Materials and methods
Main Experiment
Participants
An a priori power analysis was conducted using G_Power version 3.1.9.7
(Faul et al., 2007) to determine the minimum sample required to test the
study hypothesis. The effect size was set at 0.25 considered to be
medium using Cohen’s criteria (Cohen, 1992). A F-test assessing the
interaction between the condition (n = 2) and the measurements (weights;
n = 6) was applied with a significance criterion of a = 0.05 and power =
0.95. The minimum sample size needed was N = 28 to detect differences in
discrimination sensitivity between conditions and among measurements.
Thirty-five volunteers (male/females: 11/24; mean age ± SE=24.7±0.8
years) participated in the experiment. Written informed consent was
obtained from all participants before data collection. The study was
approved by the ethical committee of the University of Genoa (Comitato
Etico per la Ricerca di Ateneo, n° 2021/42) and was conducted in
accordance with the Declaration of Helsinki.
Experimental procedure
The experiment consisted of a single session where participants
performed a weight discrimination video task. The task was built using
jsPsych 6.3.0 library and performed off-line (de Leeuw, 2015). The task
was preceded by the instructions and a questionnaire collecting personal
data (i.e., gender, age, weight, height) and physical activity level
(i.e., activities performed, year of experience, weekly hours of
training) data. The protocol is represented in Figure 1A.
Video stimuli
The stimuli consisted of videos showing an actor performing a typical
everyday activity; namely, the actor (either a woman or a man, depending
on the gender of the participant, Figure 1B) moving a box (weight of the
empty box 390 g) from the chest to a shelf over the head (Concentric
condition, given that the main muscle involved in the movement, i.e.,
the anterior deltoids – based on preliminary EMG acquisitions, shortens
in this kind of movement) or bring it down from the shelf to the chest
(Eccentric condition, given that the deltoid anterior lengthens). In
both conditions, the box was filled with varying amounts of sheets of
paper in such a way as to assume 7 different additional weights (0, 2.5,
5, 7.5, 10, 12.5, 15 kg). Actors were informed about the weight of the
box. Videos, whose durations range from 1.3s to 3.7s, were acquired on
the same day with a video-camera positioned to record the execution of
the lifting movement laterally. The face of the actor was blurred to
cover facial expressions, as these reflected the level of exertion
during a physical task (de Morree & Marcora, 2010).
Task
Participants (hereafter ‘observers’) sat in front of a laptop with a
16-inch LCD screen positioned on a table, at a distance of approximately
60 cm. They were required to perform a two-interval forced-choice (2IFC)
task (Duarte et al., 2018). Each trial consisted of a sequence of two
videos, a reference, and a comparison stimulus/video. After observing
both, the observer had to indicate in which video the box was heavier.
In particular, she/he had to press the left arrow key to answer “the
First” and the right arrow key to answer “the Second”. The
7.5kg-video was shown in every trial as it represented the reference
stimulus. The 0kg-, 2.5kg-, 5kg-, 7.5kg-, 10kg-, 12.5kg-, and 15kg-video
were the comparison stimuli (the 7.5kg-video was used both as a
reference and as comparison stimuli). In each trial, the order of
appearance of the reference and the comparison stimuli was random. Each
comparison video was displayed 12 times in random order for each
condition, for a total number of trials corresponding to 168 (7 box
weights, 12 repetitions, 2 conditions). The total duration of the
experiment was about 30 minutes.
— Figure 1 here —
Data analysis
The discrimination sensitivity (dˈ ) was evaluated at each
comparison stimulus (except 7.5 kg) using signal-detection theory as
described in Norman et al. (Norman et al., 2009). The d’ was calculated
according to one-dimensional classification experiments, following the
procedure adopted by Norman et al. (Norman et al., 2009). A “hit”
occurred when the first weight was heavier and the observers correctly
responded, “first was heavier.” A “false alarm” occurred when the
first weight was lighter, but the observers incorrectly responded,
“first was heavier.” The hit rate was computed by dividing the number
of hits by the total number of trials in which the first weight was
heavier; the false-alarm rate was obtained by dividing the number of
false alarms by the total number of trials in which the first weight was
lighter (Norman et al., 2009). The log-linear adjustment method was used
to adjust for extreme values of hits and false alarms (Macmillan &
Creelman, 2004). The higher the dˈ values, the better the ability
to discriminate between the object’s weight. Then data were classified
as Light (0kg, 2.5kg, 5kg) and Heavy (10kg, 12.5, 15kg), andLHmean dˈ was obtained in the two conditions by averaging
the answers to Light and Heavy weights. Finally, the mean dˈ was calculated considering all the weights.
