What is true about the relationship between semantic memory and aging?

Journal Article

Martin Lövdén,

Address correspondence to Martin Lövdén, Center for Lifespan Psychology, Max Planck Institute for Human Development, Lentzeallée 94, D-14195 Berlin, Germany. E-mail:

Search for other works by this author on:

Accepted:

16 December 2003

• PDF
• Split View
• • Article contents
  • Figures & tables
  • Video
  • Audio
  • Supplementary Data

• Cite

  Cite

  Martin Lövdén, Michael Rönnlund, Åke Wahlin, Lars Bäckman, Lars Nyberg, Lars-Göran Nilsson, The Extent of Stability and Change in Episodic and Semantic Memory in Old Age: Demographic Predictors of Level and Change, The Journals of Gerontology: Series B, Volume 59, Issue 3, May 2004, Pages P130–P134, https://doi.org/10.1093/geronb/59.3.P130

  Close

• • Email
  • Twitter
  • Facebook
  • More

Close

Navbar Search Filter Microsite Search Term Search

Abstract

CROSS-SECTIONAL data suggest that memory for specific events (i.e., episodic memory) deteriorates gradually from the age of 20 and onward, whereas small age-related declines are observed on memory for facts (i.e., semantic memory) at least until very old age (e.g., Lindenberger & Baltes, 1997; Nilsson et al., 1997; Park et al., 2002). A recent longitudinal investigation (Rönnlund, Nyberg, Bäckman, & Nilsson, 2003), conducted as part of the Betula Study (Nilsson et al., 1997), from which the present data were drawn, yielded a somewhat different picture than previous cross-sectional studies. Episodic memory performance was stable for participants up to the age of 60 years, whereas semantic memory improved somewhat across midlife (35–60 years). When the participants were past the age of 60, substantial and accelerating decline in both episodic and semantic memory were observed, although the rate of deterioration was less pronounced for semantic memory than for episodic memory. Rather than focusing on averaged patterns of stability and change, the general objective of the present study was to examine structural stability and stability (and change) of interindividual differences in declarative memory (i.e., episodic and semantic memory) within the age range displaying longitudinal mean-level decline in memory performance (age > 60). We addressed three current research themes in cognitive aging.

First, we examined the extent of interindividual differences in memory change. Reliable interindividual differences in change would spark interest in person-specific factors of cognitive aging, and such differences are prerequisites for between-subject approaches to the study of associations between constructs. Available studies report small but significant amounts of variance in change (e.g., Hultsch, Hertzog, Dixon, & Small, 1998; Wilson et al., 2002). For example, Hultsch and colleagues (1998) reported 6-year test–retest correlations estimated in latent space (i.e., stability coefficients) for a sample of older adults (55–86 years old) of.92 for vocabulary and.94 for fact recall (i.e., semantic memory). For word recall and text recall (i.e., episodic memory), the estimates were somewhat lower (.85 and.89). Variance in change is statistically independent of averaged change. However, the developmental gradients of declarative memory suggest that semantic memory is less affected by the negative factors behind cognitive aging. Assuming that partly the same factors are responsible for interindividual variability in change, and empirically based on the tendency from Hultsch and colleagues (1998), we expected higher stability of interindividual differences for semantic memory than for episodic memory.

Second, cross-sectional results from the Betula study (Nyberg et al., 2003) show that separate episodic and semantic memory factors are better than a single factor in accounting for the relations among declarative memory measures, both in young and old groups. Furthermore, the unstandardized loadings of the declarative memory measures do not differ in strength or constellation between younger and older age groups (i.e., metric invariance; Meredith, 1993). These findings demonstrate that the structural stability of declarative memory is high and confirm that the factors reflect the same attributes across age groups. However, the results also revealed a tendency for dedifferentiation (Reinert, 1970); that is, there are age-related increases in the correlation between the constructs. This study addresses whether these cross-sectional findings generalize to longitudinal data.

