Journal Pre-proof

Psychological and neural responses to architectural interiors

Alexander Coburn, Oshin Vartanian, Yoed N. Kenett, Marcos Nadal, Franziska

Hartung, Gregor Hayn-Leichsenring, Gorka Navarrete, José Luis González Mora,
Anjan Chatterjee

PII: S0010-9452(20)30033-2
DOI: https://doi.org/10.1016/j.cortex.2020.01.009
Reference: CORTEX 2822

To appear in: Cortex

Received Date: 30 March 2019

Revised Date: 21 August 2019
Accepted Date: 10 January 2020

Please cite this article as: Coburn A, Vartanian O, Kenett YN, Nadal M, Hartung F, Hayn-Leichsenring
G, Navarrete G, González Mora JL, Chatterjee A, Psychological and neural responses to architectural
interiors, CORTEX, https://doi.org/10.1016/j.cortex.2020.01.009.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2020 Published by Elsevier Ltd.

Psychological and neural responses to architectural interiors
Alexander Coburn1,3, Oshin Vartanian2, Yoed N. Kenett3, Marcos Nadal4, Franziska Hartung3,
Gregor Hayn-Leichsenring3, Gorka Navarrete5, José Luis González Mora6, Anjan Chatterjee3

1
Department of Architecture, University of Cambridge
2
Department of Psychology, University of Toronto
3
Center for Cognitive Neuroscience, University of Pennsylvania
4
Department of Psychology, University of the Balearic Islands
5
School of Psychology, Universidad Adolfo Ibáñez
6
School of Medicine, Universidad de la Laguna

Corresponding author:
Alexander Coburn
624 Parnassus Avenue
San Francisco, CA 94122
[email protected]
(617) 306-1134
 Abstract
People spend considerable time within built environments. In this study, we tested two
hypotheses about the relationship between people and built environments. First, aesthetic
responses to architectural interiors reduce to a few key psychological dimensions that are
sensitive to design features. Second, these psychological dimensions evoke specific neural
signatures. In Experiment 1, participants (n = 798) rated 200 images of architectural interiors on
16 aesthetic response measures. Using Psychometric Network Analysis (PNA) and Principal
Components Analysis (PCA), we identified three components that explained 90% of the variance
in ratings: coherence (ease with which one organizes and comprehends a scene), fascination (a
scene’s informational richness and generated interest), and hominess (extent to which a scene
reflects a personal space). Whereas coherence and fascination are well-established dimensions
in response to natural scenes and visual art, hominess emerged as a new dimension related to
architectural interiors. In Experiment 2 (n = 614), the PCA results were replicated in an
independent sample, indicating the robustness of these three dimensions. In Experiment 3, we
reanalyzed data from an fMRI study in which participants (n = 18) made beauty judgments and
approach-avoidance decisions when viewing the same images. Parametric analyses
demonstrated that, regardless of task, the degree of fascination covaried with neural activity in
the right lingual gyrus. In contrast, coherence covaried with neural activity in the left inferior
occipital gyrus only when participants judged beauty, whereas hominess covaried with neural
activity in the left cuneus only when they made approach-avoidance decisions. Importantly, this
neural activation did not covary in relation to global image properties including self-similarity
and complexity scores. These results suggest that the visual brain harbors sensitivities to
psychological dimensions of coherence, fascination, and hominess in the context of architectural
interiors. Furthermore, valuation of architectural processing in visual cortices varies by
dimension and task.

1
 1. Introduction
People in materially developed cultures spend over 90% of their lives in buildings (Evans &
McCoy, 1998). Every day, the architecture we inhabit envelopes our mind and body and
influences how we feel and behave (Ellard, 2015). The design of our built environment can
modulate how comfortable (Baker & Standeven, 1995; Brager, Paliaga, & De Dear, 2004) or
focused (Mehta & Zhu, 2009) we feel in a given moment and can influence hormonal patterns
(Fich et al., 2014; Küller & Lindsten, 1992), speed of recovery from surgery (Ulrich, 1984), and
long-term cardiac health (Kardan, Gozdyra, et al., 2015).

Given that the brain mediates human responses to architecture, scientific interest in the
neuroscience of architecture has surged in recent years (Choo, Nasar, Nikrahei, & Walther,
2017; Coburn, Vartanian, & Chatterjee, 2017; Marchette, Vass, Ryan, & Epstein, 2015; Robinson
& Pallasmaa, 2015; Vartanian et al., 2013). However, relatively little empirical research has been
conducted on the psychology of architecture (L. T. Graham, Gosling, & Travis, 2015), aside from
a limited body of architecture-focused literature within the field of environmental psychology
(see for instance, Baum & Davis, 1980; Coburn et al., 2019; Imamoglu, 2000; Ulrich, 1984).
Unlike other areas of neuroscience, such as neurolinguistics and neuroaesthetics,
neuroarchitecture lacks an extensive behavioral literature from which to construct
neurophysiological models and generate predictions (Coburn et al., 2017).

In the context of the built environment, important research has emerged indicating the
potential psychological benefits of nature-like, i.e. biophilic, design patterns in architecture
(Africa et al., 2019; Alexander, 2002; Joye, 2007; Kellert, 2005; Salingaros, 2007, 2015). This
literature hypothesizes that biophilic architectural patterns (see for instance Alexander, 2002;
Kellert, 2005; Salingaros, 2015) and design indices (Salingaros, 2019, under review) may confer
beneficial effects such as improved mood, reduced stress, and enhanced overall wellbeing
(Coburn et al., 2019; Joye, 2007; Ryan et al., 2014; Ryan & Browning, 2018; Salingaros, 2015).
However, researchers have yet to identify the precise neural and psychological mechanisms that
may mediate the predicted long-term “healing” effects of biophilic architecture (Coburn et al.,
2019; Ryan et al., 2014). Furthermore, more general frameworks of architectural psychology
and aesthetics (i.e., empirical frameworks outlining the various types of acute mental states that
diverse architectural environments can induce) have yet to be established. Here, we seek to
advance the psychology of architecture in order to lay the groundwork for a more robust line of
research on the neuroscience of architecture. We hypothesize that interactions with
architectural scenes can be explained by a limited number of underlying psychological
constructs. This hypothesis is motivated by past studies that have identified latent psychological
dimensions underlying aesthetic responses to visual stimuli in other contexts. Examples include
the “preference matrix” of landscape aesthetics outlined by Kaplan & Kaplan (1989) and the
core dimensions of novelty and complexity as related to arousal identified by Berlyne in his
empirical investigations of aesthetic responses (Daniel E. Berlyne, 1970, 1971, 1974). To our
knowledge, no such framework has been identified yet for architecture. We also hypothesize
that salient design features of curvature, ceiling height, and enclosure can modulate these key
dimensions of architectural experience. These features have been found to influence aesthetic
responses to architecture in prior experiments (Vartanian et al., 2013, 2015). Finally, we test the
exploratory hypothesis that these psychological dimensions correspond to specific patterns of
neural activity in response to viewing images of architectural interiors.

2
 1.1. Aesthetic response measures
Viewing architectural spaces elicits a broad range of aesthetic experiences, from feelings of
comfort and excitement to judgments of a building’s age and style. However, few theoretical
models have been developed to frame empirical research on the aesthetics of architecture.
Recently, we outlined a neuroscientific model of architectural experience to serve as a
foundational framework. According to the aesthetic triad model (Figure 1), aesthetic
experiences in the built environment are mediated by three large-scale neural systems:
knowledge-meaning, emotion-valuation, and sensorimotor systems (Chatterjee & Vartanian,
2014; Coburn et al., 2017).
These neural systems align approximately with three important domains of psychological
processing: cognition, emotion, and behavior (Izard, Kagan, & Zajonc, 1988; Lench, Darbor, &
Berg, 2013; Stangor, 2015). Using this adapted terminology, we propose that architectural
encounters produce three general classes of psychological experiences: cognitive judgements
associated with knowledge-meaning systems, emotional responses derived from emotion-
valuation systems, and behavioral-motivational responses linked to sensorimotor activation.
Within this psychological framework, we applied sixteen aesthetic rating scales that capture
important aspects of architectural experience (e.g. complexity; see Table 1). These response
measures have featured prominently in previous environmental psychology and empirical
aesthetics research.

Figure 1: The aesthetic triad and associated psychological domains.

1.1.1. Cognitive judgements of architecture

When people enter buildings, they often make cognitive judgments about the spaces around
them. We define cognitive judgments as informed by top-down knowledge people bring to
evaluations made about external qualities of their surroundings, rather than self-reflective
evaluations of their own inner states of being. This distinction is based on past research
suggesting that extrospective and introspective evaluations likely involve dissociable neural
circuitry (Di Dio, Macaluso, & Rizzolatti, 2007; Leder, Oeberst, Augustin, & Belke, 2004). Here,
we discuss five key measures of cognitive judgement in the built environment: complexity,
organization, modernity, naturalness, and beauty.

3
Visual complexity has drawn attention from many architectural theorists (Alexander, 2002a;
Kroll, 1987; Salingaros, 2007; Venturi, Scully, & Drexler, 1977), environmental psychologists (R.
Kaplan & Kaplan, 1989; S. Kaplan, Kaplan, & Wendt, 1972; Ulrich, 1983), and aesthetics
researchers (Daniel E. Berlyne, 1971; Frith & Nias, 1974). Visual complexity refers to “the
volume of information present in a space” (Dosen & Ostwald, 2016, p. 3) and the informational
“richness” of a scene (R. Kaplan & Kaplan, 1989, p. 53). Positive, linear correlations between
complexity and preference have been found in various contexts, including the evaluation of
artwork (Day, 1967; Leder et al., 2004; Taylor, Micolich, & Jonas, 1999), natural landscapes (S.
Kaplan, 1987; Ulrich, 1977, 1983), and built environments (Ç. Imamoglu, 2000; S. Kaplan et al.,
1972). In some cases, preference ratings have been found to follow an inverted U-shaped curve
when plotted as a function of stimulus complexity (Daniel E. Berlyne, 1970, 1971; Güçlütürk,
Jacobs, & van Lier, 2016; Taylor et al., 1999). This relationship often depends on how complexity
is operationalized (Nadal, Munar, Marty, & Cela-Conde, 2010a), which may explain the
variability in findings.
Organization is also critical to the psychology of architecture. Visual order implies both an
absence of randomness (Tullett, Kay, & Inzlicht, 2015) and the presence of predictable patterns
like symmetry (Alexander, 2002a; Reber, Schwarz, & Winkielman, 2004; Salingaros, 2007) and
structural redundancy in scenes (Kinchla, 1977; Kotabe, Kardan, & Berman, 2016b). The
psychological effects of visual organization have been discussed extensively in architectural
theory (Alexander, 2002a; Salingaros, 2007; Vitruvius Pollio, Morgan, & Warren, 1914) and art
aesthetics literature (Birkhoff, 1933; Eysenck, 1957; Reber et al., 2004). Perception of order can
also be modulated by a building’s age, condition, and architectural style. These variables have
been captured in past studies by measuring participants’ perceptions of modernity in the built
environment (Acking & Kuller, 1973; Ç. Imamoglu, 2000; V. Imamoglu, 1979).
Interacting with natural environments enhances many aspects of psychological functioning
(Berman et al., 2012; Berto, 2005; Bratman, Daily, Levy, & Gross, 2015; S. Kaplan, 1995; Ryan,
Weinstein, Bernstein, & Brown, 2010). Naturalness appears to be a salient measure of
environmental judgement (Berman et al., 2014; Kotabe, 2016) that correlates highly with scene
preference ratings (Kardan, Demiralp, et al., 2015). Recent studies also show that the perception
of naturalness is not merely determined by natural content (e.g., recognition of trees and
vegetation) but is also predicted by specific low-level visual patterns that can occur in both
natural and man-made objects and environments (Berman et al., 2014; Coburn et al., 2019;
Kardan, Demiralp, et al., 2015; Kotabe, 2016). For instance, Graham and Field (2007) found
certain man-made paintings have similar low-level visual properties as natural scenes. Indeed,
several scholars propose that nature-like aesthetic qualities are present, to varying degrees, in
the built environment, and that naturalistic architectural spaces may confer some of the same
psychological benefits as natural landscapes (Alexander, 2002a; Joye, 2007; Kellert, 2003;
Salingaros, 1998).
Beauty, which is perhaps the most global measure of aesthetic judgment, is among the most
frequently measured qualities in empirical aesthetics (Chatterjee, 2013; Ishizu & Zeki, 2011;
Leder & Nadal, 2014; Nadal et al., 2010). Beauty has long been regarded as an important quality
of architectural design in cultures around the world (Mak & Thomas Ng, 2005; Patra, 2009;
Vitruvius Pollio et al., 1914). Efforts to understand environmental beauty have gained traction in
both environmental psychology (Cooper, Burton, & Cooper, 2014; S. Kaplan, 1987; Zhang, Piff,
Iyer, Koleva, & Keltner, 2014) and architectural research (Kirk, Skov, Christensen, & Nygaard,
2009; Vartanian et al., 2013, 2015), perhaps because of the growing view that “attractiveness is
a key element in how the built environment affects our wellbeing” (Cooper & Burton, 2014), as

4
well as the primary role that beauty plays in our desire to live in a place (Ritterfeld & Cupchik,
1996).
Although we provisionally categorized these five response measures as cognitive judgments,
they likely depend on input from all three nodes of the aesthetic triad, rather than from
cognitive processing alone. For instance, low-level spatial and color features of environmental
scenes significantly predict subjective ratings of complexity, order, and naturalness (Berman et
al., 2014; Kardan, Demiralp, et al., 2015; Kotabe et al., 2016b; Kotabe, Kardan, & Berman, 2017),
even when the semantic content of scenes is removed (Kotabe, Kardan, & Berman, 2016a;
Kotabe et al., 2016b), suggesting that these measures can be shaped by low-level sensory input.
Furthermore, the experience of beauty likely involves complex interactions among sensory,
emotional, and cognitive inputs (Chatterjee & Vartanian, 2014; Leder & Nadal, 2014; Leder et
al., 2004).

