About Gale

Title List Changes

Business Development

Press Room

Outside U.S. and Canada

Product Information:

Customer Service:

Customer Resource Center:

Free Resources:

Reference Reviews

Péter's Digital Reference Shelf

January 2007

Title: Web of Science
Publisher: Thomson ISI
URL: http://www.isinet.com
Cost: depends on licensed content
Tested: December 10-31, 2006

The Context

It was more than two years ago that I reviewed the Web of Science service of ISI Thomson in this column. It was featured in several of my other articles, conference papers, and workshops on citation enhanced databases, often in comparison with Scopus and Google Scholar. For some of these, see the lengthy article in Current Science, the paper presented at the International Conference on Asian Digital Libraries in Bangkok, or the PowerPoint slides of the keynote address about the comparative content evaluation of citation databases at the annual conference of the Japan Society for Information Science and Technology and of the plenary closing session about The Illiteracy and Innumeracy of Google Scholar presented at the Annual Conference of the UK Serials Group .

Of course, there are many substantial and good reviews and comparative evaluations by others about WoS (often in context of Scopus and Google Scholar), such as the open access preprint version of the paper by Meho and Young to be published in JASIS&T later in 2007, which has a current bibliography on the new generation of citation indexes.

Since the earlier review of WoS in mid-2004, two very important events happened that reshaped the context of citation-enhanced databases: the release of Scopus and the launch of Google Scholar. I had in-depth reviews of both in this column, the most recent one for Scopus is here, and for Google Scholar is here.

To put things into perspective, the articles in Online Information Review about the Deflated, Inflated, and Phantom Citation Counts, and about Dubious Hit Counts and Cuckoo's Eggs and the PowerPoint slides of the plenary session presentation at the UKSG conference mentioned above could be useful for those who don't wish to take at face value the so-often-reported but rarely corroborated hit and citation counts of Google Scholar. These often turn out to be fool's gold, and can embarrass the researchers who refer to them without verification.

There are several other very current reviews of both Scopus and Google Scholar, representing different opinions and experiences, such as the one by Howard M. Dess for the former in the December, 2006 issues of ISTL, and the free e-print version of Bruce White's in-depth review of Google Scholar's in the first issue (under the new title) of the open access New Zealand Library and Information Management Journal.

The Content

The content of WoS depends on how the users' institutions license the database. The largest one is the Century of Science edition. Its Science module goes back to 1900, the social science module provides coverage from 1956, the Arts & Humanities module from 1975. For the 1900-1944 period papers from science journals were collected and processed, to the tune of close to 1 million articles from more than 250 of the most influential journals. Naturally, ISI took a scientific approach to determine the influence of journals, described in an intellectually interesting PR-material, which is quite a rarity.

The first round of short-listing was based on the number of citations the journals received for items published in the first 45 years of the last century – with a minimum threshold of 50 citations. This yielded a set of about 2,000 journals. Restricting the search to journals that had at least 1 article that was cited more than 100 times created a second set.

The two sets were combined and limited to those journals that published 5 or more articles with more than 100 citations OR had a total of 1,500 citations. This set was than whittled down considering also the language and country of publication) to determine the 200 most influential journals.

It would be very interesting to see the creation of the list of top journals based on the above criteria (but with higher minimum citation thresholds, varying on the discipline) for the 1945-1990, and the 1991-2006 period (although 2004 would be more realistic, as typically the first citation "hump" shows up in the second year after the publication of the article).

Although this is not (yet) possible because there is a 100,000 item limit on sets that can be created (quite an increase from the pre-2004 limit of 500 records), representing 10% of the typical intake of WoS in a single year, there are many superb new indicators to gauge the influence of journals, articles, individual authors, group authors, and organizations which I will describe in the software section. While these indicators are the most important new elements of the synergy of content and software, the traditional content enhancements are also important.

There is no published data available which editions of WoS are the most widely licensed, but a casual search on the Web suggests that the most popular version may be the one which provides coverage from 1945 onward for the science module, and the rest is the same as mentioned above. This is the version that I also used for this review.

