Chen et al.

Cognitive Research: Principles

Cognitive Research: Principles and Implications (2023) 8:55
https://doi.org/10.1186/s41235-023-00510-0 and Implications

ORIGINAL ARTICLE Open Access

Aesthetic preferences for prototypical

movements in human actions
Yi‑Chia Chen1* , Frank Pollick2   and Hongjing Lu1,3   

Abstract
A commonplace sight is seeing other people walk. Our visual system specializes in processing such actions. Nota‑
bly, we are not only quick to recognize actions, but also quick to judge how elegantly (or not) people walk. What
movements appear appealing, and why do we have such aesthetic experiences? Do aesthetic preferences for body
movements arise simply from perceiving others’ positive emotions? To answer these questions, we showed observ‑
ers different point-light walkers who expressed neutral, happy, angry, or sad emotions through their movements
and measured the observers’ impressions of aesthetic appeal, emotion positivity, and naturalness of these move‑
ments. Three experiments were conducted. People showed consensus in aesthetic impressions even after control‑
ling for emotion positivity, finding prototypical walks more aesthetically pleasing than atypical walks. This aesthetic
prototype effect could be accounted for by a computational model in which walking actions are treated as a sin‑
gle category (as opposed to multiple emotion categories). The aesthetic impressions were affected both directly
by the objective prototypicality of the movements, and indirectly through the mediation of perceived naturalness.
These findings extend the boundary of category learning, and hint at possible functions for action aesthetics.
Keywords Category, Prototype, Aesthetics, Biological motion, Emotion

Introduction impressions are not only quick, but also impact impor-
Living in a world full of objects and events, there is one tant aspects of our lives (e.g., dating and hiring decisions,
kind of stimuli that captivates most of us: other people. Marlowe et al., 1996).
Our visual systems specialize in processing sights related How do these aesthetic impressions of other people
to other people, including their eyes (Emery, 2000), faces arise? Extensive research has uncovered various percep-
(Kanwisher & Yovel, 2006), bodies (Peelen & Downing, tual factors that determine facial and body attractiveness,
2007), and even human-designed objects (Lopez-Brau including shape averageness, symmetry, and sexual dys-
et al., 2021). Upon seeing other people, in addition to morphism (Thornhill & Gangestad, 1999; Fan et al., 2004,
recognizing identity, emotion, and gender, a wealth of 2005). However, if the aim is to fully understand our
other subjective impressions also arise naturally. Most aesthetic impressions of other people, many researchers
notably, we are very quick to notice how attractive oth- have noted some missing pieces (Fink et al., 2015; John-
ers appear—a mere glance gives rise to an aesthetic son & Tassinary, 2007; Morrison et al., 2018): Most of
experience (Willis & Todorov, 2006). These aesthetic this past research has used static images or illustrations
as stimuli. However, we do not often see completely static
*Correspondence: people in our lives, especially not so in the evolution-
Yi‑Chia Chen ary past. People move, and these movements often sig-
[email protected] nal critical social information, such as emotional states
1
Department of Psychology, University of California, Los Angeles, USA
2
School of Psychology and Neuroscience, University of Glasgow, (Pollick et al., 2001), goals (Csibra et al., 1999), and social
Glasgow, UK intentions (Barrett et al., 2005; Colombatto et al., 2020).
3
Department of Statistics, University of California, Los Angeles, USA

© The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this
licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/.
Chen et al. Cognitive Research: Principles and Implications (2023) 8:55 Page 2 of 13

In this spirit, we aimed to explore how body movements man-made objects (Landwehr et al., 2011; Whitfield &
give rise to aesthetic experiences. Slatter, 1979), abstract shapes (Solso & Raynis, 1979), and
Understanding how people’s dynamic “looks” give rise dot patterns (Posner & Keele, 1968). However, most of
to aesthetic experiences is important for solving several these past explorations focused on static stimuli (except
real-world problems: First, to curb the (sometimes unde- for a few pioneering studies, Ackermann & Adams, 2004;
sirable) impact of appearances on our personal and pro- Sparrow et al., 2002). Thus, we asked: do dynamic events
fessional lives, it is critical to first understand what kind like movements in walking actions also lead to prototype
of aesthetic experiences can arise upon seeing a person. representations and preferences? With this approach, we
This study aimed to fill the gap between the vast scientific also asked what the potential function of categorical pro-
explorations on static stimuli and dynamic real-world cessing for actions may serve. Do the categorical repre-
experiences. Second, knowledge of action aesthetics will sentations serve to recognize different kinds of actions,
aid the design and development of the virtual world, or could it be further involved in perceiving the social
such as those in animations, avatars in virtual reality, and characteristics these actions may imply? To probe this
robotics. Third, being able to evaluate human movements question, we further assessed whether the potential pro-
plays critical roles in medical diagnosis and rehabilitation totype effects reflect a unified category of human walking
(e.g., Sparrow et al., 2002). We aimed to develop neces- actions, or multiple subcategories conditional on dif-
sary computational models that can provide quantitative ferent emotion states that often associate with different
measures in these clinical needs. underlying social intentions.
Beside the practical needs of understanding aesthetics Third, the choice to investigate a possible aesthetic
in movements, the present study also contributes theo- prototype effect also represents a new approach to
retically to three fields: aesthetic perception, categorical understand biological motion processing. Explanations
processing, and biological motion perception. First, we of aesthetic experiences from actions proposed in past
assessed the extent to which aesthetic experiences exhibit studies have been based almost exclusively on domain-
systematic regularities from body movements. We used a specific processes for human movements. For exam-
type of dynamic stimuli that is frequently experienced— ple, men’s dance movements are linked to attractiveness
human walking. Also, to further understand what kind because dances are used as demonstrations of physical
of perceptual processing gives rise to these aesthetic abilities (Hugill et al., 2009; McCarty et al., 2013), and
experiences, we included walking actions indicative of women’s gait patterns are linked to attractiveness as they
different emotion states. This allowed us to tease apart signal their fertile period around ovulation (Fink et al.,
effects of motion perception from social perception of 2012). Specialized processing of biological motion has
emotion. The choice to use everyday stimuli like human also been proposed to explain a preference for consist-
walking was intentional, as we aimed to research a dif- ency between body shape and movements (Klüver et al.,
ferent aesthetic experience from what previous research 2016). These studies suggested that the specialized per-
has focused on. Specifically, for the aesthetics of move- ceptual processes for biological motions underly specific
ment, there exists an interdisciplinary field of research aesthetic preferences that serve unique functions. Here,
on the aesthetics of dance (for a review, see Christensen we asked a new question: Beyond specialized aesthetic
et al., 2013). This aesthetic research has examined artis- effects, can domain-general aesthetic effects such as the
tic movements explicitly designed to communicate with prototype effect be observed in human actions? Answer-
and elicit various emotional and aesthetic experiences ing this question helps situate the role of general percep-
in an attentive and interested audience (for a discussion, tual processing in seeing biological motion.
see Orlandi et al., 2020). Our goal, however, was to look We addressed these real-world needs and theoretical
at the spontaneous aesthetic experiences that arise from questions by conducting three behavioral experiments
seeing everyday stimuli, which naturally carry biological to measure observers’ aesthetic impressions of walking
and social information without artistic or communicative actions from different actors expressing different emo-
intentions (for a discussion on the distinction of art and tions, and by constructing computational models based
aesthetics research, see Palmer et al., 2013). on the prototypicality of human walking to account for
Second, we used an aesthetic phenomenon—the pref- observed aesthetic experiences. To isolate the effect of
erences for category prototypes—as a lens to study how action dynamics, we used point-light displays from a
people organize representations of actions into different motion capture dataset (Ma et al., 2006) to remove the
categories: Prototype preferences underly all sorts of vis- influence of body shape appearance on the walkers. In
ual categories including human faces (Galton, 1878; Lan- the first experiment, we asked how much consensus and
glois & Roggman, 1990), artificial or realistic biological systematicity there are in aesthetic experiences from
organisms (Halberstadt & Rhodes, 2003; Younger, 1990), seeing other people walk. In the second experiment, we
Chen et al. Cognitive Research: Principles and Implications (2023) 8:55 Page 3 of 13

