to 3-year-olds (M = 0.31 ± 0.03 ms), p_bonf < .001, d = ˗1.23, and compared to 1.5-year-olds (M = 0.39 ± 0.05 ms), p_bonf < .001, d = ˗1.73.

For movement units, we found a main effect of trials, F(1, 610.83) = 20.85, p < .001, η_p² = 0.03. Movement units were reduced with subsequent trials. A main effect of age, F(2, 663.56) = 3.79, p = .023, η_p² = 0.01 was found, with adults having fewer movement units compared to 3-year-olds (p < .001) and compared to 1.5-year-olds (p < .001). No interaction between age and trials was detected, F(2, 610.82) = 0.21, p = .813, η_p² < 0.01 (Fig. 5c).

(II) Deadaptation

Figure 6 presents the averaged speed profiles of the drawer movement within the 12 trails of the deadaptation block compared to the averaged baseline trials and the averaged last three adaptation trials for all three age groups.

Aftereffects Between First Deadaptation Trial and Baseline

To investigate whether aftereffects of the adaptation block existed when reducing the drawer resistance back to baseline level, we conducted for each dependent variable a 3 (age) x 2 (aftereffect: baseline vs. first deadaptation trial) repeated measures ANOVA.

Peak speed showed a significant main effect of aftereffects, F(1, 53) = 10.09, p = .002, η²_p = 0.16. Thus, peak speed in the first trial of deadaptation (M = 424.97 ± 19.84 mm/s) overshot peak speed of the baseline (M = 358.57 ± 11.84 mm/s) significantly. A main effect of age was observed, F(2, 53) = 5.12, p = .009, η²_p = 0.16, with 3-year-olds (M = 346.96 ± 18.56 mm/s) having lower peak speed than 1.5-year-olds (M = 439.40 ± 38.96 mm/s), p_bonf = .007, d = ˗0.79. No significant interaction between age and aftereffect was revealed, F(2, 53) = 0.63, p = .537, η²_p = 0.02 (Fig. 4a).

Movement time exhibited a main effect of aftereffects, F(1, 52) = 5.32, p = .025, η²_p = 0.09. Movement time at the first deadaptation trial (M = 844.73 ± 44.60 ms) undershot the baseline average (M = 971.80 ± 33.43 ms). A main effect of age was shown, F(2, 52) = 7.56, p = .001, η²_p = 0.23, indicating that adults had shorter movement time (M = 774.30 ± 30.26 ms) compared to 3-year-olds (M = 957.58 ± 44.45 ms), p_bonf = .014, d = ˗0.66, and compared to 1.5-year-olds (M = 1000.35 ± 97.98 ms), p_bonf = .002, d = ˗0.81. No significant interaction between age and aftereffect was detected, F(2, 52) = 0.31, p = .734, η²_p = 0.01 (Fig. 4b).

Movement units did not reveal a main effect of aftereffect, F(1, 54) = 1.14, p = .290, η²_p = 0.02. However, a main effect of age was present, F(2, 54) = 11.49, p < .001, η²_p = 0.30, with fewer movement units in adults (M = 1.10 ± 0.09) compared to 3-year-olds (M = 1.47 ± 0.11), p_bonf = .001, d = ˗0.88, and compared to 1.5-year-olds (M = 1.54 ± 0.10), p_bonf < .001, d = ˗1.06. There was no significant interaction between age and aftereffect, F(2, 54) = 0.65, p = .527, η²_p = 0.02 (Fig. 4c).

Residual Error in Deadaptation Block

To investigate whether participants deadapted their movements to the baseline level at the end of deadaptation, we used again 3 (age) x 2 (residual error: mean of last three deadaptation trials vs. baseline) repeated measures ANOVAs for each dependent variable.

Peak speed showed a main effect of residual error, F(1, 55) = 10.45, p = .002, η²_p = 0.16, with still higher peak speed at the end of deadaptation (M = 392.71 ± 11.88 mm/s) compared to baseline (M = 360.20 ± 11.49 mm/s). A significant main effect of age was shown, F(2, 55) = 4.67, p = .013, η²_p = 0.15, with lower peak speed in 3-year-olds (M = 335.54 ± 14.91 mm/s) than in 1.5-year-olds (M = 409.56 ± 27.29), p_bonf = .012, d = ˗0.87. No significant interaction was revealed between age and residual error, F(2, 55) = 0.25, p = .780, η²_p < 0.01 (Fig. 4a).