Database size and growth

By virtue of covering additional journals, the yearly intake of records kept increasing in WoS (with some wavering in 2002. It exceeded the 1 million mark per year in 1994, then the 1.5 million mark in 2005. The record count for 2006 is close to 1.4 million as of the last day of 2006, but the coverage of the current year is completed in every indexing/abstracting database only several months into the new year, so by the Spring of 2007 there is likely to be more than 1.5 million records for documents published in 2006.

Currently (as of the last week of 2006), there are 37.3 million records in WoS (for the edition covering 1945-2006). For comparison, the current size of Scopus is 29.3 million records, but an addition of more than 7 million pre-1996 records for papers in chemistry, physics, math and social sciences is imminent and expected by early 2007. This will make WoS and Scopus go neck to neck in terms of the total size of their databases. (Both WoS and Scopus have additional components for patents and Web sites, which are not included in my numbers above).

In spite of the similar size, the shape of the two databases is as different as is the body of a typical 300 pound football player and of a 300 pound basketball player. Symbolically speaking, WoS is "taller" than Scopus, but Scopus is "wider" for the most current 7 years. This is because WoS goes back to 1945 (in the 1945-2006 edition), while Scopus (which has a single edition) goes back to only 1965 (with a smattering of records before that year), but it has a broader source base and thus wider item coverage from 1999 onward.

Of course, the absolute size of WoS decreases if the time span licensed decreases. For example, the total size is about 35 million records (1965-2006), 27.8 million (1980-2006), 23.83 million (1985-2006), or 14.7 million (1995-2006), and 7.85 million in the excellent educational subset (2001-2006) available for accredited library schools.

Cited references

The inclusion of all the cited references from the source documents processed is the highlight of WoS. WoS records have been enhanced by cited references from the very beginning, while in Scopus, citations were added to records from 1995 onward. (The PR materials in Scopus indicate 1996 as the start year of enhancement by cited references, but I found about 100,000 records for items published in 1995 which were enhanced by cited references).

The gap between WoS and Scopus in terms of the total number of cited references is not necessarily proportional to the gap between the total number of records in the two databases throughout the years, simply because not all articles have cited references, and many articles have only a single cited reference.

The WoS software does not offer an option to determine the number of such records, but the Dialog implementation of the subset of the three citation indexes can help. It is to be noted, however, that about 4.5% of the records appear in two or three of the citation indexes on DIALOG, and the excellent duplicate removal function of Dialog cannot be applied to the very large subset I used for testing. (The WoS software automatically removes the duplicates when searching two or three components of WoS). Therefore the numbers quoted below are only estimates, but rather close ones.

Having analyzed a nearly 33 million record subset of WoS on Dialog for the period between 1972 and 2006, I found that more than 7 million source documents (22%) had no cited references, and more than 3.5 million source documents (11%) had only one cited reference. The former ratio slightly varies between the three citation index components. In the fields of sciences and social sciences 77-78% of the source documents have cited references, in arts & humanities this ratio is 84%.

As for the source documents with a single cited reference, the ratio is widely different among the three main disciplinary categories. In the sciences, only 2.75% of the source documents belong to this category, while in the social sciences this ratio is 28%, and in the arts & humanities the ratio is 54%. The overall average reflects the predominance of science articles in the 1945-2006 edition of WoS in the science category, which is about 77% versus 14% and 9% of papers from the Social Sciences and Arts & Humanities, respectively.

Abstracts

The presence of abstract is not a strong point in WoS. The citation indexes were not meant to be indexing/abstracting services, and the huge print volumes would have multiplied by the inclusion of abstracts, not to mention the increase of other costs of adding abstracts. With the proliferation of online access, and by making the abstracts searchable, ISI decided to include abstracts when there was an editorial or author abstract in the source documents. This policy was applied only for records which were added from 1991 to the Science Citation Index, from 1992 to the Social Sciences Citation Index, and from 2000 to the Arts & Humanities Index. No wonder that less than 35% of my close to 33 million record sample had abstract; about 10.5 million for the science, 1 million for the social sciences, and a puny 56.7 thousands for the arts and humanities sources.

