When a Picasso is a “Picasso”: The entry point in the identification of visual art Acta Psychologica 133 (2010) 191–202 Contents lists available at ScienceDirect Acta Psychologica j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / a c t p s y When a Picasso is a ‘‘Picasso”: The entry point in the identification of visual art B. Belke a,*, H. Leder a, G. Harsanyi b,c, C.C. Carbon b a Faculty of Psychology, University of Vienna, Liebiggasse 5, 1010 Vienna, Austria b Department of Psychology, University of Bamberg, Markusplatz 3, 96047 Bamberg, Germany c Institute of Psychology, Humboldt University, Berlin, Rudower Chaussee 18, 12489 Berlin, Germany a r t i c l e i n f o Article history: Received 30 December 2008 Received in revised form 10 November 2009 Accepted 13 November 2009 Available online 24 December 2009 PsycINFO classification: 2323 2340 2610 Keywords: Art perception Classification Identification Style-based recognition Aesthetic experience 0001-6918/$ - see front matter � 2009 Elsevier B.V. A doi:10.1016/j.actpsy.2009.11.007 * Corresponding author. E-mail address: benno.belke@gmail.com (B. Belke) a b s t r a c t We investigated whether art is distinguished from other real world objects in human cognition, in that art allows for a special memorial representation and identification based on artists’ specific stylistic appearances. Testing art-experienced viewers, converging empirical evidence from three experiments, which have proved sensitive to addressing the question of initial object recognition, suggest that identi- fication of visual art is at the subordinate level of the producing artist. Specifically, in a free naming task it was found that art-objects as opposed to non-art-objects were most frequently named with subordinate level categories, with the artist’s name as the most frequent category (Experiment 1). In a category-ver- ification task (Experiment 2), art-objects were recognized faster than non-art-objects on the subordinate level with the artist’s name. In a conceptual priming task, subordinate primes of artists’ names facilitated matching responses to art-objects but subordinate primes did not facilitate responses to non-art-objects (Experiment 3). Collectively, these results suggest that the artist’s name has a special status in the memo- rial representation of visual art and serves as a predominant entry point in recognition in art perception. � 2009 Elsevier B.V. All rights reserved. 1. Introduction People have an astonishing ability to identify objects in a fast, automatic, and very efficient way. This behavior is an evolution- based adaptation for accessing memory representations involving the identification of objects and the corresponding demands for specific actions (Rosch, 1975). In their seminal studies, Rosch, Mer- vis, Gray, Johnson, and Boyes-Braem (1976) determined the princi- ples by which humans divide up the world the way they do. There is now a substantial body of research suggesting that most objects are initially classified at a particular level of abstraction, called the basic level (for a review, see Rosch et al., 1976). The basic level is defined as (a) the level at which most knowledge is organized and the highest level of abstraction at which a single mental image can be formed, (b) category members share a similar shape, and (c) similar motor actions are used to interact with category members (Tanaka, 2001). Rosch (1975) argued that the basic level of catego- rization provides the entry point in human classification and repre- sents the level at which objects are first recognized and proposed: ‘‘in the perceived world, information rich bundles of perceptual or functional attributes occur that form natural discontinuities, and ll rights reserved. . that [. . .] basic cuts in categorization are made at these discontinu- ities” (p. 31). The authors empirically demonstrated the special sta- tus of the basic level by interpreting participants’ naming preferences and category membership verification times as indi- cating that people first identify objects at the basic level and then access the superordinate level or subordinate level categories. According to Rosch et al. (1976), basic level categories are noted by a balance between informativeness or ‘‘cue validity” (the num- ber of attributes the concept conveys) and ‘‘cognitive economy” (a sort of summary of the important attributes that distinguish it from other categories). Informativeness is lacking at the highest le- vel because few attributes are conveyed, and economy is missing at the lowest level because too many attributes are conveyed. Later, Jolicoeur, Gluck, and Kosslyn (1984) proposed that the en- try point in recognition corresponds to the level where ‘‘contact is made first with semantic memory” (p. 272) and where the percep- tual stimulus first makes contact with its underlying memorial representation. Importantly, Jolicoeur et al. (1984) noted that the entry point often corresponds to the basic level, but in many in- stances of object identification it does not. Research on human ob- ject identification demonstrated that the entry point could be modulated by at least two factors: domain-specific expertise and typicality of an exemplar for its corresponding basic level. Con- cerning the former, expertise in a particular field is likely to shift http://dx.doi.org/10.1016/j.actpsy.2009.11.007 mailto:benno.belke@gmail.com http://www.sciencedirect.com/science/journal/00016918 http://www.elsevier.com/locate/actpsy 192 B. Belke et al. / Acta Psychologica 133 (2010) 191–202 entry level of many objects towards the subordinate level (Rosch et al., 1976). Tanaka and Taylor (1991), for example, found that a bird expert is able to recognize a picture of a sparrow as a ‘‘spar- row” as quickly as he recognizes it as a ‘‘bird”. In the domain of face perception, Tanaka (2001) suggested a general basic level shift in human perception towards subordinate classification of familiar faces. Thus, a face will more likely be identified at the unique level of identity (e.g., as ‘‘Bill Clinton” rather than as a ‘‘face” at the basic level). Similar downward shifts in recognition were found by Gau- thier and Tarr (1997) after participants were trained in the identi- fication of artificial objects. On the basis of these findings, Jolicoeur et al. (1984) concluded ‘‘for many objects (and perhaps many situ- ations) we use identification routines at levels other than the basic level” (p. 272). 1.1. Special features of art-objects What might be the entry point in the identification of visual art? Up to this point research in the domain of object recognition has been concerned with object classes such as furniture (Rosch et al., 1976), faces (e.g., Tanaka, 2001), every-day objects (Op de Beeck & Wagemans, 2001) and even artificial objects (e.g., Gauthier & Tarr, 1997) but very little is known about the representation and initial identification of visual art. For instance, what might be the first access to semantic memory when a person identifies Pablo Pi- casso’s ‘‘Dora Maar seated”? This painting might be initially recog- nized as a deformed depiction of a ‘‘woman’s head” or simply as a ‘‘human”. Alternatively, it might be identified as ‘‘Dora Maar seated”, a ‘‘Picasso”, a ‘‘Cubist painting”, or even more general, as a ‘‘work of art”. These numerous classification options reveal dif- ferent taxonomies applicable to art-objects. We propose that visual art is distinguishable from many other object classes, since it matches at least two taxonomies that may affect its identification. The first is based on depictive content (what is represented) and the second on stylistic representation (how it is represented), such as the corresponding art school or artist’s name (cf. Augustin, Le- der, Hutzler, & Carbon, 2008). In this respect, artworks are multidi- mensional stimuli that can be placed in numerous classifying contexts and therefore provide a challenge for initial recognition. The idea of multiple classification options for art-objects was proposed within a framework model of aesthetic experience, in which explicit identification is one of the cumulative stages of information processing that are involved in aesthetic experiences (Leder, Belke, Oeberst, & Augustin, 2004). The model states that, depending on the viewer’s experience, explicit classification could be based on style or representational content. For instance, in the case of René Magritte’s painting ‘‘Ceci n’est pas une pipe” it is pos- sible to identify either a depiction of a ‘‘tobacco pipe” or a painting by ‘‘René Magritte”, with both referring to a similar level of abstrac- tion. Consequently, the investigation of a possible entry point in vi- sual art must take into account at least two conceptual hierarchies, one content-related and the other art-specific, which allow for hier- archical relationships between categories (superordinate, basic, and subordinate level categorization). Moreover, in many cases of modern and contemporary art (e.g., Abstract Art, Minimalism) ex- plicit semantic references are often completely absent. In these cases, the identification has to rely on syntactic qualities (Berlyne, 1971) such as stylistic or expressive information rather than its semantic content (e.g. subject matter, represented object). Analyses of modern and contemporary art reveal that a persis- tent feature of art is stylistic variety, which refers to single artists as the main proponents (Leder et al., 2004). The stylistic diversity in art is propelled by artists’ attempts to distinguish themselves from others in the art market. In order to be successful artists’ styles have to be distinctive as well as novel or innovative. This strong individualization makes the art world special in that production structures still are very individual, but create forces that not only call for idiosyncratic objects, such as single artworks, but also encourage the establishment of individual styles – which are recognizable and attributable to