The language preference among sociology scholars

THE ‘LANGUAGE PREFERENCE’ IN SOCIOLOGY : MEASURES OF 'LANGUAGE SELF-CITATION', 'RELATIVE OWN-LANGUAGE PREFERENCE INDICATOR’ AND ‘MUTUAL USE OF LANGUAGES' Moshe Yitzhaki Department of Information and Library Studies Bar-Ilan University, Ramat-Gan, ISRAEL ABSTRACT Although between one-third to one-half of world social sciences research literature is published in languages other than English, studies show very scant use of it by American and English scholars. Almost all studies, however, were conducted from the Anglo-Saxon perspective, limiting the scope of the study to English-published sources or English-speaking scientists and research workers. The present study aimed at assessing the scope of the language preference in a social sciences field, not only among American and British scholars, but among German and French ones as well, using the technique of citation analysis. Samples including mostly 50-60 original research articles were drawn from the 1985-1994 volumes of nine leading sociology journals published in the US, UK, Germany and France and the references appended to each were scrutinized in order to determine the frequency distribution of the languages cited in each periodical. Findings clearly showed a strong preference of writers to cite material in their own language. However, the extent of this bias differed from journal to journal. The American and British writers rank first, with close to 99% of their references being in English. German scholars rank next, preferring German sources in 75% of the cases, and French scholars quote French sources in only 66% of their references. In order to calculate the new refined measure of 'relative own-language preference' (ROLP) indicator, the proportions of ‘language self-citation’ were related to the estimated proportions of these languages in the existing body of sociology research. This measure reveals that German sociologists have the strongest bias towards their mother-tongue, their ratio of references in German exceeding almost 12 to 28 times the expected figure according to the German language share in sociology research. Next come French sociologists (8 to 14 times) while American and British ones display the lowest own-language bias, only slightly higher than expected. Further analysis of the foreign languages preference of each group, according to a 'mutual-use' matrix, shows a relative low use of German and French sources by British-American sociologists. INTRODUCTION Most writers agree on the overwhelming importance of English in the worldwide transfer of scientific information, comprising between 40 to 80 per cent of the total international communication in the sciences, depending on the discipline (Wood 1967; Holmstrom 1973; Engelbert 1976; Michel 1982; Inman 1983). Replacing the German language, English has become the ‘lingua franca’ of the scientific world (Thorp a.o. 1989). However, increasing amounts of primary scientific material, especially in the applied and technical sciences, are published in many other languages, mainly Russian, German, French and Japanese (Wood 1967; Wellisch 1973; Michel 1982; Jagodzinski-Sigogneau 1982; Bradely 1993; Regaunt 1993). Studying scientific publications in geology, Regaunt (1994) claims that other languages rather than English dominate some important subfields in this discipline. Actually, some argue that the best scientific research of non-English speakers is published in English anyway (Inman 1983, Garfield 1980), but this view is still highly disputed (Michel 1982; Chan 1976). However, almost all studies, including those in the sciences and the social sciences, were conducted from an Anglo-Saxon perspective, limiting their scope to English-language sources or English-speaking scientists and research workers. The reason might be simply the natural preference of a researcher to study his own domain. Also since English has become the lingua franca of the scientific world today, the main question for research became, what languages are used by English-speaking scholars. This assumption, however, is less true of the social sciences whose published research feature more non-English languages than in the sciences. Moreover, very little research was devoted to the language preference among non-English-speaking scholars, as compared to their English-speaking colleagues, and no attempt was made to apply additional measures, besides the simple measure of the percentages of each language used. LITERATURE REVIEW Studies conducted recently in science and technology by Hanson (1962), Wood (1967), Hutchins and the Sheffield Group (1971), Morgan (1977), Leck (1978), Ellen (1979), Sherwood (1979), Schmidt (1979) and Large (1983) indicate that English-speaking scientists and research workers use very little foreign language material (mainly French and German), approximately 8 to 15 per cent of the material read or cited. In the social sciences the situation is more extreme. Although one-third to one-half of social science research literature is published in non-English languages (Inman 1983; Wood 1967; Wellisch 1973) all recent studies reveal scant use of it by American and British scholars. Studies in various fields and years found this use to range from zero to 12%, with an average of 4% and a median of 3% (Louttit 1955; Guttsman 1966; Earle and Vickery 1969; Wood and Bower 1969; Stewart 1970; Hutchins 1971; Finison and Whittemore 1975; Ellen 1979; Fitzgibbons 1980). Studies done in humanities fields do not provide a uniform picture, the rate of foreign language material use by English-writers varying from 0.8% to 67% (!). Studying the languages cited by scholars in biblical research in 1950 and 1980, Yitzhaki (22) suggested using, beyond the common measure of 'language self-citation', also the measures of 'relative own-language preference’ (indicating some degree of so-called ‘linguistic isolation') and 'language mutual-use matrix', discussed later (in: Methodology). PURPOSE OF THE STUDY The objective of the present study was to assess by means of citation analysis the extent of the language self-citation (or language preference) in the field of sociology, assuming that this phenomenon results, at least partly, from a ‘language barrier’. More specifically, our study aimed: 1. to gather empirical data concerning the language preference among American, British, German and French sociologists publishing in their own scholarly journals. 2. to determine the exact extent of this preference by means of various measures, some of them new and rarely used before. METHODOLOGY Between 50 to 60 regular original research articles were drawn from recent volumes of each of the following nine different-nationality sociology periodicals: 1. The American Sociological Review (USA) : 1992-93. 2. Sociology (UK) : 1990-94. 3. The Sociological Review (UK) ; 1990-94. 4. British Journal of Sociology (UK) : 1990-95. 5. Theory, Culture and Society (UK) : 1985-93. 6. Koelner Zeitschrift fuer Soziologie (Germany) : 1990-94.. 7. Sozialwissenschaftliche Informationen (Germany) : 1989-92. 8. Revue francaise de Sociologie (France) : 1990-94. 9. L'Annee Sociologique (France) : 1989-94. The samples included only articles written originally in the language of the country of the periodical's publication (excluding translated papers) and by authors affiliated with institutions in that country only. Thus, Swiss authors were omitted from the German and French samples, and Australian and Dutch ones from the American and British samples. Articles written by British authors but published in the American periodical or vice versa, were also excluded from the samples. General categories of articles excluded from the samples were : book reviews, articles on topics of explicit local interest, articles 'in honor of...' where naturally references are heavily biased towards the honored person, and articles with less than 10 references. The references appended to each article were extracted according to a simple sampling procedure, depending on their number. If more than 30 references were listed, every second or third one (if more than 60) was taken, sampled systematically with a random starting point. Author self-citations were ignored and repeated references to the same bibliographical item in one single article were counted only once. In all, the nine journals yielded a total of more than 12000 references whose language was recorded. The language of the cited item was determined as it appeared in the reference, that is, a translated item was not recorded according to its original language, and a paper in German, for example, published in a French collection of essays, was counted as German. MEASURES USED Language analysis of cited references usually aims to reveal the languages upon which the research in the investigated field depends, as well as the relative use of each. Providing that the sample of source articles is not confined to a certain single language, but rather comprises several sub-samples published in various languages, it is possible to calculate the rate of 'language self-citation' (or ‘language preference’) for each language group. The measure of 'language self-citation' can be simply defined as the proportion of references written in the same language as the citing source. 