Reader characteristics, behavior, and success in fiction book search

Reader Characteristics, Behavior, and Success in Fiction Book Search Mikkonen Anna School of Information Science, University of Tampere, Kalevantie 4, Tampere 33100, Finland. E-mail: anna.mikkonen@uta.fi Vakkari Pertti School of Information Science, University of Tampere, Kalevantie 4, Tampere 33100, Finland. E-mail: pertti.vakkari@uta.fi We examined the search behaviors of diverse fiction readers in different search scenarios. The aim was to understand how fiction readers with varied reading preferences are selecting interesting novels in library catalogs. We conducted a controlled user study with 80 participants. Two reader groups were elicited according to similar reading preference patterns. The readers enjoyed the entertainment, escape, and comfort that reading as a pleasurable activity offered. The aesthetic readers valued the artistic and aesthetic pleasures, widening vocabulary, and ability to express oneself through fiction books. We compared the search queries and search actions between the 2 reader groups. Our results demonstrated that preference patterns were associated with readers’ search behavior, that is, the number of viewed search result pages, opened book pages, dwell time on book pages, and the type of search queries. Based on the findings, we present 3 search tactics for fiction books in library catalogs: i) focused querying, ii) topical browsing, and iii) similarity-based tactic. The most popular search tactic in each search scenario was “focused querying” with known author in both reader groups. Introduction While the Internet and modern technologies are substantially changing people’s reading behavior (Bergstrom & Hoglund, 2014; Shimray, Chennupati, & Chennupati, 2015; Zhang & Kudva 2014), the book still remains the medium of choice for millions of people for leisure reading (Zhang Received August 3, 2015; revised November 23, 2016; accepted November 26, 2016 C 2017 ASIS&T  Published online 0 Month 2017 in Wiley Online V Library (wileyonlinelibrary.com). DOI: 10.1002/asi.23843 & Kudva, 2014). In various studies (i.e., Maryl, 2008; Ooi & Liew, 2011, Usherwood & Toyne, 2002) fiction readers have tried to articulate their love of reading by describing a range of motives for reading: reading purely for enjoyment, reading to seek specific emotions, relaxation or escape, or reading for personal development and enhancing societal consciousness. Either way, the habit of reading for pleasure is thriving (The Reading Agency, 2013). Public libraries are a major source for fiction, providing both printed fiction books and expanding fiction ecollections for diverse readers (The Reading Agency, 2013; The State of America’s Libraries, 2014). Fiction collections have become increasingly available in library catalogs and digital libraries: readers are searching for and reserving known titles and aiming at serendipitous book discoveries. Instead of asking a librarian, hints of interesting fiction books are searched from the user-generated content and enriched metadata in online library catalogs (Tang, Sie, & Ting, 2014; Vakkari, Luoma, & P€ontinen, 2014). Given the growing availability of fiction e-collections and the increasing number of readers with varied search competencies accessing library catalogs, an understanding of fiction readers’ search behavior is essential to create functioning interfaces for fiction. Previous research (i.e., Goodall, 1989; Pejtersen, 1989; Saarinen & Vakkari, 2013) has provided some information on how readers select fiction in public libraries. However, we have only a few studies (i.e., Tang et al., 2014; Vakkari et al., 2014) examining fiction readers’ search behavior in library catalogs. The purpose of this study is to create a fiction reader typology and to examine the search behaviors of diverse fiction readers. The aim is to understand how fiction readers with varied reading preferences are selecting interesting novels in different search scenarios. JOURNAL OF THE ASSOCIATION FOR INFORMATION SCIENCE AND TECHNOLOGY, 00(00):00–00, 2017 The study addresses the following research questions: RQ 1. Is there an association between reader characteristics and search actions in different search tasks? If yes, how? RQ 2. Is there an association between reader characteristics, search actions, and search success in different search tasks? If yes, how? RQ 3. Is there an association between reader characteristics and the type of queries in different search tasks? If yes, how? RQ 4. What are the typical search tactics applied in fiction searching? A search action is an identifiable basic action (i.e., conducting a keyword search, a click to view a search result page or a book page) during the search process (Bates, 1990). A search task is a sequence of actions with the goal of finding specified information (Ingwersen & J€arvelin, 2005). A query is a formulation of a precise request for information in the form required by the information retrieval technique (Ingwersen & J€arvelin, 2005). To address the study objectives, we conducted an experiment with fiction readers working on various search scenarios to find fiction books. The study provides information on the search behaviors of readers with varied reading interests. In addition, the study enhances understanding of typical search tactics for fiction books in library catalogs. Related Work Searching for Fiction in Library Catalogs The theoretical background of fiction book search is associated with everyday-life information seeking and with the serious leisure perspective (Stebbins, 2007). In the serious leisure perspective (SLP), leisure is defined as an unforced, contextually framed activity engaged in leisure time, which people do in either a satisfying or a fulfilling way. Casual leisure is a form of leisure encompassing informationseeking behavior that is a pleasurable activity requiring little training to enjoy it (Stebbins, 1997, 2007). Searching for fiction books may be considered as a casual leisure activity that is not motivated by an actual information need or a task to be solved. Instead, personal and hedonistic needs engage fiction readers in often exploratory searching behavior that combines querying and browsing strategies for investigating interesting books to read (Elsweiler, Wilson, & Kirkegaard Lunn, 2011; Marchionini, 2006). Previous studies examining fiction readers’ search behavior have focused on identifying typical search tactics in fiction retrieval (Pejtersen, 1989), on examining visual and social browsing tools in a fiction book search (Jiang, 2013; Thudt, Hinrichs, & Carpendale, 2012), and on comparing fiction readers search behaviors between a traditional and an enriched library catalog (Oksanen & Vakkari, 2012; Vakkari & P€ ontinen, 2015). Pejtersen (1989) identified four search tactics in fiction searching in online catalogs. A bibliographical search tactic is used when readers are searching for a 2 known item or author. An analytical search tactic is used when readers wish to access books about a topic such as multiculturalism. A