| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
RESEARCH ARTICLE |
a Baycrest Centre for Geriatric Care, Toronto, Canada
b York University, Toronto, Canada
Angela K. Troyer, Department of Psychology, Baycrest Centre for Geriatric Care, 3560 Bathurst Street, Toronto, Ontario, Canada M6A 2E1 E-mail: atroyer{at}baycrest.org.
Decision Editor: Toni C. Antonucci, PhD
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionAppendix ENDIXReferences |
|---|
IN a national survey of nearly 15,000 adults aged 55 and older, 15% reported having had trouble remembering things frequently during the past year (Cutler and Grams 1988
). Examination of a wide range of variables from that survey indicated that potential correlates, such as sex, age, educational attainment, and health factors, accounted for very little variance in the frequency of self-reported memory problems. As this large-scale survey suggests, self-report measures of memory function provide important information about aspects of subjective memory abilities in aging. Self-appraisals of memory add a unique dimension to the assessment of memory performance that is not captured by objective memory testing alone. Self-report tools allow the measurement of everyday memory problems rather than more artificial, laboratory-based memory problems. Consequently, such tools allow estimation of the clinical significance of a documented memory problem and can guide the development of intervention programs aimed at improving everyday memory function.
In addition, self-reports provide information about meta-memory, or insight and awareness of an individual's own memory functioning. Such insight is known to differ among memory-disordered patients of differing etiologies (Schacter 1991). Moreover, even among healthy individuals, self-report measures often reveal inaccurate beliefs about memory and aging influenced by culturally based negative stereotypes (Cutler and Grams 1988; Hertzog and Hultsch 2000). This aspect of metamemory may be associated with a poor appraisal of one's own memory efficacy, which in turn may lead to a self-propagating cycle of lowered expectations, reduced effort on everyday tasks involving memory, and, ultimately, impaired performance (i.e., a self-fulfilling prophecy of age-related memory reductions). Given the potential significance of metamemory, there is a need for well-developed self-report memory measures that are reliable, valid, and applicable to clinical and research settings.
Several self-report memory questionnaires have been developed for use with research participants and brain-injured clinical patients. The Metamemory in Adulthood Questionnaire (MIA; Dixon, Hultsch, and Hertzog 1988) contains 108 items that describe memory issues and changes relevant to healthy aging. The seven subscales provide information about an individual's knowledge of general memory processes and tasks, frequency of memory strategy use, self-rated memory ability, perceptions of memory stability over time, anxiety regarding memory, memory and achievement motivation, and locus of control in memory abilities. Good psychometric data have been presented for the factor structure, reliability, and validity of this test (Dixon et al. 1988; Dixon and Hultsch 1983a, Dixon and Hultsch 1983b). Another metamemory scale, the Memory Functioning Questionnaire (MFQ; Gilewski, Zelinski, and Schaie 1990), contains 64 items distributed across four subscales that reflect subjective appraisals of frequency of forgetting in different situations, the seriousness of the consequences of forgetting in these situations, comparison of present and past memory functioning, and frequency of memory strategy use. Psychometric evidence supports the factor structure, reliability, and validity of this test (Gilewski et al. 1990; Zelinski, Gilewski, and Anthony-Bergstone 1990). Additional self-report instruments provide information primarily or exclusively about the frequency with which various memory mistakes occur. These include the Memory Assessment Clinics Self-Rating Scale (Crook and Larrabee 1990, Crook and Larrabee 1992), the Everyday Memory Questionnaire (Martin 1986), and the Cognitive Failures Questionnaire (Broadbent, Cooper, FitzGerald, and Parkes 1982). Recently, Lachman, Bandura, Weaver, and Elliott 1995 developed a metamemory questionnaire, the Memory Controllability Inventory, focusing on perceptions of control over one's memory.