For movement time, we found a significant main effect of residual error, F(1, 55) = 17.61, p < .001, η²_p = 0.24, with still higher movement time at the end of deadaptation (M = 986.64 ± 34.02 ms) compared to baseline (M = 855.32 ± 28.83 ms). A main effect of age, F(2, 55) = 9.69, p < .001, η²_p = 0.26, showed that adults (M = 777.18 ± 25.82 ms) had shorter movement time than 3-year-olds (M = 1011.86 ± 48.54 ms), p_bonf < .001, d = ˗1.08, and shorter movement time than 1.5-year-olds (M = 981.47 ± 67.08 ms), p_bonf = .003, d = ˗0.94. The interaction between age and residual error was not significant, F(2, 55) = 1.33, p = .272, η²_p = 0.05 (Fig. 4b).

Movement units displayed a significant main effect of residual error, F(1, 55) = 24.20, p < .001, η²_p = 0.31, with fewer movement units at the end of deadaptation (M = 1.17 ± 0.03) compared to baseline (M = 1.33 ± 0.04). A main effect of age was determined, F(2, 55) = 40.34, p < .001, η²_p = 0.60, with fewer movement units in adults (M = 1.00 ± 0.00) compared to 3-year-olds (M = 1.33 ± 0.05), p_bonf < .001, d = ˗1.62, and compared to 1.5-year-olds (M = 1.43 ± 0.06), p_bonf < .001, d = ˗2.16. The interaction between age and residual error was significant, F(2, 55) = 6.34, p = .003, η²_p = 0.19. Adults and 1.5-year-olds had fewer movement units at the end of adaptation compared to baseline, while 3-year-olds had slightly more movement units (M_{diff_adults} = 0.00 ± 0.00, M_{diff_3−yo} = 0.27 ± 1.33, M_{diff_1.5−yo} = ˗0.22 ± 1.434) (Fig. 4c).

Linear Mixed Model Trial-wise Deadaptation Analysis

Linear mixed models were applied to each dependent variable, with trials (25–36) and age as fixed effects, and subject as a random factor.

For peak speed, we found a significant main effect of trials, F(1, 613.52) = 12.27, p < .001, η_p² = 0.02, indicating that peak speed was significantly deadapted. No significant main effect of age, F(2, 653.55) = 3.01, p = .050, η_p² < 0.01 was detected. The interaction between age and trials, F(2, 613.52) = 1.22, p = .295, η_p² < 0.01 was not significant. This suggests that peak speed was not reduced differently with subsequent trials between age groups (Fig. 5a).

For movement time, we found no main effect of trials, F(1, 599.77) = 0.23, p = .634, η_p² < 0.01. This suggests that movement time was not deadapted. We also found no main effect of age, F(2, 644.76) = 0.90, p = .408, η_p² < 0.01. Furthermore, the interaction between trials and age was not significant, F(2, 599.74) = 0.75, p = .475, η_p² < 0.01 (Fig. 5b).

However, a main effect of trials on the number of movement units, F(1, 616.97) = 9.49, p = .002, η_p² = 0.02 was found. Thus, the number of movement units decreased with subsequent trials. No main effect of age, F(1, 651.05) = 2.06, p = .128, η_p² < 0.01 was revealed as well as and no interaction between age and trials, F(2, 616.96) = 0.49, p = .610, η_p² < 0.01 (Fig. 5c).

This study explored how 1.5- and 3-year-olds adapt their internal models compared to adults using a naturalistic force adaptation task. Participants opened a drawer with varying resistance across three blocks: baseline (normal resistance), adaptation (increased resistance), and deadaptation (baseline resistance). To our knowledge, this is the first study focusing on force adaptation in children under 4 years. This approach isolated the sensory error between predicted and actual action effects as primary driver of adaptation and deadaptation processes.

Force Adaptation Behavior Across all Age Groups

The unexpected increase of drawer resistance significantly perturbed the drawer opening in all age groups, indicated by lower peak speed, longer movement time and more movement units. No differences in initial errors across age groups suggest that the adjusted weights (500 g for adults, 250 g for children and 150 g for infants) successfully and comparably perturbed drawer dynamics (cf. Supplementary Material).

