Introduction

“Pass the salt” or “switch on the light” are all actions performed to reach an object. Reaching is essential in everyday life. For example, imagine that you are in a pub and the waiter brings you a beer. You must know which actions allow you to reach the beer (Weast & Proffitt, 2018). Similar situations are often present in everyday life, in which adaptive behaviour depends on the ability to grasp action possibilities or “affordances” (Bhargava et al., 2020; Chemero & Turvey, 2007; Harrison et al., 2016; Tosoni et al., 2021).

Affordance perception

According to Gibson’s ecological approach to perception, action possibilities, called affordances, are perceived directly from coupling between the observer and the environment (Gibson, 1979), with perception embodied in the biomechanics of action (Warren, 1984). The capacity to perceive action boundaries has been related to the concept of affordance. Action boundary corresponds to the bound beyond which an environmental feature does not afford an action according to body capacity. Biomechanics of action constrain both real and perceived action boundaries. Environmental units are reported to body features, posture and behaviour (Gibson, 1979). This kind of relative evaluation of the environment involves “intrinsic measurements” rather than extrinsic measurements (i.e., standard physical measure). In their Experiment 1, Carello et al. (1989) divided the perceived maximum distance for which participants were able to reach the target (i.e., a physical feature of the environment) by arm length (i.e., an observer’s bodily feature) to assess the perceived reachability boundary. This ratio is called the estimated critical point (πc). The actual πc (i.e., the ratio between the actual maximum distance for which participants were able to reach the target and the arm length) is generally also computed to assess the motor reachability boundary. Carello et al. (1989) found that the difference between long-armed and short-armed participants for the perceived maximum distance was eliminated when it was body-scaled. This was consistent with an ecological view on reachability perception: The distance at which a target is reachable is finely perceived in one’s own scale (Carello et al., 1989). Nevertheless, the authors found a general reachability overestimation (for similar effects see also Graydon et al., 2012; Mark et al., 1997; Rochat & Wraga, 1997; Weast & Proffitt, 2018). The number of degrees of freedom during the reaching act (Carello et al., 1989), the effector used for estimations (Weast & Proffitt, 2018), the involvement of analytical reflective judgement during the task (Heft, 1993) and postural stability (Gabbard et al., 2005) are all factors that can contribute to influence estimation. Reachability perception then appears to be a complex phenomenon that depends upon several factors (Gabbard et al., 2005), some of which could extend to field affective processes.

Affective influences on affordance perception

Besides biomechanical features, there are non-visual factors affecting perception (Laurent, 2014; Proffitt, 2006). The influence of affective processes on perception of action capacities has been empirically studied through concepts such as anxiety. Bootsma et al. (1992) showed that the pickup of information specifying reaching affordance was less accurate in the presence of induced anxiety. Pijpers et al. (2006) found that induced anxiety reduced the perceived reaching height in wall climbing, as long as induced anxiety reduced actual action capabilities. Graydon et al. (2012) also found more conservative judgements in reachability perception following anxiety induction. They concluded that anxiety influences the perception of reaching affordance.

Mood is another affective process that can influence the perception of affordance. Riener et al. (2011) have suggested that negative mood is used as embodied information, leading to a more conservative perception of affordance. Vegas and Laurent (2022) consistently reported that experimentally induced mood led to decreasing sitting affordance perception following happy and sad, but not neutral, mood inductions in the laboratory. Mood shapes the active motor relationship between the environment and the observer by influencing affordance perception (Vegas & Laurent, 2022). According to the authors, mood could modify available energy levels and/or disrupt a participant’s attunement to optical variables that play key roles in action guidance and/or perceptual-motor calibration.