'Language self-citation' is analogous to 'journal self-citation' and 'subject self-citation' (Earle and Vickery 1969; Nicholas and Ritchie 1978; Yitzhaki 1989) and its rate indicates the degree to which researchers in a field draw upon the literature published in their own language. A low rate of 'language self-citation' is assumed to indicate a considerable use of, and dependence on, the literature in other languages, while a high rate should indicate self-sufficiency, and independence of foreign language research literature. Seemingly, the closer the rate is to 100%, the less dependent this group of scholars is on foreign language material. However, this measure alone is inadequate and may create a distorted picture, since it does not indicate the quantity of foreign language material thus lost by those scholars, or its importance. Understandably, in a field in which English comprises a vast majority of the total world research output, the loss of information by English-writers with a high rate of 'language self-citation' is much smaller than in field in which most research material is published in languages other than English. The 'importance' of research material in 'foreign' languages is very difficult to assess and is highly disputed, as indicated above. The other problem, however, is easier to solve providing one has data regarding the estimated proportion of the total published research, shared by each language in a specific subject field. The actual 'language self-citation' rate should be related to an expected rate, which is based on the assumption that, by and large, with no language barriers and material availability problems existing, scholars would cite publications in a certain language according to its share in the total research output of that field. This refined measure of the relative language-self-citation, which might be also called the 'relative own-language preference’ (ROLP) indicator, for a group of researchers in a certain field who publish in a certain language, could be represented by the following ratio : their actual self-citation rate of that language expected self-citation rate of that language when the expected rate is expressed by an estimated proportion of this language in the total published research output cumulated hitherto in that field. A ratio equaling 1 means a 'balanced' state in which the authors in that field cite publications in their own language proportionately to its share in the existing body of research literature. The higher the ratio above 1, the higher the degree of 'self-language preference' and disuse of a larger body of foreign language material. The smaller it is than 1, the greater the relative use of foreign language literature. FINDINGS AND DISCUSSION Table 1 (Parts I and II) presents the frequency distribution of the cited languages in articles published in each of the nine journals, as well as the appropriate ratios of 'relative own-language preference'. Clearly, the highest ratio of 'language self-citation' is found among the American and British sociologists (around 99 %), with only about 1% of their references in foreign languages. The two German journals display considerably different figures : 85.5 % in the Sozial. Infor. vs. only 63.3 % in the Koelner Zeit. f. Soz., which might be explained by a more local focus of the former. It seems that French sociologists are least impeded by the ‘language preference factor’, since their articles display the highest ratio of foreign-language use (32.5%-35.5%) with only 64.5% to 67.5% of their references in French. However, the situation and the conclusions are entirely different once we apply the refined measure of 'relative own-language preference’ (ROLP indicator) described above. Table 2 presents the estimates needed for calculating these indicators.  T A B L E 1 ; P a r t I FREQUENCY DISTRIBUTION (IN %) OF LANGUAGES OF CITED REFERENCES IN FIVE ENGLISH-LANGUAGE SOCIOLOGY JOURNALS (1985-1994) AND THEIR DERIVED INDICATORS OF ‘RELATIVE OWN-LANGUAGE PREFERENCE’ (ROLP) Language of cited references American Sociological Review Sociology (UK) The Sociological Review (UK) British J. of Sociology (UK) Theory, Culture and Society (UK) English German French Other 98.6 0 0.1 1.1 98.8 0.4 0.8 0 99.3 0.7 0 0 99.5 0.5 0 0 89.6 4.9 5.4 0.1 Total 100.0 % 100.0 % 100.0 % 100.0 % 100.0 % Total no. of references checked 2374 1468 1446 1278 1197 N (=Total no. of articles checked) 100 50 50 50 60 Ref. mean Ref. median 47.4 44.2 51.8 45.5 35.2 28.8 32.3 31.5 33.0 31.3 Estimat. prop. of that lang. in sociological research (a) 81.4 % (b) 68.7 % (a) 81.4 % (b) 68.7 % (a) 81.4 % (b) 68.7 % (a) 81.4 % (b) 68.7 % (a) 81.4 % (b) 68.7 % Indicator of ‘Relative Own- Language Preference’ (a) 98.6 = 1.2 81.4 (b) 98.6 = 1.4 68.7 (a) 98.8 = 1.2 81.4 (b) 98.8 = 1.4 68.7 (a) 99.3 = 1.2 81.4 (b) 99.3 = 1.4 68.7 (a) 99.5 = 1.2 