Archivesof Clinical Neuropsychology,Vol. 10, No. 1, pp.

47-56, 1995
Pergamon Copyright© 1994 NationalAcademyof Neuropsychology
Printed in the USA. All rights reserved
0887-6177/95 $9.50 + .00
0887-6177(93)E0003-T

The Rey-Osterreith Complex Figure Test:

Norms for Young Adolescents and an
Examination of Validity
Richie G. Poulton I and Terrie E. Moffitt 2

l Clinical Research Unit for Anxiety Disorders, University of New South Wales at St.
Vincent's Hospital, Sydney, Australia, and
2University of Wisconsin, Madison, Wl

Although recent studies have investigated theoretically relevant aspects of perfor-

mance and psychometric properties of the Rey-Osterreith Complex Figure Test
(ROCFT), there is no documentation of the instrument's construct and predictive
validity in an unselected, nonclinical population. This is necessary because
knowledge o f base rates o f p o o r performance in the general population is
required to evaluate the significance o f performance deficits in clinical popula-
tions. The R O C F T was administered to 740 children aged 13 years who are
members of an unselected birth cohort, representative of the general population.
Normative findings are presented. Correlational analysis indicated that perfor-
mance on the R O C F T was closely related to performance on the Block Design
and Object Assembly subtests o f the WISC-R. The first documentation was
obtained of the ROCFT's predictive validity. Subjects in an unselected nonclini-
cal sample who had a history o f central nervous system health problems scored
below the rest of the sample. Implications f o r the continued use of this instru-
ment are discussed.

This work was supported by U.S. Public Health Service Grants from the Violence and
Traumatic Stress Branch (MH-45070 and MH-45548 to T. Moffitt) of the National
Institute of Mental Health. Richie Poulton is supported by an Australian Postgraduate
Research Award. The Dunedin Multidisciplinary Health and Development Research
Unit is supported by the New Z e a l a n d Health Research Council. We t h a n k Mrs.
Kathleen Campbell for her assistance with this research. We are grateful to the mem-
bers of our longitudinal sample, and to their families.
Address correspondence to: Richie Poulton, Clinical Research Unit for Anxiety
Disorders, University of New South Wales at St. Vincents Hospital, Sydney, Australia.
47
48 R. G. Poulton and T. E. Moffitt

The Rey-Osterreith Complex Figure Test (ROCFT) is an instrument widely

used by clinicians to investigate visual perceptual organization and visual
memory in brain damaged patients (Lezak, 1983). In the recent literature,
studies have appeared that have provided normative data for children's copy
and recall of the figure; a reliable and valid method for quantifying goodness
of organization and style (Waber & Holmes, 1985, 1986), and reliable scoring
criteria (Loring, Martin, Meador, & Lee, 1990).
In their large normative study, Waber and Holmes (1985) found that nearly
total accuracy of copy was achieved by age 9, with little change thereafter. They
found that left and right differences elicited by horizontal administration, as
Osterreith (1944) had presented it, were of considerable developmental signifi-
cance. By age 8, 64% of the children executed the copy from left to right. This
proportion increased to 80% between ages 9 and 12 and by age 13, reached 90%.
Direction of execution was not systematically related to handedness, although in
every age group, right-handed children produced better organised designs than
did left-handed children. This study will describe the normative performance of
an unselected sample of 13 year-old children with special regard to providing
replication of the descriptive results of Waber and Holmes (1985). In particular,
differences in performance due to gender or handedness will be examined.
In one of the few studies examining construct validity of the ROCFT,
Wood, Ebert, and Kinsbourne (1982) examined the pattern of intercorrelation
among tests of visual spatial ability in a clinical population. In their sample,
delayed recall performance on the ROCFT was strongly related to intellectu-
al skills unrelated to memory. The Block Design subtest of the WAIS, gener-
ally recognised as the best measure of visuospatial organisation in the
Wechsler scales (Lezak, 1983, p. 279), accounted for a significant proportion
of the variance on this purported memory test. These findings emphasise the
need to consider cognitive strategies and the way in which they operate on
memory in any theoretical or practical understanding of memory deficits in
clinical populations.
A construct validity study is therefore required to examine the relationship
between performance on the ROCFF and other theoretically similar measures
of cognitive function in an unselected nonclinical population. The determina-
tion of such a relationship in this sample will provide a better understanding of
the cognitive abilities measured by this test and provide normative data
against which productions by brain damaged persons can be evaluated.
Because the ROCFT measures visuo-spatial, constructional functions, and
nonverbal memory (Lezak, 1983), it should be more strongly related to the
performance scale of the WISC-R, which measures similar constructs, than to
the Verbal scale of the WISC-R which is concerned primarily with verbal abil-
ities. It was hypothesised that Performance IQ (PIQ) would correlate more
strongly than Verbal IQ (VIQ) with both copy and recall scores.
In clinical populations it has been found that for many brain damaged indi-
viduals the ROCFT is too complex to appreciate in its entirety. In order to
 Norms and Validity of the Complex Figure 49