Depression could also influence affordance perception. According to the Diagnostic and Statistical Manual of Mental Disorders 5th Edition (DSM-5), a depressed mood (e.g., sad, empty, hopeless) is one of the main symptoms of the major depressive disorder (American Psychiatric Association (APA), 2013). Although depression and anxiety are distinct disorders, there is an overlap between them (Richter et al., 2020). They are highly comorbid and it is difficult to separate their symptoms (Kalin, 2020). It has been proposed that both negative mood and anxiety could influence affordance perception (Graydon et al., 2012; Vegas & Laurent, 2022). Moreover, in depression, the body can shut the door on affordances in the personal surrounding (Fuchs, 2005; Fuchs & Schlimme, 2009). The phenomenal space would no longer be correctly embodied. The body would become an obstacle in being experienced as conspicuous, heavy, and solid (Fuchs, 2005; Fuchs & Schlimme, 2009; Zatti & Zarbo, 2015), which would disturb affordance perception. Affordances would be less salient during depression (de Haan et al., 2013; Kiverstein et al., 2020). A central feature in the aetiology of depression is unresolvable stress (Pizzagalli, 2014). According to the learned helplessness (Pryce et al., 2011; Seligman, 1975) and action inhibition (Laborit, 1979; Li et al., 2015) models, depression is a result of the learning of lack of control over a situation, without action gratification, leading to feelings of helplessness and reduced action. Depressed people often cannot get rid of the experience of body failure (Fuchs & Schlimme, 2009). Depression, which has sometimes been conceived as a pathology of action (Canbeyli, 2010; Laborit, 1979, 1982) and as a pathology of gratification anticipation (Keren et al., 2018; Pizzagalli, 2014; Smoski et al., 2009), could be associated with modifications in affordance perception (de Haan et al., 2013; Kiverstein et al., 2020). Although depression is one of the most prevalent psychiatric conditions worldwide (World Health Organization, 2017) embodied in perception-action coupling (e.g., through oculomotor processes, Carvalho et al., 2015), and associated with psychomotor disturbances (Canbeyli, 2010), no experimental studies have investigated the relationship between affordance and mood disorders to date.

Aim and hypotheses

We addressed this issue by assessing the relationships between depressive symptoms and reachability perception. We tested the following hypotheses.

  1. 1.

    Perceived reachability boundary (estimated πc) is different between participants with depressive symptoms and healthy participants (key hypothesis). Because of the limited amount of available data in this specific study field, this hypothesis was exploratory and was not associated with directional predictions.

  2. 2.

    The modification of perceived reachability boundary associated with depressive symptoms is explained by: (a) mood (depressive symptoms have a mood component and mood is known to influence affordance perception); (b) and/or anxiety (anxiety is known to influence affordance perception and there is a close relationship between anxiety and depression).

  3. 3.

    Reachability is body-scaled at actual (motor) and estimated (perceived) levels in participants without depressive symptoms. Actual and estimated reachability boundaries (actual πc, estimated πc) do not depend upon arm length. Given the disturbance of embodiment in depression (i.e., the body could become an obstacle and be experienced as conspicuous, heavy and solid; Fuchs & Schlimme, 2009), this independence might be challenged in participants with depressive symptoms.

Method

Sample

Sampling plan

We calculated the required number of participants by an a priori power analysis. The sample size was determined for the planned 2 × 2 repeated-measures ANOVA with the Superpower R package (Lakens & Caldwell, 2019). ANOVA_exact function was used with a 2 × 2 mixed design including group (healthy participants vs. participants with depressive symptoms) as a between-participant factor and the assessment mode (estimated vs. actual) as a within-participant factor. The correlation among within-participant factors was set at .5 (i.e., correlation coefficient corresponding to large in terms of magnitude of effect sizes (Cohen, 1988); we expected a strong relationship between estimated reachability and actual reachability). Calculations were based on an alpha error set at .05 and the estimated πc difference between healthy participants and participants with depressive symptoms set at .09. No published study had measured the relationship between depressive symptoms and perceived reachability boundary. We based our predictions on studies that showed an influence of anxiety on reachability perception. Graydon et al. (2012) showed that participants who were subject to anxiety overestimated less reachability than control participants. The difference between these two groups was .09. Thus we expected that there would be a difference in πc of .09 between groups composed of healthy participants and those with participants with depressive symptoms. Due to the limited number of studies in the domain of depression and affordances, this procedure seemed to be the best. Power analysis indicated that eight participants were needed per group (16 overall) for a statistical power of 90% and the interaction between both factors.

Participants

Participants were recruited using student mailing lists from the University of Franche-Comté where the experiment took place. We recruited 21 participants who were all students.

Three participants were excluded from the initial sample and the analyses. Two participants did not complete all the affective scales and one had both estimated and actual outlier πc. Outlier values were screened with a boxplot according to Tukey’s rule (Howell, 2010), which defines outliers as data points that are either < 1.5 times the interquartile range from the first quartile or > 1.5 times the interquartile range from the third quartile. This participant’s results were excluded from inferential statistical analyses. Observed outlier values did not seem to be extreme legitimate values. Indeed, this participant was the outlier for estimated πc and actual πc, but not for actual critical distance and arm length values. The final sample included 18 participants (see Table 1 for sociodemographic and affective characteristics).

Table 1 Affective and sociodemographic characteristics of the study sample
Full size table

The study was carried out following the ethics principles of both the American Psychological Association (American Psychological Association, 2017) and the Declaration of Helsinki (World Medical Association, 2013). Participants provided written informed consent before the beginning the experiment.

