Vernellia R. Randall , The Myers-Briggs Type Indicator, First Year Law Students and Performance , 26 Cumb. L. Rev. 63 - 101 (1995) 

At orientation, all members of the entering first year class of law students were asked to participate. They were asked to complete the MBTI Form G, which has ninety-two self-scoreable items. Each item offers a forced choice between two opposing answers that equate to one of the four opposing preferences.⁽⁷¹⁾ The choices are between every day events and word pairs selected to evoke a choice between the competing preferences.⁽⁷²⁾ A person's type is based on which pole of the four preferences the person prefers, with the four preferences combining to render sixteen possible types.⁽⁷³⁾ 

The items are weighted 0-2, and scores are given for each pole based on the points totaled from the responses. The result allows a person to determine not only which pole is preferred, but by how much.⁽⁷⁴⁾ The level of a person's preference can be slight (1-9), moderate (11-19), clear (21-39), or very clear (41 or higher).⁽⁷⁵⁾ For example, a person might score 43 on extravert items and 20 on introvert items. His score would be E 23. The 

E indicates that he has a preference for extraversion, and the 23 indicates that his preference is "clear."⁽⁷⁶⁾ 

Data Interpretation 

Throughout this paper, I compare different groups, such as: males v. females, white students v. students of color, and extraverts v. introverts. When comparing groups, the question arises whether the groups really represent populations that are different from each other. It is possible, some may say even probable, that different groups given the same treatment (i.e., extraverted law students versus introverted law students) could make different grades merely by chance. That is, any observed difference could result merely from sampling error. Thus, as a researcher, I wanted to test the null hypothesis that the groups being compared are really only two samples from the same population and any observed difference is due to chance or sampling error. In short, the null hypothesis establishes that the real difference between groups being compared is zero.

o the question becomes: How large does an observed difference have to be before a researcher is justified in rejecting the null hypothesis?⁽⁷⁷⁾ I used Tests of Statistical Significance to answer those questions. Significance is designated with the symbol " p". Most social scientists treat results with a statistical significance of .05 or less as "significant," or meaningful, and treat a statistical significance of .01 as very significant.⁽⁷⁸⁾ A statistical significance of .05 means that only five times out of a hundred will the observed result come from chance or some random process. A statistical significance of .01 means that only one time out of a hundred will the observed result come from chance. Consequently, lower significance levels indicate a higher probability of real or reliable results. 

While conventional research reports significance at three levels (<.05 or <. 01 or <.001), I reported the actual probability. I did so, in part because I believe conventional significance levels may be too conservative in interpreting the practical significance of differences in grade point averages. While results greater than .05 are not as statistically reliable as results meeting the .05 test, they do indicate possible non-random differences in the population. Such differences would be extremely important in a population where even very small differences in grades can result in substantial differences in treatment in the job market, in selection to law review, and most importantly, in being placed on probation or being dismissed. ⁽⁷⁹⁾ 

Description of Students 

Initially, all 170 students in the entering class completed the Myers-Briggs Type Indicator. However, the study was limited to the 154 students who had first semester grade point averages (FSGPA).⁽⁸⁰⁾ The students were overwhelmingly white,⁽⁸¹⁾ male⁽⁸²⁾ 

71. FN71. Myers & McCaulley, supra note 57, at 3. 

72. FN72. Id. at 3. 

73. FN73. Id. at 2-3. 

74. FN74. Id. at 58. However, this does not necessarily show how strongly the preference is felt, how well developed the skills associated with the preference are, or the problems a person may have using that preference in inappropriate contexts. Id. at 52-61. 

75. FN75. Id. at 58. 

76. FN76. Id. 

77. FN77. John L. Phillips, Jr., How to Think About Statistics 109-10 (1982). 

78. FN78. Id. 

79. FN79. For example, a student with a 1.79 will be dismissed from the University of Dayton law school at the end of the first year, while a student with a 1.80 (a difference of .01) will be placed on probation. 

80. FN80. Interesting questions arise as to why sixteen students withdrew before finals, some even before the completion of orientation, and whether learning styles would have any predictive value. 

81. FN81. Of the 154 students in the first year class, 88.3% were white, 7.8% were African-American, 1.9% were Asian-American/Pacific Islander, 1.3% were Hispanic-American and .6% were international students. Appendix, Table A1. A slightly larger percentage of the students in the study was students of color than was represented by the students who were not in the program. However, the difference was not statistically significant (p=.908). 

82. FN82. Of the 154 students, 58.4% (90) were male and 41.6% (64) were female. Appendix, Table A2. While the study had a higher percentage of female students than was represented by the students not in the program, the difference was not significant (p.424) 

83. FN83. The mean age of the students was 26.4 years; the youngest student was 22; the oldest student was 59. Appendix, Table A3. The students in the study were younger than students who were not in the study. The difference was statistically significant (p.033). 

84. FN84. The mean undergraduate grade point average (UGPA) was 3.069, while the minimum was 1.600; the maximum GPA was 3.890. Appendix, Table A4. While the students in the study had a slightly higher UGPA than the students who did not participate (2.89), the difference was not statistically significant (p=.1913).