Speed of Person Perception Affects Immediate and Ongoing Aesthetic

$FWD3V\FKRORJLFD ²

Contents lists available at ScienceDirect

Acta Psychologica

journal homepage: www.elsevier.com/locate/actpsy

Speed of person perception aﬀects immediate and ongoing aesthetic 7 evaluation

Mel W. Khawa,⁎, Phoebe Nicholsb, David Freedbergc a Center for Cognitive Neuroscience & Duke Institute for Brain Sciences, Duke University, USA b Department of Neuroscience & Visual Arts Department, Bowdoin College, USA c Department of Art History and Archaeology & Italian Academy for Advanced Studies, Columbia University, USA

ARTICLE INFO ABSTRACT

Keywords: Recent studies have shed light on how aesthetic judgments are formed following presentations lasting less than a Aesthetics second. Meanwhile, dedicated neural mechanisms are understood to enable the rapid detection of human faces, Empirical aesthetics bodies, and actions. On the basis of cognitive studies of: (i) the speed and acuity of person perception, and (ii) Neuroaesthetics preferential attention given to human imagery (e.g., faces and bodies), we hypothesize that the visual detection Person perception of humans in portraits increases the magnitude and stability (i.e., similarity to later responses) of aesthetic Face perception ratings. Ease of person perception is also expected to elicit longer durations of preferential viewing time, a surplus Attention measure of viewing behavior that should be positively related to subsequent ratings. To test these ideas, we use a set of cubist portraits previously established to be more or less categorizable in terms of the aggregate time required to perceive the depicted person. Using these images, we track aesthetic judgments made following short and unconstrained presentations; in an intervening task, we measure viewing behavior when subjects are able to selectively reveal regions of these images. We ﬁnd that highly categorizable artworks (those that require less time to identify the ﬁgure as human) elicit higher and more predictive aesthetic ratings following 30 ms presentations while also eliciting longer viewing durations. Changes in ratings throughout the task are positively correlated with cumulative viewing time; critically, an image's categorizability level further moderates the strength of this relationship. These results demonstrate that a particular kind of visual object recognition – the recognition of human forms – modulates aesthetic preferences at a glance, subsequent viewing patterns, as well as rating changes over time.

1. Introduction Hutzler, & Carbon, 2008). Earlier studies of the timecourse of aesthetic judgments focused on the collative properties of artworks and their re- Recent results in empirical aesthetics highlight the brief time lation to arousal levels and hedonic experiences (Bachmann & Vipper, window from which meaningful aesthetic judgments are declarable. For 1983; Berlyne, 1960). In one such study, verbal judgments of ‘com- example, judgments of beauty following glimpses of stimuli (e.g., fol- plexity’ and ‘orderliness’ are obtainable after 50 ms glances at pattern lowing presentations lasting 30–500 ms) are positively correlated with (synthetic) and painting (genuine) artistic images (Cupchik & Berlyne, analogous judgments following unrestricted viewings (Verhavert, 1979). The present study also focuses on aesthetic evaluations at short Wagemans, & Augustin, 2018). This type of investigation has also been timescales. In addition, we examine a specific form of visual object performed with more dynamic stimuli; accurate judgments of liking can recognition and its relevance toward quick aesthetic evaluations: the be made after listening to brief passages of music – durations as brief as visual detection of people. This form of visual perception, along with 500 ms suffice for select genres of music (Belfi, Kasdan, Rowland, downstream social impressions (of gender, race, personality, etc.), is Vessel, & Starr, 2018). Studies utilizing brief time windows have also studied in the wider cognitive sciences as person, people, or social focused on specific visual attributes relevant to aesthetic processing. perception (Phillips, Weisbuch, & Ambady, 2014; Rutherford & For instance, content-wise similarity in representational artworks can Kuhlmeier, 2013). Our investigation looks at ratings and viewing be- be discriminated upon following 10 ms views, while similarity in style havior elicited by a set of cubist portraits previously studied by Hekkert requires durations upward of 50 ms to disambiguate (Augustin, Leder, and van Wieringen (1990). Within this set of portraits, Hekkert and van

⁎ Corresponding author at: Duke Institute for Brain Sciences, Duke University, 308 Research Dr, Durham, NC 27710, USA. E-mail address: [email protected] (M.W. Khaw). https://doi.org/10.1016/j.actpsy.2019.05.006 Received 5 October 2018; Received in revised form 6 May 2019; Accepted 9 May 2019 $YDLODEOHRQOLQH-XQH (OVHYLHU%9$OOULJKWVUHVHUYHG M.W. Khaw, et al. $FWD3V\FKRORJLFD ²