draw the figure, these persons must break it down into its components, thereby
simplifying the task by adopting a piecemeal approach with more opportunity
for performance errors. Despite the popularity and widespread use of the
ROCFT for over 50 years, there are no reports investigating the instrument's
predictive validity for differentiating children with a history of central nervous
system (CNS) health problems and normal children in the general population.
This may, in part, be explained by the methodological difficulties inherent in
attempting this type of validation study.
The present study attempts to overcome these difficulties by using a com-
plete birth cohort of children taking part in a longitudinal health study. It
will provide the first documentation of the ROCFT's predictive validity
using an unselected normal sample. It was hypothesised that subjects who
have a history of CNS health problems will produce defective copy and
recall productions compared to normals. That is, the CNS health problem
group will have significantly lower copy scores and/or significantly lower
recall scores.

METHOD

Subjects

Sampling and research design. We studied a complete birth cohort of ado-

lescents involved in the Dunedin Multidisciplinary Health and Development
Study. The cohort's history has been described in detail by Silva (1990). Briefly,
the study is a longitudinal investigation of the health, development, and
behavior of a complete cohort of all consecutive births between April 1, 1972
and March 31, 1973, in the city of Dunedin, New Zealand. At the time of birth
perinatal data were obtained. When the children were traced for follow-up at 3
years of age, 1,139 children were deemed eligible for inclusion in the
longitudinal study because they were still living in the province. Of these,
1,037 (91%, 501 girls and 536 boys) were assessed, forming the sample for
the longitudinal study.
The sample has been reassessed with a diverse battery of psychological,
medical, and sociological measures every 2 years since the children were age
3. Data were collected for 991 subjects at age 5, 954 at age 7, 955 at age 9,
925 at age 11,850 at age 13,976 at age 15, and 1008 at age 18 in 1990-1991.
The neuropsychological tests reported in this article were administered during
the age-13 assessment, which took place within 60 days of each child's 13th
birthday. Of the participants at 13, 740 children received the ROCFI" under
standardized testing conditions at the research unit laboratory, but 110 chil-
dren had missing data for the ROCFT because they lived too far away to come
to the unit. Elsewhere we have published evidence that the effects of missing
data at each wave of data collection do not bias the representativeness of this
50 R. G. Poulton and T. E. Moffitt

sample for studies of IQ, neuropsychological test scores, or socio-economic

status (Frost, Moffitt, & McGee, 1989; Silva, 1990).
The New Zealand cohort constitutes a representative genera[ sampling
design, and thus may be used to calculate estimates of population norms. With
regard to social origins, the children's fathers were representative of the social
class distribution in the general population of similar age in New Zealand.
Regarding racial distribution, members of the sample are predominately of
European ancestry (less than 7% identified themselves as culturally or ethnic-
ally Maori or Polynesian at age 18).