Third, we examined the association between semantic and episodic memory change. The degree to which changes in different cognitive functions are independent of each other is an important issue in cognitive aging research. Support for a high degree of dependency comes from cross-sectional studies showing that age differences in different cognitive abilities are well predicted by the shared variance among these abilities (e.g., Salthouse & Czaja, 2000; Verhaeghen & Salthouse, 1997; but see, e.g., Allen et al., 2001; Lindenberger & Pötter, 1998). Results from longitudinal studies show that correlations among changes in different cognitive abilities are high (e.g., Hultsch et al., 1998; Wilson et al., 2002). In this study, we examined the strength of the association between change in episodic and semantic memory.

Methods

For information concerning the design of the Betula study, detailed participant characteristics, and the full battery of measures included, see Nilsson and colleagues (1997). Only methodological features relevant to this study are described in the paragraphs that follow.

Participants

At baseline, or T1, the considered sample consisted of 500 participants, 100 in each cohort (60, 65, 70, 75, and 80 years old). Participants were randomly drawn from the population registry in Umeå, a city of about 100,000 inhabitants in northern Sweden. They were screened for dementia, severe sensory handicaps, and mental retardation. The general impression from attrition analyses is that the sample is highly representative of the target population (Nilsson et al., 1997).

Five years later, at Time 2, or T2, the sample was reduced to 361 participants. Excluded were 82 participants who did not return for retesting (55 were deceased, 19 refused to participate, 3 had moved, and 5 were ill or not available), 17 participants who were diagnosed as demented after T1, 16 participants with a T2 score below 23 on the Mini-Mental State Examination (Folstein, Folstein, & McHugh, 1975), and 24 participants with scattered missing values at T2 on the measures included in this study. Characteristics of the effective longitudinal sample are summarized in Table 1.

Because this study focuses on variability and associations, we explored selectivity of the effective sample for variances and correlations (for selectivity analyses on means, see Rönnlund et al., 2003). Separately for the total T1 sample (N = 500) and the effective sample (n = 361), Table 2 summarizes the standard deviations for, and correlations among, chronological age, gender, education, a unit-weighted composite of the indicators of episodic memory (see the paragraphs that follow), and a semantic memory composite at T1. The effective sample tends to be less variable than the total sample, and the correlations are somewhat attenuated in the effective sample. However, these effects are relatively small; for example, the important correlation between semantic and episodic memory was.58 in the effective sample and.62 in the total sample.

Procedure and Measures

At both occasions, the measures were obtained during two test sessions, each lasting for 1.5 to 2 hr for each participant. The first session consisted mainly of health examinations and questionnaires. The second session included an extensive battery of cognitive tasks (Nilsson et al., 1997). We selected five measures of episodic and semantic memory, respectively (cf. Rönnlund et al., 2003).

Episodic Memory

Recall of actions

Participants were presented with two consecutive lists imperatives (e.g., roll the ball) presented at a rate of 8 s/item. The nouns in the sentences were dividable into eight semantic categories. During study of one of the lists, participants were requested to perform the action described by the imperatives, using external objects corresponding to the nouns. The other list was studied without enactment. A free recall test followed after each list, and the measure was the number of sentences recalled in the enacted condition (RCAC in all further descriptions).

Cued recall of nouns

After the free recall test described herein, participants were presented with the eight categories into which the nouns could be divided and told that these categories serve as retrieval cues. The number of nouns recalled from the enacted (CRAC) and the nonenacted (CRNO) encoding conditions yielded two indicators of episodic memory.

Item recall

Participants were presented with 20 fictitious facts about famous and nonfamous people. During the test phase, each statement was presented in the form of a question to which the participant should respond. These questions were mixed with 10 questions that could be answered only on the basis of general knowledge and 10 fabricated question about nonfamous people. The number of correct answers corresponding to studied facts was used in the analyses (RCIT).