Wieringen (1990) identified varying levels of image categorizability – Furthermore, inferences about a person's attractiveness and trust- this metric was based on the amount of time subjects required to worthiness (Willis & Todorov, 2006), as well as (above-chance) classi- identify the depicted human figure within each portrait. Based on fications of sexual orientation (Rule & Ambady, 2008; Rule, Ambady, & physiological and behavioral findings surrounding person perception, Hallett, 2009) can be elicited following presentation windows of 100 ms we predict that the speed of forming such percepts – as reflected by and below. Critically, for the study here, images of faces and people are known categorizability levels – positively affects the magnitude and also known to elicit preferential judgments, decisions, and attention. On stability of aesthetic ratings over time. In what follows, we briefly re- this issue, we restrict the focus of our study to person perception in- view: (i) relevant timing benchmarks during the extraction of visual dependent of varying levels of personal beauty or attractiveness. At this information; (ii) social judgments and attention modulation using basic level, perceptual realizations of faces embedded in abstract pat- human imagery; and, (iii) links between person perception and other terns have been shown to coincide with increases in aesthetic appre- theoretical constructs in aesthetic processing. ciation (Muth & Carbon, 2013). Furthermore, experiments in economic The general timecourse of visual information extraction contains decision-making demonstrate that people readily trade work effort and several relevant benchmarks for aesthetic processing. In agreement money for opportunities to look at faces (Hayden, Parikh, Deaner, & with rapid content-based judgments of artworks (Augustin et al., 2008), Platt, 2007). The inclination to perform such tradeoffs has since also categorical detection of faces or animals in natural scenes can be per- been demonstrated in monkeys, where viewing their conspecifics is formed well within 250 ms (Rousselet, Macé, & Fabre-Thorpe, 2003). “paid for” using consumables (Deaner, Khera, & Platt, 2005). In these The visual accumulation of such information resembles a continuous decision-making experiments, such tradeoffs are taken to suggest that process – using a series of precisely-timed masks, behavioral accuracy images of faces are associated with valuable information; such visual and associated brain activity both plateau following 40–60 ms of pre- stimuli might provide information about social hierarchies, member- mask exposure time (Bacon-Macé, Macé, Fabre-Thorpe, & Thorpe, ship, and attributes relevant to social interactions (Klein, Shepherd, & 2005). Indeed, semantic details regarding people and objects are re- Platt, 2009). Indeed, as highlighted previously, the perception of hu- portable from glances at complex natural scenes following views as mans is often accompanied by secondary inferences, such as immediate short as 27 ms (Fei-Fei, Iyer, Koch, & Perona, 2007). Despite the quick judgments of personality traits from faces (Todorov, Olivola, Dotsch, & flow of coarse object information (Riesenhuber & Poggio, 2000), more Mende-Siedlecki, 2015) and emotional affect from postures (de Gelder, nuanced behavioral judgments nonetheless vary depending on proces- De Borst, & Watson, 2015). Using minimalistic figures of human mo- sing type and the amount of processing that is allowed to take place. For tion, it has also been shown that socially-interacting figures are more example, color similarity judgments are driven primarily by retinal likely to be consciously perceived compared to isolated or scrambled color information after 200 ms exposure and declines with further ex- figures, when these “options” are presented to individual eyes (Su, van posure time (Schulz & Sanocki, 2003). Select forms of fundamental Boxtel, & Lu, 2016); the same type of human figures in motion triggers perceptual grouping are also more or less difficult to perform. The reflexive attentional orienting when used as task-irrelevant distractors grouping of small pattern elements and individuation of large ones (Shi, Weng, He, & Jiang, 2010). Person perception can thus be expected appear to be intrinsic and immediate, while the opposite ability de- to influence aesthetic evaluation by facilitating the visual transfer of velops between ages 5–10 (Kimchi, Hadad, Behrmann, & Palmer, social information – information potentially impacting aesthetic ap- 2005). Crucially, the integration of configural elements into im- praisal (this link is discussed in further detail below). In social neu- pressions of people is relatively quick and arguably innate. Sufficient roscience, humanized perception – the recognition of others' mental representations of people include 8–12 joints of a moving figure at lives – has recently been identified as a promoter of pro-social behavior, 150 ms of viewing time (Johansson, 1973), while specific face re- along with familiarity and similarity to the sense of self (see Harris and cognition takes place within 100–200 ms of exposure (Jacques & Fiske (2009) for a review). From this perspective, person perception in Rossion, 2006; Sugase, Yamane, Ueno, & Kawano, 1999). In addition, artworks might engender thoughts about those depicted, and their re- geometric shapes in face-like arrangements are capable of eliciting levance to the observer. Indeed, the assessment of self-relevancy has preferential gaze from newborn infants (Morton & Johnson, 1991). been postulated to be a central component of aesthetic processing The sub-second speed of recognizing faces and making social in- (Vessel, Starr, & Rubin, 2013). Given what might be a common visual ferences (Johnson, 2005) are putatively enabled by sub-cortical nuclei preference for humanized content, we furthermore hypothesize that such as the amygdala (Vuilleumier, Armony, Driver, & Dolan, 2001), in aesthetic ratings (and subsequent viewing time) will favor images with tandem with a cortical network centered at the fusiform gyrus easily-perceived human subjects, even following brief initial glances. In (Kanwisher, McDermott, & Chun, 1997; Kanwisher & Yovel, 2006). relation to existing theoretical frameworks in empirical aesthetics, the Visual recognition of human portrayals might also involve brain regions role of person perception/recognition ties into two major topics: (i) selectively dedicated to action recognition and observation (Cross, processing fluency and familiarity, as well as (ii) the elicitation of Hamilton, Kraemer, Kelley, & Grafton, 2009; Grafton, Arbib, Fadiga, & empathetic responses. “Fluent” stimuli – those that are more easily Rizzolatti, 1996; Rizzolatti, Fadiga, Gallese, & Fogassi, 1996). While the processed cognitively – are theorized to obtain greater degrees of aes- subliminal influence of faces on abstract art has been explored (Flexas thetic appreciation; theoretical mechanisms include positive affect et al., 2013), the influence of person perception per se, as well its effects arising from said processing (Reber, Schwarz, & Winkielman, 2004), as on aesthetic evaluations over time, are unknown. Seeing as minimal well as deviations from expectations (regarding fluency) resulting in visual detail and exposure time are required for the detection of faces pleasure and/or interest (Graf & Landwehr, 2015). At present, data- and bodies, time-constrained assessments of artworks should be parti- driven measures have been used to identify general determinants (e.g., cularly impacted by the emergence of such percepts. In particular, the visual contrast, symmetry, etc.) of fluency (Mayer & Landwehr, 2018). relative ease of detecting portrait subjects should strengthen potential In relation to our study, fluency is also driven by factors that aid in the correlations between aesthetic first impressions and evaluations made categorization of stimuli, such as the prototypicality of shapes following extensive viewing. (Winkielman, Halberstadt, Fazendeiro, & Catty, 2006), prior exposure Given the perceptual benchmarks above, the ability to produce to stimuli (Bornstein & D'Agostino, 1994), and matching perceptual spontaneous high-level judgments toward people might appear less primes (Reber, Winkielman, & Schwarz, 1998). In this respect, and as surprising. Locher, Unger, Sociedade, and Wahl (1993) show that the perceptual literature above makes clear, the detectability of people steoreotypic factors of attractiveness (e.g., masculinity, femininity) within portraits is predicted to increase an image's level of conceptual exert their influence on judgments following 100 ms-long presentations. or semantic fluency (Reber et al., 2004; Whittlesea, 1993). In our study, Bar, Neta, and Linz (2006) demonstrate that stable personality im- the presence of people within artworks is expected to imbue glances pressions are declarable within the first 40 ms of viewing faces. with readily interpretable information; faces and bodies likely facilitate

