Norms for CERAD Constructional Praxis Recall

Abstract

Recall of the 4-item constructional praxis measure was a later addition to the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) neuropsychological battery. Norms for this measure, based on cognitively intact African Americans age ≥70 (Indianapolis-Ibadan Dementia Project, N=372), European American participants age ≥66 (Cache County Study of Memory, Health and Aging, N=507), and European American CERAD clinic controls age ≥50 (N=182), are presented here. Performance varied by site; by sex, education and age (African Americans in Indianapolis); education and age (Cache County European Americans; and only age (CERAD European American controls). Performance declined with increased age, within age with less education, and was poorer for women. Means, standard deviations, and percentiles are presented separately for each sample.

Introduction

The neuropsychological battery developed by the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) to assess stage of Alzheimer’s disease (AD) includes five measures: verbal fluency; abbreviated Boston Naming test (Kaplan, Goodglass, & Weintraub, 1983); Mini-Mental State Examination (MMSE; Folstein, Folstein, & McHugh, 1975); 10-item Word List Learning, Recall, and Recognition test; and constructional praxis (CP; Rosen, Mohs, & Davis, 1984). Several years after the initial battery was developed, a CP recall measure was added in response to concern about disproportional focus on assessment of verbal abilities. Clinicians and researchers requested praxis memory in addition to verbal memory, to better describe the characteristics and course of AD. With heightened interest in different forms of mild cognitive impairment (Petersen, 2004), and proposed DSM 5 categories of neurocognitive decline, information on praxis memory becomes increasingly important.

Norms for the original CERAD measures have been published for different settings, languages and cultures (e.g., Berres, Monsch, Bernasconi, Thalmann, & Stahelin, 2000; Ganguli, Ratcliff, & DeKosky, 1997; Guruje et al., 1995; Karrasch & Laine, 2003; Lee et al., 2004; McCurry et al., 2001; Stewart, Richards, Brayne, & Mann, 2001; Unverzagt et al., 1999; Welsh et al., 1994), but have not always provided norms on the CP recall test. Exceptions includeBerres et al. (2000), for speakers of German in Europe;Lee et al. (2004), for Korea;McCurry et al. (2001) for Japanese Americans (speakers of Japanese or English); and Welsh-Bohmer et al. (2009), for English speakers in the U.S. With the exception of the study by Lee et al. (2004), focus has been on investigator-identified community residents with normal cognition. The present paper complements these and adds to the Lee et al. (2004), clinically-based study by providing CP recall norms for carefully evaluated, cognitively normal epidemiologically identified African Americans and European Americans, as well as for volunteer control subjects followed at tertiary medical center memory disorders clinics. The information provided is intended to be useful in determining the status of community residents who are being evaluated for epidemiological or clinical purposes. Information on European American volunteer control subjects indicates the level of performance that may be expected in Memory Disorders Clinic volunteers – a comparison group frequently used in clinical studies. Information on Memory Disorders Clinic volunteers comes from the original English-speaking CERAD sample (Heyman et al., 1997; Morris et al., 1989), and completes information on norms for the CERAD neuropsychology battery based on that sample.

Methods

Samples

The performance of cognitively normal older persons on the CP recall test was available from two epidemiological surveys designed to assess the prevalence and incidence of Alzheimer’s disease and other dementing disorders. These were the Indianapolis-Ibadan Dementia Project (Hendrie et al., 1995; 2001), from which we selected the African Americans in Indianapolis, and the Cache County Study of Memory, Health and Aging (Welsh-Bohmer et al., 2009), carried out in Utah with an overwhelmingly (99.6%) European American sample. In addition information was available from a small group of cognitively normal European American control subjects entered into CERAD (Morris et al., 1989). Thus we are able to compare older, cognitively intact, epidemiologically identified African Americans with epidemiologically identified European Americans, and the latter with European American clinic volunteers.