Participants were explicitly informed that they could withdraw from the experiment at any moment and with no need for justification. Eligible participants were at least 18 years old and did not have uncorrected visual disorders. A debriefing was given to all participants at the end of the experiment and all participants who were found to have depressive symptomsFootnote 1 were informed about mental healthcare possibilities.

Measure and operationalization of reachability affordance

A reaching movement (Fig. 1A) can be represented by a triangle in which arm length (L), shoulder height (H), and distance between target and participant (D) were the sides (Fig. 1B). To control shoulder height across participants, we suppressed the D- and L-constituted angle, by horizontally aligning the shoulder and the target (Fig. 1C). Thus, participants had their reaching arm parallel to the floor and L and D became equal, so that the critical (i.e., maximal) distance corresponded to arm length (Fig. 1D). Following Warren's (1984) study, the critical point (i.e., the reachability-boundary beyond which participants could not reach the target) was defined as the ratio between the critical distance (i.e., the extrinsic measure corresponding to the maximal distance to reach the target in cm) and arm length. In other words,

Fig. 1
figure 1

Affordance measure. Panel A. Reaching movement. Panel B. During a reaching movement, the arm length (L), the distance between the observer and the target (D), and shoulder height (H) shape a triangle. Panel C. To control shoulder height, we removed the D- and L-constituted angle, by horizontally aligning shoulder and target. Panel D. Once the angle has been removed, D and L are equal. Therefore, reachability depended upon arm length. Note. L: Arm Length; H: Shoulder height; D: Participant-target distance

Full size image
$${\pi}_c=\frac{d_c}{L}$$
(1)

where πc is the critical point, dc is the critical distance, Lis the arm length.

Materials

Apparatus

Participants sat in an adjustable chair facing a 90-cm high table. The shoulder blades remained in contact with the backrest. A button response pad (4 cm height at the level of the first button) was used as the target to be reached. The target was moved along two graduating strips that were placed at the participant’s shoulder width. We used a 94-cm high stick to adjust shoulder height so that the arm was parallel to the floor (Fig. 1). We placed the stick in the middle of the shoulder to determine the correct shoulder height. The experimenter reported collected data (e.g., actual and estimated critical distances, arm lengths) on a computer running Inquisit 5 software (Milliseconds, 2018). A picture of the apparatus is available in the Online Supplementary Materials (OSM; https://osf.io/um2kw/).

Affective scales

Depressive symptoms were assessed with a French version of the Beck Depression Inventory-Fast Screen (BDI-FS; Alsaleh & Lebreuilly, 2017; Beck et al., 2000). The BDI-FS is a seven-item instrument. It is faster to administer (Alsaleh & Lebreuilly, 2017; Poole et al., 2009), and has better validity, reliability, fidelity, sensitivity and specificity to assess depression compared to the Beck Depression Inventory-II (Alsaleh & Lebreuilly, 2017). These very good psychometric properties (Alsaleh & Lebreuilly, 2017) have been demonstrated in a clinical pain population (Poole et al., 2009), a geriatric population (Scheinthal et al., 2001), a general population in Germany (Kliem et al., 2014) and a French student population (Alsaleh & Lebreuilly, 2017). The latter population is the sample in the present study. The BDI-FS excludes somatic symptoms to avoid overlap with somatic diseases (Kliem et al., 2014; Poole et al., 2009) and an increase sensitivity in a student population, which have common somatic symptoms (Alsaleh & Lebreuilly, 2017). Thus, the BDI-FS is a well-adapted and relevant tool to detect depressive symptoms in the current study. The total BDI-FS score (ranging between 0 and 21) is an index of individual depressive symptoms. A value of 4 is the cut-off score for depressive symptoms. In the final sample of this study, seven participants out of 18 included in the analyses had depressive symptoms. The remaining participants (i.e., 11 participants) reported minimal levels of depressive symptoms (scores less than 4).

Anxiety was assessed with the French version of the Spielberg’s State-Trait Anxiety Inventory (STAI-Y; Gauthier & Bouchard, 1993; Spielberg, 1983). The STAI-Y is a 20-item instrument commonly used by researchers and practitioners to measure trait and state anxiety (STAI-Y1 and STAI-Y2, respectively). Participants answer on a four-point scale ranging from 1 (not all) to 4 (extremely).

We evaluated mood with the French version of the Positive and Negative Affective Schedule scale (PANAS; Crawford & Henry, 2004; Gaudreau et al., 2006; Watson et al., 1988). PANAS is one of the most widely used scales to measure mood (Bali & Jaggi, 2015). It is a 20-item scale with ten items measuring positive mood and ten items measuring negative mood. Participants answered on a five-point Likert scale ranging from 1 (very slightly or not at all) to 5 (extremely).