M.W. Khaw, et al. $FWD3V\FKRORJLFD ² social inferences more easily compared to other perceptual groupings gauge the processing weight, as indicated by viewing time, put on such as shapes, textures, and inanimate objects. Indeed, using a similar specific information (e.g., time and money amounts) when economic set of cubist paintings, the explicit recognition of depictions resulted in choices are made (Khaw, Li, & Woodford, 2018; Payne, 1976; Reeck, pupil dilation, with preferences given toward easily construed land- Wall, & Johnson, 2017). In this study, mouse-tracking allows us to track scapes and portraits (Kuchinke, Trapp, Jacobs, & Leder, 2009). Fur- viewing time preferentially distributed toward specific images, while thermore, individuals possess an inclination to actively associate geo- enabling us to subsequently visualize particular areas of interest within metric shapes with person-like characteristics – a bias in favor of these images. Finally, we utilize a conventional ratings task at the end perceiving animacy, originally reported by Heider and Simmel (1944). of the experiment where presentation durations were not limited. It should be noted, however, that the present study is by no means intended to reduce the merits of abstract, inanimate, or otherwise 2. Method challenging art; indeed, the process that leads to the developed liking of such art has recently been documented empirically (Belke, Leder, & 2.1. Observers Carbon, 2015). Nonetheless, we intuit that the clarity with which people are depicted likely aids the short-term assignment of aesthetic 20 adults (11 female) participated in all phases of the main study values, during which relative familiarity might encourage valuation consecutively. The average age of this sample was 22.53 (SD = 3.36). (Alter & Oppenheimer, 2008). Familiarity (having prior experience Participants were recruited from the Columbia University community with a particular stimuli) is itself a likely promoter of aesthetic en- using advertisements placed throughout the university's main campus. gagement in our task. This general prediction stems from the ‘mere Participants completed each task at an individual computer station in exposure’ effect (Zajonc, 1968) – where prior experience increases about 45 min. During post-task interviews, none of these 20 partici- aesthetic liking (Martindale, Moore, & West, 1988), possibly through pants reported having an a priori familiarity with the specific images the selective priming of different image attributes (Martindale & Moore, used, though several noted a familiarity with the cubist style more 1988). Finally, human portrayals also tie in with theories that empha- generally. All participants provided informed consent and were com- size the role of embodied imagery and empathetic responses in art (e.g. pensated with $15 for their participation. Crozier & Greenhalgh, 1992; Freedberg & Gallese, 2007; Lipps, 1935), Following a reviewer's suggestion, we recruited a further 17 adults though once again this does not exclude the possibility of projecting (9 female) from the Duke University community for a follow-up ex- human and embodied forms onto abstract shapes. To illustrate, the periment to validate the categorizability levels assigned by Hekkert and observation of actions in art elicits motor responses that have been van Wieringen (1990). The average age of this sample was 23.63 interpreted as empathetic reactions (e.g., Battaglia, Lisanby, & (SD = 3.46). These participants also provided informed consent before Freedberg, 2011). Such reflexive motoric reactions also arise from participating, and were compensated with $8 for their participation. viewing abstract visual information, such as textures and brush strokes None of these additional participants reported familiarity with the in artworks (Freedberg & Gallese, 2007; Sbriscia-Fioretti, Berchio, images that were presented to them during the task. Freedberg, Gallese, & Umilta, 2013). If the basis of empathetic responses lie in these patterns of matching neural responses (to some 2.2. Stimuli inferred activity or action), then human figures should represent a di- rect source of such action-relevant information. The neural intersection The stimuli from this experiment were cubist paintings originally between embodied responses and aesthetic appreciation has also been selected by Hekkert and van Wieringen (1990) in their study of pro- reviewed extensively by Ticini, Urgesi, and Calvo-Merino (2014) as totypicality. We replaced one painting that was ambiguously titled in well as by Kirsch, Urgesi, and Cross (2016). Overall, the relevance of the High Categorizability category, with the full list of works appearing human portrayal in theories are wide and varied, though the general in Table 1. Based on the average time it took for their subjects to find prediction would be that such content plays a significant positive role in the human figure within each painting, Hekkert and van Wieringen aesthetic engagement and the assignment of aesthetic value. (1990) divided the paintings into three categories: Low Categorizability In the present study, we examine the assignment of aesthetic value (LC), Intermediate Categorizability (IC), and High Categorizability on a set of known cubist portraits featuring differing levels of cate- (HC). In that study, these images were also rated on ‘beauty,’‘proto- gorizability. These artworks were originally used by Hekkert and van typicality,’ and ‘complexity’. The authors concluded that aesthetic Wieringen (1990), where the authors defined 3 levels of categoriz- preferences for abstract works are more influenced by complexity, ability (low, intermediate, and high) based on the time it took for their while seemingly representational works are more influenced by proto- participants to identify the human figure in each painting. We hy- typicality. pothesize that these pre-defined levels affect aesthetic evaluations fol- In order to preserve subjects' lack of familiarity with these images lowing short presentation durations as well as conventional, un- (in accordance with other studies of initial, first impressions-based, constrained presentations. These differences are expected to manifest in aesthetic ratings), we did not conduct a confirmatory pre-test of image terms of rating magnitudes, correlations between measures, as well as categorizability. Furthermore, a post-task assessment of categorizability rating changes as a function of viewing time. In addition to ratings, we was also not possible with these subjects as they would have learned the also predict that preferential viewing time – time freely dedicated to- relevant spatial locations within various images. Thus, we employ the ward images beyond initial glances – will be allocated favorably toward categorizability ranges defined by Hekkert and van Wieringen (1990) highly categorizable images. Such a prediction would be consistent which we later validate with an independent sample of subjects (see with the desirability of discovering faces (Muth & Carbon, 2013), and below). In their report, LC images required mean response times above viewing socially-relevant imagery (Hayden et al., 2007). Notably, this 7000 ms for the identification of the human figure. Similarly, IC images prediction lies in opposition to the cost of time incurred by effortful necessitated average response times between 3000 and 7000 ms; lastly, processing of artworks (e.g., Kuchinke et al., 2009; Reber et al., 2004), HC images were those that required less than an average of 3000 ms for whereby longer processing times are found to be negatively related to figure identification. preference ratings. To test these predictions, we employ a study with 3- In order to control for potential differences in low-level visual phases. First, we employ a short presentation ratings task akin to studies properties between categories, all images underwent a set of pre-pro- of the micro-genesis of aesthetic preferences (Augustin et al., 2008; cessing procedures. We equated images' histograms in terms of their Verhavert et al., 2018). Second, we use a mouse-tracking task where luminance, contrast, and average amplitude spectra using the SHINE subjects selectively reveal portions of images (using their mouse cursor) Toolbox (Willenbockel et al., 2010). The average response times using from different categories. This measure was used in previous studies to this set of normalized images, relative to the original set, as well as the