explored how emotion recognition and human form through the windows and you are relaxed and well rested.
influences these aesthetic experiences. In the third exper- You are waiting for your flat mate since the two of you are
iment, we examined how action prototypicality and aes- going out shortly. While waiting you start pacing, more
thetic impressions are linked causally. from habit than anything else.” All actors were given the
same scripts to read before performing actions. Details
of creating the motion capture dataset and instruc-
Method tions were included in the paper by Ma and colleagues
Participants (2006). For the present study, each walking video was
For each of the three experiments, 50 naive observers created by using the 5-s excerpt after 8.3 s (500 frames
(Experiment 1: 34 females and 16 males; Experiment 2: in 60 Hz sampling rate) in the motion-capture film. We
33 females and 17 males; Experiment 3: 40 females and then down-sampled the video from 60 to 30 Hz, which
10 males; all with normal or corrected-to-normal vision) is a more typical frame rate for videos displayed online.
from the University of California, Los Angeles (UCLA) Fifteen joints were used to create each point-light display
community completed an individual 30-min experimen- (650px × 350px), and each joint was depicted as a white
tal session online in exchange for course credit. A total dot (12px in diameter) on a uniform black background
of 40 additional observers participated (15 in Experi- (Fig. 1a). The videos were then mirrored to create the 80
ment 1; 15 in Experiment 2; 10 in Experiment 3) but were mirrored videos.
removed based on predetermined criteria (see details in
the Observer exclusions section below). The sample size
Spatially scrambled creature videos for Experiment 2
was predetermined arbitrarily, preregistered, and fixed
To isolate factors related to local joint motion from the
throughout all experiments. The study was approved by
holistic processing and explicit recognition of emotion
the UCLA Institutional Review Board.

Stimuli
Because the stimuli were rendered on observers’ own
web browsers, viewing distance, screen size, and display
resolutions could vary depending on computer moni-
tors used by observers; hence, we report visual stimulus
dimensions using pixel (px) values.

Intact walker videos for Experiment 1 and 3

From the motion capture dataset (Ma et al., 2006), we
created 80 point-light displays from 20 unique walkers
(10 female, 10 male) expressing four different emotions
(happy, neutral, angry, sad) while walking back and forth
between left and right. For each emotion, actors read a
script that depicted different emotional scenarios for
them to express through their movements. For example,
the script for performing the happy walking action was
“It’s Friday evening and you feel great, because earlier you
handed in your final year project. Your supervisor was
very pleased, he complimented you on it and hinted that
you’re going to get excellent marks for it. You just talked
you your flatmate who suggested you go out to celebrate
and now you are just waiting for him to finish getting
ready. As you are getting more excited, you start pacing
around the room, this is going to be such a good night
and you can’t believe that you are almost finished with Fig. 1 a The image on the left is a sample frame from one
your degree. You almost want to start skipping with joy!”. of the intact walker videos used in Experiment 1 and 3. The image
The same procedure also applied to the neutral walk- on the right depicts a few frames of movements after removing
ing action, with the script read: “It is a sunny Saturday the global body displacements in the video. b A sample frame
of the same walker from (a) in a scrambled video used in Experiment
morning and you are in your flat, the sun is streaming in
2 and its corresponding movement depiction
Chen et al. Cognitive Research: Principles and Implications (2023) 8:55 Page 4 of 13

expressions, we created spatially scrambled videos so use the word prototypicality directly in the instructions,
that the configuration of body shapes no longer resem- but asked the observers to rate the walking styles’ natu-
bled a human walker. This was achieved by randomizing ralness (“……how natural you find each walking style to
initial positions of each of the 15 joints in each walker be. In other words, how common/usual do you think the
while keeping the trajectory patterns of the joints intact walking style looks”). Again, they used a 6-point scale
(Fig. 1b; Cutting, 1981). We selected random initial x with naturalness labels (certainly natural, probably natu-
and y positions separately. For y positions, we first found ral, guess natural, guess not natural, probably not natural,
the full y movement ranges throughout the 5-s video for and certainly not natural).
each walker and each of their 15 joints. We then ran- In the second part of all three Experiments, the observ-
domly selected a new y position for each joint, with the ers rated the emotion positivity of the walkers (Experi-
constraint that the full movement range of the joint stay ment 1 and 3) or the creatures (Experiment 2) in the
inside the full movement range for the walker. For x posi- videos (“……how positive you find each walker/creature’s
tions, we calculated the joints’ relative x positions from emotion to be. In other words, how positive of a mood do
the center of each walker (mean x position of all joints) you think the walker/creature is in”), using a 6-point scale
in each frame of the 5-s videos. This way we temporar- with emotion positivity labels (certainly positive, prob-
ily removed the global horizontal motion of the walk- ably positive, guess positive, guess not positive, probably
ers for the randomization process. We then performed not positive, and certainly not positive).
the same position selection used for y positions. Finally, The videos displayed intact walkers in Experiment 1
we added the horizontal movements back to all joints in and 3, and spatially scrambled walkers in Experiment 2.
each frame. The videos were then mirrored to create the Each of the 80 videos was displayed once in each block
80 mirrored videos. These spatially scrambled point-light in different random orders. Whether the original or the
displays were referred to as “creatures” in the experiment. mirrored version were shown was randomly decided for
each video and each observer but kept the same across
Experiment procedure the two rating tasks.
Observers were directed to a website where stimulus At the end of the experiment, observers answered a
presentation and data collection were controlled via cus- series of debriefing questions to ensure they had com-
tom software written in HTML, CSS, JavaScript, JQuery, pleted the experiment without any issues.
and PHP. Observers were not allowed to participate
using phones or tablets. After completing a CAPTCHA Observer exclusions
task (using the hCaptcha service: https://​www.​hcapt​ In addition to the 150 observers whose data were ana-
cha.​com/), they were asked to maximize their window lyzed, 40 observers (15 in Experiment 1; 15 in Experi-
size, informed about their task, and quizzed about their ment 2; 10 in Experiment 3) participated and were
understanding of the instructions, and then, they pro- excluded based on criteria decided before data collection
vided their consent. During the instructions, they were began, with some observers triggering more than one
shown 4 videos from different walkers and different criterion. For Experiment 1, five observers reported that
emotional expressions, to help them gauge the range of they did not follow the instructions or did not take the
experiences they would have. They then performed one experiment seriously; one observer failed the instruction
practice trial to get familiar with the rating scale. quiz more than once; one observer spent less than 0.5 s
In the first part of Experiment 1, observers were asked on at least one page of the instructions; one observer
to rate “how visually pleasing you find each walking had a browser viewport smaller than 800px × 600px; one
style to be”, and “In other words, how good/beautiful do observer had at least one trial with the video not fully in
you think the walking style looks/movements look”. The view during the rating task; one observer gave the same
same goes for the first part of the Experiment 2, except rating to more than 15 consecutive trials; six observ-
the phrase “walking style” was replaced with “creature’s ers hid the experiment browser tab more than three
movement”. Observers rated each video on a 6-point lik- times during the trials; and four observers took too long
ert scale with labels (certainly pleasing, probably pleas- to complete the experiment (two SDs longer from the
ing, guess pleasing, guess not pleasing, probably not mean duration from all observers in the same experiment
pleasing, and certainly not pleasing). For Experiment 3, before exclusions).
instead of aesthetic ratings, the observers provided rat- For Experiment 2, seven observers reported that
ings to indicate their subjective impression of each walk- they did not follow the instructions or did not take the
ing style’s prototypicality. To avoid jargon, we did not experiment seriously; one observer spent less than 0.5 s
Chen et al. Cognitive Research: Principles and Implications (2023) 8:55 Page 5 of 13