African Americans in Indianapolis

Data on African Americans comes from the Indianapolis component of the Indianapolis-Ibadan Dementia Project, a study of the prevalence and incidence of Alzheimer’s disease and dementia (Hendrie et al., 1995; 2001). A sample of African Americans 65 years of age and over was drawn from a random sampling of 60% of the households in 29 contiguous census tracts in Indianapolis in which 80% of the residents were African American according to 1990 census data. Of the invited participants, 85% agreed to enter the study. In addition, all African Americans in a sample of representative nursing homes in the same area were included in the study. The cohort was formed in 1992–1993 with a sample size of 2,212 and followed longitudinally at 2–3 year intervals. A two-stage design was used, with screening via the Community Screening Instrument for Dementia (Hall et al., 1993). All who fell below the screen’s cutpoint, a randomly selected 50% in the intermediate range, and 5% in the good performance group received clinical evaluations which consisted of an examination by a clinician (physician or nurse), structured informant interview, and cognitive testing. The clinician examination included physical, neurological, and mental status examinations. The informant interview was a structured interview adapted from the CAMDEX (Hendrie, et al., 1988), and conducted by a research nurse with a family member who knew the subject well, to probe for cognitive symptoms and current performance in daily functioning. Cognitive testing included the CERAD neuropsychology battery (Morris et al., 1989). Diagnoses (normal cognition, mild cognitive impairment, and dementia) were made in a consensus conference of geriatric psychiatrists, neurologists, and neuropsychologists using standard diagnostic criteria. Stage of dementia was assessed by the Clinical Dementia Rating scale (Morris, 1993), on which a rating of 0 indicates cognitively normal; 0.5 indicates questionable dementia, very mild dementia, or now mild cognitive impairment; and values of 1 and greater indicate increasing levels of impairment in dementia.

Data for the current study on CP recall come from 1997–2004 when the delayed recall portion of this measure was first introduced in the Indianapolis-Ibadan study. Sample members were then 70 years of age and older; 79 had been administered CP copy at a prior wave (19 of them at two prior waves). Of the 395 participants identified with normal cognition, 372 had a CDR rating of 0. Of the remaining 23, one had no CDR score, one had a CDR rating of 1, and 21 had a CDR rating of 0.5. Only persons with a CDR score of 0 were included in analysis to insure unquestionable normal cognition. Complete information was available on all 372 participants, three of whom were excluded since their CP recall scores were out of range. The analysis sample therefore consists of 369 participants, 11 of whom had prior exposure to CP copy.

Cognitively intact Cache County Study participants

In 1994, all persons 65 years of age and older, living in Cache County, Utah, were invited to participate in the study; 90% (5092/5067), did so. In 1995, these participants received an initial screening based on an adapted version of the modified Mini-Mental State Exam (3MS-R; Tschanz et al., 2002), or a proxy interview using the Informant Questionnaire for Cognitive Decline (IQCODE; Jorm & Jacomb, 1989). Those 90 years of age and over, all persons scoring below predetermined cutoffs on the 3MS-R and IQCODE, and randomly selected individuals above the cutoff, matched in terms of gender, 5-year age group and APOE ε4 status (in a ratio of approximately 2 normals:1 anticipated AD case), were selected for clinical evaluation.

Clinical evaluations were done in the home by a nurse and psychometrist, who determined blood pressure, performed physical and neurological examinations, administered an extended CERAD neuropsychological battery, reviewed cognitive symptoms, took a medical history, and determined medication use. Individuals suspected of having a dementing disorder were further examined by a board certified geropsychiatrist, and standard laboratory studies and neuroimaging were carried out where feasible. Diagnosis of dementia and AD followed DSM-III-R and NINCDS-ADRDA criteria, with final determination by a consensus panel of neurologists, geriatric psychiatrists, neuropsychologists, and behavioral neuroscientists (Breitner et al., 1999). Of 993 persons who were fully clinically evaluated, 507 were rated as clinically normal, i.e., dementia, and other cognitive impairments including mild cognitive disorders were determined to be absent. Using this sample of 507, norms have previously been published for the measures of the CERAD neuropsychological battery used in the Cache County Memory Study (Welsh-Bohmer et al., 2009); information on CP recall was available for 484 participants. The CP recall data have been re-analyzed here, using age categories that permit direct comparison across the three data sets represented in the present paper.

