The Effects of Posture on Body Part Width Representations

The Effects of Posture on Body Part Width Representations

Lettie Wareing

2023

Despite the ubiquity of our bodily experiences, our representations of our body’s size are not geometrically accurate. For example, when estimating the length of body parts using the hand as a metric, consistent patterns of distortions across body parts are observed. Given the presence of these distortions, some have proposed that representations of length and width emerge directly, or indirectly, from the organisation of somatotopic maps in somatosensory cortex, rather than from their actual relative dimensions. However, whilst length representations are well researched with respect to this notion, less is known about representations of body part width across the body. Moreover, it is unclear from previous research whether body part width representations may be confounded by participants’ posture. Specifically, individuals have shown an enhanced tendency to overestimate body part width when seated upon a chair, suggesting that the chair may become incorporated into the body representation. Consequently, the aim of the current investigation was to further elucidate how width is represented across body parts and whether posture moderates these representations. Participants estimated how many hands widths made up the width of the back, shoulders, hips, torso, thigh, and head in one of three conditions: standing (n = 37), seated upon a chair (n = 33), or seated upon a backless stool (n = 39). Whilst estimates did differ across body parts, no effect of posture was observed. Moreover, the patterns of distortions observed differed from those seen in previous investigations. Results therefore indicate that body part width representations are neither accurate nor fixed, rather, they show distortions which vary across individuals and contexts. It is proposed that inter-individual heterogeneity in width representations may result from humans possessing alternative perceptual mechanisms for judging aperture passability. Therefore, maintaining fixed width representations is unnecessary, and hence too energetically costly to maintain.