1.5- and 3-year-olds exhibited trial-to-trial force adaptation that was not different to adults in peak speed and movement units. Peak speed increased while movement units decreased within subsequent adaptation trials. This extends existing literature that also children as young as 1.5 years show force adaptation comparable to adults^⁠8–10. Further, it supports evidence from weight perception studies^⁠1–5 that internal models of everyday movement help to select appropriate action and facilitate adapting actions to unpredictable changes. Moreover, it aligns with motor adaptation studies on walking^24–26 and sitting³⁶ before four years of age. Our study showed that children adapt their internal models of force application between expected and actual sensory outcomes through sensory prediction errors, without the need for reward (prediction) errors. Previous research indicates that sensorimotor adaptation can occur without reward prediction errors, but it is less extensive and slower with only sensory prediction errors than with both sensory and reward prediction errors⁴³.

Despite similarities in adaptation, there were age differences in movement time. Adults adapted their movement time quickly within the first adaptation trials, whereas 1.5- and 3-year-olds needed longer but therefore showed stronger adaptation processes, possibly due to their higher movement time at the beginning of the experiment and adaptation block. Further, children’s trial-to-trial variability was higher compared to adults. This supports other findings that children (> 4 years) show more variable and slower trial-to-trial adaptation than adults^⁠8–10, potentially due to less fine-tuned internal models^⁠9 and ongoing nervous system development¹⁰. This result highlights that variability appears to be an important component of motor learning and development^28,44. Indeed, motor development is characterized by variability in trajectory, velocity, amplitude, and duration of movement, typically including multiple submovements that can be indicated by more movement units directed toward the object and an indicator for the typical development of motor abilities^28,36,40,45. In our study, children were likely still using a more explorative control policy to strengthen their motor skills, which might have resulted in higher trial-to-trial variability, especially in 1.5-year-olds. Additionally, adults showed higher peak speed, shorter movement time and fewer movement units than children, indicating that children’s motor performance still differed from adult levels. This is in line with the understanding that motor control fine-tuning continues to develop through childhood, influenced by ongoing changes in the corticospinal tract^46,47.

At the end of adaptation, across all age groups peak speed and movement time returned to baseline, indicating successful adaptation. In contrast, adaptation in terms of the number of movement units varied across age groups. Adults maintained the same number of movement units (one movement unit – celling effect), while 3-year-olds had more, and 1.5-year-olds had fewer movement units compared to baseline. Movement units indicate motor planning quality, with fewer units suggesting advanced forward planning and more movement units suggesting reliance on online control mechanisms^30,39,40. In our drawer-opening task, using one or multiple movement units could be interpreted differently. Ideally, one movement unit with a soft stop shows optimal motor planning. However, one unit can also result from excessive force with a harsh stop, which is not necessarily more advanced than pulling, pausing, and continuing gently, leading to multiple units. For instance, 1.5-year-olds might use one unit with a harsh stop, while 3-year-olds use more units with a soft stop. Therefore, fewer movement units do not always indicate better motor planning in our task. Thus, we consider movement units a weaker indicator of motor planning quality compared to peak speed and movement time.

Force Deadaptation Across all Age Groups

All age groups showed aftereffects in peak speed and movement time when drawer resistance returned to baseline, indicating successful internal model adaptation. In response to the aftereffect, all age groups deadapted their peak speed and movement units to a comparable extent with subsequent trials. However, by the end of deadaptation, residual errors in peak speed, movement time, and movement units remained different from baseline across all age groups. Peak speed was still higher by the end of deadaptation, while movement time was still shorter than in the baseline, indicating incomplete deadaptation. While previous studies suggested 10–12 trials per block were sufficient for adaptation and deadaptation^⁠8,9, more trials might be necessary for a complete deadaptation of all parameters (50 trials/block¹⁰, 40 trials/ block¹⁶). The number of movement units was lower than in the baseline. This reduction may be attributed to improved motor planning with repetitions³⁹.

Limitations and Future Implications

The current study has some limitations that should be considered in future research and when interpreting the results. The drop-out rate was relatively high, especially among 1.5-year-olds who tend to have higher drop-out rates due to lack of cooperation (e.g., distraction by technical equipment, behind-the-scene curiosity or movement restlessness). The number of trials was relatively low due to the methodological adjustments necessary for testing the young age groups. Thus, we cannot conclude whether the deadaptation process was not completed due to the low number of trials or for other reasons. Unlike other force adaptation tasks^⁠8–10, we did not provide a specific task goal or outcome feedback. Instead, we used the individual baseline average as individual adaptation goal and relied solely on sensor prediction errors to drive adaptation. Thus, our task was more naturalistic and individualized. However, in further studies, a specific task goal with the possibility of task errors may be included to investigate how young children adapt in more restricted circumstances⁴³. Further studies may also consider the muscle pre-activation (e.g., measured with EMG) to obtain data about the predicted force application and the correction of muscle activation during the movement.

