Successful Reference Transactions:

A Multivariate Analysis Based on a Review of the Reference Literature

Dr. John V. Richardson Jr., UCLA Professor of Information Studies

Since the late 1960s, researchers studying the library reference transaction have suggested a host of possible variables which may influence the accuracy rates as part of a successful outcome. Yet, the nature and extent of these variables was still largely unknown until recently. In the late 1990s, Matthew Saxton posited a multivariate analysis of the factors thought to affect the reference transaction; as such, this study would be a significant contribution to the improvement of professional practice, especially in light of the so-called half-right (50% accuracy rates found) reference service supposedly being provided.

Hence, the purpose of this paper is to review (as well as to identify, describe and analyze) the reference transaction literature for evidence which might account for the situation mentioned above. Strictly speaking, then, this review does not address user satisfaction or the utility of the results given to a library user. Rather, the ultimate goal is to predict the successful reference transaction by constructing a testable factor analytic model that takes auto-correlations into consideration.

Two research questions are driving this report: what concepts are employed and how are they operationalized as variables? But, most importantly, what might a statistical model for a hierarchical linear analysis look like?

Potential Variables

First, this section identifies numerous concepts and prior attempts to operationalize these concepts as variables. The following variables are listed in no particular order; all are potential candidates for causation.

1. Performance (i.e., accuracy, efficiency, and time taken)

In each of the following studies, answers to questions were simply categorized as accurate, partially accurate, or inaccurate. Based on a test set of ten test questions, Goldhor (1967) first reported that results takes about twenty minutes, but accuracy (see extensive discussion of this concept in chapter two and three of Saxton and Richardson, 2002, cited below) was only 55%. At about the same time, Bunge (1967) argued that efficiency could be defined as “the rate of accuracy over the time needed to answer a query”; in other words, measured by dividing the number of correctly answered questions by time in minutes taken to answer the set of questions. In 1971, a major advance in question scoring was posited by Childers that questions are answered: 1) wholly correct; 2) wholly correct but with some uncertainty; 3) partially correct; 4) wholly incorrect or no answer; and 5) no attempt to answer at all.

2. Staff ability (i.e., adequate staffing, trained or untrained, with education + experience)

Bunge further hypothesized that “professionally trained reference librarians would outperform untrained staff members in terms of their ability to answer reference questions.” However, he expressed the relationship in negative terms “lack of education could be offset by experience.” In fact, Bunge (1969) found that “although trained reference librarians were not necessarily more accurate than untrained staff in answering questions, they were faster and more efficient than other staff.”

3. Library collection (including physical facilities, supplies and equipment)

Crowley (1971) hypothesized that “those libraries which were able to spend more could build larger collections (resources) and hire and train more competent staff (responders).” Intuitively, larger collections should contain more resources to answer more questions and experienced staff ought to cost more so that libraries with deeper pockets ought to secure such staff more readily. Powell (1975) determined “that reference collection size had a curvilinear relationship to accuracy in that the size of the effect diminished as collection size exceeded 3500 volumes.”

4. Budget (i.e., expenditures)

Similarly, Crowley (1971) found that budgets, defined as “expenditures per capita” influenced the outcome of reference transactions. For instance, “high” budgets allowed libraries to answer more questions correctly, although his findings were not statistically significant.

Findings

Based on the above cited studies, one can summarize the research as follows; attention is focused on simple correlations which should not be used in discussions of causation. Experience influences successful reference outcomes, whether the staff member is trained (i.e., ALA-accredited MLIS holder or those who have taken reference or bibliography courses, in particular) versus untrained (i.e., the library clerk or paraprofessional). The former are more efficient (i.e., accuracy/time) and specifically in their case, experience correlated with performance. Yet, most studies have neglected the role of “age, interests, imagination, and tenacity” of the librarian.

Summary

In short, based on the preceding findings the following statistical model might be tested in the future:

DV        IV₁ + IV₂ + IV₃ + IV₄ + IV₅ + IV₆ + IV₇ + IV₈ where

the dependent variable (DV) is the presence or absence of an accurate reference transaction and the independent variables (IV_n) are performance (i.e., accuracy, efficiency, and time taken), staff ability (i.e., adequate staffing, trained or untrained, with education + experience), library collection (including physical facilities, supplies and equipment), and budget (i.e., expenditures) as operationalized in the reference literature as discussed above.

ACKNOWLEDGMENT AND REFERENCES

This report is based on Matthew L. Saxton and John V. Richardson Jr.’s Understanding Reference Transactions: Transforming an Art into A Science (New York: Academic Press, 2002). All citations can be found and are classified by format (i.e., journal articles, books and reports, and dissertations and theses) and further subarranged alphabetically by first author’s last name starting on page 191ff.