CERAD European American control cases

Cases and controls enrolled in CERAD by the then extant Alzheimer’s Disease Research Centers and other major AD centers in the U.S., had to meet strict inclusion/exclusion criteria (Morris et al., 1989). At entry, control subjects had to be cognitively intact, and had to meet the same health criteria as AD cases (be ambulatory, have no health condition that could affect cognitive performance). In addition they could not be blood relatives of an AD case. Stage of disease was determined by the Clinical Dementia Rating scale (CDR; Hughes, Berg, Danziger, Coben, & Martin, 1982). Between 1987 and 1995, 429 European American, English-speaking control subjects were entered into CERAD. CP recall was not part of the original battery, but was added in 1993. Of those then enrolled in CERAD, 182 European American controls were eligible to be administered the CP recall test, with response obtained from 177. (Data were only available on 11 African American controls, a number too small for consideration here.) All but 1 subject had been administered the CP copying test on one or more previous occasions.

Each study was approved by the IRB at each site.

CERAD Neuropsychology Battery

The order of administration of the CERAD neuropsychology measures is as follows: verbal fluency (naming as many animals as possible in 60 seconds), abbreviated 15-item Boston Naming test, Mini-Mental State Examination (MMSE), 10-item Word List Learning test (administration of a 10-item word list on three consecutive occasions, each time using a different randomized order with immediate recall after each administration), CP copy, Word List Recall (unprompted recall of the 10-item word list), Word List Recognition (identification of the 10 original word list items in a list of 20 words), and CP recall. The time interval between CP copy and CP recall is 2 to 2.5 minutes.

Constructional Praxis tasks

The CP copying task consists of four figures (circle, diamond, overlapping rectangles, Necker cube). Each figure is printed on the top half of a regular size sheet of paper, and presented individually to the subject, who is asked to copy the figure in the space below. The scoring range is 0–11 (criteria are given in Table 1). The subject is not informed about the later recall task, which is thus an incidental learning task.

Table 1.

Scoring criteria for Recall of CP

Item	No. of points	Criteria
Circle	2	Closed within 1/8” Circular shape
Diamond	3	Draws 4 sides Closes all angles, within 1/8” Sides of approximately equal length
Overlapping rectangles	2	Both rectangles are 4-sided Overlap resembles the original
Necker’s cube	4	Figure is 3-dimensional Frontal face correctly oriented (may be right or left oriented) Internal lines correctly drawn Opposites sides are parallel (within 10°)

Uniform test administration

At all sites, testers were trained to administer and score CP recall according to procedures developed by CERAD, and their competence to do so was evaluated.

Statistical Analysis

Descriptive statistics were used to characterize the sample. SAS Proc GLM was run separately for each site to determine whether there were significant differences in performance as a function of sex, education, or age. All three were significant for Indianapolis, education and age were significant for Cache county, and only age was significant for CERAD. SAS Proc GLM was also used to determine whether there were differences in performance across the three sites after sex, age, and education were controlled when (a) all data from each site was entered, (b) when data was restricted to that in common across all sites (i.e., age range 70–94, education 9–15 years), and (c) comparing the two European American sites (age range restricted to 70–94, education unrestricted since they had a similar education distribution).

In calculating norms, age was categorized in overlapping decades starting at 5-year intervals (e.g., 70–79, 75–84, 80–89, 85–94, but with the oldest age category adjusted, where relevant, to reflect the highest age in the sample), maximizing the clinical usefulness of the data (Pauker, 1988). Education was categorized as 0–8, 9–11, 12–15, and ≥16 years. While this represented the educational range across all three samples, distribution was skewed; the lowest educational category was most strongly represented by the Indianapolis sample, and the highest by the Cache county sample. Means and standard deviations and percentiles were determined for each cell using SAS Proc Means, and are presented only when at least 10 people are included in a given cell. While only whole scores are recorded for CP recall, Proc Means provides percentile scores that sometimes end in .5. Such scores were rounded down to the next integer, to give poor performers the benefit of the doubt. All analyses were run in SAS version 8.2 (SAS Institute, Cary, NC).

Results

The three samples differed in terms of ethnicity (all but two Cache county sample members were European American, they were retained to maintain comparison with the original study (Welsh-Bohmer et al., 2009), gender distribution, education, and age (Table 2). Just over two thirds of the Indianapolis and CERAD samples were female, compared to just over half for Cache County. Indianapolis sample members reported fewer years of education. Age reflected study entry criteria, and time during the study when CP recall was introduced.