Despite the considerable number of self-report memory questionnaires described in the literature, there are several drawbacks associated with their use, especially in clinical settings. First, because some of these questionnaires were developed primarily for research use, the items included do not necessarily reflect aspects of memory that are amenable to clinical memory intervention. For example, items concerning self-evaluation of past memory ability and those involving memory for very remote events would not reflect typical targets of clinical interventions. Rather, clinical interventions tend to focus on problems with recent memory (e.g., remembering new names) and on strategies relevant to everyday life (e.g., repetition, written aids). Additional items pertaining to contentment with one's memory, which is relevant to clinical intervention, are not included in current measures. Second, several of the questionnaires include items that are not applicable to some individuals, such as public speaking, reading novels, driving, or working. It can be difficult to interpret the results of a questionnaire when not all items are answered. Third, some questionnaires, such as the MIA, contain a very large number of items. The considerable amount of time required to complete such a questionnaire could adversely affect compliance, especially in older or cognitively impaired individuals. Finally, although the MIA and MFQ address the multidimensional aspects of metamemory, several existing questionnaires do not. For example, unidimensional questionnaires, such as those that address only the frequency with which memory mistakes occur, do not provide as much information about self-reported memory as does a multifactorial questionnaire. Some aspects of metamemory are not addressed at all in the existing questionnaires; for example, although there are subscales assessing anxiety about memory (e.g., Dixon et al. 1988), there are no scales assessing other common emotional reactions to memory, such as frustration, irritation, unhappiness, or pleasure.
In this article, we present psychometric data on a new questionnaire, the Multifactorial Memory Questionnaire (MMQ), that assesses three dimensions of self-reported memory, including overall contentment or satisfaction with one's own memory ability (MMQ-Contentment), perception of everyday memory ability (MMQ-Ability), and use of everyday memory strategies and aids (MMQ-Strategy). To increase clinical utility, we did not include items unlikely to be amenable to clinical interventions (e.g., comparison of present memory ability to performance when younger) or of little relevance to everyday functioning (e.g., use of laboratory-based mnemonics such as the peg-word system) in the questionnaire. To increase compliance, we made the scales as short as possible (i.e., requiring about 10 min to complete all three) without compromising either completeness of content or statistical reliability and validity. Thus, the MMQ is an improvement over other memory self-report instruments because it encompasses a number of features in combination (i.e., multidimensionality, clinical relevance, brevity, and ease of administration) that are found only partially in any single existing questionnaire. In addition, the Contentment scale of the MMQ includes items pertaining to aspects of memory-related affect, which are not typically included in existing questionnaires.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendix ENDIXReferences |
|---|
) of the Telephone Interview for Cognitive Status (TICS; Brandt, Spencer, and Folstein 1988; n = 40) and/or a 10-item word-list memory task (n = 101). Participants were required to earn at least 30 out of 50 possible points on the modified TICS, a cutoff score recommended by Welsh and colleagues. In addition, participants were required to recall at least two items following a 30-s distraction interval on the word-list memory task. This cutoff score was derived from scores (M = 4.8, SD = 1.6) obtained by an independent sample of 24 older adults (Troyer and Craik 1995) who were screened both for cognitive decline with the Mini-Mental State Examination (Folstein, Folstein, and McHugh 1975) and for medical problems that could affect cognition. Fifteen participants were excluded from the sample because of low scores on the memory task alone (n = 11), low scores on both the modified TICS and the memory task (n = 1), or invalid data on the memory task because of failure to follow instructions (n = 3). This resulted in a total of 115 participants who met the inclusionary criteria. (Although excluded from the main analyses, the 12 participants with low memory were retained in a separate sample for one of the validity analyses, described in the Results section.) In the main sample, the mean age was 71.7 years (SD = 9.9, range = 40–91), and the mean level of education was 13.8 years (SD = 2.8, range = 3–21). A majority (i.e., 79%) of the participants were female.
Memory Questionnaire Scales
The original version of MMQ-Contentment contained 21 items addressing a variety of emotions and perceptions that participants may have about their current memory ability. Statements address positive emotions (e.g., confidence, satisfaction), negative emotions (e.g., embarrassment, irritation), and subjective ability ratings (e.g., comparison to peers, belief that one has a serious memory problem). Respondents rated their level of agreement with each statement on a 5-point scale (i.e., strongly agree, agree, undecided, disagree, and strongly disagree) according to how they felt over the past 2 weeks. For each item, 0 to 4 points were given on the basis of level of agreement such that higher scores indicated greater contentment.