Recognition of nouns

Participants were presented with a list of 32 nouns; 16 of these were from the enacted or nonenacted encoding conditions already described, and 16 served as distractors. There were eight targets from each encoding condition and eight distractors that emanated from the semantic categories (see earlier text) in the nonenacted condition and eight distractors emanating from the categories in the enacted condition. Number of hits minus number of false alarms from the nonenacted condition was used (RNNO).

Semantic Memory

Vocabulary

This measure was derived from a 30-item multiple-choice synonym test. Participants selected a synonym to each target word from five alternatives. The number of correctly selected synonyms in 7 min yielded an indicator of semantic memory (VOC).

General knowledge

This measure emanated from the test labeled Item Recall. The 10 questions that were possible to answer only on the basis of general knowledge served as the measure (GENK).

Word fluency

Three measures of word fluency were used (FLUA, FLUM, and FLUB). The first of these measures was the number of words with the initial letter A generated in 1 min. The second measure was the number of professions with the initial letter B generated in 1 min. The third measure represented the number of five-letter words with the initial letter M produced in 1 min.

Results & Discussion

We analyzed the data with structural equation modeling. We used maximum likelihood estimation. The critical ratio for all significance tests was 1.96 (i.e., p <.05). All observed measures had acceptable skewness (–0.61–0.63) and kurtosis (–0.70–0.19).

First, we specified a standard longitudinal factor model (e.g., Hertzog & Schaie, 1986) including four factors, that is, episodic memory and semantic memory at T1 and T2, respectively. Chronological age, gender, and education were included in the model. All factors and demographic variables were allowed to covary, and all residuals had autocovariances over time. We expected correlated residuals caused by shared method variance between RCAC and CRAC and FLUA and FLUM (cf. Nyberg et al., 2003). This model had an acceptable fit, χ2 = 354.1, df = 198, comparative fit index (CFI) =.958, root mean square error of approximation (RMSEA) =.047. Against this model we tested a model forcing longitudinal invariance in the estimated factor loadings. This specification did not degrade the model significantly, χ2 = 365.3, df = 206, CFI =.957, RMSEA =.046; χ2diff = 11.3, df = 8, p >.18. Hence, we found longitudinal metric measurement invariance to be an acceptable hypothesis.

Next, we tested for longitudinal difference in the factor covariances between episodic and semantic memory. The model forcing these T1 and T2 covariances to be equal was rejected, χ2 = 376.7, df = 207, CFI =.954, RMSEA =.048; χ2diff = 11.3, df = 1, p <.001. The covariances increased over time. A model forcing equal variances in the memory factors over time also fit significantly worse than the accepted model, χ2 = 378.6, df = 208, CFI =.954, RMSEA =.048; χ2diff = 13.2, df = 2, p <.001. The variances increased over time, suggesting that the increase in covariance was in part a result of increasing variances. In the accepted model, the correlations between the declarative memory factors were.71 (confidence interval or CI = 0.63–0.79) at T1 and.76 (CI = 0.68–0.84) at T2. From the 95% confidence intervals, we conclude that the correlations differed from unity but not from each other. The stability coefficients were.95 (CI = 0.91–0.99) for semantic memory and.87 (CI = 0.81–0.93) for episodic memory. From the confidence intervals we conclude that the correlations differed from unity and from each other.

The next step was to reconfigure the accepted standard factor model into a Latent Difference Score Model (McArdle & Nesselroade, 1994). This model allows for direct estimation of variance in change and has the same fit as the previously accepted model. Age, gender, and education were implemented as predictors of level and change in the memory constructs. Figure 1 displays the model and the accompanying estimates. In the paragraphs that follow, these results are discussed together with the findings from the factor model. First, however, we note that the carefully screened sample of relatively healthy nondemented individuals ensured that interindividual differences in memory performance could be examined without confounding age-related memory change with major, but pathology related, sources of interindividual and intraindividual differences in change (e.g., dementia). In addition, the results in Table 2 suggest that attenuation of variances and associations caused by selective attrition is relatively modest. Thus, the data provided a solid ground for our three main empirical interests—we discuss these in turn.