Table 2.

Demographic characteristics of the samples

	Indianapolis (African Americans) (N = 369)		Cache County (European Americans) (N = 507)		CERAD controls (European Americans) (N = 182)
	N	%	N	%	N	%
Ethnicity   African American   White   Asian   Other	369 0 0 0	100.0 0 0 0	0 505 1 1	0 99.6 0.2 0.2	0 182 0 0	0 100.0 0 0
Sex   Male   Female	112 257	30.4 69.6	227 280	44.8 55.2	58 124	31.9 68.1
Education    0–8 years    9–11 years   12–15 years   ≥16 years Mean (standard deviation)	107 104 144 14 10.23	29.0 28.2 38.0 3.8 (3.33)	22 65 286 134 13.41	4.4 12.8 56.4 26.4 (2.94)	12 7 116 77 14.30	6.6 3.8 47.3 42.3 (3.12)
Age (years)   50–59   60–69   70–79   80–89   90+ Mean (standard deviation)	0 0 209 138 22 79.27	0.0 0.0 56.6 37.4 6.0 (5.64)	0 40 217 189 61 79.76	0.0 7.9 42.8 37.3 12.0 (7.54)	23 76 70 10 3 65.87	12.6 41.8 38.5 5.5 1.6 (7.98)

Analyses unrestricted on age and education indicated that each site was significantly different from each other site. Least squares means (adjusted for demographic characteristics, reported since this is an unbalanced design) and (standard errors) for Indianapolis, Cache county, and CERAD were 6.25 (0.14), 7.87 (0.12), and 6.80 (0.22) respectively, the significance of the difference between the Indianapolis and CERAD scores was P <.05, while that between the other comparisons was P <.0001. Analyses restricted to a common age and education range gave slightly different results. The significance of the difference between Indianapolis and Cache county remained (P < .0001), but the Cache County/CERAD difference declined to P <.002, and there was no longer a significant difference between Indianapolis and CERAD. The least squares means (Indianapolis, Cache county, CERAD) were 5.98 (0.16), 7.69 (0.15), 6.22 (0.44). Finally, analysis comparing Cache county with CERAD, where age range was restricted to be the same for both groups but restriction was not needed for education, found a statistically significant difference (P < .0001), with least squares means (standard errors) being 8.12 (0.11) and 7.09 (0.23) respectively. Based on these findings, information is presented separately for each sample.

Simultaneous consideration of sex, education and age showed that their associations with CP recall score varied by study. All three were statistically significant for the Indianapolis sample (education, P<0.001; age, P=0.0003; sex, P=0.042). For the Cache county sample only education (P<0.0001) and age (P<0.0001) were significantly associated. For CERAD controls only age (P<0.0001) was significantly associated with CP recall scores.

For all three samples, age was categorized in overlapping decades, increasing by 5-year intervals. Ages 50–59, 55–64, and 60–69 are present only for CERAD, and 66–69 only for Cache County. The youngest and oldest age categories vary for each group, as indicated in Tables 3 and 4, otherwise the age categories are comparable across the three sites. Only age categorization was used for CERAD, since neither education nor sex was related to score in this sample. The same education categories were used for the Indianapolis and Cache county samples, capturing the lower educational experience of the former group, and the higher educational experience of the latter. Education was not categorized for CERAD since it was not associated with score in this sample. Means, standard deviations, and percentiles are given separately for each group (Table 3 for Indianapolis, Table 4 for Cache county and CERAD).

Table 3.

Constructional Praxis 4-item recall – norms for Indianapolis sample (African American): means, standard deviations, and percentile scores by sex, education and age categories (Indianapolis – epidemiological sample).