search by analogy is generated when readers want something similar to a novel they have previously read. A browsing tactic is applied in situations when readers have only a vague idea of what they would like to read (Pejtersen, 1989). In The Bohemian Bookshelf, introduced by Thudt et al. (2012), serendipitous discoveries in a library catalog were enabled through information visualization offering various perspectives to the fiction collection. The system was found to especially support a browsing search tactic. Jiang (2013) examined the users of a social library system and noticed that in addition to the internal search engine, catalog browsing, associative browsing, and social browsing were popular search strategies. Oksanen and Vakkari (2012) found that effort invested in inspecting search results and book pages instead of querying was an essential factor for finding interesting novels in browsing situations. Vakkari and P€ontinen (2015) evaluated users’ search result page (SERP) browsing patterns between an enriched and a traditional library catalog. The results showed that the enriched catalog supported users to identify potentially clickable items on the results list sooner and more effectively compared to a traditional library catalog. Fiction Reader Characteristics Previous studies (i.e., Miesen, 2003; Ooi & Liew, 2011; Ross, 2001; Usherwood & Toyne, 2002; Yu & O’brien, 1999) have revealed a wide variety of motives for reading imaginary literature. Reading literary novels has been detected to fulfill affective needs, such as enjoyment and entertainment (Miesen, 2003; Ooi & Liew, 2011). Ross (2001) suggested that an affective dimension was involved in book selection from choosing a book according to mood, to valuing a book for its emotional support in providing confirmation, courage, or self-acceptance. Alongside the enjoyment of the reading process, utilitarian outcomes such as reading fiction for learning and practical knowledge appear to be equally important (Miesen, 2003; Usherwood & Toyne, 2002). In addition, escapism has been named as the most conscious perception that readers derive from the act of pleasure reading. For readers, escapism means actively being involved in other worlds, but also escape through the aesthetic pleasure that the works of literature offered (Usherwood & Toyne, 2002.) Saarinen and Vakkari (2013) distinguished between entertainers, esthetes, and realists when studying fiction readers’ book selection in a public library. Entertainers were seeking relaxation and distraction from daily routines from pleasure reading. Entertainers expected that the plot of the novel would carry along, and the characters and actions would be fascinating. Esthetes were looking for the aesthetic pleasure that the works of literature offered. The artistic and esthetic experience produced by a novel was the major criterion of a good novel for the esthetes. Realists were reading JOURNAL OF THE ASSOCIATION FOR INFORMATION SCIENCE AND TECHNOLOGY—Month 2017 DOI: 10.1002/asi to deepen their knowledge about certain subjects. Realists preferred especially realistic texts and historical novels. Research Method Participants In order to achieve a sample of participants with fiction reading interest, 80 persons were recruited in public libraries, in fiction reading groups, and in writing and literature classes in the Open University of Finland. The Snowball method and a newspaper advertisement were used. A movie ticket was offered as compensation for participation. Despite the expenditure of a great deal of effort invested in recruiting participants, 80 participants were reached to take part in the study. The 80 participants were considered a group large enough to conduct various between-group analysis. Thus, the number of participants was cut off at 80. If a participant had not read a single fiction book during the previous 12 months and reported personal disinterest in fiction reading, they were excluded from the study due to lack of genuine fiction reading interest. Participants were randomized into two groups. Half of the participants completed four search tasks in a typical public library catalog, Satakirjastot (https://www.satakirjastot. fi), while the other half used BookSampo (www.kirjasampo. fi), a catalog designed for fiction search. Background variables (i.e., gender and experience in using online catalogs) did not significantly differ between the groups. In both groups, 18% of the participants were males and 82% females. Eighty-one percent of Satakirjastot (Sata) users and 58% of BookSampo (Sampo) users utilized online catalogs for fiction searching at least once in every 3 months. The age distribution of participants varied from 18 to 78 in both groups (Figure 1). Sata users averaged 34 years of age, whereas Sampo users averaged 42 years of age. The age distribution of participants in Sampo was significantly higher compared to the age distribution of participants in Sata (p 5 .015). However, as there was no significant difference in participants’ experience in using online catalogs between the groups, the difference unlikely affected the results. Search Scenarios Three search tasks were designed based on previous research (Goodall, 1989; Pejtersen, 1989; Spiller, 1980; Thudt et al., 2012; Yu & O’brien, 1999). In one search tasks the participants were given an author with which to begin the search process. The remaining two search tasks reflect the idea of individual information needs, as the participants were asked to proceed without any given topic. Simulated search tasks were as follows: Known author search: A friend of yours recommends that you should get to know Olli Jalonen’s novels. Find Olli Jalonen’s books and choose two novels that you find interesting. FIG. 1. Age distribution of participants. Open-ended browsing: Find three novels that interest you that you would like to read. Search by analogy: Think of and mention one novel that you have read and found interesting recently. Now search for three novels that you would consider similarly interesting as the one you mentioned. The Catalogs Used As a traditional catalog, Satakirjastot service was used (https://www.satakirjastot.fi). As an enriched catalog the BookSampo service was used (www.kirjasampo.fi). Satakirjastot (Sata) is the web service of the city libraries of the Satakunta region in Finland. The service consists of a library catalog and an information retrieval system from given databases. In Sata, the metadata for fiction contains bibliographic information added with subject terms from the fiction thesaurus Kaunokki. Cover images and blurbs from recently published books are also available. BookSampo (Sampo) is an enriched web service for fiction in Finland. Compared to the catalog Sata, Sampo provides fiction-related metadata that is more varied and accounts for different access points to literature. In Sampo, the associations between the works of literature are realized by semantic web technologies such as the ontologization of the fiction thesaurus Kaunokki (Hypen & M€akel€a, 2011). At the front page of Sampo, users can begin the