Chapman L J, Chapman J P
J Abnorm Psychol. 1989 Nov;98(4):357-66. doi: 10.1037//0021-843x.98.4.357.
The resolution of the heterogeneity of schizophrenics and their relatives using cognitive tasks requires measures of individual differences--usually of differential ability, which is inferred from differential performance. Tasks that are suited for establishing a group's differential ability are often unsuited for measuring individual differences. The problem is that for any group for which Tasks A and B have different mean accuracy, the (A - B) difference scores of individual subjects have an artifactual curvilinear relation to (A + B) overall accuracy, with the largest (A - B) scores occurring at about 50% (A + B) overall accuracy for a dichotomously scored free-response task. Two possible solutions are (a) to convert Task B scores to residualized scores, using the regression of Task B scores on Task A scores as determined for normal subjects; or (b) to titrate overall (A + B) accuracy of each subject to a constant level by manipulating a variable.
使用认知任务来解决精神分裂症患者及其亲属的异质性问题，需要测量个体差异——通常是差异能力，这是从不同表现中推断出来的。适合确定一组人差异能力的任务往往不适合测量个体差异。问题在于，对于任务A和任务B平均准确率不同的任何一组人来说，个体受试者的（A - B）差异分数与（A + B）总体准确率存在人为的曲线关系，对于二分计分的自由反应任务，最大的（A - B）分数出现在总体准确率约为50%（A + B）时。两种可能的解决方案是：（a）使用正常受试者确定的任务B分数对任务A分数的回归，将任务B分数转换为残差分数；或者（b）通过操纵一个变量，将每个受试者的总体（A + B）准确率调整到一个恒定水平。