on at least one page of the instructions; three observ- sequence: [(0,0), (0,0), (1,2), (3,3)] and B sequence: [(0,0),
ers had a browser viewport smaller than 800px × 600px; (2,2), (3,3), (3,3), (3,4)], the algorithm would correspond
one observer had more than four trials with response A1 (i.e., the coordinates in the first frame of A sequence)
times longer than two minutes in at least one block; two and A2 (the second frame of A sequence) both to B1 (the
observers hid the experiment browser tab more than first frame of B sequence), and A3 to B2, and finally A4 to
three times during the trials; one observer provided B3, B4, and B5. With this correspondence mapping over
a non-sensical answer to one of the debriefing ques- time, the algorithm can minimize the total dissimilarity,
tions; and three observers took too long to complete the calculated by the distance between matched coordinates
experiment. (0 + 0 + 1 + 0 + 0 + 1 = 2). This procedure allowed us to
For Experiment 3, one observer reported technical map corresponding frames between two clips in an auto-
issues, three observers reported that they did not under- matic and data-driven manner.
stand the instructions or did not take the experiment
seriously; one observer spent less than 0.5 s on at least Prototypicality scores
one page of the instructions; one observer had at least To model prototypicality, we performed pairwise com-
one trial with the video not fully in view during the rat- parison between all 80 videos. For each action clip in
ing task; three observers hid the experiment browser tab each video, we applied DTW algorithm multiple times
more than three times during the trials; and two observ- to find the best matching sequence in all other 79 videos
ers took too long to complete the experiment. following these steps: (a) We measured the clip length in
frame number and dropped any action clip that had less
Prototypicality model than six frames (200ms) as they contained too little infor-
Trajectory preprocessing mation for meaningful mapping. (b) For each of the rest
Each video was preprocessed separately following these of the action clips, we found the best matching sequences
steps. (a) We first subtracted the head’s x position in each within the full length of each of the other 79 videos. We
frame from all joints’ x positions to remove the global did so by defining the max and min clip length for the
horizontal movement. (b) The mean x and y positions for possible matching sequence as ± 12 frames from its own
each joint across all the frames were calculated and sub- clip length, with the constraint that the length should be
tracted from the joint’s x and y positions in each frame. no longer than 150 frames (the full video) and no shorter
This way, all joints’ trajectories ended up centering at than seven frames. (c) We looped through all allowed
coordinate (0, 0). (c) The video was segmented manu- clip lengths for the possible matching sequences. For
ally into multiple action clips based on the type of move- each clip length, there will be multiple possible matching
ments the walker was performing—walking from left to sequences in each of the other 79 videos. For example,
right, turning around on the right, walking from right to a clip length of 5 frames includes sequences like Frame
left, or turning around on the left. The number of action 1–5, Frame 2–6, Frame 3–7, and so on, from another
clips differed between videos and ranged from 1 to 5 video. We performed DTW between the action clip and
clips. each of these possible matching sequences to find the
one that had the smallest dissimilarity across clip length.
Dynamic time warping Thus, we found the best corresponding sequence and its
Since the same type of movements in different action dissimilarity to the action clip for each of the other 79
clips was performed in each walker’s own walking speed videos. Note that the DTW was performed considering
and rhythm. Comparing different clips required a method all 15 joints’ coordinates at the same time.
to map the corresponding frames in the gait cycle across The total dissimilarity between two videos was then cal-
walkers. For example, the frame where Walker A raised culated by summing over the dissimilarity of each action
their right foot in Clip A should be compared with the clip in one video and their best corresponding sequences
frame where Walker B raised their right foot at a differ- in the other video. This dissimilarity score was then nor-
ent time point in Clip B. We used dynamic time warp- malized to the standard video length of 150 frames (since
ing algorithms (DTW) to measure dissimilarity between a few action clips might be dropped from finding a cor-
walking sequences. DTW was performed by “warping” responding sequence because of its short clip length).
the temporal sequences of coordinates nonlinearly in Note that this process yielded asymmetrical dissimilarity
time to find the optimal (i.e., least dissimilar) correspond- scores when we find the matching sequences from Video
ence between two sequences. We use a simple example A to Video B, and Video B to Video A. We simply took
to illustrate this algorithm: Considering computing dis- the smaller value to represent two videos’ dissimilar-
similarity between two sequences of 2D coordinates—A ity. A video’s prototypicality score was calculated by one
Chen et al. Cognitive Research: Principles and Implications (2023) 8:55 Page 6 of 13