search process with a basic search (similar to baseline catalog) or with various visual and social browsing options. For a detailed description of the catalogs used, see Mikkonen & Vakkari (2015). In Mikkonen & Vakkari (2015, 2016a), differences in participants’ search behaviors were compared between the two catalogs. In these studies, participants’ search success JOURNAL OF THE ASSOCIATION FOR INFORMATION SCIENCE AND TECHNOLOGY—Month 2017 DOI: 10.1002/asi 3 TABLE 1. Four stages of the data analysis. Stage of the analysis Data Analysis method 1. Creating the reader types 2. Analyzing search queries Pre-questionnaire. An excel file containing query terms of each issued query identified from search logs. 3. Coding the variables measuring search process 4. Analyzing search actions and search queries by reader type Search logs from on-screen activity. Measures from previous stages: Reader groups Variables for search actions Variables for search queries Statistical analysis: K-means cluster analysis. Qualitative content analysis Classification of the search queries into  six categories according to query content  two categories according to query success. Qualitative preanalysis. Manual coding of the variables measuring the search process. Statistical analysis Search actions: frequencies, means, standard deviations in each reader group, t-tests between the reader groups. Search queries: frequencies and proportions of queries in each query category by reader type, v2-tests between the reader groups. was found to be equal in both systems. In this study, the limitations inherent in the number of participants prevented a simultaneous comparison of 80 participants between two catalogs and two reader groups, as it would have reduced the number of cases in each group overly small to conduct a reliable statistical analysis. Thus, the data from the two user groups were merged and the search behaviors were compared between reader groups. For a detailed description of differences in search behaviors between catalogs, see Mikkonen & Vakkari (2015, 2016a). Experimental Procedure The experimental setting shared similarities with the work of Vakkari and P€ontinen (2015) and Vakkari et al. (2014). In those studies, the instructions and posttask questions for fiction search tasks were presented in a similar fashion as in this study. Information of the above-mentioned experiment was used when designing the experimental setting of this study. The experimental setting was pretested with one participant to gain information on the duration of the test as a whole and to see if the instructions were unambiguous enough. The search logs were saved with a Morae Recorder (http://www.techsmith.com/morae.html). The Morae Recorder captures audio, video, and on-screen activity during a research session. Variables measuring the search actions were manually calculated from on-screen activity after conducting the user test. The audio was used in analyzing and classifying participants’ search queries. In each task the participants were asked to search for three novels that were of interest to them except in “known author task.” In this task the participants were asked to find only two interesting novels, because the production of the selected author was relatively limited. “Known author” task functioned as a training task at the beginning of the experiment. The experiment consisted of the following steps: 1. Prequestionnaire (demographic questions, participant’s search experience in online catalogs, questionnaire for reading preferences). 4 2. 3. 4. 5. 6. Introduction to the experiment. Demo of the retrieval system. Execution of search tasks. Posttask questionnaire after each completed search task. Postsession questionnaire after the completion of all search tasks. 7. Brief posttask interview. Based on the findings by Miesen (2003) and Saarinen and Vakkari (2013), a prequestionnaire measuring participants’ reading preferences and motives for fiction reading was designed (Appendix). In the posttask questionnaire, participants were asked to rank the novels found according to how much they were of interest to them on a scale from 1 to 3, where 1 was “only a little interesting,” 2 was “quite interesting,” and 3 was “very interesting.” Scoring 0 was used if an interesting novel was not found. The time for completing the tasks was not limited. Each participant completed the tasks individually. Latin square rotation was used with the tasks. The Latin square technique ensures that each task appears at each position equally often and that each task precedes and follows every other task equally often (Carter & Lubinsky, 2016). During the experiment, the researcher was present to help in case technical problems occurred. Analysis Both qualitative and quantitative data were analyzed. The four stages of the data analysis are given in Table 1. Profiles of Reader Groups Reader groups were created based on the reading preferences questionnaire (Appendix). To identify groups of participants sharing similar responses on the questionnaire, and to distinguish different preference patterns in fiction reading, we conducted a cluster analysis. The aim of cluster analysis is to identify homogenous groups based on their shared characteristics. The clustering was guided by the procedures outlined by Hair, Anderson, Tatham, and Black (1998). First, the cases with missing data were excluded from the analysis JOURNAL OF THE ASSOCIATION FOR INFORMATION SCIENCE AND TECHNOLOGY—Month 2017 DOI: 10.1002/asi TABLE 2. Cluster means for the two-cluster solution of the k-means cluster analysis. Variables Mean Entertainment Artistic experiences A possibility to experience beauty Intellectual stimulation Widens reader’s view of the world Insight into foreign culture Practical solutions to everyday life Understanding about the world New perspectives Rich and elaborate language Interesting characters Thought-provoking An original style Cluster characteristics Cluster age Cluster n Males n (%) Females n (%) Sata users n (%) Sampo users n (%) Cluster 1 “An entertained reader” Cluster 2 “An aesthetic-utilitarian reader” 3.85 2.42 2.21 2.39 3.30 3.09 1.70 2.52 2.88 2.91 2.70 3.15 2.06 3.58 3.36 3.40 3.62 3.98 3.84 2.87 3.56 3.82 3.51 3.53 3.78 2.73 35.1 33 7 (21.2) 26 (78.8) 18 (54.5) 15 (45.5) 45.6 45 8 (17.8) 37 (82.2) 20 (44.4) 25 (55.6) (two cases). Since the variables used in the cluster analysis did not vary in range, standardizing prior to clustering was not needed. Then, a k-means cluster analysis was conducted. A k-means cluster analysis is a procedure that is applied when the number of clusters is predefined (Hair et al., 1998). Considering the n-number and findings from previous research (Miesen, 2003; Saarinen & Vakkari, 2013; Yu & O’brien, 1999), two clusters was hypothesized as a rational solution. First, each variable in the preference questionnaire was included in the analysis. Then, the variables that did not significantly (p > .05 in the variance analysis) contribute to grouping the participants to clusters were excluded from the analysis. The final two-cluster solution contained 