“Liking Words As a Function of the Experienced Frequency of Their Occurrence”, 1980-02 (; similar):
A hypothetical inverted-U curve is postulated linking liking of stimuli to familiarity with them.
An experiment using a special procedure was carried out in which the relationship was investigated for words, ranging from very unfamiliar to very familiar, between favorability and familiarity.
The results conformed to the theoretical curve.
This indicated that the positive correlation between the variables reported by several researchers (eg. Zajonc) and the negative correlation found by others (eg. Cantor) should be regarded as complementary rather than contradictory.
…When the stimulus words were roughly split into 2 groups, the relatively unfamiliar and the relatively familiar, liking was found to be positively related to familiarity in the former case (as in Zajonc-type studies) and negatively related to familiarity in the latter case (as in Cantor-type studies). The function that properly fitted the familiarity-favorability relationship over the full range of the familiarity variable was found to be curvilinear, first rising and then falling. Thus the result contained both the Zajonc-type and the Cantor-type effects, showing them to be complementary rather than contradictory. We undoubtedly achieved this by using a very wide spread of the independent variable; and this was made possible by the particular experimental procedure adopted.