Time span

As for the importance of extended retrospectivity, it varies from disciplines to disciplines, and even for different search topics within a discipline. It can be measured by the cited half-life, which indicates the number of years from the current year that account for half of all the citations received. This is calculated at the journal level by ISI for the Journal Citation Reports, and is also aggregated for the journals in a given subcategory.

Obviously, in disciplines where the cited half life is less than 6 years, such as in Immunology, Genetics, Microbiology or Applied Physics, the 1995-2006 edition may suffice. On the other hand, in disciplines where the cited half life is longer than a decade, as it is in History, Religion, Mineralogy, Zoology or Mathematics, at least the WoS edition of 1980-2006 is needed.

Source Base

The source base of WoS also kept increasing. It is especially welcome that more than 300 open access journals are now covered by WoS, which represents about 10% of the open access journals currently available in the world. This category certainly should be increased, especially because studies have shown that they tend to be more often cited than the subscription-based journals. In the category of Information and Library Science, for example, there is only one open access journal covered by WoS, the excellent Information Research. D-Lib Magazine and First Monday would much more deserve coverage by WoS than 2-3 of the currently included titles which have lower impact factor, and much lower immediacy index than either of those two open access journals. As for the proportion of the journals, the science journals dominate with 6,598 titles (68%), followed by the 1,950 social sciences journals (20%), and the 1,149 arts & humanities journals (12%). This is somewhat different, however, from the item-level composition of WoS mentioned earlier.

In spite of the common belief that WoS covers only journals, there are tens of thousand of chapters of monographic series, as well as conference papers in WoS, but they are indeed dwarfed by the journals. (It is to be noted that not all proceedings are conference proceedings, such as PNAS, the Proceedings of the National Academy of Sciences of the U.S.A., one of the most influential journal). It gives a sense of importance of this document type when you learn that in the separate ISI Proceedings database there 4.5 million records for papers presented at conferences, congresses, symposia and national meetings from 1990 onward. More than 90% of them are in the fields of science and technology.

This genre along with books should be a growing part of the source base, as they are important scholarly media as well as rich sources of cited references, especially the books, which enjoy their second wind due to the large scale digitization which help in discovering the power of citation searching. (To the credit of Google, after the review of Google Book Search in this column, this special service was very significantly improved and turned from a frog into a prince with only a few warts left).

The Software

There have been many upgrades that have made the software more user-friendly, more intuitive and more in line with the practice of web wide search engines. Some have been simple changes, others are very significant, and a group of them are more than just evolutionary changes.

For a simple but important example, in July 2006, WoS switched from interpreting the space in a query as a phrase construction, i.e. big island meaning big followed by island directly, to interpreting the space character as a Boolean AND operator, i.e. big AND island. This was a smart move because with the earlier interpretation WoS handicapped itself, as almost all the web wide search engines use the Boolean AND operation, returning far more hits for the same query. Phrase searching is possible, but it now requires the use of double quotes as in "big island", again reflecting the common practice.

Increased set size and clustering

The set limit that I complained about in the earlier review of WoS was extended also for the sorting operation by source title, first author and —most importantly— the Times Cited By (citedness frequency) element from a paltry 300 records to 100,000. In the Analyze function, which allows the ranking of result sets (alphabetically or by frequency) by author, institution name, document type, language, publication, source title and subject category, the limit was increased from 2,000 to 100,000.

There is now a similar function which provides clustering automatically by the same data elements (although only in decreasing frequency order) and allows the users to restrict the search to cluster elements marked through check-boxes. This way, the result set of searching for example by the word stress can be restricted to materials science, mechanics and applied physics. Alternatively, results retrieved from sources in these disciplines can be easily excluded to focus on psychological stress.

There is one issue in the clustering by sources that needs refinement. In one test, the cluster did not show the name of one of the citing sources, Lecture Notes in Computer Science (LNCS), the well regarded Springer book series. Actually it does appear in the cluster, but it display the title of the book within the series: Advances in XML Information Retrieval and Evaluation.