13 variables and seemed to be a good fit to the data (Table 2). In the final cluster solution, the cluster means yielded a difference in reading preferences between Clusters 1 and 2. Participants in Cluster 1 seemed to appreciate the entertainment provided by fiction reading, while participants in Cluster 2 were high in aesthetic, artistic, and utilitarian aspects offered by fiction reading. Therefore, Cluster 1 was named “An entertained reader” and Cluster 2 “An aestheticutilitarian reader.” To further test the validity of the cluster solution, we analyzed the differences in reading preferences between the clusters with an independent samples t-test. The results showed significant differences between the two clusters on the dependent measures (Table 3). The data confirmed the differences between clusters and the validity of the cluster solution was supported. The two reader groups identified were the entertained and the aesthetic-utilitarian readers, referred to as the entertained and the aesthetics in the remainder of the paper. Readers in both groups shared similar interests in terms of TABLE 3. Significance testing of cluster differences in reading preferences. Variable A possibility of daydreaming Entertainment Artistic pleasures Escapism Aesthetic pleasures Experiencing beauty Widens vocabulary and ability to express oneself Intellectual stimulation Insight into foreign cultures Practical solutions Understanding of the world New perspectives Factual information about the world Rich and elaborate language Gripping plot Interesting characters Thought-provoking Original style Challenge for its reader Entertained mean Aestheticutilitarian mean p 2.6 3.5 2.4 3.6 2.7 2.2 2.5 3.0 3.2 3.4 3.8 3.5 3.4 3.6 .066 .079 .000 .410 .000 .000 .000 2.2 3.1 1.7 2.5 2.9 1.7 2.8 3.8 2.9 3.6 3.8 1.5 .008 .000 .000 .000 .000 .210 2.9 3.7 2.7 3.2 2.1 2.2 3.5 3.5 3.5 3.8 2.7 2.9 .000 .416 .000 .000 .002 .003 entertainment, the possibility of daydreaming and escapism provided by fiction reading. When reading novels, readers in both groups valued novels offering factual information about the world and a gripping plot. However, the artistic and aesthetic pleasures, intellectual stimulation, and experiencing beauty were highlighted significantly more among the aesthetics compared to the entertained. In addition, widening vocabulary and developing a better understanding about the world through fiction books were emphasized more among the aesthetics compared to the entertained. The aesthetics preferred novels containing interesting characters and original style combined with rich and elaborate language. They sought challenging oneself and one’s attitude towards everyday life through reading thought-provoking novels. The entertained particularly enjoyed escape and comfort that reading as a pleasurable activity offered. They preferred a gripping plot, interesting characters, and having an insight into foreign cultures while reading a novel. The entertained enjoyed reading simply to have enjoyment without a necessity to analyze a text or a language in detail. Search Queries The query terms were manually identified and saved in the search logs. The number of terms in each query was calculated and the query was classified either as a successful one or as an unsuccessful one. A query was considered successful if it resulted in a selection of a novel. A query was considered unsuccessful if it resulted in no selections. Next, the type of each query was qualitatively analyzed and classified in one of the following categories: author, title, genre, topic, person, place, or time frame. Each query was placed JOURNAL OF THE ASSOCIATION FOR INFORMATION SCIENCE AND TECHNOLOGY—Month 2017 DOI: 10.1002/asi 5 TABLE 4. The measures for search actions and search success. Measures for search actions # of search actions # of search queries # of visits to SERPs # of opened book pages Time per a query Time per a SERP visit Time per a book page Measure for search success Reader type Average book scores per task in one category only. The categories were taken from the Finnish fiction thesaurus Kaunokki (http://onki.fi/en/ browser/overview/kauno), which was used in fiction indexing in both catalogs used in the experiment. The classification of search queries was conducted by a single coder. Intracoder reliability, which refers to the consistent manner by which the researcher codes, was considered during the coding process. In order to test the ability of the coding protocol to result in a consistent categorization of content, the classification of search queries was conducted two times by the same coder. During the coding, it was assessed whether any incoherence had occurred in the single coder’s understanding or application of the protocol definitions. In addition, data sessions with fellow researchers were held during the coding process. In these sessions, the coding protocol was assessed and discussed with other researchers. The type of search queries was typically unambiguous, as the majority of the queries contained query terms from one query category only. In cases where contradictions occurred, the query was placed in a category that was interpreted as representing a searcher’s main target by listening to the audio data. The search queries were statistically analyzed in each task and in both reader groups. The proportion of queries in each category was separately calculated for successful and unsuccessful queries. v2-tests were used to test for significant differences in query type between the reader groups. Measures for Search Actions and Search Success Search action was defined as an identifiable basic action during the search process (Bates, 1990). Search actions in the 80 user tests were manually coded by a single coder. The identified search actions were measured by their frequency and duration (Table 4). Success in search tasks was measured by the books’ interest grading given by participants. Average book scores per task were calculated. An independent samples t-test was used to test differences in search actions between reader types in each search scenario. Reader Characteristics, Search Actions, and Search Queries Known Author Task In “known author task,” the aesthetics selected more interesting novels with fewer search actions compared to the 6 TABLE 5. Search actions in “known author task” by reader type. Search actions # search actions # queries # SERP visits # book pages Time per a query (sec) Time per a SERP visit (sec) Time per a book page (sec) Average book scores Entertained mean (SD) 15 1.3 5.0 5 18 35 21 3.4 (10) (0.6) (3.7) (5.2) (14) (20) (14) (0.7) Aesthetics mean (SD) 10 1.2 3.5 3 25 54 24 3.8 (6) (0.4) (2.7) (2.9) (18) (40) (20) (0.6) t-test p .013 .210 .041 .032 .051 .008 .586 .012 TABLE 6. Search actions in “open-ended browsing” by reader type. Reader type Search action # search actions # queries # SERP visits # book pages Time per a query (sec) Time per a SERP visit (sec) Time per a book page (sec) Average book scores Entertained mean (SD) 24 2.7 6.8 6.5 31 50 14 6.0 (13) (2.8) (4.5) (4.9) (24) (113) (10) (0.8) Aesthetics mean (SD) 