MMQ-Ability contains 20 everyday memory situations, such as remembering appointments, names, and telephone numbers. Items are phrased as memory failures (e.g., forgetting an appointment) and were based in part on previously published questionnaires (Broadbent et al. 1982; Crook and Larrabee 1992; Dixon et al. 1988; Gilewski et al. 1990; Jennings and Hay 1994). Respondents indicated the frequency with which each mistake occurred over the last 2 weeks on a 5-point scale (all the time, often, sometimes, rarely, never). For each item, 0 to 4 points were given on the basis of self-reported frequency, with higher scores indicating better subjective memory ability (i.e., fewer memory mistakes).
The original version of MMQ-Strategy contained 20 items describing different memory aids and strategies applicable to everyday memory tasks, such as writing appointments on a calendar and repeating information to oneself. Items were based in part on previously published questionnaires and interviews (Dixon et al. 1988; Harris 1980; Intons-Peterson and Fournier 1986; Lovelace and Twohig 1990; Park, Smith, and Cavanaugh 1990). Respondents indicated the frequency with which each strategy was used over the last 2 weeks using a 5-point scale (never, rarely, sometimes, often, all the time). For each item, 0 to 4 points were given on the basis of frequency of use, with higher scores indicating more frequent use of memory aids and strategies.
A few participants omitted some questionnaire items. For these scales, scores were prorated on the basis of the responses to completed items, and this prorated score was used in the analyses when possible. This method was used only if very few items (i.e., four or fewer) were omitted, which occurred on 26 out of 344 scales. There were no scales with more than four items omitted.
Other Questionnaires and Tests
In addition to the MMQ, some participants also completed previously validated self-report measures of memory (n = 52) or mood (n = 44). Memory questionnaires included the MIA (Dixon et al. 1988) and the MFQ (Gilewski et al. 1990). The entire questionnaires were administered, although we selected only certain subscales to compare to the MMQ. The MIA Anxiety and Change subscales and the MFQ Seriousness of Forgetting subscale were selected as measures of memory-related affect and perceived change in memory functioning expected to be related to MMQ-Contentment. The MIA Capacity subscale and the MFQ General Frequency of Forgetting subscale were selected as measures of subjective memory ability expected to be similar to MMQ-Ability. The MIA Strategy subscale and the MFQ Mnemonics subscale were selected as measures of memory strategy use expected to be similar to MMQ-Strategy. Measures of general mood included the Geriatric Depression Scale (Yesavage et al. 1983) and the Anxiety scale from the Symptom Checklist-90-Revised (Derogatis 1994).
Participants completed three objective memory tasks. Because of the group setting, it was not possible to use clinical memory tasks; thus, memory tasks that allowed simultaneous visual and oral presentation of information as well as a self-paced written test were created. The word-list memory test, as described previously, was used as a laboratory-based measure of recent memory. The score was the number of words correctly recalled after a 30-s delay. A name-memory task was used as an everyday measure of recent memory. For this test, two three-part name strings (i.e., first, middle, and last) were presented via slide projection for 10 s and were simultaneously read aloud. Free recall was tested following a 5-min delay filled with lecture. Points were awarded for both accuracy (i.e., correct name and spelling) and placement (i.e., correct sequencing of names within a string and no confounding of names between the two strings). As an everyday measure of prospective memory, participants were asked to make two telephone calls at specified times and to provide specific information (i.e., name and identification number). Points were awarded both for the promptness of the calls and for the information provided.
Participants also completed two tests of attention. As a speeded test of attention, a symbol-digit matching task was used. Each participant was given a sheet of paper with six symbol-digit pairs above rows of symbols alone. Participants were instructed to write the appropriate digit beneath each symbol as quickly as possible for 60 s. Scores were calculated as the number of correct symbol-digit matches. A digit span task was also created in which participants listened to strings of three to eight digits as they were read aloud and wrote down the digits immediately after presentation. The score was the longest span of digits written in correct sequence.
