The Jackson Vocational Interest Survey (JVIS) is the product of years of careful research. Information about this research, as well as more details about the psychometric properties of the JVIS can be found in the JVIS Manual, available from SIGMA Assessment Systems. A brief version of this manual, the JVIS Quick Manual, is also available for download.

Scale Construction
Definition of the JVIS scales was based on a reconceptualization of occupational preferences in terms of work roles and work styles. Work roles refer to relatively homogeneous sets of activities relevant to occupations. A work role may strongly relate to certain occupations such as scales for Medical Service or Law, or cut across particular occupations and be relevant to a variety of careers, as in the case of Supervision or Human Relations Management. Work styles refer to a preference for certain kinds of work environments. For example, computer programmers and physical scientists are often required to work long hours to find solutions to difficult problems. Their scores on the Stamina scale are consistently high.

Another feature of the JVIS is that it places equal emphasis upon the measurement of interests of women and men. The JVIS was standardized in such a way that an equal number of males and females contributed to the selection of items and scales, and that items were required to show discrimination for each sex separately. The format allows males and females to be measured in terms of a common set of interest dimensions which do not make discriminations on the basis of traditional “male” and “female” occupations. Counselors who wish to give individuals an equal opportunity to consider occupations traditionally associated with only one sex can do so by employing the JVIS.

Perhaps the most unique feature of the JVIS is its method of scale construction. Each scale was designed to measure the interest designated by the scale name and to be relatively unrelated to other scales. To accomplish this result, careful attention was paid to the preparation of a large pool of items reflecting basic work role and work style dimensions. This was followed by administration of these items to large samples of males and females, well over a thousand of each. Final item selection involved a series of multivariate psychometrically-based procedures designed to select items most clearly related to the interests being assessed, to suppress response biases, and to minimize the redundancy between scales.

Norms
The most recent normative sample of JVIS profiles was collected in 1999. These consist of the responses of 1750 males and 1750 females from Canada and the U.S. This sample of 3500 individuals includes the responses of 2380 secondary school students (1190 males and 1190 females) and 1120 adults (560 males and 560 females). The adult sample consists of university and college students as well as adults seeking career interest assessment.

Reliability of JVIS Basic Interest Scales
The JVIS manual presents test-retest coefficients for two distinct samples. The first sample is a group of 172 university students who completed the JVIS one week apart, as part of introductory psychology research participation requirements. These test-retest reliability's range from .91 for Social Service to .72 for Independence, with a median of .84. The second sample is from a study by Berk (1988) assessing dimensions of person reliability in the context of vocational assessment. A group of 95 first year university students, 43 men and 52 women, completed the JVIS on two occasions separated by four to six weeks. Test-retest reliability's range from .92 for Social Service to .69 for Independence and Academic Achievement, with a median of .82. Internal consistency coefficients for JVIS Basic Interest scales are also presented in Table 4-2. Coefficients in the third column are based on a sample of 1573 high school students, 799 males and 774 females, who were administered the JVIS during school hours. These values range from .70 to .91, with a median of .81. Listed in the fourth column of Table 4-2 are reliability coefficients (coefficients alpha) for the normative sample of 1750 males and 1750 females. Coefficient alpha values range from .88 for Mathematics and Medical Services to .54 for Professional Advising with a median of .72. Lower values are indicative of scales assessing a number of facets.

Reliability of JVIS General Occupational Theme Scales
Internal consistency reliability's for the 10 General Occupational Themes based on the normative sample have a median value of .875. The test-retest reliability's from two respondent samples have a respective median values of .885 and .895.

Reliability of Individual JVIS Profiles
Reliability has traditionally been employed to describe tests or scales, not individuals. But in counseling decisions the question of the reliability of a single individual’s responses is just as much a matter of concern as is the question of the reliability of the test being completed. Accordingly, a method was devised (Jackson, 1976) for appraising individual reliability that has been incorporated into computer scoring. This score appears in the Administrative Indices section of each report as the Response Consistency Index. It is simply an odd-even reliability computed on a single individual across many scales, rather than the more usual odd-even reliability coefficient calculated on a single scale across many individuals. Thus, items in each scale were ranked in order and numbered sequentially. The total scores received by an individual for the odd-numbered items on each of the 34 scales were assigned to the x variable and the total score received for the even-numbered items to the y variable. A product moment correlation (R) is then calculated across the 34 scales and corrected by the Spearman-Brown formula. Berk and Fekken (1990) administered the JVIS to a sample of 95 university students on two occasions separated by four to six weeks. The mean reliability of individual profiles (corrected by the Spearman-Brown formula) was .84 on the first occasion and .87 on the retest occasion. The theoretical expected value of purely random responding, confirmed by Monte Carlo studies, is 0.00, with a standard deviation of about .18. Given the distributions of real and of random individual reliability coefficients, one can reasonably assume that individuals on whom a value of less than .20 is obtained can be categorized as probably primarily attributable to careless, non-purposeful, and/or unarticulated responding. The higher the individual R coefficient, the more confidence one can have in the reliability of the profile.