Stability coefficients were high for both semantic and episodic memory. However, there was also evidence of small but significant interindividual differences in change; that is, stability coefficients were lower than unity and the variances in the change factors were significant (see Figure 1). Stability coefficients were higher for semantic than for episodic memory, suggesting that the factors behind interindividual differences in change may have differential impact or differ in constellation for the two declarative memory constructs. Presumably, interindividual differences in cognitive change might be higher in very old age when powerful sources of individual differences, such as preclinical dementia (e.g., Sliwinski, Lipton, Buschke, & Stewart, 1996) and terminal decline (see Small & Bäckman, 1999, for an overview), may increasingly dominate over normative aging-related changes even in carefully screened and relatively select samples. However, in normal aging, interindividual differences in declarative memory appear to be very stable across relatively short measurement intervals such as 5 years.

The results concerning structural stability demonstrated longitudinal configural and metric invariance, but also evidence for increasing variances and covariances, as well as a numerical tendency for dedifferentiation in terms of increasing correlations between the declarative memory constructs. These results are consistent with previous cross-sectional data (Nyberg et al., 2003). The configural invariance conveys that the division of declarative memory into episodic and semantic memory is statistically sound also over time in old age. The invariance in the factor loadings confirms the assumption that the factors reflect the same attributes across time; this finding supports the validity of conclusions based on mean-level changes (Rönnlund et al., 2003) and covariances across time (this study).

The association of semantic memory change with episodic memory change was strong (.68; see Figure 1). Because the relatively limited variances in change might have attenuated the strength of the relation, this estimate should be considered as conservative. Thus, the finding indicates strong couplings between changes in episodic and semantic memory in old age. From a purely statistical viewpoint, this strong association between changes in the memory constructs, together with the existence of interindividual differences in change, should inevitably lead to dedifferentiation (Hofer & Sliwinski, 2001). The nonsignificant nature of the pattern of dedifferentiation is therefore probably attributable to lack of power to detect increases in correlations over the relatively short measurement interval. However, at the heart of the substantive dedifferentiation hypothesis is the notion of an increased association between cognitive changes (e.g., Lövdén & Lindenberger, in press). Future studies should directly address the practicability of this dynamic dedifferentiation notion rather than focus on occasion-based relations between cognitive abilities. However, this demands careful methodological considerations to ensure that limited interindividual differences in change are not differentially attenuating associations across different periods of the life span.

Finally, some additional findings in Figure 1 should be highlighted. Age was a uniquely significant predictor of declarative memory change, indicating greater episodic and semantic memory decline with advancing age (see also Hultsch et al., 1998). More years in formal education were associated with better memory performance, and particularly with better semantic memory performance (see also Lindenberger & Baltes, 1997). Gender was related to level of episodic memory performance (in favor of women), but not to semantic performance (see also Herlitz, Nilsson, & Bäckman, 1997). However, none of these latter two demographic variables were associated with change (see also Hultsch et al., 1998).

To conclude, we find that this study demonstrated longitudinal configural and metric invariance of a declarative memory model, a strong association between changes in semantic and episodic memory, a tendency for dedifferentiation, evidence for small but reliable interindividual differences in change in declarative memory, and higher stability coefficients for semantic than for episodic memory. Together, these findings depict relatively high degrees of structural stability and stability of interindividual differences in declarative memory performance in a sample of older adults relatively free from pathology.

Decision Editor: Margie E. Lachman, PhD

Figure 1.