Body perception, affordances, somatosensation, visual perception

Method
Participants
Ethical approval for this study was obtained from Lancaster University Psychology Department on 31st May 2023.
As this study aimed to investigate body part width representations in healthy populations, only participants aged 18-55 years without any physical, or mental impairment were included in the study. However, as previous research (Readman et al., 2021) using the same paradigm for length estimates has shown no influence of anxiety or depression on body part estimates, participants with diagnoses of these conditions were not excluded. Participants were excluded if they had any current or historic diagnosis of cognitive impairment, as this can affect instruction comprehension (Han et al., 2011), or visual impairment, to ensure difficulties in seeing the body parts did not confound findings. Furthermore, given the associations between other psychiatric impairments (e.g., Priebe & Röhricht, 2001), neurological impairments (e.g., Blanke et al., 2004), or eating disorders (Mölbert et al., 2017) with distorted body perceptions, individuals with a current or historic diagnosis of a condition falling within any of these categories were excluded.
A total of 123 (61 females) participants ranging from 18 to 68 years (M = 28.80 years, SD = 10.79) were recruited via opportunity sampling for this study. Participant recruitment was ended before the required N = 150 due to time constraints. All participants were entered into a draw to win one of two £25 vouchers as an expression of goodwill. A total of 15 participants were excluded for failing to meet the inclusion criteria, leaving a final sample of N = 108 (50 females). Participants were aged 18 to 55 years (M = 27.98 years, SD = 9.56); the majority of participants were right-handed (n = 99) and over half the participants had normal vision (52.78%), with the remaining participants having corrected-to-normal vision.
Reasons for exclusion included a current or historic psychiatric impairment (n = 2) or eating disorder (n = 4), falling outside the study age restrictions (n = 3), visual impairment (n = 2), being pregnant (n = 1), failing to provide demographic information needed to determine eligibility (n = 2), and a self-reported misunderstanding of task instructions (n = 1).
Design
This study constituted a 3x6 mixed design with condition (standing, chair, or stool) as the between-subjects variable and body part (torso, hips, shoulders, back, thigh, or head) as the within-subjects variable. The dependent variable was participants’ accuracy ratios for each body part (actual size/ estimated size) where an accuracy ratio of over 1.0 indicated overestimation, and under 1.0 indicated underestimation of body part width.
Materials and Procedure
After providing their consent, participants completed a self-report demographic and clinical questionnaire administered via Qualtrics (Qualtrics, Provo, UT) which asked about participants’ age, biological sex, preferred hand, and details regarding their neurological, cognitive, and psychiatric history.
Following this, participants were randomised to one of the three conditions (Standing, Chair, or Stool). In each condition, participants were asked to estimate how many hand widths of their dominant hand made up the width of six different body parts: the torso, shoulders, hips, back, head, and thigh. Participants were instructed to be as accurate as possible, using fractions where necessary. They were asked to refrain from touching the body part with their hand, or basing estimates off estimates for previous body parts if the two body parts were proportionally related. The researcher defined each body part verbally and pointed to their endpoints on their own body prior to the participant making their estimate.
Participants in the standing condition performed all estimates whilst stood upright, without leaning on any surfaces. In the chair condition, participants were seated upon a standard desk chair with a high back and no arm rests. In the stool condition, participants were seated upon a fixed height bar stool with no back. The condition completed by participants was counterbalanced, and the order of body parts estimated was randomised.
After making their estimates, the researcher used a soft tape measure to measure the actual width of the cued body parts before debriefing participants. The study took around 10 minutes to complete.
Analysis
Prior to conducting the analysis, outliers were removed using the median absolute deviation (MAD) approach. This procedure involves removing participants whose accuracy ratios deviated more than three absolute deviations from the median for a given body part. The MAD approach was chosen as it is more robust than traditional methods of outlier detection based upon standard deviations from the mean (Jones, 2019; Leys et al., 2013).
To calculate the dependent variable of accuracy ratios, first, participants’ hand estimates for each body part were converted to centimetres by multiplying their estimate in hands by their measured hand width. After this, estimates for each body part were divided by the actual width of the body part to produce an accuracy ratio.
To test the study hypotheses, data was analysed using a 3x6 mixed ANOVA using the rstatix package available from RStudio (Version 4.2.1). Body Part was entered as the within-subjects variable, and Condition as the between-subjects variable. The assumption of normality was checked using the Shapiro-Wilks test, and the sphericity assumption via Mauchly’s test. Partial eta-squared was used as a measure of effect size.
Though frequently used in analysis, frequentist statistics are not without limitations. It is typically assumed that a p-value of <.05 is evidence for the alternative hypothesis, however this value only represents the probability of obtaining results as extreme as those observed, if the null is true (Wagenmakers et al., 2018). Therefore, data which is unusual under the null hypothesis is not automatically any less unusual under the experimental hypothesis (Wagenmakers et al., 2017). Moreover, a non-significant finding in frequentist analyses cannot be taken as evidence in favour of the null hypothesis (Kruschke & Liddell, 2018). In this regard, Bayesian statistics have several advantages over frequentist statistics including the ability to incorporate prior knowledge, quantify the degree of uncertainty surrounding the existence of an effect, and the ability to quantify the strength of evidence in favour of the null, or alternative hypotheses (see Wagenmaker et al., 2018 for a discussion).
Consequently, to provide further support for conclusions drawn using frequentist analyses, a Bayesian Mixed ANOVA was conducted using the anovaBF function from the BayesFactor available in RStudio (Version 4.2.1). Default priors were used given that these reflect average effect sizes observed across all psychological experiments, and hence are likely to be more reliable than priors drawn from a single, potentially methodologically flawed, study (Rouder et al., 2012).
Where a significant main effect of Body Part or Condition was observed, Holm-Bonferroni adjusted frequentist, and Bayes Factor, pairwise t-test comparisons were conducted to determine the pattern of differences underlying these effects.
In addition, to determine whether body part width estimates differed significantly from 1.0 (i.e., an unbiased estimate), Holm-Bonferroni adjusted frequentist, and Bayes Factor, one-sample t-tests were conducted for each body part.
To judge the strength of evidence provided by the Bayes Factor analyses, Kass and Raftery (1993) criteria was used. By this criteria, Anecdotal evidence is regarded as inconclusive. Percentage error (a measure of certainty in the estimate) was reported alongside Bayes Factors, where <20% is regarded as an acceptable level of uncertainty (Van Doorn et al., 2021).

Lancaster University

Data/Excel.csv

Wareing2023

Leanna Keeble

Open

None

English

Data

LA1 4YF

Dr Sally Linkenauger

MSc

Cognitive, perception

123

ANOVA, Bayesian Analysis, T-Test