81.4 (b) 99.5 = 1.4 68.7 (a) 89.6 = 1.1 81.4 (b) 89.6 = 1.3 68.7  T A B L E 1 ; P a r t II FREQUENCY DISTRIBUTION (IN %) OF LANGUAGES OF CITED REFERENCES IN FOUR GERMAN AND FRENCH SOCIOLOGY JOURNALS (1985-1994) AND THEIR DERIVED INDICATORS OF ‘RELATIVE OWN-LANGUAGE PREFERENCE’ Language of cited references Koelner Zeitschrift f. Soziologie Sozialwissen. Information. Revue francaise de Sociologie L’Annee Sociologique English German French Other 35.2 63.3 1.3 0.2 11.0 85.5 3.6 0 35.5 0 64.5 0 29.6 1.7 67.5 1.2 Total 100.0 % 100.0 % 100.0 % 100.0 % Total no. of references checked 1273 1064 1267 1315 N (Total no. of articles checked) 60 60 60 60 Ref. mean Ref. median 43.1 39.0 19.8 18.2 28.2 26.1 35.1 29.3 Estimat. prop. of that lang. in sociological research (a) 3.0 % (b) 5.0 % (a) 3.0 % (b) 5.0 % (a) 4.8 % (b) 8.1 % (a) 4.8 % (b) 8.1 % Indicator of ‘Relative Own- Language Preference’ (a) 63.3 = 21.1 3.0 (b) 63.3 = 12.7 5.0 (a)85.5 = 28.5 3.0 (b) 85.5 =17.1 5.0 (a) 64.5 = 13.4 4.8 (b) 64.5 = 8.0 8.1 (a)67.5= 14.1 4.8 (b) 67.5 = 8.3 8.1  T A B L E 2 ESTIMATED FREQUENCY DISTRIBUTION OF LANGUAGE OF SOCIOLOGY RESEARCH LITERATURE (1965-1995) ---------------------------------------------------------------- ESTIMATE (a) ESTIMATE (b) ---------------------------------------------------------------- LANGUAGE No. % No. % ---------------------------------------------------------------- ENGLISH 193,991 81.4 193,991 68.7 GERMAN 7,085 3.0 14,170 5.0 FRENCH 11,510 4.8 23,020 8.1 OTHER 25,697 10.8 51,394 18.2 ---------------------------------------------------------------- TOTAL 238,283 100.0% 282,575 100.0% Estimate (a) : According to SOCIOFILE CD-ROM 1/74 - 12/95. Estimate (b) : Assuming that only 50% of the non-English items are included in this database. SOCIOFILE 1/74-12/95 CD-ROM database (by Sociological Abstracts, inc.) covers sociology and related disciplines, theoretical and applied. It contains information from approximately 1600 journals in 30 different languages from about 55 countries, by including bibliographical citations and abstracts (from Sociological Abstracts and from SOPODA - Social Planning/Policy and Development Abstracts) and enhanced dissertation citations (from Dissertations Abstracts International, 1986- ). Journals published by various sociological organizations and periodicals containing the term 'sociology' in their titles, are abstracted fully, irrespective of language or country of publication, while other journals in related fields are scanned for relevant articles. Actually, the abovementioned CD-ROM file includes also over 11,000 items published in 1965-1973. Estimate (a) in Table 2 shows the frequency distribution of the languages of all items included in this file. However, since this file is likely to be biased towards items published in English, an adjustment was made by doubling the quantity of non-English items, assuming they are underrepresented in the file. The corrected proportions, represented by estimate (b), might be a better estimate of the entire body of research literature existing in the field. The 'relative own-language preference’ (ROLP) indicators presented in Table 1 were calculated by dividing the ratio of 'language self-citation' by the estimated ratio of that language in the existing body of research literature of that field. These ROLP indicators reveal that the highest degree of 'own-language preference' (i.e. the strongest bias towards their mother-tongue) is found among the Germans, whose ratio of references in the German language exceeds 21 to 28 times (according to estimate a) the amount expected according to the German language share in world sociology research. Next come French sociologists whose use of their own language exceeds about 14 times the expected ratio, while American and British ones display a much better situation, with coefficients slightly higher than 1, indicating the lowest own-language bias, being only slightly higher than expected. Calculating the ROLP indicators according to estimate b, which gives higher proportions to other languages in the existing body of research literature at the expense of English, naturally shows slightly higher rates of relative language self-citation among the American and British sociologists (respectively 1.4 and 1.3 vs. 1.2 and 1.1 according to estimate a). Understandably, estimate b improves the German and French indicators, which drop considerably (from 21-28.5 to 13-17 and from 13-14 to 8 respectively), but it remains relatively much higher, compared to their Anglo-Saxon colleagues, still indicating a strong ‘bias’ (or preference) towards one's mother-tongue, 8 to 17 times higher than expected according to its share. Mutual Use of Languages It is actually possible