This could be appropriate if it were a consistent practice. However, in the various item display formats, the series title, Lecture Notes in Computer Science, is shown, not the book title. It adds further to the confusion that when you click on the link to show the journal impact factor of LNCS it links you to Lecture Notes in Artificial Intelligence (which happens to be a subseries of the LNCS series, which in turn has its own JCR record). The actual paper has nothing to do with the LNAI subseries. No wonder that LNAI and LNCS have a surprisingly low impact factor because very often they are not identified at all or not appropriately in the cited references, which are quite difficult to construct correctly for papers published in these otherwise very important (sub)series.

The only reason I know these problems is because I managed a computer science library in my prior life, and learned how convoluted and illogical the titles and the relationships between the members (subseries) of the huge family of LNCS series can be. If you want an intellectual challenge and a frustrating morning, go to the Springer Web site and try to make sense bibliographically of LNCS, its subseries and individual volumes.

The only serious set limit that was not changed is the maximum 500 records of marking and downloading records, which is particularly inconvenient for users who are doing comprehensive searches, bibliometric and scientometric searches, and need to post-process data using various statistical programs. It would be very useful to have a much higher limit for saving records, to spare the error-prone donkey work of cycling savings operations in 500-item batches for researchers who may need to handle large sets.

On the other hand, WoS recently introduced the free EndNote Web service. It allows saving into an online library selected references (up to 10,000 but in max 500 items steps) using a light version of the EndNote bibliography management software. It is very convenient and more than sufficient for most of the users. Most importantly, it can save the agony of formatting a bibliography to one of the 2,300 formatting style each of which has its own peculiar syntax rules, representing the bane of publishing for no good reason. Cited references may be syntactically correct and please the publishers, but a large proportion of them are semantically incorrect and/or incomplete.

Tools to reduce citation errors

There are some new tools and features in WoS which help to alleviate the consequences of the rather sloppy citation behavior of authors. While the inaccuracies are not as apparent to the naked eye, they significantly reduce the effectiveness of citation matching. Of course, these software tools and features are no panacea. They don't compensate for the misspelled versions of the name of the cited source, cited author, or for the wrong volume and starting page numbers, not to mention the incomplete and correct but inconsistent variations of these.

The completeness and consistency of these data elements are quintessential for getting correct citedness scores for authors, organizations, articles and journals. WoS has gone out of its way to help the user to locate these variants through the Cited Reference Search mode, and to include those in their search. However, from the statistics and samples of the many evaluative and comparative articles and conference papers about WoS, I see that most of the users never go out of their way to look up and browse the precious index of cited references. There should be additional links to encourage users to look up also the index entries and to include the obviously valid if partially inaccurate citing references.

I analyzed citations for the highly ranked Annual Review of Information Science and Technology, as most authors write only one chapter, so it is plausible to herd together the variants jus by eyeballing. This is, for example the master record for Marcia J Bates' chapter. It reports 31 citing papers. Looking up the index generated from the cited author's last name and initial(s), the cited work, the year of publication, volume number and starting page number illustrates the typical reasons for no automatic matching of citing and cited references, the ones which are not blue and whose occurrence is not counted automatically in the citedness score of the cited item, but could be added manually by the researchers.

They include such reasons as omission of the middle initial, wrong or missing volume number, and starting page number, wrong source abbreviation. I marked these with red in the captured images.

Tracing down the first non-matching citing article, it is clear that the errors are in the citing article itself, where at least the absurdly wrong volume number (vol 60) should have been a red flag. ARIST just reached its 40th volume in 2006. In addition, two entries down the bibliography there is a correct volume number for an ARIST chapter – this is exactly the consequence of focusing on the format of one of the 2,300 (!) citation styles in use - rather then on the citation content. (I know from practice because the excellent bibliography editor of ARIST identified and corrected many sloppy citations in my manuscript for a chapter. It's another question that the dedication of my bibliography editor and of Information Today, Inc. -the actual publisher of ARIST- to ensure accurate citations is exceptional).

True, in this case "only" seven citing references are not accounted for, but the ratio may be worse, much worse. The chapter by Nan Lin and W. D. Garvey shows up with 13 citing articles. Looking up the index show an additional 17 articles, obviously citing this ARIST chapter – incorrectly or incompletely for automatic matching.

