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