1.1.2. Emotional responses to architecture

In addition to eliciting external judgments, architectural spaces modulate affect, emotions, and
other inner states of being. Alexander (2002) emphasized the importance of judging a building
not only via detached observation of its appearance, but also by examining the degree to which
it “touches us in our humanity” (Alexander, 2002a, p. 300) and “stirs our feelings, our passion”
(Alexander, 2002a, p. 302). Several other writers have also highlighted the introspective
dimension of architectural experience (Bachelard, 1994; Heidegger, 2013; Linnet, 2012; Tanizaki,
2001). Eight measures of emotional experience in the built environment are outlined below:
personalness, hominess, relaxation, comfort, stimulation, uplift, vitality, and valence.
The degree of personal feeling that a building generates is an important consideration in
architectural design (Alexander, 2002a; L. T. Graham et al., 2015; Sommer, 1969; Wiking, 2017).
Personal spaces feel warm and intimate (L. T. Graham et al., 2015; Sommer, 1969) and generate
feelings of “depth, tenderness, and longing” (Alexander, 2002a, p. 302), whereas impersonal
spaces often feel cold and standardized (Linnet, 2012). A related measure, the degree to which
an architectural space makes a person feel cozy or “at home” (Daniels, 2015; L. T. Graham et al.,
2015; Ritterfeld & Cupchik, 1996), is captured by the Canadian concept of hominess (Linnet,
2012; Wiking, 2017). Considerable emphasis has been placed on the degree of stress or,
conversely, relaxation that people experience in response to environmental design (Baum &
Davis, 1980; Fich et al., 2014; L. T. Graham et al., 2015; Tullett et al., 2015; Tyrväinen et al.,
2014; Ulrich et al., 1991). Comfort is also a salient measure of occupant experience that
abounds in architectural research (Baker & Standeven, 1995; Brager et al., 2004; Fanger, 1973;
Nicol & Humphreys, 2002; Thorsson, Honjo, Lindberg, Eliasson, & Lim, 2007).
Researchers have taken interest in understanding how design parameters modulate the degree
of physiological stimulation that occupants experience (Acking & Kuller, 1973; L. T. Graham et
al., 2015; Ritterfeld & Cupchik, 1996). A related measure is the extent to which a place feels
uplifting, on the one extreme, and depressing, on the other (Evans, 2003). This scale may be
particularly relevant to wellbeing, as the frequency of daily uplifts a person experiences predicts
long-term health measures like stress and depression (Kanner, Coyne, Schaefer, & Lazarus,
1981; Vitaliano, Scanlan, Ochs, & Syrjala, 1998). Scholars have also measured the impact of
environmental design on vitality (Ryan et al., 2010; Tyrväinen et al., 2014), which covaries with
important physiological and psychological health measures (Ryan & Deci, 2008; Ryan &
Frederick, 1997). Vitality has been defined as “a positive sense of aliveness and energy” (Nix,
Ryan, Manly, & Deci, 1999, p. 530) and is closely related to the Chinese concept of chi, which Nix
and colleagues defined as a source of calm energy that “can be more or less accessed by

5
individuals depending on their lifestyles and personal practices” (Nix et al., 1999, p. 268). A
related but broader measure, valence, describes the degree to which an architectural space
makes an occupant feel good or bad. Valence is among the most frequently studied affective
measures in empirical aesthetics and is closely related to other common measures such as
preference, liking, and pleasantness (Acking & Kuller, 1973; Daniel E. Berlyne, 1970; Di Dio et al.,
2007; Leder et al., 2004).
Although these affective response scales are associated with neural networks regulating
pleasure and emotion, it is likely that cognitive and sensory processes also influence emotional
responses to architecture. For instance, hominess ratings are likely modulated by cognitive
evaluations based on an individual’s culture, upbringing, and memories of home. Pleasure
responses to architectural scenes have also been shown to depend on education and expertise
(Kirk et al., 2009), suggesting that valence may be influenced by top-down cognitive processing.

1.1.3. Behavioral-motivational responses to architecture

The final class of aesthetic response scales encompasses the psychological measures of
behavior, movement, and motivation, which may be to a first approximation linked to
sensorimotor processing in the brain. Here, we focus on three behavioral measures: interest,
approachability, and explorability.
Interest, an important response measure in empirical aesthetics (Daniel E. Berlyne, 1971; Day,
1967; Silvia, 2005, 2012) and environmental psychology (R. Kaplan & Kaplan, 1989; Ulrich,
1983), is closely linked to sensory perception (Day, 1967) and motivation (Silvia, 2008). James
(1892) described interest as an automatic psychological process that enables us to identify and
attend to sensory stimuli that are important for our welfare. Environmental psychologists later
applied this idea to landscape perception by proposing that sensory features of the
environment are more likely to capture human interest if they prove beneficial or detrimental to
our species’ survival over the course of evolutionary history (Appleton, 1975; S. Kaplan, 1987;
Wilson & Kellert, 1995).
Interest can also motivate motor responses to physical surroundings (Joye & Dewitte, 2016; R.
Kaplan & Kaplan, 1989; Ulrich, 1983), including fundamental decisions to approach or avoid
architectural spaces (Ritterfeld & Cupchik, 1996; Vartanian et al., 2015, 2013). Another
important behavioral response to architecture is “the need to explore, to find out more about
what is going on in one’s surroundings” (R. Kaplan & Kaplan, 1989, p. 51). Although these
response measures are associated with sensorimotor processing, they likely involve input from
cognitive and affective domains discussed previously. Despite being strongly influenced by
sensory content, interest has often been described as a measure of emotion (Silvia, 2005, 2008,
2012), and could be categorized as an affective response measure. Like valence and beauty,
approachability describes a global psychological response that is likely modulated by cognitive
and emotional processes.
These 16 aesthetic response measures have been widely studied in environmental psychology
and represent important aspects of architectural experience. In the next section, we introduce
three salient architectural variables that have previously been shown to modulate neural and
behavioral responses to the built environment.

1.2. Architectural variables

1.2.1. Ceiling Height

6
Research suggests that ceiling height can affect psychological responses to architectural
interiors. On average, preferences for ceiling height peak around 10 feet across a range of
spatial functions (Baird, Cassidy, & Kurr, 1978). In a recent study investigating the effect of
ceiling height on aesthetic perceptions and neural activity, spaces with high ceilings received
higher beauty ratings than those with low spaces. Functional magnetic resonance imaging
(fMRI) results showed that rooms with high ceilings differentially activated neural structures
involved in visuospatial attention and exploration, such as the left middle frontal gyrus and left
precuneus (Vartanian et al., 2015). These findings were consistent with previous research
indicating that high ceilings increase perceptions of spaciousness (Stamps, 2011) and prime
thoughts of freedom, whereas low ceilings are more likely to prime thoughts of confinement
(Meyers-Levy & Zhu, 2007).

1.2.2. Enclosure

Spatial enclosure has been found to modulate aesthetic and psychological responses to building
interiors. Appleton’s prospect-refuge theory (1975) proposed that humans have evolved innate
preferences for environments that offer opportunities to see (i.e., points of prospect) without
being seen (i.e., points of refuge). Such places, he argued, have historically proven beneficial to
our species survival by enabling humans to see and hide from threats (Appleton, 1975). In
support of this theory, evidence suggests that humans generally feel safer in more open spaces
(Stamps, 2005) and also tend to prefer interior environments that afford greater visual
connection with external surroundings (Vartanian et al., 2015), when controlling for other
factors.
In a study of psychological and neural responses to open and enclosed architectural interiors,
participants were more likely to want to approach open rooms and to rate those rooms as
beautiful in comparison to enclosed interiors. Open spaces also activated neural areas
associated with perceived visual motion, whereas enclosed surroundings activated neural
regions involved in fear processing (Vartanian et al., 2015). This finding was theoretically
consistent with results from a previous study indicating that enclosed spaces, relative to open
environments, increase vulnerability to stress and prolong an occupant’s stress response
following exposure to an induced stress test (Fich et al., 2014).

1.2.3. Curvature

Geometric contour, or curvature, has generated much interest from aesthetics and architectural
researchers. In many contexts, people exhibit greater preferences for curvilinear than rectilinear
objects (Bar & Neta, 2006; Dazkir & Read, 2012; Leder & Carbon, 2005). Rectilinear shapes and
patterns also evoke more unpleasant emotions compared to curvilinear forms (Hevner, 1935;
Lundholm, 1921; Poffenberger & Barrows, 1924). These perceptual effects may extend to the
built environment. People prefer airport passenger areas that embody curvilinear rather than
rectilinear design (Van Oel & van den Berkhof, 2013). In our study on the perception of
architectural curvature, for instance, we found that curved building interiors were judged as
more beautiful than rectilinear spaces. Curved buildings also activated key areas of the visual
cortex, including the lingual and calcarine gyrus, when participants made approach-avoidance
decisions (Vartanian et al., 2013). It has been theorized that people prefer curved forms over
rectilinear forms in the built environment because curved forms are more commonly found in
nature and thus feel inherently more natural (Coburn, 2019; Kellert, 2005; Salingaros, 2015).
Supporting this idea, the density of curved edges has been found to correlate positively with
perceptions of naturalness and aesthetic preference for images of outdoor spaces, whereas the

7
density of straight edges has been shown to correlate negatively with perceptions of
naturalness and preference for such spaces (Berman et al., 2014; Ibarra et al., 2017; Kardan et
al., 2015). Curvature has consequently been identified as an example of nature-like or “living”
structural pattern in architecture (Salingaros, 2015).
Although these three variables do not exhaustively capture the diversity of architectural
geometry, they represent a useful starting point for investigating psychological responses to the
built environment. The next section outlines the research questions and hypotheses that
motivated our three experiments.

1.2.4. Global Image Properties

In addition to the above-mentioned architectural variables, we also tested (in Experiment 2)

whether key Global Image Properties (GIPs) of architectural scenes correlated with the principal
psychological components. GIPs are computed measures of global psychophysical properties of
scenes. They capture quantitative information about whole scenes, thereby complementing the
qualitative architectural variables (ceiling height, enclosure, and curvature).
Two GIPs were measured: self-similarity and complexity. Self-similarity implies that an image as
a whole is structurally similar to its parts. Complexity represents the amount of detail in an
image. These GIPs were chosen because both have consistently been found to correlate highly
with aesthetic preference ratings in studies of visual art, architecture and landscapes (Redies et
al., 2012; Mullin et al., 2015; Hayn-Leichsenring, Kenett, Schulz & Chatterjee, unpublished data).
Architectural scholars have also emphasized the importance of self-similarity and complexity as
key patterns that contribute to the beauty of architectural design (self-similarity: Alexander,
2002; Capo, 2004; Crompton, 2002; Goldberger, 1996; Salingaros, 2007; complexity: Alexander,
2002; Salingaros, 2007; Venturi et al., 1966).

1.3. Research questions

Three research questions motivated the following experiments: 1) Can aesthetic responses to
architectural scenes be reduced to a few key psychological dimensions? 2) Are these dimensions
sensitive to salient design features and GIPs? 3) Do these psychological dimensions correlate
with neural activation patterns, and to what extent are these correlations modulated by task?
We hypothesized that 1) a few key psychological dimensions would explain much of the
variance underlying diverse aesthetic response measures, and that 2) these dimensions would
be sensitive to ceiling height, enclosure, and curvature. Furthermore, we predicted that 3) each
latent psychological dimension would be linked to a distinct pattern of neural activation.

2. Experiment 1
In Experiment 1, we asked participants to rate images of building interiors on 16 aesthetic
measures that capture important aspects of architectural experience. We then carried out two
complementary approaches, psychometric network analysis (PNA) and principal components
analysis (PCA), to identify whether or not the original 16 measures could be reduced to a few
latent psychological dimensions. The stimuli were counterbalanced on three architectural
variables of interest (ceiling height, enclosure, and curvature), and we examined the degree to
which these spatial properties influenced psychological response measures. Finally, we
investigated correlations between psychological responses and two Global Image Properties
(GIPs) of architectural scenes, Self-Similarity and Complexity. These GIPs were chosen because

8
both have consistently been found to correlate highly with aesthetic preference ratings in
studies of visual art, architecture and landscapes.

2.1. Methods

2.1.1. Materials

The stimuli for this experiment were 200 images of interior architectural spaces (the complete
stimulus set is displayed in S1 and S2 of Supplementary Materials). These same images were
previously used in three studies (Vartanian et al., 2015, 2013, 2019). The stimuli were selected
from image databases accessed by the Department of Architecture, Design, and Media
Technology of Aolborg University and The Royal Danish Academy of Fine Arts School of
Architecture. Specifically, two architects independently rated every image on (a) perceived
enclosure (open, closed), (b) ceiling height (high, low), and (3) contour (round, square). The
image set included in the study consisted only of those images on which the two independent
raters reached 100% agreement regarding its standing on each of those three dimensions. Thus,
the spaces selected for the study varied on three environmental parameters. Half of the rooms
were enclosed, while the other half were open. Half had high ceilings and half had low ceilings.
Finally, half of the interiors had curvilinear edges (“round” condition), while the other half were
rectilinear (“square” condition). This setup yielded the eight experimental conditions outlined in
Figure 2 (n = 25 per condition): closed square low, closed square high, closed round low, closed
round high, open square low, open square high, open round low, and open round high.

9
Figure 2: Eight experimental conditions (n = 25 per condition) were generated by counterbalancing three
architectural variables (ceiling height, enclosure, and curvature) across the stimulus set.

2.1.2. Participants

We recruited 798 US-based adults (391 women, 401 men, 6 other) from Amazon’s Mechanical
Turk to participate in this study. Sample size was determined by our goal of obtaining
approximately 50 ratings per image on each of the sixteen aesthetic rating scales. Ages ranged
from 18 to 75 years (M = 38.06, SD = 11.96), and education level ranged from 5 to 22 years (M =
15.04, SD = 2.11). Participants were compensated $4.00 for their participation and the
experiment took approximately 40 minutes to complete. Informed consent was obtained from
each participant and the study was approved by the IRB of the University of Pennsylvania. Four
participants repeated the study twice. For each of these participants, data from the second
round of testing were excluded from analysis.