Procedure

See Fig. 2 for a general view of steps of the procedure.

Fig. 2
figure 2

Procedure of the current experiment. A. Pre-tasks. B. Verbal perceptual judgment task. The box moved discreetly from trial to trial (towards or away) in 5-cm steps following two paths (left or right). We adjusted the paths to shoulder width such that each path was in the frontline of the corresponding shoulder. For each location, participants judged whether they could reach the target button of the box. C. Actual motor task. Participants made two reaches (one for each arm). D. Post-tasks. Note. Black crosses indicate the target button for each path on the figure. We did not display such crosses during experiment. For an example of two trials with the judgement task with a GIF, please the Online Supplementary Material (https://osf.io/98rxm/)

Full size image

Pre-tasks

The experiment took place in a laboratory room. Participants provided consent and filled out a general information form, without seeing the apparatus at this stage. Participants’ shoulder width was measured to adjust the distance between the two adjustable strips on the table (see LGS document in the OSM). Participants sat on the adjustable chair and we adapted shoulder height with the stick so that the arm was parallel to the floor. Next, the seated participants faced the apparatus in the middle of the table, until their chest touched the table. Participants were told to sit comfortably on the chair and move as little as possible. They were asked to keep their back and shoulder blades in contact with the backrest.

Verbal perceptual judgement task

In the perceptual judgement task, participants verbally estimated their perceived reachability boundary. The target was either moved towards or away from the participant to the left path or the right path in increments of 5 cm for each trial. When the target was moved towards the participant, the target movement followed one of the two sided-paths (i.e., left or right) and moved toward the edge of the table that was in touch with participant. When moving the target away from the participant, the target movement started from the edge of the table that was in touch with them and moved following one of the two-sided paths (i.e., left or right) towards the opposite edge of the table. Participants had to tell to the experimenter whether or not the target could be reached with their hands (without moving). For ‘left path trials’, the target was the first left button of the box. For ‘right path trials’, the target was the first right button of the box. When making their predictions participants had to take into account that both their shoulder blades and back were supposed to remain in contact with the backrest. All participants completed 12 trials. Among these 12 trials, half (i.e., six) followed the ‘left path’, and the other half (i.e., 6) followed the ‘right path’. Within each path condition, half (i.e., three) presented distance increments from the nearest position (‘away’) and the other half presented distance decrements from the farthest position (i.e., ‘towards’). The trial order was randomized across participants with an Inquisit 5-based randomization process and was different for each participant.

At the beginning of each trial, the experimenter referred to the computer, which randomly determined the presentation order of the trials. He moved the target in increments of 5 cm, either toward or away the participant until they said “yes, I can reach it” or “no, I cannot longer reach it”, respectively. Participants could adjust their estimations by movements in increments/decrements of 1 cm to estimate the perceived reachability boundary as accurately as possible. The experimenter reported the critical distance on Inquisit 5 and then launched the next trial.

Actual motor task

After completing the perceptual estimations, participants were asked to effectively reach the target. They made two reaches: one reach for each arm (left and right arms for the left and right paths, respectively). We determined the actual motor reachability boundary. Their shoulder blades and their back remained in contact with the backrest of the chair. Participants did not receive verbal feedback about their perceptual or motor performances from the experimenter. They only could collect exteroceptive and proprioceptive feedbacks about their movement in the motor performance condition. There was no reaching movement in the perceptual condition.

Post-tasks measurement and debriefing

Bodily measures

Estimated and actual outcomes only provided us with table-target distances. Thus, the experimenter measured the table-shoulder distance on each side. To obtain the critical distance, those distances were added to the estimated and actual data for each side, respectively. The experimenter measured arm lengths, and asked the participant to provide information about their dominant hand.

Affective assessment

Finally, STAI-Y, BDI-FS and PANAS affective questionnaires were administered.

Debriefing

At the end of the experiment, participants were debriefed by providing them with information about the purpose of the study. The experimenter was also available to answer any questions. Participants with depressive symptoms (i.e., score ≥ 4 on the BDI-FS; Alsaleh & Lebreuilly, 2017; Beck et al., 2000) were informed about the possibility of receiving support from care units.

Data analysis

Data preparation

We collected arm lengths, results of the judgement task (i.e., 12 verbal estimations) and two motor assessments (one for each arm) for each participant. We computed the means of each condition of estimation (i.e., left path-towards, left path-away, right path-towards, right path-away). The mean critical points (πc) were calculated for each estimated and actual result by dividing the critical distances by the associated arm length for each participant. The mean actual πc and the estimated πc were computed for each participant and for each estimated condition.