20 3.5 9.0 4.7 30 37 25 6.0 (10) (2.6) (4.7) (3.5) (20) (30) (13) (0.7) t-test p .172 .205 .048 .057 .891 .335 .001 .868 entertained (Table 5). The aesthetics made significantly fewer SERP visits, but devoted significantly more time per SERP compared to the entertained. Moreover, the aesthetics made the selections by opening significantly fewer book pages compared to the entertained. The posttask questions showed that the aesthetics were more familiar with the given author’s literary production compared to the entertained. The difference was nearly significant (t-test p-value .058). Open-Ended Browsing In “open-ended browsing,” the reader groups were equally successful in selecting good books (Table 6). The aesthetics visited search results significantly more frequently, but a single SERP visit was shorter compared to the entertained. Moreover, the aesthetics opened fewer book pages, but invested significantly more time to a single book page compared to the entertained. It seems that the aesthetics skimmed through a larger amount of SERPs more quickly, but opened fewer titles for a longer time for a closer examination compared to the entertained. Search by Analogy In “search by analogy,” the reader groups were equally successful in selecting interesting novels. Neither of the search actions differed between the two reader types (Table 7). In both reader groups, searchers tended to begin the task by typing the name of a previously read title as a first entry term. They proceeded to examine the subject terms in the JOURNAL OF THE ASSOCIATION FOR INFORMATION SCIENCE AND TECHNOLOGY—Month 2017 DOI: 10.1002/asi metadata of the previously read title, and then used the subject terms as an entry at the subsequent query. Both reader groups made several SERP visits and opened many book pages in order to identify similarity between the titles. Both reader groups devoted more time to SERPs compared to book pages. more often compared to the entertained. In both reader groups, 15% of the successful queries consisted of a genre, such as “crime fiction.” In “search by analogy,” the majority of the queries were successful among the entertained. Among the aesthetics, half of the queries were successful. In both groups, the most popular approaches in finding similar novels were to issue either a topic or an author of the novel as an entry term. When an author was used as an entry term, the searchers aimed at similarity between novels by selecting novels by the same author. When a topic was used as an entry term, the searchers aimed at similarity between novels by selecting titles covering similar subjects. A typical search approach in both reader groups was to pick interesting topics from the metadata of previously read novels and proceed to search for titles with similar topics. An observation of participants revealed that the subsequent search actions were different between the reader groups: the aesthetics tended to proceed by issuing new queries with the identified topics as an entry, while the entertained preferred proceeding without issuing new queries. Instead, the entertained utilized automatic recommendations to continue the search process. Search Queries and Success In total, the data consisted of 770 queries, of which 445 were successful and 325 unsuccessful. In each task the queries contained typically two query terms. In “known author task,” the majority of the queries were successful (over 90% in each group). In each query the author was used as an entry term. In unsuccessful queries the query contained solely the author’s surname, which produced results also from other authors with the same surname. These queries were reformulated by adding the author’s first name into the query, which resulted in success. In “open-ended browsing,” the majority of the queries were successful in each reader group. The most popular approach in both groups was to issue an author as an entry term (Table 8). A greater proportion of the queries containing an author name resulted in success among the aesthetics compared to the entertained. Interestingly, among the aesthetics queries with a title did not lead to book selections, whereas 14% of the title queries were successful among the entertained. The aesthetics searched for good books by issuing topical queries (such as “friendship” or “universities”) Discussion Search Behaviors of the Aesthetics and the Entertained The findings of this study are consistent with previous research on fiction reader characteristics. As in previous studies (i.e., Miesen, 2003; Ooi & Liew, 2011; Saarinen & Vakkari, 2013), the participants in this study enjoyed the entertainment, escape, aesthetic pleasure, and utilitarian aspects that reading as a pleasurable activity offered. We elicited different preference patterns in fiction reading and created two fiction reader types: the entertained and the aesthetics. Our typology for fiction readers is in line with Saarinen and Vakkari (2013), where fiction readers were divided into entertainers, esthetes, and realists. Previous research on the search behavior of different fiction readers in library catalogs is scarce. Our study revealed significant differences in search behaviors between the two fiction reader groups. The major differences in search actions between the entertained and the aesthetics occurred TABLE 7. Search actions in “search by analogy” by reader type. Reader type Entertained mean (SD) Search action # search actions # queries # SERP visits # book pages Time per a query (sec) Time per a SERP visit (sec) Time per a book page (sec) Average book scores 25 3.1 8.0 7.2 31 40 21 6.0 (16) (3.4) (5.6) (6.0) (20) (41) (16) (0.6) Aesthetic mean (SD) 21 3.0 7.5 5.6 30 38 21 6.0 (14) (2.2) (4.4) (4.6) (23) (31) (9) (0.8) t-test p .254 .910 .697 .761 .889 .820 .972 .767 TABLE 8. Success and query types in “open-ended browsing” and “search by analogy” (%). Open-ended browsing Successful queries Search by analogy Unsuccessful queries Successful queries Unsuccessful queries Query type E (n 5 51) A (n 5 82) E (n 5 34) A (n 5 60) E (n 5 48) A (n 5 59) E (n 5 35) A (n 5 62) Author Title Topic Place Genre Time frame Total v2 p-value 41 14 21 10 12 2 100 42 0 42 1 15 0 100 49 15 18 3 15 0 100 26 9 45 0 19 1 100 25 27 32 5 11 0 100 37 8 37 2 14 2 100 17 20 26 5 29 3 100 23 21 41 0 10 5 100 .001 .044 .066 .087 E 5 the entertained; A 5 the aesthetics. JOURNAL OF THE ASSOCIATION FOR INFORMATION SCIENCE AND TECHNOLOGY—Month 2017 DOI: 10.1002/asi 7 in the number of SERP visits and opened book pages, and in the time devoted to SERPs and opened titles. Saarinen and Vakkari (2013) found that accessing novels of a known author was perceived equally easy among occasional and avid readers. Our study demonstrated that in “known author task,” the aesthetics’ familiarity with the given author’s literary production supported them to select more interesting novels with less effort compared to the entertained. Instead of opening several book pages, the aesthetics devoted much time to examining SERPs and tended to make the selections based on the information provided in the results