Procedure

Before the teenagers came to the Dunedin Unit for assessment, their par-
ents were asked to complete a detailed questionnaire concerning family back-
ground, and health and behavioral reports. Details were provided to the par-
ents outlining the nature of the assessment and the parents were asked to sign
a consent form. Each teenager was assessed over the course of 1 day with a 6
hour protocol of health, cognitive, educational, attitudinal, behavioral, and
emotional measures. No child had taken the ROCFT before.
All of the performance measures used in the present study were adminis-
tered in the morning, in 4 50-minute sessions that were counterbalanced in
order and separated by 10 minute breaks. Each child was tested by one of
three testers trained in psychological assessment who presented the task to the
subjects according to a preestablished protocol. All response forms were
scored by TM, who was blind to the subjects' identity.
The ROCFT was administered as part of a research battery of neuropsy-
chological tests which also included the Grooved Pegboard, the WISC-R
Mazes, the Rey Auditory Verbal Learning Test, the Trail Making Test, the
Controlled Word Association Test, and the Wisconsin Card Sort test. A factor
analysis of all the neuropsychological tests administered produced five fac-
tors labelled Verbal, Visuospatial, Verbal Memory, Visual Motor Integra-
tion, and Mental Flexibility (Frost, Moffit, & McGee, 1989). The ROCFT
copy score and delayed recall score loaded highly on the visuospatial
factor (r = 0.73 and r = 0.66 respectively), and neither ROCFT score loaded
> r = 0.20 on any other factor.
The time delay between Copy and Recall administrations of the ROCFT
was determined by the minutes that elapsed between the subject finishing the
ROCFT copy, completing an intervening task (Grooved Pegboard Test) and
being asked to draw the ROCFT figure from memory. The delay period
ranged from 2 to 9 minutes (68% of subjects' delay was 4 minutes, 98% of
subjects' delay was between 3 and 5 minutes inclusive. Previous research has
shown that, within the limits of an hour or so, the length of the delay is appar-
ently of little consequence for ROCFT memory scores (Wood, Ebert, and
Kinsbourne, 1982).
 Norms and Validity of the Complex Figure 51

Instruments and Scoring

Rey-Osterreith complex figure accuracy score (range 0-36, mean 32,

standard deviation 4.2). Following the instructions given by Osterreith (1944,
p. 2 2 1 - 2 2 2 , also see Lezak, 1983, p. 400), the figure was broken into 18
elements and each element was scored for both accuracy of formation and for
spatial location on the figure. Since the reproduction of each unit can earn as
many as 2 score points the m a x i m u m score was 36, with a higher score
indicating greater accuracy. In the event of an element receiving a one half
point (that is, it was malformed and in the wrong place), the final score was
rounded down to the nearest number left of the decimal place for ease of
computerised data coding. On the copy production, scores were skewed in the
direction o f the m a x i m u m score. A log t r a n s f o r m a t i o n was conducted,
therefore, to correct for skew before further statistical analysis.

Delayed Rey-Osterreith recall score. Scored as for copy (see accuracy score
above): range 0-36, mean 19.2, standard deviation 6.7.

Number of omissions at copy and recall. This was simply a count o f the
number of elements from Osterreith's original 18 omitted from the subject's
copy and recall performance. For each item omitted, two score points were
deducted from the overall score.

Number of distortions at copy and recall. A count of items not omitted, but
distorted on the copy and recall administrations was made. Distortions were
elements that had been scored as 1 or 1.5 using Osterreith's (1944) scoring
protocol. That is, either form or location distortions entered these counts.

Intelligence Quotient. The W I S C - R was administered at age 13 according to

its standard protocol with two exceptions: Two of the 10 subtests were omitted
because of time constraints, then scores were prorated according to the test
manual (Wechsler, 1974). Certain information subtest items that ask about
American history in the American version of the W l S C - R have been modified
for local relevance in the New Zealand version. Descriptive statistics for the
New Zealand sample were: Full Scale IQ (M = 108.2, SD = 15.1), Verbal IQ
(M = 104.2, SD = 14.9), and Performance IQ (M = 111.0, SD = 15.4). The
New Zealand mean for the W I S C - R is somewhat higher than the mean of the
American standardization sample (100) which is generally attributed to greater
racial homogeneity in New Zealand than in the USA. The standard deviation
in the two countries is the same (15).

Grooved pegboard. This was measured using the Grooved Pegboard Test. This
task m e a s u r e s m o t o r speed, c o o r d i n a t i o n , steadiness, and v i s u o m o t o r
52 R. G. Poulton and T. E. Moffitt

integration. It consists of a small board containing a 5 x 5 set of slotted holes

angled in different directions. The subject is timed while inserting notched
pegs into the board with each hand.

