Differential item functioning on the Myers-Briggs type indicator

Abstract

Differential item functioning on the Myers-Briggs Type Indicator (MBTI) was examined in regard to gender. The Myers-Briggs has a differential scoring system for males and females on its thinking/feeling subscale. This scoring system preserves the 60 % thinking male and 30 % thinking female proportion that is implied by the Jungian theory underlying the Indicator. The MBTI's authors contended that the sex-based differential scoring system corrects items that subjects at a certain level of a latent trait either incorrectly endorse or leave blank.

This reasoning is the classical definition of differential item functioning (DIF); consequently, the non differentially scored items should exhibit DIF. If these items do not show DIF, then there would be no reason to use a differential scoring system.

Although the Indicator has been in use for several decades, no rigorous item response theory (IRT) item-level analysis of the Indicator has been undertaken. IRT analysis allows for mean differences in subgroups to occur, independent of the question of DlF. Linn and Harnisch's (1981) pseudo-lRT analysis was chosen to test for the presence of DlF in the MBTl items because it is best for tests of relatively small length. The Myers-Briggs subscales range from 22 to 26 items, which is relatively small by lRT standards. lRT analyses conducted on N=1887 subjects indicated that no items on the thinking/feeling subscale showed evidence of DIF. Out of 94 items, only one extraversion/introversion item and one judging/perception item showed evidence of DIF; no Thinking/Feeling items showed DIF. It is recommended that sex-based differential MBTI scoring be abandoned, and that the distribution of type in the population be examined in future studies.