2.1.3. Procedures

Participants collectively rated 200 images of architectural interiors on 16 aesthetic rating scales
(Table 1). Approximately 50 ratings were collected per image for each scale. The stimuli were
divided into four blocks of 50 images. Each image block contained an even distribution of
images from each of the eight architectural conditions, with 6-7 randomly selected stimuli
represented from each condition per block (see S3 in Supplementary Materials). This blocking
scheme ensured that participants had approximately equal exposure to each architectural
condition for each rating task they completed. The aesthetic rating scales were also divided into
rating groups, with four response measures in each group. Sixteen rating groups were created,
each containing a unique combination of four rating scales (see S4 in Supplementary Materials).

At the start of the experiment, participants were presented with a slideshow of all 200 images
shown in random order. This was intended to familiarize them with the full range of stimuli

10
before they rated any images. Participants were subsequently assigned, at random, to one of
the sixteen rating groups. They were then presented with one of the four image blocks and
were asked to rate every image within that block on one of the four ratings scales from their
assigned rating group. Next, they rated images from a second image block on a second rating
scale, images from a third block on a third rating scale, and images from the final block on the
fourth rating scale. Ratings were entered on a 7-point sliding semantic differential scale
displayed below the image. Prompts and scale anchors are shown in Table 1. The presentation
order of the four image blocks and the assigned order of the four rating tasks were randomized.
Images within each block were also presented to participants in a randomized sequence. This
design allowed participants to experience a variety of rating tasks while minimizing the cognitive
demands of frequent task switching (Monsell, 2003). It also ensured that images received an
equal number of ratings on each scale and minimized ordering effects by assigning diverse
combinations of rating task sequences to different participants. After completing the study,
participants were asked to fill out a brief demographics questionnaire.

Table 1: Prompts and end anchors of 7-point rating scales for the aesthetic rating scales.
Aesthetic Rating Scale Rating Prompt Low Anchor High Anchor
Complexity This room looks… Simple Complex
Organization This room looks… Disordered Organized
Naturalness This room looks… Artificial Natural
Beauty This room looks… Ugly Beautiful
Personalness This room looks… Impersonal Personal
Interest This room looks… Boring Interesting
Modernity This room looks… Aged Modern
Valence This room makes me feel… Bad Good
Stimulation This room makes me feel… Bored Excited
Vitality This room makes me feel… Lifeless Alive
Comfort This room makes me feel… Uncomfortable Comfortable
Relaxation This room makes me feel… Stressed Relaxed
Hominess This room makes me feel… Alienated At home
Uplift This room makes me feel… Diminished Uplifted
Approachability If I saw this room, I'd… Leave Enter
Explorability If I saw this room, I'd… Ignore it Explore it

2.2. Analysis & Results

Data analysis was carried out at the item level. This analysis was achieved by calculating the
average rating for each image on every aesthetic rating scale. To identify the principal
psychological components of architectural experience, we applied two complementary
approaches: a psychometric network analysis (PNA; Christensen, Kenett, Aste, Silvia, & Kwapil,
2018) and principal component analysis (PCA; Field, Miles, & Field, 2014). Next, three-way
factorial ANOVAs were calculated to determine the degree to which the three architectural
variables predicted principal component scores.

2.2.1. Psychometric network analysis (PNA)

A novel approach to studying multidimensional psychological constructs is through network

science (Christensen et al., 2018). This approach is applied at the cognitive and psychological
levels to study cognitive phenomena (Baronchelli, Ferrer-i-Cancho, Pastor-Satorras, Chater, &
Christiansen, 2013; De Deyne, Kenett, Anaki, Faust, & Navarro, 2016; Isvoranu, Borsboom, van

11
Os, & Guloksuz, 2016; Isvoranu, van Borkulo, et al., 2016; Karuza, Thompson-Schill, & Bassett,
2016). Recent research has applied psychometric network analysis (PNA; Christensen et al.,
2018; Epskamp, Maris, Waldorp, & Borsboom, 2016) to investigate the intricate interactions of
psychopathology and personality (Costantini et al., 2017). The network approach defines
psychological constructs as complex systems—phenomena that emerge from the causal
interactions between dimensions of a multidimensional construct (Borsboom & Cramer, 2013;
Schmittmann et al., 2013). Such an approach can offer a unique perspective for examining the
psychological components of architectural experience, by defining the 16 measures of
architectural experience as nodes in an “aesthetic network” and examining the interaction
between these nodes. Constructing this network allows us to examine the structure of the
network, to investigate how the nodes cluster into “communities,” and to map interactions
between nodes.

Figure 3: Correlation matrix of 16 aesthetic rating scales. This figure was created using the stats (R Core
Team, 2016) and corrplot (Wei & Simko, 2016) packages in R.

We first constructed a psychological response network. In this network, nodes represent the
sixteen measures of architectural experience and edges represent the rating associations
between items, i.e., the similarity of average ratings across items for different measures. To
prepare the data for network analysis, a correlation matrix (Figure 3) was plotted across the
sixteen aesthetic response measures using the stats (R Core Team, 2016), corrplot (Wei &
Simko, 2016), and psych (Revelle, 2016) packages in R (R Core Team, 2016). The rating
12
correlation matrix was examined as an adjacency matrix of a weighted, undirected network.
With this approach, each aesthetic response measure represents a node in the network and the
edges between two measures represent the correlation between them. The weight (i.e.,
correlation) of the edge is indicated by the correlation between two nodes. Therefore, an
adjacency (or connectivity) matrix corresponds to an n x n matrix, where n is the number of
measures (nodes) and each cell represents a correlation between two measures. Most of the
edges will have small values or weak correlations, which represents noise in the network. To
minimize such noise and possible spurious correlations, the Planar Maximally Filtered Graph
(PMFG) method was used, which constructs a sub-graph, capturing the most relevant
information (i.e., removal of spurious connections and retaining high correlations) within the
original network (Kenett, Kenett, Ben-Jacob, & Faust, 2011; Tumminello, Aste, Di Matteo, &
Mantegna, 2005).
To visualize the networks, we applied the force-directed layout of the Cytoscape software
(Shannon et al., 2003). In these 2D visualizations (Figure 4), nodes (i.e., aesthetic response
measures) are represented as circles and edges between them are represented by lines. Since
these networks are unweighted and undirected, the links merely convey symmetrical (i.e.,
bidirectional) relations between two nodes. Analyzing the structure of the network, three
communities were found. Community 1 was closely associated with three aesthetic response
measures (organization, modernity, and beauty); five measures comprised Community 2
(naturalness, personalness, relaxation, hominess, and comfort); and four communities clustered
onto Community 3 (explorability, complexity, interest, and stimulation). The four remaining
aesthetic measures (uplift, valence, vitality, and approachability) were grouped at the
intersection of the three communities. The discovery of these three communities motivated a
PCA to further identify the psychological dimensions of architectural experience.

13
 Figure 4: Visualization of the aesthetic network.

2.2.2. PCA of aesthetic response measures

A PCA was carried out to identify the principal components underlying the sixteen aesthetic
response measures. The correlation matrix that was plotted in the previous analysis (Figure 3)
revealed a high degree of covariance across many of the original aesthetic measures. The
determinant of the correlation matrix (DCM) was calculated using the stats R package (R Core
Team, 2016), yielding a value of 6.3 x 10-14. This value was substantially below the minimum
threshold of 1x10-5 recommended by Field et al. (2014), indicating that the multicollinearity
among the dependent variables was too high to perform an accurate factor analysis. To remedy
this problem, six variables were excluded from factor analysis because each exhibited high
bivariate correlations (above 0.9) with at least one of the retained variables. The excluded
variables were vitality (0.92 correlation with valence), uplift (0.96 correlation with valence),
comfort (0.91 correlation with valence), relaxation (0.91 correlation with valence), stimulation
(0.93 correlation with interest), and explorability (0.92 correlation with interest). Modernity was
also excluded from factor analysis to further reduce redundancy, and because it was deemed
the least theoretically relevant of the remaining 10 rating scales. After excluding these variables
from the analysis, the DCM for the nine retained measures yielded a value of 4.8 x 10-6, which
was within an acceptable range of the recommended threshold (Field et al., 2014). For further
discussion of the methodological reasons for excluding redundant variables in PCA, see (Field et
al., 2014, Chapter 17).
PCA was performed on the 9 retained variables with oblique (oblimin) rotation, using the
“principal” function in the psych R package (Revelle, 2016). The Kaiser-Meyer-Olkin (KMO) index
score of 0.83 confirmed the sampling adequacy for the PCA, and all KMO values for individual
variables were above 0.63. KMO values were calculated using the “KMO” function of the psych
package in R (Revelle, 2016). Bartlett’s sphericity test indicated that correlations between
variables were sufficiently high for PCA (χ2= 2392, p < .001). Bartlett’s test was run using the
“cortest.bartlett” function of the psych package in R (Revelle, 2016). An initial PCA was carried
out with 9 components retained to determine eigenvalues for each component in the data. The
first three components were retained, given that all three had eigenvalues exceeding Jolliffe’s
criterion of 0.7 and together explained 90% of the variance. The decision to retain three
components was also consistent with the identification of three communities in the preceding
network analysis (Figure 3).
Table 2 shows the factor loadings, eigenvalues, and explained variance for each of the three
retained principle components after oblimin rotation. The variables that cluster on each
component suggest that PC1 represents a sense of coherence, PC2 represents the feeling of
hominess, and PC3 captures the experience of fascination. In any PCA analysis, the naming of
components is a challenging task, as this process requires some degree of interpretation from
the authors. We chose the terms coherence and fascination as names for PC1 and PC3,
respectively, because these were two important terms use frequently in the Kaplans’
foundational research on landscape aesthetics (Berman, Jonides, & Kaplan, 2008; S. Kaplan,
1995; S. Kaplan & Berman, 2010; S. Kaplan & R. Kaplan, 1989). We felt that the particular
combination of variable loadings for these components suggested a close alignment with these
two concepts that the Kaplans had previously identified as important dimensions of landscape
experience (see discussion). Naming PC2 was more challenging because there was no clear term
in architectural scholarship or environmental psychology that clearly unified the concepts of
naturalness, personalness, and hominess. We ultimately chose one of the loading variable

14
names, hominess, as the component name. Hominess is a term frequently used in Canadian
culture to describe an intimate environmental experience, and we felt that this concept would
be straightforward for most English speakers to understand. One of us (AC) has presented these
data several times to non-neuroscientist and non-architect audiences. Anecdotally, from his informal
querying of these audiences, people resonate with and understand hominess more easily than
naturalness and personalness. Alternative component names and concepts for PC2 are further
addressed in the discussion.
Figure 5 displays these PCA results in graphical form. Each arrow represents a discreet
psychological variable, and each axis represents a principal component. The size and direction of
the arrows indicates the proximity of the original variables to the latent principal components.
Finally, Figure 6 displays the correspondence of the network structure identified in the aesthetic
network and the 3 PCs. The arrows display each PC overlaid on the corresponding “community”
(i.e. cluster of nodes) of the network.

Table 2: Factor loadings on the three principal components.

Aesthetic PC1 PC2 PC3
Rating Scale (Coherence) (Hominess) (Fascination)
Complexity -0.08 -0.06 1.02
Organization 1.04 -0.19 -0.12
Naturalness -0.31 0.90 -0.04
Beauty 0.76 0.19 0.24
Personalness 0.10 0.83 0.11
Interest 0.37 0.08 0.71
Valence 0.74 0.22 0.26
Hominess 0.49 0.73 -0.09
Approachability 0.69 0.21 0.34

Eigenvalue 3.57 2.39 2.16

Variance Explained 40% 27% 24%
Cumulative Variance 40% 66% 90%

15
Figure 5 (Top): Factor loadings on PC1 (X-axis) and PC2 (Y-axis). (Bottom): Factor loadings on PC1 (X-axis) and
PC3 (Y-axis). Graphics were created using the “biplot” function of the stats R package (R Core Team, 2016).

16
Figure 6: Diagram of 3 principal components overlaid onto the aesthetic network. Each blue arrow
represents a principal component corresponding to a community (cluster of nodes) in the network. Note
that seven of the variables included in the PNA were excluded from the PCA due to multicollinearity.

2.2.3. Architectural variables predicting principal component scores

3-way factorial ANOVAs were carried out using the stats (R Core Team, 2016) and ez (Lawrence,
2016) R packages to determine the relationship between principle component scores and the
three architectural variables of interest. Graphical and statistical results of this analysis are
displayed in Figure 7. There were significant main effects of ceiling height [F (1,192) = 13.56, p <
.001, ηp2 = 0.08], enclosure [F (1,192) = 5.21, p = .024, ηp2 = 0.03], and curvature [F (1,192) =
14.94, p < .001, ηp2 = 0.09] on PC3 (fascination) as well as significant main effects of enclosure
on PC1 [coherence; F (1,192) = 6.39, p = .012, ηp2 = 0.03] and PC2 [hominess; F (1,192) = 10.94, p
= .001, ηp2 = 0.05]. No significant interaction effects were found among the three architectural
variables.

17
Figure 7: 3-way factorial ANOVA results and plots of principal component scores as a function of
architectural variables. Error bars represent 95% confidence intervals of PC scores for each condition.
Visualizations were created using JASP statistical software (Wagenmakers, 2016).

2.2.4. Global Image Properties (GIPs) of scenes predicting principal component scores

In the next analysis, we set out to determine whether key Global Image Properties (GIPs) of
architectural scenes correlated with the three principal psychological components identified in
the preceding sections. Two GIPs were measured in this analysis: self-similarity and complexity.
Self-similarity implies that an image as a whole is structurally similar to its parts. Complexity
represents the amount of detail in an image.
Quantitative measures of self-similarity and complexity were calculated for the 200
architectural images. Self-similarity was measured using the Pyramidal Histogram of Oriented
Gradients (PHOG) method. For every image, mean strength of luminance gradients is binned
over orientations resulting in histograms of oriented gradients (HOGs). Then, the image is
divided into 4 (level 1), 16 (level 2) and 64 (level 3) rectangles of similar size. To obtain a
measure for self-similarity, the HOG features of the entire image are compared with the HOG
features of the sub-images (see Dalal & Triggs, 2005; Bosch, Zisserman, & Munoz, 2007).
Complexity was measured using the HOG Complexity method. Here, the mean strength of the
gradients across all orientations is used as measure for image complexity (for a detailed
description of both methods, see Braun et al., 2013). These calculated values were then

18
regressed on principal component scores of individual images in order to determine correlations
between GIPs and psychological components of architectural experiences.
The results of this analysis are shown in Table 3. Self-similarity was found to correlate positively
with PC3 (Fascination) scores (r = .242, p < .001) and negatively with PC1 (Coherence) scores (r =
-.198, p = 0.005). In other words, images exhibiting a high degree of self-similarity were
perceived as significantly more fascinating but less coherent, relative to those with a low degree
of self-similarity. Complexity correlated positively with PC2 (Hominess) scores (r = .175, p = .013)
and PC3 (Fascination) scores (r = .519, p < .001), indicating that more complex scenes were
perceived as more home-like as well as more fascinating, compared to less complex scenes.