single artists over different works of art (Grasskamp, 1994). This style-based appearance may be reflected in human recog- nition and presumably allows for a style-based type of processing (Leder, 2003; Leder et al., 2004; Winston & Cupchik, 1992). As a re- sult, the initial identification might yield cognitive processing that differs from other object domains (but might be comparable to the domain of architecture or design as well, see Carbon & Leder, 2005). Memory representations of artworks may be closely linked in semantic memory with the creating artist, as single works often contain elements which share ill-defined features of style that are similar throughout one artist’s style (Hartley & Homa, 1981). These artist’s specific stylistic features could serve as ‘‘basic-cuts” (Rosch et al., 1976) in categorization and provide an entry point (Tanaka, 2001) of recognition for artworks. We assumed that the impor- tance of style-based processing and stylistic diversity in the visual arts makes it likely that artworks might most efficiently be recog- nized with respect to single artists, e.g. a particular painting being considered a ‘‘Picasso”. This means that a Picasso painting might not initially be recognized as a ‘‘portrait” nor as a ‘‘Cubist painting”, but rather as a painting ‘‘by Picasso” or a ‘‘Picasso-esque” painting. Alternatively, from the way that people gather knowledge and expertise in art (see Parsons, 1987), it might be possible that art- works are initially identified on a more general level of abstraction. Art historians propose the use of art-styles to categorize artworks of the same historical period (e.g., ‘‘Expressionism”), of groups of closely related artists (e.g., ‘‘Bauhaus”), or of a common visual ap- proach to depiction (e.g., ‘‘Pointillism”). These concepts applied by art historians to categorize art might not be reflected in the initial human identification of artworks. We assume that they are of min- or relevance compared with a more specific artist-related categori- zation (Hasenfus, Martindale, & Birnbaum, 1983). 1.2. The present study We investigated whether there is a particular entry point in the identification of visual art that is different from other object classes, and if so, whether this entry point is at a medium (basic) level of abstraction as proposed for many object classes. In three experi- ments, initial identification of artworks was investigated by testing the relationship between art-related categories and art identifica- tion by relative art experts. Using tasks such as speeded naming, category verification, and visual matching, paradigms were applied that had been predominantly used in the domain of object and face identification (e.g., Rosch et al., 1976; Segui & Fraisse, 1968; Tanaka, 2001). Performances on works of art were tested against every-day objects, for which entry points are well investigated by a substan- tial body of research. According to our hypotheses about style- based representation in art, it was assumed that an entry point is based on recognition of style and shows up at a subordinate level of the creating artist with highest frequency proportion in naming (Experiment 1), with the fastest verification speed (Experiment 2), with the largest amount of priming in a visual matching task (Experiment 3). Alternatively, according to a structural definition of a basic level, artworks should be identified first at a more general level (such as artistic genre, e.g. ‘‘landscape”) and are more fre- quently named as such, are verified faster and yield higher priming gains on such basic level categories than at subordinate level cate- gories (of the artist’s name). In contrast to these two hypotheses, art — due to its complexity, ambiguity and stylistic representation — may only allow for idiosyncratic recognition, so that a general entry point as revealed for many other object domains (e.g., Rosch et al., 1976; Segui & Fraisse, 1968) may simply not exist. B. Belke et al. / Acta Psychologica 133 (2010) 191–202 193 For methodological reasons, the three experiments required participants to have some professional art training, as art recogni- tion relies on previous experience and knowledge. For example, a person who has never encountered a painting by Gerhard Richter and who is not familiar with his name would neither be able to classify it as such in a naming task nor to identify Richter’s name in a matching task nor respond to it in a priming task. Therefore, it was necessary to test participants with some academic back- ground in art history, as ensured by at least one year of study. 2. Experiment 1: Free Naming task A free naming task (Rosch et al., 1976; Tanaka & Taylor, 1991) was carried out as the initial experiment. Participants were asked to name each object as fast as possible with the first noun that came instantaneously to mind. In order to analyze a broad spec- trum of categories and concepts applied in the domain of art, paintings from a wide range of art-historic genres, artistic styles, and artists were used. Non-art-objects served as control stimuli. These consisted of artifacts and natural object classes, such as cars and dogs, for which effects of identification and explicit categoriza- tion are already well established and for which identification at a ‘‘basic level” was proposed and expected (e.g., Jolicoeur et al., 1984; Mervis & Rosch, 1981; Murphy & Brownell, 1985; Rosch et al., 1976; Tanaka, 2001; Tanaka & Taylor, 1991). By analyzing naming frequencies, the aim of this experiment was to reveal at which level of abstraction artworks are initially identified compared with every-day objects. Previous research has shown that participants use basic level names (e.g., ‘‘bird”, ‘‘dog”, ‘‘chair”, and ‘‘hammer”) when asked to spontaneously iden- tify pictures of common objects (e.g., Tanaka & Taylor, 1991). This finding has been used as evidence that the most accessible level of abstraction for categorizing objects is the ‘‘basic level”. For free naming frequencies it was assumed that art-objects might be iden- tified based on their corresponding artistic styles that are either at level of the individual artists or according to art-schools. Through- out the analyses we will refer to every-day objects as ‘‘non-art- objects” and to works of art as ‘‘art-objects”. Moreover, findings were taken into account that demonstrated effects of domain-specific expertise on object recognition. These studies showed that experts are more likely to classify objects in their domain of expertise at a more specific level of abstraction (e.g., Tanaka & Taylor, 1991). Therefore, each participant’s level of art expertise, as reflected by academic training in the arts, was assessed to reveal potential influences on the level of identification and nature of concepts applied to categorize art-objects. 2.1. Method 2.1.1. Participants Twenty students (17 female) studying for a MA degree in art history at the Freie Universität Berlin took part in the experiment. Mean age was 24.0 years (SD = 2.73) ranging from 21 to 33 years. Students had been enrolled between two semesters (1 year) and 12 semesters (6 years) (Mean = 6.45, SD = 3.36). Each participant was tested individually and received 5€ for participation. All partic- ipants had normal or corrected-to-normal vision. 2.1.2. Stimuli The stimuli consisted of 42 pictures, half of which were from art and half from non-art categories. Seven pictures were selected from the non-art categories ‘‘dog” (Saint Bernard, Boxer, Collie, Dalmatian, Fox terrier, Poodle, and German Shepherd Dog), ‘‘car” (Audi A8, Mercedes-Benz E Class, BMW 5 series, Volkswagen Golf, Porsche 911, MCC Smart, and Volkswagen Polo), and ‘‘home fur- nishings” (office chair, bread bin, dining table, folding chair, desk, desk lamp, and table lamp). No furnishings by well-known design- ers were used (such as Eames chairs) to prevent familiarity effects with those objects. For the same reason, the trademarks and logos of cars were erased from the pictures. The criterion for the classes of non-art-objects was that they had been used in previous studies on human object identification studies (e.g., Rosch et al., 1976). Moreover, they belong to the most common (most frequently men- tioned) categories of nouns in Germany. Exemplars of ‘‘dog” and ‘‘home furnishings” categories were among the ten most fre- quently mentioned category norms in Mannhaupt (1983) (the Ger- man equivalent of Battig & Montague, 1969) and selected car brands were among the ten most familiar brands in Germany as determined by a recent market study of AC-Nielsen (2004). Art-objects were selected from various artists in the genres ‘‘landscape” (Paul Cézanne, André Derain, Caspar-David Friedrich, Wassily Kandinsky, Claude Monet, Vincent van Gogh, and John Constable), ‘‘portrait” (Max Beckmann, Paul Klee, Gustav Klimt, Roy Lichtenstein, Edvard Munch, Pablo Picasso, and Gerhard Rich- ter), and ‘‘still life” (Salvador Dali, Albrecht Dürer, Gerhard Richter, Vincent van Gogh, Andy Warhol, and two paintings by René Mag- ritte). See Appendix 1 for a list of titles and artists. Four additional pictures served as practice trials (i.e. wristwatch, racing bike and painting by Edgar Degas and Lyonel Feininger), which were se- lected from other object classes that were used in the experimental trials. Each picture was standardized to 380 square centimeters with the original width-to-height ratio maintained. 