			Percentile
	N	Mean (SD)	95	90	75	50	25	10	5
Education: 0–8 years
Women
Age 70–79	35	4.71 (1.89)	8	8	6	4	4	2	2
Age 75–84	47	4.62 (2.44)	9	8	6	4	3	1	1
Age 80–89	31	4.35 (2.73)	9	8	6	4	2	1	1
Age 85–96	19	3.26 (2.47)	8	7	5	4	1	0	0
Men
Age 70–74	10	5.40 (2.91)	11	9*	8	5	3	2	0
Age 70–79	24	5.17 (2.76)	10	9	7	5	3	2	2
Age 75–84	20	5.40 (2.89)	10*	9*	7*	5	3	2	1
Education: 9–11 years
Women
Age 70–74	20	5.75 (2.57)	9	8*	8	6	4	2	1
Age 70–79	40	5.93 (2.58)	9*	9	8	6	4	2	2
Age 75–84	40	5.95 (2.65)	10	9	8*	6	4	2*	1
Age 80–89	26	5.88 (2.50)	9	9	8	6	5	3	1
Men
Age 70–79	25	6.44 (2.62)	10	9	9	6	5	3	2
Age 75–84	22	5.68 (2.48)	9	9	8	6	4	2	2
Age 80–89	10	4.30 (2.16)	8	7*	5	4*	3	1*	1
Education: 12–15 years Information on 16+ years of education is not provided because of sample size
Women
Age 70–74	21	5.95 (3.41)	10	10	9	7	2	2	1
Age 70–79	56	6.16 (2.87)	10	9	8	7	4	2	1
Age 75–84	66	6.27 (2.71)	10	10	8	7	5	2	1
Age 80–89	44	5.34 (3.01)	10	10	8	5	3	2	0
Age 85–96	19	3.21 (2.02)	8	6	4	3	2	0	0
Men
Age 70–74	10	7.90 (2.38)	11	10*	10	8	7	4*	3
Age 70–79	20	7.60 (2.59)	11	11	10	7	6	3*	3
Age 75–84	24	6.79 (3.01)	11	11	9	6*	5	3	3
Age 80–89	16	6.13 (3.07)	11	11	8*	6	4	3	0

N = number of respondents in group. In the Indianapolis sample there were no non-respondents.

(SD) = (Standard deviation)

^*indicates that score has been rounded down to the next whole number

Table 4.

Constructional Praxis 4-item recall -- norms for Cache County (epidemiological sample, European American) and controls at CERAD-participating memory disorders clinics (European American): means, standard deviations and percentile scores by education and age categories (Cache County), and age categories only (CERAD).

			Percentile
	N (group size)	Mean (SD)	95	90	75	50	25	10	5
Cache county Education: 0–8 years No information provided, sample too small (n = 22) Education: 9–11 years
Age 66–74	13 (13)	8.08(2.60)	11	10	10	9	7	5	2
Age 70–79	19 (19)	7.79 (2.12)	10	10	9	8	7	5	2
Age 75–84	23 (24)	7.48 (2.29)	10	10	9	8	7	5	2
Age 80–89	24 (25)	6.67 (2.44)	9	9	9	7*	6	3	2
Age 85–94	22 (24)	5.77 (2.79)	9	9	8	6	4	2	1
Age 90–102	12 (16)	5.25 (3.33)	9	9	8*	5	2*	1	0
Education: 12–15 years
Age 66–69	23 (23)	8.83 (2.08)	11	11	11	9	8	6	5
Age 66–74	81 (81)	8.63 (1.87)	11	11	10	9	7	6	5
Age 70–79	129 (129)	8.40 (2.06)	11	11	10	9	7	6	5
Age 75–84	132 (135)	8.13 (2.07)	11	11	10	8	7	6	4
Age 80–89	97 (105)	7.38 (2.34)	11	10	9	7	6	4	3
Age 85–94	53 (62)	6.62 (2.61)	11	10	8	7	5	3	2
Age 90–102	23 (29)	6.30 (2.60)	11	9	8	6	4	4	3
Education: 16+ years
Age 66–69	12 (12)	9.33 (1.61)	11	11	11	9*	8	8	6
Age 66–74	47 (47)	9.11 (1.73)	11	11	11	9	8	6	6
Age 70–79	65 (65)	8.89 (1.74)	11	11	10	9	8	6	6
Age 75–84	61 (61)	8.57 (2.00)	11	11	10	9	7	6	5
Age 80–89	50 (50)	8.24 (2.21)	11	11	10	8	7	5	4
Age 85–94	23 (26)	7.61 (2.27)	11	11	9	8	6	4	4
Controls at CERAD Memory Disorders Clinics (The values below apply to 6–21 years of education; education was not associated with score in this group)
Age 50–54	12 (12)	9.17 (3.24)	11	11	11	11	8	7	0
Age 50–59	23 (23)	9.30 (2.55)	11	11	11	10	8	7	6
Age 55–64	33 (33)	8.76 (3.15)	11	11	11	10	8	6	0
Age 60–69	70 (76)	8.74 (2.94)	11	11	11	10	8	5	0
Age 65–74	88 (99)	8.51 (2.55)	11	11	11	9	7	5	3
Age 70–79	60 (70)	7.85 (2.75)	11	11	10*	8	6	4	3
Age 75–84	29 (34)	6.62 (3.68)	11	11	11	7	4	1	1
Age 80–89	11 (11)	4.73 (4.15)	11	9	9	4	1	0	0