Table 3: GIPs predicting Principal Component Scores

Global Image Property PC1 PC2 PC3
(Coherence) (Hominess) (Fascination)
Self-Similarity r = -.198; p = .005 r = .112; p = .113 r = .242; p < .001
Complexity r = -.063; p = .378 r = .175; p = .013 r = .519; p < .001

2.3. Summary
In Experiment 1 we applied two data-driven approaches—PNA and PCA—to identify latent
psychological dimensions underlying aesthetic ratings of architectural scenes. PNA revealed
three communities in an “aesthetic network” showing the relationships between 16 aesthetic
response measures. These communities motivated our PCA analysis which replicated the three-
component structure. We interpret these three components as coherence (PC1), hominess
(PC2), and fascination (PC3). Furthermore, analyses revealed that responses along these three
dimensions were sensitive to key qualitative variables of architectural design (ceiling height,
enclosure, and curvature), as well as to quantitative global image properties (self-similarity and
complexity).
In order to prevent participant fatigue in Experiment 1, we had each participant rate all images
on only four of the 16 aesthetic measures. This study design may have reduced the reliability of
the PCA given that each participant was exposed to only a subset of the dependent variables.
This design also required us to undertake item-level analysis for the network analysis and PCA.
We were therefore unable to account for within-participant ratings across items in the PCA,
which reduced the variability of our data set. In order to address these limitations, we designed
a replication study (Experiment 2) in which participants rated a subset of images across all
dependent variables. This study is described in the next section.

19
 3. Experiment 2
In Experiment 2, we conducted a replication study to investigate the robustness of our findings
from Experiment 1, by determining if a different experimental design would yield the same
three principal components. Whereas each participant in the Experiment 1 rated all 200
architectural images on a subset of 4 aesthetic rating scales, participants in Experiment 2 were
asked to rate a subset of architectural images on all 9 non-redundant aesthetic rating scales.
This new design enabled us to perform a more robust PCA that accounted for each participant’s
within-participant ratings for each architectural condition across all of the dependent measures
of interest.

3.1. Methods

3.1.1. Participants

614 American adults (305 women, 307 men, 2 other) were recruited from Amazon’s Mechanical
Turk to participate in this study. Data from 12 additional participants was excluded from analysis
due to non-adherence to experimental instructions. Sample size was determined by our goal of
obtaining approximately 50 ratings per image on each of the nine aesthetic response measures.
Ages ranged from 19 to 72 years (M = 35.68, SD = 10.87), and education level ranged from 2 to
26 years (M = 15.26, SD = 2.31). Participants were compensated $2.40 for their participation and
the experiment took approximately 20 minutes to complete. Informed consent was obtained
through the IRB of the University of Pennsylvania.

3.1.2. Procedures

The 200 architectural images were divided into the eight experimental conditions shown in
Figure 2 (n = 25 per condition): closed square low, closed square high, closed curved low, closed
curved high, open square low, open square high, open round low, and open round high. Each of
these conditions was then split into a low-beauty group and a high-beauty group, based on the
images’ beauty scores from Experiment 1, yielding a total of 16 groups of images. Images that
received the median beauty score within each 25-image condition were alternately assigned to
either the low-beauty group or the high-beauty group for that condition. This median split was
introduced along the beauty dimension to ensure that each participant was exposed to
examples of both high and low beauty scenes within each architectural condition.
Each participant rated a batch of 16 images on all nine dependent psychological measures.
Batches were created by randomly selecting one image from each of the 16 groups. This design
ensured that each participant rated one low-beauty image and one-high beauty image from
each experimental condition. Participants rated all 16 images on one dependent measure
before moving onto the next rating task to minimize fatigue from frequent task-switching
(Monsell, 2003). The order of image presentation was randomized within each individual rating
task, and the order in which the nine ratings tasks were assigned was also randomized within
each participant. After completing the study, participants were asked to fill out a brief
demographics questionnaire.

20
 3.2. Analyses & Results

3.2.1. PCA of aesthetic response measures

Correlations among the nine psychological measures were analyzed using the stats (R Core
Team, 2016), corrplot (Wei & Simko, 2016), and psych (Revelle, 2016) packages in R (R Core
Team, 2016). The correlation matrix (Figure 8) yielded a DCM value of 7.7 x 10-3. This was above
the recommended minimum threshold of 1x10-5 (Field et al., 2014), indicating that
multicollinearities among the psychological variables were sufficiently low to perform a reliable
principal components analysis.

Figure 8: Correlation matrix of 9 psychological variables from Experiment 2.

We performed a PCA on the 9 dependent variables with oblique (oblimin) rotation. As in

Experiment 1, all analyses were completed using the psych package in R (Revelle, 2016). The
Kaiser-Meyer-Olkin (KMO) index score was 0.9, confirming the sampling adequacy for the PCA.
KMO values for all individual variables were above 0.86. Bartlett’s sphericity test showed that
correlations among variables were sufficiently high for PCA (χ2= 47843, p < .001). An initial PCA
was carried out with 9 components retained to determine eigenvalues for each component in
the data. The first three components had eigenvalues exceeding Jolliffe’s criterion of 0.7 (Field
et al., 2014) and together explained 76% of the variance. These three components were
retained.
Table 3 displays the factor loadings, eigenvalues, and variance explained for each of the three
retained principle components after oblimin rotation. A similar factor structure emerged as we
found previously in Experiment 1. In the replication, PC1 captured the feeling of hominess, PC2
represented coherence, and PC3 described the experience of fascination. Thus, Experiment 2
closely replicated the PCA results of Experiment 1, with the exception that hominess explained
more of the overall variance than coherence in the follow-up study.

21
In order to quantify the similarity of factor structures, we identified coefficients of factor
congruence using the ‘factor.congruence’ function in the psych package of R. Factor
congruences are the cosines of pairs of vectors defined by the loadings matrix. Across the two
PCAs, each component of the replication PCA had a factor congruence of 0.98 with the
corresponding component of the original PCA (accounting for the fact that PC1 and PC2 of the
original study were “flipped” in the replication study). This calculation suggests a very high
degree of similarity between the factor structures of the two PCAs.

Table 3: Factor loadings on the three principal components.

Aesthetic PC1 PC2 PC3
Rating Scale (Hominess) (Coherence) (Fascination)
Complexity -0.10 -0.10 0.98
Organization -0.15 0.97 -0.13
Naturalness 0.90 -0.15 -0.19
Beauty 0.28 0.57 0.31
Personalness 0.79 -0.04 0.14
Interest 0.24 0.27 0.59
Valence 0.42 0.53 0.20
Hominess 0.74 0.20 0.04
Approachability 0.37 0.52 0.26

Eigenvalue 2.73 2.28 1.83

Variance Explained 30% 25% 20%
Cumulative Variance 30% 56% 76%

3.2.2. Determining the influence of architectural variables on psychological ratings

3-way factorial ANOVAs were carried out using the ANOVA function in JASP statistical software
(Wagenmakers, 2016) to determine the effect of the three architectural variables on principle
component scores (Figure 9). There were significant main effects of ceiling height [F (1, 9816) =
15.23, p < .001, ηp2 = 0.002], enclosure [F (1, 9816)= 118.43, p < .001, ηp2 = 0.012], and curvature
[F (1, 9816) = 20.95, p < .001, ηp2 = 0.002] on PC1 (hominess). For PC2 (coherence), there were
also significant main effects of ceiling height [F (1, 9816) = 28.25, p < .001, ηp2 = 0.003],
enclosure [F (1, 9816) = 180.39, p < .001, ηp2 = 0.018], and curvature [F (1, 9816) = 13.58, p <
.001, ηp2 = 0.001]. Finally, significant main effects were found for ceiling height [F (1, 9816] =
243.00, p < .001, ηp2 = 0.024), enclosure [F (1, 9816) = 61.21, p < .001, ηp2 = 0.006], and
curvature [F (1, 9816) = 232.83, p < .001, ηp2 = 0.023] on PC3 (fascination). No significant
interaction effects were found.

22
Figure 9: 3-way factorial ANOVA results and plots of principal component scores as a function of
architectural variables. Error bars represent 95% confidence intervals of PC scores for each condition.
Visualizations were created using JASP statistical software (Wagenmakers, 2016).

3.3. Summary
In Experiment 2, we replicated the findings of Experiment 1 using a different study design in
which participants rated a subset of architectural images on all nine non-redundant aesthetic
response measures. Importantly, our findings replicate the three latent dimensions identified by
the PNA and PCA analyses in Experiment 1 (Figure 6). These dimensions are coherence (ease
with which one organizes and comprehends a scene), fascination (a scene’s informational
richness and generated interest), and hominess (extent to which a scene feels like a personal
space). Furthermore, analyses revealed that responses along these three dimensions were
sensitive to key variables of architectural design (ceiling height, enclosure, and curvature). The
directions of correlation between the three architectural variables and the three psychological
dimensions were the same in the replication experiment as in the original study, although all
nine combinations of these correlations were statistically-significant in the replication study,

23
compared to only five of nine in the original study. The next section investigates the neural
correlates of the three latent psychological dimensions.

4. Experiment 3
In Experiment 3, we reanalyzed data from two previously published fMRI studies (Vartanian et
al., 2015, 2013) in relation to the principal components of architectural experience identified in
Experiments 1-2. In those previous studies, the authors investigated the effects of curvature
(Vartanian et al, 2013), ceiling height, and perceived enclosure (Vartanian et al., 2015) on
aesthetic judgements and neural activity. Participants were shown the images in the fMRI
scanner and were asked to make beauty judgements or approach-avoidance decisions in two
separate tasks (i.e., runs). The authors then identified neural regions that covaried with the
architectural variables (curvature, ceiling height, and enclosure) for each aesthetic judgment
task. The motivation behind those original studies was to identify regions of the brain that were
sensitive to variations in curvature, ceiling height, and enclosure in architecture. In addition, the
authors predicted that people’s responses to those features and the neural correlates that
accompany them would vary as a function of the task they were instructed to perform (i.e.,
beauty judgment vs. approach-avoidance decisions). For example, whereas people might find
curvilinear spaces beautiful, they may not necessarily indicate a willingness to enter them.
Indeed, the authors of those previous studies found that although participants judged spaces
that had high ceilings and were curvilinear as more beautiful, they did not demonstrate a
corresponding willingness to enter them. In turn, the authors found differences in the neural
correlates of curvature, ceiling height, and perceived enclosure as a function of the task under
consideration. Based on independent samples of architects and undergraduates and using a
subset of the same images, the same authors have since replicated the moderating effect of the
task (i.e., beauty judgment vs. approach-avoidance decisions) on ratings in a follow-up study
(Vartanian et al., 2019).

In the present investigation, we reanalyzed the neural data from the two previous studies
described above to address a different research question altogether. Specifically, we sought to
determine whether the three latent psychological dimensions (i.e., principal components)
identified in the preceding experiments would correlate with specific patterns of neural
activation. Embedded within that primary question, we also tested whether these correlational
patterns would vary as a function of task. Correlations between principal components and
neural activity would indicate the brain’s sensitivity to core aspects of our psychological
responses to architectural scenes. Because processing of the three latent dimensions under
consideration (i.e., coherence, hominess, and fascination) necessarily relies at least in part on
visual inspection of scenes, we suspected that we would observe sensitivity in relation to these
dimensions in dissociable regions within the visual cortex. Given that variations in the neural
correlates of architectural features (i.e., contour, ceiling height, and enclosure) as a function of
task had been observed in previous studies (Vartanian et al., 2013, 2015), we were interested in
determining whether similar task-based differences would be observed for coherence,
hominess, and fascination in the present context. Finally, since the results of Experiment 1
demonstrated that GIPs (i.e., complexity and self-similarity) predicted coherence, hominess, and
fascination scores, we also tested to see whether there would be any overlap between the
neural correlates of coherence, hominess, and fascination and those related to the GIPs under
consideration. If so, one might conclude that any covariation observed between coherence,
hominess, and fascination and neural activity might be in part driven by a shared neural
architecture that is also sensitive to variation in GIPs.