divided by its average dissimilarity score with all other Transparency and openness
videos in a category (i.e., the other 79 videos for the one- In the above sections, we reported all data exclusions, all
category model). Hence, higher prototypicality score of a manipulations, and all measures. Experiment 1’s design
video indicates that the video is more representative of a and analysis plan were preregistered and can be viewed
category as it shows higher average similarity to all exem- here: https://​aspre​dicted.​org/​KJD_​33W. All materials,
plars in this category. code, and data can be downloaded here: https://​doi.​org/​
10.​17605/​OSF.​IO/​87G3E. A demonstration of the experi-
ments can be viewed online here: https://​yi-​chia-​chen.​
Mediation analysis
github.​io/​walker-​proto​type-​demo-​expt/.
To understand the causal path between objective pro-
totypicality and aesthetic impression, we conducted a
mediation analysis to separate direct effect and indi- Results
rect effect through subjective prototypicality. Beside Behavioral findings: emotion expressions influence
the standard Sobel test, we performed the “permutation aesthetic impressions
confidence interval for ab” method (Taylor & MacKin- We first addressed two questions regarding the patterns
non, 2012) to address the caveats of the Sobel test. Using in the behavioral measures: Is there a systematic consen-
both the z-scores of the ratings and the residuals after sus in people’s aesthetic impressions evoked from point-
regressing out emotion positivity ratings, we followed light walkers? How do the emotion expressions relate to
these steps: (a) We first calculated the predicted aes- aesthetic experiences?
thetic rating for each video according to a linear regres-
sion model including both modeled prototypicality and
Systematic consensus in action aesthetics
naturalness rating as predictors. (b) The aesthetic ratings’
To gauge the consensus on aesthetic impressions, we used
residuals from these predicted values were calculated. (c)
data from Experiment 1 and calculated each observer’s
We calculated the predicted naturalness rating for each
“aesthetic taste typicality” (Chen et al., 2022a) by correlating
video according to a linear regression model including
their rating z-scores for each video to the average z-scores
only modeled prototypicality as the predictor. (d) The
of the other 49 observers for each video. This measure
naturalness ratings’ residuals from these predicted values
revealed how similar an observer’s aesthetic taste was
were calculated. (e) We iterated the following steps (f-j)
compared to an average taste from all other observers—
for 10,000 times. (f ) The residuals calculated from step b
hence, how typical one’s aesthetic taste was. All observers
were permutated and led to a permutated set of aesthetic
showed positive taste typicality, except for one observer
ratings. (g) The same permutation was applied to natural-
who showed a weak negative taste typicality (Experiment 1:
ness residuals and led to a permutated set of naturalness
M = 0.536, SD = 0.176, Range = [− 0.099, 0.788]; Experiment
ratings. (h) Using a linear regression model with model
2: M = 0.272, SD = 0.139, Range = [− 0.010, 0.532]). This pat-
prototypicality as the predictor to predict permutated
tern indicates substantial consensus across observers and
naturalness ratings, we found the permutated slope for
systematic variations in aesthetic impressions across differ-
prototypicality (denoted by a*). (i) Using a linear regres-
ent walks.
sion model with both modeled prototypicality and the
original naturalness rating as predictors to predict per-
Emotion positivity correlated with positive aesthetic
mutated aesthetic ratings, we found the permutated slope
impressions
for naturalness (denoted by b*). (j) We then multiplied
For both Experiment 1 (intact walkers) and Experiment 2
a* and b* to get the permutated indirect effect. (k) After
(scrambled creatures), we depict each video’s mean aes-
10,000 iterations, we got a distribution of a* x b* values
thetic and emotion positivity ratings in Fig. 2a, c. There
and found the confidence interval based on the 2­ 50th and
is a clear relationship between aesthetic ratings and emo-
the ­9751st values after sorting.
tion positivity ratings: The more emotionally positive a

(See figure on next page.)

Fig. 2 a Each intact walker video’s mean aesthetic rating from Experiment 1 plotted against its emotion positivity rating. b Mean aesthetic ratings
of videos in four emotion categories from Experiment 1. All error bars are within-subject 95% confidence intervals (computed after subtracting
individual overall means from the individual’s means in four categories). c Each scrambled creature video’s mean aesthetic rating from Experiment 2
plotted against its emotion positivity rating. d Mean aesthetic ratings of scrambled creatures in four emotion categories from Experiment 2. e Each
intact walker video’s mean naturalness rating from Experiment 3 plotted against its modeled prototypicality. f Mean naturalness ratings of videos
in four emotion categories from Experiment 3
Chen et al. Cognitive Research: Principles and Implications (2023) 8:55 Page 7 of 13

a. b.
Experiment 1 Experiment 1
6 6

5 5
Aesthetic Rating

Aesthetic Rating
4 4

3 3

2 2

1 1
1 2 3 4 5 6 Happy Neutral Angry Sad
Emotion Positivity Rating
c. d.
Experiment 2 Experiment 2
6 6

5 5
Aesthetic Rating

Aesthetic Rating

4 4

3 3

2 2

1 1
1 2 3 4 5 6 Happy Neutral Angry Sad
Emotion Positivity Rating

e. f.
Experiment 3 Experiment 3
6 6

5 5
Naturalness Rating

Naturalness Rating

4 4

3 3

2 2

1 1
2 3 4 5 Happy Neutral Angry Sad
Prototypicality (x10-5)
Fig. 2 (See legend on previous page.)
Chen et al. Cognitive Research: Principles and Implications (2023) 8:55 Page 8 of 13

walk appeared, the more aesthetically pleasing it looked. (0.6), Msad = 3.3 (0.5); main effect with z-scores: F(3,
This pattern was confirmed by a significant positive 147) = 18.7, p < 0.001, η2p = 0.276).1 This finding suggested
by-video correlation calculated with group averages of that the pattern of results observed in intact walkers
z-scores (Intact walkers: r(78) = 0.662, p < 0.001; Scram- depends on explicitly recognizing human actions, as a
bled creatures: r(78) = 0.627, p < 0.001; all tests reported different result pattern emerged when the moving entity
were two-tailed tests), and further supported by a one- no longer appeared to be a human being.
sample t-test of comparing correlations between the Why do we aesthetically prefer neutral walkers? If
two ratings at the individual level to zero (Intact walk- categorical representations based on prototypes exist
ers: M = 0.321, SD = 0.186, Range = [− 0.206, 0.713]), for human walkers, it is possible that neutral walkers
t(49) = 12.18, p < 0.001, Cohen’s d = 1.72; with 48/50 appeared the most prototypical, and thus, the preference
observers showing positive correlations, p < 0.001; Scram- is a result of an aesthetic prototype effect. To test this
bled creatures: M = 0.186, SD = 0.174, Range = [− 0.168, idea, we used a computational model to quantify objec-
0.729]), t(49) = 7.59, p < 0.001, Cohen’s d = 1.07; with tive prototypicality in walking stimuli, and in Experiment
42/50 observers showing positive correlations, p < 0.001). 3 we measured observers’ subjective prototypicality.