2.1.3. Procedure The experiment consisted of four practice and 42 experimental trials. At the beginning of each trial, a short instruction appeared on the screen asking participants to name each of the stimuli as fast as possible with the first noun that comes to mind. After a 1000 ms interval, the written instruction was replaced with a 800 ms blank screen, which was then followed by a 2000 ms pic- ture-stimulus (either a non-art-object or art-object), which in turn was followed by another blank screen. After 1500 ms, the partici- pants were asked to start the next trial by pressing any key on the keyboard. The stimulus order was randomized. The experimenter sat behind the participant and noted down the verbal responses for each experimental trial. Instructions and stimuli were presented on a Macintosh Powerbook with a 15” monitor (resolution 1024 � 768 pixels). The procedure was con- trolled by the software PsyScope 1.2.5 (Cohen, MacWhinney, Flatt, & Provost, 1993). Viewing distance to the screen was approxi- mately 70 cm. Visual angles ranged from approximately 10.93� to 13.36�. In order to exclude the possibility that basic level categories were used due to a lack of familiarity with subordinate level cate- gories, a post-experimental questionnaire explicitly tested whether participants were able to identify each stimulus on a very specific (subordinate) level. In this paper and pencil questionnaire, participants were asked to name each object on a printed-paper version at a specific level of categorization. For example, they were asked to specify the particular kind of car (e.g. BMW or Audi). For art-objects, participants were asked to indicate the artist’s name, painting’s title, or specific artistic style. Stimuli that could not be identified correctly at a subordinate level were excluded from the main analysis for the corresponding participant. 2.2. Results and discussion 2.2.1. Data correction Before the actual analyses, all incorrect responses were elimi- nated according to three criteria. First, verbal classifications were excluded from analyses if a person could not name an object cor- Object Domain P er ce nt R es po ns es Art Non-Art 0 10 20 30 40 50 60 70 80 Basic level Subordinate level Fig. 1. The results from Experiment 1 showing the percentage of basic level and subordinate level labels used to name art-objects and non-art-objects. Error bars represent within-subjects confidence intervals (Loftus & Masson, 1994). 194 B. Belke et al. / Acta Psychologica 133 (2010) 191–202 rectly at a subordinate level in the post-experimental question- naire. In the case of art-objects, this task required labeling of the specific painting with the artist’s name, artistic style, or title of the painting. If participants could not provide any of these informa- tion the corresponding trial was omitted (e.g., if the painting ‘‘Hampstead Heath” by Constable could not be labeled as ‘‘a Con- stable”, ‘‘Hampstead Heath”, or ‘‘English Romantic”). Second, if an object was named wrongly in the experiment, the response was considered as incorrect (e.g. a painting by ‘‘Klee” labeled with ‘‘von Stuck”). Third, verbal classifications, which were not in accor- dance with the instructions as decided by two judges, such as adjectives (e.g., ‘‘yellow” and ‘‘linear”) or free associations (e.g., ‘‘my favorite painting”) were eliminated from the data set. Given these three exclusion criteria, participants responded to 93.0% of art-objects and 95.5% of non-art-objects. Thus, participants were very familiar with the subordinate level terms of the objects. All correct responses were post hoc classified into three levels of abstraction (i.e., superordinate, basic, and subordinate levels). Re- sponses to non-art-objects were classified similar to those classi- fied by Tanaka and Taylor (1991). For example, the term ‘‘dog” was regarded as a basic level concept and ‘‘German Shepherd” as a subordinate level concept. While the level of abstraction for non-art-objects is quite clear by virtue of the level of inclusiveness and a host of empirical research (e.g., Jolicoeur et al., 1984; Mervis & Rosch, 1981; Murphy & Brownell, 1985; Rosch et al., 1976; Tana- ka, 2001; Tanaka & Taylor, 1991), the level of abstraction for art- object categories is less evident. Therefore, a special rating proce- dure was necessary in which naming responses were classified according to their level of abstraction. Five independent expert rat- ers received a non-redundant list of all categorizations in order to rate naming responses as referring to superordinate, basic, or sub- ordinate levels. A concordance criterion (degree of inter-rater agreement) was set to .8 of response-categorization. In other words, four of five raters had to agree that the given verbal re- sponse belonged to a superordinate, basic, or subordinate level. In 74.59% of post hoc response labeling, there was an inter-rater- agreement of at least 80%. The remaining naming responses were classified through expert discussion.1 2.2.2. Analysis of variance The dependent variable of interest was percentages of frequen- cies. Independent variables were object domain (i.e., art-object or non-art-object) and level of categorization (i.e., superordinate, ba- sic, or subordinate). Fig. 1 shows that participants classified non- art-objects in 54% of the trials at the basic level and 44% at the sub- ordinate level. Artworks were classified in 35% of the trials at the basic level and 65% at the subordinate level. No verbal responses were given at the superordinate level for art-objects, and only 2% of verbal classifications at superordinate level for non-art-objects. Given the lack of superordinate level re- sponses, verbal responses were analyzed by 2 � 2 analysis of vari- ance with object domain (art-object vs. non-art-object) and level of categorization (basic vs. subordinate). As expected, the effect of le- vel of categorization in this analysis was not significant, F(1, 19) < 1, p = 0.414, n.s. Critically, the object domain � level of cate- gorization was significant, F(1, 19) = 18.9, p < 0.001, g2p = 0.498. The interaction indicated that non-art-objects were more fre- quently named at the basic level than were art-objects, F(1, 1 Notably, the inter-rater agreement about response classifications for style-related categories (such as ‘‘Impressionism”) was below the consistency criteria of .8. After expert-discussion it has been decided that style-related classifications were to be regarded as subordinate level categories, since the criteria for basic level categories according to Rosch’s definition (see this article page 3) did not apply to style categories, but seemed more specific. However, it seems worth mentioning that the pattern of results in Experiment 1 did not significantly differ when artistic style was treated as a basic level category. 19) = 18.2, p < 0.001, g2p = 0.489, whereas art-objects were more frequently named at the subordinate level than non-art-objects, F(1, 19) = 17.8, p < 0.001, g2p = .4983. The interaction between level of categorization and object domain confirms previous findings, namely that the basic level (Rosch et al., 1976) is important in the identification of every-day objects. In contrast, participants cat- egorized art-objects more specifically with subordinate categories rather than applying basic level categories. In a more detailed analysis, possible influences of relative art expertise were considered as indicated by the duration of academic training in the arts. Based on the number of semesters of enroll- ment in a MA degree in art history a post-hoc median split divided participants into groups of low and high art expertise. The median for semesters of enrolment for both groups was 5.5. The resulting medians were 3.5 and 9.5 for the low expertise and high expertise group, respectively. To reveal possible differences due to art exper- tise, we ran a 2 � 2 � 2 analysis of variance with object domain (art-objects and non-art-objects) and level of categorization (basic and subordinate) as within-subjects factors and expertise (high and low) as between-subjects factors. Again, the only significant effect was the object domain � level of categorization interaction, F(1, 18) = 18.6, p < 0.001, g2p = 0.507. Neither a main effect of art expertise nor any interaction with any other factor was found (all Fs < 1.4). The lack of art expertise effects suggested that the preferred level of categorization of art-objects was not directly af- fected by differences in academic training. 2.2.3. Type of categories applied to categorize art-objects To further explore the nature of concepts participants applied in naming of art-objects, in subsequent analyses art-related classifi- cations were distinguished from content-related classifications. B. Belke et al. / Acta Psychologica 133 (2010) 191–202 195 These analyses of frequency were carried out independently of the level of abstraction. Art-related classifications comprised using the artist’s name, artistic genre, artistic style, or title of the painting. Content-related classifications consisted of naming referring to the depicted object of the painting (e.g., ‘‘eye”, ‘‘pipe”). In total, par- ticipants used 162 (61%) art-related and 117 (39%) content-related categories, v2 (1) = 7.3, p < 0.007. Furthermore, specific types of art-related naming were investigated. Among art-related catego- ries, the artist’s name was most frequently mentioned (135 in- stances, 83.3%), followed by the title of the work (20 instances, 12.4%), style (5 instances, 3.0%), and genre (2 instances, 1.2%). Thus, the artist’s name was the predominant art-related category, which was mentioned significantly more frequently than the works’ titles, t(19) = 5.9, p 6 .001. Titles were more frequently mentioned than artistic style, t(19) = 3.0, p = 0.007. The difference between style and genre was not significant, t(19) = 1.4, p = 0.174, n.s. In more detailed analyses, we also investigated whether relative levels of art expertise influenced the level of specificity within art- related classifications. Specifically, we tested whether free naming would shift towards the more particular level of titles with in- creased academic training in the arts. For identifying art-objects, more art-trained students used the artist’s name in 83 trials (86.5%) and applied titles in 13 trials (13.5%), while less art-trained students used the artist’s name in 52 trials (88.1%) and titles in se- ven trials (11.9%). Percentages of frequencies of naming were sub- jected to a mixed design ANOVA with