N = number responding; number in parentheses is total group size.

(SD) = (Standard deviation)

^*indicates that score has been rounded down to the next whole number

In all samples, level of performance decreased with increase in age, and within age declined as education decreased (where education was relevant). For the Indianapolis sample only, there was further difference by sex, with women typically performing more poorly than men.

Finally, we examined the effect of prior exposure to the CP copy test using data from the 11 Indianapolis participants who had responded to the test previously. They did not differ significantly from the nonexposed group on demographic characteristics or in CP Recall performance. We were unable to confirm this in the CERAD sample since all but one person had received prior exposure.

Discussion

To the extent that sex, education, and age were relevant, the pattern of scores was comparable across the three samples (poorer performance with increased age, less education, and by women). The samples differed in the extent to which these demographic characteristics were associated with CP recall score, and in level of performance, with Cache County European Americans being consistently better. Differences in performance between European Americans and African Americans have generally been considered to reflect non-equivalence of similarly labeled years of education (Manly, Jacobs, Touradji, Small, & Stern, 2002). Other factors, such as geographic region may also be involved, since there is evidence that older community residents in the south may perform more poorly than those living elsewhere (Fillenbaum, Unverzagt, Ganguli, Welsh-Bohmer, & Heyman, 2002; Wadley, Unverzagt, McGuire, et al., 2011). Why the performance of Cache County European Americans was better than that of the CERAD clinic volunteers is unclear.

All the main comparison studies (Berres et al., 2000; Lee et al., 2004; McCurry et al., 2001), also looked at the impact of age, education, and sex on CP recall performance. All found that age, sex, and gender (and sometimes also two-way interactions among these) to be important. Berres et al. (2000), based on information from 617 cognitively normal German speaking participants from the chemical-pharmaceutical industry (53–92 years old; 432 men, 185 women; 8–20 years of education), found all three demographic characteristics to be important determinants of score on the CERAD neuropsychology measures, but do not specifically indicate whether this holds for CP recall. Lee et al. (2004), based on 618 volunteers in Korea (60–90 years old; 209 men, 409 women; 0–20 years education), found age, sex and education, and also the two way interactions between these characteristics, to be significant. The AddNeuroMed study (Paajanen et al., 2010), with information on 223 cognitively intact subjects from Finland, France, Greece, Italy, Poland and the United Kingdom, found that score varied with demographic characteristics. Other studies on CP recall tend to have smaller samples (Karrasch, Sinerva, Gronholm, Rinne, & Laine, 2005; Spangenburg, Henderson, & Wagner, 1997; Yuspeh, Vanderploeg, & Kershaw, 1998), or re-analyze a larger study already referenced above (Zehnder et al., 2009). However, we found sex differences only for the Indianapolis (African American) sample. Differences between studies may reflect unexamined cultural factors: Berres et al. (2000) used information from German speakers in the chemical-pharmaceutical industry; the Lee et al. (2004) data come from Korea; the McCurry et al. (2001) data come from a Japanese-American sample, in which nearly half were tested in Japanese and the AddNeuroMed study reported performance differences across countries..