Emotion categories revealed sophisticated effects Computational modeling: an aesthetic prototype effect
If the effect of emotion positivity was unidimensional in human walks
(differing only with respect to how positive the expres- We first constructed a single category model: Using the
sions appeared), the relationship between aesthetic and dynamic time warping algorithm (DTW) to compute the
emotion positivity ratings described above would predict similarity of joint movements from two actions (Gavrila
that happy walkers/creatures would yield most positive & Davis, 1995), we calculated pairwise similarity across
aesthetic impressions compared to neutral, angry, and walking videos and computed average similarity for each
sad walkers/creatures. However, we found a different walker (to the other 79 walks) as an index of objective
patterns of results: In Experiment 1 with intact walk- prototypicality. Body movements closer to the proto-
ers, when different emotional categories were examined typical walking sequence would show greater similarity
(Fig. 2b), the neutral walkers were rated the most aes- to the other walks, resulting in greater prototypicality
thetically pleasing, even higher than the happy walkers index values. We then examined the relation between the
(Mhappy = 3.5 (0.6), Mneutral = 3.9 (0.5), Mangry = 2.9 (0.6), objective prototypicality index for each of the walks and
Msad = 3.1 (0.6); and showed a significant main effect of the aesthetic ratings from Experiment 1 for intact point-
emotion category on aesthetic ratings (F(3, 147) = 62.0, light walkers: We found that the more prototypical a walk
p < 0.001, η2p = 0.559; all post hoc comparison with neu- was, the more aesthetically pleasing it appeared both
tral walkers: ts > 4.4, ps < 0.001 after Bonferroni correc- before and after regressing out the emotion positivity
tion; see Additional file 1 for details). Moreover, this ratings (Fig. 3a; aesthetic z-scores average: r(78) = 0.566,
finding was not due to misclassification of the walkers’ p < 0.001, aesthetic residuals average: semipartial cor-
emotion expressions, as the happy walkers were still relation r(78) = 0.546, p < 0.001). Critically, the objective
rated the most emotionally positive (Mhappy = 4.0 (0.4), prototypicality index revealed that the neutral walkers
Mneutral = 3.7 (0.4), Mangry = 3.4 (0.5), Msad = 2.6 (0.4); one- indeed were the most prototypical (Fig. 3b), followed by
way ANOVA main effect with z-scores: F(3, 147) = 185,
p < 0.001, η2p = 0.791; all post hoc comparison with happy
walkers: ts > 5.9, ps < 0.001 after Bonferroni correction).
Furthermore, the effect of neutral walkers having higher 1
The results of Experiment 2 here appeared to rely on how fast the points
aesthetic ratings than other emotion categories of walk- moved—the faster the points moved, the more emotionally positive and
ers remained to be significant after regressing out the aesthetically pleasing the walker looked. We supported this observation
emotion positivity ratings from the aesthetic ratings by calculating the sum speed of the joints for each walker following these
steps: (a) The x position of the head in each frame were subtracted from
(one-way ANOVA main effect with aesthetic residuals: the x positions of all 15 joints. (b) Each joint’s travel distance in each frame
F(3, 147) = 41.0, p < 0.001, η2p = 0.455; all post hoc com- was calculated based on the joint’s x and y positions. (c) The sum of dis-
parison with neutral walkers: ts > 5.8, ps < 0.001 after tance traveled by all 15 joints throughout the video divided by the total
duration of the video (5 s) gave us the sum speed of all joints. This speed
Bonferroni correction). In Experiment 2, when body con- measure was positively correlated with the group average of both emo-
figuration was removed by spatial scrambling, this effect tion positivity and aesthetic z-scores (emotion positivity: r(78) = 0.704,
was not observed (Fig. 2d): Observers found happy and p < 0.001; aesthetics: r(78) = 0.346, p = 0.002), as well as individual’s emotion
positivity and aesthetic ratings (emotion positivity: M = 0.112, SD = 0.150,
angry creatures more aesthetically pleasing than neutral Range = [− 0.201, 0.365]), t(49) = 5.77, p < 0.001, Cohen’s d = 0.82; with 40/50
creatures, and sad creatures the least aesthetically pleas- observers showing positive correlations, p < 0.001; aesthetics: M = 0.113,
ing (Mhappy = 3.8 (0.4), Mneutral = 3.6 (0.4), Mangry = 3.8 SD = 0.286, Range = [− 0.528, 0.675]), t(49) = 2.79, p = 0.007, Cohen’s
d = 0.39; with 33/50 observers showing positive correlations, p = 0.015).
Chen et al. Cognitive Research: Principles and Implications (2023) 8:55 Page 9 of 13

happy, sad, and angry walkers (one-way ANOVA main 1898). After regressing out the emotion positivity ratings
effect: F(3, 57) = 6.91, p < 0.001, all post hoc comparison from the aesthetic ratings, the single category model still
with neutral walkers: ts > 2.9, ps < 0.05 after Bonferroni out-performed the emotion category model (semipartial
correction). Thus, the computational model provides an correlation r(78) = 0.440, p < 0.001) numerically (single
account for the aesthetic prototype effect, supporting the category model: semipartial correlation r(78) = 0. 546;
hypothesis that dynamic events involve representations comparison: p = 0.138). Thus, the additional emotion-
of a category. based categories worsened the model predictions of
Walking stimuli in our study might be represented as aesthetic ratings for the walkers, suggesting that the aes-
exemplars of a single category (walking), or as exemplars thetic judgments are better explained by assuming that
of multiple categories depending on the different emo- human walks form a single category.
tions the walkers were expressing (e.g., happy walking,
sad walking). Would a model with more refined emo- Objective and subjective prototypicality
tional categories more accurately predict human aes- After observing the strong correlation between proto-
thetic judgments relative to the parsimonious model with typicality and aesthetic impressions, we next ask: what
a single category of walking? To address this question, is the causal relationship between these variables? In
we tested the emotion category model: For each walk, general, correlational data afford multiple causal inter-
instead of computing the overall average similarity to pretations. However, in this case an additional con-
the other 79 walks, we computed average similarity only straint is apparent: model-derived prototypicality is
within the same expression (to the other 19 walks).2 We an objective statistical measure that solely depends on
then examined the relation between the emotion cate- the distribution of stimulus exemplars, whereas aes-
gory objective prototypicality index for each of the walks thetic impressions were measured by subjective human
and the aesthetic ratings from Experiment 1 in the same judgments. It seems logically impossible for a subjec-
way as for the single category model: we again found that tive measure to have a causal impact on an objective
the more prototypical a walk was, the more aestheti- statistical measure on stimuli. A remaining question,
cally pleasing it appeared (r(78) = 0.382, p < 0.001). How- however, is whether the causal path from objective pro-
ever, this correlation from the emotion category model totypicality to aesthetic impression operates by a direct
was weaker than that from the single category model causal path, indirect path via some other variables (e.g.,
(emotion category model: r(78) = 0.382, single category perceive naturalness of actions), or both.
model: r(78) = 0.566; comparison: p = 0.012; all compari- While it is often possible in general that a third fac-
sons between correlations were conducted with cocor, tor could cause both objective prototypicality and aes-
Diedenhofen & Musch, 2015, based on Pearson & Filon, thetic experiences, it is paradoxical in this specific case:
the third factor would need to influence the objective
prototypicality, i.e., causing the estimate to be higher or
lower, yet not itself be part of the objective prototypical-
2 ity, i.e., not changing the estimate. For example, perhaps
We focused on the emotion categories rather than the gender categories
because a separate experiment with a new group of 50 subjects showed that actors find it easier to perform neutral walks compared
the emotion categories were more discriminable than the gender categories to happy, angry, and sad emotional walks, which led to
(for emotion discrimination between four emotions: chance = 25%, Mac- differences in naturalness in the movements. However,
curacy = 48.4%, ­SDaccuracy = 7.7%, t(49) = 21.42, p < 0.001, Cohen’s d = 3.03; for
gender discrimination between two gender: chance = 50%, Maccuracy = 59.9%, this could not be considered as a third factor that could
­SDaccuracy = 6.3%, t(49) = 11.03, p < 0.001, Cohen’s d = 1.56). However, we cause both the differences in objective prototypicality
still constructed two additional models considering the gender categories and aesthetic experiences, as the differences in natural-
for additional information: the gender category model and the emotion-
plus-gender category model. For the gender category model, we computed ness manifest in the movements themselves and thus,
the average similarity only within the same gender (to the other 39 walks). would be part of the measure of objective prototypicality,
For the emotion-plus-gender category model, we computed the aver- rather than acting as a third factor that could influence
age similarity only within the same expression and gender (to the other 9
walks). We again examined the relation between the model prototypicali- the objective prototypicality.
ties and the aesthetic ratings from Experiment 1. In both models, we again We first examined the relationship between objective
found that the more prototypical a walk was, the more aesthetically pleas- prototypicality derived from the single category model
ing it appeared (for the gender category model, aesthetic z-scores average:
r(78) = 0.564, p < 0.001, aesthetic residuals average: r(78) = 0.551, p < 0.001; and subjective prototypicality (i.e., naturalness ratings)
for the emotion-plus-gender category model, aesthetic z-scores average: provided by observers from Experiment 3. Each video’s
r(78) = 0.347, p = 0.002, aesthetic residuals average: r(78) = 0.423, p < 0.001). mean naturalness ratings and model-derived prototypi-
The gender category model’s performances did not differ significantly
from the single category model (aesthetic z-scores average: p = 0.946; aes- cality are depicted in Fig. 2e. There was a clear positive
thetic residuals average: p = 0.866), while the gender-plus-emotion category correlation between the two measures, both before and
model performed worse before regressing out the emotion positivity ratings after regressing out the emotion positivity ratings (with
(p = 0.007), and not significantly different after (p = 0.124).
Chen et al. Cognitive Research: Principles and Implications (2023) 8:55 Page 10 of 13