M.W. Khaw, et al. $FWD3V\FKRORJLFD ²

Table 1 cursor where they perceived the portrait subject immediately following List of paintings used for the study, originally selected by Hekkert and van their response. Wieringen (1990).

Category Artist Painting 2.3.2. Rating task In the first rating phase, subjects were presented with members of High categorizability Braque Standing Female Nude (1907 each stimulus category in randomized order. Subjects were instructed - 08) to rate on a continuous slider bar how pleasing they perceived the on- Capek Cubist figure (1913) Gleizes Portrait of Jacques Nayral screen stimuli, ranging from 0 (not pleasing) to 100 (most pleasing). In (1910 - 11) order to discourage rote memorization of prior ratings, ratings were Gleizes Man on a balcony (1912) displayed with a precision of two decimal places; in addition, subjects Gris Portrait of Picasso (1912) were not limited in the time available to declare their ratings. In the Gris Portrait of Josette (1916) fi Gris Harlequin with guitar (1917) rst version of the rating phase, presentation durations were con- Gris Harlequin seated beside a strained to 30 ms. This specific duration was chosen as it meets ex- table (1919) perimental benchmarks of aesthetic, social, and visual judgments. Larionov Woman walking on the Following the findings of Verhavert et al. (2018), reliable judgments of boulevard (1912) beauty (in terms of significant correlations with unconstrained ratings) Léger Woman sewing (1909) Metzinger Lady at her dressing table are obtainable from 30 ms onward, with the average correlation uti- (1916) lizing 40 ms appearing to be only marginally greater. Second, as de- Picasso Clovis Sagot (1909) scribed in more detail within the Introduction section, high-level social Picasso Nude woman in an armchair judgments of face stimuli such as of emotion (Morris, Öhman, & Dolan, (1909) Picasso Wilhelm Uhde (1910) 1998), sexual orientation (Rule et al., 2009; Rule & Ambady, 2008), and Picasso Seated female nude (1910) personality (Carney, Colvin, & Hall, 2007) are obtainable within the Picasso Harlequin (1918) 30–50 ms range. The 30 ms constraint thus represents a duration that Intermediate categorizability Braque Woman playing a mandolin allows us to replicate prior observations regarding aesthetic judgments, (1910) while being in a range that allows for the visual extraction of person- Braque Female figure (1910 - 11) Braque La musicienne (1913) related semantic associations. Subjects subsequently completed a Braque The musician (1917 - 18) viewing procedure (Spotlight task, described below). In the following Gleizes Dancer (1917) final rating phase, images were left on the screen for as long as subjects Léger Seated woman (1913) deliberated on their judgments. In both rating phases, each image was Léger Soldier smoking (1916) McDonald- Synchrony in purple (1917) presented 3 times for a total of 120 trials, comprising 40 unique images Wright (Table 1). Picasso Nude with draperies (1907) Picasso Portrait of D.H. Kahnweiler 2.3.3. Spotlight task Picasso Nude (1910) During the spotlight task, participants were presented with stimuli Villon Portrait of madame Y.D. (1913) only viewable through a small circular aperture. The position of the Low categorizability Braque Seated man with a guitar aperture is controlled by mouse movement in order to uncover specific (1911) portions of the stimuli. The aperture was a Gaussian transparency filter Duchamp Nude descending a staircase featuring a standard deviation of 33 pixels, centered on a square mask (1911 - 12) with a height and width of 200 pixels (Scarfe, 2015). Subjects were able Léger The typographer (1919) Mondrian Female figure (1912) to reveal any on-screen region of their choice through this “spotlight” Picasso Female nude (1910 - 11) for 5 s during each trial. Before subjects were allowed to maneuver the Picasso Clarinet player (1911) mouse, stimuli were presented briefly for 30 ms in their respective lo- Picasso Man with a pipe (1915) cations (one image at a time, in random order). All subjects completed Picasso Man with a mandolin (1911) Picasso Ma jolie (1911 - 12) 48 trials, comprising 3 repetitions of each complete set of images. Picasso Man with a violin (1912) Images in the IC and LC categories received 4 additional repetitions Picasso Man leaning on a table (1916) each, as their categories each included 4 fewer images than the HC set. Udalstova At the piano (1914) These additionally repeated images were randomly selected from each category so that 16 triplets were presented in each repetition. The spatial location of each category's images were randomized during each relative rankings of image discriminability are plotted in Fig. 2 and trial (one image from each category occupied either the middle, right, detailed in the Results section (Categorizability Speeds). or left on-screen position; Fig. 1B). During the entirety of the 5 s provided for each trial, the momentary locations of participant's mouse's 2.3. Procedure movements (in the form of {x,y} screen coordinates) were recorded. A circular timer above the screen indicated the portion of 5 s remaining 2.3.1. Categorizability task on each trial. Following a reviewer's suggestion, we sought to replicate the ro- bustness of categories defined by Hekkert and van Wieringen (1990) 3. Results with the digitally normalized images that were utilized in the main study. Toward this end, we employed an analogous recognition task 3.1. Categorizability speeds using an independent sample of participants. Subjects on this task were instructed to press the spacebar key (with their finger constantly posi- Based on a reviewer's suggestion, we ran a follow-up study to va- tioned on the keyboard) as soon as they perceived the depicted portrait lidate the categorical boundaries proposed by Hekkert and van subject. The onset of images was cued by a 3-second countdown which Wieringen (1990). We first confirm that the average times for subjects was always triggered by the subject (to prevent presentations before the to declare recognition (of a person in each portrait) are not significantly subject was ready). To further encourage subjects to produce mean- different between the previous and current study (Fig. 2A). Here we ingful responses, subjects were instructed to indicate with the mouse perform paired two-tailed t-tests between the previously reported