Interestingly, non-English author names may make the citing references less error prone because probably the unusual names puts on guard the authors and editors. There are misspellings in the items citing the ARIST chapter by Efthimis N. Efthimiadis, such as Efhimiadis, Efthimiades, Eftimiadis but not an excessive proportion.

Author Finder service

The recently introduced Author Finder service may address when the name and first initial, or even the middle initial of the different authors are the same, but it is not a panacea. It does help in the case of Marcia J. Bates, as there are authors in different disciplines identified as Bates MJ. The subject list displayed may make you wonder whether she is the one under Arts & Humanities (as she has authored and co-authored several articles about the Getty project), and/or the one under Social Sciences (as she wrote extensively about library science topics). You may wonder even about Bates MJ under Physics and Biomedicine, but the next step after choosing any of the above mentioned subject fields, displays the author affiliations which makes it clear that those authors are not "our" Marcia J Bates, while the affiliations under Multidisciplinary Science & Technology clearly allude through the top three affiliations (UCLA, U Washington and the Getty Project) to Marcia J Bates' (co-) authorship (if you know that before joining UCLA she taught at University of Washington.

The Author Finder service, however, may backfire, making users bypass relevant records, such as those of mine where the author affiliation is described as Roosevelt University and University of Haifa, which I had nothing to do with, and I had no co-author from those universities. These are likely to be data entry errors.

This takes me to another issue—misspellings in authors' last names and/or first initials. Although ISI has been diligent in correcting the two dozen misspellings of my last name in WoS, they remained in the implementation of the citation indexes in the Dialog system). The indexers and/or data entry operators of ISI are relentless in their efforts to spell my last name as Jasco, and in 2005 and 2006 they enhanced their repertoire by misspelling my first initial as A and N. Some peers wondered if I stopped writing since 2005.

I would not mention this if it had not hit me hard with my review of Google Scholar in 2005, which is one of the four papers attributed to Jacso N. While I applaud the correct spelling of may last name, the misspelled first initial just deprived me of precious citations received, and will further lower my modest citation measures in the long run. This brings me to the splendid new set of citation measures offered by WoS, which happens to be the first database to implement and integrate the new, widely accepted h-index (in addition to several other essential bibliometric indicators).

Citation Report

The Citation Report, introduced in December 2006 is far the most important software development in WoS, which actually also creates precious new content. The Citation Report provides a compact, highly informative profile about authors, journals, organizations and topics. It shows what was the publication productivity and citedness of an entity. It can help put these entities in perspective, and indicate how certain topics evolved and have been addressed in the past decades or century – depending on the extent of your subscription.

The Hirsch-index

Probably the Hirsch-index (also known as h-index or h-number) of authors covered by WoS will be the most popular. It has been already the topic of several articles since its introduction a good year ago, and received wide acceptance. Suffice it to say here that the h-index was proposed by professor Jorge E. Hirsch in a short article published in PNAS (here is the open access version). As the title of Hirsch's paper indicates the h-index (also known as the Hirsch number or h-number) was meant to be a simple index to quantify an individual's scientific research output. The shortest abstract of a scholarly article that I have ever seen sums it up well: "I propose the index h, defined as the number of papers with citation number ≥h, as a useful index to characterize the scientific output of a researcher".

For example, the h-index 18 for professor Tefko Saracevic indicates that he has at least 18 papers which have been cited at least 18 times, and marks that item in with a green line. in the result list. Actually he has 18 papers with 20 or more citations, but his 19th and 20th papers (listed in decreasing citedness order) were cited "only" 17 times, each.

The h-index eliminates most of the disadvantages of the other methods used for measuring the scholarly performance of the authors. It measures productivity and citedness simultaneously, it reduces distortions caused by often cited literature review and methodology papers, and it is fairly easy to determine in systems which allow sorting the results by citedness score, such as WoS and Scopus. I hope that the hosts of other citation enhanced databases, such as CSA and Ebsco will also sit up and take notice and finally offer the sorting of results by the citedness of the papers in, say, PsycINFO and PsycArticles, one of my two evergreen pleas.