Participants were allocated to either the group of participants with depressive symptoms or the group of healthy participants according to their BDI-FS scores, based on a cut-off of 4 (Alsaleh & Lebreuilly, 2017; Beck et al., 2000). Participants with a score ≥ 4 were allocated to the depressive symptoms group, while those with a score < 4 were allocated to the healthy group. We conducted the following analyses with seven participants in the depressive symptoms group and 11 participants in the healthy group.

Statistical analyses were performed with JASP (JASP Team, 2020). We also used R (R Studio Team, 2019) to generate linear regression plots.

Bayesian and frequentist statistics

When hypotheses were based on a null effect, we used Bayesian statistics and reported the Bayes factor (BF). BF01 is the ratio between the likelihood of the data given the null hypothesis (H0) and the likelihood of the data given the alternative hypothesis H1 (Jarosz & Wiley, 2014). A BF01 > 1 indicates H0 is more likely than H1, while a BF01 < 1 indicates H1 is more likely than H0. As BF01 increases, there is more evidence in favour of H0 and less in favour of H1 (Jarosz & Wiley, 2014). In the present case, Bayesian statistics seemed to be more appropriate than frequentist analyses. The former quantifies evidence in favour of H0 (Rouder et al., 2016), while the latter does not (Amrhein et al., 2019; Fisher, 1935; Wagenmakers et al., 2017). We also performed other Bayesian analyses. Because these analyses were tedious and lengthy, they are presented in the OSM (https://osf.io/9gf3w/). The main findings of this study can be understood without being given all the details of these analyses.

Reachability boundary perception and depressive symptoms

A mixed ANOVA was determined for the average critical point using the assessment mode (actual, estimated) as a within-participant factor and the group (participants with depressive symptoms, healthy participants) as a between-participant factor. Pairwise comparisons were computed with Holm’s correction (Holm, 1979). We reported a 95% confidence interval for the mean difference, corrected using Bonferroni’s method.

Body-scaled perception: Linear regressions

If reachability perception is body-scaled, then the estimated πc should not depend upon the participants’ arm lengths. We evaluated whether arm length predicted intrinsic outcomes (i.e., actual and estimated critical points) and extrinsic outcomes (i.e., actual and estimated critical distances) by linear regression. The assumption was that arm length would predict critical distances but would not predict critical points. We also reported BF01 to test whether critical points were predicted by arm length. This latter analysis corresponded to a Bayesian linear regression using the JZS method with a default Cauchy prior with r = .354 (e.g., Rouder & Morey, 2012).

Additional analyses

Overlaps between affective factors

Mood (Riener et al., 2011; Vegas & Laurent, 2022) and anxiety (e.g., Graydon et al., 2012), factors that influence perception, are often associated with depressive disorders. Therefore, we performed hierarchical regressions to assess the possible overlap between these affective factors and to determine whether a depression effect could result from a covariate effect. Each model included a ‘depressive symptoms’ (as assessed by the BDI-FS) predictor in addition to one of these predictors (level 2): anxiety-trait, anxiety-state, positive mood or negative mood (respectively assessed with STAI-Y1, STAI-Y2, PANAS-PA or PANAS-NA). These models were compared to the null model including only the ‘depressive symptoms’ predictor (level 1). The supplementary variance explained in addition to that already explained by the ‘depressive symptoms’ predictor was computed through R2 change. We also reported BF01 to quantify evidence for null models. A BF01 > 1 indicated that the null model including the ‘depressive symptoms’ predictor alone was more likely than alternative models including either anxiety-trait or anxiety-state or positive mood or negative mood in addition to the ‘depressive symptoms’ predictor. We also performed hierarchical regressions to compare null models including either anxiety-trait or anxiety-state or positive mood or negative mood (level 1) to the model including ‘depressive symptoms’ predictor (level 2) in addition to another affective predictor. Supplementary variance explained in addition to anxiety or mood was computed through R2 change. We also reported BF01 to quantify evidence for the null model. A BF01 > 1 indicated that the null model including only one other affective predictor was more likely than the alternative model including the ‘depressive symptoms’ predictor in addition to another affective predictor. We applied the JZS method with a default Cauchy prior with r = .354 (e.g., Rouder & Morey, 2012).

Depressive symptoms and the gap between estimated πc and actual πc

We also performed exploratory linear regression between the BDI-FS score and ∆πc (i.e., the difference between estimated πc and actual πc) in order to test whether the level of depressive symptoms could predict the gap between estimated πc and actual πc.

Results

Reachability boundary perception and relationship with depressive symptoms

Data are summarized in Fig. 3 (see Table S1 in the OSF for descriptive statistics).

