eJUSt - Stanford University Libraries and Highwire Press
About eJUSt
Research Methodology
Research Team
Research Findings
Expert Workshops
Privacy Policy
Frequently Asked Questions
  Research Findings

MARCH 2002


I.1. List of Participating Societies

Members of 20 professional societies in life sciences were sampled for the Scientific Journal User survey May 22 to June 20, 2001. The following table shows the names of societies and sample sizes per society.

Society Sample size*
American Society for Microbiology 2,279
Radiological Society of North America 1,443
American Society for Biochemistry and Molecular Biology 1,249
American Physiological Society 1,091
Biophysical Society 873
The Endocrine Society 701
American Society of Plant Physiologists 679
American Association of Immunologists 658
American Association for Clinical Chemistry 616
Society for Study of Reproduction 567
American College of Chest Physicians 455
American Society for Pharmacology & Experimental Therapeutics 451
American Society for Nutritional Sciences 352
American Heart Association 350
Botanical Society of America 339
The Society of Investigative Dermatology 168
Genetics Society of America 153
American Psychosomatic Society 145
American Society of Hematology 22
Royal Society of Medicine Press 18

*The number exceeds total sample size (12,465) because some respondents belong to multiple societies

I.2. Sampling Methods:

The first survey was conducted online; data were collected from a Web-based questionnaire, recorded in the form of TSV files, by a third-party specialist firm called Perseus. The purpose of the survey (as described in greater detail elsewhere) was to uncover the perceptions of scientists about e-journals.

Target Population:

Life scientists and medical professionals who read life and medical science journals, especially HighWire journals in either paper or e-journal format (both users and nonusers of e-journals)

Survey Design:

  • Sampling: We contacted 70 scientific societies affiliated with HighWire Press to request membership information such as e-mail addresses. Twenty societies that broadly represent life sciences–biological sciences, health sciences, and agricultural sciences–agreed to release their member information (limited to name, e-mail address, and membership status–e.g., active, student, retired) for the survey.
  • Online questionnaire: This is best viewed (though not active for data collection) at http://ejust.stanford.edu/firstsurvey-linked.htm
  • Solicitation methods: We sent 108,774 e-mail solicitations, requesting members to respond via the questionnaire on the Web. Approximately 13,903 addresses returned "undeliverable" or vacation messages, resulting in a contact group of approximately 94,871.
  • Survey implementation and data collection period: May 22 to June 20, 2001
  • Survey response rate: 12,453 net responses were received during the period; the survey response rate is 13.14%

Sample Population:

  • Sample size: The number of sample observations was 12,465
  • Sample distribution (demographic characteristics):
    • Gender: 30% of the respondents are female.
    • Age: The average (mean) age of respondents is 47; the median age is 48.
    • Research fields: 60% of respondents reported their research field as biological sciences, 34% health sciences, 4% agricultural, and 2% other research fields related to life sciences.
    • Occupation: 38% of respondents are faculty members in academic institutions, 12% students and post-doctorate researchers, 11% researchers in academic institutions, 11% researchers in the private sector, 6% government agency, 17% medical doctors, and 3% retirees.
    • Residence: Respondents are from 99 different countries. The majority reside in the United States (72%), and the rest are from Canada and countries in Europe, Asia, Africa, Latin America, and Australia.


II. 1. Econometric Models for Analyses

The study has three dependent variables to (1) e-journal usage; (2) scientists' preferences regarding printed copy versus online editions; (3) scientists' perception about the impact of e-journals on their research activities.

Ordered probit regression is used to measure equation (1), e-journals usage, and logistic regressions for equations (2) and (3), preference regarding journal formats and perception about e-journals.


where EJU is a degree of e-journal usage with 5 categories (see Table 2); PREF is a dependent variable, preferences regarding journal format, printed copy vs online versions; and PEJ represents perception about the impact of e-journals on research activities. DEMO1 is a set of demographic characteristics such as AGE, MD, BIOLOGY, MALE, USA-CAN which represent respondents' age, whether a respondent is a medical doctor, whether a respondent's field of research is biology, whether a respondent is male, and an indicator that a respondent resides in either the U.S. or Canada, respectively (refer Table 1); INTERNET is a variable to measure Internet familiarity–weekly hours spent on the Internet; SUBS represents the resources available through personal subscriptions–numbers of journal subscriptions; FACILITY and ACCESS represent available resources such as PC, printers, and Internet access and scientific journal access through institutions; PAPER1, the numbers of papers submitted in the past year, PAPER2, the number of papers published or accepted in the journals during the past year, which measure respondents' research productivity. DEMO2 represents a set of variables in DEMO1 and JOBEXP, job experience measured by years (refer Table 1). P^[EJU] is a predicted probability of e-journal usage from equation (1).