Fig. 3 a Each intact walker video’s mean aesthetic rating from Experiment 1 plotted against its single-category objective prototypicality. b
Mean single-category objective prototypicalities from intact walker videos in four emotion categories. c The modeled prototypicality’s effect
on the aesthetic z-scores was partially mediated by the naturalness z-scores. Here, s represents the slope. The results of regressions with the indirect
effect of naturalness removed are reported in parentheses

naturalness z-score: r(78) = 0.556, p < 0.001; with natural- F(3, 147) = 58.4, p < 0.001, η2p = 0.544; all post hoc com-
ness residuals after regressing out the emotion positivity: parison with neutral walkers: ts > 6.4, ps < 0.001 after
semipartial correlation r(78) = 0.540, p < 0.001), suggest- Bonferroni correction). For emotion positivity ratings, we
ing that typical movements were indeed associated with replicated the finding in Experiment 1 that happy walks
more natural impressions. This result was further sup- were perceived as the most positive (Mhappy = 3.9 (0.4),
ported by the similar patterns found with naturalness Mneutral = 3.5 (0.4), Mangry = 3.4 (0.5), Msad = 2.5 (0.4); one-
ratings (Fig. 2f ) and modeled prototypicality (Fig. 3b) way ANOVA main effect with z-scores: F(3, 147) = 227,
when we separated the results based on emotion catego- p < 0.001, η2p = 0.823; all post hoc comparison with happy
ries: Neutral walkers were perceived as the most natural, walkers: ts > 9.0, ps < 0.001 after Bonferroni correction).
followed by happy, sad, and angry walkers (Mhappy = 3.5 Next, we asked if a movement’s model-derived proto-
(0.4), Mneutral = 4.0 (0.5), Mangry = 3.1 (0.5), Msad = 3.5 typicality exerts its effect on aesthetic experience through
(0.5); one-way ANOVA main effect with z-scores: F(3, a subjective impression of prototypicality. We conducted
147) = 55.6, p < 0.001, η2p = 0.531; all post hoc comparison mediation analyses to examine both the direct effect
with neutral walker: ts > 6.0, ps < 0.001 after Bonferroni (modeled prototypicality directly influenced aesthetic
correction). Moreover, the effect persisted after regress- experience), and the indirect effect (subjective proto-
ing out the emotion positivity ratings from the natural- typicality mediated the effect on aesthetic experience).
ness ratings (one-way ANOVA main effect with residuals: Because the same stimuli were used in Experiments 1 and
Chen et al. Cognitive Research: Principles and Implications (2023) 8:55 Page 11 of 13

3, this analysis included both aesthetic judgments (from possible that other aspects of action perception are sen-
Experiment 1) and naturalness judgments (from Experi- sitive to emotion and other social intentions underlying
ment 3). With a Sobel test and the method of permuta- actions.)
tion confidence interval (Taylor & MacKinnon, 2012), At the same time, the demonstration of an aesthetic
using data from both Experiments 1 and 3, we found a prototype effect in human walks also constitutes a unique
partial indirect causal relationship (Fig. 3c) both before approach to understanding human action aesthetics. In
and after regressing out the emotion positivity: Objec- contrast to the focus on domain-specific explanations
tive prototypicality influenced the subjective impression for aesthetic experiences based on dance movements
of the prototypicality of walks (i.e., naturalness ratings), (Cross et al., 2011) or sexualized features in human walks
which in turn influenced the aesthetic impression of (Meskó et al., 2021; Morris et al., 2013), we showed that
the walks (z-score analyses: the Sobel test, t(77) = 4.84 human actions are subject to the same domain-general
p < 0.001; the permutation CI = [0.229, 0.538], which aesthetic processes that operate over static categories
did not include zero; residual analyses: the Sobel test, (Ryali et al., 2020). This way, the general explanations of
t(77) = 5.18, p < 0.001; the permutation CI = [0.274, 0.604], aesthetic preferences can also apply to human actions.
which did not include zero). The gross direct relationship As prototypical actions could reflect health and develop-
between the objective prototypicality and the aesthetic mental stability in body movements (Møller & Swaddle,
impression (z-score: slope = 0.566, p < 0.001; residuals: 1997), an aesthetic preference for prototypes may have
slope = 0.546, p < 0.001) was substantially weakened after general functional value (Chen et al., 2022a, 2022b; Hal-
removing the indirect effect of the subjective impression berstadt & Rhodes, 2003; Unkelbach et al., 2008; Vogel
of prototypicality (z-score: slope = 0.183, p = 0.027; resid- et al., 2021; Zajonc, 2001). Any particular general account
uals: slope = 0.102, p = 0.115). remains speculative, and future work is needed to fur-
ther examine the exact functions of these prototypical
Discussion preferences.
Using a combination of behavioral experiments and These discoveries inform several practical fields.
computational models, we made four main findings: (1) The discovered clear consensus of how good a walk
People share substantial consensus on how aesthetically looks suggests a new source of bias to navigate in our
pleasing a walk looks. (2) Human walks look more aes- personal and professional lives. At the same time,
thetically pleasing when they expressed positive emo- the prototype effect introduces a design principle for
tions; this preference depends on holistic processing and depicting animated humans, developing virtual agents,
explicit recognition of human body configurations. (3) and creating robots. The computational models used
Aesthetic prototype effects can be observed in human in this study can also be used for machine assistance
actions: People find prototypical walks more aesthetically in medical diagnosis, rehabilitations, and prosthetic
pleasing than atypical walks. (4) This effect was caused limbs design (Pitkin, 2013), such as developing an early
both directly by the prototypicality of the walk itself screening test for abnormal gaits, an evaluative scor-
and indirectly through the mediation of the subjective ing program for improvements from rehabilitation, or
impression of prototypicality. automatic design evaluations for how prosthetic limbs
The observed prototype effects indicate that human affect movements.
walks expressing different emotion states form a single
category within a representational space, which opens
Abbreviations
a new dimension in the exploration of categorical pro- DTW Dynamic time warping algorithm
cessing. Beyond static objects and animals, dynamic px Pixel
events can form categories and afford the same kind of
representational structures that lead to prototype effects. Supplementary Information
These categories likely form at the basic level (Rosch The online version contains supplementary material available at https://​doi.​
et al., 1976), as the emotional expressions and gender of org/​10.​1186/​s41235-​023-​00510-0.
the walkers does not divide the action representations
Additional file 1. Posthoc tests for all main effects of emotion category
into multiple categories (at least in the context of form- across three experiments and the modeling results were reported here.
ing general impressions such as aesthetic impressions).
This finding also suggested that categorical processing
Acknowledgements
of walking is not involved in differentiating emotional For annotation and helpful conversations, we thank Jeff Chang, Anika
states underlying walking actions and may mainly serve Vaishampayan, and Felix Chang.
the function of action recognition. (Note that it is still
Chen et al. Cognitive Research: Principles and Implications (2023) 8:55 Page 12 of 13