list. The entertained clicked several items on SERPs to view the details, and made the selections based on the metadata provided on title pages. It seems that the result list provided enough information for the aesthetics to make the selections, whereas title page examination was needed more among the entertained to assess the novels’ interest level. This hints that a reader’s literary knowledge influences search actions when selecting novels of a given author, as knowledge of the author’s literary production provides an advantage to identify more interesting novels compared to selecting titles of a previously unknown author. In “open-ended browsing,” the reader groups were equally successful in selecting good books. The entertained often picked a familiar author’s novels, whereas the aesthetics preferred to search for previously unknown novels of a particular topic. This seems to be in line with Tang et al. (2014), who found that highly involved fiction readers were willing to explore novel book recommendations and expand their reading horizon, whereas readers with high knowledge of their reading preferences preferred an author browsing strategy. The entertained identified interesting titles by few SERP visits and numerous opened titles, whereas the aesthetics made several short SERP visits and opened few titles for closer examination. It seems that when browsing for good books, the aesthetics made their preselections from a larger group of search result snippets, but focused in detail on fewer book pages compared to the entertained. In “search by analogy,” search actions and search success were similar in both reader groups. The first query was often formulated to access the metadata of a previously read book for discovering suitable entry terms. Both reader groups made several SERP visits and opened many book pages in order to identify similarity between the titles. Both reader groups devoted more time to SERPs compared to book pages. Identifying familiar authors or titles, or appealing unknown titles, were the key motives for entering book pages. In book pages, the similarities between titles were identified from metadata. The query analysis revealed that the most popular search approaches in each search scenario were searching by an author or a topic in both reader groups. This confirms the findings of Koolen, Bogers, van den Bosch, and Kamps (2015), who noticed that two dominating aspects in complex search queries for books in social media were the content of the book and the familiarity of the desired reading experience. Familiarity was described as searching for similar 8 books as known ones or books related to a previous reading experience. If searching by an author, the aim is often to select unread titles from a familiar author based on previous reading experiences. Searching by a topic relates to searching by the content of a book as the aspect of content was described as the topic, plot, genre, or style of a book (Koolen et al., 2015). Our findings showed that the majority of the search queries were one-dimensional, containing solely a single aspect such as an author, a topic, or a genre. This seems to contradict the findings of Pejtersen (1989), who noticed that readers’ requests for fiction were multidimensional, containing various aspects such as subject matter, setting, author’s intention, and accessibility. This difference might be due to the fact that readers in Pejtersen’s study presented the requests to librarians, while participants in our study used online catalogs for searching. It is likely that the requests for librarians are more versatile in comparison with the search queries in online catalogs. When browsing for good books or novels similar to a previously read one, the aesthetics tended to find novels of a certain topic, whereas the entertained preferred unknown titles from familiar authors. This is in line with Saarinen and Vakkari (2013), who noticed that reader’s literary competence was associated with browsing behavior when selecting novels: esthetes with interest in high-standard novels and a broad variety of novels accessed novels by open-ended browsing more compared to participants reading for entertainment. Participants reading for entertainment knew in advance the titles they were seeking and open-ended browsing was not a common search strategy for them. Search Tactics for Fiction Books in Library Catalogs When initially coding the variables for search actions, particular patterns of search actions were detected. By watching on-screen activity, participants’ actual search paths were monitored and each possible search action during the search process was detected in each task. Search actions were found to form typical sequences occurring in a similar fashion. By integrating the information from the qualitative analysis of search queries and the participants’ typical use patterns, it was possible to elicit three suggestive search approaches that gave an insight into fiction readers’ navigational behavior in library catalogs. Each of these search approaches were named a search tactic, which refers to a set of search moves that are temporally and semantically related (Wildemuth, 2004). Three suggestive search tactics for fiction books were i) focused querying, ii) topical browsing, and iii) similarity based-tactic. Readers’ typical search behaviors and the factors influencing the book selection were different in the identified search tactics. Each search tactic is next presented in detail. Focused Querying Tactic Figures 2 and 3 present two typical scenarios for the “focused querying tactic.” JOURNAL OF THE ASSOCIATION FOR INFORMATION SCIENCE AND TECHNOLOGY—Month 2017 DOI: 10.1002/asi FIG. 2. Focused querying with known title. In “focused querying tactic,” searcher’s literary competence and previous reading experiences function as a starting point for the search process. A searcher begins by issuing a quick query with either a previously known title or author. If a known title functions as an entry term, the first SERP visit is often quick (Figure 2). When examining the results, available metadata elements are the key factors for book selection: the richer the metadata is in SERPs, the more it supports quick book selections. If a known author functions as an entry term, the emphasis in SERP examination is on identifying unread titles (Figure 3). Again, rich metadata in SERPs supports book selections from the results list. If metadata is found to be limited, or if a searcher wants to examine an appealing title more in detail, the title page is opened. Title page examination results either in an engagement and selection, or in rejection followed by a new SERP visit or a subsequent query. Our findings showed that the “focused querying tactic” was an effective and a well-functioning search tactic for novels regardless of the search scenario. However, previous knowledge of authors and titles is required in order to apply the “focused querying tactic.” An enriched catalog might support “focused querying” more over a traditional one, as ideas for authors and titles are often provided