type of subordinate classification (artist’s name vs. title) as within-subjects factor and art expertise (high vs. low expertise) as between-subjects factor. The analyses confirmed the above reported finding that artists’ names were significantly more often used than titles, F(1,18) = 34.1, p = 0.001, g2p = 0.654. The main effect of art expertise indi- cated that the differences of art expertise were marked by a higher frequency of art-related categories when compared to less trained participants, F(1,18) = 6.2, p = 0.023, g2p = 0.255. Critically, the interaction was not significant, F(1,18) < 1, p = 0.406, g2p = 0.039, n.s., indicating that the artist’s name was the predominant category in both expertise groups. To summarize the findings of the free naming task, it was found that speeded naming of art-objects differed from non-art-objects. As expected, non-art-objects were classified at a more general (ba- sic) level of abstraction, while art-objects were named at a more specific (subordinate) level, with the artist’s name as the predom- inant category. In respect to the kind of categories, it was found that participants used art-related categories (artist’s name, style or genre terms) more often than content-related categories. More- over, advanced academically trained students used more art-re- lated classifications compared to less academically trained students, while both groups showed equally high frequencies of identifying art-objects with their corresponding artists’ names. Re- sults are in accordance with the assumption that art-objects allow for a special kind of identification (and memory representation) based on individual artists’ styles that may serve as an entry point in recognition. In order to show that the findings of the free naming task did not simply reflect naming preferences of participants or social convention rather than initial memory access, the accessibil- ity of representations related to the artists’ names was explicitly tested in a speeded verification task (Experiment 2) and in category verification and priming facilitation tasks (Experiment 3). 2 It seems noteworthy to mention that portrait paintings differ from portrait photography, as their visual representation is more strongly shaped by a style. Therefore, findings of face recognition might not apply to recognition of portrait paintings. 3. Experiment 2: Speeded category-verification task Experiment 2 employed a speeded category-verification task similar to that used by Tanaka (2001, Experiment 2) in the domain of face recognition. Participants were presented with a superordi- nate, basic, or subordinate level category term and were subse- quently shown a picture, and were asked to indicate whether the picture was an exemplar of that category. The results were compared between two different object domains: Portraits as a representative of art-objects and cars as a sample of non-art-ob- jects. The choice of stimuli classes and selection of verbal catego- ries was oriented on the findings of Experiment 1. Portraits were selected because they are considered one of the most important genres in fine art.2 Moreover, according to a post hoc questionnaire of Experiment 1, the set of portrait paintings received the highest familiarity with the corresponding subordinate level terms which ensured that par- ticipants were familiar with the corresponding artists’ names or ti- tles (participants correctly named portraits, still life and landscape paintings at subordinate level in 84%, 77%, and 58%, respectively). Cars were selected as the appropriate contrast category because they represent a widely known class of every-day objects, for which specific familiarity data of a current survey were at hand (see AC-Nielsen, 2004). The task required participants to verify portraits and cars at the superordinate level (‘‘artwork” or ‘‘vehicle”), basic level (‘‘portrait” or ‘‘car”), and subordinate level (‘‘artist’s name” or ‘‘brand mark”). Reaction times were measured as the dependent variable. Accord- ing to a ‘‘basic-first” hypothesis (Tanaka, 2001), it could be rea- soned that portraits, like cars, should be identified at the basic level first and therefore categorized faster at a medium (basic) le- vel of abstraction than on a subordinate level. This would mirror findings obtained for many everyday object classes (cf. Rosch et al., 1976), namely that the entry point of recognition occurs at a basic level of abstraction. Alternatively, according to Experiment 1 and our hypothesis on an artist-specific recognition of art, we ex- pected participants to verify portraits faster on the subordinate le- vel (with the artist’s name) than on the basic level (as a ‘‘portrait”). 3.1. Method 3.1.1. Participants Sixteen (14 female) students of art history participated in this experiment. The mean age was 25.7 (SD = 3.94) ranging from 22 to 36 years. Participants were enrolled between three semesters (1.5 years) and 12 semesters (6 years) (mean = 6.63, SD = 2.75). Participants were tested individually and received payment for participation. None of them participated in Experiment 1. Each had normal or corrected-to-normal vision. 3.1.2. Stimuli Pictures were taken from two categories, cars and portraits. The car category consisted of eight brands: Audi, Mercedes Benz, BMW, Volkswagen, Fiat, Ford, Peugeot, and Opel. According to AC-Nielsen (2004), these car brands are among the ten most familiar brands within the German population. Eight portraits by the artists Leo- nardo da Vinci, Paul Klee, Gustav Klimt, Roy Lichtenstein, Anselm Feuerbach, Pablo Picasso, Jean-Honoré Fragonard, and Vincent van Gogh represented the portraits category. It was expected that paint- ings by these artists would be familiar to students of art history. In order to reduce expectancy effects, pictures from two contrast- ing categories, ‘‘bicycles” and ‘‘still life”, were used as filler trials. The contrast categories shared the same superordinate level as the target categories (i.e., cars and bicycles are vehicles and portraits and still lifes are artworks). The bicycle pictures consisted of eight different types of bicycles: mountain, racing, trekking, BMX, collapsible, fit- ness, women’s, and children’s bicycles; as still life paintings depic- 196 B. Belke et al. / Acta Psychologica 133 (2010) 191–202 tions were used from Pierre Chardin, Albrecht Dürer, Jean Metzinger, René Magritte, Gerhard Richter, Vincent van Gogh, Andy Warhol and Georges Braque (for a complete list of artists and works see Appen- dix 2). Additionally, four pictures from categories other than those used for the experimental trials were selected as practice trials. All pictures were standardized to 380 square centimeters with the ori- ginal width-to-height ratio of the paintings maintained. 3.1.3. Procedure At the beginning of the experimental session, participants re- ceived a written list of subordinate level terms for all of the 16 tar- get exemplars. Terms were presented in a random order one after the other on a monitor for 2500 ms each (after a fixation cross). Subsequently to an inter-stimulus interval of 800 ms, a fixation cross marked the beginning of each trial which remained for 1000 ms. Next, a blank screen appeared for 1000 ms, followed by a category word that remained for 2500 ms. After a 500 ms blank interval, the category name was replaced with a picture. Partici- pants were instructed to verify whether the preceding word label matched the picture (e.g., ‘‘Does the following picture show a por- trait?”) and to give their response (true or false) by pressing the corresponding ‘‘true” or ‘‘false” buttons. The picture remained on screen until participants responded. Assignments of ‘‘false” or ‘‘true” to a left or right key on the keyboard were fully balanced across participants. Trial order was fully randomized. Each of the eight portraits and eight cars was shown with two response types (true and false) and three levels of categorizations (superordinate, basic, and subordinate) resulting in 96 experimental trials. In the superordinate level and true condition, the category-word was either ‘‘artwork” or ‘‘vehicle”. In the basic level and true re- sponse condition, ‘‘portrait” or ‘‘car” and in the subordinate level and true condition the corresponding artist’s name (e.g., ‘‘Picasso” and ‘‘Van Gogh”) or the brand of the car (e.g., ‘‘VW and ‘‘Porsche”). In the false conditions category words were taken from a different exemplar of the same higher-order level category. For instance, a ‘‘Porsche” letter string and an ‘‘Audi” picture stimulus were paired, with both referring to the more inclusive level category ‘‘car”. In the basic level condition, a false word label that shared the same super- ordinate category was provided (e.g., the letter string ‘‘landscape painting” was presented with a ‘‘portrait” picture stimulus, with both referring to the superordinate category ‘‘artwork”). False trials were designed with the restrictions that each word-picture combi- nation at the subordinate level would appear only once during the experiment and each word within a level of categorization would appear with the same frequency in order to prevent response bias. In addition to the 96 experimental trials, the procedure con- tained 32 filler trials. The filler trials contained the foil categories bicycle and still life at the basic level of categorization and at two response types (true and false). True/false responses and reac- tion times were recorded. 