Journal of Experimental Psychology 1973, Vol. 101, No. 2, 301-306 ON THE REPRESENTATION OF MEANINGS OF GENERAL TERMS J RICHARD C. ANDERSON 2 AND BARRY McGAW s University of Illinois The 5s were exposed to sentences containing general terms, and then attempted to recall the last word of each sentence given single-word retrieval cues. Among 2 matched low associates of a general term, the cue referring to a case that resembled the most probable exemplar of the category named by the general term evoked the greater recall. The results seem to indicate that people use exemplars to represent the meanings of general terms, which was interpreted as support for an imagery theory of meaning. Several lines of evidence make an im- agery theory of meaning attractive. "Image-evoking value" is a potent deter- miner of the learnability of words (Paivio, 1968). Instructions to construct inter- active images dramatically facilitate paired- associate learning (e.g., Bower, 1971) and similarly facilitate the learning of sen- tences (Anderson, 1971; Anderson & Hidde, 1971). There are, however, seemingly in- surmountable problems with an imagery theory of meaning. It does not appear possible to represent in an image the meaning of an abstract term such as TRUTH or CREATIVITY. A similar problem arises with respect to general, concrete terms. By a "general" term we mean a word that refers to a class that contains more than one member, that is, a category label. Consider the term CONTAINER, for example. A person can form an image of an egg carton, a gallon milk jug, or a cardboard box, but an image of a partcular instance could not represent the properties common to all containers. There are several possible approaches to dealing with the problem of general terms within an imagery framework. The first is to concede, as Paivio (1969) has ap- parently done, that the meaning of general 1 The first author held a Fulbright-Hayes Fellow- ship at the Computer Based Learning Project, University of Leeds, Leeds, England, when this article was written. 2 Requests for reprints should be sent to Richard C. Anderson, Department of Educational Psychol- ogy, University of Illinois, Urbana, Illinois 61801. a Now at the Queensland Department of Educa- tion, Brisbane North Quay, Australia. terms cannot be represented in images. This is too great a concession. Only the trivial case of singular terms is left to be explained by imagery. Another possibility is to suppose that the meaning of a general term is represented in the image of an abstracted prototypical instance, so to speak, an image of the Platonic ideal. There are persuasive objections to this view. A third possibility is that general terms are encoded, at least in part, with images of particular instances. We feel that this latter possibility furnishes the only defensible ground upon which an imagery theory of meaning could be erected. The purpose of the present research was to investigate whether people in fact en- code general terms on the basis of the properties of particular instances. The strategy was to construct sentences con- taining the names of categories listed in the Battig and Montague (1969) norms. The most probable exemplars of the cate- gories should be the ones named by the most frequent associates in the norms. For example, DOG is the most frequent associate of ANIMAL, so it is probable that the meaning of ANIMAL would be repre- sented in part in the image of a dog. Also selected from the norms were 2 equi- probable low associates of each category name. One of these was judged to name an instance bearing a greater "resem- blance" than the other to the instance named by the most frequent associate. In the case of ANIMAL the 2 low associates 301