Significance statement Colombatto, C., Chen, Y.-C., & Scholl, B. J. (2020). ‘Gaze deflection’ reveals how
From dating to hiring, aesthetic impressions of people have profound (and gaze cueing is tuned to extract the mind behind the eyes. Proceedings of
sometimes undesirable) impacts on our lives. For example, judgements of the National Academy of the United States of America, 117, 19825–19829.
people’s abilities are biased by irrelevant factors from their appearances. The Cross, E. S., Kirsch, L., Ticini, L. F., & Schütz-Bosbach, S. (2011). The impact of
urgent needs to understand and mitigate such biases are reflected in the aesthetic evaluation and physical ability on dance perception. Frontiers in
abundance of research on attractiveness. However, our current understand‑ Human Neuroscience, 5(102), 1–10.
ing of what looks good has been detached from everyday experience in an Csibra, G., Gergely, G., Bı́ró, S., Koos, O., & Brockbank, M. (1999). Goal attribution
important way: Unlike the static stimuli used in past research, people move. without agency cues: The perception of ‘pure reason’ in infancy. Cogni-
Here, we used a kind of dynamic stimuli people see every day—human tion, 72, 237–267.
walking to examine the aesthetic consensus. The results revealed an aesthetic Cutting, J. E. (1981). Coding theory adapted to gait perception. Journal of
prototype effect, where viewers find prototypical actions more aesthetically Experimental Psychology: Human Perception and Performance, 7, 71–87.
pleasing. This finding contributes theoretically by linking action aesthetics Diedenhofen, B., & Musch, J. (2015). cocor: A comprehensive solution for the
to category processing. At the same time, the newly developed model for statistical comparison of correlations. PLoS One, 10(e0121945), 1–12.
action similarity provides essential evaluative tools for both identification and Emery, N. J. (2000). The eyes have it: The neuroethology, function and evolu‑
rehabilitation of abnormal movements. This new understanding of consensual tion of social gaze. Neuroscience & Biobehavioral Reviews, 24, 581–604.
aesthetic experiences from human actions can also aid developments of Fan, J., Dai, W., Liu, F., & Wu, J. (2005). Visual perception of male body attractive‑
pleasant animation, VR, and robotics technology. Most importantly, this study ness. Proceedings of the Royal Society B: Biological Sciences, 272, 219–226.
enables future explorations to understand how dynamic “looks” may have Fan, J., Liu, F., Wu, J., & Dai, W. (2004). Visual perception of female physical
profound impacts on our personal and professional lives. attractiveness. Proceedings of the Royal Society of London Series B: Biological
Sciences, 271, 347–352.
Author contributions Fink, B., Hugill, N., & Lange, B. P. (2012). Women’s body movements are a poten‑
All authors designed the research and wrote the manuscript. YC prepared the tial cue to ovulation. Personality and Individual Differences, 53, 759–763.
materials with inputs from HL and FP. YC conducted the experiments. YC and Fink, B., Weege, B., Neave, N., Pham, M. N., & Shackelford, T. K. (2015). Integrat‑
HL analyzed the data. ing body movement into attractiveness research. Frontiers in Psychology,
6(220), 1–6.
Funding Galton, F. (1878). Composite portraits. Journal of the Anthropological Institute of
This study was funded by National Science Foundation BSC-1655300 awarded Great Britain & Ireland, 8, 132–142.
to HL. Gavrila, D. M., & Davis, L. S. (1995). Towards 3-D model-based tracking and
recognition of human movement: A multi-view approach. In International
Availability of data and materials Workshop on Automatic Face and Gesture Recognition, 3, 272–277.
Experiment 1’s design and analysis plan were preregistered and can be viewed Halberstadt, J. B., & Rhodes, G. (2003). It’s not just average faces that are attrac‑
here: https://​aspre​dicted.​org/​K JD_​33W. All materials, code, and data can be tive: Computer-manipulated averageness makes birds, fish, and automo‑
downloaded here: https://​doi.​org/​10.​17605/​OSF.​IO/​87G3E. A demonstration biles attractive. Psychonomic Bulletin & Review, 10, 149–156.
of the experiments can be viewed online here: https://​yi-​chia-​chen.​github.​io/​ Hugill, N., Fink, B., Neave, N., & Seydel, H. (2009). Men’s physical strength is
walker-​proto​type-​demo-​expt/. associated with women’s perceptions of their dancing ability. Personality
and Individual Differences, 47, 527–530.
Johnson, K. L., & Tassinary, L. G. (2007). Compatibility of basic social perceptions
Declarations determines perceived attractiveness. Proceedings of the National Academy
of Sciences, 104, 5246–5251.
Ethics approval and consent to participate Kanwisher, N., & Yovel, G. (2006). The fusiform face area: A cortical region spe‑
All experiments were approved by the UCLA Institutional Review Board, and cialized for the perception of faces. Philosophical Transactions of the Royal
all observers indicated informed consent. Society B: Biological Sciences, 361, 2109–2128.
Klüver, M., Hecht, H., & Troje, N. F. (2016). Internal consistency predicts attrac‑
Consent for publication tiveness in biological motion walkers. Evolution and Human Behavior, 37,
Not applicable. 40–46.
Landwehr, J. R., Labroo, A. A., & Herrmann, A. (2011). Gut liking for the ordinary:
Competing interests Incorporating design fluency improves automobile sales forecasts. Mar-
The authors declare that they have no competing interests. keting Science, 30, 416–429.
Langlois, J. H., & Roggman, L. A. (1990). Attractive faces are only average.
Psychological Science, 1, 115–121.
Received: 6 February 2023 Accepted: 5 August 2023 Lopez-Brau, M., Colombatto, C., Jara-Ettinger, J., & Scholl, B. J. (2021). Atten‑
tional prioritization for historical traces of agency [Conference presenta‑
tion]. In Annual Meeting of the Vision Sciences Society, online.
Ma, Y., Paterson, H. M., & Pollick, F. E. (2006). A motion capture library for the
study of identity, gender, and emotion perception from biological
References motion. Behavior Research Methods, 38, 134–141.
Ackermann, B. J., & Adams, R. (2004). Interobserver reliability of general Marlowe, C. M., Schneider, S. L., & Nelson, C. E. (1996). Gender and attractive‑
practice physiotherapists in rating aspects of the movement patterns of ness biases in hiring decisions: Are more experienced managers less
skilled violinists. Medical Problems of Performing Artists, 19, 3–11. biased? Journal of Applied Psychology, 81, 11–21.
Barrett, H. C., Todd, P. M., Miller, G. F., & Blythe, P. W. (2005). Accurate judgments McCarty, K., Hönekopp, J., Neave, N., Caplan, N., & Fink, B. (2013). Male body
of intention from motion cues alone: A cross-cultural study. Evolution and movements as possible cues to physical strength: A biomechanical analy‑
Human Behavior, 26, 313–331. sis. American Journal of Human Biology, 25, 307–312.
Chen, Y.-C., Chang, A., Rosenberg, M. D., Feng, D., Scholl, B. J., & Trainor, L. J. Meskó, N., Őry, F., Csányi, E., Juhász, L., Szilágyi, G., Lubics, O., Putz, Á., & Láng, A.
(2022a). ‘Taste typicality’ is a foundational and multi-modal dimension of (2021). Women walk in high heels: Lumbar curvature, dynamic motion
ordinary aesthetic experience. Current Biology, 32, 1837–1842. stimuli and attractiveness. International Journal of Environmental Research
Chen, Y.-C., Pollick, F., & Lu, H. (2022b). Aesthetic preferences for causality in and Public Health, 18, 299.
biological movements arise from visual processes. Psychonomic Bulletin & Møller, A. P., & Swaddle, J. P. (1997). Asymmetry, developmental stability and
Review, 29, 1803–1811. evolution. Oxford University Press.
Christensen, J. F., & Calvo-Merino, B. (2013). Dance as a subject for empirical Morris, P. H., White, J., Morrison, E. R., & Fisher, K. (2013). High heels as super‑
aesthetics. Psychology of Aesthetics, Creativity, and the Arts, 7, 76–88. normal stimuli: How wearing high heels affects judgements of female
attractiveness. Evolution and Human Behavior, 34, 176–181.
Chen et al. Cognitive Research: Principles and Implications (2023) 8:55 Page 13 of 13