already at the entry page of an enriched catalog. Topical Browsing Tactic Typical search behaviors in “topical browsing tactic” are provided in Figure 4. In “topical browsing tactic,” a starting point for the search process is a desired topic or genre of a novel. A searcher begins the search process by reflecting possible interesting topics or genres. In the beginning of “topical browsing,” much time is devoted to formulating queries, as searchers aim at selecting unfamiliar and novel titles. SERPs were examined in detail in order to identify appealing titles or familiar authors. If the results were found unsatisfying, the searcher entered a subsequent query. If an interesting title was encountered, the searcher opened the title page for closer examination. In title pages, the reading experience provided by the novel is assessed particularly by inspecting the subject headings and the blurb. If the content description provokes enough interest, the novel is selected. If the novel is rejected, the searcher returns to SERPs or issues a subsequent query. FIG. 3. Focused querying with known author. Our findings showed that the “topical browsing tactic” was a popular search approach for previously unknown novels. Ability to reflect and identify desired topics and to issue them as query terms is required to apply the topical browsing tactic. In addition, the ability to identify appealing titles from SERPs presenting topical novels is essential in identifying interesting titles effectively. Similarity Based Tactic Typical search behaviors in the similarity-based tactic are provided in Figure 5. The “similarity- based tactic” is applied when a reader wishes to identify titles similar to a previously read interesting one. The starting point for the search is a previously read novel that has provided a good reading experience. The searcher begins by issuing a query with the previously read title as an entry term, which is followed by a quick SERP visit. The title page is opened and the metadata is examined in order to identify the elements that are essential in the finding of new, similar titles. Usually, the searcher examines the index terms in order to identify the novel’s topics and settings. The blurb and cover image are also paid attention to. Occasionally, the similarity is not associated with topics. Instead, an elaborate language or an author’s style may function as the most desirable aspects in the novel. Thus, text samples (if provided) are also read. After identifying the desired features in the title page, they are expressed as a subsequent query. When examining SERPs, they are compared against the previous reading experience and similarities between novels are assessed. In title pages, the similarities are assessed particularly by inspecting the subject headings and the blurb. If the content description provokes enough interest, the novel is selected. If the novel is rejected, the searcher returns to SERPs. If SERPs are found unsatisfying, the searcher usually returns to the original title page for new ideas for suitable entry terms. Alternatively, a searcher may rely on the features offered by the system to find similar items. Instead of issuing new queries, the search proceeds by following the system’s features: in the title page, index terms and automatic recommendations are examined and utilized to achieve new SERPs. If interesting titles appear in new SERPs or automatic recommendations, the similarities are again assessed by inspecting the subject headings and the blurb in new title pages. JOURNAL OF THE ASSOCIATION FOR INFORMATION SCIENCE AND TECHNOLOGY—Month 2017 DOI: 10.1002/asi 9 FIG. 4. Search behaviors in topical browsing. FIG. 5. Similarity-based tactic. In our study, the similarity-based tactic was applied particularly in tasks where the searchers proceeded according to their genuine reading interests. Compared with “topical browsing,” the departure of the search is more limited: a searcher has in mind an appealing title, which functions as a reference point of SERPs. The query terms are identified based on the previously read or already found interesting title, and the results are examined with relation to the known title. Limitations A few limitations should be noted. First, the sample was biased towards females and the highly educated. Second, as the questionnaire for the reading preferences was used for the first time, the validity and the reliability of the instrument should be considered. Participants’ answers to the questionnaire might have been influenced by factors such as time or mood. Third, the limited number of participants might have influenced the final cluster solution. We suspect that the cluster analysis would have yielded a solution with 10 three to four clusters with a larger number of participants. In addition, the limited number of participants prevented a simultaneous comparison of search behaviors between two catalogs and reader groups. This could have had an association to further elaborate the results. Finally, the participants conducted search tasks in an experimental setting. Thus, it is not reasonable to generalize the findings to the population at large. Analyzing real log files instead of users in an experimental setting could have produced more generalizable results. However, as the aim of this study was to understand fiction readers’ search behaviors and search success in library catalogs, the measuring of search success had to be considered. In analyzing real log files, the search success and users’ criteria for selecting particular novels would have been challenging to detect. Observing, interviewing, and analyzing users and their real search behaviors enabled the investigation of search success as rated by users. As the search tasks reflected the major browsing tactics for fiction retrieval, it could be assumed that the typical search approaches of fiction readers in a natural setting would share some characteristics with our results. JOURNAL OF THE ASSOCIATION FOR INFORMATION SCIENCE AND TECHNOLOGY—Month 2017 DOI: 10.1002/asi Conclusion The aim of this study was to investigate fiction readers’ search actions and search queries in different reader groups. We created a typology of fiction readers based on reading preferences. Our first main contribution was in demonstrating that different preference patterns for fiction books were clearly associated with readers’ search behavior, that is, the number of viewed SERPs, opened book pages, dwell time on book pages, and the type of search queries. Our second main contribution was in presenting three search tactics for fiction books in library catalogs: i) focused querying, ii) topical browsing, and iii) similarity-based tactic. We demonstrated that the search scenario determined the use of the identified search tactics. Searchers’ literary knowledge and previous reading experiences were associated with the selection of a suitable search approach in each search scenario: at all times the starting point for the search was either a known title or an author, an interesting topic or genre, or a previously read well-known title. The new works of literature were assessed in relation to the earlier reading experiences. References Bates, M.J. (1990). Where should the person stop and the information search interface start? Information Processing and Management, 26, 575–591. Bergstrom, A., & Hoglund, L. (2014). A national survey of early adopters of e-book reading in Sweden. Information Research, 19. paper 621. Retrieved from http://InformationR.net/ir/19-2/paper621.html. Carter, R., & Lubinsky, J. (2016). Rehabilitation research. principles and actions. 