Latent Difference Score Model (McArdle & Nesselroade, 1994) for episodic and semantic memory, including age, gender (sex; women = 0 and men = 1), and years of formal education (edu) as predictors of level and change. Estimates are in standardized form; only significant ones (p <.05) are displayed. Variances (V) and their standard errors (in parentheses) are shown only for the change factors. The unstandardized loadings are set equal across measurement occasions, but the corresponding standardized estimates may vary across time. Estimates for the autocovariances are not shown. EPM = episodic memory; SEM = semantic memory; CH = change; RCAC = recall of actions; CRAC and CRNO = cued recall of nouns from enacted and nonenacted conditions, respectively; RCIT = item recall; RNNO = recognition of nouns; GENK = general knowledge; VOC = vocabulary; T1 and T2 = Times 1 (baseline) and 2 (5 years later), respectively. Word fluency: FLUA = initial letter A; FLUM = five-letter words, initial letter M; FLUB = professions, initial letter B

Table 1.

Participant Characteristics for the Effective Sample.

Notes: Standard deviations are given in parentheses. MMSE = Mini-Mental State Examination (Folstein et al., 1975); T1 = Time 1 (baseline); T2 = Time 2 (5 years later); M/F = male or female. Education shows years in formal education from T1. Health shows self-reports of the presence or absence of 20 psychological (e.g., anxiety) and physiological (e.g., back pain) conditions.

Table 1.

Participant Characteristics for the Effective Sample.

Notes: Standard deviations are given in parentheses. MMSE = Mini-Mental State Examination (Folstein et al., 1975); T1 = Time 1 (baseline); T2 = Time 2 (5 years later); M/F = male or female. Education shows years in formal education from T1. Health shows self-reports of the presence or absence of 20 psychological (e.g., anxiety) and physiological (e.g., back pain) conditions.

Table 2.

Standard Deviations and Correlations Among Age, Gender, Years in Education, Episodic Memory, and Semantic Memory in the Total T1 sample and in the Effective Longitudinal Sample at T1.

Notes: With each entry, values for the total T1 sample (first; N = 500) and the effective longitudinal sample (second; n = 361) are separated by slashes. T1 = Time 1 (baseline); T2 = Time 2 (5 years later). For gender, women were coded as 0 and men as 1. Education shows years in formal education.

Table 2.

Standard Deviations and Correlations Among Age, Gender, Years in Education, Episodic Memory, and Semantic Memory in the Total T1 sample and in the Effective Longitudinal Sample at T1.

Notes: With each entry, values for the total T1 sample (first; N = 500) and the effective longitudinal sample (second; n = 361) are separated by slashes. T1 = Time 1 (baseline); T2 = Time 2 (5 years later). For gender, women were coded as 0 and men as 1. Education shows years in formal education.

This research was supported by grants to Lars-Göran Nilsson from the Bank of Sweden Tercentenary Foundation, the Swedish Council for Planning and Coordination of Research, the Swedish Council for Research in the Humanities and Social Sciences, and the Swedish Council for Social Research. Lars Bäckman, Lars Nyberg, and Ake Wahlin were supported by grants from the Swedish Science Council. We thank Ulman Lindenberger and three anonymous reviewers for comments on previous drafts of this manuscript. Martin Lövdén is now at the Max Planck Institute for Human Development.

References

Allen, P. A., Hall, R. J., Druley, J. A., Smith, A. F., Sanders, R. E., Murphy, M. D.

2001

. How shared are age-related influences on cognitive and noncognitive variables?

Psychology and Aging

,

16

,

532

-549.

Folstein, M. G., Folstein, S. E., McHugh, P. R.

1975

. “Mini-mental state”: A practical method for grading the cognitive state of patients for the clinician.

Journal of Psychiatric Research

,

12

,

189

-198.

Herlitz, A., Nilsson, L.-G., Bäckman, L.

1997

. Gender difference in episodic memory.

Memory & Cognition

,

25

,

801

-811.

Hertzog, C., Schaie, K. W.

1986

. Stability and change in adult intelligence: 1. Analysis of longitudinal covariance structures.

Psychology and Aging

,

1

,

159

-171.

Hofer, S. M., Sliwinski, M. J.

