vankampen.p65 28 College & Research Libraries January 2004 Development and Validation of the Multidimensional Library Anxiety Scale Doris J. Van Kampen This article reports on the development and validation of the Multidi­ mensional Library Anxiety Scale, which was designed to assess six di­ mensions of an individual’s perception of an academic library and the information search process (ISP). A two-part study was undertaken for the development of the instrument. In part one, twenty-one participants completed a pilot questionnaire that assisted the researcher in develop­ ing the instrument. In part two, 278 participants completed a revised questionnaire consisting of a 54-item Likert-type scale that assessed levels of library anxiety experienced by students enrolled in a doctoral degree–granting program at an urban southeastern university. rustration over how to begin a search for a topic and how to find information related to the topic are recurrent themes when students write about their library experiences.1–3 Library anxiety has been reported as a characteristic among gradu­ ate and undergraduate students, with an estimated 95 percent of college students engaging in frequent academic procras­ tination because of library anxiety.4, 5 This phenomenon has been discussed in the literature and quantitatively measured by a number of researchers over the past decade using the Library Anxiety Scale (LAS) developed by Sharon Bostick. The Library Anxiety Scale was created “to cat­ egorize and measure this concept in col­ lege students.”6 The first formalized study of library anxiety as a phenomenon occurred in the mid-1980s, when Constance Mellon con­ ducted a two-year study of undergradu­ ate beginning composition classes at a southern university.7 Twenty English in­ structors assigned and collected personal writings from students describing their library search experiences. The diary-like entries described how students under­ took their searches and their feelings about the process. The major themes that emerged were confusion, fear, and the feeling of being lost. Four strands con­ cerning the “lost in the library” theme were noted: “the size of the library; not knowing where things were; not know­ ing what to do; and not knowing how to begin the [library] research process.”8 From her research, Mellon developed a theory about library anxiety, which stated, “When confronted with the need to gather information in the library … many students become so anxious that they are unable to approach the problem logically or effectively.”9 Mellon’s theory of library anxiety was qualitatively tested, but the question re­ mained as to whether a valid and reliable Doris J. Van Kampen is Systems Librarian at Saint Leo University; e-mail: doris.vankampen@saintleo.edu. 28 mailto:doris.vankampen@saintleo.edu Development and Validation of the Multidimensional Library Anxiety Scale 29 instrument could be developed to quan­ titatively measure library anxiety. In 1992, Sharon Bostick developed and validated the LAS, which was designed to catego­ rize and measure library anxiety in col­ lege students at two- and four- year insti­ tutions.10 The LAS measures five compo­ nents: (1) barriers with staff, (2) affective barriers, (3) comfort with the library, (4) knowledge of the library, and (5) me­ chanical barriers. “Barriers with staff” refers to a library patron’s perception of library staff as intimidating and unap­ proachable as well as being too busy to help. “Affective barriers” measures the respondent’s feeling of adequacy when using the library, and “comfort” reflects how safe, welcoming, and nonthreaten­ ing the library is perceived to be. Feelings of inadequacy and discomfort regarding the library have been acknowledged as a limiting factor to students’ research ef­ forts in the library. “Knowledge of the li­ brary” reflects student perspectives on how familiar they think they are with the library and its resources, and “mechani­ cal barriers” examines the feelings that emerge as a result of student reliance on library equipment. However, when the LAS was devel­ oped, scant information and few theories were available on the user’s feelings dur­ ing the research process itself; the Internet was not yet widely used as a research tool, and database access was limited prima­ rily to the physical confines of the library. User studies were limited mostly to sta­ tistics reflecting usage and how the user interacted with the library as a system as opposed to examining user perspectives of the research process. Even now, the ma­ jority of current user research focuses on levels of satisfaction with services or on how patrons interact with the library’s database or Web site, rather than on how a person may feel about the library, tech­ nology, and the process of searching for information. “Most of the current patron user research in academic libraries focuses on levels of satisfaction with current ser­ vices and on how patrons are navigating the library OPAC or website.”11 A 2001 sampling of reports by this author from user satisfaction surveys in library re­ search studies did not find