Google Scholar used to list the results in decreasing order of citedness, but it does not anymore as you can see from this sample , so just eyeballing the beginning of its result list to determine the h-index may lead to wrong numbers. This is especially true if the author has a large number of publications, or Google Scholar splits the citation counts among variants of the same records it scraped from various sources, but did not consolidate into one, such as the records for this paper. These variants, of course, don't line up so neatly in a subject search. They are scattered, and as you can see here, too, they are definitely not listed in any sensible order.

These are not the only wrong numbers in Google Scholar, which plays fast and loose with both its hit and citation numbers in more ways than one). The citation counts, nevertheless, will make most researchers happy as the numbers will always be higher than those reported by WoS and Scopus which limit their citation counts to scholarly, academic and/or professional sources that they clearly identify versus Google Scholar which counts citations from any sources its special crawler finds on the web, including Psychology 101 assignment papers.

There are already tools available for calculating the h-index from Google Scholar's resuts, but it would be prudent to think twice before you include your h-index calculated on Google Scholar-based tools in your job application. Of course, the h-index produced by WoS is limited to the sources covered by WoS, but the list of sources is readily available even for non-subscribers, to gauge the scope of sources. The same is true for Scopus. Google Scholar does not offer a list of its sources. It certainly has the right and the reason to remain silent, but it would help its credibility and scholarly qualifier to identify its sources.

Other bibliometric measures

WoS calculates and displays the h-factor automatically when clicking on the Citation Report link, along with many other enlightening information. A citation report can be generated for various entities (authors, groups of authors, journals, organizations with which the authors are affiliated, or even topics).

For example, here is the Citation Report summary for the journal Scientometrics, which has the 6 th highest h-index among the 53 information and library science journals monitored by ISI. (I excluded from the ranking two journals, The Scientists and Library Journal, which had more then 10,000 items, the maximum limit allowed by the Citation Report function in WoS. Once again, it would be of great help if the limit would be 100,000 as for the regular search process. Obviously, the larger the set the longer it would take to create the charts and the other elements of the Citation Report, but they are worth the wait).

The small charts showing the number of publications and the number of citations for the entity, provide a compact and revealing profile of the publication and citation pattern for the past 20 years at a glance. (The caption of the charts refers to the first 20 years, but in reality it is the last 20 years. It would be better to refer to these as the most current 20 years). When years ago I plotted these same chart from data downloaded for 55 journals in library and information science it took me several days, now it is a question of minutes, and the time frames can be dynamically changed for recalculating the measures.

Sometimes, the chart could not be produced, as was the case with the citations/year compact chart of Bulletin of the Medical Library Association, the predecessor of the Journal of the Medical Library Association. However, the fuller chart -which covers the entire period since the journal has been covered by ISI- helped out.

Users, of course must be very careful in creating the final set used for the Citation Report. They may need to create different sets to take care of the different variants for the entity names and then combine them, if needed. In my tests, there was no need to that for, say, Carol Tenopir as all of her works appear under one name format, Tenopir C, and there is no other author with this last name and first name initial, and only one author with the same last name and TJ as the first and middle initial.

However, in case of Marcia J Bates, distinct sets had to be created to pick the items entered under Bates M and the ones under Bates MJ – eliminating the other authors with the Bates M and Bates MJ name formats, to come up with a correct set. You can't pick or de-select items from a set in Google Scholar, as it does not even have a temporary, let alone a permanent set - except for the temporary result list of maximum 1,000 items. Yes, I know, it is free..

A seemingly simple and unambiguous word like Islam may distort the topical search results if one does not realize that the first three most cited articles (and many others) have nothing to do with the religion apart from the fact that the abstracts refer to an author whose last name is Islam.

It is to be emphasized that the h-index and the other measures may depend heavily on the edition and the search parameters used for creating the set to be analyzed. They should be always identified when reporting the measures. In the samples below the 1945-2006 edition was used, and the query is shown in the top part of the screen.