2001

. Understanding ageing: An evaluation of research designs for assessing the interdependence of ageing-related changes.

Gerontology

,

47

,

341

-352.

Hultsch, D. F., Hertzog, C., Dixon, R. A., Small, B. J.

1998

. Memory change in the aged. New York: Cambridge University Press.

Lindenberger, U., Baltes, P.B.

1997

. Intellectual functioning in old and very old age: Cross-sectional results from the Berlin Aging Study.

Psychology and Aging

,

12

,

410

-432.

Lindenberger, U., Pötter, U.

1998

. The complex nature of unique and shared effects in hierarchical linear regression: Implications for developmental psychology.

Psychological Methods

,

3

,

218

-230.

Lövdén, M., Lindenberger, U., in press Development of intellectual abilities in old age: From age gradients to individuals. In O. Wilhelm and R. W. Engle (Eds.), Understanding and measuring intelligence. Thousand Oaks, CA: Sage.

McArdle, J. J., Nesselroade, J. R.

1994

. Using multivariate data to structure developmental change. In S. H. Cohen & H. W. Reese (Eds.), Life-span developmental psychology: Methodological contributions (pp. 223–267). Hillsdale, NJ: Erlbaum.

Meredith, W.

1993

. Measurement invariance, factor analysis and factorial invariance.

Psychometrica

,

58

,

525

-543.

Nilsson, L.-G., Bäckman, L., Erngrund, K., Nyberg, L., Adolfsson, R., Bucht, G., et al.

1997

. The Betula prospective cohort study: Memory, health, and aging.

Aging, Neuropsychology, and Cognition

,

4

,

1

-32.

Nyberg, L., Maitland, S. B., Rönnlund, M., Bäckman, L., Dixon, R. A., Wahlin, Å., et al.

2003

. Selective adult age differences in an age-invariant multi-factor model of declarative memory.

Psychology and Aging

,

18

,

149

-160.

Park, D. C., Lautenschlager, G., Hedden, T., Davidson, N. S., Smith, A. D., Smith, P. K.

2002

. Models of visuospatial and verbal memory across the adult life span.

Psychology and Aging

,

17

,

299

-320.

Reinert, G.

1970

. Comparative factor analytic studies of intelligence throughout the life span. In L. R. Goulet & P. B. Baltes (Eds.), Life-span developmental psychology: Research and theory (pp. 476–484). New York: Academic Press.

Rönnlund, M., Nyberg, L., Bäckman, L., Nilsson, L-G.

2003

. Stability, improvement, and decline in adult life-span development of declarative memory: Cross-sectional and longitudinal data from a population-based study. Manuscript submitted for publication.

Salthouse, T. A., Czaja, S. J.

2000

. Structural constraints on process explanations in cognitive aging.

Psychology and Aging

,

15

,

44

-55.

Sliwinski, M., Lipton, R. B., Buschke, H., Stewart, W.

1996

. The effects of preclinical dementia on estimates of normal cognitive functioning in aging.

Journal of Gerontology: Psychological Sciences

,

50B

,

P162

-P170.

Small, B. J., Bäckman, L.

1999

. Time to death and cognitive performance.

Current Directions in Psychological Science

,

8

,

168

-172.

Verhaeghen, P., Salthouse, T. A.

1997

. Meta-analyses of age-cognition relations in adulthood: Estimates of linear and non-linear age effects and structural models.

Psychological Bulletin

,

122

,

231

-249.

Wilson, R. S., Beckett, L. A., Barnes, L. L., Schneider, J. A., Bach, J., Evans, D. A., et al.

2002

. Individual differences in rates of change in cognitive abilities of older persons.

Psychology and Aging

,

17

,

179

-193.

The Gerontological Society of America

The Gerontological Society of America

What is true about the relationship between semantic memory and aging quizlet?

What is the relationship between aging and memory?

Which of the following statements about the relationship between semantic memory and aging?

Does semantic knowledge decrease with age?