any that indi­ cated that the library in question was de­ termined to be unsatisfactory in any sig­ nificant area. The vast majority of surveys rated library satisfaction levels as high or very high. However, many of the survey results noted an increased demand for more full-text databases and online ser­ vices. The researcher was interested in whether doctoral students, who should have had experience with the ISP and us­ ing the library, showed evidence of library anxiety, and if they did, whether anxiety levels varied over time. Other questions of interest to the researcher concerned how gender might influence use of tradi­ tional or online use of resources and whether doctoral students who exhibited higher levels of anxiety stated a prefer­ ence for using online resources. However, before these questions could be answered, an instrument needed to be located or de­ veloped for this purpose. Because of the age of the original LAS and other devel­ opments in the field of library user re­ search, it was deemed appropriate to cre­ ate a new instrument based on the LAS, named the Multidimensional Library Anxiety Scale (MLAS), which would take these factors and the researcher ’s ques­ tions into consideration.12 Although this instrument in no way invalidates the LAS, it does take into consideration off-cam­ pus use of library resources and doctoral students’ attitudes during the course of their dissertations. Statements regarding respondents’ preferences concerning use of the library online or in person and statements that reflected comfort level with computers were included. The initial set of potential statements concerning online resources was e-mailed or mailed using the U. S. postal service with a cover letter to librarians and fac­ ulty from academic libraries and to Sharon Bostick. Codes were developed for use in SPSS 10.0 using Bostick’s LAS as a reference tool for placement of the variable within a general classification http:consideration.12 http:tutions.10 30 College & Research Libraries January 2004 scheme. Following revisions to the sur­ vey, a pilot study was conducted to de­ termine the potential validity of the in­ strument. The pilot study instrument consisted of fifty-seven statements and was con­ ducted during the summer semester of 2001 at a large metropolitan-based uni­ versity in the South. The sample popula­ tion consisted of eighteen doctoral stu­ dents enrolled in EDG 7931, Structural Equation Models, an elective course for all graduate education majors in the Col­ lege of Education. The sample population selected was based on willingness of fac­ ulty members to allow students to have time in class to participate. A five point Likert–type scale was used for the pilot survey, with available an­ swers being “strongly agree,” “agree,” “undecided,” “disagree,” and “strongly disagree.” The pilot instrument state­ ments were printed on a word processor and photocopied. Directions were printed at the top, and a business card was in­ cluded in the packet inviting the students to e-mail comments or questions about the survey to the researcher. A test–retest method was used for the pilot to improve reliability of the instru­ ment; this method is used when a re­ searcher is interested in whether the par­ ticipants will answer a set of statements the same way more than once over a pe­ riod of time. The pilot study participants were given the survey first with coded envelopes to identify respondents and again three weeks later to determine the consistency of their answers. The direc­ tions asked students to respond to the statements by circling the number that best correlated with their feelings regard­ ing the statement. Upon completion of the pilot study, data were input into SPSS 10.0. Negatively worded statements were reverse-scored so that all the statements were scored in the same direction. Exploratory Factor Analy­ sis (EFA) was utilized on the data collected to extract latent variables and examine cor­ relations. According to Kevin Kieffer, “EFA is an analytic technique in which the pri­ mary concern is to reduce a larger set of variables into a smaller and more manage­ able set based in the consistency of the data.”13 EFA assists with detecting and analyzing possible patterns based on the correlations among the variables. It is used when there is “initial uncertainty as to the number of factors being measured.”14 The purpose was to summarize the interrela­ tionships among the fifty-three variables (statements) in order to assist in the analy­ sis and conceptualization of possible cat­ egories. The extraction method used was principle components, with a varimax pro­ cedure, excluding cases listwise and sort­ ing factors by size. Using this method, vari­ ables with an absolute value of less than .30 were excluded. The reason for exclud­ ing variables below this threshold was to eliminate variables with lower correlations and to retain variables with higher corre­ lations, aiding in the interpretation of the results. If one examines similar research using EFA, one will find that most studies use .30 as a rule of thumb for determining “noteworthy