Procedure
Questionnaires and tests were administered in a fixed order to groups of 5 to 15 participants, as follows: MMQ Contentment, Ability, and Strategy scales; then the objective memory tests (word-list memory, name memory, and prospective memory telephone task); then the tests of attention (symbol-digit followed by digit span). For participants who were tested on two occasions, only scores from the first session were used unless otherwise indicated. Although all 115 participants completed the three memory questionnaire scales at least once, some did not complete all of the other tests and questionnaires. The MIA, MFQ, and/or the mood measures were administered at the end of the session to the subsample of participants who completed them.
For the purpose of content validation, the 61 MMQ items were randomly ordered and were worded with similar terminology (i.e., all items began with the phrase "Do you ..."). We recruited 12 memory experts from the psychology department and the research division at Baycrest Centre for Geriatric Care to rate questionnaire items. Each expert had a PhD in psychology; 6 were licensed clinicians who work with memory-disordered patients, and 6 were professionals who primarily conduct memory research. We asked raters to classify each item as belonging to one of three metamemory domains: emotions and perceptions of one's own memory, everyday memory mistakes or problems, or memory strategies or aids.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendix ENDIXReferences |
|---|
Factor Structure
We performed a principal components analysis with varimax rotation using the responses on the remaining 60 items from all 115 participants. We forced the solution to three components and, with few exceptions, each item loaded most highly onto the expected component. Eigenvalues of the three components ranged from 3.6 to 15.3. Eighteen of the 20 items from MMQ-Contentment loaded most highly onto a single component. The loadings from these 18 items ranged from .35 to .79, with a mean loading of .66. All 20 items from MMQ-Ability loaded most highly onto another component, with loadings ranging from .45 to .74 and a mean loading of .59. Nineteen of the 20 items from MMQ-Strategy loaded most highly onto a third component. The loadings from these items ranged from .27 to .70 with a mean loading of .50. The 3 items that did not load onto the expected components were eliminated from the subsequent analyses, for a total of 57 remaining items. These items and their component loadings are listed according to their respective scales in the Appendix. All loadings are positive because each item was scored such that higher scores indicated higher levels of contentment. In all, the three factors accounted for 40.4% of total score variance.
We retabulated scores for each participant using the retained items. Pearson correlations calculated between the three questionnaires indicated a strong relation only between MMQ-Contentment and MMQ-Ability, r(115) = .62, p < .001. MMQ-Strategy was not highly correlated with either MMQ-Contentment, r(114) = -.17, p = .078, or MMQ-Ability, r(114) = -.15, p = .104.
Reliability
We examined 4-week test-retest reliability using data from 24 participants who were tested on two occasions and did not receive any memory interventions. Correlations indicated highly reliable scores on MMQ-Contentment, r(24) = .93, p < .001; MMQ-Ability, r(24) = .86, p < .001; and MMQ-Strategy, r(24) = .88, p < .001. Unlike other correlation coefficients (which need to be squared), reliability coefficients are interpreted directly as the proportion of variance accounted for (Anastasi and Urbina 1997; McNemar 1969). Thus, 86–93% of the total test-retest score variance was due to true variance rather than to the effects of random daily changes in participant or environmental conditions.
We examined internal consistency using Cronbach's alpha on data from all participants. These analyses indicated highly reliable scores on MMQ-Contentment (
= .95), MMQ-Ability ( = .93), and MMQ-Strategy ( = .83). Thus, at least 83–95% of the total within-test score variance was due to true score variance rather than to item content heterogeneity or poor item quality.
Construct Validity
We tested convergent validity by examining effect sizes (Cohen 1988) based on correlations between scores obtained on the present memory questionnaires, other self-report memory and mood questionnaires, and objective memory tests. For MMQ-Contentment, correlations showed a large effect size with both the Anxiety and Change subscales of the MIA, rs(50) = -.57 and .61, respectively, ps < .001; medium effect sizes with the Seriousness of Forgetting subscale of the MFQ, r(51) = .45, p = .001, and with the Geriatric Depression Scale, r(43) = -.41, p = .007; and a small effect size with the Anxiety subscale of the Symptom Checklist-90-Revised, r(43) = -.27, p = .079. For MMQ-Ability, correlations showed large effect sizes with the Capacity subscale of the MIA, r(50) = .60, p < .001, and the General Frequency of Forgetting subscale of the MFQ, r(51) = .70, p < .001. For MMQ-Strategy, correlations showed large effect sizes with the Strategy subscale of the MIA, r(50) = .64, p < .001, and the Mnemonics subscale of the MFQ, r(51) = .66, p < .001.
