the target items in learning', both tests use category cues to elicit cued recall (CR). In controlled learning for the. CCR test, the category cues are used to search ...
Journal of the International Neuropsychological Society (1995), 1, 483-493. Copyright © 1995 INS. Published by Cambridge University Press. Printed in the USA.
Aging, encoding specificity, and memory change in the Double Memory Test
HERMAN BUSCHKE, MARTIN SLIWINSKI, GAIL KUSLANSKY, AND RICHARD B. LIPTON The Saul R. Korey Department of Neurology and Rose F. Kennedy Center for Mental Retardation and Human Development, Albert Einstein College of Medicine, Bronx, NY 10461 (RECEIVED December 14, 1994; ACCEPTED February 2, 1995)
Abstract Aged and young adults were tested by category cued recall after learning with category cues (CCR) or with item cues (ICR). CCR was about twice ICR for both aged and young adults. The aged recalled less than the young and did not benefit as much from greater encoding specificity and deeper processing in CCR. ICR and CCR were correlated, so that expected CCR can be predicted from ICR. The regression of CCR on ICR was linear for young adults, but was piecewise linear for the aged, showing that the relationship between ICR and CCR was not uniform for the aged adults. Lower than expected CCR by a subset of aged without clinical dementia may be a sign of preclinical dementia. (JINS, 1995, /, 483-493.) Keywords: Aging, Encoding-specificity, Preclinical dementia
Introduction Memory decline occurs both as a concomitant of normal aging and as a hallmark of early dementia (Friedland, 1993). As the population ages, the incidence of dementing illness has risen and is expected to continue to rise (Drachman, 1994; Ebly et al., 1994). The aged fear that their memory symptoms may be harbingers of impending dementia. New treatments for Alzheimer's disease are likely to be most beneficial early in the course of the disease. Therefore, describing the nature of the memory decline in normal aging and developing tests for early diagnosis of dementia have emerged as urgent public health priorities. Previous research suggests that at least some of the memory decline in normal aging and dementia may be accounted for by reduced ability to benefit from encoding specificity (Craik & Tulving, 1975; Craik & Simon, 1980; Craik et al., 1987; Grober & Buschke, 1987; Puglisi et al., 1988; Bird & Luszcz, 1991). Encoding specificity is the principle that "specific encoding operations performed on what is perceived determine what is stored and what is stored determines what retrieval cues are effective in
Reprint requests to: Herman Buschke, Department of Neurology, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, NY 10461.
483
providing access to what is stored" (Tulving & Thomson, 1973, p. 369). Encoding specificity means that encoding and retrieval must be coordinated for accurate assessment of memory and sensitive detection of memory impairment (Fischer & Craik, 1977; Schacter & Tulving, 1982; Buschke, 1984a). Although the principle of encoding specificity is well established, most memory tests are not designed to optimize encoding specificity. One memory test specifically intended to coordinate the conditions of encoding and retrieval is the free and cued selective reminding test (FCSRT; Buschke, 1984b), an extension of the selective reminding test (Buschke, 1973). The FCSRT coordinates encoding and retrieval by using the same cues to control learning as well as retrieval. In controlled learning for the FCSRT, subjects are given a series of category cues and are asked to search the set of to-be-remembered items for the item that belongs to each semantic category. For items that are not retrieved by free recall, the same category cues are presented to elicit cued recall. Cued recall in the FCSRT by normal aged subjects adds substantially to their free recall, and patients with dementia also benefit from such cued recall (Buschke, 1984b), although their total recall is less than that of normal aged subjects (Grober & Buschke, 1987). The FCSRT is useful for detecting memory impairment associated with dementia (Buschke, 1984b; Grober
484
et al., 1988; Tuokko & Crockett, 1989; Petersen et al., 1994). However, the FCSRT is limited by ceiling effects because cued recall with one item for each cue is so effective. Ceiling effects preclude accurate assessment of the full benefit of encoding specificity, limit comparisons of young adults, healthy aged, and aged with dementia, and limit detection of memory impairment. Furthermore, the FCSRT cannot assess the benefit of encoding specificity because the FCSRT cannot separate the effect of coordinating encoding and retrieval from the effect of cued recall alone. A control test without category cues in learning is needed to separate these effects and assess the benefit of encoding specificity (Bird & Luszcz, 1991). The Double Memory Test (DMT) is a revision of the FCSRT designed to prevent ceiling effects and include the necessary control test. The Double Memory Test The DMT consists of two conditions (tests) that are identical except for the level of processing (Craik & Lockhart, 1972; Craik & Tulving, 1975) and encoding specificity (Tulving & Thomson, 1973) induced during the learning phase. The retrieval phases are identical. Both conditions use cued recall of four items by each category cue to eliminate ceiling effects by "cue-overload" (Earhard, 1967; Tulving & Pearlstone, 1966; Watkins, 1979). Category cues are used for learning as well as for retrieval in the CCR experimental test but cues are used only for retrieval in the ICR control test. The names of these tests (Category and Item) refer to the kind of cue used to search for the target items in learning', both tests use category cues to elicit cued recall (CR). In controlled learning for the CCR test, the category cues are used to search for and identify each to-be-remembered item; in controlled learning for the ICR test the items themselves