coefficient magnitudes.”15 The initial factor analysis yielded eleven factors in twenty-three iterations. For the pilot study, a bivariate correlation analy­ sis using Pearson’s r was completed after breaking each factor into clusters, and then the statements and the correlations were examined. The researcher reflected on the software program’s placement and de­ cided to move several items from one fac­ tor to another until most items were cor­ relating at .45 or higher. This was done based on consultation with an expert in the field and on the researcher’s own opin­ ion as to where the items would make the best fit because “analytic results can inform the definitions we wish to create.”16 Cor­ relations were rerun on each factor after the items were moved to examine the in­ teraction among the items. Most correla­ tions were significant at the .01 level, and all but a few items were significant at the .05 level. Because of the small size of the pilot sample, it was decided to keep sev­ eral of the items with correlations more than .30 and less than .45 to determine whether the correlations were due more Development and Validation of the Multidimensional Library Anxiety Scale 31 due to the implementation of several strategies, which have been reported to increase response rates for mail surveys.17– 19 The cost for printing and mailing the initial packet with four first-class stamps, two envelopes, five sheets of paper, and the bookmark was $2.10 per mailing, for a total of $1,163.40. The cost of purchas­ ing, printing, and mailing the postcard re­ minders was 49 cents per postcard ($131.32 total), with 268 postcards being mailed on January 19, 2002. After ten weeks, a total of 278 surveys had been completed and returned, bringing the cost rate to $ 4.65 per survey. Data from survey responses were in­ put into SPSS 10.0. Negatively worded variables were reverse-scored. Measures of central tendency and dispersion were explored after a new factor was created using SPSS 10.0 Syntax COMPUTE [ ] command, which compiled the total score for each case, excluding cases with miss­ ing data. Distribution of the data was symmetric, as depicted in the stem & leaf plot. A stem & leaf plot is a frequency table that has been graphically depicted, simi­ lar to a bar graph rotated clockwise. (See figure 1.) EFA was utilized on the survey data to extract latent variables and examine corre- FIGURE 1 Stem & Leaf Plot Frequency Stem & Leaf 2.00 16 . & 4.00 17 . && 16.00 18 . 134579& 17.00 19 . 13389&& 22.00 20 . 02345568& 38.00 21 . 022334455677788999& 45.00 22 . 00122344456677888999 41.00 23 . 000134445556677899& 33.00 24 . 1111235567789& 23.00 25 . 023344557& 6.00 26 . 04& 1.00 27 . & Stem width: 10.00. Each 1eaf: 2 case(s) & denotes fractiona1 1eaves to the sample’s size and less to the item itself. Revisions to the instrument were based the following criteria: • lack of correlation with other items (item removed); • perceived lack of clarity based on participant feedback (item rewritten); • relocation within the instrument to reflect intercorrelations; • perception of respondents (based on how they answered) that the item ap­ peared to measure some other construct than the one the researcher had believed it would. Each variable retained the same code that had been assigned during the pilot study. The researcher decided that it would be better to retain the same codes for each statement rather than create new ones so that it would be easier to refer back to the pilot data, if needed. Data Collection: Survey The full study was conducted by mailing questionnaires to the entire doctoral population of 554 who had been enrolled at the institution during the spring of 2001. The questionnaires were mailed from January 1 through February 23, 2002. A letter explaining the purpose of the sur­ vey was sent with the instrument. Each letter was personalized and signed by hand, with one statement in red ink to emphasize the need for each student’s participation. All envelopes were hand­ written, with a collector ’s stamp on each envelop and a self-addressed stamped envelope included with each mailing. A bookmark was enclosed with each survey as a token gift. Poten­ tial respondents were informed that if they returned a completed survey by January 21, 2002, they would be eligible for several small prizes. A follow-up postcard was mailed a little over two weeks after the initial mailing. Sending a duplicate packet was considered but because of the cost of each packet and the diminishing return rate, the re­ searcher decided not to resend the pack­ ets. It was estimated that a high percent­ age (over 50%) would return the survey http:1,163.40 http:surveys.17 32 College & Research Libraries January 2004 lations. The extraction method used was principal compo­ nents, rotated orthagonally FIGURE 2 Scree Plot, Full Study using a varimax procedure. “Principle components are very closely tied to the origi­ nal variables, with each subject’s score on a principle component being a linear com­ bination of his or her scores on the original variables.”20 Factors with an absolute value of less than .30 