Central nervous system health problems. Repeated medical examinations of

the c h i l d and p a r e n t a l r e p o r t s o f the c h i l d ' s h e a l t h w e r e o b t a i n e d
systematically at every biyearly assessment. Medical record searches were
also conducted for every child. These records provided a complete health
record for the first 13 years of each child's life. Using these archives, cases
were selected by the second author based on a criterion of clear record of
neurological diagnosis or symptomatology by at least one physician. In all
cases, the health problems were reported by more than one source or at more
than one assessment age. Because the few cases of discrete syndromes could
not support independent analysis, these cases were combined to form a single
dichotomously-coded variable. Nineteen children had experienced relatively
severe head injuries; for example, one girl struck by a motor vehicle at age
five suffered a residual left side hemiplegia, speech disorder, disinhibited
social behavior, and grand real seizures. Congenital syndromes having
neurological symptoms, such as Cerebral Palsy, Down's Syndrome, and
cortical blindness, characterized 17 children. Seizure disorders that required
medication (febrile seizures were excluded) were recorded in 15 medical
histories. Ten children were known to evidence obvious hemiplegia or
strongly l a t e r a l i z e d n e u r o l o g i c a l findings. M a j o r gross n e u r o l o g i c a l
abnormalities assessed in early childhood characterized 14 children. Seven
children were under the care of a physician for severe, longstanding migraine
headaches. The cumulative number of subjects who had a record of significant
neurological health problems who had ROCF'I" test scores was 51. Many of
these cases met more than one criterion for inclusion such as the case
previously described with severe head injury, hemiplegia, and seizure
disorder. Not all of the impaired children received the Rey-Osterreith because
some of the most severely handicapped children had become inaccessible for
testing by age 13. At least one institutionalized child was untestable because
of severe behavioral/neurological problems.

RESULTS

Table 1 shows the means and standard deviations for the girls and boys, and
total sample on the Rey-Osterreith variables. T-tests revealed two statistically
significant differences between boys and girls on these measures. Girls produced
more omissions than boys and their overall recall score was marginally lower.
However, because the differences were negligible and not clinically significant,
it was decided to treat the boys and girls as one group for further analyses.
 Norms and Validity of the Complex Figure 53

TABLE 1
Means and Standard Deviations for the Rey-Osterreith Measures

Whole Sample Boys Girls

N = 740 N = 383 N = 357

Me~ SD Me~ SD Mean SD t

Copy Score (T) 32.41 4.18 32.43 3.97 32.40 4.40 -0.10
Recall Score (T) 19.24 6.71 19.89 6.49 18.54 6.87 -2.75*
Copy Omissions 0.36 0.85 0.37 0.72 0.34 0197 0.56
Copy Distortions 2.78 2.83 2.70 2.70 2.85 2.93 0.74
Recall Omissions 5.34 2.78 5.02 2.65 5.85 2.94 3.24*
Recall Distortions 5.41 2.65 5.43 2.78 5.39 2.50 -0.20

*p < 0.05.
T = Due to skewed distributions these variables underwent a log transformation for the purpose of