5th ed. St. Louis, MO: Elsevier. Elsweiler, D., Wilson, X., & Kirkegaard Lunn, B. (2011). Understanding casual-leisure information behaviour. In A. Spink & J. Heinstr€ om (Eds.), New directions in information behaviour (pp. 211–241). Library and Information Science, Vol. 1. Bingley, UK: Emerald Group Publishing. Goodall, D. (1989). Browsing in the public libraries. LISU Occasional paper No 1. Loughborough, UK: Library and Information Statistics Unit. Hair, J.F., Anderson, R.E., Tatham, R.L., & Black, W.C. (1998). Multivariate data analysis. 5th ed. Upper Saddle River, NJ: Prentice-Hall. Hypen, K., & M€akel€a, E. (2011). An ideal model for an information system for fiction and its application: Kirjasampo and Semantic Web. Library Review, 60, 279–292. Ingwersen, P., & J€arvelin, K. (2005). The turn: Integration of information seeking and retrieval in context. Dordrecht, The Netherlands: Springer. Jiang, T. (2013). An exploratory study on social library system users’ information seeking modes. Journal of Documentation, 69, 6–26. Koolen, M., Bogers, T., van den Bosch, A., & Kamps, J. (2015). Looking for books in social media: An analysis of complex search requests. In Proceedings of ECIR’15. LNCS 9022 (pp. 184–196). Berlin: Springer. Marchionini, G. (2006). Exploratory search: From finding to understanding. Communications of the ACM, 49, 41–46. Maryl, M. (2008). The apology of popular fiction. Everyday uses of literature in Poland. In S. Zyngier, M. Bortolussi, A. Chesnokova, & J. Auracher (Eds.), Directions in empirical literary studies (pp. 317– 328). Amsterdam: John Benjamins Publishing Company. Miesen, H.W.J.M. (2003). Predicting and explaining literary reading: An application of the theory of planned behavior. Poetics, 31, 189–212. Mikkonen A. & Vakkari P. (2015). Books’ interest grading and fiction readers’ search actions during query reformulation intervals. Proceedings of Joint Conference on Digital Libraries 2015 in Knoxville, Tennessee, ACM, New York, NY, 27–36. Mikkonen A. & Vakkari P. (2016a). Finding fiction: Search moves and success in two online catalogs. Library and Information Science Research, 38, 60–68. Mikkonen A. & Vakkari P. (2016b). Readers’ interest criteria in fiction book search in library catalogs. Journal of Documentation, 72, 696– 715. Oksanen, S., & Vakkari, P. (2012). Emphasis on examining results in fiction searches contributes to finding good novels. In Proceedings of JCDL’12 (pp. 199–202). New York: ACM. Ooi, K., & Liew, C.L. (2011). Selecting fiction as part of everyday life information seeking. Journal of Documentation, 67, 748–772. Pejtersen, A.M. (1989). The Bookhouse: Modelling user’s needs and search strategies as a basis for system design. Risø report M-2794. Finland: Roskilde Risø National Laboratory. Ross, C.S. (2001). Making choices: What readers say about choosing books to read for pleasure. The Acquisition Librarian, 13, 5–21. Saarinen, K., & Vakkari, P. (2013). A sign of a good book: Readers’ means of accessing fiction in the public library. Journal of Documentation, 69, 736–754. Shimray, S., Chennupati, K., & Chennupati, K.R. (2015). An overview of mobile reading habits. Journal of Library & Information Technology, 35, 344–354. Spiller, D. (1980). The provision of fiction for public libraries. Journal of Librarianship and Information Science, 12, 238–266. Stebbins, R.A. (1997). Casual leisure: A conceptual statement. Leisure Studies, 16, 17–25. Stebbins, R.A. (2007). Serious leisure: A perspective for our time. New Brunswick, NJ: Transaction. Tang, M.-C., Sie, Y.-J., & Ting, P.-H. (2014). Evaluating books finding tools on social media: A case study of aNobii. Information Processing and Management, 50, 54–68. The Reading Agency. (2013). Library facts. Retrieved from http://readingagency.org.uk/news/library-facts004/. The State of America’s Libraries. (2014). A report from the American Library Association. Retrieved from http://www.ala.org/news/sites/ala. org.news/files/content/2014-State-of-Americas-Libraries-Report.pdf. Thudt, A., Hinrichs, U., & Carpendale, S. (2012). The Bohemian Bookshelf: Supporting serendipitous book discoveries through information visualization. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 1461–1470). New York: ACM. Usherwood, B., & Toyne, J. (2002). The value and impact of reading imaginative literature. Journal of Librarianship and Information Science, 34, 33–41. ontinen, J. (2014). Book’s interest grade Vakkari, P., Luoma, A., & P€ and dwell time in metadata in selecting fiction. In Proceedings of Information Interaction in Context 2014 (pp. 28–37). New York: ACM. Vakkari, P., & P€ ontinen, J. (2015). Result list actions in fiction search: An eye-tracking study. In Proceedings of JCDL’15 (pp. 7–16). New York: ACM. Wildemuth, B.M. (2004). The effects of domain knowledge on search tactic formulation. Journal of the American Society for Information Science & Technology, 55, 246–258. Yu, L., & O’brien, A. (1999). A practical typology of adult fiction borrowers based on their reading habits. Journal of Information Science, 25, 35–49. Zhang, Y., & Kudva, S. (2014). E-books versus print books: Readers’ choices and preferences across contexts. Journal of the American Society for Information and Technology, 65, 1695–1706. Appendix: Reading Preferences Questionnaire Please select how important the following aspects are when reading fiction books: JOURNAL OF THE ASSOCIATION FOR INFORMATION SCIENCE AND TECHNOLOGY—Month 2017 DOI: 10.1002/asi 11 Fiction reading provides. . . A possibility for daydreaming Entertainment Artistic experiences Escapism Aesthetic pleasures An opportunity for relaxation A possibility to experience beauty Widening of vocabulary and ability to express oneself Intellectual stimulation Widening the reader’s view of the world Insight into foreign culture Practical solutions to everyday life Understanding of the world New perspectives A novel offers/a novel is. . . A portrayal of a reality A chance for identification Unfamiliar subject Entertaining Rich and elaborate language Gripping plot Text that resonates with feelings Easy to read Surprising features of narration Interesting characters Thought-provoking Challenge for its reader Factual information about the world Historical events An original style A novel has won an award 12 Not important A little important Somewhat important Very important 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 4 4 4 4 4 4 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 JOURNAL OF THE ASSOCIATION FOR INFORMATION SCIENCE AND TECHNOLOGY—Month 2017 DOI: 10.1002/asi