Copyright 2004 Psychonomic Society, Inc. 402 The purpose of this article is to make the largest free as- sociation database collected in the United States available to interested researchers and scholars. More than 6,000 participants produced nearly three quarters of a million re- sponses to 5,019 stimulus words. What Do Free Association Norms Measure? With high degrees of average reliability (r = .89), free as- sociation response probabilities index the likelihood that one word can cue another word to come to mind with minimal contextual constraints in effect (Nelson, McEvoy, & Dennis, 2000). Free association probabilities provide a relative, rather than an absolute, index of what is generally called for- ward strength, because they are sensitive to other associates activated by the cue word that compete for production (Gillund & Shiffrin, 1984; Nelson, Dyrdal, & Goodmon, in press). In short, given a cue, free association probabilities index the relative accessibility of related words in memory. We assume that free association taps into lexical knowl- edge acquired through world experience. Such experience creates associative structures involving the representations of words and the links that bind them together. These struc- tures conform to the general constraints underlying small- world networks (Steyvers & Tenenbaum, in press). Like the semantic networks of WordNet, associative networks are sparse, they exhibit strong local clustering, and they have short average path lengths between words. An aver- age of only three associative steps is required to get from any one word in our norms to any other. Associative structures capture the shared lexical expe- riences of many people. They are dynamic because they are derived from the stream of everyday experience that changes gradually over time but that can produce precipi- tous temporary effects; for example, because of the movie, the primary free association response in our norms to in- stinct is basic. Norms must be used in conjunction with knowledge about current trends and local culture. They are also dynamic because they are sensitive to experience that deviates from the norm. For example, for words related to drug and alcohol use, the associative structures of sub- stance abusers are different from the norm (Reich & Gold- man, in press; Stacy, 1997). Although the language of association theory is used to describe the procedures and findings of experiments that rely on free association probabilities, this task appears to capture both associative knowledge and aspects of mean- ing. They index the knowledge that lemons are sour, that birds fly, and so on. Many researchers have attempted to draw a distinction between the effects of association and meaning by creating materials that are semantically but, ostensibly, not associatively related (e.g., for reviews, see Hutchison, 2003; Lucas, 2000). However, given the small- This research was supported by Grant MH16360 from the National Institute of Mental Health to D.L.N. and by Grants MH45207 and AG13973 to C.L.M. In order of appearance on the project, our special thanks go to David Brooks, Joesph Wheeler, Jr., Richard Borden, Maria- Teressa Bajo, Pepe Canas, Charlotte Hall, Patricia Holley, Leilani Good- mon, and Ami Willbanks for helping us. Correspondence concerning this article should be addressed to D. L. Nelson, Department of Psy- chology, University of South Florida, PCD Building, Room 4120, Tampa, FL 33620-8200 (e-mail: dnelson2@chuma1.cas.usf.edu). The University of South Florida free association, rhyme, and word fragment norms DOUGLAS L. NELSON and CATHY L. MCEVOY University of South Florida, Tampa, Florida and THOMAS A. SCHREIBER University of Kansas, Lawrence, Kansas Preexisting word knowledge is accessed in many cognitive tasks, and this article offers a means for indexing this knowledge so that it can be manipulated or controlled. We offer free association data for 72,000 word pairs, along with over a million entries of related data, such as forward and backward strength, number of competing associates, and printed frequency. A separate file contains the 5,019 normed words, their statistics, and thousands of independently normed rhyme, stem, and fragment cues. Other files provide n  n associative networks for more than 4,000 words and a list of idiosyn- cratic responses for each normed word. The database will be useful for investigators interested in cuing, priming, recognition, network theory, linguistics, and implicit testing applications. They also will be useful for evaluating the predictive value of free association probabilities as compared with other measures, such as similarity ratings and co-occurrence norms. Of several procedures for measuring preexisting strength between two words, the best remains to be determined. The norms may be down- loaded from www.psychonomic.org/archive/. Behavior Research Methods, Instruments, & Computers 2004, 36 (3), 402–407