Fig. 3
figure 3

Critical point (πc as a function of assessment mode [actual, estimation] and group [healthy participants, participants with depressive symptoms]). Error bars represent ± 1 standard error of the mean

Full size image

ANOVA revealed a main effect of assessment mode F(1, 16) = 20.49, p = .0003, ω2 = .23 (i.e., the estimated πc was higher than the actual πc), a main effect of group, F(1, 16) = 12.43, p = .003, ω2 = .25 (i.e., the πc in healthy participants was higher than the πc in participants with depressive symptoms). Main effects should be interpreted with caution (Howell, 2010) due to the significant assessment mode × group interaction, F(1, 16) = 8.43, p = .01, ω2 = .10. Holm-corrected post hoc tests were used to follow up the group × assessment mode interaction. This showed that there was a significant difference in the estimated πc between participants with depressive symptoms and healthy participants (pholm = .0005,95 % CI = [−.12, −.03]), but not in the actual πc (pholm = .42,95 % CI = [−.07, .02]). For healthy participants, there was a significant difference between actual πc and estimated πc (pholm = .0001,95 % CI = [−.1, −.03]). For participants with depressive symptoms, the comparison between actual πc and estimated πc did not reach the significance threshold (pholm = .629,95 % CI = [−.06, .03]).

Additional analyses

Overlap

Hierarchical regression analyses (Table 2) showed that depressive symptoms (BDI-FS) predicted estimated πc variation (level 1) (R2 = .39, p = .005. Adding anxiety traits or anxiety state or positive mood or negative mood (STAI-Y1, STAI-Y2, PANAS-PA or PANAS-NA, respectively) to the second level did not lead to a better prediction of estimated πc variation (all ps > .30). Similar results were obtained with Bayesian analysis. It showed ‘anecdotal levels’ of evidence supporting the null models including depressive symptoms (level 1) (all 1 < BF01 < 3).

Table 2 Results from hierarchical regressions assessing overlaps between affective factors
Full size table

Conversely, when anxiety traits or anxiety states or positive mood or negative mood were included in the null model without depressive symptoms (level 1), only negative mood significantly predicted estimated πc variation (Table 2). Adding depressive symptoms to the second level led to a better prediction of estimated πc variation when the predictor included in the null model (level 1) was anxiety traits or anxiety states or positive mood (all ps < .02). Similar results were obtained with Bayesian analysis. It showed ‘anecdotal levels’ of evidence in favour of alternative models including depressive symptoms (level 2) (all BF01 < 1). Adding depressive symptoms to the second level did not lead to a better prediction of πc when the predictor included in the null model (level 1) was negative mood. The latter point should be moderated by the results of Bayesian analysis indicating that the model including depressive symptoms (level 2) was more likely than the null models including only the negative mood predictor (BF01 = .853) to predict estimated πc.

Depressive symptoms and the gap between estimated πc and actual π

There was a significant correlation between BDI-FS scores and the ∆πc (b =  − .006, t(16) =  − 3.119, p = .007 (Fig. 4). No outliers were detected that could deform the regression line (Cook’s distance .5, for each observation). R2 = .378 (adjustedR2 = .339), showing that BDI-FS scores explained 37.8% of ∆πc variation.

Fig. 4
figure 4

Linear regression between the Beck Depression Inventory-Fast Screen-France (BDI-FS-Fr) score and the difference between the estimated πc and the actual πc (∆πc)

Full size image

Body-scaled perception: Linear regressions

We computed the linear regressions for motor πc and dc in the entire sample, and for estimated πc and dc in the healthy group and the depressive symptoms group separately.Footnote 2

Linear regression showed that arm length predicted actual dc (Fig. 5A). However, when actual dc was body-scaled, it was not predicted by arm length (Fig. 5B).Footnote 3