Morrison, E. R., Bain, H., Pattison, L., & Whyte-Smith, H. (2018). Something in
the way she moves: Biological motion, body shape, and attractiveness in
women. Visual Cognition, 26, 405–411.
Orlandi, A., Cross, E. S., & Orgs, G. (2020). Timing is everything: Dance aesthet‑
ics depend on the complexity of movement kinematics. Cognition,
205(104446), 1–12.
Palmer, S. E., Schloss, K. B., & Sammartino, J. (2013). Visual aesthetics and
human preference. Annual Review of Psychology, 64, 77–107.
Pearson, K., & Filon, L. N. G. (1898). Mathematical contributions to theory of
evolution: IV. On the probable errors of frequency constants and on the
influence of random selection and correlation. Philosophical Transactions
of the Royal Society of London Series A, 191, 229–311.
Peelen, M. V., & Downing, P. E. (2007). The neural basis of visual body percep‑
tion. Nature Reviews Neuroscience, 8, 636–648.
Pitkin, M. (2013). What can normal gait biomechanics teach a designer of
lower limb prostheses? Acta of Bioengineering and Biomechanics, 15, 3–10.
Pollick, F. E., Paterson, H. M., Bruderlin, A., & Sanford, A. J. (2001). Perceiving
affect from arm movement. Cognition, 82, B51–B61.
Posner, M. I., & Keele, S. W. (1968). On the genesis of abstract ideas. Journal of
Experimental Psychology, 77, 353–363.
Rosch, E., Mervis, C. B., Gray, W. D., Johnson, D. M., & Boyes-Braem, P. (1976).
Basic objects in natural categories. Cognitive Psychology, 8, 382–439.
Ryali, C. K., Goffin, S., Winkielman, P., & Angela, J. Y. (2020). From likely to likable:
The role of statistical typicality in human social assessment of faces.
Proceedings of the National Academy of Sciences, 117, 29371–29380.
Solso, R. L., & Raynis, S. A. (1979). Prototype formation from imaged, kinestheti‑
cally, and visually presented geometric figures. Journal of Experimental
Psychology: Human Perception and Performance, 5, 701–712.
Sparrow, W. A., Shinkfield, A. J., Day, R. H., Hollitt, S., & Jolley, D. (2002). Visual
perception of movement kinematics and the acquisition of “action proto‑
types.” Motor Control, 6, 146–165.
Taylor, A. B., & MacKinnon, D. P. (2012). Four applications of permutation
methods to testing a single-mediator model. Behavior Research Methods,
44, 806–844.
Thornhill, R., & Gangestad, S. W. (1999). Facial attractiveness. Trends in Cognitive
Sciences, 3, 452–460.
Unkelbach, C., Fiedler, K., Bayer, M., Stegmüller, M., & Danner, D. (2008). Why
positive information is processed faster: The density hypothesis. Journal
of Personality and Social Psychology, 95, 36–49.
Vogel, T., Ingendahl, M., & Winkielman, P. (2021). The architecture of prototype
preferences: Typicality, fluency, and valence. Journal of Experimental
Psychology: General, 150, 187–194.
Whitfield, T. A., & Slatter, P. E. (1979). The effects of categorization and proto‑
typicality on aesthetic choice in a furniture selection task. British Journal
of Psychology, 70, 65–75.
Willis, J., & Todorov, A. (2006). First impressions: Making up your mind after a
100-ms exposure to a face. Psychological Science, 17, 592–598.
Younger, B. (1990). Infant categorization: Memory for category-level and
specific item information. Journal of Experimental Child Psychology, 50,
131–155.
Zajonc, R. B. (2001). Mere exposure: A gateway to the subliminal. Current Direc-
tions in Psychological Science, 10, 224–228.

Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in pub‑
lished maps and institutional affiliations.