Explanations and examples

The Help file of WoS well explains the meaning of the other performance indicators. I illustrate them here on the sample record for Tefko Saracevic. His summary measures indicate that there are master records for 111 papers he authored or co-authored. These in turn were cited 1,480 times, yielding an average of 13.3 citations per paper. There are 1,011 citing papers listed if you click on the View link. This is smaller than the total number of citations simply because some of the citing papers may (and do) cite more than one of his papers. Only 32 of those papers 1,011 are (co-)authored by Tefko Saracevic – reflecting a very modest self-citation rate. It would be more obvious for the casual user to label these two values explicitly (instead of using the View and View without self citation links). It would be the best to generate these measures automatically, along with the others. Of course, this would put some additional burden on the server, even if some users may not be interested in the details. Then again, processors get faster and cheaper every day, and such powerful features and computing wizardry well deserve them.

The citations received per year by any of the 111 papers is 35.24, a very good ratio in library and information science. The year-by-year breakdown of all the citations per items is displayed in 5-year moveable windows. It could only be better if the window could be set to 10 years, as it is in Scopus, even at the expense of showing only partial bibliographic details of the articles in the result matrix. It is to my heart content to see the average citations per year measure at the individual item level as I indicate here on the topical search for attention deficit disorder. This has been my other evergreen plea through several papers and directly to WoS, Scopus, Ebsco and CSA.

As mentioned before, such illuminating and factual information are available in less than a minute for sets smaller than 2,000. Sets of 8-10,000 items may take longer, and occasionally the process may need to be re-started by another click on the Citation Report button to produce the summary.

Sometimes intermittent failures occurred in my tests. For example, when viewing the citing articles for Information Research, WoS reported to have found 24 citing papers. This is 2 less than the sum of times cited value, but showed only 17 of the 24 articles. (I tried to re-run the search, even logged in again, but the same error persisted). The good news is that before I submitted the manuscript, I made a last try and the glitch is gone.

One error that was not intermittent was an error message in displaying the list of citing items and in producing the compact graph to show the citations per year received by the papers of Eugene Garfield. The complete graph, however, compensated for the latter, and the summary did justice to the father of citation indexing (and the founder of ISI) by displaying far the highest yearly citation rate of 96.09 in the information and library science field, and –not surprisingly- far the highest h-index index field with an outlier value of 30, when the rest of the best have a h-index of 18. True, Garfield has been publishing for the longest time, with his seminal paper published more than 50 years ago in Science, and the h-index favors those who have been publishing for a long time.

It should be pointed out, however, that even the younger generation can achieve high h-indexes, as Amanda Spink did. She started publishing in the early 1990s, has 104 articles in WoS. The per year citation of her articles is an impressively high 69.12 in our profession, and she has the h-index of 18 as do Christine Borgman, Marcia J Bates and Tefko Saracevic, who was her professor and has been her most common co-author.

This splendid new WoS feature of Citation Report also gives justice to the next generation's most likely citation leader in the field of information science and technology, Mike Thelwall. He published 85 papers between 2000 and 2006 (and this number does not include all his conference papers) in all the right sources. These papers received 653 citations from 174 articles, producing a very impressive overall average of 96.09 citations per year, and an h-index of 14, likely to rise to 15 or even 16 by the time you read this.

There is nothing better for customers than fierce competition for market share. ISI has kept adding new content and software features through regular updates. The latest services clustering of results set by several criteria, the instant calculation and superbly informative and compact visualization of new citation measures, such as the sum of times a paper was cited (including and excluding self-citation, the average citations per item, the Hirsch-index, the almost instant display of charts for the distribution of articles and citations per year by authors, journals, organizations or topic, the exporting of these details into a spreadsheet format, or downloading to a free Web version of EndNote, represent more than a series of evolutionary steps. It is a breakthrough for those interested in citation analysis, but did not have the resources to calculate key citation performance measures, or did not have the software to format them to the whims of the publishers' manuscript guidelines. The launching of Google Scholar and the sophisticated features of Scopus undoubtedly triggered these important additions to WoS. Scopus certainly has its aces up its sleeve for the next duel for novel and useful features.