RESEARCH ON REPEATED EXPOSURE EFFECTS OF THE STIMULI

Abstract

The logarithmic relationship between frequency of stimulus exposure and liking of those stimuli is referred to as the mere exposure phenomenon. Moreland & Zajonc asserted that there is an independent effect of exposure frequency on liking that is not operated by subjective recognition; in other words, there is a by-pass from stimulus frequency to liking, besides the pass through subjective recognition. Birnbaum & Mellers viewed familiarity and liking ratings as measures of subjective recognition that is a presupposed underlying process, and claimed that one factor intervenes the effect of exposure frequency on liking and stimulus recognition. According to Birnbaum's notion, the small discrepancy between theoretical and obtained correlation matrix among variables originated in one-mediator model may be caused by the fact that some of the dependent variables are non-linearly related to the single factor.After having carried out a series of linear structural relationships analyses by maximum likelihood methods, Moreland & Zajonc verified that the two-mediator model provided a significantly better degree of fit to their obtained data than the one-mediator model did.For the purpose of resolving these opposite arguments, two kinds of experiments were done. Subjects were confronted with prior exposures of 5 nonsense figures of 0, 1, 3, 9 and 27 frequencies in the first experiment, and then 5 nonsense words were given to them using the same frequencies in the second experiment. Consideration was paid to the fact that the subjects had never seen the same stimulus at five kinds of frequencies twice. Following these exposures, the subjects made stimulus recognition and liking ratings for the same stimuli as those that were presented the last time and for different stimuli from the ones that were showed for the first time.In the experiment of nonsense figures, the improvement in attractiveness associated with repeated stimulus exposure brought with it an increase of stimulus familiarity. Relating the rated attractiveness and recognition of the stimuli in the prior series to their frequencies, the correlation between the rating measures was somehow inflated, relative to Moreland & Zajonc data. On the other hand, the correlation between exposure frequency (log [raw frequency+1]) looked small compared with Moreland & Zajonc data.For nonsense words, the relationship between familiarity and frequency remained almost monotone, and there was a tendency toward reverse U-shaped relationship among affect and frequency.With respect to the correlation matrix of the obtained data, a computer program, LISREL IV was run. As a result, it looked as if the onemediator model worked better than the two-mediator model, although the differences weren't terribly large in the case of nonsense figures. Moreover, it was found that the promise of the two-mediator model could still hold in the case of nonsense words, together with the former case. It was concluded that properties of the stimuli which need either the one-mediator or two-mediator should be specialized.