In summary, this study provides evidence that 1.5- and 3- year-olds applied force adaptation to reduce sensory prediction errors, thereby matching their expected and actual movement outcomes. This suggests that internal models are built and adapted at a very young age when the environment changes unpredictably, making this process a very early component of motor control. To understand whether this is an even earlier skill, we look forward to future research that further reduces the age of study and examines force adaptation in infants during the first year of life.

Acknowledgements

We thank Anna Grab and Federica Thompson with recruitment and data collection, and all the families and students participating in this study. This research was funded by the Hessian Ministry for Higher Education, Research, Science and the Arts (Cluster Project “The Adaptive Mind”). JT is supported by the Johanna Quandt Foundation.

Author contribution

LF conception, design, acquisition, analysis, interpretation of data; writing and editing draft; JF analysis, interpretation of data, revision, editing; FL revision, editing; YLS revision, editing; JT revision, editing; GS design, interpretation of data, revision, editing, supervision

Data availability statement

Data and analysis code related to this study can be publicly found at: DOI 10.17605/OSF.IO/B4H9V

Additional information

The authors declare no competing interests.

Upshaw, M. B. & Sommerville, J. A. Twelve-month-old infants anticipatorily plan their actions according to expected object weight in a novel motor context. Front. Public Health 3, 1–7; 10.3389/fpubh.2015.00032 (2015).
Marshall, P. J., Saby, J. N. & Meltzoff, A. N. Infant Brain Responses to Object Weight: Exploring Goal-Directed Actions and Self-Experience. Infancy 18, 942–960; 10.1111/infa.12012 (2013).
Schwarzer, G. & Jovanovic, B. Infants' predictive minds: The role of motor experience. Child Dev. Perspect., 1–6; 10.1111/cdep.12506 (2024).
Mash, C. Object Representation in Infants' Coordination of Manipulative Force. Infancy 12, 329–341; 10.1111/j.1532-7078.2007.tb00246.x (2007).
Gottwald, J. M. & Gredebäck, G. Infants' prospective control during object manipulation in an uncertain environment. Exp. Brain Res. 233, 2383–2390; 10.1007/s00221-015-4308-7 (2015).
Leite, C. M. F., Campos, C. E., Couto, C. R. & Ugrinowitsch, H. An internal model approach for motor behavior. BJMB 15, 356–371; 10.20338/bjmb.v15i5.273 (2021).
Wolpert, D. M., Ghahramani, Z. & Jordan, M. I. An internal model for sensorimotor integration. Science 269, 1880–1882; 10.1126/science.7569931 (1995).
Konczak, J., Jansen-Osmann, P. & Kalveram, K.-T. Development of force adaptation during childhood. J. Motor Behav. 35, 41–52; 10.1080/00222890309602120 (2003).
Jansen-Osmann, P., Richter, S., Konczak, J. & Kalveram, K.-T. Force adaptation transfers to untrained workspace regions in children: evidence for developing inverse dynamic motor models. Exp. Brain Res. 143, 212–220; 10.1007/s00221-001-0982-8 (2002).
Takahashi, C. D. et al. Neuromotor noise limits motor performance, but not motor adaptation, in children. J. Neurophysiol. 90, 703–711; 10.1152/jn.01173.2002 (2003).
Miall, R. C. & Wolpert, D. M. Forward Models for Physiological Motor Control. Neural. Netw. 9, 1265–1279; 10.1016/s0893-6080(96)00035-4 (1996).
Shadmehr, R. & Mussa-Ivaldi, F. A. Adaptive representation of dynamics during learning of a motor task. J. Neurosci. 14, 3208–3224; 10.1523/jneurosci.14-05-03208.1994 (1994).
Shadmehr, R., Smith, M. A. & Krakauer, J. W. Error correction, sensory prediction, and adaptation in motor control. Annu. Rev. Neurosci. 33, 89–108; 10.1146/annurev-neuro-060909-153135 (2010).