To date, comparison of CP recall performance across studies is problematic. Depending on sample characteristics, data may be presented separately by age group (with age categories inconsistent across studies); by sex and age; by age and education (with both categories differing across studies); or by sex, education, and age. We have tried to maintain consistency in both age and education categories, using conventional groupings of each. Overall, however, older people (regardless of how grouped) perform more poorly than younger persons, women perform more poorly than men, possibly reflecting gender-related spatial abilities (see review by Herlitz, Lovén, Thilers, & Rehnman, 2010), and those with more years of education perform better than those with fewer. Within these general similarities, specific scores may differ notably. For instance (and assuming a rough education matching where such is possible), the CERAD mean CP recall score for persons age 80 and over is 4.3, but the score for the Japanese American sample is 4.9, that for the Korean sample is 5.7 (women) and 7.1 (men); for Cache County it is 7.1; and for the German-speaking sample it is 7.7 (women), and 8.3 (men). Differences are unlikely to be attributable to small sample sizes. Sample sizes were 201 (Japanese-Americans; McCurry et al., 2001); 372 (African Americans; Hendrie et al., 2001); 507 (European Americans; Welsh-Bohmer et al., 2009), 617 (German-speaking; Berres et al., 2002), and 618 (Korean; Lee et al., 2004). The reasons for score differences in older age are unclear, possibly education, environmental characteristics, and survival issues are at play. Differences, however, are unlikely to be due to the Flynn effect (Flynn 1987), since all studies were done within a decade of each other, and all persons were quite elderly. The score differences suggest caution in using norms developed in a setting other than that for which comparison is desired.

The additional value of CP recall is not yet known. CP copy has been found to discriminate particularly in the later stages of AD (Wang et al., 2004; Welsh, Butters, Hughes, Mohs, & Heyman, 1992; Zec et al., 1992). There has been interest in using CP tasks to identify pathology in specific areas of the brain (Nielsen et al., 1996), to distinguish late onset depression from AD (Künig et al., 2006), and to examine the impact of vitamins and homocysteine on cognitive function (Riggs, Spiro III, Tucker, & Rush, 1996). It remains to be seen whether CP recall can further aid in these endeavors.

This study has certain limitations. While the participants in both epidemiology-based studies can be expected to be more representative of persons their age than volunteers, the manner of selection which ensured that they were cognitively intact may have favored inclusion of persons with lower cognition. (Among others, all persons with a score below a cutpoint indicative of cognitive impairment were selected for further evaluation; some were deemed cognitively intact.) All but one CERAD subject and 11 in the Indianapolis group had been administered the CP copy task previously. While the current data focus on performance at the first administration of CP recall, these participants had seen the figures at least once a year or two before, and several had seen them on multiple prior occasions. Although prior exposure might be expected to improve performance, no improvement was observed in the present case. Unfortunately, unlike the other CERAD measures, for which 1-month test-retest data were obtained, we have no comparable information for CP recall. Volunteers at tertiary care memory disorders clinics have been found to perform better on neuropsychology measures than the general population their age (Fillenbaum, et al., 2002), although that does not seem to hold here.

The norms presented here are based on data gathered by uniformly trained testers from carefully evaluated cognitively normal epidemiologically-based and volunteer-based subjects. The extent to which they apply to persons in other countries, with different educational systems and gender expectations is unclear. Similarly, it is appropriate to be cautious in generalizing to other geographic areas in the U.S., which may have different emphases on education. These limitations notwithstanding, present data complete the norms developed for the CERAD Neuropsychology Battery based on the original CERAD sample of control subjects, and provide age-consistent information on a broad variety of older persons, particularly persons 80 years of age and older. They also provide norms useful in determining the level of performance of older African Americans (by comparing with Indianapolis data), and European Americans (by comparing with Cache county data) who are evaluated for either epidemiological or clinical purposes, e.g., to ascertain whether intervention is desirable to improve status or prevent decline, to determine eligibility for clinical studies. In making such comparison, however, it is critical to bear in mind that these data are not nationally representative, and may reflect unknown effects that are peculiar to each site. The CERAD data indicate the levels of performance that may be expected of European American volunteers to memory disorders clinics, a group frequently used as control subjects in studies of patients at memory disorders clinics.