24
 4.1. Methods
To undertake this task, we re-analyzed data from Vartanian et al. (2015, 2013). In those studies,
healthy participants (n = 18, 12 females, 6 males, average age = 23.39 years, SD = 4.49) viewed
the same 200 photographs of architectural interiors used in Experiments 1 and 2 during a
functional MRI scan. Participants viewed the images under two different conditions
(administered as counterbalanced runs): in the beauty judgment condition, they were
presented with 100 stimuli, and on each trial indicated with a button press whether the image
was “beautiful” or “not beautiful.” In the approach-avoidance condition, they were presented
with the remaining 100 stimuli, and on each trial indicated with a button press whether they
would opt to “enter” or “exit” the space. The details of the neuroimaging acquisition
parameters are reported in S5 of Supplementary Materials (see also Vartanian et al., 2015,
2013).
The original analyses conducted by Vartanian et al. (2015, 2013) required the use of categorical
contrasts. Our focus here was different because we tested the hypothesis that there would be a
correlation between neural activation and variations in coherence, fascination, and hominess
scores associated with each image (derived from Experiment 2). To test this hypothesis, we
conducted parametric analyses of fMRI data using Statistical Parametric Mapping (SPM12)
software. Specifically, the presentation of each stimulus was treated as an event, coupled with
its coherence, fascination, hominess, self-similarity and complexity ratings associated with it as
the five parameters of interest. Specifically, for each of the 200 stimuli in the dataset, ratings of
coherence, fascination, and hominess were derived from the PCA of aesthetic response
measures (Experiment 2). In turn, for the same 200 stimuli, ratings of self-similarity and
complexity were derived from the analysis of GIPs (Experiment 1). These five ratings per
stimulus were entered as the five parameters in the parametric analysis. Because we were
interested in testing linear relationships, the parameters were entered as 1st order polynomial
expansions into the model. In turn, the motor response and the inter-stimulus-interval (ITI)
were entered into the analysis and treated as events of no interest (this enabled us to model
but remove brain activation associated with the presentation of the “+” in the ITI and motor
movement associated with a button press). Furthermore, because the participants had
completed the original study under two different task conditions, we conducted the
aforementioned parametric analysis separately for the beauty judgment and approach-
avoidance decision runs.
This analytic strategy enabled us to (a) determine whether the same brain regions would exhibit
sensitivity to variations in scores associated with coherence, fascination, and hominess under
different contexts (i.e., tasks), (b) conduct a conjunction analysis to see if there was a
statistically significant overlap between the regions associated with any of the three
components under both conditions, and (c) to identify any possible overlap between the neural
correlates of coherence, hominess, and fascination and the neural correlates of the two GIPs
under consideration (i.e., self-similarity and complexity). In terms of (a) our null hypothesis was
that no region would exhibit covariation with coherence, fascination, or hominess ratings when
the task involved beauty judgment or approach-avoidance decisions. In terms of (b) our null
hypothesis was that there would be no statistically significant overlap between the regions
associated with any of the three components under both conditions of beauty judgment and
approach-avoidance decisions. In terms of (c) our null hypothesis was that there would be no
statistically significant overlap between the regions that exhibit covariation with coherence,
hominess, or fascination and those that exhibit covariation with self-similarity or complexity. For
the reporting of our results, we adopted a combination of voxel-level and cluster-size correction

25
to control against false-positives. Specifically, using a random effects analysis, we report
activations that survived whole-brain voxel-level intensity threshold of p < 0.001 (uncorrected
for multiple comparisons), and a cluster-level correction of p < .05 (corrected for multiple
comparisons using whole-brain family-wise error). All brain regions are reported in relation to
the MNI (Montreal Neurological Institute) coordinate system.

4.2. Results
Fascination. The results demonstrated that in approach-avoidance decisions brain activation in
the left lingual gyrus covaried with fascination scores, whereas in beauty judgments brain
activation in the right lingual gyrus covaried with fascination scores. Note that under both
approach-avoidance and beauty judgment conditions, the clusters of activation were bilateral. A
follow-up conjunction analysis demonstrated that brain activation encompassing the right
lingual gyrus covaried with fascination scores under both beauty judgment and approach-
avoidance conditions (Table 4 and Figure 10a). These clusters were also bilateral.
Hominess. The results demonstrated that when making approach-avoidance decisions, brain
activation encompassing the left cuneus covaried with hominess scores. Note that the clusters
of activation were bilateral (Table 4 and Figure 10b). In turn, when making beauty judgments,
brain activation did not covary with hominess scores. Here no conjunction analysis was
conducted because in the case of beauty judgments no statistically significant effect was
discovered.
Coherence. The results demonstrated that when making approach-avoidance decisions, brain
activation did not covary with coherence scores. By contrast, when making beauty judgments,
brain activation encompassing the left inferior occipital gyrus covaried with coherence scores
(Table 4 and Figure 10c). The extent of this cluster was also bilateral. Here no conjunction
analysis was conducted because in the case of approach-avoidance decisions no statistically
significant effect was discovered.
GIPs. The results demonstrated that when making beauty judgments or approach-avoidance
decisions, brain activation did not covary with self-similarity or complexity scores.

Table 4: The neuroanatomical correlates of Fascination, Hominess, and Coherence. (BJ = beauty
judgment; AA = approach-avoidance decisions; CJN = conjunction analysis).

Dimension Task Structure Coordinates (x, y, z) T-score Cluster size (KE)

Fascination BJ Lingual gyrus 26, 80, 6 6.70 3154
AA Lingual gyrus 14, 100, 8 9.04 6145
CJN Lingual gyrus 26, 80, 8 6.65 2819

Hominess AA Cuneus 10, 92, 4 5.45 2040

Coherence BJ Occipital gyrus 34, 88,  5.08 332

26
27
Figure 10. Brain regions where activation covaried in relation to (a) Fascination, (b) Hominess, and (c)
Coherence. Regions of the brain that exhibited covariation under the beauty judgment condition appear in
red, whereas regions of the brain that exhibited covariation under the approach-avoidance condition
appear in green. Regions of the brain that exhibited covariation under both conditions (conjunction analysis)
appear in yellow. The mosaic slices shown in the left column were selected at the following z coordinates: -
50, -27, -4, 18, 41, and 64. Images were generated using MRIcroGL. For illustration purposes, the images
depicted in MRIcroGL show activations that extend beyond those that survived our whole-brain family-wise
correction for multiple comparisons reported in the manuscript.

4.3. Summary
Parametric analyses demonstrated that, regardless of the task, the degree of fascination
covaried with neural activity in the right lingual gyrus. In contrast, coherence covaried with
neural activity in the left inferior occipital gyrus only when participants judged beauty, and
hominess covaried with neural activity in the left cuneus only when they made approach-
avoidance decisions. Importantly, neural activation in the aforementioned regions did not
covary in relation to GIPs including self-similarity and complexity scores. Our results suggest that
the valuation of architectural processing in the visual cortices varies by dimension, as well as by
task in the case of coherence and hominess dimensions. These imaging results build on our
previous behavioral results by demonstrating that the brain exhibits sensitivity to the three
dissociable psychological dimensions of architectural experience identified in Experiments 1 and
2. Furthermore, the brain’s sensitivity to these dimensions may vary by function of task.

5. General Discussion
5.1. Summary of results
In three experiments, we set out to identify key psychological dimensions that are important
aspects of architectural experience. We further investigated how these dimensions relate to
brain activity. Specifically, we tested three hypotheses: 1) aesthetic responses to architectural
scenes can be reduced to a few latent psychological dimensions, 2) these dimensions are
sensitive to design variables of ceiling height, enclosure, and curvature, and 3) each
psychological dimension evokes a distinct neural response in the brain.
We used two complementary data-driven approaches to test the first hypothesis: PNA and PCA.
PNA enabled us to map out relationships between 16 aesthetic measures in Experiment 1 as
nodes within an “aesthetic network.” This analysis revealed that the aesthetic measures
clustered into three distinct communities. We further analyzed the relationships between
aesthetic rating scales using PCA in Experiments 1 and 2 and found that three principal
components – coherence, hominess, and fascination – explained most of the variance in ratings
across two independent samples.1 These components closely resembled the communities from
the PNA (see Figure 6). We interpreted these three communities/components as representing
latent psychological dimensions in response to architectural scenes. ANOVAs were also
conducted in Experiments 1 and 2, revealing that the three psychological dimensions were
sensitive to salient architectural features of ceiling height, enclosure, and curvature. In
Experiment 1, these dimensions were also found to be sensitive to two GIPs of the architectural
images: self-similarity and complexity.

1
The replication of this factor structure in two independent samples is notable given that people generally exhibit
less agreement when making aesthetic judgments of architectural stimuli compared to natural scenes (Vessel,
Maurer, Denker, & Starr, 2018).

28
In Experiment 3, parametric and conjunction analyses of fMRI data revealed that fascination
scores covaried systematically with neural activity in the right lingual gyrus regardless of
whether participants were engaged in beauty judgments or approach-avoidance decisions. In
contrast, coherence scores covaried with neural activity in the left inferior occipital gyrus only
when participants were judging beauty, whereas hominess scores covaried with neural activity
in the left cuneus only when they made approach-avoidance decisions. Together, these results
suggest that aesthetic responses to architectural scenes are explained by a few psychological
dimensions that are associated with distinct neural signatures in the brain. Importantly, we
were able to show that neural activation in the aforementioned regions did not covary in
relation to GIPs including self-similarity and complexity scores. In other words, the observed
patterns of neural activity are more likely to be driven by psychological responses that are not
mediated directly by psychophysical properties of the images.

5.2. Psychological dimensions of architectural experience

The analyses in Experiments 1 and 2 yielded three latent psychological dimensions underlying
aesthetic ratings of architectural scenes: Coherence, Hominess, and Fascination. Coherence
accounted for 40% and 25% of the variance in image ratings for Experiments 1 and 2,
respectively. Organization, beauty, valence, and approachability loaded on this component in
both studies.2 The close relationship between organization and these three global response
measures is consistent with fluency theory, which argues that ordered arrangements of a
scene’s composition – including structural redundancy, balance, and symmetry – heighten
aesthetic appeal by increasing the efficiency, or fluency, of information processing in the visual
system (Arnheim, 1971; D. J. Graham & Redies, 2010; Oppenheimer & Frank, 2008; Palmer,
Schloss, & Sammartino, 2013; Ramachandran & Hirstein, 1999; Reber et al., 2004). Previous
empirical work has indeed demonstrated that order and related constructs are reliable
predictors of aesthetic responses to visual art (Birkhoff, 1933; Eysenck, 1957; Oppenheimer &
Frank, 2008; Palmer et al., 2013) and landscapes (R. Kaplan, 1973; R. Kaplan & Kaplan, 1989; S.
Kaplan, 1987). Environmental disorder, by contrast, is linked to heightened anxiety (Tullett et
al., 2015), increased rule-breaking behavior (Kotabe et al., 2016b), reduced cognitive
performance (Evans, Gonnella, Marcynyszyn, Gentile, & Salpekar, 2005), and a diminished sense
of meaning in life (Heintzelman & King, 2014). Building on these past findings, our results
suggested that the coherence component was primarily driven by the perception of organization
but also involved multiple domains of psychological processing, including cognitive, affective,
and behavioral responses to architectural scenes.
The hominess component explained 27% and 30% of the variance in image ratings for
Experiments 1 and 2, respectively. In both studies, three psychological measures converged on
this component: personalness, hominess, and naturalness. These aesthetic measures relate to
the Danish concept of hygge, which describes “a feeling of coziness, warmth, and togetherness”
(Wiking, 2017, p. 25) that is often felt in the presence of intimate spaces and social settings.3
Environments that generate this mood generally feel “personal and authentic” (Linnet, 2012, p.
403) and “echo the feeling of home” (Wiking, 2017, p. 24). Hygge relates closely to the concept
of wholeness, which Alexander described as a spatial quality that makes occupants feel more

2
Four of the excluded measures – vitality, uplift, comfort, and relaxation – proved to be nearly redundant measures
of valence in Experiment 1 and were therefore most closely associated with this first principal component.
3
Although this concept has received particular emphasis in Danish culture, hygge has close translations in many
languages, including the Canadian hominess, the Dutch Gezelligheid, the Norwegian koselig, and the German
gemütlichkeit (Linnet, 2012; Wiking, 2017)

29
intimately connected to their surroundings and more liberated to express their authentic
personalities (Alexander, 1977, 1979, 2002a). Linnet described a similar phenomenon of
“rooting,” or increased connectedness, that occurs in the presence of hygge (Linnet, 2012, p.
407). Like wholeness, hygge has both social and spatial connotations. Spaces that create
hyggelig4 atmospheres often feel “organic” and “not strongly controlled” (Linnet, 2012, p. 405),
qualities that align with the measure of naturalness in our study. Wholeness has similarly been
linked to naturalistic visual patterns in architecture (Alexander, 2002a) and to loose, organic
construction processes (Alexander, 2002b). Thus, the experience of hominess may depend on
interactions between sensory inputs (i.e., naturalistic stimuli) and affective processing
mechanisms (i.e., feelings of belonging).
The third component, fascination, explained 24% and 20% of the variance in image ratings in
Experiments 1 and 2, respectively. In both studies, this component represented the vector sum
of two variables, complexity and interest. In Experiment 1, explorability and stimulation also
exhibited such high bivariate correlations with interest that they were considered redundant
variables. The close relationships that emerged between these four measures are consistent
with previous research. Interest ratings of visual art have been shown to correlate closely with
stimulus complexity (Daniel E. Berlyne, 1971; Silvia, 2005, 2012). Complexity has also been
found to predict stimulation responses to both art and architectural images (Daniel E. Berlyne,
1970, 1971; Heath, Smith, & Lim, 2000; Taylor et al., 2005). In response to the widespread
proliferation of minimalism in post-war Western architecture, several architectural theorists
emphasized the importance of visual complexity and ornament for generating interest and
excitement in the built environment (Alexander, 2002a; Salingaros, 2007; Venturi et al., 1977).
Kaplan and Kaplan (1989) proposed that complex landscapes provide a richness of information
that triggers visual interest and motivates exploration. Early studies in empirical aesthetics also
revealed close associations among these four response measures (D. E. Berlyne, 1963; Day,
1967). Here, we extend these past findings to the built environment by reporting that
complexity, interest, stimulation, and exploration all loaded on one multi-modal dimension of
architectural experience.
The three-part factor structure that emerged from our studies on architectural interiors is
reminiscent of the pivotal psychological framework that Kaplan and Kaplan (1989) proposed for
outdoor environments. Their seminal “preference matrix” outlined two psychological
dimensions that contribute to aesthetic experiences of outdoor landscapes: understanding and
exploration. Understanding describes “the need to make sense of what is going on” (R. Kaplan &
Kaplan, 1989, p. 51) in a landscape and is influenced by environmental features such as
“coherence” (how ordered a scene looks) and “legibility” (how easily a scene can be recognized,
interpreted, and remembered). This psychological dimension aligns closely with the coherence
component of our study, which describes how easily information in an architectural scene can
be processed. The Kaplans’ exploration dimension encompasses the human desire to “find out
more about what is going on in one’s surroundings” (R. Kaplan & Kaplan, 1989, p. 51).
Environmental features that stimulate exploration include complexity (the informational
richness of a scene) and mystery (the promise of hidden information waiting to be revealed).
This dimension echoes the component we described as fascination, a term that S. Kaplan later