to construct a coefficient matrix, by calculating for each group of sociologists (i.e. each periodical) the proportion of their use of a certain language, in relation to its part in the existing body of research literature. Evidently, only one journal (Theory, Culture and Society - UK) out of the five included in the British-American group displays a more 'balanced' use of both German and French sources. Its use of references in German and French mostly exceeds (or equals) the expected proportion (i.e. their estimated share) by a factor of 1.6 and 1.1 respectively (or slightly lower if estimate b is used). Next come the French whose use of English sources range between 0.36 and 0.44 of the latter's estimated share in the research literature of sociology, and it is slightly higher if estimate b is used. The two French journals differ, however, concerning the use of German sources : none in RFS vs. factor of 0.57 in l’Annee Soc., probably due to the fact that the latter is a yearbook reviewing also international trends in sociology. The coefficients for the two German journals indicate low to fairly balanced use of French sources (0.3-0.75 of the expected values), but a relatively lower use of sources in English, only a small fraction (0.13 or 0.43) of the expected according to the English estimated share. This finding corroborates the one mentioned above (in Table 1) which showed the German journals to have the highest ROLP indicator. The mutual-use coefficients for the British-American group (except for TCS) puts it in the 'worst' position relative to the German and French journals, since their references indicate only a fractional use of German sources (0.23 or less) and even lower use (0.17 or less) of French ones. This finding corroborates former ones by Louttit (1955) and Finison (1975) concerning the field of psychology. Possible Explanations 1. To be sure, the ‘self-citation’ phenomenon is not limited, as is well known, to the language dimension, but rather exists in some other parameters like authors, journals, countries. institutions, etc. There seem to be more than one reason accounting for this phenomenon, a detailed discussion of which is beyond the scope of the present paper. 2. One may rightly assume, however, that the relatively greater availability and accessibility of certain publications, compared to others, is probably one of the reasons. Obviously, this factor enhances the use and citation of own-language material, which is usually easier to obtain. Many, if not most, academic libraries in the US and UK do not hold the full array of French and German scholarly journals, not to mention monographs, and to some extent a similar situation exists in French and German academic institutions regarding scholarly journals and monographs in English. 3. One may rightly assume that the leading foreign journals and the important monographs in each of these three languages are being held by academic libraries in these countries, and most others may be obtained via inter-library loan. Thus, in view of the figures in Table 1 (Parts I and II) there is no doubt that the ‘language barrier’ is at work here, besides the ‘immediate availability’ factor. It is likely to assume that a many scholars will prefer publications written in their mother-tongue and will not go into excessive efforts in order to get material in foreign language, especially if it has to be obtained from distant libraries by a special procedure. 4. Scholars, apparently, tend to use and cite sources with which they became familiar by learning and reading, during their academic studies and later. Since local universities put relatively more emphasis on local publications it increases the proportion of own-language sources in the list of references appended to a paper. 5. Some of the topics discussed in sociology (and other) journals deal with local issues, for which most of the sources are written in the local language. CONCLUSIONS 1. To determine the ‘relative own-language preference’ (ROLP) existing among groups of scholars in a certain field one should relate the proportion of 'language self-citation' of a group of scholars publishing in a certain language, to some estimate of this language share in the existing body of research literature of this field. 2. Similarly, 'mutual-use' coefficients can be constructed by relating this group's proportional use of foreign languages to each language share in the research literature. 