Fig. 5
figure 5figure 5

Linear regressions to assess the body-scaling of reachability. Triangle dots represent extrinsic measures and circle dots represent intrinsic measures. White dots represent the entire sample, black dots represents Participants’ with depressive symptoms group and grey dots represent healthy participants’ group. A. Entire Sample. Linear regression between actual critical distance (cm) and arm length (cm). Analysis showed a significant link between actual critical distance and arm length (𝑏 = 1.032, 𝑡(16) = 8.307, 𝑝 = .0000003). No outliers were detected (Cook'sdistance < .3 for each observation). 𝑅2 = .812;𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑅2 = .812, meaning that arm length explained 81.2% of actual critical distance variations. B. Entire Sample. Linear regression between the actual critical point (πc) and arm length (cm). Analysis did not show a significant link between the actual πc and arm length (𝑏 = .0002, 𝑡(16) = .092, 𝑝 = .928). No outliers were detected (Cook's distance < .4 for each observation). 𝑅2 = .001, (𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑅2 = ―.062). 𝐵𝐹01 = 2.425. C. Depressive symptoms. Linear regression between the estimated critical distance (cm) and arm length (cm). Analysis showed a significant link between the estimated critical distance and arm length (𝑏 = 1.343, 𝑡(5) = 8.643, 𝑝 = .0003). No outliers were detected (Cook's distance < .5 for each observation). 𝑅2 = .937 (𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑅2 = .925), meaning that arm length explained 93.7% of estimated critical distance variations. D. Depressive symptoms. Linear regression between the estimated critical point (πc) and arm length (cm). Analysis did not show a significant link between the estimated πc and arm length (𝑏 = .004, 𝑡(5) = 1.957, 𝑝 = .108). No outliers were detected (Cook's distance < .6 for each observation). 𝑅2 = .434 (𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑅2 = .321) . However, Bayesian linear regression showed evidence for H1 (𝐵𝐹01 = .782). E. Healthy. Linear regression between the estimated critical distance (cm) and arm length (cm). Analysis showed a significant link between the estimated critical distance and arm length (𝑏 = .864, 𝑡(9) = 4.352, 𝑝 = .002). No outliers were detected (Cook's distance < .3 for each observation). 𝑅2 = .678 (𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑅2 = .642), meaning that arm length explained 67.8% of estimated critical distance variations. F. Healthy. Linear regression between the estimated critical point (πc) and arm length (cm). Analysis did not show a significant link between the estimated πc and arm length (𝑏 = ―.003, 𝑡(9) = ―1.114, 𝑝 = .294). No outliers were detected (Cook's distance <.3 for each observation). 𝑅2 = .121 (𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑅2 = .024). 𝐵𝐹01 = 1.433

Full size image

For healthy participants and participants with depressive symptoms, linear regressions showed that arm length predicted estimated dc (Figure 5C, E). However, when estimated dc was body scaled, frequentist analyses did not show any predictive relationship between arm length and estimated πc whatever the group. Bayesian analyses confirmed that arm length did not predict estimated πc in healthy participants, while in participants with depressive symptoms, the model including arm length was found to be more likely than the null model (Fig. 5D and F).

Discussion

Relationship between perceived reachability boundary and depressive symptoms

The goal of the present study was to assess the relationship between depressive symptoms and reachability perception. The key prediction was that perceived reachability boundary (estimated πc) is different between participants with depressive symptoms and healthy participants. Results confirmed this hypothesis 1. Participants with depressive symptoms made more conservative estimations of reachability than healthy participants. Furthermore, an increased level of depressive symptoms was associated with a decrease in perceived reachability overestimation.

However, we do not conclude that participants with depressive symptoms were “more accurate” than healthy participants. As mentioned earlier, overestimation bias could result from experimental constraints. The whole body hypothesis suggests that the engagement of the whole body is anchored in estimations (Fischer, 2000; Rochat & Wraga, 1997). Thus, it might be that participants do not really overestimate reachability (Graydon et al., 2012), but rather calibrate their judgement of what is reachable with more degrees of freedom than prescribed in the experimental instructions and permitted by the experimental setup (Graydon et al., 2012). In the current study, participants were not allowed to lean forward during reaching trials, which they spontaneously do when trying to reach an object (Rochat & Wraga, 1997). Participants would have taken into account these multiple degrees-of-freedom in their estimations (Rochat & Wraga, 1997). During actual trials, if participants had been allowed to lean forward, then the actual πc could have increased, reducing overestimation. In their study, Graydon et al. (2012) provided a similar interpretation of reachability overestimation. Carello et al. (1989) showed that participants were more accurate in estimating reachability boundary in the multiple-degrees-of-freedom condition (e.g., leaning forward during reaching trials) than in the single-degree-of-freedom condition. Furthermore, the involvement of analytical reflective judgement in the task results in overestimation (Heft, 1993). Relatedly, Weast and Proffitt (2018) found that action-based (i.e., perceptual-motor) judgements were more accurate than verbal (i.e., analytical reflective) judgements. If there were a global overestimation effect due to experimental constraints, depressive symptoms would be associated with an underestimation of action possibilities. This is supported by other findings in the present experiment, which showed that when the level of depressive symptoms increased, the overestimation decreased.