Tzvi, E., Loens, S. & Donchin, O. Mini-review: The Role of the Cerebellum in Visuomotor Adaptation. Cerebellum 21, 306–313; 10.1007/s12311-021-01281-4 (2022).
Reuter, E.-M., Booms, A. & Leow, L.-A. Using EEG to study sensorimotor adaptation. Neurosci. Biobehav. Rev. 134, 104520; 10.1016/j.neubiorev.2021.104520 (2022).
Ruitenberg, M. F. L., Koppelmans, V., Seidler, R. D. & Schomaker, J. Developmental and age differences in visuomotor adaptation across the lifespan. Psychol. Res. 87, 1710–1717; 10.1007/s00426-022-01784-7 (2023).
Avraham, C. & Nisky, I. The effect of tactile augmentation on manipulation and grip force control during force-field adaptation. J. Neuroeng. Rehabil, 1–19; 10.1186/s12984-020-0649-y (2020).
Rezazadeh, A. & Berniker, M. Force field generalization and the internal representation of motor learning. PloS one 14, e0225002; 10.1371/journal.pone.0225002 (2019).
Schween, R., McDougle, S. D., Hegele, M. & Taylor, J. A. Assessing explicit strategies in force field adaptation. J. Neurophysiol. 123, 1552–1565; 10.1152/jn.00427.2019 (2020).
Reuter, E.-M., Mattingley, J. B., Cunnington, R., Riek, S. & Carroll, T. J. Pushing attention to one side: Force field adaptation alters neural correlates of orienting and disengagement of spatial attention. Eur. J. Neurosci. 49, 120–136; 10.1111/ejn.14266 (2019).
Carius, D. et al. Quantifying motor adaptation in a sport-specific table tennis setting. Sci. Rep. 14, 601; 10.1038/s41598-023-50927-2 (2024).
Deng, K., Chan, J. S. Y. & Yan, J. H. Explicit Aiming Strategy Decreases the Differences in Visuomotor Adaptation between Children and Young Adults. Dev. Neuropsychol. 44, 495–512; 10.1080/87565641.2019.1675664 (2019).
Ferrel-Chapus, C., Hay, L., Olivier, I., Bard, C. & Fleury, M. Visuomanual coordination in childhood: adaptation to visual distortion. Exp. Brain Res. 144, 506–517; 10.1007/s00221-002-1064-2 (2002).
Rossi, C., Bastian, A. J. & Therrien, A. S. Mechanisms of proprioceptive realignment in human motor learning. Curr. Opin. Physiol. 20, 186–197; 10.1016/j.cophys.2021.01.011 (2021).
Musselman, K. E., Patrick, S. K., Vasudevan, E. V. L., Bastian, A. J. & Yang, J. F. Unique characteristics of motor adaptation during walking in young children. J. Neurophysiol. 105, 2195–2203; 10.1152/jn.01002.2010 (2011).
Vasudevan, E. V. L., Torres-Oviedo, G., Morton, S. M., Yang, J. F. & Bastian, A. J. Younger is not always better: development of locomotor adaptation from childhood to adulthood. J. Neurosc. 31, 3055–3065; 10.1523/JNEUROSCI.5781-10.2011 (2011).
Corbetta, D., Wiener, R. F. & Thurmann, S. L. Learning to Reach in Infancy. In Reach-to-Grasp Behavior. Brain, Behavior and Modelling Across the Life Span, edited by D. Corbetta & M. Santello (Routledge, New York and London, 2019), pp. 18–41.
Hadders-Algra, M. Variation and variability: key words in human motor development. Phys. Ther. 90, 1823–1837; 10.2522/ptj.20100006 (2010).
Adolph, K. E. & Robinson, S. R. Motor Development. In Handbook of Child Psychology and Developmental Science, edited by Lerner, R.M., Liben L.S., Mueller, U. 7th ed. (John Wiley & Sons Inc., Hoboken, NJ, 2015), Vol. 7, pp. 113–157.
Preißler, L. et al. Effects of visual and visual-haptic perception of material rigidity on reaching and grasping in the course of development. Acta Psychol. 221, 103457; 10.1016/j.actpsy.2021.103457 (2021).
Jovanovic, B. & Schwarzer, G. Learning to grasp efficiently: the development of motor planning and the role of observational learning. Vis. Res. 51, 945–954; 10.1016/j.visres.2010.12.003 (2011).
Konczak, J. & Dichgans, J. The development toward stereotypic arm kinematics during reaching in the first 3 years of life. Exp. Brain Res. 117, 346–354; 10.1007/s002210050228 (1997).