3. When applied to sociology these measures show the German and French sociologists, publishing in the studied periodicals, to have a higher ROLP, biased towards their own mother-tongue, compared to American and British ones. 4. The 'mutual-use' matrix gives one British journal the highest 'grades' in using literature in German and French, followed by the French journals and the German ones, while the American journal and the three other British ones display only a fractional use of existing German and French publications. * I am grateful to my student, Ms. Ida Isaac, for her part in collecting and processing the data for four of the journals. I am also grateful to two anonymous referees, whose comments helped me considerably in shaping the final version of this paper. R E F E R E N C E S Bradely D. (1993) European elites envy American cohesion. Science 260 : 1738-1739. Chan, G.K.L. (1976) The foreign language barrier in science and technology. International Library Review 8 : 317-325. Earle, P. and Vickery, B. (1969) Social science research literature use in the UK as indicated by citations. Journal of Documentation 25 : 123-141. Ellen, S.R. (1979) Survey of foreign language problems facing the research worker. Interlending Review 7(2): 31-41. Engelbert, H. (1976) The terminological barrier to the understanding of scientific and technological literature. Translation News 6(2) : 2-10. Finison, L.J. and C.L. Whittemore. (1975) Linguistic isolation of American social psychology. American Psychologist 30 : 513-516. Fitzgibbons, S.A. (1980) Citation analysis in the social sciences. in: Stuart, R.D. and Miller, G.B. eds., A Collection Development in Libraries; a Treatise. Greenwich, Conn. JAI,. Part B, 291-344. Garfield, E. (1980) Do French scientists who publish outside of France and/or in English do better research? in: Garfield, E. Essays of an Information Scientist. vol. 3. Philadelphia, ISI Press , 498-503. Guttsman, W.L. (1966) The literature of the social sciences and provision for research in them. Journal of Documentation 22 : 186-194. Hanson, C.W. (1962) The Foreign Language Barrier in Science and Technology. London, Aslib. Holmstrom, J.E. (1973) The foreign language barrier, in: Sherrod, J. and Hodina, A. eds., Reader in Science Information. Washington, D.C.,Microcard, , 94-103. Hutchins, W.J. a.o. (1971) The Language Barrier. University of Sheffield. Inman, C.R.H. (1983) Scientific Publications in English and developing countries: a report of a survey of scientists' experiences. Journal of Information Science 6 : 159-164. Jagodzinski-Sigogneau, M. a.o. (1982) How to measure the degree of independence of a research system? Scientometrics 4 : 119-133. Large, J.A. (1983) The Foreign-Language Barrier ; Problems in Scientific Communication. London, Andre Deutsch. Leck, C.F. (1978) Foreign languages and biologists today. Bioscience 28 : 367. Louttit, C.M. (1955) The use of foreign languages by psychologists, American Journal of Psychology 68 : 484-486. Michel, J. (1982) Linguistic and political barriers in the international transfer of information in science and technology. Journal of Information Science 5 : 131-135. Morgan, B.A. (1977) National bias in reference citation in English language agricultural engineering journals. IAALD Quarterly Bulletin 22(3-4) : 60-64. Nicholas, D. and Ritchie, M. (1978) Literature and Bibliometrics. London, Clive Bingley. Ralph, A. (1982) Language and information retrieval in the social sciences. Aslib Proceedings 34 : 394-405. Regaunt, S. (1993) What lingua franca? Nature 361 : 107. Regaunt, S. (1994) English as lingua franca in geological scientific research ; a bibliometric study. Scientometrics 29 : 335-351. Schmidt, J.M. (1979) Translation of periodical literature in plant pathology. Special Libraries 70 : 12-17. Sherwood, B.A. (1979) Universal language requirement. Physics Today 32 : 9. Stewart, J. (1970) The literature of politics: a citation analysis. International Library Review 2 : 329-353. Thorp, R.G. a.o. (1989) The foreign language barrier : a study among pharma ceutical research workers. Journal of Information Science 14 : 17-24. Wellisch, H. (1973) Linguistic and semantic problems in the use of English-language information services in non-English-speaking countries. International Library Review 5 : 147-162. Wood, D.N. (1967) The foreign-language problem facing scientists and technologists in the U.K.-report of a recent survey. Journal of Documentation 23(2) : 117-130. Wood, D.N. and Bower, C.A. (1969) The use of social science periodical literature. Journal of Documentation 25 : 108-122. Yitzhaki, M. (1988) The language barrier in the humanities:the case of biblical studies. INFORMETRICS 87/88; Select Proc. of the First International Conf. on Bibliometrics, Diepenbeek, Belgium, Aug. 1987. Amsterdam. Elsevier, 301-314. Yitzhaki, M. (1989) Subject dispersion and discipline affiliation. Scripta Hierosolymitana (Publications of the Hebrew University of Jerusalem) 29 : 217-249. PAGE 13