The ratio of responses in which the comparison stimulus was judged
“Heavier” than the reference stimulus at each box weight was computed
for each observer in the two conditions (Concentric and Eccentric) to
build a Psychometric function. The observers’ psychometric curves were
obtained by finding the best-fitting logistic functions usingpsyphy and quickpsy R package (Linares & López-Moliner,
2016; Yssaad-Fesselier & Knoblauch, 2006). The lower and upper
asymptotes, threshold, and just noticeable difference (JND), were
estimated for each psychometric function (Knoblauch & Maloney, 2012).
Lower asymptote (ALOW) and upper asymptote
(AUP) were computed according to Oh et al. (Oh et al.,
2016). The lower/higher ALOW/AUP, the
better the ability to discriminate low/high weights. The threshold
corresponds to the curve point crosses 0.5 on the y-axis and indicates
the point of subject equality (Kopec & Brody, 2010). JND is considered
the smallest weight that produces changing in perception and is
calculated as the half difference between the weights at which the
psychometric function equals 0.75 and 0.25, respectively (von Sobbe et
al., 2021). A lower JND indicated a better ability to discriminate the
stimuli.
Statistical analysis
Sensitivity (dˈ ) at each comparison stimulus (Norman et al.,
2009), mean sensitivity (mean dˈ , obtained by averagingdˈ at the different comparison stimuli except 7.5-kg) (Maguinness
et al., 2013), ALOW, AUP, threshold, and
JND, were considered as outcome parameters. Shapiro-Wilk test was
applied to evaluate data distribution and Levene’s test was used to
evaluate the equality of variances. LHmean dˈ and mean dˈ were normally distributed, whilst dˈ ,
ALOW, AUP, threshold, JND, and heavier
probability at each comparison stimulus (including 7.5 kg) were not.
Statistical analyses were chosen based on data distribution.
Concerning the sensitivity analysis, Wilcoxon tests were applied to
compare dˈ values at each comparison stimulus between Concentric
and Eccentric conditions. Within each condition, Friedman tests,
followed by post hoc, were used to assess differences among dˈ at
each comparison stimulus (0kg, 2.5kg, 5kg, 10kg, 12.5kg, and 15kg). An
ANOVA was applied on LHmean dˈ with Amount-of-weight (2
levels, Light and Heavy) and Condition (2 levels, Concentric and
Eccentric) as within subject factors. Bonferroni post hoc tests were
applied in case of significant interaction. Then, a t-test was performed
to statistically compare mean dˈ between Concentric and
Eccentric conditions.
Concerning the psychometric function, all parameters were statistically
evaluated using the Wilcoxon test to assess differences between the two
conditions.
Normally distributed data are reported as mean value ± standard error
(SE), while not-normally distributed data are given as median
[interquartile range, IQR]. The significance level was set at 0.05,
except for dˈ , where Bonferroni correction was applied due to
multiple comparisons. Statistical analyses were performed with SPSS
Statistics 26 software.
Control experiment
A Control Experiment was conducted to assess the role of the actor’s
perceived effort in the ability to discriminate the weights of moving
objects.
Participants
Thirteen volunteer observers (females/males: 7/6, mean age ± SD: 29.5 ±
4.9 years), not involved in the Main Experiment, were recruited in the
Control Experiment.
Experimental procedures
Observers were involved in a single session experiment during which they
were required to perform a video task lasting about 30 minutes. Video
stimuli were the same used in the Main Experiment (Figure 1A). In a
single trial, a video was shown and then observers were asked to rate
the actor’s effort using a Visual Analog Scale (VAS). The VAS consisted
of a line approximately 10 cm (378 pixels) long, with “No effort” and
“Maximal Effort” as initial and final anchor points, respectively. The
score was expressed on a scale from 0 to 100 (Delgado et al., 2018).
Each video was displayed 10 times, in random order, for a total number
of 140 trials (7 box weights, 10 repetitions, 2 conditions).
Statistical analysis
The difference between the perceived effort of the actor at each
comparison stimulus and at the reference stimulus (i.e., 7.5kg),
expressed as an absolute value (ΔVAS Effort; e.g., abs[VAS
Effort(7.5kg) – VAS Effort(12.5kg)]), was calculated. Furthermore,
based on the value of the comparison stimulus for which ΔVAS Effort was
calculated, ΔVAS Effort values were classified as Light (0kg, 2.5kg,
5kg) and Heavy (10kg, 12.5, 15kg), and LHmean ΔVAS Effort was
obtained for the two conditions.
Shapiro-Wilk test was performed to evaluate if the parameters were
normally distributed. Levene’s test was used to evaluate the equality of
variances. ΔVAS Effort was analyzed via a 2 x 6 ANOVA with Condition (2
levels: Concentric and Eccentric) and Box-Weight (6 levels: 0kg, 2.5kg,
5kg, 10kg, 12.5kg, and 15 kg) as within subject factors. Also,LHmean ΔVAS Effort was analyzed via 2x2 ANOVA with Condition (2
levels: Concentric and Eccentric) and Amount-of-weight (2 levels, Light
and Heavy) as within subject factors. Bonferroni post hoc tests were
applied in case of significant interaction.