As expected, the relation between self-reported memory function and performance on objective memory tests was weaker. Small to negligible effect sizes were obtained from correlations between MMQ-Ability and performance on the word-list memory test, r(101) = .10, p = .309; name memory test, r(101) = .09, p = .398; and the telephone task, r(94) = .14, p = .166.
To further examine the relation between self-reported memory and objective memory performance, we examined MMQ-Contentment and MMQ-Ability obtained by a group of participants with documented low memory performance (i.e., more than 2 standard deviations below the mean on the word-list memory task) who were excluded from the main sample, as described previously. Each individual from this low-memory sample was matched with 1 or 2 (whenever possible) individuals from the main sample according to sex, age (within 3 years), and education (within 2 years). The low-memory (n = 12) and normal-memory (n = 20) groups did not differ in age (Ms = 74.9 and 74.5 years), level of education (Ms = 12.6 and 12.8 years), or sex (67% and 65% female), p > .80. Scores on MMQ-Contentment were lower in the low-memory group (M = 30.7) than the normal-memory group (M = 43.1), t(30) = 3.10, p = .004. Similarly, scores on MMQ-Ability were lower in the low-memory group (M = 37.3) than the normal-memory group (M = 46.2), t(30) = 3.19, p = .003.
We examined discriminant validity by calculating correlation coefficients between the memory questionnaires and tests of attention. Scores on the memory questionnaires were unrelated (i.e., showed negligible effect sizes) to the symbol-digit task, rs = -.02 to .00, ps > .85, and the digit span task, rs = .01 to .07, ps > .50.
Relation to Demographic Variables and Mental Status
Scores on MMQ-Contentment, MMQ-Ability, and MMQ-Strategy were not related to age, rs = -.13 to .07; education, rs = .02 to .07; or sex, rs = -.13 to .13, all ps > .170. Moreover, MMQ scores were not significantly related to performance on the modified TICS, rs = .03 to .20, ps > .220.
Descriptive Statistics
Means, standard deviations, and percentiles based on performance by all 115 participants are provided in Table 1 . To examine the shape of the score distributions, we calculated skewness (S) and kurtosis (K) coefficients and tested them for significance using z scores. Scores from the three questionnaire scales did not show significant skewness, Ss = -0.43 to 0.21, zs = -1.89 to 0.94, ps > .05. Scores from MMQ-Contentment and MMQ-Strategy did not show significant kurtosis, Ks = -0.80 to -0.03, zs = -1.78 to -0.07, ps > .05, but scores from MMQ-Ability were significantly peaked, K = 1.69, z = 3.79, p < .001.
|
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendix ENDIXReferences |
|---|
On the basis of previous reports indicating the multidimensional nature of metamemory (e.g., Dixon and Hultsch 1983a; Gilewski et al. 1990), the present scales were developed so that they could be administered separately for the assessment of separate dimensions of memory appraisal (i.e., Contentment, or feelings about one's memory; Ability, or the subjective impression of one's memory capabilities; and Strategy, or one's reported frequency of use of various memory aids). The final assignment of items to each questionnaire scale is supported in several ways. First, there is excellent interrater agreement among 12 memory experts from both research and clinical settings, which demonstrates content validity for our initial distribution of items across the various scales.
Second, principal components analysis yields three meaningful factors, each of which is associated with high loadings by the items from one scale but not the other two. This pattern of loadings provides a measure of factorial validity that confirms our a priori division of items into scales representing different dimensions of metamemory. Correlational analysis reveals a strong, positive relation between MMQ-Contentment and MMQ-Ability. As one might expect intuitively, self-ratings of high competence regarding one's memory abilities in performance-demanding situations are associated with high ratings of satisfaction with one's memory. Conversely, individuals with feelings of low competence naturally are less satisfied with their memory. The lack of a relation between self-reported strategy use and ratings of either contentment or ability is also not surprising. Individuals with perceived memory deficiencies might use strategies as an intentional compensatory technique, in which case frequency of strategy use would correlate negatively with self-reported memory abilities and contentment. However, those who rate their memory ability positively and feel satisfied with it may achieve such memory success precisely because they use strategies frequently, in which case reported frequency of strategy use would correlate positively with ratings on the other two scales.