4
Translation: “hygge-like” (Wiking, 2017)

30
adopted in his research on Attention Restoration Theory (Berman, Jonides, & Kaplan, 2008; S.
Kaplan, 1995; S. Kaplan & Berman, 2010).5
Intriguingly, the Kaplans’ framework for landscape aesthetics offers no equivalent to our
hominess component, suggesting that this dimension of psychological experience may be
specifically relevant to architectural interiors. Perhaps owing to the widespread influence of
Kaplans’ work, psychological measures related to coherence (e.g., fluency, order) and
fascination (e.g., complexity, interest) have been widely studied in environmental psychology
and empirical aesthetics research. Hominess and related constructs (e.g., personalness,
coziness) have received less attention.
Our results suggest that the experiences of coherence, hominess, and fascination all depend on
multiple domains of psychological processing, indicating that the most salient psychological
experiences in the built environment are likely generated by the integration of cognitive,
emotional, and sensory information. Furthermore, in both experiments, beauty, valence, and
approachability loaded moderately on all three principal components. This finding suggests that
the most global measures of architectural experience (how beautiful a room looks, for instance)
may be influenced by all three of these underlying psychological constructs. The near
orthogonality of order and complexity in the two PCA studies also supports previous theoretical
claims that order and complexity are consistently perceived as independent dimensions of the
physical environment (Alexander, 2002a; R. Kaplan & Kaplan, 1989; Salingaros, 2007). This
finding suggests that order and complexity are perceptually salient qualities of the built
environment that can be manipulated independently in architectural design strategies.
It is important to note that spatial enclosure had the strongest impact on psychological
responses. In both experiments, open spaces received significantly higher scores than enclosed
spaces on all three principal components, thus replicating past findings that open environments
are often perceived as more beautiful (Vartanian et al., 2015), safer (Stamps, 2005; Fich et al.,
2014), and more likely to stimulate movement and exploration (Vartanian et al., 2015). These
results also support Appleton’s theory that humans prefer environments with greater
affordances of visual prospect (Appleton, 1975) and Hildebrand’s hypothesis that our evolved
landscape preferences extend to the built environment (Hildebrand, 1999; Vartanian et al.,
2013). Furthermore, our results suggest that previously reported aesthetic preferences for high
ceilings and curved interiors may be driven by sensory experiences related to visual interest,
simulation, and exploration. These hypotheses are consistent with past fMRI findings that high
ceilings and curved spaces differentially activate neural structural associated with visuospatial
exploration and attention (Vartanian et al., 2015).
Our analyses also yielded unexpected results. We were surprised to find that open spaces and
high ceilings were associated with higher hominess scores. Since this psychological construct is
typically associated with feelings of “cozy interiority” and with spaces that create “a strong
sense of being inside” (Linnet, 2012), we expected rooms with low ceilings to be linked with this
component. However, many environmental variables contribute to a hyggeligt ambiance,
including lighting, surface textures, color, and furniture arrangement (Linnet, 2012; Wiking,
2017). Since we did not control for these other variables when selecting our stimuli, it is possible
that they affect ratings above and beyond the effects of enclosure and ceiling height. It is also

5
According to Attention Restoration Theory (ART), environments that are inherently fascinating are restorative,
because they capture involuntary attention in an automatic, bottom-up fashion and allow directed attention
mechanisms, which are controlled in top-down fashion, a chance to replenish (Berman, Jonides, & Kaplan, 2008; S.
Kaplan, 1995).

31
possible that the low ceilings in our stimuli conveyed a sense of confinement and claustrophobia
rather than coziness.
Finally, it is revealing that participants’ responses along the three psychological components
proved to be sensitive to two GIPs of architectural images, self-similarity and complexity. The
largest effect sizes were found for fascination component. More self-similar and more complex
scenes were perceived as significantly more fascinating than less self-similar and less complex
scenes. Interestingly, the fascination component exhibited the largest effect sizes in response to
the qualitative architectural variables of Ceiling Height, Curvature, and Enclosure, as well as for
the quantitative GIPs. These results suggest that the specific visual features we measured in this
study may modulate psychological responses along the fascination dimension more than they
modulate responses along the other two dimensions. It remains to be seen whether perceptions
of fascination are also highly sensitive to other architectural variables and visual patterns
beyond those measured in this study. We were also surprised to find that the self-similarity
measure correlated negatively with the coherence dimension, as we expected that more self-
similar scenes would be perceived as more organized and coherent than less self-similar scenes.
One possible explanation is that participants perceived images with strong focus points as more
coherent, as these images are generally less self-similar.

5.2. Neural responses to architectural scenes

Our imaging results demonstrated that various regions in the visual cortex are differentially
sensitive to core dimensions of our psychological responses to architectural scenes. The fact
that in all cases neural activity was observed in the visual cortex is not surprising, given that
processing the three latent dimensions under consideration (i.e., coherence, hominess, and
fascination) necessarily relies at least in part on visual inspection of scenes. Interestingly, the
degree of fascination drove neural activity in the right lingual gyrus regardless of whether
participants were engaged in beauty judgments or approach-avoidance decisions. In contrast,
the degree of coherence evoked neural activity in the left inferior occipital gyrus only when
participants judged beauty, whereas the degree of hominess evoked neural activity in the left
cuneus only when they made approach-avoidance decisions.
We underscore two important points about these findings. It has been known that parts of
visual cortex evaluates objects in addition to classifying them into such categories as faces,
places, and objects (Chatterjee, Thomas, Smith, & Aguirre, 2009). Our findings suggest that this
kind of neural response is likely a top-down influence on object processing that segregates along
psychological dimensions. The second point is about the stability of these neural responses.
Fascination evokes relatively stable responses, given that the associated neural activity was not
affected by the behavioral task in which people were engaged. By contrast, coherence only
modulated visual responsiveness when people made beauty judgments, whereas hominess only
modulated visual responsiveness when people made approach-avoidance decisions. Our data
do not allow us to infer why the effect of fascination was stable across tasks, whereas coherence
and hominess exerted different degrees of influence on neural processing in the context of
beauty judgments vs. approach-avoidance decisions. In addition, despite the fact that all the
regions identified in the present study reside within the visual cortex (Wandell, Dumoulin, &
Brewer, 2009), we suspect that these neural signatures represent top-down influences on visual
valuation rather than bottom-up psychophysical properties of the visual images.
One final point is worth emphasizing. Our psychological components were derived from
responses by participants in the U.S.. The neural data were derived from participants in the
Canary Islands a few years earlier, indicating considerable generalization of our claims. At the

32
time of fMRI data collection, we were ignorant of the psychological components now used to
model the data. Yet, these responses were present in the brains of our participants while being
hidden from us because we did not know to ask the question of the relationship between
fascination, hominess, and coherence and neural responses.

5.3. Limitations
We used images of architectural interiors as our stimuli in order to limit our focus to visual
perception of interior spaces and to expose participants to a wide variety of architectural spaces
within a reasonable timeframe. However, we relied on two-dimensional stimuli, which limits the
generalizability of our findings to three-dimensional built spaces. Future research could leverage
more immersive technologies like virtual reality to answer similar questions using more lifelike
simulations of architectural environments (see Banaei, Hatami, Yazdanfar, & Gramann, 2017;
Shemesh et al., 2017). We also chose to focus our study on visual perception of architecture. In
doing so, we are agnostic about the contribution of nonvisual senses to architectural
experiences. Finally, our studies considered three basic architectural design variables, which
together capture a limited proportion of a building’s visual properties. Indeed, it is likely that a
more complete understanding of the impact of architectural design variables on human thinking
and behavior will require examining more design variables and psychological outcome measures
of interest.

5.4. Conclusions
Here, we investigated the primary psychological dimensions of architectural experience. In a
pair of studies, we found and replicated the observation that three latent psychological
constructs – coherence, fascination, and hominess – collectively explained most of the variance
across a range of aesthetic response measures. The first two components align closely with the
psychological dimensions outlined in the Kaplans’ pivotal “preference matrix” of landscape
aesthetics (R. Kaplan & Kaplan, 1989). Indeed, coherence and fascination are well-established
dimensions in assessing natural scenes and visual art. Hominess, however, emerged as a new
dimension in relation to architectural interiors that has received scant attention to date in
empirical research. In our third study we found that variations in the three latent psychological
constructs were associated with brain activation in dissociable regions within the visual cortex.
These results provide new insights on how architectural design influences subjective human
experiences and reveal that the visual cortex is sensitive to specific psychological valuations in
our encounters with architectural interiors.
These findings have several practical implications for architectural design. First, it would be
useful for architects to test the psychological impact of proposed design schemes (before they
are constructed) along the dimensions of the three principal components identified in these
studies (coherence, fascination, and hominess). These dimensions offer a specific framework for
incorporating behavioral feedback into design iterations before a building is constructed.
Secondly, these components could be used for post-occupancy evaluations of buildings.
Architects and researchers could test occupant responses along these components (e.g. by
having occupants rate a space along each of the three psychological dimensions) and use this
behavioral feedback to guide future decisions related to interior design and construction.
Finally, architects might weigh these components differently, depending on the kind of building
being designed. The optimal balance of these components for a home, hospital, library, or a
museum might be different. More generally, the identification of these three psychological

33
components and their neural signatures advance our understanding of how people experience
interior spaces. This has far-reaching implications for architectural design and research alike.

34
 Acknowledgements & further information
This study was supported by generous funding from the Smith Family Fund and the Cambridge
Trusts as well as grants from Comisión Nacional de Investigación Científica y Tecnológica
(CONICYT/FONDECYT Regular N° 1171035).
These sponsors were not involved in study design, data collection, analysis, or interpretation.
They played no role in writing the manuscript or in deciding where to submit the manuscript.
We thank Jenna Crocker and Jane Cai for preparing scripts in Excel used to analyze fMRI data in
Experiment 3. We also acknowledge Cristián Modroño’s contributions to the protocol design of
the original fMRI data collection, data which were reanalyzed in this paper.
No part of the study procedures or analysis plans was pre-registered prior to the research being
conducted. We report how we determined our sample size, all data exclusions, all
inclusion/exclusion criteria, whether inclusion/exclusion criteria were established prior to data
analysis, all manipulations, and all measures in the study. The data, R script, and stimuli for this
research have been uploaded to Open Science Framework. All files for the study can be
accessed using the following link:
https://osf.io/mc3aj/?view_only=e2d881780fcb4e66ab2d210dc5d5e0ec

35
 References
Acking, C. A., & Kuller, R. (1973). Presentation and Judgement of Planned Environment and the
Hypothesis of Arousal. Environmental Design Research, 1.

Africa, J., Heerwagen, J., Loftness, V., & Ryan Balagtas, C. (2019). Biophilic design and climate
change: performance parameters for health. Frontiers in Built Environment, 5.

Alexander, C. (1977). A pattern language: towns, buildings, construction. New York: Oxford
University Press.

Alexander, C. (1979). The timeless way of building. New York: Oxford University Press.

Alexander, C. (2002a). The phenomenon of life: an essay on the art of building and the nature of
the universe (Vol. 1). Berkeley, Calif: Center for Environmental Structure.

Alexander, C. (2002b). The process of creating life: an essay on the art of building and the nature
of the universe (Vol. 2). Berkeley, CA: Center for Environmental Structure.

Appleton, J. (1975). The experience of landscape. London: John Wiley & Sons. Retrieved from
http://www.openbibart.fr/item/display/10068/937492

Arnheim, R. (1971). Entropy and Art. Berkeley: University of California Press.

Bachelard, G. (1994). The Poetics of Space. Beacon Press.

Baird, J. C., Cassidy, B., & Kurr, J. (1978). Room preference as a function of architectural features
and user activities. Journal of Applied Psychology, 63, 719–727.

Baker, N. V., & Standeven, M. A. (1995). A behavioural approach to thermal comfort assessment
in naturally ventilated buildings. In Proceedings CIBSE National Conference. Eastbourne.

Banaei, M., Hatami, J., Yazdanfar, A., & Gramann, K. (2017). Walking through Architectural
Spaces: The Impact of Interior Forms on Human Brain Dynamics. Frontiers in Human
Neuroscience, 11. doi:10.3389/fnhum.2017.00477

Bar, M., & Neta, M. (2006). Humans Prefer Curved Visual Objects. Psychological Science, 17,
645–648.

Baronchelli, A., Ferrer-i-Cancho, R., Pastor-Satorras, R., Chater, N., & Christiansen, M. H. (2013).
Networks in cognitive science. Trends in Cognitive Sciences, 17, 348–360.

Baum, A., & Davis, G. E. (1980). Reducing the stress of high-density living: An architectural
intervention. Journal of Personality and Social Psychology, 38, 471.

36
Berlyne, D. E. (1963). Complexity and incongruity variables as determinants of exploratory
choice and evaluative ratings. Canadian Journal of Psychology; Old Chelsea, Quebec, 17,
274–290.

Berlyne, Daniel E. (1970). Novelty, complexity, and hedonic value. Perception & Psychophysics,
8, 279–286.

Berlyne, Daniel E. (1971). Aesthetics and psychobiology. New York: Appleton-Century-Crofts.

Berlyne, Daniel E. (1974). Studies in the new experimental aesthetics: Steps toward an objective
psychology of aesthetic appreciation. Hemisphere.

Berman, M. G., Hout, M. C., Kardan, O., Hunter, M. R., Yourganov, G., Henderson, J. M., …
Jonides, J. (2014). The Perception of Naturalness Correlates with Low-Level Visual
Features of Environmental Scenes. PLOS ONE, 9, e114572.

Berman, M. G., Jonides, J., & Kaplan, S. (2008). The Cognitive Benefits of Interacting with
Nature. Psychological Science, 19, 1207–1212.

Berman, M. G., Kross, E., Krpan, K. M., Askren, M. K., Burson, A., Deldin, P. J., … Jonides, J.
(2012). Interacting with nature improves cognition and affect for individuals with
depression. Journal of Affective Disorders, 140, 300–305.

Berto, R. (2005). Exposure to restorative environments helps restore attentional capacity.

Journal of Environmental Psychology, 25, 249–259.

Birkhoff, G. D. (1933). Aesthetic Measure. Cambridge, MA: Harvard University Press.

Borsboom, D., & Cramer, A. O. (2013). Network analysis: an integrative approach to the
structure of psychopathology. Annual Review of Clinical Psychology, 9, 91–121.