3.2. Results The analyses were based on reaction times of correct true and (separately) correct false responses. Responses to the cars category showed that participants correctly responded ‘‘true” to 97%, 98%, and 77% of the trials for superordinate, basic, and subordinate lev- els, respectively. For false trials, participants responded correctly to 99%, 98%, and 87% of the trials for superordinate, basic and subor- dinate levels, respectively. To portraits, participants responded correctly to 99%, 83%, and 95% of the trials for superordinate level, basic level and subordinate level, respectively. For false trials, par- ticipants falsified correctly to 98%, 97% and 98% for superordinate, basic, and subordinate level categorizations, respectively. Reaction times were adjusted by setting boundaries to elimi- nate outliers. The lower boundary was set to 300 ms and the upper boundary was set to 3000 ms, which is equivalent to approxi- mately 2.5 standard deviations from the mean (MRT = 1098 ms, SDRT = 807 ms). Thus, 3.32% of correct experimental trials (in the true and false conditions) were regarded as outliers. 3.2.1. Analysis of true responses Mean adjusted RTs were analyzed as the dependent variable in a 2 x 3 repeated-measures analysis of variance with object domain (portrait and car) and level of categorization (superordinate, basic, and subordinate) as independent variables. The main effect for ob- ject domain was significant, F(1, 15) = 7.6, p = 0.015, g2p = 0.336, indi- cating that portraits were verified faster than cars. The main effect for level of categorization was also significant, F(2, 30) = 26.8, p < 0.001, g2p = 0.641, indicating slower responses for a more specific level of categorizations. Importantly, the critical domain � level interaction was also significant, F(2, 30) = 21.6, p < 0.001, g2p = 0.590. As shown in Fig. 2, the interaction indicates that cars were categorized faster at the basic level than at the subordinate level, F(1, 15) = 32.8, p < 0.001, g2p = 0.686. On the other hand, RTs for por- traits at the subordinate level (i.e., the artist’s name) and basic level did not differ significantly, F(1, 15) = 1.4, p = 0.254, g2p = 0.086, n.s. The interaction indicates that participants were faster to verify cars on the basic level than on the subordinate level, which confirms the assumption of a general basic level advantage (Rosch et al., 1976) for non-art-objects. In contrast, a different pattern was found for works of art. At the subordinate level (the artist’s name) works of art were categorized as fast as at the basic level (artistic genre). In sum, this task demonstrated that art-objects were verified differently from non-art-objects. People are just as fast to catego- rize artworks (with the artists’ names) at a level subordinate to the basic level, as they are to categorize artworks at the basic level. By contrast, verification times in the contrast category (‘‘car”) were faster at the basic level than at the subordinate level. This pattern of results is similar to findings in face perception, where verifica- tion times were as fast on the sub – as on the basic level for faces, while objects in the contrast category (‘‘dog”) were verified faster on the basic level (Tanaka, 2001). The results indicated that repre- sentations of artworks are highly accessible at a specific level of abstraction related to the artists’ names. This provides further evi- dence for an artist-specific style-based recognition of art-stimuli. 3.2.2. Analysis of false responses The results of responses in the false condition were in accordance with effects obtained in the true condition. Reaction times of correct falsification responses depending on object domain and level of cat- egorization are displayed in the lower panel of Fig. 2. The resulting mean reaction times were analyzed in a repeated-measures analysis of variance with object domain (portrait and car) and level of cate- gorization (superordinate, basic, and subordinate) as within-partic- ipant factors. The main effect of object domain was significant, F(1, 15) = 7.1, p = 0.019, g2p = 0.321, indicating that portraits were veri- fied faster than cars. The main effect of level of categorization was also significant, F(2, 30) = 11.4, p < 0.001, g2p = 0.431, indicating slower responses for more specific level of categorizations. The crit- ical object domain � level of categorization interaction was also sig- nificant, F(2, 30) = 35.6, p < 0.001, g2p = 0.704. This interaction indicates that cars were categorized faster at the basic level than at the subordinate level, F(1, 15) = 50.6, p < 0.001, g2p = .772, whereas portraits were categorized faster at the subordinate level than at the basic level, F(1, 15) = 9.8, p = 0.007, g2p = 0.393. 4. Experiment 3: Conceptual priming task The results from Experiments 1 and 2 suggested that partici- pants have detailed perceptual representations related to single Fig. 2. The results from Experiment 2 showing the mean response times collapsed by participants for categorizing portraits and cars at the superordinate, basic, and subordinate level in the true and false conditions. 3 According to biological taxonomies spiders are of course not insects but arachnids. However, it is common usage to identify spiders as insects in Germany , as documented in production norms (Mannhaupt, 1983). B. Belke et al. / Acta Psychologica 133 (2010) 191–202 197 artists that can be quickly accessed during initial processing. In Experiment 3, such perceptual representations of art-objects were directly examined using a conceptual priming task. Participants were presented with a word prime, a basic level prime, a subordi- nate level prime, or a neutral category prime (consisting of the letter string ‘‘blank”). Two simultaneously presented pictures followed the word primes. Participants were asked to judge whether the two pictures were identical or different. As suggested by Tanaka (2001), cross-modal priming effects were measured by the difference in reaction time between primed trials and neutral trials in the same picture conditions. According to Posner (1969) and Posner and Mitchell (1967), such an identity-priming paradigm allows for the investigation of participants’ visual repre- sentations that are activated by the word prime. The stronger the priming is, the shorter the reaction times will be. The degree of facilitation depends on the match between mental representations, as elicited by the word stimulus and its correspondence with the physical picture stimulus. Thus, ‘‘the closer the match between the mental representation and the visual percept, the faster the matching response” (Tanaka, 2001, p. 540). In comparison with Experiment 2, Experiment 3 considered a broader range of catego- ries. Art stimuli were selected from four artistic genres (portraits, landscapes, nudes, and still life paintings) and compared with four classes of every-day objects (cars, insects, furniture and dogs). Assuming that participants have developed elaborated repre- sentations related to artworks of single artists priming effects with artworks should be strongest at the subordinate level. For non-art- objects such effects were not expected. 4.1. Method 4.1.1. Participants Fourteen (12 female) students in art history from the Freie Uni- versität Berlin participated in the experiment. Mean age was 25.7 (SD = 2.02) ranging from 23 to 30 years. They were enrolled be- tween eight semesters (4 years) and 11 semesters (5.5 years) (mean = 9.92, SD = 1,20) in a MA degree in art history. None of the students participated in Experiment 1 or 2. Participants were tested individually and received allowance for participation. Each participant had normal or corrected-to-normal vision. 4.1.2. Stimuli Art stimuli were selected from four art categories, each contain- ing four paintings: nudes (Francois Boucher, Lucas Cranach the El- der, Rembrandt, and Peter Paul Rubens), landscapes (Paul Cézanne, Caspar-David Friedrich, Claude Monet, and Vincent van Gogh), por- traits (Leonardo da Vinci, Anselm Feuerbach, Jean-Honoré Frago- nard, and Pablo Picasso) and still life paintings (Jean-Baptiste Chardin, Albrecht Dürer, René Magritte, and Andy Warhol); see Appendix 3 for a full list of artists and titles. Non-art stimuli were selected from four categories, cars (Audi, Mercedes, BMW, and Volkswagen), arthropods (ant, fly, butterfly, and spider3) furniture (bed, couch, cupboard, and table) and dogs (Collie, Dalmatian, Poo- dle, and German Shepherd). Exemplars of the categories furniture, dogs, and insects were among the ten most mentioned, as indicated in the linguistic production norms for the German population (Mannhaupt, 1983). Car stimuli were examples of the ten most men- tioned brands in a recent market analysis (AC-Nielsen, 2004). Four exemplars were selected from each category resulting in 32 picture stimuli. Stimuli were standardized to 320 square centimeters with original width-to-height ratio maintained. 4.1.3. Procedure and apparatus Participants were instructed to judge whether two simulta- neously presented stimuli were identical or different. Afterwards, they performed eight practice trials followed by 192 experimental trials. Each trial began with the presentation of a fixation cross in the center of the screen for 1000 ms, which was then replaced with a word prime for 2500 ms. Subsequently, a 300 ms blank screen interval was shown and then followed by the simultaneous appear- ance of two pictures, which remained on screen until participants pressed one of two marked buttons (indicating that paintings were Table 1 Mean reaction times in milliseconds (and standard errors of the mean) depending on prime type and object category. Object category Neutral Basic level Subordinate level Cars 857 (69) 837 (46) 934 (68) Dogs 685 (42) 653 (46) 636 (35) Furniture 687 (31) 635 (29) 621 (38) Insects 738 (65) 655 (45) 681 (43) Landscape paintings 708 (47) 745 (53) 708 (53) Nude paintings 745 (31) 707 (46) 750 (47) Portraits 744 (49) 702 (44) 690 (55) Still lifes 824 (95) 791 (70) 660 (26) Domain A m ou nt o f P rim in g (m s) Art Non-Art 0 10 20 30 40 50 60 70 80 Fig. 3. The results of Experiment 3 showing the amount of facilitation for basic and subordinate level words for art-objects and non-art-objects. Error bars represent within-subjects confidence intervals (Loftus & Masson, 1994). 