A generalized model of factor analysis, called transform factor analysis, was proposed so as to provide useful guidelines for further developments in factor analytic methodology. Topics discussed were: a generalized model, existing models as its special cases, appropriate transformations, analysis of all forms of association measures and variance-covariance structures of original variables. A numerical example was given to illustrate applicability of transform factor analysis to a problem handled normally by other methods.

View full abstract

The present study was performed to investigate the effects of the inter-relatedness of informations and the rating order of the test trait upon the context effect in personality impression formation. Ss were presented 16 sets of 4 personalitytrait adjectives, one test trait and three context traits, and were then asked to rate favorableness for the test trait under a variety of inter-relatedness conditions: High-relatedness (HR), Mild-relatedness (MR), and Low-relatedness (LR). Test-trait ratings were made before the total ratings (T-O), or after the total ratings (O-T), On the basis of 2×3 factorial design, 294 female students were randomly assigned to six experimental groups. First, results indicate that the context effect occurred regardless of whether Ss were asked to rate the test trait before the total ratings, or after the total ratings. Second, results showed that the context effect increases with the inter-relatedness among the stimulus trait adjectives in a set, and that the differences of the context effect among three inter-relatedness levels increased as a function of discrepancy between the test trait likableness and the context likableness. Third, results indicate that the relation between the rating of the test trait and the total rating was not linear; that is, the total ratings were polarized as only function of the normative likableness of the context traits. Results of the present study, considered as a whole, are clearly consistent with the meaning-change hypothesis of the context effect.

View full abstract

Two experiments were conducted in order to search for the proper control procedure in the on-the-base-line type of CER experiment, using rats as subjects and licking behavior as the criterion response. In the first experiment, both the TRC and the traditional control procedure in which the US (shock) was explicitly unpaired were shown to be inadequate control procedures, for the reason that both procedures contributed to the different basal emotional level of Ss from that of experimental group. In addition, there was evidence of conditioning of “inhibition” in the traditional control group. In the second experiment, the adequacy of Hammond's ‘random control’ procedure was confirmed.

View full abstract

Two experiments were run for 25 randomly generated shapes varying in the number of turns. Similarity judgments, made by 36 undergraduate students, were scaled using non-metric multidimensional scaling technique. As the results, 3 dimensions were obtained, representing dimensions of compactness, complexity, and symmetry. As a second experiment, judgments were made on 30 semantic differential scales by 40 undergraduate students. As the results of analysis, 3 factors were extracted, representing dimensions of activity, potency, and evaluation. These solutions were compared using canonical correlation method, which yielded only one significant relation between them. The total variance shared by the two spaces was only 22.1%.

View full abstract

The purpose of this experiment was to investigate how classification preference depend on the hierarchical structures of stimulus sets and response alternatives. Stimulus sets were constructed using 4, 5, 6, and 8 stimuli drawn from a total set of eight equally spaced one-dimensional stimuli. The Ss classified these sets into groups in any way they wished (free classification) and into specified numbers of groups (restricted classification). The results indicated that: a) classification was identical for hue and value sets; b) numerical factors and metric stimulus properties acted in a hierarchic fashion in determining classification. Sequential partitions of classifications on the basis of numerical balance and nominal, ordinal, and interval metric properties lead to increased prediction of classification; c) the number of groups used in free classification, a function of the number and range of the stimuli in a set, was independent of the preferred classification for each number of groups. These results suggests that classification should be conceptualized as involving independent hierarchic structures.

View full abstract

Rats were trained 15 trials per day to avoid electric shock in a shuttle box to a 13/15 criterion either under D (20mg/kg chlordiazepoxide, p. o.) or under N (placebo). Reaching the criterion, some Ss, immediately, and others, after overtraining, were trained to the same criterion under the shifted drug condition. The drug state was further changed twice with no other precedural differences.
Results indicated a faster acquisition of the avoidance response and a marked increase in intertrial responding (ITR), which were explained by the drug's disinhibition effect rather than by its possible fear-reducing action. As in previous studies, drug-learning dissociation was more marked with the D-to-N shift than with the reverse shift, and this state-dependency was somewhat weakened for overlearned behavior and clearly reduced in the second and the third shifting. ITR was not state-dependent.

View full abstract