M.W. Khaw, et al. $FWD3V\FKRORJLFD ²

Fig. 1. (A) Participants first rated each image after they were presented individually for 30 ms. In the A B last phase of the study, each image was presented again for as long as subjects were producing their rating. Each rating phase involved 3 repeated presentations, with a shuffled sequence of images presented three times (i.e., the same image could not appear twice in a row). (B) In the spotlight task, subjects were presented with one randomly selected member of each image category briefly for 30 ms. The images were then obscured with an opaque mask – subjects were required to hover over the image with their mask to reveal portions of the image. The spotlight was a clear circular aperture centered at the current mouse coordinate.

response times and those measured in our replication sample. No sig- the original category assignments are still effective at eliciting broad nificant differences were found for HC (t(15) = −0.82, p = 0.42), IC (t differences in the speed of person perception. (11) = −0.32, p = 0.75), nor LC (t(11) = 0.11, p = 0.91) images. In fi ff order to con rm that these category assignments a ected the speed of 3.2. Ratings task(s) recognition on a within-subjects basis, we computed the subject-specific rankings that were assigned to each image based on their individual We first sought to test the hypothesis that HC (least abstract) images response times (i.e., ranks 1 through 40, comprising fastest to slowest elicited the greatest magnitude of ratings between subjects. First, a one- response for each subject). With these distributions of relative rankings, way between subjects ANOVA confirmed that ratings were differen- fi we can con rm – using two-sample t-tests – that HC images occupy tially distributed across the three image categories. higher rank values than the neighboring category of IC images (t This was true in both the 30 ms presentation condition (F (2, (474) = −16.09, p < 0.001) while IC images occupy higher ranks 2397) = 9.55, p < 0.001) and the subsequent unlimited duration than LC images (t(406) = −5.99, p < 0.001). The distribution of these condition (F (2, 2397) = 13.47, p < 0.001). Additional post-hoc two- fi ranks and their interquartile ranges can be seen in Fig. 2B. As a nal sample t-tests confirm the predicted directionality of this ordering comparison to the response times reported by Hekkert and van (Fig. 3A). Across both viewing conditions, HC images received higher Wieringen (1990), we compare the distribution of our replication ratings than IC images (t(1678) = 2.15, p < 0.05 for the 30 ms view- samples' response times to their reported means for each image (Fig. ings; t(1678) = 2.65, p < 0.01 for unconstrained viewings). Mean- fi S2A, B, C). We nd that 34/40 of their reported averages fell within the while, IC images received greater ratings than LC images (t 95% C.I. bounds of our measured distributions of response times. The (1438) = 2.06, p < 0.05 for the 30 ms viewings; t(1438) = 2.37, averages that fell outside of these bounds were associated with 3 images p < 0.05 for unconstrained viewings). Interestingly, we also observe from the HC category, 2 from the IC category, and 1 LC image. In the that ratings decrease between each mode of presentation across all absence of further data from the original experiment by Hekkert and three categories – an observation that is covered in more detail in the van Wieringen (1990), we are unable to speculate further on these in- analysis below, on viewing times and subsequent changes in ratings. dividual image discrepancies. We are nonetheless able to conclude that The differences between categories were observed despite the fact

10 Fig. 2. (A) The average response times for the re- A Hekkert & van Wierengen data cognition of depicted portrait subjects are similar 8 between our replication sample and the averages reported by Hekkert and van Wieringen (1990). ﬁ 6 Dashed lines denote the category boundaries de ned in the original study. Error bars denote the asso- 4 ciated standard error of each mean. (B) By computing the relative rank for each image within sub- ﬁ 2 jects, we furthermore con rm that HC images are Response Time (s) generally fastest at being recognized (often being the 0 top 1–10 fastest images of being recognized), fol- CH CI CL lowed by IC and LC images, occupying the middle Image Category and lower ranges respectively.

High Cat. Int Cat. Low Cat. 25 25 25 B Median I.Q.R. 20 20 20

15 15 15

10 10 10

Percent Observed 5 5 5

0 0 0 10 20 30 40 10 20 30 40 10 20 30 40 Relative Speed Rank (1 - fastest; 40 - slowest)

M.W. Khaw, et al. $FWD3V\FKRORJLFD ²

Fig. 3. (A) Ratings favor the high-categorizability images for both unconstrained and constrained ratings task; this is despite heterogeneous preferences regarding the ranking of each particular image. (B) The between subjects agreement in within-category agreement is practically zero for HC and IC images, suggesting for individual-specific tastes for these images. There is a modest yet statistically-significant level of agreement for LC images. All error bars denote the standard error surrounding each mean. (C) Ratings produced following rapid presentations were significantly correlated to those under unlimited viewing times, replicating previous experiments; high categorizability images are associated with the greatest average correlation, indicating that preferences are more reliably formed at short durations for such images.

that the preference orderings for each image were highly specific to predictive ratings following brief presentations. each individual. To confirm this, we computed the average Spearman rank-order correlation between each subject and the remaining subjects 3.3. Spotlight task in the sample. The Spearman rank-order correlation was used here as a measure of agreement in the ordering of each category's images. We first confirmed that the selective revelation of individual images Preference ranks were used instead of ratings in order to account for was concentrated at known regions of interest, as demonstrated in si- subject-specific baseline ratings and rating differences between ranks. milar eye-tracking exercises in the past (e.g., Yarbus, 1967). As has Each image's rank was determined based on the ordering of the mean been previously shown, the areas of images that elicited greater viewing rating provided to each image (recall that each image was presented 3 densities were generally the face and bodies of the persons being por- times each during each phase). The distribution of these rank-order trayed (Figs. 3 and S1, S2, S3). In order to visualize the density of correlations (comprising 19 observations per subject) was not sig- mouse-coordinates that were associated with each image, we summed nificantly different from zero for either the HC (t(379) = −0.88, the overall density of mouse-coordinates recorded from each subject in p = 0.38 for 30 ms views; t(379) = −0.41, p = 0.68 for unlimited individual pixel-sized bins (belonging to each image's presentation duration views) or IC images (t(379) = 1.59, p = 0.11 for 30 ms views; frame size of 333 × 572 pixels). The resulting density counts were t(379) = 1.58, p = 0.11 for unlimited duration views), supporting the smoothed using a Gaussian kernel (a convolution between the density notion of individual-specific preference orderings. For the LC images, counts and a square Gaussian filter matrix of 20-pixels, featuring a there was a statistically-significant, albeit modest level of agreement standard deviation of 5 pixels). The resulting heatmaps are plotted for between subjects in both presentation conditions. The average visualization purposes in Fig. 4, with the full range of images available Spearman's correlation ranged between 0.06 and 0.12; Fig. 3B(t in Supplemental Figs. 1–3. The heatmaps are overlaid above an edge- (379) = 3.88, p < 0.001 for the 30 ms condition while t(379) = 8.39, detected version of each image, in order to best visualize the regions p < 0.001 for the unlimited duration condition). that correspond to areas of greater viewing density. Next, we sought to replicate the previously reported correlations in The first quantitative hypothesis we sought to test using the spot- studies of briefly presented artworks (Verhavert et al., 2018). We do light paradigm was whether HC images elicited greater amounts of this by computing Pearson correlations between each subjects' 30-ms relative attention between categories; we measure this in a non-para- and unlimited viewing judgments (for each image). We replicate the metric manner using the overall density of recorded mouse coordinates, range of correlations previously reported – with the greatest aggregate collected within each frame. Here we are interested in testing the correlations being observed in the HC category of images (Fig. 3C). specific hypothesis that HC images are preferentially viewed over both Each distribution of correlation coefficients were significantly greater IC and LC images. Again, we began with a one-way between subjects than zero (t(19) = 10.82, p < 0.001; t(19) = 5.29, p < 0.01; and t ANOVA performed on the average time dedicated to each image cate- (19) = 2.99, p < 0.01 for HC, IC, and LC images respectively). A one- gory across trials (Fig. 5A). There is a significant difference between way between subjects ANOVA furthermore confirms that the means of mean viewing times across categories (F (2, 2877) = 13.46, these correlations between groups were not equal (F (2, 57) = 4.41, p < 0.001). Subsequent post-hoc two-sample t-tests confirm that HC p < 0.05). Further post-hoc two-sample t-tests indicate that the cor- images receive greater viewing time compared to IC images (t relations in the HC category were greater than those in the LC category (1918) = 4.02, p < 0.001); while IC images receive more viewing time (t(38) = 3.17, p < 0.001), though there was not a significant differ- than LC images (t(1918) = 4.80, p < 0.001). ence between HC correlations and its neighboring IC category (t Given that there is consistently one depicted person per image, we (38) = 1.72, p = 0.09, in a two-tailed test). Overall, all three image also examined whether the above viewing trend is robust over time. In sub-types replicate significant rating correlations following brief other words, do subjects keep choosing to view HC images even after viewing durations; in addition, the average correlations obtained fol- repeated viewings? To answer this question, we compare the fraction lows the predicted ordering, with HC images eliciting the most highly- viewing time between the first and last blocks of trials (each comprising

