研究人際知覺的新取向與工具：階層類別分析法

A New Approach to The Study of Person Perception: The Hierarchical Classes Analysis

本文的主要目的在於簡介Rosenberg和De Boeck（1988）所研發的階層類別分析法（HICLAS）及其電腦應用（HICLAS，2.01版，1992）。HICLAS的電腦程式主要是分析兩個向度、兩個變項的資料，即如個體（或物體〕和其特質的二分法之矩陣資料。HICLAS的結果圖為一多階層的結構，包括一組個體（或物體〕的多階層關係圖、一組特質的多階層關係圖，以及兩組間之關係。此一完整的結構特色，使其優於其它的分類法，如MDS等。為使此新工具暨其概念較易於被理解，本文提供一假設的矩陣資料為例，並進行其概念與結構之說明。再者，HICLAS應用在「被虐待兒童」之研究成果亦為此文之實徵證例。本文希望達到拋磚引玉的效果，期待HICLAS能在本土心理學的研究開發上有所貢獻。

The main purpose of this paper is to introduce the hierarchical classes analysis (HICLAS) developed by Rosenberg and DeBoeck in 1988 and revised in 1992. The HICLAS algorithm is designed to analyze a set of two-way, two-mode (i. e., targets by attributes) data. Its graphic re presentation is a se t-theoretical structure that includes a hierarchical structure of targets (or objects) and a hierarchical structure of attributes (or features) as well as the relationship between these two structures. HICLAS with this novel feature is superior to other similar techniques, such as MDS, hierarchic al clustering, and block modeling. A hypothetic al data from person perception is applied to illustrate the HICLAS representation and interpretations using both disjunctive and conjunctive ana lyses. Moreover, an empirical ex ample from the child abuse study is described. The characteristics of the abused mothers (those who were physically abused by their parents when they were young) were revealed by applying a free-response method and the HICLAS. The explanation of the cycle of abuse is proposed based on these results. Furthermore, applying HICLAS in the development of indigenous psychology as well as the study of ambivalence are addressed.