were excluded. This was done based on the researcher ’s readings in this field and af­ ter discussion with an expert in the field on where an ap­ propriate cut-off point should be located. The initial factor analysis yielded sixteen compo­ nents after thirty-one iterations. The Scree plot leveled off after five factors. (See fig­ ure 2.) After examining the Scree plot, eigen­ values, components matrix, and correla­ tions matrix, the researcher decided to force the factor analysis into seven com­ ponents, as all but one of the components factoring on factors 8–16 were also factor­ ing on components 1–7. The seven factors fell generally into the following categories: 1. perceived knowledge of the library and confidence concerning the ISP; 2. the ISP and general library anxiety; 3. perceived importance of the library and constraints to use; 4. comfort level with technology, in­ teraction with staff; 5. library independence; 6. comfort level while inside the li­ brary building; 7. perceived barriers concerning staff. Upon further examination of the fac­ tors, the researcher decided factors seven and four had too many overlapping vari­ ables (staff issues) to be considered as separate factors and forced the factor analysis into six factors to avoid attenu­ ating the results. Each component re­ tained had at least three significant load­ ings and an eigenvalue greater than 1.0. Each retained component was considered a major component and, as such, would probably be of interest to other investiga­ tors.21 Once again, the extraction method used was principal components. Table 1 represents the initial eigenval­ ues of the six factors, the rotated (trace) eigenvalues for each factor, the percent of variance and cumulative variance, and the number of items factoring on each component as derived from the forced factor analysis. To test for internal consistency, Cronbach’s Alpha was computed on the fifty-three statements. Cronbach’s Alpha is a common index of reliability. When using this measure of reliability, the closer the re­ sults are to 1.0, the higher the estimate of reliability. The resultant alpha coefficient of .88 for all fifty-three items provided evi­ dence of adequate internal consistency. A new factor was derived for each of the six components using the COMPUTE fac­ tor [ ] syntax command in SPSS 10.0. Keep in mind that a factor is not a variable; it is “the reduction of a larger set of variables into a smaller and more manageable set based on the consistency of the data.”22 Fac­ tors were labeled as follows: 1. KNOW: Comfort and confidence (li­ brary independence) when using the li­ brary; Development and Validation of the Multidimensional Library Anxiety Scale 33 2. ISPLIB: The In­ TABLE 1 formation Search Pro­ Forced Factor Analysis of Survey cess and general Li­ brary Anxiety; Factor # Initial Variance % of Number of 3. STAF: Perceived Eiegen. Value % Cum. % Trace Items barriers concerning 1 11.529 staff; 2 3.865 4. IMPLIB: Per­ 3 3.169 ceived importance of 4 2.931 understanding how 5 1.921 to use the library; 6 1.756 5. TECH: Comfort level with technology *rotated **lUIotatedand as it applies to the library; 6. BUIL: Comfort level while inside the library building. Labels developed in this manner were designed to correspond with the forced factor analysis. The rotated components matrix was examined for overlapping variables, and some overlapping vari­ ables were removed from one or more of the factors for better analysis. The deci­ sion as to which items to remove was made after examining the rotated compo­ nents matrix, the correlations matrix, de­ scriptive statistics within Cronbach’s Al­ pha for each factor ’s item, scale, and scale, if item removed. Along with overlapping variables on some factors, two variables did not manifest on any factor. If an over­ lapping variable was considered of inter­ est to the researcher ’s questions in more than one area, a judgment was made as to whether to leave it on two factors. No variable was allowed to overlap on more than two factors. Cronbach’s Alpha was then run on each separate factor to test for internal consistency within each factor. Score reli­ ability, as measured by coefficient alpha, for each subscale was as follows: 1. Comfort and confidence when us­ ing the library = .86 2. Information Search Process and gen­ eral Library Anxiety = .87 3. Barriers concerning staff = .73 4. Importance of understanding how to use the library = .79 5. Comfort level with technology and as it applies to the library = .73 5.698 19.87 5.698 21* 31** 6.664 26.541 5.377 21* 14** 5.463 32.004 4.302 11* 18** 5.053 37.057 3.820 10* 10** 3.313 40.369 3.165 9* 8** 3.027 43.396 2.808 10* 6** 6. Comfort level while inside the li­ brary building = .74 Although the average interitem corre­ lation was high, several items were re­ moved after Cronbach’s Alpha had been run. This is considered a normal proce­ dure. Some items were removed in view of the fact that, although they were giving adequate correlations and did not detract from the Alpha scores, they were not en­ hancing the scores and overlapped with another factor. One item (would rather use the library in person) on factor 5 was re­ moved from factor 5, but not from the scale when it was noted that it was depressing the Alpha score of that factor by .18 (be­ fore removal: .5513; after removal: .7322). This was done because there was insuffi­ cient evidence that the item was measur­ ing the same underlying construct as the other items within that factor. This item would perhaps show a different correla­ tion with regard to the other factors if the survey were given to another population. The following conclusions were drawn concerning the survey instrument: • Library anxiety in an academic li­ brary setting can be measured using this survey instrument. • Intercorrelations for all fifty-three factors were sufficient to ensure internal consistency among the items. • Intercorrelations for each factor were sufficient to ensure internal consis­ tency among the items within each factor. • If used on a similar population, the instrument should be sufficiently stable 34 College & Research Libraries January 2004 to produce results that measure the six dimensions of library anxiety identified during factor analysis. The first phase of research, developing an instrument, was considered complete. The survey had been returned by a suffi­ cient number of students, the factor analy­ sis did not reveal any significant problems, and further analysis could begin. Summary This developmental study of the MLAS indicated that the questionnaire scale showed good internal consistency and construct validity and that the scale has the potential to be a useful tool for deter­ mining what aspects of the library and the information search process are per­ ceived to be barriers by graduate stu­ dents. Researchers interested in graduate students as subjects might wish to use the scale, particularly with master’s- and doc- toral-level subjects involved in a thesis or dissertation. Further publication by the author concerning the actual results of the study are forthcoming. Researchers inter­ ested in obtaining a copy of the scale are encouraged to contact the author. Addi­ tional research into the usefulness and stability of the scale for similar or dissimi­ lar populations at other institutions is in­ dicated. Notes 1. Constance Mellon, “Library Anxiety: A Grounded Theory and Its Development,” College & Research Libraries 47 (Mar. 1986): 160–65. 2. ———, Naturalistic Inquiry for Library Science: Methods and Applications for Research, Evalu­ ation, and Teaching (New York: Greenwood Pr., 1990). 3. Carol C. Kuhlthau, Seeking Meaning: A Process Approach to Library and Information Services (Norwood, N.J.: Ablex, 1993). 4. Sharon Bostick, “The Development and Validation of a Library Anxiety Scale” (Detroit, Mich.: Wayne State University, 1992). Abstract in Digital Dissertation. 5. Anthony Onwuegbuzie and Qun Jiao, “I’ll Go to the Library Later: The Relationship be­ tween Academic Procrastination and Library Anxiety,” College & Research Libraries 61(1): 45-54. Available online from http://library.ucf.edu. (Retrieved 1 October 2001.) 6. Bostick, “The Development and Validation of a Library Anxiety Scale.” 7. Mellon, “Attitudes: The Forgotten Dimension in Library Instruction,” Library Journal 113 (Sept. 1988): 137–39. 8. Ibid. 9. Ibid. 10. Bostick, “The Development and Validation of a Library Anxiety Scale.” 11. Doris J. Van Kampen, Library Anxiety, the Information Search Process, and Doctoral Students’ Use of the Library (Orlando, Fla.: University of Central Florida, 2003). 12. MLAS © 2003 Doris J. Van Kampen and Sharon Bostick. To obtain a copy of the MLAS, please contact the author at doris.vankampen@saintleo.edu. 13. Kevin Kieffer, “An Introductory Primer on the Appropriate Use of Exploratory and Con­ firmatory Factor Analysis,” Research in Schools 6 (fall 1999): 75–92. 14. Joseph Capelleri and Robert Gerber, Encyclopedia of Biopharmaceutical Statistics (Cambridge, Mass.: Cambridge Biostatistics and Data Management, Inc., 2003). 15. Kieffer, “An Introductory Primer on the Appropriate Use of Exploratory and Confirma­ tory Factor Analysis,” 80. 16. Bruce Thompson and Larry Daniel, “Factor Analytic Evidence for the Construct Validity of Scores: A Historical Overview and Some Guidelines,” Educational and Psychological Measure­ ment 56 (Apr. 1996): 197–208. 17. Earl R. Babbie, Survey Research Methods (Belmont, Calif.: Wadsworth, 1990). 18. Francis Buttle, “Questionnaire Color and Mail Survey Response Rate,” Journal of the Mar­ ket Research Society 39 (Oct. 1997): 625–26. 19. David Dodd and Barbara Markwiese, “Survey Response Rate as a Function of Personal­ ized Signature on Cover Letter,” Journal of Social Psychology 127 (Feb. 1987): 97–98. 20. Richard Harris, Primer of Multivariate Statistics (London: Lawrence Erlbaum Assoc., 2001). 21. William Zwick and Wayne Velicer, “Comparison of Five Rules for Determining the Num­ ber of Components to Retain,” Psychological Bulletin l99 (1986): 432–42. 22. Kieffer “An Introductory Primer on the Appropriate Use of Exploratory and Confirma­ tory Factor Analysis,” 76. mailto:doris.vankampen@saintleo.edu http:http://library.ucf.edu