In this section of the supplement, we provide further information concerning the secondary analysis of the University of South Florida Free Association Norms, described in the introduction to Experiment 2, beginning with material (in italics) extracted from the (unedited) paper.

The findings of Experiment 1 were consistent with the hypothesis that the AMI is mediated by horizontal (coordinate) associations rather than vertical (subordinate) ones. Given the important role of backward associations in producing the AMI (Roediger, Watson et al., 2001), however, it is possible that the effect revealed by this experiment is an artifact of differences in the backward association values of horizontal/coordinate and vertical/subordinate links. That is to say, the backward associative strengths between horizontal associates such as thread, pin, and eye, and the critical lure needle, may simply be higher than the corresponding strengths between vertical associates such as apple, orange, and kiwi, and the critical lure fruit. If so, then the allegedly qualitative difference between types of associative links, horizontal/coordinate and vertical/subordinate, dissolves into a merely quantitative difference in backward associative strength.

A secondary analysis of the University of South Florida (USF) Free Association Norms (Nelson, McEvoy, & Schrieber, 1998) made this possibility appear remote. Two of the authors classified all the responses to the 12 critical targets employed in this experiment as either horizontal/coordinate word associations or vertical/subordinate category instances (93.6% agreement, with disagreements resolved through discussion and arm-twisting). When we compared the two types of responses for each of the 12 lists, taken separately, we found that in only one case (the insect list) was there a significant difference between them (independent-groups t tests, p < .05). Even this difference was anomalous, as it was based on a comparison of only 3 horizontal/coordinate associates against 20 vertical/subordinate associates. When the 12 lists were combined, there was no overall difference in backward, mediated, or overlapping strength between associates and instances, all t < 1.

Unfortunately, the classic free-association  and category-instance norms (Battig & Montague, 1969; Russell & Jenkins, 1954) do not provide information about backwards associative strength.  Moreover, the information about forward associative strength is based on quite different instructions.  The USF norms used a single set of instructions, for free associations, and provide information about backward as well as forward associative strength, not to mention other comparative measures. Fortunately, the free-association instructions elicited a fair number of category instances -- although, admittedly, the resulting instances are probably unrepresentative of what is generated by category-instance instructions.  Still, we did the best we could with what was available.  Even if some of our category instances are a little unusual, they still permit a test of the hypothesis that the difference in AMI elicited by associates and by instances is an artifact of listwise differences in backwards associative strength.

The USF Norms

The USF norms list provide information on a number of different properties of free associations:

Animal List

For the animal list, we found 59 horizontal associates (e.g., zoo) and 97 vertical instances (e.g., aardvark).  The following table compares associates and instances on the major variables in the USF norms.

For the black list, we found 122 horizontal associates (e.g., white) and 7 vertical instances (e.g., coal, as in coal black).