Taken together, these results suggest that depressive symptoms are associated with a modification in affordance perception with more conservative estimations (Fig. 6). More conservative estimations of participants with depressive symptoms are consistent with a decrease in affordance salience in depression (de Haan et al., 2013; Kiverstein et al., 2020; Maiese, 2018). For similar actual πc in both groups, a given target could be perceived as affording a reach in healthy participants, but not in participants with depressive symptoms. Participants with depressive symptoms may make conservative estimations of reachability because the farthest targets do not invite them to act (i.e., affordances were less salient). Relationships between depressive symptoms and action possibilities perception could be also explained by learned helplessness (Pryce et al., 2011; Seligman & Maier, 1967) and action inhibition (Laborit, 1979; Li et al., 2015) models. Affordance has been defined by the fit between organism capacities and environmental features (Fajen & Turvey, 2003; Gibson, 1979). One of the aetiologies of depression is the incapacity to act (e.g., Pryce et al., 2011). Learned helplessness has been associated with negative predictions concerning one’s own ability to have control over environment (Pryce et al., 2011). Depressed people anchor and cannot untie the experience of body failure (Fuchs & Schlimme, 2009). They might recalibrate their perception to suit the modifications related to their body and action capacities. Such a recalibration could change the perceived action boundary. A general belief of incapacity is consistently found in depressed individuals (Hiroto & Seligman, 1975; Ratcliffe, 2010; Schneider, 2006). This belief could be anchored in recurrent perception of action (im)possibilities.

Fig. 6
figure 6

Summary of findings. Depressive symptoms are associated with a modification of reaching affordance perception. This modification is characterized by more conservative estimations of reachability and a disturbance of reaching body-scaling. Note. Dotted lines represent hypothetical links. Solid lines represent links supported by the data

Full size image

Overlap between affective factors

Negative mood seems to be involved in the modification of affordance perception associated with depressive symptoms. We found an overlap between negative mood and depressive symptoms in the prediction of reachability perception.

This overlap was consistent with the widely shared standpoint according which mood dysregulation is a central feature of depression symptomatology (APA, 2013). Functionally, negative mood is a basic dimension of depression that could partially predict motor capacity perception. This result is also consistent with others published by Vegas and Laurent (2022). The authors showed that experimentally induced happy and sad moods led to underestimating sitting abilities in healthy individuals. The ‘mood component’ of depression could similarly play an important role in the modification of affordance perception in individuals with depressive symptoms

Body-scaling hypothesis

The current study also revealed body-scaling of reachability at the actual (motor) level in the entire sample. Reachability was also body-scaled at the perceptual level in healthy participants. Such findings support the notion that the distance at which a target is reachable is perceived according to one’s own scale (Carello et al., 1989).

Bayesian analysis was consistent with hypothesis 3, because it revealed that there was no body-scaling of reaching at the perceptual level in participants with depressive symptoms only. This result is consistent with the view that embodiment disturbance can be found in depression (Fuchs, 2005; Fuchs & Schlimme, 2009; Zatti & Zarbo, 2015). The body should be integrated in the phenomenal space to allow individuals to perceive what they are capable of doing in the environment (Fuchs, 2005; Fuchs & Schlimme, 2009). In depression, the body could become an obstacle and be experienced as conspicuous, heavy and solid (Fuchs, 2005; Fuchs & Schlimme, 2009; Zatti & Zarbo, 2015). This disturbance of body-scaling observed in the current study could also explain the modification of perceived reachability boundary (Fig. 6). However, a Bayesian approach leads us to avoid making strong general conclusions regarding that matter given the small level of evidence provided (BF01 < 3).

Limitations and implications

The limitations of this research are twofold. First, the study sample was exclusively composed of student participants, which limits the generalizability of our conclusions. Second, the relationships between depression and perception have only been approached through a psychometric evaluation of depressive symptoms and not following a clinical characterization of depression.

The current study strengths are also twofold. First, it was the first experiment that empirically assessed the relationships between depressive symptoms and affordance. Experimental evidence was reported, which supports previous conceptual or theoretical approaches (De Haan et al., 2013) and paves the way to future rapprochements between psychiatry or psychopathology and embodied approaches to perception. Second, we introduced Bayesian statistics to the affordance literature. These statistics allow us to more stringently test the null hypothesis. Therefore, they are well designed to assess body-scaling.

Concluding remarks

Our study showed that depressive symptoms are associated with a modification of reaching affordance perception. Participants with depressive symptoms had more conservative estimations than their healthy counterparts. At the perceptual level, body-scaling of environmental properties could be impaired in participants with depressive symptoms. This impairment could constitute an embodied component of the inhibitory regulation usually reported in depression. Negative mood likely plays a central role in this modification of affordance perception. The present research could contribute to bridging the gap between theoretical approaches to depression as a ‘pathology of action’ and research on affordances and embodied perception.