M.W. Khaw, et al. $FWD3V\FKRORJLFD ²

Fig. 4. Visualizations of filtered mouse-tracking density data gathered from example (edge detection-filtered) images belonging to HC, IC, or LC categories (pictured respectively from left to right). Attention appears distributed focally toward known regions of interest, e.g., facial features of portrayed subjects. presentations of the complete image sets). We find no statistically-sig- changes? To inspect this, we computed the average rating change for nificant differences between blocks for either HC (t(651) = 1.54, images according to their cumulative total viewing times (Fig. 6). Since p = 0.12) or LC images (t(651) = −1.43, p = 0.15), categories where the total viewing time is unique to each trial, subject, and image, we such effects are most likely to occur. We note however, the apparent binned this variable into bins of equal 0.25 s widths; this interval re- trend in which viewing time lowers for HC images and increases for LC presents increases of 5% of the total possible viewing time. We then images. In order to visualize and report this potentially emerging effect, computed the average change in ratings associated with these cumu- we computed the fraction viewing time over each trial and averaged lative viewing time bins. Consistent with the directionality of a positive this metric across individual trials (Fig. 5B). familiarity bias (wherein greater amounts of viewing time should correspond to greater degrees of appreciation), we find a significant po- 3.4. Viewing time and subsequent changes in ratings sitive correlation between total viewing time spent on an image, and the magnitude of the subsequent rating change by the end of the task Finally, we sought to explain the variance in ratings observed be- (r = 0.60, n = 60, p < 0.001). We also hypothesize that HC images tween the beginning of the study (with the 30 ms ratings task) and the will be more adept at fostering familiarity-based increases in ratings. end of the study (with the unlimited viewing task), as a function of We find a supportive trend for this, with HC viewing time exhibiting the intervening viewing time. We first observe above that ratings generally strongest positive association with subsequent increases in ratings, decline by the end of the study, regardless of the image's category followed by IC and LC viewing time respectively (Fig. 6). When testing (Fig. 3A). Given that the main intervening experiences of the subject for this trend at the individual level, we find that the correlation (between the two rating blocks) lie in the spotlight task, we ask: what is coefficients for HC and IC images (between unique cumulative viewing the relation between cumulative viewing time, and subsequent rating times, and their ensuing changes in ratings) were not significantly

Fig. 5. (A) Viewing time in the spotlight task favors A images belonging to the high categorizability cate- 500 1st Block gory, followed by images classiﬁed as medium and 2nd Block low categorizability. (B) The proportion of viewing 3rd Block time dedicated to each category is not statistically- 450 diﬀerent over blocks of trials, but does appear to trend in the following manner: LC images increase in 400 viewing time, followed by a concomitant decrease in viewing time for HC images. Continuous lines in grey depict the averages of binned trials in 10 trial bins. 350 Total Viewing Time (s) giH i hg .tnI L wo Categorizability High Cat. Int. Cat. Low Cat. B 0.6 0.6 0.6 0.5 0.5 0.5

0.4 0.4 0.4

0.3 0.3 0.3

0.2 0.2 0.2

0.1 0.1 0.1 Fraction Viewing Time Fraction Viewing Time Fraction Viewing Time 0 0 0 050 050 050 Trials Elapsed Trials Elapsed Trials Elapsed

M.W. Khaw, et al. $FWD3V\FKRORJLFD ²

High Cat. Int. Cat. Low Cat. 80 80 80 Avg. Viewing Time 60 60 60

40 40 40

20 20 20

0 0 0 Rating Change Rating Change Rating Change -20 -20 -20

-40 r = 0.67 -40 r = 0.44 -40 r = 0.29 p = 0.00 p = 0.00 p = 0.04 -60 -60 -60 0 10 20 30 0 10 20 30 0 10 20 30 Total Viewing Time (s) Total Viewing Time (s) Total Viewing Time (s)

Fig. 6. Subsequent changes in the evaluations of each image are positively associated with the amount of viewing time spent in the intervening spotlight task. In addition to the finding that viewing time is generally associated with an increase in aggregate ratings, this relation is strongest for stimuli in the high categorizability category, and weaker as categorizability decreases. different (t(38) = −1.12, p = 0.27); however, IC correlations were of real-world vs. fractal images in Vessel and Rubin (2010), and of significantly greater than those measured over LC images (t(38) = 2.66, various abstract vs. representational art in Schepman et al. (2015). p < 0.05). Departing from distinct categories of image content, our study focuses on the level of abstraction specifically within depictions of people. As in the case of the results from Schepman et al. (2015), we observe that 4. Discussion perceived representativeness is preferred over abstraction (Fig. 3A). In terms of agreement between subjects, Vessel and Rubin (2010) and We firstly observe that despite heterogeneous tastes for specific Schepman et al. (2015) also found that their respective real world and images, highly categorizable portraits were generally rated as more representational images elicited greater degrees of inter-observer visually pleasing. This is true following first-time views at short pre- agreement. Despite this, in the present study, the greatest (albeit sentation durations, as well as by the end of the task when unlimited modest) level of rankwise agreement was found between ostensibly viewing time was allowed. We also find that viewing time is pre- abstract (LC) images (Fig. 3B). One interpretation of this pattern is that ferentially devoted toward highly categorizable images. In terms of the perceived realism of human figures facilitates individual-specific rating changes over time, we find that ratings generally decrease by the appraisals of facial or personal attractiveness (Germine et al., 2015; end of the task, an effect common to all categories of images. However, Hönekopp, 2006; Rhodes, 2006). Notably, Leder, Goller, Rigotti, and as we predicted, preferential viewing time is positively associated with Forster (2016) using the method of Hönekopp (2006) find that 40% of subsequent changes in unconstrained ratings. Critically, the relation face attractiveness variability can be ascribed to private (individual- between ratings and viewing time also appears to depend on categor- specific) taste, relative to 75% for abstract art. On this topic, our study izability: highly categorizable images elicit stronger associations be- raises a new observation regarding abstract faces/bodies, for which tween viewing time and rating changes. Taken together, we find that there might be an increased level of shared beauty standards; though, the relative ease of person perception changes how aesthetic evaluation further study is warranted to overcome the small number of images begins and proceeds. These differences manifest within first im- (12–16) used within each category here. Vessel and Rubin (2010) no- pressions, viewing time allocation, and subsequent rating changes. In tably suggested that collectively shared semantic associations lead to relation to existing frameworks in empirical aesthetics, these three greater levels of aesthetic agreement. The relatively low degree of types of effects can be seen to correspond to different stages of aesthetic agreement for HC images suggest that easily-perceived people does not processing (Chatterjee, 2003; Leder, Belke, Oeberst, & Augustin, 2004). necessarily elicit a high level of inter-rater agreement. This might be Firstly, brief duration or single-glance preferences are likely driven by due to the rich set of social inferences – e.g., of emotions, identity, race, coarse categorical information obtainable during sub-second exposures, etc. – that can form downstream of person perception. Future exten- thus favoring highly categorizable images. The brief (30 ms) time sions of the approach we set out here could involve testing abstract vs. window from which this effect occurs suggests that visual object re- realistic depictions of varying subject matter, e.g., of inanimate objects. cognition – not necessarily arising from artworks – highly influences In addition, the boost in aesthetic appreciation when humans are easily aesthetic judgments made at these timescales. Further work should detectable ought to be investigated in other types of visual stimuli. examine whether other types of semantic information (e.g., of animate Previously studied categories of images that might be susceptible to the and inanimate objects) can influence early aesthetic judgments. Sec- effects we observe here include natural scene photographs with and ondly, the spotlight task demonstrates that viewing time is continuously without people (e.g., Fei-Fei et al., 2007), as well as abstract patterns and preferentially given to highly categorizable portraits, supporting embedded with human-like details (e.g., Muth & Carbon, 2013). the view that personified content is desirable beyond immediate object Our results validate short presentation durations and apparent le- recognition (Crozier & Greenhalgh, 1992). Finally, the effects of vels of abstraction as comparable constraints on aesthetic processing. viewing time seen here underscores a difference between processing We observe that the range of intra-subject correlations (between 30 ms viewing time (negatively related to preferences) and preferential or and unconstrained viewings) is similar to those reported in Verhavert surplus viewing time (positively related to preferences), discussed fur- et al. (2018). Interestingly, our HC images exhibit an average correla- ther below. tion coefficient comparable to their 100 ms and unconstrained view- Our results also broaden the range of previously described effects ings. By the same token, correlation coefficients measured using IC and regarding realism and aesthetic valuation (Schepman, Rodway, Pullen, LC images might be seen as comparable to their 50 ms and 10 ms & Kirkham, 2015; Vessel & Rubin, 2010). This study presents a com- treatments respectively. Our results thus suggest that categorizability parison of a different kind than those explored in previous reports, e.g.,

