Judgment Analysis and Learning Disabilities: An Empirical Procedure for Construct Explication

Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Teaching & Learning


Purpose: The purpose of this study is the exp1ic ion of the learning disability (LD; construct. Construct explication is a process whereby an abstract variable is made explicit in terms of concrete variables. The purpose of this construct explication is to provide a description as opposed to a cause of the phenomena of learning disabilities.

Research: This study was designed to answer the following research questions: 1. Does a relationship exist between the selected profile variables and each criterion? Which variables are most useful for prediction by this group of LD specialists? 2. Does a relationship exist between the three criteria? Furthermore, does c relationship exist between the three criteria, collectively, and the profile variables? 3. Among the group of LD specialists does more than one policy about learning disabilities exist? If so, do any of these policies differ significantly from the standard conceptual definition of learning disabilities?

Method: Thirty-four learning disabilities specialists from North Dakota rated twenty-five profiles, each containing twer /-four variables. For each profile the LD specialists were to make three decisions: 1) level of LD; 2) general type of remedial program; and 3) setting for remediation.

Statistical Procedure: The principal techniques used were based on multiple regression analy sis. Multiple correlation, stepwise backwards regression, and setwise backwards regression were used to ascertain which variables were most use- rul for prediction of each criterion. Canonical correlation was used to determine the relationship between the multiple predictor set and the multiple criteria.

The analysis centered on the clustering of the LD specialists. Judgment analysis was used to form groups of raters based on the homogeneity of their prediction equations.

Results: Concerning research question one, no univariate correlations ' re significant between the 24 profile variables and each criterion (mean level c! LD, modal type of remedial program, and modal setting for remedia tion). As a set of predictors the 24 variables were highly correlated to each criteria (R = .97, .98. .92). Using setwise and stepwise h:„k- wards procedures, the number of profile variables was reduced to sever which accounted for sixty percent of the variance in the mean level of LD criterion. Seven variables were also selected for the other two criteria.

For research question two, the three criteria were significantly intercorrelated. Also a canonical correlation was found between the seven predictors of mean level of LD and the three criteria (Rc = .83).

The central question of this study dealt with the clustering of the LD specialists. The judges (the LD specialists) that formed a group had similar regression equations. In judgment analysis, these regression equations are called judgmental policies. They also express an empirical definition of the construct under study.

The LD specialists formed two cluster. One group, the largest, empirically defined learning disabilities as low achievement in language and mathematics. The other group was more a splinter rather than a distinct cluster. They defined learning disabilities as low achievement in language and spelling.

Conclusions: The conclusions of this study can be summarized as follows: The learning disabilities construct, when empirically defined, is a much more limited construct than when conceptually defined. Although the LD specialists can be hignly predictive using the profile variables, they do so by using only a few variables consistently. The instruct of learning disabilities offers little now descriptive information to school learning difficulties. It was suggested, by the investigator, that further use of the construct was not warranted.

This document is currently not available here.