Journal oj Experimental Psychology 1974, Vol. 102, No. 6, 1116-1122 RETRIEVING ATTRIBUTE AND NAME INFORMATION FROM SEMANTIC MEMORY 1 ELIZABETH F. LOFTUS» AND WILLIAM COLE University of Washington In two experiments, 5s retrieved instances of well-learned semantic categories (e.g. ANIMALS). In Experiment I, two types of restrictors were used: Some- times the instance had to possess a certain attribute (e.g., it had to be small), and sometimes its name was restricted (e.g., it had to begin with the letter M). Prior knowledge about the type of restrictor differentially influenced the retrieval of these two types of items. In Experiment II, 5s produced category instances that simultaneously satisfied both an attribute and a name restric- tion (e.g., ANIMAL-SMALL-M). 5s invariably took longer to respond when the name restrictor was presented before the attribute restrictor rather than after- ward. A dictionary-network model, which embodies the notion that intrinsi- cally different processing strategies operate in the retrieval of attribute vs. name information, accounts nicely for these results. One technique for studying semantic memory is to ask 5s simple questions about material they already know and to measure how long it takes them to answer these questions. The reaction times (RTs) ob- tained are then used to make inferences about the processes used for retrieval. In several experiments that have used this technique, 5 was required to produce a member of a category that satisfied some restriction. For example, he might be asked for the name of & fruit that is yellow, or an animal that begins with the letter Z (e.g., Freedman & Loftus, 1971; Grober & Loftus, in press; Loftus & Suppes, 1972). In the Freedman and Loftus study, cate- gory-adjective and category-letter pairings (hereafter referred to as adjective and letter stimuli, respectively) were randomly "mixed" together in a long sequence of presentations. In the Grober and Loftus study, this mixed condition was included, but, in addition, 5s received two nonmixed conditions: one in which they saw only adjective stimuli, and the other in which they saw only letter stimuli. It was thus 1 This research was supported by a grant from the National Institute of Mental Health. The authors are grateful to Earl B. Hunt, Colin MacLeod, and Steve Poltrock for comments and to Elaine Botti for running the 5s in Experiment II. 2 Requests for reprints should be sent to Elizabeth F. Loftus, Department of Psychology, University of Washington, Seattle, Washington 9819S. possible to compare the speed of producing a fruit-yellow when it occurred in a mixed vs. a nonmixed condition. A major finding was that RT to an adjective stimulus was independent of whether 5 knew that he would be seeing an adjective; that is to say, RT to an adjective was equal in mixed and nonmixed conditions. However, 5's re- sponses to letter stimuli were dramatically affected by his expectations; he responded significantly faster in the mixed condition. Because mixing had a differential effect on adjective and letter stimuli, two retrieval processes were thought to occur; that is, the process of retrieving a category member when the restriction is on a quality or at- tribute of that member (e.g., it must be yellow) was thought to be different from the process of retrieving a member when the restriction is on the name of that mem- ber (e.g., the name must begin with the letter Z). An alternative interpretation was pro- posed to us by a grant review panel at the National Institute of Mental Health. The panel argued that the adjective-letter dif- ference might not be the result of intrinsi- cally different processing strategies, but rather of the fact that the letter and adjec- tive cues impose different degrees of con- straint or uncertainty. When 5 is in a nonmixed-letter condition, he first sees a category for some interval of time, and then a letter appears. He must respond as 1116