Reliability of the MMQ scales is demonstrated in two ways. Test-retest reliability in a subgroup of participants who completed the questionnaires in two sessions 4 weeks apart (with no intervention) is excellent for all three scales. Moreover, the intratest reliability of each of the questionnaires is acceptable, as indicated by measures of internal consistency.
Construct validity is also assessed in several ways. As expected, convergent validity is demonstrated by medium-to-large effect sizes of correlations between the present instruments and their corresponding scales on other multidimensional metamemory questionnaires and subjective mood scales. Specifically, feelings about one's memory (contentment vs dissatisfaction) are related to ratings of anxiety (on the MIA and Symptom Checklist-90-Revised), perceived change in memory functioning (on the MIA), feelings of seriousness of forgetting (on the MFQ), and endorsement of depressive symptoms (on the Geriatric Depression Scale). The fact that a much larger effect size is observed for the correlation between MMQ-Contentment and the Anxiety scale of the MIA than for its correlation with depressive symptomatology or generalized feelings of anxiety further validates the construct of feelings about one's memory as a specific dimension in a multidimensional model of metacognition. Similarly, scores on MMQ-Ability and MMQ-Strategy, which reflect one's perceived memory capabilities and reported frequency of use of memory aids, respectively, are related to scores on corresponding scales of the MIA (Capacity and Strategy) and MFQ (General Frequency of Forgetting and Mnemonics).
Both the MIA and MFQ are well-established multidimensional questionnaires with good psychometric properties (Dixon et al. 1988; Gilewski et al. 1990), and both have been used frequently in studies of cognitive aging. Thus, it is a testament to the construct validity of the three scales developed and reported in this study that scores on them are significantly related to scores on the corresponding scales of these other instruments. At the same time, it is gratifying to note that the largest correlation observed (i.e., between MMQ-Capacity and MFQ General Frequency of Forgetting) was .70, which accounts for only half the variance in scores on the two scales. That scores on the present scales are significantly but not excessively related to scores on other instrument subscales indicates that the new questionnaire scales are similar to existing questionnaires yet provide additional information about unique aspects of memory self-appraisals.
Among the main sample of healthy, elderly individuals who participated in this study, questionnaire ratings show very low correlations with objective memory performance across both traditional experimental tasks, such as word-list memory, and everyday memory tasks, such as name memory and a prospective memory telephone task. Other studies have also found self-reports of memory ability to be highly reliable but not very accurate in predicting actual memory performance (e.g., Little, Williams, and Long 1986; Rabbitt and Abson 1990; see Herrmann 1982, Herrmann 1990, for reviews). Others (Hertzog, Saylor, Fleece, and Dixon 1994) have found that the relationship between metamemory and objective memory performance is mediated by a number of variables, such as age, type of memory task, and experience with the task. Differences between memory self-appraisals and objective memory performance may arise from several sources. For social reasons, memory failures are not necessarily pointed out to people by their friends; on the contrary, reports by others may be deliberately misrepresented in order to flatter—or to intimidate.
If people cannot rely on their friends to give them truthful feedback about their performance, they likely do no better to rely on themselves either. When reporting all the words one can recall on a memory test, for example, one may not be aware of those that are not recalled. These factors may contribute to inaccuracies in evaluating one's own memory performance. In addition, to the extent that variation in self-reported memory failures is driven by general beliefs about memory and aging rather than monitoring of specific incidents of faulty memory, self-appraisal of memory ability would be uncorrelated with objective memory performance. Posing questionnaire items as failures (as opposed to an "agree-disagree" format) might elicit implicit theories or stereotypes about age-related memory declines in general. If so, the format of the MMQ may have contributed to the lack of relation between reported and actual memory abilities.