M.W. Khaw, et al. $FWD3V\FKRORJLFD ² modifies the stability of aesthetic first-impressions on a similar scale to viewing effects. Our data suggest that the attribution of viewing time having different time constraints. Although Augustin et al. (2008) might further change over a larger number of trials than those tested found evidence for content-based aesthetic processing following pre- here (Fig. 5B). Indeed, the maximum possible viewing time per non-HC sentations as short 10 ms, the results here demonstrate that percep- image in our task (25 s) is less than the mean viewing time (< 30 s) tually ambiguous (LC) images limit aesthetic processing in the same measured in a live museum setting (Smith & Smith, 2001). Although manner as constrained viewing time does. Related to the issue of pro- there are no significant differences in viewing time distribution be- cessing constraints, the finding that time-limited ratings are greater tween repeated viewings, the trend in Fig. 5B (as well as the ordering of than their unlimited counterparts (Fig. 3A) might appear to be incon- means in Fig. 5A) suggests that LC views might eventually exceed HC gruent with the effect of preferential viewing time (in the Spotlight views after sufficient exposure. This is a natural trend to expect, if one task). Without further duration conditions, we are unable to determine assumes that subjects eventually seek out additional perceptual dis- whether fixed initial presentations benefit from shorter durations. Such coveries (i.e., “a-ha” moments as by Muth & Carbon, 2013) within an effect would be consistent with the known effects of increased pro- seemingly abstract images (once the content of HC and IC images are cessing time on decreased liking (e.g., Kuchinke et al., 2009), discussed processed), or if the ostensibly more challenging art was meta-cogni- and disambiguated from preferential viewing time below. In addition, tively chosen to be viewed/appreciated later (Kahnx, Ratner, & although we replicate the scale of within-subject correlations docu- Kahneman, 1997). Further questions on this issue will center on ex- mented by Verhavert et al. (2018), their aggregate ratings appear to not tended presentation and viewing durations (beyond the range our significantly change across their range of tested durations. A simple subjects performed), to gauge the larger range of possible associations albeit study-specific explanation of this observation is that brief aes- between viewing time, presentation time, and rating changes. thetic judgments incorporate additional assumptions about images re- Overall, our results highlight how one kind of visual object relative to unrestricted viewings. In the former, holistic “gist” impressions cognition – person perception – influences the assignment and devel- (Oliva, 2005; Oliva & Torralba, 2006) of the portraits we selected might opment of aesthetic ratings. These effects occur at several different influence early judgments, taking into account assumed features (e.g., timescales, ranging from ratings made following 30 ms of initial views, assumed symmetry or simplicity) or sensory-level information (e.g., to those made without any duration constraints. We also document how shading and shape) that are updated upon extended viewing. this form of perceptual categorizability modifies viewing patterns in When it comes to the resulting changes in ratings, we are able to tandem with rating changes per unit time of viewing. The detectability attribute a significant portion of this variability to a positive familiarity of human figures appears to boost aggregate evaluations as well as the bias. That is, images that accrue more viewing time tend to be rated rate of evaluative changes over time. Further research is warranted to more highly by the end of the experiment. A causal link between fa- determine whether these effects are specific to depictions of people or miliarity and processing fluency was initially established under fluency- generalizable to other focal content in artworks. In the latter case, attribution accounts (Jacoby & Dallas, 1981), whereby memory traces aesthetic evaluation might be generally facilitated by levels of ease in of previously experienced stimuli facilitate performance on a task (e.g., forming conceptual or semantic associations, consistent with broader word recognition). The positive effect of viewing time here contrasts accounts of processing fluency. If so, further experiments of this kind results that pair processing time against liking. Measures such as clas- will be able to identify other visual attributes that independently affect sification time (Reber et al., 2004) and recognition time (Kuchinke early and late phases of aesthetic processing, a distinction emphasized et al., 2009) are negatively related to preferences, as longer times in in other recent studies (e.g., Belke et al., 2015; Graf & Landwehr, 2015; such cases represent a reduced ease of processing. Critically, this ex- Thömmes & Hübner, 2014). periment involves several passive exposures prior to exploratory viewing during the spotlight procedure. In this way, time distributed Appendix A. Supplementary figures freely toward stimuli – beyond requisite processing time – should be predictive of higher liking. Given this distinction between processing Supplementary data to this article can be found online at https:// and surplus viewing time, future studies ought to delineate when this doi.org/10.1016/j.actpsy.2019.05.006. distinction occurs and how this process aligns with the visual recognition of focal subject matter. Furthermore, although it is difficult to tell References from our data alone whether increasing aesthetic value causally pro- motes increased viewing time (or the other way around), this result is Alter, A. L., & Oppenheimer, D. M. (2008). Easy on the mind, easy on the wallet: The roles consistent with other non-aesthetics studies demonstrating that: (i) gaze of familiarity and processing fluency in valuation judgments. Psychonomic Bulletin & Review, 15(5), 985–990. positively modulates resulting preferences for faces (Shimojo, Simion, Augustin, M. D., Leder, H., Hutzler, F., & Carbon, C. C. (2008). Style follows content: On Shimojo, & Scheier, 2003), (ii) long-term familiarity determines liking the microgenesis of art perception. Acta Psychologica, 128(1), 127–138. for face views (Mita, Dermer, & Knight, 1977), and (iii) multiple sub- Bachmann, T., & Vipper, K. (1983). Perceptual rating of paintings from different artistic styles as a function of semantic differential scales and exposure time. Archiv für liminal exposures lead to increased positive affect (Monahan, Murphy, Psychologie, 135(2), 149–161. & Zajonc, 2000). Interpreted collectively, these results suggest that ei- Bacon-Macé, N., Macé, M. J. M., Fabre-Thorpe, M., & Thorpe, S. J. (2005). The time ther a fluency-based (Winkielman & Cacioppo, 2001) or general fa- course of visual processing: Backward masking and natural scene categorisation. miliarity effect is operating here, with HC portraits cultivating the most Vision Research, 45(11), 1459–1469. Bar, M., Neta, M., & Linz, H. (2006). Very first impressions. Emotion, 6(2), 269. positive increases in ratings following a fixed duration of views. Given Battaglia, F., Lisanby, S. H., & Freedberg, D. (2011). Corticomotor excitability during that these effects are likely driven by the perceptual discovery of sin- observation and imagination of a work of art. Frontiers in Human Neuroscience, 5(79). fi gular portrait subjects, future studies might investigate whether the Bel , A. M., Kasdan, A., Rowland, J., Vessel, E. A., Starr, G., & D, P. (2018). Rapid timing of musical aesthetic judgments. Journal of Experimental Psychology: General. same results hold for non-naïve individuals already familiar with such Belke, B., Leder, H., & Carbon, C. C. (2015). When challenging art gets liked: Evidences artworks. For instance, subtle features of portraits such as profile angles for a dual preference formation process for fluent and non-fluent portraits. PLoS One, (Humphrey & McManus, 1973; Schirillo, 2000), and pose similarity 10(8), e0131796. Berlyne, D. E. (1960). Conflict, arousal, and curiosity. with the viewer's posture (Bertamini, Byrne, & Bennett, 2013) are Bertamini, M., Byrne, C., & Bennett, K. M. (2013). Attractiveness is influenced by the known to affect viewer appreciation – such effects ought to arise after relationship between postures of the viewer and the viewed person. i-Perception, 3(3), the initial process of person perception. Thus, the effects of viewing 170–179. Bornstein, R. F., & D'Agostino, P. R. (1994). The attribution and discounting of perceptual time here can be understood as describing short-term aesthetic value fluency: Preliminary tests of a perceptual fluency/attributional model of the mere assignment (under 30 s), following person recognition in untrained exposure effect. Social Cognition, 12(2), 103–128. viewers. Carney, D. R., Colvin, C. R., & Hall, J. A. (2007). A thin slice perspective on the accuracy of first impressions. Journal of Research in Personality, 41(5), 1054 1072. There are also indicators regarding limits on these preferential –