198 B. Belke et al. / Acta Psychologica 133 (2010) 191–202 either ‘‘the same” or ‘‘different”). Word primes appeared either at basic level, subordinate level, or in a neutral condition. Basic level primes for art-objects were ‘‘landscape”, ‘‘nude painting”, ‘‘por- trait”, and ‘‘still life”. Basic level primes for non-art-object were ‘‘car”, ‘‘furniture”, ‘‘dogs”, and ‘‘insects”. In the neutral word prime condition, the word ‘‘blank” was presented on the screen. The two pictures presented simultaneously in the ‘‘same” con- ditions were either two paintings or two non-art-objects. In the ‘‘different” conditions, the two pictures shared the same basic level (e.g., two different landscapes or two different cars) with the restrictions that each combination appeared only once and all stimuli appeared with equal frequency. Participants were asked to judge ‘‘as accurately and as fast as possible” whether the stimuli were the same or different by pressing one of two buttons with the right or left index finger. Assignment of ‘‘same” or ‘‘different” responses to the left or right key was balanced among participants. Key responses and reactions times were recorded by the experi- mental control software PsyScope 1.2.5 (Cohen et al., 1993), running on a Macintosh G4 with a 19” monitor (resolution 1024 � 768 pixels). Participants were tested individually. Thirty- two stimuli (art-objects and non-art-objects) combined with three levels of categorization primes (basic, subordinate, and neutral) in two response conditions (same or different) yielded a total of 192 experimental trials altogether. The visual angle was approximately 6.64� with a viewing distance of about 70 cm. 4.2. Results The following analyses were based on priming scores for correct ‘‘same” responses. Reaction times were adjusted based on data cor- rection as described in Experiment 2. Table 1 shows mean reaction times depending on prime level and object category. To measure the amount of priming effects, differences in reaction times were calculated between responses in the neutral conditions and correct the same responses in the priming conditions. To obtain individual priming scores for each object category and participant, differences in reaction times were calculated between the neutral condition and either the basic level condition or subordinate level condition. Mean priming scores for art- and non-art-objects were analyzed and compared. As carried out by Tanaka (2001, Experiment 4, p. 541) differ- ences were tested among the four art categories. Mean priming scores for category at the two levels of abstraction were subjected to a 4 (category: nude, landscape, portrait, or still life) � 2 (level of categorization: basic or subordinate) within-participants ANOVA. The main effect of level of abstraction was significant, F(1, 13) = 5.1, p = 0.042, g2p = 0.280. Neither the main effect for category, F(1, 13) < 1, p = 0.512, g2p = 0.057, n.s., nor was the interaction sig- nificant, F(3,39) = 2.1, p = 0.124, g2p = 0.136, n.s. Due to the lack of the main effect for category and interaction, the four art categories were collapsed to obtain one individual mean priming scores for the art-domain. To test differences among the four non-art-catego- ries, mean priming scores at the two levels of abstraction were submitted to a 4 (cars, dogs, furniture, insects) � 2 (basic, subordi- nate) within-participants ANOVA. Neither any main effects nor the interaction was significant (all Fs < 1.37). Consequently, non-art- object categories were collapsed to obtain individual mean prim- ing scores for the non-art domain. Priming scores (differences in RTs between neutral and treat- ment conditions) were collapsed and analyzed across participants. The amount of facilitation (based on reaction times) in the differ- ent conditions is shown in Fig. 3. A 2 � 2 repeated-measures analysis of variance was performed with object domain (art-object, non-art-objects) and level of catego- rization (basic, subordinate). This revealed a significant domain � level of categorization interaction by participants F(1, 13) = 8.3, p = 0.013 g2p = .391. No other effects were significant. The interaction indicates that additional priming effects were found at the subordi- nate level for art-objects but not for non-art-objects. Consistently, a post hoc comparison of priming effects at subordinate level and basic level primes showed a significant difference for art-objects, t(13) = 2.3, p = 0.043, g2p = 0.280. For non-art-objects the difference was not significant, t(13) = �1.1, p = 0.292, g2p = 0.085, n.s. Thus, peo- ple recognized artworks fastest with the artist’s names and were able to access elaborated artist-related visual representations when primed with a matching artist’s name. To summarize the findings of Experiment 3, subordinate prim- ing of the artist’s name facilitated a visual comparison task for the corresponding paintings by the artists, but when participants were primed with subordinate terms of non-art-objects, no such additional facilitation on recognition was observed. According to the logic of the word-picture priming task (the stronger the match between word and picture the faster the visual comparison reaction time) this result indicates that participants activated art- ist-specific visual representations triggered by the artist’s name. These primed representations could be either based on an artist’s characteristic style or alternatively, given that the selected art- works were typical examples by each of the painters, consist of ico- nic representations of individual works. Taking into account that a painter such as Picasso has produced a substantial range of well- known works, and participants could not anticipate which painting by the artist is being shown, it is reasonable to conclude that participants activated style-related representations that facilitated visual comparison. This explanation is in line with the notion that artists try to establish highly recognizable individual styles that are B. Belke et al. / Acta Psychologica 133 (2010) 191–202 199 also represented in the cognitive structure of the beholder (for a theoretical explanation, see Leder et al., 2004). In sum, consistent with the results of Experiments 1 and 2, the results of Experiment 3 suggest that art-objects are identified at a specific level of iden- tity and that detailed representations related to a single artist could be rapidly assessed. In order to provide additional empirical evidence, absolute priming effects for each priming condition were analyzed with four one-sample t-tests (sampled over subjects). The pattern of results confirmed the results of analysis of variance described above. Prim- ing facilitation for non-art-objects at the basic level was significant t (13) = 2.4, p < 0.001 g2p = 0.312 one-tailed, 4 whereas priming facilita- tion for non-art-objects at subordinate level was not t (13) < 1, p = .150, g2p = 0.058, n.s., one-tailed. Thus, every-day objects were identified at a basic level of abstraction. Critically, priming facilitation for art-objects at subordinate level (artist’s name) was significant t (13) = 1.9, p = 0.009, g2p = 0.199, one-tailed, whereas priming facilita- tion at basic level (artistic genre) was not, t (13) = 1.0, p = 0.124, g2p = 0.067, n.s., one-tailed. This test provided further empirical evi- dence that priming of artist’s name produces a significant amount of priming and suggested that participants possess fine-grained vi- sual representations linked to individual artists’ names. 5 The definition of an entry point as an object concept in this study is based on literature on conceptual hierarchies in the tradition of Rosch et al. 1976 (and many followers). In line with these approaches, the levels of abstraction and entry points in the categorization of objects were consistently defined and investigated as cognitive reference points. However, it seems likely, that such explicit classifications are preceded by perceptual analysis and implicit memory integration effects (see Leder et al., 2004 for a framework model of hierarchical processing stages) that may be linked to evaluations of the target stimulus (e.g., Scherer, 2003). Therefore, the early analysis of shape, form, content and style could be affectively experienced as e.g. ‘‘tender, disturbing, and fascinating”. The informational value of such affective evaluations for the recognition of artistic objects could be that they may serve as prototypical or 5. General discussion The purpose of the study was to present empirical evidence for a special art-related entry point in the identification of visual art. Converging empirical evidence from three experiments, which have proved sensitive to addressing the question of object identifi- cation suggested that initial recognition is at the level of the pro- ducing artist. In particular, the results of a speeded naming task (Experiment 1) showed that participants categorized a familiar painting with its matching artist’s name (subordinate level) more often than with any other art-related concept. In a category-verifi- cation task (Experiment 2), it was shown that works of art were processed as fast with the artist’s name (at the subordinate level) as with artistic genre (at the basic level). Verification times for ar- tists’ names were even faster compared to verification times of cat- egories at the subordinate level for non-art-objects. In a conceptual priming experiment (Experiment 3), artists’ names produced stronger RT facilitation compared to priming of artistic genre (basic level) terms. In this priming task, word-primes served as semantic cues in which visual representations specific to the subsequent stimulus had to be activated, in order to facilitate the visual match- ing response (cf. Posner, 1969; Posner & Mitchell, 1967; Tanaka, 2001). Therefore, the results suggested that participants possess of elaborate, fine-grained visual representations linked to individ- ual artists’ names. Given that each of the painters investigated in the present study has produced a substantial oeuvre of well-known single works of art, it seems implausible to assume that priming ef- fects