Book Citation Index (BCI) Overview

• Book Citation Index (BCI) is a multidisciplinary citation database designed to index monographs and book chapters, especially in Humanities and Social Sciences.
• It compiles extensive data with over 29K books and 379K chapters at launch, using metrics like FNCI and activity index to evaluate publisher and book impact.
• BCI faces challenges such as citation fragmentation, language bias, and publisher concentration, which affect accurate evaluation and bibliometric analysis.

The Book Citation Index (BCI), launched by Thomson Reuters in October 2010, represents the first large-scale, multidisciplinary citation database explicitly designed to index monographs and book chapters, particularly for fields underrepresented by journal-centric citation tools such as the Science Citation Index (SCI), Social Sciences Citation Index (SSCI), and Arts & Humanities Citation Index (A&HCI). The BCI and its analytics—such as Book Publishers Citation Reports (BPCR)—enable new forms of quantitative bibliometric analysis in the Humanities and Social Sciences, but they also introduce field-specific and technical complications surrounding coverage, citation practices, bibliometric indicators, and evaluation practices.

1. Scope, Coverage, and Composition of the Book Citation Index

At launch in October 2010, the BCI indexed 29,618 books and 379,082 book chapters, with a pronounced expansion by May 2012 to cover 28,805 books and 367,616 chapters from 2006–2011, of which 17,005 books and 202,830 chapters (55%) fell within the Humanities & Arts and Social Sciences & Law (Torres-Salinas et al., 2012). Subject coverage is determined by assignment of records to Web of Science (WoS) subject categories, which are internally regrouped for bibliometric aggregation—for example, 19 disciplines in BPCR, or 15 macro-disciplines in coverage analyses (Torres-Salinas et al., 2013). Temporal coverage initially prioritized material from 2005 onward, with selective backfilling to 2003.

Selection criteria explicitly favor English-language publications, resulting in a strong language and regional bias (96% English, 75% US/UK), with many non-Anglophone and regional publishers absent or underrepresented [(Torres-Salinas et al., 2012); (Torres-Salinas et al., 2013)]. The publisher distribution is highly concentrated, with 33 publishers accounting for 90% of all books in the index; Springer, Palgrave Macmillan, and Routledge alone account for over 50% (Torres-Salinas et al., 2013). The database encompasses both monographs and edited volumes, with every book chapter indexed as a separate entity regardless of book type [(Leydesdorff et al., 2012); (Torres-Salinas et al., 2013)]. Document types are distinguished as “Book” and “Book Chapter,” with the latter assigned even to single-author monographs.

Book chapters may be cross-classified into multiple disciplines (≈12% of chapters), complicating partitioned analyses and necessitating careful normalization (Torres-Salinas et al., 2013).

2. Organizational Structure and Publisher Analytics

Analogous to Journal Citation Reports, BCI data permit the assembly of Book Publishers Citation Reports (BPCR) that rank publishers within disciplines according to bibliometric indicators of output and impact. Publishers’ names are normalized to merge variants; rankings are primarily based on total output, supplemented by impact metrics (Torres-Salinas et al., 2012). The activity index (AI) is employed to gauge field specialization for publishers: $\mathrm{AI}_{p,n} = \frac{B_{p,n} / B_p}{B_n / B}$ where $B_{p,n}$ is the number of books by publisher $p$ in field $n$, $B_p$ is the total books by $p$, $B_n$ is total books in $n$, and $B$ is total books in BCI (Torres-Salinas et al., 2013).

A critical complication is the separation of citation attribution: book–level and chapter–level citations are not aggregated. Publishers with numerous edited volumes thus accrue high output and citation counts, while those specializing in monographs—a typical profile for university presses—are systematically disadvantaged if aggregation is not performed.

3. Bibliometric Indicators and Statistical Models

Production and impact are assessed using six primary indicators (Torres-Salinas et al., 2012):

• Total Items ($P$): $B_{p,n}$0
• Books ($B_{p,n}$1): $B_{p,n}$2, $B_{p,n}$3 if record $B_{p,n}$4 is a book
• Book Chapters ($B_{p,n}$5): $B_{p,n}$6, $B_{p,n}$7 if record $B_{p,n}$8 is a chapter
• Total Citations ($B_{p,n}$9): $p$0
• Average Citations per Item ($p$1): $p$2
• Percentage of Non-Cited Items (NC): $p$3

The citation distributions in BCI display Lotkaian (power-law) behavior, high uncitedness (74–92% depending on discipline), and extreme right-skew: a power-law exponent $p$4 estimated at 1.28–1.89 across major areas (Torres-Salinas et al., 2013). Statistical comparisons—such as Kullback–Leibler divergence—are used to measure the information gain between citation histograms of publishers against disciplinary baselines. Visualization methods such as "heliocentric clockwise maps" summarize similarity to baseline, output, and average citation in polar coordinates (Torres-Salinas et al., 2013).