are used to search for and identify each item. Controlled learning, with the same cues that will be used to elicit CR, results in deeper processing and greater encoding specificity in the CCR test than in the ICR test. Controlled learning also assures attention and induces the same kind of effective processing by all subjects in each condition, so that performance on the CCR test provides an estimate of memory performance under optimal conditions for learning and recall (Buschke, 1984b, 1987). The difference between the number of words recalled on the CCR and ICR memory tests provides an estimate of the benefits obtained from greater encoding specificity and deeper processing in the CCR test. Because the ICR and CCR tests are identical except for the use of category cues in controlled learning, performance on these control and experimental tests should be correlated. Performance on the CCR test can therefore be predicted from performance on the ICR test, using the regression of CCR on ICR obtained from normal control subjects (Buschke et al., in press). Performance on the CCR test that is significantly less than expected, given
H. Buschke et al. ICR performance, is interpreted as evidence for a specific kind of memory impairment—a deficit in the ability to benefit from increased encoding specificity. To assess the relative benefit of increased encoding specificity for aged and young adults, the regressions of CCR scores on ICR scores of aged and young adults will be compared. Homogeneity of memory changes in aging Before the DMT can be used to detect memory impairment in individuals, it must be used to assess the group differences between aged and young adults. In any study of group differences between aged and young adults, it is essential to determine if observed age effects are homogeneous (i.e., constant throughout the distribution of all aged subjects). When age differences in memory are investigated, it is especially important to assess homogeneity because differential memory changes in a subset may be due to unrecognized preclinical dementia (Sliwinski et al., in prep., 1995a, 1995b). In the present study, the regression of CCR scores on ICR scores is used to assess the homogeneity of specific memory changes in aging. If memory changes in the aged are due to generalized ageassociated memory impairment in all members of the aged group, the regression should reveal homogeneity. If dementia or some other factor influenced the performance of a subset of the aged group, the regression should reveal heterogeneity. Age group effects can be homogeneous, or constant, in two distinct ways. If the effects of age group are additive, so that all individuals are affected to the same extent, then the CCR-ICR regression lines of the young and aged will differ by a constant value; that is, the intercepts will differ, but the slopes will be the same. If the effects of age group are multiplicative, so that low-performing individuals are affected more than high-performing individuals, then the regression lines of the young and aged will differ by a constant proportion; that is, the slopes will differ. Either of these outcomes indicates homogeneous aging effects. In either case, if a subset of aged individuals exhibits a disproportionately large age group effect, that is, if the differences between the young and aged are not constant throughout the aged sample, then the aging effects in that sample are heterogeneous. Identification of differential memory impairment in a subset of supposedly normal aged is important because (1) the subset may have memory impairment associated with unrecognized preclinical dementia, and (2) memory changes associated with aging alone must be distinguished from memory change due to other factors, such as preclinical dementia, to investigate memory changes associated with normal aging. This study compares performance on the DMT by samples of aged and young adults to answer the following questions: (1) Do young and aged adults benefit from encoding specificity as reflected by the contrast between ICR and CCR? (2) Are the benefits of increased encod-
Double Memory Test
485
ing specificity of similar magnitude for young and aged? (3) Are ICR and CCR correlated for young and aged adults? (4) Are age differences in the ICR and CCR contrast heterogeneous? (5) Does the nature of the heterogeneity suggest the possibility of an unrecognized subset of aged individuals with preclinical dementia? Method
Subjects All subjects gave informed consent approved by the Committee on Clinical Investigations of Albert Einstein College of Medicine (AECOM). The 219 aged (131 female and 88 male) subjects were selected from a sample of nondemented, healthy community-residing participants in a longitudinal study of memory and cognition in aging (Project 2 of the Einstein Aging Study). These subjects were aged 64-89 yr. The 120 young adult (72 female and 48 male) subjects were healthy college students paid $20.00 for participation and 6 secretarial and nonfaculty staff at AECOM. The young adults were aged 17-35 yr. Demographic data are shown in Table 1. Aged and young subjects were excluded if there was evidence of any of the following: disturbance of consciousness, medical or neurologic disease causing cognitive impairment, head injury with loss of consciousness for more than 1 hr, current psychiatric disorder, alcohol or drug dependence, endocrine or hematologic disease
or malignancy not in remission for more than 2 yr, or current use of psychotropic or antidepressant drugs. To screen for clinical dementia, aged individuals were excluded from this study if they made 9 or more errors on the Blessed Information-Memory-Concentration mental status test (BIMC; Blessed et al., 1968; Katzman et al., 1983) or if their recall on the FCSRT was below the cutscore for dementia (Grober et al., 1988). Table 1 compares the cognitive status of the young and aged subjects. The aged had less education than the young adults; 98.1