resulted from the priming of single iconic representations of single artworks. Rather, the pattern of cross-modal priming effects is best explained by the activation of artist-specific memorial rep- resentations, which presumably consist of a kind of style-related visual prototype linked to individual artists. This interpretation is in accordance with a central assumption of a model of aesthetic experience – that processing of visual art comprises processing of style, which separates art perception from many other domains of perception (Belke & Leder, 2006; Leder et al., 2004). The idea that the identification of art is at the subordinate level of the corresponding artist is further sustained when considered in 4 One-sample t-tests were analyzed one-tailed because the priming hypothesis clearly predicts the direction of priming facilitation (decrease in RTs as compared to neutral conditions). light of a ‘‘cognitive economy” argument often made for the human cognitive system (e.g., Rosch et al., 1976). Such a classification is in between an ‘‘entry-level” of unique identity (Tanaka, 2001) such as titles (which might be too demanding, see Leder, Carbon, & Ripsas, 2006), and more inclusive art categories, such as genres and broad art schools (which might be inefficient for identifying a particular piece of art). Therefore, identification at the level of the producing artist may allow for an optimum level. Moreover, the often-high de- gree of semantic ambiguity (or even absence of any representa- tional information) often defies a clear semantic determinacy, which is why explicit classification based on artists’ specific styles might provide the most efficient in the identification of art. Given the wide range of artistic ‘strategies’ for depiction such as alien- ation, distortion and abstraction and therefore often highly stylized representations (of objects, themes or subject matters), the recogni- tion of visual art may most efficiently rely on reoccurring, salient stylistic features. These artist-specific prototypical style features might serve as ‘‘discontinuities” (Rosch et al., 1976) on which ‘‘ba- sic-cuts” in perception are made, and which provide an entry point in recognition. Due to these arguments and findings of the study, we propose that art has a special status amongst external-world ob- jects since it allows for a memorial representation based on stylistic features that are linked in semantic memory to the creating artist.5 Another contributing factor as to why the subordinate level pro- duced superior performance in the paradigms applied may be that perceptual similarity for art-objects does not increase with the level of specificity in the same way as it does for every-day object classes. For every-day objects research has shown that subordinate-level classifications require additional perceptual processing to extract features that are needed to go beyond the basic-level of processing (Op de Beeck, Béatse, Wagemans, Sunaert, & Van Hecke, 2000; Gau- thier, Anderson, Tarr, Skudlarski, & Gore, 1997). Compared to per- ceptual similarity within common object classes utilized in our study, paintings from a particular artistic genre might be generally more distinctive from each other. This increased distinctiveness might lower perceptual effort to discriminate amongst works of art and as a result favors a subordinate level of identification. Our study revealed a processing advantage on the subordinate level of the creating artist for viewers with a considerable range of academic training in the visual arts, varying from one year to six years of full-time studies. This result provides some evidence that an artist-based recognition reflects a processing characteristic that is not restricted to advanced levels of art expertise, but may be a hallmark of art perception itself. However, due to methodological reasons, performances of the free naming, category verification and visual identity task were not compared with performances by art- lay people, since these tasks required participants to possess of some declarative knowledge about artists, genres and styles. Although one could argue that viewers with only a passing famil- iarity with artists, styles or genres would show a classic basic level advantage (and identify an artwork at a medium level of abstrac- defining attributes (e.g., a painting by Mark Rothko may appear affectively ‘‘warm”, ‘‘floating” and spatially ‘‘immersive”, which might be crucial attributes for the artist’s memorial representation as a ‘‘Rothko”). If such affective and evaluative information is accessible in early processing stages, it might be possible that these affective attributes play a significant role in the identification of artistic objects. 200 B. Belke et al. / Acta Psychologica 133 (2010) 191–202 tion based on membership of a certain genre, broad art style or art school), we propose that the mental representation of visual art (due to the above-mentioned specificities) is generally organized around the subordinate level of the producing artist, even when art expertise is limited.6 This assumption is supported by Supplementary data which were obtained from people without formal art training and which we have not reported in this study. In a feature listing task people had to list as many attributes of given art-related categories as they could think of (see Rosch et al., 1976, for details of the method). The results indicated that art-lay people list significantly more items for the subordinate level terms of the artist’s name and spe- cific art-styles compared to basic level terms such as the artistic- genre. This result is in accordance with the assumption that art- specific declarative knowledge may be generally organized at the subordinate level of the producing artist. Clearly, future research is needed to investigate this hypothesis further. Moreover, relative levels of art expertise did not show up as a contributing factor on the finding that the artist’s name was the predominant category in naming art-objects in Experiment 1. Therefore, art expertise effects were not investigated further in Experiments 2 and 3, since they were not the primary focus of this study. However, it cannot be excluded that more advanced art expertise may modify the entry-level and provoke a downward shift in recognition (subsidiary to the artist’s name) to the level of a painting’s title, as the level of unique identity (Tanaka, 2001). The assumption of titles as an alternative (art-related) entry point was not investigated in Experiments 2 and 3, as frequencies of categorizing art-objects with their corresponding titles were rather marginal in Experiment 1, regardless of the level of partici- pant’s background in academic training in the arts. Future research may address the possible occurrence of such a downward shift more closely. The results also indicated that expert viewers were able to ac- cess artist-specific representations as fast as basic-level represen- tations of every-day objects, meaning that a painting by Picasso was identified as a ‘‘Picasso” within the same time frame as a depiction of a car was identified as a ‘‘car”. Although converging evidence of the three experiments suggested the artist’s name as a candidate for an entry point in art-recognition, we did not address the exact time-course involved, whether identification on the level of the artist precedes the identification of depictive content or vice versa. Therefore, we cannot rule out that the ini- tial contact between stimulus and semantic memory might be on the level of depictive content (Augustin et al., 2008). However, this might be possible for representational art but is less likely with Abstract, Minimal or Conceptual art, which often lack unam- biguous semantic references and often disrupt processing rou- tines that are prevalent in every-day perception, such as object recognition (Leder et al., 2004). Furthermore, art-related genre la- bels provided in Experiments 2 and 3 contained explicit refer- 6 It is important to note that exposure to art, unlike encounters with every-day objects, is a rather exclusive event and often restricted to a museum or gallery context; hence art can be regarded as an expertise domain per se. For many object categories a non-expert would still be able to name and identify an instance of such categories with an appropriate basic level and subordinate level concept. This is unlikely in the case of art, which represents a more specialised domain of knowledge. This study did not address the question how people without relevant art-specific categories and whose memorial representations have not been shaped by previous encounters with instances of a particular artist, genre or style, would identify an artwork. However, it seems likely that under these conditions, identification may more strongly rely on recognition of external-world object references (provided that representational information can be extracted from the work) and this may characterize a transfer of everyday processing strategies into the realm of art (Cupchik & Gebotys, 1988). Presumably, repeated exposure to works of art may foster the fast implicit learning of styles (Gordon & Holyoak, 1983) and the formation of style based memorial representations (Belke, Leder, & Augustin, 2006), which may evolve around single artists. ences to the paintings’ depictive content (e.g., the category ‘‘landscape painting” contained the word landscape) but were of lesser importance for identification than artists’ names. Future studies may address the exact time-course involved in content vs. style-based recognition and may reveal additional evidence (or boundary conditions) for the status of a subordinate artist- based recognition as the predominant entry point in the identifi- cation of visual art. Acknowledgements This research was supported by a Grant to Leder (SFB 626 C5) from the Deutsche Forschungsgemeinschaft (DFG) and by a Grant to Leder and Carbon by the FWF ‘‘Fonds zur Förderung