F o r the total sample the Copy score had a possible range o f 0 - 3 6 , a mean
o f 32.4 and a standard deviation of 4.2. The mean Recall score was 19.2 with
a standard deviation o f 6.7. Copy distortions were the major reason for the
deduction o f points on the Copy trial while both copy distortions and omis-
sions contributed to the Recall performance score.
A t-test ( p < 0.05) was conducted to determine whether handedness had an
effect on Copy and Recall scores. A nonsignificant result was obtained, indi-
cating that handedness did not have an effect on the accuracy o f the copy (RH
mean 32.49, standard deviation 4.0; L H mean 32.13, standard deviation 3.52)
or delayed recall productions of the figure (RH mean 19.29, standard devia-
tion 6.71; L H mean 18.72, standard deviation 6.87). It was found that 88% o f
the sample executed the copy production beginning on the left side. That is,
12% o f the subjects began their drawings right o f the midline.
A correlation analysis was conducted between the ROCb-T measures and a
series o f cognitive variables including the VIQ and PIQ scales o f the W l S C - R
and the G r o o v e d Pegboard measure o f visual m o t o r integration. As can be
seen in Table 2, the C o p y score was m o d e r a t e l y c o r r e l a t e d with the V I Q
(r = 0.37) but strongly correlated with the PIQ (r = 0.49). A similar pattern of
relationship was observed between the recall score and V I Q (r = 0.32) and
PIQ (r = 0.43), respectively. A more detailed analysis o f the Performance IQ
subtests revealed that Copy and Recall Omission and Distortion scores were
most strongly related to the Object A s s e m b l y and Block Design among the
subtests o f the W l S C - R . The results indicate that subjects who p e r f o r m e d
poorly on the R O C F T also performed poorly on the Block Design and Object
A s s e m b l y tasks o f the W l S C - R . Mild correlations were found between Copy
and Recall performance and the Grooved Pegboard Test.
A t-test ( p < 0.05) was c o n d u c t e d to c o m p a r e the p e r f o r m a n c e on the
R e y - O s t e r r e i t h variables o f children with a history o f CNS health problems
and those without such a history. As Table 3 indicates, there were significant
54 R. G. Poulton and T. E. Moffitt

TABLE 2
Matrix of Correlations Between ROCFT Measures and Cognitive Variables

Grooved Block Object

Pegboard VIQ PIQ Design Assembly

Total Copy Score 0.37 0.49

Total Recall Score 0.32 0.43
Copy Omissions 0.31 -0.26 -0.34 -0.38 -0.25
Copy Distortions 0.28 -0.30 -0.43 -0.48 -0.38
Recall Omissions 0.25 -0.25 -0.32 -0.33 -0.30
Recall Distortions 0.11 -0.18 -0.26 -0.28 -0.23

Note. A correlation in excess of 0.12 is significant atp < 0.001.

differences for the C o p y and R e c a l l scores, C o p y Distortions, and Recall

Omission measures between the two groups. Those with a history o f CNS
health problems obtained lower scores on their Copy and Recall productions
and greater C o p y Distortions and Recall Omission scores. Other measures did
not show significant differences.

DISCUSSION

The R e y - O s t e r r e i t h C o m p l e x Figure Test is purported to measure visual

spatial organization and long-term figural memory. Waber and Holmes (1985)
found that right-handed children produced better organised designs than did
left-handed children. Specifically, left-handed children produced more part-
oriented designs implying p o o r strategy (not using the base rectangle as an
organizational unit). The present study found that handedness was not related
to copy or recall performance, which does not support previous findings of
differences in accuracy due to handedness (Waber & Holmes, 1985). This dis-

TABLE 3
T-Test of the Rey-Osterreith Measures with History of Central
Nervous System Health Problems

Group I Group IV
Normal CNS Health Problems
Variables N= 625 N= 51

Copy Score 32.61 27.38*

Recall Score 19.47 13.29"
Copy Omissions 0.33 0.93
Copy Distortions 2.69 5.14"
Recall Omissions 5.27 7.50*
Recall Distortions 5.36 6.21

*p < 0.05.
Note. Nontransformed scores are reported in this table.
 Norms and Validity of the Complex Figure 55

crepancy may be due to the different measures of performance that were used
in these two studies. Waber and Holmes developed a more complex and
detailed measure of organisation whereas in the present study, performance
was determined by the use of the original (and most widely-used) scoring sys-
tem (Lezak, 1983). Consistent with Waber and Holmes (1985), we found that
by age 13, approximately 90% of the children executed their copy from left to
right, supporting their contention that directional preferences can be viewed as
diagnostically significant only at age nine or after.
By investigating its relationship with other cognitive measures, the figure's
construct validity was examined. The hypothesized link between the ROCFI"
measures and the PIQ score was supported. A strong relationship emerged
(r = 0.49 for Copy and r = 0.43 for Recall) which warranted further investiga-
tion. It was found that the Copy and Recall Omission and Distortion scores were
most strongly related to the Block Design and Object Assembly subtests of the
WlSC-R. They were not significantly related to the Picture Completion or
Coding subtests. The Block Design subtest is generally regarded as a good mea-
sure of visuospatial organization (Lezak, 1983) and the Object Assembly subtest
emphasizes the ability to form visual concepts quickly and translate them into
rapid hand responses. The results obtained suggest that the ROCFT was measur-
ing a general visuo-practic factor and was strongly related to the nonverbal cog-
nitive functions most similar in structure to those measured by the Block Design
and Object Assembly Subtests. This association may be partly explained by the
role of strategy (visuo-spatial organisation and planning) required for successful
completion of both tests. Importantly however, the ROCb-T provides a measure
of nonverbal memory for complex stimuli, unlike the WISC-R subtests, thereby
eliciting unique information about these cognitive processes.
Mild correlations on copy and recall performance with the Grooved
Pegboard measure of visual motor integration were also obtained providing
further evidence for a general visuo-practice ability underlying performance
on the ROCFI'.