M.W. Khaw, et al. $FWD3V\FKRORJLFD ²

Chatterjee, A. (2003). Prospects for a cognitive neuroscience of visual aesthetics. Bulletin fluency theory: Four algorithmic measures for antecedents of aesthetic preferences. of Psychology and the Arts, 4(2), 55–60. Psychology of Aesthetics, Creativity, and the Arts, 12(4), 399. Cross, E. S., Hamilton, A. F. D. C., Kraemer, D. J., Kelley, W. M., & Grafton, S. T. (2009). Mita, T. H., Dermer, M., & Knight, J. (1977). Reversed facial images and the mere-ex- Dissociable substrates for body motion and physical experience in the human action posure hypothesis. Journal of Personality and Social Psychology, 35(8), 597. observation network. European Journal of Neuroscience, 30(7), 1383–1392. Monahan, J. L., Murphy, S. T., & Zajonc, R. B. (2000). Subliminal mere exposure: Specific, Crozier, W. R., & Greenhalgh, P. (1992). The empathy principle: Towards a model for the general, and diffuse effects. Psychological Science, 11(6), 462–466. psychology of art. Journal for the Theory of Social Behaviour, 22(1), 63–79. Morris, J. S., Öhman, A., & Dolan, R. J. (1998). Conscious and unconscious emotional Cupchik, G. C., & Berlyne, D. E. (1979). The perception of collative properties in visual learning in the human amygdala. Nature, 393(6684), 467. stimuli. Scandinavian Journal of Psychology, 20(1), 93–104. Morton, J., & Johnson, M. H. (1991). Conspec and conlern: A two-process theory of infant de Gelder, B., De Borst, A. W., & Watson, R. (2015). The perception of emotion in body face recognition. Psychological Review, 98(2), 164. expressions. Wiley Interdisciplinary Reviews: Cognitive Science, 6(2), 149–158. Muth, C., & Carbon, C. C. (2013). The aesthetic aha: On the pleasure of having insights Deaner, R. O., Khera, A. V., & Platt, M. L. (2005). Monkeys pay per view: Adaptive va- into gestalt. Acta Psychologica, 144(1), 25–30. luation of social images by rhesus macaques. Current Biology, 15(6), 543–548. Oliva, A. (2005). Gist of the scene. 251–256. Fei-Fei, L., Iyer, A., Koch, C., & Perona, P. (2007). What do we perceive in a glance of a Oliva, A., & Torralba, A. (2006). Building the gist of a scene: The role of global image real-world scene? Journal of Vision, 7(1), 1–29. features in recognition. Progress in Brain Research, 155,23–36. Flexas, A., Rosselló, J., Christensen, J. F., Nadal, M., La Rosa, A. O., & Munar, E. (2013). Payne, J. W. (1976). Task complexity and contingent processing in decision making: An Affective priming using facial expressions modulates liking for abstract art. PLoS One, information search and protocol analysis. Organizational Behavior and Human 8(11), e80154. Performance, 16(2), 366–387. Freedberg, D., & Gallese, V. (2007). Motion, emotion and empathy in esthetic experience. Phillips, L. T., Weisbuch, M., & Ambady, N. (2014). People perception: Social vision of Trends in Cognitive Sciences, 11(5), 197–203. groups and consequences for organizing and interacting. Research in Organizational Germine, L., Russell, R., Bronstad, P. M., Blokland, G. A., Smoller, J. W., Kwok, H., & Behavior, 34, 101–127. Wilmer, J. (2015). Individual aesthetic preferences for faces are shaped mostly by Reber, R., Schwarz, N., & Winkielman, P. (2004). Processing fluency and aesthetic environments, not genes. Current Biology, 25(20), 2684–2689. pleasure: Is beauty in the perceiver's processing experience? Personality and Social Graf, L. K., & Landwehr, J. R. (2015). A dual-process perspective on fluency-based aes- Psychology Review, 8(4), 364–382. thetics: The pleasure-interest model of aesthetic liking. Personality and Social Reber, R., Winkielman, P., & Schwarz, N. (1998). Effects of perceptual fluency on af- Psychology Review, 19(4), 395–410. fective judgments. Psychological Science, 9(1), 45–48. Grafton, S. T., Arbib, M. A., Fadiga, K., & Rizzolatti, G. (1996). Localization of grasp Reeck, C., Wall, D., & Johnson, E. J. (2017). Search predicts and changes patience in representations in humans by positron emission tomography. Experimental Brain intertemporal choice. Proceedings of the National Academy of Sciences, 201707040. Research, 112(1), 103–111. Rhodes, G. (2006). The evolutionary psychology of facial beauty. Annual Review of Harris, L. T., & Fiske, S. T. (2009). Social neuroscience evidence for dehumanised per- Psychology, 57, 199–226. ception. European Review of Social Psychology, 20(1), 192–231. Riesenhuber, M., & Poggio, T. (2000). Models of object recognition. Nature Neuroscience, Hayden, B. Y., Parikh, P. C., Deaner, R. O., & Platt, M. L. (2007). Economic principles 3(11s), 1199. motivating social attention in humans. Proceedings of the Royal Society of London: Rizzolatti, G., Fadiga, L., Gallese, V., & Fogassi, L. (1996). Premotor cortex and the re- Biological Sciences, 274(1619), 1751–1756. cognition of motor actions. Cognitive Brain Research, 3(2), 131–141. Heider, F., & Simmel, M. (1944). An experimental study of apparent behavior. The Rousselet, G. A., Macé, M. J. M., & Fabre-Thorpe, M. (2003). Is it an animal? Is it a human American Journal of Psychology, 57(2), 243–259. face? Fast processing in upright and inverted natural scenes. Journal of Vision, 3(6), Hekkert, P., & van Wieringen, P. C. (1990). Complexity and prototypicality as determi- 440–455. nants of the appraisal of cubist paintings. British Journal of Psychology, 81(4), Rule, N. O., & Ambady, N. (2008). Brief exposures: Male sexual orientation is accurately 483–495. perceived at 50 ms. Journal of Experimental Social Psychology, 44(4), 1100–1105. Hönekopp, J. (2006). Once more: Is beauty in the eye of the beholder? Relative con- Rule, N. O., Ambady, N., & Hallett, K. C. (2009). Female sexual orientation is perceived tributions of private and shared taste to judgments of facial attractiveness. Journal of accurately, rapidly, and automatically from the face and its features. Journal of Experimental Psychology: Human Perception and Performance, 32(2), 199. Experimental Social Psychology, 45(6), 1245–1251. Humphrey, N. K., & McManus, I. C. (1973). Status and the left cheek. New Scientist, 59, Rutherford, M. D., & Kuhlmeier, V. A. (Eds.). (2013). Social perception: Detection and in- 437–439. terpretation of animacy, agency, and intention. MIT Press. Jacoby, L. L., & Dallas, M. (1981). On the relationship between autobiographical memory Sbriscia-Fioretti, B., Berchio, C., Freedberg, D., Gallese, V., & Umilta, M. A. (2013). Erp and perceptual learning. Journal of Experimental Psychology: General, 110(3), 306. modulation during observation of abstract paintings by Franz Kline. PLoS One, 8(10), Jacques, C., & Rossion, B. (2006). The speed of individual face categorization. e75241. Psychological Science, 17(6), 485–492. Scarfe, P. (2015). Psychtoolbox demos. Retrieved from http://peterscarfe.com/ Johansson, G. (1973). Visual perception of biological motion and a model for its analysis. ptbtutorials.html. Perception & Psychophysics, 14(2), 201–211. Schepman, A., Rodway, P., Pullen, S. J., & Kirkham, J. (2015). Shared liking and asso- Johnson, M. H. (2005). Subcortical face processing. Nature Reviews Neuroscience, 6(10), ciation valence for representational art but not abstract art. Journal of Vision, 15(5), 766. 1–10. Kahnx, B., Ratner, R., & Kahneman, D. (1997). Patterns of hedonic consumption over Schirillo, J. A. (2000). Hemispheric asymmetries and gender influence rembrandt's por- time. Marketing Letters, 8(1), 85–96. trait orientations. Neuropsychologia, 38(12), 1593–1606. Kanwisher, N., McDermott, J., & Chun, M. M. (1997). The fusiform face area: A module in Schulz, M. F., & Sanocki, T. (2003). Time course of perceptual grouping by color. human extrastriate cortex specialized for face perception. Journal of Neuroscience, Psychological Science, 14(1), 26–30. 17(11), 4302–4311. Shi, J., Weng, X., He, S., & Jiang, Y. (2010). Biological motion cues trigger reflexive Kanwisher, N., & Yovel, G. (2006). The fusiform face area: A cortical region specialized attentional orienting. Cognition, 117(3), 348–354. for the perception of faces. Philosophical Transactions of the Royal Society of London: Shimojo, S., Simion, C., Shimojo, E., & Scheier, C. (2003). Gaze bias both reflects and Biological Sciences, 361(1476), 2109–2128. influences preference. Nature Neuroscience, 6(12), 1317. Khaw, M. W., Li, Z., & Woodford, M. (2018). Temporal discounting and search habits: Smith, J. K., & Smith, L. F. (2001). Spending time on art. Empirical Studies of the Arts, Evidence for a task-dependent relationship. Frontiers in Psychology, 9(2102). 19(2), 229–236. Kimchi, R., Hadad, B., Behrmann, M., & Palmer, S. E. (2005). Microgenesis and onto- Su, J., van Boxtel, J. J., & Lu, H. (2016). Social interactions receive priority to conscious genesis of perceptual organization: Evidence from global and local processing of perception. PLoS One, 11(8), e0160468. hierarchical patterns. Psychological Science, 16(4), 282–290. Sugase, Y., Yamane, S., Ueno, S., & Kawano, K. (1999). Global and fine information coded Kirsch, L. P., Urgesi, C., & Cross, E. S. (2016). Shaping and reshaping the aesthetic brain: by single neurons in the temporal visual cortex. Nature, 400(6747), 869. Emerging perspectives on the neurobiology of embodied aesthetics. Neuroscience & Thömmes, K., & Hübner, R. (2014). A picture is worth a word: The effect of titles on Biobehavioral Reviews, 56–68. aesthetic judgments. Proceedings of the twenty-third biennial congress of the International Klein, J. T., Shepherd, S. V., & Platt, M. L. (2009). Social attention and the brain. Current Association of Empirical Aesthetics (pp. 599–603). . Biology, 19(20), R958–R962. Ticini, F., Urgesi, C., & Calvo-Merino, B. (2014). Embodied aesthetics: Insight from Kuchinke, L., Trapp, S., Jacobs, A. M., & Leder, H. (2009). Pupillary responses in art cognitive neuroscience of performing arts. Aesthetics and the embodied mind: Beyond appreciation: Effects of aesthetic emotions. Psychology of Aesthetics, Creativity, and the art theory and the Cartesian mind-body dichotomy (pp. 103–115). Dordrecht: Springer. Arts, 3(3), 156. Todorov, A., Olivola, C. V., Dotsch, R., & Mende-Siedlecki, P. (2015). Social attributions Leder, H., Belke, B., Oeberst, A., & Augustin, D. (2004). A model of aesthetic appreciation from faces: Determinants, consequences, accuracy, and functional significance. and aesthetic judgments. British Journal of Psychology, 95(4), 489–508. Annual Review of Psychology, 66. Leder, H., Goller, J., Rigotti, T., & Forster, M. (2016). Private and shared taste in art and Verhavert, S., Wagemans, J., & Augustin, M. D. (2018). Beauty in the blink of an eye: The face appreciation. Frontiers in Human Neuroscience, 10, 155. time course of aesthetic experiences. British Journal of Psychology, 109(1), 63–84. Lipps, T. (1935). Empathy, inner imitation, and sense-feelings. New York: Holt and Vessel, E. A., & Rubin, N. (2010). Beauty and the beholder: Highly individual taste for Company. abstract, but not real-world images. Journal of Vision, 10(2), 1–14. Locher, P., Unger, R., Sociedade, P., & Wahl, J. (1993). At first glance: Accessibility of the Vessel, E. A., Starr, G. G., & Rubin, N. (2013). Art reaches within: Aesthetic experience, physical attractiveness stereotype. Sex Roles, 28(11−12), 729–743. the self and the default mode network. Frontiers in Neuroscience, 7(258). Martindale, C., & Moore, K. (1988). Priming, prototypicality, and preference. Journal of Vuilleumier, P., Armony, J. L., Driver, J., & Dolan, R. J. (2001). Effects of attention and Experimental Psychology: Human Perception and Performance, 14(4), 661. emotion on face processing in the human brain: An event-related fmri study. Neuron, Martindale, C., Moore, K., & West, A. (1988). Relationship of preference judgments to 30(3), 829–841. typicality, novelty, and mere exposure. Empirical Studies of the Arts, 6(1), 79–96. Whittlesea, B. W. (1993). Illusions of familiarity. Journal of Experimental Psychology: Mayer, S., & Landwehr, J. R. (2018). Quantifying visual aesthetics based on processing Learning, Memory, and Cognition, 19(6), 1235.

M.W. Khaw, et al. $FWD3V\FKRORJLFD ²

Willenbockel, V., Sadr, J., Fiset, D., Horne, G. O., Gosselin, F., & Tanaka, J. W. (2010). Journal of Personality and Social Psychology, 81(6), 989. Controlling low-level image properties: The shine toolbox. Behavior Research Methods, Winkielman, P., Halberstadt, J., Fazendeiro, T., & Catty, S. (2006). Prototypes are at- 42(3), 671–684. tractive because they are easy on the mind. Psychological Science, 17(9), 799–806. Willis, J., & Todorov, A. (2006). First impressions: Making up your mind after a 100-ms Yarbus, A. L. (1967). Eye movements during perception of complex objects. Boston, MA: exposure to a face. Psychological Science, 17(7), 592–598. Springer177–211. Winkielman, P., & Cacioppo, J. T. (2001). Mind at ease puts a smile on the face: Zajonc, R. B. (1968). Attitudinal eﬀects of mere exposure. Journal of Personality and Social Psychophysiological evidence that processing facilitation elicits positive aﬀect. Psychology, 9(2p2), 1.