der wissens- chaftlichen Forschung” (National Austrian Scientific Fund; P18910). We would like to thank Johan Wagemans, Slobodan Markovic and two anonymous reviewers for their valuable com- ments on an earlier version of this article. Appendix 1 List of paintings used in Experiment 1. Artist Title Style Paul Cézanne ‘‘Mont Saint-Victoire” Postimpressionism John Constable ‘‘Hampstead Heath” Realism André Derain ‘‘Landscape in southern France” Fauvism Caspar-David Friedrich ‘‘Solitary tree” Romanticism Claude Monet Detail of ‘‘Water lilies 1906” Impressionism Vincent van Gogh ‘‘Starry Night” Postimpressionism Wassily Kandinsky ‘‘Murnau” The Blue Rider Max Beckmann ‘‘Self-portrait with a cigarette” New Objectivity (Neue Sachlichkeit) Paul Klee ‘‘Senecio” Expressionism Gustav Klimt Detail of ‘‘Judith” Art Nouveau Roy Lichtenstein Detail of ‘‘That’s The Way—It Should Have Begun! But It’s Hopeless.” Pop-Art Edvard Munch ‘‘The Scream” Symbolism Pablo Picasso ‘‘Portrait de Dora Maar”1937 Synthetic Cubism Gerhard Richter ‘‘Emma” Photorealism Salvador Dali ‘‘Persistence de la Memoire” Surrealism Albrecht Dürer ‘‘The Hare” Renaissance René Magritte ‘‘Le Faux Miroir” Surrealism René Magritte ‘‘Ceci n’est pas une pipe” Surrealism Gerhard Richter ‘‘Skull 1983” Photorealism Vincent van Gogh ‘‘Sunflowers” Postimpressionism Andy Warhol ‘‘Campbell’s Soup” Pop-Art Jean-Honoré Fragonard ‘‘Diderot” Rococo Pablo Picasso ‘‘Portrait de Dora Maar” Synthetic Cubism Roy Lichtenstein Detail of ‘‘That’s The Way – It Should Have Begun! But It’s Hopeless” Pop-Art (continued on next page) B. Belke et al. / Acta Psychologica 133 (2010) 191–202 201 Appendix 1 (continued) Artist Title Style Leonardo da Vinci ‘‘De Benci” High-Renaissance Gustav Klimt ‘‘Judith” (part of it) Art Nouveau Anselm Feuerbach ‘‘Nanna” Classicism Paul Klee ‘‘Senecio” Expressionism Vincent van Gogh ‘‘Armand Roulin” Postimpressionism Appendix 2 List of paintings used in Experiment 2. Artist Title Style Jean-Honoré Fragonard ‘‘Diderot” Rococo Pablo Picasso ‘‘Portrait de Dora Maar” Synthetic Cubism Roy Lichtenstein Detail of ‘‘That’s The Way – It Should Have Begun! But It’s Hopeless.” Pop-Art Leonardo da Vinci ‘‘De Benci” High-Renaissance Gustav Klimt Cut-out of ‘‘Judith” Art Nouveau Anselm Feuerbach ‘‘Nanna” Classicism Paul Klee ‘‘Senecio” Expressionism Vincent van Gogh ‘‘Armand Roulin” Postimpressionism Appendix 3 List of paintings used in Experiment 3. Artist Title Style Francois Boucher ‘‘Reclining Girl” Rococo Lucas Cranach the Elder ‘‘Venus 1532” Early Renaissance Painting Rembrandt ‘‘Bathsheba” Baroque Peter Paul Rubens ‘‘Leda and the Swan” Baroque Paul Cézanne ‘‘Mont Saint- Victoire” Postimpressionism Caspar-David Friedrich ‘‘Solitary Tree” Romanticism Claude Monet ‘‘Water lilies 1906” Impressionism Vincent van Gogh ‘‘Starry Night” Postimpressionism Leonardo da Vinci ‘‘De Benci” High-Renaissance Anselm Feuerbach ‘‘Nanna” Classicism Fragonard ‘‘Diderot” Rococo Pablo Picasso ‘‘Portrait de Dora Maar” Synthetic Cubism J. B. S. Chardin ‘‘Silver Tumbler” Rococo Albrecht Dürer ‘‘The Hary” Renaissance René Magritte ‘‘Ceci n’est pas une pipe” Surrealism Andy Warhol ‘‘Campbell’s Soup” Pop-Art Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.actpsy.2009.11.007. References AC-Nielsen (2004). ‘‘Markenwert PKW 2003” – Eine Studie der Zeitungsgruppe BILD [‘‘Brand value of private cars 2003” – A study of the newspaper association BILD]. Augustin, M. D., Leder, H., Hutzler, F., & Carbon, C. C. (2008). Style follows content. On the microgenesis of art perception. Acta Psychologica, 128, 127–138. Battig, W., & Montague, W. (1969). Category norms of verbal items in 56 categories: A replication and extension of the Connecticut category norms. Journal of Experimental Psychology, 80, 1–46. Belke, B., & Leder, H. (2006). Annahmen eines Modells der ästhetischen Erfahrung aus kognitionspsychologischer Perspektive [trans. Assumptions of a model of aesthetic experience from a cognitive scientific perspective]. In Sonderforschungsbereich 626, Ästhetische Erfahrung: Gegenstände, Konzepte, Geschichtlichkeit [trans. Special research division 626, Aesthetic experience: Objects, Concepts, Historicity]. Belke, B., Leder, H., & Augustin, M. D. (2006). Mastering style. Effects of explicit style-related information, art knowledge and affective state on appreciation of abstract paintings. Psychology Science, 48(2), 115–134. Berlyne, D. E. (1971). Aesthetics and psychobiology. New York: Appleton-Century- Crofts. Carbon, C. C., & Leder, H. (2005). The Repeated Evaluation Technique (RET). A method to capture dynamic effects of innovativeness and attractiveness. Applied Cognitive Psychology, 19(5), 587–601. Cohen, J. D., MacWhinney, B., Flatt, M., & Provost, J. (1993). PsyScope: A new graphic interactive environment for designing psychology experiments. Behavioral Research Methods, Instruments, and Computers, 25(2), 257–271. Cupchik, G. C., & Gebotys, R. J. (1988). The search for meaning in art: Interpretive styles and judgments of quality. Visual Arts Research, 14(2), 38–50. Gauthier, I., Anderson, A. W., Tarr, M. J., Skudlarski, P., & Gore, J. C. (1997). Levels of categorization in visual recognition studied with functional MRI. Current Biology, 7, 645–651. Gauthier, I., & Tarr, M. J. (1997). Becoming a ‘greeble’ expert: Exploring mechanisms for face recognition. Vision Research, 37(12), 1673–1682. Gordon, P., & Holyoak, K. (1983). Implicit learning and generalization of the ‘mere exposure’ effect. Journal of Personality and Social Psychology, 45(3), 492–500. Grasskamp, W. (1994). Die unbewältigte Moderne. Kunst und Öffentlichkeit. [The unaccomplished Modernity. Art and Public]. München: Beck. Hartley, J., & Homa, D. (1981). Abstraction of stylistic concepts. Journal of Experimental Psychology: Human Learning and Memory, 7, 33–46. Hasenfus, N., Martindale, C., & Birnbaum, D. (1983). Psychological reality of cross- media artistic styles. Journal of Experimental Psychology, 9(6), 841–863. Jolicoeur, P., Gluck, M. A., & Kosslyn, S. M. (1984). Pictures and names – Making the connection. Cognitive Psychology, 16(2), 243–275. Leder, H. (2003). Familiar and fluent! Style-related processing hypothesis in aesthetic appreciation. Empirical Studies of the Arts, 21(2), 165–175. Leder, H., Belke, B., Oeberst, A., & Augustin, D. (2004). A model of aesthetic appreciation and aesthetic judgments. British Journal of Psychology, 95, 489–508. Leder, H., Carbon, C. C., & Ripsas, A.-L. (2006). Entitling Art: Influence of different types of title information on understanding and appreciation of paintings. Acta Psychologica, 121, 176–198. Loftus, G. R., & Masson, M. E. J. (1994). Using confidence intervals in within-subjects designs. Psychonomic Bulletin & Review, 1, 476–490. Mannhaupt, H. R. (1983). Produktionsnormen für verbale Reaktionen zu 40 geläufigen Kategorien [German category norms for verbal items in 40 categories]. Sprache & Kognition, 2, 264–278. Mervis, C. B., & Rosch, E. (1981). Categorization of natural objects. Annual Review of Psychology, 32, 89–115. Murphy, G. L., & Brownell, H. H. (1985). Category differentiation in object recognition: Typicality constraints on the basic category advantage. Journal of Experimental Psychology: Learning, Memory, and Cognition, 11(1), 70–84. Op de Beeck, H., Béatse, E., Wagemans, J., Sunaert, S., & Van Hecke, P. (2000). The representation of shape in the context of visual object categorization tasks. NeuroImage, 12, 28–40. Op de Beeck, H., & Wagemans, J. (2001). Visual object categorization at distinct levels of abstraction: A new stimulus set. Perception, 30, 1337–1361. Parsons, M. J. (1987). How we understand art: A cognitive developmental account of aesthetic experience. Cambridge, UK: Cambridge University Press. Posner, M. (1969). Abstraction and the process of recognition. In G. H. In, G. H. Bower, & J. T. Spence (Eds.), The psychology of learning and motivation. Oxford, England: Academic Press. Posner, M., & Mitchell, R. (1967). Chronometric analysis of classification. Psychological Review, 74(5), 392–409. Rosch, E. (1975). Principles of categorization. In E. Rosch & E. Lloyd (Eds.), Cognition and categorization (pp. 27–48). Hillsdale, NJ: Erlbaum. Rosch, E., Mervis, C., Gray, W., Johnson, D., & Boyes-Braem, P. (1976). Basic objects in natural categories. Cognitive Psychology, 8(3), 382–439. Scherer, K. (2003). Introduction: Cognitive components of emotion. In R. J. Davidson (Ed.), Handbook of affective sciences (pp. 563–673). Oxford: Oxford University Press. Segui, J., & Fraisse, P. (1968). Le temps reaction verbale. III Réponses spécifiques et réponses catégorielles à des stimulus objets Verbal reaction times. III Specific http://dx.doi.org/10.1016/j.actpsy.2009.11.007 202 B. Belke et al. / Acta Psychologica 133 (2010) 191–202 responses and categorical responses to stimulus objects]. Année Psychologique, 68(1), 69–82. Tanaka, J. W. (2001). The entry point of face recognition: Evidence for face expertise. Journal of Experimental Psychology: General, 130(3), 534–543. Tanaka, J. W., & Taylor, M. (1991). Object categories and expertise: Is the basic level in the eye of the beholder? Cognitive Psychology, 23(3), 457–482. Winston, A. S., & Cupchik, G. C. (1992). The evaluation of high art and popular art by naive and experienced viewers. Visual Arts Research, 18, 1–14. When a Picasso is a “Picasso”: The entry point in the identification of visual art Introduction Special features of art-objects The present study Experiment 1: Free Naming task Method Participants Stimuli Procedure Results and discussion Data correction Analysis of variance Type of categories applied to categorize art-objects Experiment 2: Speeded category-verification task Method Participants Stimuli Procedure Results Analysis of true responses Analysis of false responses Experiment 3: Conceptual priming task Method Participants Stimuli Procedure and apparatus Results General discussion Acknowledgements Appendix 1 Appendix 2 Appendix 3 Supplementary data References