Another reason for the poor relationship between memory appraisals and performance is that most memory tests are not predictive of everyday memory performance. People may use compensatory strategies (such as written aids) in everyday life that are not used in laboratory testing. They may also have more practice with the types of everyday memory situations presented on questionnaires than with those in laboratory tasks. The name-memory and telephone tasks used in the present study appear to be more relevant to everyday memory than word-list learning in the sense that learning names and making telephone calls are everyday tasks; nevertheless, they were contrived with standardized administration and scoring procedures and likely do not tap the same underlying construct as learning the name of someone to whom one has been introduced recently or remembering to make a phone call that is meaningful in one's everyday life. Moreover, self-report questionnaires are not intended to be used as screening instruments for actual memory problems; instead, researchers should use the two types of tools in a thorough assessment of memory to provide information about both subjective memory beliefs and objective memory performance.
The lack of relation between the MMQ scales and the telephone task, which is a prospective memory task, is consistent with the results of several other studies indicating that subjective memory assessment is unrelated (or is negatively related) to performance on prospective memory tasks (Dobbs and Rule 1987; Sunderland, Watts, Baddeley, and Harris 1986; Zelinski et al. 1990). Some studies have found modest correlations between self-reported memory and objective memory performance. For example, Dixon and Hultsch 1983b found that metamemory scores on the Task and Strategy subscales of the MIA were the best overall predictors of memory for text in adulthood. However, consistent with the present results, Dixon and Hultsch found no relation among older adults between memory performance and any of the three MIA scales (i.e., Anxiety, Capacity, and Strategy) that correspond to the MMQ scales. Conversely, there does seem to be a good correspondence between self-reports and memory performance when the self-reported aspects of metacognition or metamemory are directly related to the outcome measure. For example, self-reported strategy use on a recall test is positively related to recall performance on that test (Hertzog, McGuire, and Lineweaver 1998). Similarly, self-reported memory efficacy in both younger and older adults, as measured by predicted memory performance on a specific task, is related to recall performance on that task, thereby demonstrating age-invariant accuracy in metacognitive predictions (Connor, Dunlosky, and Hertzog 1997). Thus, it appears that the relation between metamemory and objective memory performance differs depending on whether the self-reported memory reflections refer to general aspects of memory (i.e., abilities or strategy use) or, instead, are directed to the specific test used as the outcome measure of memory performance.
Although memory self-appraisals are unrelated to memory performance in our main sample of healthy, elderly participants, scores on two of the three scales do differ as a function of objective memory performance when participants with documented poor memory abilities (who were excluded from the main analyses) are included. Specifically, participants with poorer memory have lower scores on both the MMQ-Contentment and MMQ-Ability scales relative to the main sample. The fact that the scales discriminate between older adults with low versus normal objective memory performance suggests that the failure to find a relation between questionnaire ratings and memory performance in this and other studies may be due in part to the standard practice of excluding individuals with memory impairments, which results in a restricted range of memory ability.
We assessed discriminant validity by comparing scores on the memory questionnaires to performance on attention measures (i.e., symbol-digit and digit span). As expected, memory appraisals and attention are unrelated, thereby demonstrating another aspect of construct validity. In addition, MMQ scores are unrelated to demographic variables and global mental status among the older adults participating in this study. It is possible, however, that a significant correlation between age and MMQ would have emerged if younger adults were included, thereby increasing the age range.
In sum, evidence for the reliability and validity of the MMQ is obtained among a sample of 115 healthy older adults. The relatively short length of the MMQ (i.e., 57 items covering three distinct scales) and the intended applicability to clinical intervention should enhance the usefulness of the questionnaires in clinical and research settings. Future studies should be conducted to determine whether the factor structure observed here remains constant across different samples (e.g., young individuals, cognitively impaired elderly persons) and longer time frames.
|
Acknowledgments |
|---|
Received for publication September 9, 1999. Accepted for publication August 28, 2000.
|
Appendix ENDIX |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendix ENDIXReferences |
|---|
Metamemory Questionnaire-Contentment
Metamemory Questionnaire-Ability
Metamemory Questionnaire-Strategies
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendix ENDIXReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