Hofsten, C. von. Development of Visually Directed Reaching: The Approach Phase. J. Hum. Mov. Stud., 160–178 (1979).
Nagai, Y. Predictive learning: its key role in early cognitive development. Philos. Trans. R. Soc. B 374, 20180030; 10.1098/rstb.2018.0030 (2019).
Gehb, G., Kubicek, C., Jovanovic, B. & Schwarzer, G. The positive influence of manual object exploration on predictive grasping for a moving object in 9-month-old infants. J. Vis. 19, 13; 10.1167/19.14.13 (2019).
Hadders-Algra, M., Brogren, E. & Forssberg, H. Ontogeny of postural adjustments during sitting in infancy: variation, selection and modulation. J. Physiolo. 493, 273–288; 10.1113/jphysiol.1996.sp021382 (1996).
Oldfield, R. C. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 9, 97–113; 10.1016/0028-3932(71)90067-4 (1971).
Vicon Motion Systems. Vicon Nexus User Guide. Nexus 2.12 documentation. Available at https://help.vicon.com/space/Nexus212/11241108/Vicon+Nexus+User+Guide.
Blischke, K. Entwicklung von Zielbewegungen. In Motorische Entwicklung, edited by Schott, N., Munzert, J. 1st ed. (Hogrefe, Göttingen, Bern, Wien, Paris), Vol. 1, pp. 103–126.
Verhaar, E., Medendorp, W. P., Hunnius, S. & Stapel, J. C. Online reach correction in 6- and 11-month-old infants. Infancy 28, 667–683; 10.1111/infa.12524 (2023).
Douglas Bates et al. Package 'lme4'. convergence, 1–129 (2015).
Kuznetsova, A., Brockhoff, P. B. & Christensen, R. H. B. lmerTest Package: Tests in Linear Mixed Effects Models. J. Stat. Soft. 82; 10.18637/jss.v082.i13 (2017).
Leow, L.-A., Marinovic, W., Rugy, A. de & Carroll, T. J. Task errors contribute to implicit aftereffects in sensorimotor adaptation. Eur. J. Neurosci. 48, 3397–3409; 10.1111/ejn.14213 (2018).
Dhawale, A. K., Smith, M. A. & Ölveczky, B. P. The Role of Variability in Motor Learning. Ann. Rev. Neurosci. 40, 479–498; 10.1146/annurev-neuro-072116-031548 (2017).
Edelmann, G. M. & Mountcastle, V. B. The mindful brain: Cortical organization and the group-selective theory of higher brain function (MIT press, Massachusetts, 1982).
Forssberg, H. et al. Development of human precision grip. II. Anticipatory control of isometric forces targeted for object's weight. Exp. Brain Res. 90, 393–398; 10.1007/BF00227253 (1992).
Eyre, J. A. Corticospinal tract development and its plasticity after perinatal injury. Neurosci. Biobehav. Rev. 31, 1136–1149; 10.1016/j.neubiorev.2007.05.011 (2007).

No competing interests reported.

SupplementaryMaterialForceAdaptationAcrossAges.doc

Download PDF

Reviewers agreed at journal
18 Nov, 2024
Reviewers invited by journal
15 Nov, 2024
Editor assigned by journal
11 Nov, 2024
Editor invited by journal
14 Aug, 2024
Submission checks completed at journal
13 Aug, 2024
First submitted to journal
05 Aug, 2024

You are reading this latest preprint version

Force Adaptation across Ages: Investigation of Internal Models in Early Childhood and Adulthood

Status:

Version 1

Abstract

Figures

Introduction

Methods

Participants

Apparatus

Experimental procedure

Data preprocessing

Data analysis

Statistical analyses

Results

(I) Adaptation

Initial Error between Baseline and First Adaptation Trial

Residual Errors in Adaptation Block

Linear Mixed Model Trial-wise Adaptation Analysis

(II) Deadaptation

Aftereffects Between First Deadaptation Trial and Baseline

Residual Error in Deadaptation Block

Linear Mixed Model Trial-wise Deadaptation Analysis

Discussion

Force Adaptation Behavior Across all Age Groups

Force Deadaptation Across all Age Groups

Limitations and Future Implications

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1