CENTRAL NERVOUS SYSTEM HEALTH PROBLEMS

In order to investigate if and how performances differed depending on the

functional integrity of the brain, a comparison was made between members of
the sample with and without a documented history of CNS health problems. It
was hypothesized that subjects who had a history of CNS health problems
would produce defective copy and recall productions compared to subjects
without such a history, that is, CNS health problems will result in a signifi-
cantly lower Copy and Recall score.
The results revealed that the heterogeneous group with a history of severe
CNS health problems obtained significantly lower scores than the remainder
of the sample. Specifically, this group of subjects with a positive clinical
56 R. G. Poulton and T. E. Moffitt

history made significantly more distortions when copying the figure and sig-
nificantly more omissions when reproducing the figure from memory, com-
pared to the rest of the children. This finding lends support to Lhermitte and
Signoret (1972) and Brooks (1972) who attributed poor recall performance
among brain damaged subjects to inefficient encoding of the information ini-
tially and importantly, provides the first documentation of the predictive valid-
ity of the ROCFT; the CNS-disordered members of an unselected general
sample scored significantly worse. Of course, the ROCFT alone would not be
used by responsible clinicians to infer a presumptive diagnosis of CNS disor-
der, but this study suggests the ROCFT can play a very useful role as one
component of a diagnostic test battery.

REFERENCES

Brooks, D. N. (1972). Memory and head injury. Journal of Nervous and Mental Disease, 155,
350-355.
Frost, A., Moffit, T., & McGee, R. (1989). Neuropsychological correlates of psychopathology in
an unselected cohort of young adolescents. Journal of Abnormal Psychology, 98, 307-313.
Lhermitte, E, & Signoret, J.-L. (1972). Analyse neuropsychologique et differenciation des syn-
dromes arrmesiques. Revue Neurologique, 126, 164-178.
Loring, D. W., Martin, R. C., Meador, K. J., & Lee, G. P. (1990). Psychometric construction of the
Rey-Osterreith Complex Figure Test: Methodological considerations and interrater reliability.
Archives of Clinical Neuropsychology, 5, 1-14.
Lezak, M. D. (1983). Neuropsychological assessment. New York: Oxford University Press.
Osterreith, P. (1944). Le test de cope d'une figure complexe. Les Archives de Psychologie, 31,
206-356.
Silva, P. A. (1990). The Dunedin multidisciplinary health and development study: A fifteen year
longitudinal study. Paediatric and Perinatal Epidemiology, 4, 96-127.
Waber, D. P., & Holmes, J. (1985). Assessing children's copy productions of the Rey-Osterreith
Complex Figure. Journal of Clinical and Experimental Neuropsychology, 7(3), 264-280.
Waber, D. P., & Holmes, J. (1986). Assessing childrens' memory productions of the Rey-
Osterreith Complex Figure. Journal of Clinical and Experimental Neuropsychology, 8(5),
563-580.
Wechlser, D. (1974). Manual of the Wechsler Intelligence Scale for Children-Revised. New York:
Psychological Corporation.
Wood, E B., Ebert, V., & Kinsboume, M. (1982). The episodic-semantic memory distinction in
memory and amnesia: Clinical and experimental observations. In L. Cermack (Ed.), Memory
and amnesia. Hillsdale, NJ: Lawrence Erlbaum Association.