372 CLARK AND PAIVIO Shoben, 1983), and as theoretical critics questioned the proposed nature of imagery (e.g., Pylyshyn, 1973). Research on item attributes continues to become more sophisticated in a variety of ways, including considera-tion of larger numbers of properties (e.g., Paivio et al., 1989; Rubin, 1980), and serious efforts to model and simulate the effects of stimulus attributes (e.g., Ellis {\&} Lambon Ralph, 2000). Computers have played a central role in both of these developments. The multivariate and simulation studies that might ultimately contribute to the emergence of theoretical models with strong empirical foundations has increased demand for large item pools and increased numbers of properties. The need for a greater number of properties and items follows from the essentially nonexperimental nature of item attribute research. Although researchers may exper-imentally assign different types of materials to different subjects, they generally do not and in some cases cannot experimentally manipulate the property or properties of interest (although such experimental approaches have been used in some cases). Rather, the properties are gen-erally measured in some fashion, and these measures will invariably correlate with other properties that might produce spurious effects or mask the effects of the target attributes. The primary way to address this problem, as in any nonexperimental research, is to identify diverse potentially contaminating constructs, obtain reliable and valid measures, and either control them in the selection of materials or include them in statistical analyses that can accommodate correlated factors (e.g., multiple re-gression, factor analysis, structural equation modeling). In addition to serving this control function, the col-lection of a large number of properties can itself provide information useful in the conceptualization of various item attributes. We can illustrate with one controversial question—the relative importance of frequency and age of acquisition in picture naming and other semantic re-trieval tasks (Morrison {\&} Ellis, 2000). Paivio et al. (1989) obtained picture naming and imagery latencies for a moderate-sized pool of pictures and their most common labels. Information on a wide range of properties, in-cluding age of acquisition, was also obtained. Factor analysis of the results indicated that age of acquisition loaded on several different factors (e.g., familiarity, con-creteness, name length), all of which contributed to pic-ture naming latencies. One interpretation of this result is that age of acquisition is a multidimensional measure that taps a number of distinct properties of words and pictures, hence its superiority to single-component pre-dictors in multiple regression analyses. More specula-tively, one might hypothesize that people rating age of acquisition are actually making judgments of how con-crete, short, and familiar items are, and that children in fact first learn words that tend to be concrete, short, and familiar. Generalization across items provides yet another rea-son to continue the development of item norms for use in cognitive research and theorizing. No single set of norms will ever suffice, because results can depend on the par-ticular pool of items that have been included in the norms. Despite the large corpus of materials on which the Ku{\v{c}}era and Francis (1967) frequency norms are based, for exam-ple, abstract words are still probably overrepresented just because of the types of text that dominate the corpus (e.g., literary and academic materials). It is therefore im-portant to continue to develop additional norms to per-mit evaluation of the generality of findings across di-verse word pools. Another facet of the generalization issue is the possi-bility of generational or cohort differences across ex-tended periods of time. With respect to word familiarity, for example, exposure to and knowledge of particular words might differ today from ratings, like those in the PYM norms, collected during the 1960s. New norms and replication of existing properties allow researchers to de-termine the continuing validity of norms collected years and in some cases decades ago. This article reports two extensions of the PYM norms. Part 1 reports a marked expansion of the number of prop-erties available for the original 925 PYM items, and Part 2 reports an expansion of the number of items for which basic properties are available. We also provide re-sults of factor analyses for both extensions, with the analysis in Part 1 being particularly informative about interrelationships among a diverse collection of word properties.
basic ratings of imagery, familiarity, and a new age of acquisition measure