Field-normalized citation metrics, such as the CWTS Crown Indicator/FNCI,

$p$5

(where $p$6 is a publisher's average citations per publication and $p$7 is the expected field-year-type average), are used to account for field and vintage effects (Torres-Salinas et al., 2013).

4. Citation Practices, Indexing Anomalies, and Methodological Issues

Citation dynamics in book-based scholarship differ substantially from journal-dominated domains. Monographs accumulate citations more slowly, with a longer citation half-life, particularly in SSH disciplines (Leydesdorff et al., 2012). Book chapters often receive more citations than whole books; for example, in SCI/SoSCI/A&HCI pre-2005, books averaged 10.5 citations, chapters 19.3 (Leydesdorff et al., 2012). In BCI, 80.5% of all records are uncited; mean citations per record is 1.67 (Torres-Salinas et al., 2013).

Significant methodological challenges include:

• Dispersion: Citations to chapters are not summed into parent book records, causing the impact of monographs to be systematically underestimated unless chapters are manually aggregated [(Leydesdorff et al., 2012); (Torres-Salinas et al., 2013)].
• Counting conventions: Counting all chapters of a monograph as individual publications inflates productivity metrics, but counting only the book penalizes authors—particularly problematic for research evaluation (Leydesdorff et al., 2012).
• Edition and translation fragmentation: Citations are partitioned by edition, reducing measured impact (Kousha et al., 2018).
• Database errors: Not all citations retrievable via WoS Cited Reference Search are included in BCI records (Torres-Salinas et al., 2013). Book–level records do not incorporate citations to chapters, distorting publisher and author impact analyses.

5. Discipline and Publisher Patterns

The representation of disciplines and publishers in the BCI is highly uneven. Over 56% of BCI book output is in Humanities & Arts (28.9%) and Other Social Sciences (27.9%), with technical fields comprising a smaller but notable share (Torres-Salinas et al., 2013). Springer (Netherlands), Palgrave Macmillan (UK), and Routledge (UK) account for >50% of books; Anglophone and large commercial publishers dominate, while university presses are more present among highly cited books but hold a much smaller share overall.

Publisher specialization can be measured using the activity index; for example, Maney Publishing and Rodopi specialize in Humanities & Arts, while Nova Science and Routledge emphasize Social Sciences. University presses (e.g., Princeton, MIT, Cambridge) consistently achieve higher field-normalized citation scores (Torres-Salinas et al., 2013).

Citation impact characteristics reveal edited volumes and books in series generally receive more citations, especially in Science, Engineering & Technology (e.g., mean citation for edited books in SCI: 35.41 vs. 10.16 for non-edited). University presses are consistently overrepresented among the top 5% of cited books in SSH areas (Humanities: 61% of highly cited books from university presses, vs. 27% in total) (Torres-Salinas et al., 2013).

6. Visualization, Comparisons, and Alternatives

"Heliocentric clockwise maps" and citation histograms are employed to visualize publisher impact, output, and the similarity of publisher citation patterns to disciplinary baselines. Most large generalist publishers cluster at low information-gain (I-gain) values, meaning their output defines the standard disciplinary citation shape, while smaller but high-impact houses are identified as outliers (Torres-Salinas et al., 2013).

Comparisons of BCI with external sources, such as Microsoft Academic (MA) and Google Books, demonstrate that BCI underestimates citation impact, in part due to edition and chapter fragmentation; MA typically aggregates across editions and finds 1.5–3.6× higher citation counts in many fields (Kousha et al., 2018). BKCI's English-language, US/UK bias and lack of edition-level aggregation limit its representativeness for non-Anglophone and interdisciplinary book impact.

7. Critical Limitations and Interpretive Cautions

The BCI exhibits pronounced coverage gaps, linguistic and regional bias, and technical anomalies, all of which affect the validity of publisher and author evaluation. Key limitations include:

• Citation neglect: Book records do not incorporate chapter citations, and not all retrievable citations are captured in record-level data (Torres-Salinas et al., 2013).
• Coverage bias: The index skews toward English-language and commercial publishers, underrepresenting university presses and regional output (Torres-Salinas et al., 2012).
• Inflated/deflated indicators: Book series and edited volumes may artificially inflate counts; conversely, monographs' impact may be under- or over-estimated depending on counting methodology (Leydesdorff et al., 2012).

Given these limitations, the use of BCI or derivatives like BPCR for high-stakes evaluation (e.g., researcher promotion) is discouraged. More robust use-cases include library collection development and preliminary mappings of publisher visibility. Aggregation across multiple bibliometric sources (e.g., Microsoft Academic, Google Books) is advised for a more comprehensive picture of book impact (Kousha et al., 2018).

---

Select References:

• (Torres-Salinas et al., 2012, Leydesdorff et al., 2012, Torres-Salinas et al., 2013, Torres-Salinas et al., 2013, Torres-Salinas et al., 2013, Kousha et al., 2018).