EJU variable consists of five categorical values (question 2 in Table 2), while values of both PREF (questions 4 and 5 in Table 3 and Table 4) and PEJ variables (questions 17, 14, 15, and 16 in Table 5, Table 6, and Table 7) are binary. Five categories for EJU are yesterday or today, within last week, within a month, longer than a month, and never. A binary value for PREF and PEJ is "Agree" and "Disagree." The binary value has created by recoding four categories, "Strongly Agree," "Somewhat Agree," "Somewhat Disagree," and "Strongly Disagree," into two categories, "Agree" and "Disagree." Table 2 shows a question on EJU and responses per category. Many different questions were used to analyze preferences regarding formats including reasons of preference. Table 3 and Table 4 describe questions for PREF variable. We measured a predicted probability of whether they prefer online to printed editions. Questions in Table 4 are analyzed condition on their responses to whether they prefer online journals to printed versions (Table 3). Equation (3) has been used to measure respondents' perception about the impact of e-journals on their research activities (questions in Table 5, Table 6, and Table 7).

II. 2. How to Read the Regression Tables

The signs of the four cut off points (a1, a2, a3, and a4) listed at the top of the column titled P[EJU] in Table 8 change from negative to positive, and their values increase. This first overall test tells us that e-journal usage frequency can be adequately "explained" by the set of independent variables we chose for the ordered probit model: demographic variables (age, occupation, fields of research, country of residence), Internet familiarity, number of subscriptions, number of papers submitted last year, and institutional access to scientific journals (see second column, Table 8).

Moving down the column to the detailed results of this regression analysis, we can see that certain groups of people use e-journals more frequently than others. Because 5 = never and 1 = yesterday or today, the EJU variable actually represents how little respondents use e-journals. Thus a positive coefficient for any of the categorical independent variables means that that category uses e-journals less often than the rest of the sample, while a negative coefficient means that that category uses e-journals more often. For continuous variables such as age or the number of subscriptions, if they are positive, then as they increase, e-journal usage frequency decreases; if they are negative, then as they increase, e-journal usage frequency increases.

If a variable is not marked with asterisks, then its relationship with e-journal usage frequency was not large enough to be statistically significant. Because all the variables are entered into the regression equation at once, each effect "controls for" all of the other variables. Consequently, we can say that a statistically significant variable has an "independent effect" on a person's e-journal usage frequency, regardless of that person's other characteristics.

II. 3. Tables

Table 1: Definitions of Selected Variables for Analyses
Table 2: E-Journal Usage Behavior
Table 3: Preference for Format (Printed Copy versus Online)
Table 4: Reasons for Format Preference
Table 5: Perception about the Future of Peer-Reviewed Journals
Table 6: Perception about E-Journals and Online Searching
Table 7: Perception about the Impacts of E-Journals on Research Activities
Table 8: Predicted Probabilities of E-Journal Usage, Preference for Format, and Perceptions about Whether E-Journals Affect Research Activities
Table 9: Logistic Regression Analyses: Probability of Various Motives for Favoring or Disfavoring Online Journal Editions over Printed Journal Editions
Table 10: Logistic Analyses of Probabilities of Perceptions about the Impact of E-Journals on Research Activities (among Those Who Answered "Yes" to Q15)
Table 11: Perception about E-Journals and Online Searching

Tables 1-8
Tables 9-11

Please share your reactions to the Survey Findings by filling out a simple form. Your feedback is a valuable component of the Electronic Journal User Study.

Last updated: 03-29-02

Home § About § Methodology § Workshops § Findings § Researchers § Privacy § Announcements § Presentations § FAQs

eJUSt home page Stanford University Libraries HighWire Press