Bosch, A., Zisserman, A., & Munoz, X. (2007). Representing shape with a spatial pyramid kernel.
Paper presented at the Proceedings of the 6th ACM international conference on Image
and video retrieval.

Brager, G., Paliaga, G., & De Dear, R. (2004). Operable windows, personal control and occupant
comfort. ASHRAE Transactions, 110, 17–35.

Bratman, G. N., Daily, G. C., Levy, B. J., & Gross, J. J. (2015). The benefits of nature experience:
Improved affect and cognition. Landscape and Urban Planning, 138, 41–50.

Braun, J., Amirshahi, S. A., Denzler, J., & Redies, C. (2013). Statistical image properties of print
advertisements, visual artworks and images of architecture. Frontiers in Psychology, 4.
https://doi.org/10.3389/fpsyg.2013.00808.

37
Chatterjee, A. (2013). The aesthetic brain: How we evolved to desire beauty and enjoy art.
Oxford University Press.

Chatterjee, A., Thomas, A., Smith, S. E., & Aguirre, G. K. (2009). The neural response to facial
attractiveness. Neuropsychology, 23, 135.

Chatterjee, A., & Vartanian, O. (2014). Neuroaesthetics. Trends in Cognitive Sciences, 18, 370–
375.

Choo, H., Nasar, J. L., Nikrahei, B., & Walther, D. B. (2017). Neural codes of seeing architectural
styles. Scientific Reports, 7, 1–8.

Christensen, A. P., Kenett, Y. N., Aste, T., Silvia, P. J., & Kwapil, T. R. (2018). Network structure of
the Wisconsin Schizotypy Scales–Short Forms: Examining psychometric network filtering
approaches. Behavior Research Methods, 1–20.

Coburn, A., Vartanian, O., & Chatterjee, A. (2017). Buildings, Beauty, and the Brain: A
Neuroscience of Architectural Experience. Journal of Cognitive Neuroscience, 29, 1521–
1531.

Cooper, R., & Burton, E. (2014). Wellbeing and the environmental implications for design. In
Wellbeing and the Environment (Vol. 2, pp. 653–668). West Sussex, England: John Wiley
& Sons Inc.

Cooper, R., Burton, E., & Cooper, C. L. (Eds.). (2014). Wellbeing and the Environment (Vol. 2).
West Sussex, England: John Wiley & Sons Inc.

Costantini, G., Richetin, J., Preti, E., Casini, E., Epskamp, S., & Perugini, M. (2017). Stability and
variability of personality networks. A tutorial on recent developments in network
psychometrics. Personality and Individual Differences.

Daniels, I. (2015). Feeling at home in contemporary Japan: Space, atmosphere and intimacy.
Emotion, Space and Society, 15, 47–55.

Day, H. (1967). Evaluations of subjective complexity, pleasingness and interestingness for a

series of random polygons varying in complexity. Perception & Psychophysics, 2, 281–
286.

Dazkir, S. S., & Read, M. A. (2012). Furniture Forms and Their Influence on Our Emotional
Responses Toward Interior Environments. Environment and Behavior, 44, 722–732.

De Deyne, S., Kenett, Y. N., Anaki, D., Faust, M., & Navarro, D. J. (2016). Large-scale network
representations of semantics in the mental lexicon. Big Data in Cognitive Science: From
Methods to Insights, 174–202.

38
Di Dio, C., Macaluso, E., & Rizzolatti, G. (2007). The golden beauty: brain response to classical
and renaissance sculptures. PloS One, 2, e1201.

Dosen, A. S., & Ostwald, M. J. (2016). Evidence for prospect-refuge theory: a meta-analysis of
the findings of environmental preference research. City, Territory and Architecture, 3, 1–
14.

Ellard, C. (2015). Places of the heart: the psychogeography of everyday life. New York: Bellevue
Literary Press.

Epskamp, S., Maris, G. K., Waldorp, L. J., & Borsboom, D. (2016). Network psychometrics. ArXiv
Preprint ArXiv:1609.02818.

Evans, G. W. (2003). The built environment and mental health. Journal of Urban Health : Bulletin
of the New York Academy of Medicine, 80, 536.

Evans, G. W., Gonnella, C., Marcynyszyn, L. A., Gentile, L., & Salpekar, N. (2005). The Role of
Chaos in Poverty and Children’s Socioemotional Adjustment. Psychological Science, 16,
560–565.

Evans, G. W., & McCoy, J. M. (1998). When buildings don’t work: The role of architecture in
human health. Journal of Environmental Psychology, 18, 85–94.

Eysenck, H. J. (1957). The Dynamics of Anxiety and Hysteria: An Experimental Application of

Modern Learning Theory to Psychiatry. London: Routledge & Kegan.

Fanger, P. O. (1973). Assessment of man’s thermal comfort in practice. British Journal of

Industrial Medicine, 30, 313–324.

Fich, L. B., Jönsson, P., Kirkegaard, P. H., Wallergård, M., Garde, A. H., & Hansen, Å. (2014). Can
architectural design alter the physiological reaction to psychosocial stress? A virtual TSST
experiment. Physiology & Behavior, 135, 91–97.

Field, A. P., Miles, J., & Field, Z. (2014). Discovering statistics using R. London: Sage.

Frith, C. D., & Nias, D. K. B. (1974). What determines aesthetic preferences. Journal of General
Psychology, 91, 163–173.

Graham, D. J., & Redies, C. (2010). Statistical regularities in art: Relations with visual coding and
perception. Vision Research, 50, 1503–1509.

Graham, L. T., Gosling, S. D., & Travis, C. K. (2015). The Psychology of Home Environments: A Call
for Research on Residential Space. Perspectives on Psychological Science, 10, 346–356.

39
Güçlütürk, Y., Jacobs, R. H. A. H., & van Lier, R. (2016). Liking versus Complexity: Decomposing
the Inverted U-curve. Frontiers in Human Neuroscience, 10, 1–11.

Heath, T., Smith, S. G., & Lim, B. (2000). Tall Buildings and the Urban Skyline: The Effect of Visual
Complexity on Preferences. Environment and Behavior, 32, 541–556.

Heidegger, M. (2013). Building Dwelling Thinking. In Poetry, Language, Thought (Later Printing
Used edition, pp. 141–160). New York: Harper Perennial Modern Classics.

Heintzelman, S. J., & King, L. A. (2014). (The Feeling of) Meaning-as-Information. Personality and
Social Psychology Review, 18, 153–167.

Hevner, K. (1935). Experimental studies of the affective value of colors and lines. Journal of
Applied Psychology, 19, 385–398.

Hildebrand, G. (1999). Origins of architectural pleasure. University of California Press. Retrieved

from
https://books.google.com/books?hl=en&lr=&id=R6MpMs3pz0oC&oi=fnd&pg=PP19&dq
=origins+of+architectural+pleasure&ots=-
o0ZnBfB7K&sig=esP8vPrILC6CDLjhsm7Ykz2WBdA

Imamoglu, Ç. (2000). Complexity, Liking and Familiarity: Architecture and Non-Architecture

Turkish Students’assessments of Traditional and Modern House Facades. Journal of
Environmental Psychology, 20, 5–16.

Imamoglu, V. (1979). Assessment of Living Rooms by Householders and Architects.

MyScienceWork. Retrieved from
https://www.mysciencework.com/publication/show/7c1facf371096225dcdb79857d7aa
80a

Ishizu, T., & Zeki, S. (2011). Toward A Brain-Based Theory of Beauty. PLOS ONE, 6, e21852.

Isvoranu, A.-M., Borsboom, D., van Os, J., & Guloksuz, S. (2016). A network approach to
environmental impact in psychotic disorder: brief theoretical framework. Schizophrenia
Bulletin, 42, 870–873.

Isvoranu, A.-M., van Borkulo, C. D., Boyette, L.-L., Wigman, J. T., Vinkers, C. H., Borsboom, D., &
Investigators, G. (2016). A network approach to psychosis: pathways between childhood
trauma and psychotic symptoms. Schizophrenia Bulletin, 43, 187–196.

Izard, C. E., Kagan, J., & Zajonc, R. B. (1988). Emotions, Cognition, and Behavior. Cambridge, UK:
Cambridge University Press.

Joye, Y. (2007). Architectural lessons from environmental psychology: The case of biophilic
architecture. Review of General Psychology, 11, 305–328.

40
Joye, Y., & Dewitte, S. (2016). Up speeds you down. Awe-evoking monumental buildings trigger
behavioral and perceived freezing. Journal of Environmental Psychology, 47, 112–125.

Kanner, A. D., Coyne, J. C., Schaefer, C., & Lazarus, R. S. (1981). Comparison of two modes of
stress measurement: Daily hassles and uplifts versus major life events. Journal of
Behavioral Medicine, 4, 1–39.

Kaplan, R. (1973). Predictors of environmental preference: designers and “clients.” In

Environmental Design Research. Dowden, Hutchinson & Ross.

Kaplan, R., & Kaplan, S. (1989). The experience of nature - a psychological perspective.

Kaplan, S. (1987). Aesthetics, Affect, and Cognition: Environmental Preference from an

Evolutionary Perspective. Environment and Behavior, 19, 3–32.

Kaplan, S. (1995). The restorative benefits of nature: Toward an integrative framework. Journal
of Environmental Psychology, 15, 169–182.

Kaplan, S., & Berman, M. G. (2010). Directed Attention as a Common Resource for Executive
Functioning and Self-Regulation. Perspectives on Psychological Science, 5, 43–57.

Kaplan, S., Kaplan, R., & Wendt, J. S. (1972). Rated preference and complexity for natural and
urban visual material. Perception & Psychophysics, 12, 354–356.

Kardan, O., Demiralp, E., Hout, M. C., Hunter, M. R., Karimi, H., Hanayik, T., … Berman, M. G.
(2015). Is the preference of natural versus man-made scenes driven by bottom–up
processing of the visual features of nature? Frontiers in Psychology, 6.

Kardan, O., Gozdyra, P., Misic, B., Moola, F., Palmer, L. J., Paus, T., & Berman, M. G. (2015).
Neighborhood greenspace and health in a large urban center. Scientific Reports, 5,
11610.

Karuza, E. A., Thompson-Schill, S. L., & Bassett, D. S. (2016). Local patterns to global
architectures: influences of network topology on human learning. Trends in Cognitive
Sciences, 20, 629–640.

Kellert, S. R. (2003). Kinship to mastery: Biophilia in human evolution and development. Island
Press. Retrieved from
https://books.google.com/books?hl=en&lr=&id=PkufpPphlz8C&oi=fnd&pg=PP15&dq=ki
nship+to+master+biophilia+in+human+evolution&ots=f6ZlC-
VdnC&sig=3SfZaMmDbhuArghgLp9y9iqZ3Gc

Kenett, Y. N., Kenett, D. Y., Ben-Jacob, E., & Faust, M. (2011). Global and local features of
semantic networks: Evidence from the Hebrew mental lexicon. PloS One, 6, e23912.

41
Kinchla, R. A. (1977). The role of structural redundancy in the perception of visual targets.
Perception & Psychophysics, 22, 19–30.

Kirk, U., Skov, M., Christensen, M. S., & Nygaard, N. (2009). Brain correlates of aesthetic
expertise: a parametric fMRI study. Brain and Cognition, 69, 306–315.

Kotabe, H. P. (2016). Disorder, Naturalness, and Their Influence on Aesthetics and Behavior.
University of Chicago, Chicago. Retrieved from
https://knowledge.uchicago.edu/handle/11417/214

Kotabe, H. P., Kardan, O., & Berman, M. G. (2016a). Can the high-level semantics of a scene be
preserved in the low-level visual features of that scene? A study of disorder and
naturalness. In Proceedings of the 38th Annual Meeting of the Cognitive Science Society
(Vol. 38, pp. 1721–1726).

Kotabe, H. P., Kardan, O., & Berman, M. G. (2016b). The order of disorder: Deconstructing visual
disorder and its effect on rule-breaking. Journal of Experimental Psychology: General,
145, 1713–1727.

Kotabe, H. P., Kardan, O., & Berman, M. G. (2017). The Nature-Disorder Paradox.

Kroll, L. (1987). An Architecture of Complexity. Cambridge, MA: MIT Press.

Küller, R., & Lindsten, C. (1992). Health and behavior of children in classrooms with and without
windows. Journal of Environmental Psychology, 12, 305–317.

Lawrence, M. A. (2016). ez: Easy Analysis and Visualization of Factorial Experiments. (Version
4.4-0). Retrieved from https://CRAN.R-project.org/package=ez

Leder, H., & Carbon, C.-C. (2005). Dimensions in appreciation of car interior design. Applied
Cognitive Psychology, 19, 603–618.

Leder, H., & Nadal, M. (2014). Ten years of a model of aesthetic appreciation and aesthetic
judgments : The aesthetic episode – Developments and challenges in empirical
aesthetics. British Journal of Psychology, 105, 443–464.

Leder, H., Oeberst, A., Augustin, D., & Belke, B. (2004). A model of aesthetic appreciation and
aesthetic judgments. British Journal of Psychology, 95, 489–508.

Lench, H. C., Darbor, K. E., & Berg, L. A. (2013). Functional Perspectives on Emotion, Behavior,
and Cognition. Behavioral Sciences, 3, 536–540.

42
Linnet, J. T. (2012). The social-material performance of cozy interiority (pp. 403–408). Presented
at the Ambiances in action / Ambiances en acte(s) - International Congress on
Ambiances, Montreal 2012, International Ambiances Network.

Lundholm, H. (1921). The Affective Tone of Lines: Experimental Researches. Psychological

Review, 28, 43–60.

Mak, M. Y., & Thomas Ng, S. (2005). The art and science of Feng Shui—a study on architects’
perception. Building and Environment, 40, 427–434.

Marchette, S. A., Vass, L. K., Ryan, J., & Epstein, R. A. (2015). Outside Looking In: Landmark
Generalization in the Human Navigational System. Journal of Neuroscience, 35, 14896–
14908.

Mehta, R., & Zhu, R. (Juliet). (2009). Blue or Red? Exploring the Effect of Color on Cognitive Task
Performances. Science, 323, 1226–1229.

Meyers-Levy, J., & Zhu, R. (2007). The Influence of Ceiling Height: The Effect of Priming on the
Type of Processing That People Use. Journal of Consumer Research, 34, 174–186.

Monsell, S. (2003). Task switching. Trends in Cognitive Sciences, 7, 134–140.

Mullin, Caitlin & Hayn-Leichsenring, Gregor & Wagemans, Johan. (2015). There is beauty in
gist: An investigation of aesthetic perception in rapidly presented scenes. Journal of
Vision. 15. 123. 10.1167/15.12.123.

Nadal, M., Munar, E., Marty, G., & Cela-Conde, C. J. (2010a). Visual complexity and beauty
appreciation: Explaining the divergence of results. Empirical Studies of the Arts, 28, 173–
191.

Nadal, M., Munar, E., Marty, G., & Cela-Conde, C. J. (2010b). Visual Complexity and Beauty
Appreciation: Explaining the Divergence of Results. Empirical Studies of the Arts, 28,
173–191.

Nicol, J. F., & Humphreys, M. A. (2002). Adaptive thermal comfort and sustainable thermal
standards for buildings. Energy & Buildings, 34, 563–572.

Nix, G. A., Ryan, R. M., Manly, J. B., & Deci, E. L. (1999). Revitalization through Self-Regulation:
The Effects of Autonomous and Controlled Motivation on Happiness and Vitality. Journal
of Experimental Social Psychology, 35, 266–284.

Oppenheimer, D. M., & Frank, M. C. (2008). A rose in any other font would not smell as sweet:
Effects of perceptual fluency on categorization. Cognition, 106, 1178–1194.

43
Palmer, S. E., Schloss, K. B., & Sammartino, J. (2013). Visual Aesthetics and Human Preference.
Annual Review of Psychology, 64, 77–107.

Patra, R. (2009). Vaastu Shastra: Towards sustainable development. Sustainable Development,

17, 244–256.

Poffenberger, A. T., & Barrows, B. E. (1924). The Feeling Value of Lines. Journal of Applied
Psychology, 8, 187–205.

R Core Team. (2016). R: A language and environment for statistical computing. Vienna, Austria:
R Foundation for Statistical Computing. Retrieved from https://www.R-project.org/

Ramachandran, V. S., & Hirstein, W. (1999). The science of art: A neurological theory of
aesthetic experience. Journal of Consciousness Studies, 6, 15–51.

Reber, R., Schwarz, N., & Winkielman, P. (2004). Processing Fluency and Aesthetic Pleasure: Is
Beauty in the Perceiver’s Processing Experience? Personality and Social Psychology
Review, 8, 364–382.

Redies C., Amirshahi S.A., Koch M., Denzler J. (2012) PHOG-Derived Aesthetic Measures
Applied to Color Photographs of Artworks, Natural Scenes and Objects. In: Fusiello
A., Murino V., Cucchiara R. (eds) Computer Vision – ECCV 2012. Workshops and
Demonstrations. Lecture Notes in Computer Science, vol 7583. Springer, Berlin,
Heidelberg.

Revelle, W. (2016). psych: Procedures for Personality and Psychological Research (Version
1.6.12). Evanston, IL: Northwestern University. Retrieved from https://CRAN.R-
project.org/package=psych

Ritterfeld, U., & Cupchik, G. C. (1996). Perceptions of interior spaces. Journal of Environmental
Psychology, 16, 349–360.

Robinson, S., & Pallasmaa, J. (Eds.). (2015). Mind in Architecture: Neuroscience, Embodiment,
and the Future of Design. Cambridge, Massachusetts: The MIT Press.

Ryan, C. O., and Browning, W. D. (2018). “Biophilic design,” Encyclopedia of Sustainability

Science and Technology, ed R. Meyers. New York, NY: Springer.

Ryan, C. O., Browning, W. D., Clancy, J. O., Andrews, S. L., and Kallianpurkar, N. B. (2014).
Biophilic design patterns: emerging nature-based parameters for health and well-being
in the built environment. Archnet International Journal of Architectural Research. 8, 62–
76.

44
Ryan, R. M., & Deci, E. L. (2008). From Ego Depletion to Vitality: Theory and Findings Concerning
the Facilitation of Energy Available to the Self. Social and Personality Psychology
Compass, 2, 702–717.

Ryan, R. M., & Frederick, C. (1997). On Energy, Personality, and Health: Subjective Vitality as a
Dynamic Reflection of Well-Being. Journal of Personality, 65, 529–565.

Ryan, R. M., Weinstein, N., Bernstein, J., & Brown, K. W. (2010). Vitalizing effects of being
outdoors and in nature. Journal of Environmental Psychology, 30, 159–168.

Salingaros, N. A. (1998). A scientific basis for creating architectural forms. Journal of

Architectural and Planning Research, 283–293.

Salingaros, N. A. (2007). A Theory of Architecture. Solingen: ISI Distributed Titles.

Salingaros, N. A. (2015). “Biophilia and Healing Environments: Healthy Principles For Designing
the Built World”. New York: Terrapin Bright Green, LLC.

Schmittmann, V. D., Cramer, A. O., Waldorp, L. J., Epskamp, S., Kievit, R. A., & Borsboom, D.
(2013). Deconstructing the construct: A network perspective on psychological
phenomena. New Ideas in Psychology, 31, 43–53.

Shannon, P., Markiel, A., Ozier, O., Baliga, N. S., Wang, J. T., Ramage, D., … Ideker, T. (2003).
Cytoscape: a software environment for integrated models of biomolecular interaction
networks. Genome Research, 13, 2498–2504.

Shemesh, A., Talmon, R., Karp, O., Amir, I., Bar, M., & Grobman, Y. J. (2017). Affective response
to architecture – investigating human reaction to spaces with different geometry.
Architectural Science Review, 60, 116–125.

Silvia, P. (2005). Cognitive Appraisals and Interest in Visual Art: Exploring an Appraisal Theory of
Aesthetic Emotions. Empirical Studies of the Arts, 23, 119–133.

Silvia, P. (2008). Interest—The Curious Emotion. Current Directions in Psychological Science, 17,
57–60.

Silvia, P. (2012). Human emotions and aesthetic experience: an overview of empirical aesthetics.
In Aesthetic Science: Connecting minds, brains, and experience (pp. 250–275). New York:
Oxford University Press.

Sommer, R. (1969). Personal Space: The Behavioral Basis of Design. Englewood Cliffs, N.J.:
Prentice Hall Trade.

45
Stamps, A. E. (2005). Visual permeability, locomotive permeability, safety, and enclosure.
Environment and Behavior, 37, 587–619.

Stamps, A. E. (2011). Effects of Area, Height, Elongation, and Color on Perceived Spaciousness.
Environment and Behavior, 43, 252–273.

Stangor, C. (2015). Principles of Social Psychology. Minneapolis: University of Minnesota.

Retrieved from http://open.lib.umn.edu/socialpsychology/

Tanizaki, J. (2001). In Praise of Shadows. Vintage Books.

Taylor, R. P., Micolich, A. P., & Jonas, D. (1999). Fractal analysis of Pollock’s drip paintings.
Nature, 399, 422–422.

Taylor, R. P., Spehar, B., Wise, J. A., Clifford, C. W., Newell, B. R., Hagerhall, C. M., … Martin, T. P.
(2005). Perceptual and physiological responses to the visual complexity of fractal
patterns. Nonlinear Dynamics Psychol. Life. Sci, 9, 89–114.

Thorsson, S., Honjo, T., Lindberg, F., Eliasson, I., & Lim, E.-M. (2007). Thermal comfort and
outdoor activity in Japanese urban public places. Environment and Behavior. Retrieved
from http://eab.sagepub.com/content/early/2007/07/13/0013916506294937.short

Tullett, A. M., Kay, A. C., & Inzlicht, M. (2015). Randomness increases self-reported anxiety and
neurophysiological correlates of performance monitoring. Social Cognitive and Affective
Neuroscience, 10, 628–635.

Tumminello, M., Aste, T., Di Matteo, T., & Mantegna, R. N. (2005). A tool for filtering
information in complex systems. Proceedings of the National Academy of Sciences, 102,
10421–10426.

Tyrväinen, L., Ojala, A., Korpela, K., Lanki, T., Tsunetsugu, Y., & Kagawa, T. (2014). The influence
of urban green environments on stress relief measures: A field experiment. Journal of
Environmental Psychology, 38, 1–9.

Ulrich, R. S. (1977). Visual landscape preference: A model and application. Man-Environment

Systems, 7, 279–293.

Ulrich, R. S. (1983). Aesthetic and affective response to natural environment. In Behavior and
the natural environment (pp. 85–125). Springer.

Ulrich, R. S. (1984). View through a window may influence recovery from surgery. Science, 224,
417–419.

46
Ulrich, R. S., Simons, R. F., Losito, B. D., Fiorito, E., Miles, M. A., & Zelson, M. (1991). Stress
recovery during exposure to natural and urban environments. Journal of Environmental
Psychology, 11, 201–230.

Van Oel, C. J., & van den Berkhof, F. D. (2013). Consumer preferences in the design of airport
passenger areas. Journal of Environmental Psychology, 36, 280–290.

Vartanian, O., Navarrete, G., Chatterjee, A., Fich, L. B., Gonzalez-Mora, J. L., Leder, H., … Skov,
M. (2015). Architectural design and the brain: Effects of ceiling height and perceived
enclosure on beauty judgments and approach-avoidance decisions. Journal of
Environmental Psychology, 41, 10–18.

Vartanian, O., Navarrete, G., Chatterjee, A., Fich, L. B., Leder, H., Modrono, C., … Skov, M.
(2013). Impact of contour on aesthetic judgments and approach-avoidance decisions in
architecture. Proceedings of the National Academy of Sciences, 110, 10446–10453.

Vartanian, O., Navarrete, G., Chatterjee, A., Fich, L. B., Leder, H., Modroño, C., Rostrup, N., Skov,
M., Corradi, G., & Nadal, M. (2019). Preference for curvilinear contour in interior
architectural spaces: Evidence from experts and nonexperts. Psychology of Aesthetics,
Creativity, and the Arts, 13(1), 110-116.

Venturi, R., Scully, V., & Drexler, A. (1977). Complexity and Contradiction in Architecture (2nd
edition). New York : Boston: The Museum of Modern Art, New York.

Vessel, E. A., Maurer, N., Denker, A. H., & Starr, G. G. (2018). Stronger shared taste for natural
aesthetic domains than for artifacts of human culture. Cognition, 179, 121–131.

Vitaliano, P. P., Scanlan, J. M., Ochs, H. D., & Syrjala, K. (1998). Psychosocial stress moderates
the relationship of cancer history with natural killer cell activity. ANNALS OF
BEHAVIORAL MEDICINE, 20, 199–208.

Vitruvius Pollio, M., Morgan, M. H., & Warren, H. L. (1914). Vitruvius: the ten books on
architecture. Cambridge, MA: Harvard University Press. Retrieved from
http://library.wur.nl/WebQuery/clc/128812

Wagenmakers, E.-J. (2016). JASP (Version 0.8 Beta 5). Amsterdam: University of Amsterdam.

Wandell, B. A., Dumoulin, S. O., & Brewer, A. A. (2009). Visual cortex in humans. Encyclopedia of
Neuroscience, 10, 251–257.

Wei, T., & Simko, V. (2016). corrplot: Visualization of a Correlation Matrix (Version 0.77).
Retrieved from https://github.com/taiyun/corrplot

Wiking, M. (2017). The Little Book of Hygge: Danish Secrets to Happy Living. William Morrow.

47
Wilson, E. O., & Kellert, S. R. (1995). The Biophilia Hypothesis. Island Press.

Zhang, J. W., Piff, P. K., Iyer, R., Koleva, S., & Keltner, D. (2014). An occasion for unselfing:
Beautiful nature leads to prosociality. Journal of Environmental Psychology, 37, 61–72.

48
 Supplementary materials
S1: Architectural images (low ceilings)
Closed Rooms Open Rooms
Curved Rooms (Curvilinear)
Square Rooms (Rectil
i near )

49
 S2: Architectural images (high ceilings)
Closed Rooms Open Rooms
Curved Rooms (Curvilinear)
Square Rooms (Rectil
i near )

50
 S3: Experiment 1 blocking scheme for image presentation
Summary of stimulus distribution within each image block. Architectural conditions are labeled as follows:
high ceilings (H), low ceilings (L), open (O), enclosed (C), square (S), round (R).

S4: Experiment 1 rating groups for aesthetic rating scales

S5: Experiment 3 neuroimaging parameters (see Vartanian et al., 2013).

A 3-Tesla MR scanner with an eight-channel head coil (Signa Excite HD, 16.0 software; General
Electric) was used to acquire T1 anatomical volume images (1.0 × 1.0 × 1.0-mm voxels). For
functional imaging, T2*-weighted gradient echo spiral-in/out acquisitions were used to produce
35 contiguous 4-mm-thick axial slices [repetition time (TR) = 2,000 ms; echo time (TE) = 21.4 ms;
flip angle (FA) = 90°; field of view (FOV) = 260 mm; 64 × 64 matrix; voxel dimensions = 4× 4 × 4.0
mm], positioned to cover the whole brain. The first 10 volumes were discarded to allow for T1
equilibration effects. The number of volumes acquired was 430 (+ 10 dummies).

51
Credit Author Statement:
Alexander Coburn: conceptualization, data curation, formal analysis, investigation,
methodology, project administration, visualization, original draft, review & editing.
Oshin Vartanian: conceptualization, data curation, formal analysis, investigation, methodology,
project administration, resources, supervision, visualization, original draft, review & editing.
Yoed Kenett: formal analysis, visualization, original draft, review & editing.
Marcos Nadal: investigation, methodology, project administration, supervision, validation,
review & editing.
Franziska Hartung: methodology, supervision, review & editing.
Gregor Hayn-Leichsenring: formal analysis, supervision, review & editing.
Gorka Navarrete: data curation, formal analysis, review & editing.
José Luis González Mora: data acquisition for Experiment 3.
Anjan Chatterjee: conceptualization, funding acquisition, investigation, methodology, project
administration, resources, supervision, original draft, review & editing.