Rey-Osterrieth Complex Figure Test
Abbreviations: ROCF, RCFT, CFT, Rey Figure
Background
The Geneva psychologist André Rey first described, in 1941, investigations using a complex figure that had to be copied first and then reproduced from memory1). The aim was a closer qualification of perceptual and memory disorders in neurological patients. The figure itself was printed in the publication, and Rey also gave instructions for quantifying errors in the reproduction from memory. His doctoral student at the time, Paul-Alexander Osterrieth, carried out further investigations with the complex figure and developed a somewhat simplified quantification method for the errors in both the initial copying and the reproduction from memory2). A more easily accessible English translation of the parts of both publications relevant to the test appeared in 1993 in The Clinical Neuropsychologist3).
Like Rey's auditory-verbal memory test, his complex figure also remains an important part of the clinical neuropsychological toolkit. It is discussed in all textbooks and handbooks of neuropsychological assessment4) 5) 6).
More recently, a widely marketed version of the RCFT has appeared that contains a detailed scoring system and can draw on a relatively large US normative sample from 19957). It can also be obtained via the German-language Testzentralen (test distributors).
Test material and copyright
Unfortunately, the Rey Figure apparently enjoys hardly any copyright protection. Entering “Rey complexe figure” into the image search of a search engine yields, among other things, an abundance of images of the original figure. On the one hand, this is bad for test security. On the other hand, author's rights still exist: André Rey, the creator of the figure, died in 1965, so the rights do not expire until 2035. It is therefore not advisable to print any of the images found on the internet and use it as a template. Meyers & Meyers8) and PAR, as publishers of their version of the RCFT, acquired the (American) rights to print the figure from the descendants of André Rey, as can be seen from the copyright statements on the back of the figure (see also the discussion page of the English Wikipedia article on the RCFT). If one does not want to go the same way and start a correspondence with the descendants of André Rey, purchasing a publisher's version is recommended. Because of the detailed scoring instructions in Meyers & Meyers9), this version is the obvious choice, even though the manual is available only in English. I am not aware of a copyright-safe German version.
Test administration
The most common way of administering the RCFT is the one described by Rey in 1941, which was also essentially used by Meyers & Meyers in the norming study. Among other recommendations, that of the diagnostics commission of the former Swiss Association for Career Guidance from 200210) stands out because it provided a detailed description of the procedure in German.
Testing starts with the complex figure being presented (A4 landscape) and the examinee being asked to copy it. Different colors can be used for this (Rey's recommendation, not that of Meyers & Meyers) so that the order in which the individual components of the figure were drawn can be reconstructed later (about 5 different ones are sufficient). The examinee must not be told that the figure will later have to be drawn from memory. Scoring the copy yields the raw score Copy, which we use accordingly.
After a maximum of 3 minutes, which can be filled with verbal tests, the examinee is asked to redraw the figure from memory. In Meyers & Meyers this trial is called Immediate Recall, and we use the same term. Different colors can be used here as well. After a further 30 minutes, again filled with verbal tasks, the figure is drawn from memory a second time (Delayed Recall).
In Meyers & Meyers, a recognition task follows, in which individual components of the figure must be correctly identified among similar ones. This subtest was not included in TDB2Online.
Test raw scores
Scoring essentially consists of quantifying the accuracy of reproduction, as worked out by Osterrieth in 1944. For this purpose, the figure is divided into 18 components, and points between 0 and 2 are awarded for the accuracy and placement of the components. The maximum score is thus 36.
The following raw scores, each between 0 and 36, result from this:
- Copy
- Immediate Recall
- Delayed Recall
In Meyers & Meyers (and also in other authors), further raw scores are collected for the copy time and for an additional recognition trial. These raw scores are not evaluated in TDB2Online.
Norming studies
To create the TDB2Online performance norms for the Rey Figure, systematic use was made of as many of the accessible studies with data from healthy adults as possible. Data from different cultural areas were initially also included. A condition for including a study was that means and standard deviations were reported for as many as possible of the three raw scores included in TDB2Online, for a sample of healthy examinees in several differentiated age ranges.
The included studies are summarized with their most important characteristics in Table 1. In total, the data of 11 studies were considered, seven from the USA, one each from Spain, Ireland, and Mexico, and a further one that reports data from 11 Latin American countries individually. For five papers the norms are available in the form of percentiles or standard scores (T scores, scaled scores), for the remaining six only in the form of means and standard deviations. Five papers included persons over a wide age range starting from young adulthood, the other six only from 30, 45, 49, 57, or only from 70 years. The educational level of the included persons varies considerably. In the US studies, the mean years of education were mostly between 14 and 16 years, in the other four studies mostly below. Not all studies reported norms for all three subtests treated here.
The test instructions were similar enough that they can be considered together. The immediate recall trials began between 20 seconds and three minutes after the end of copying, the delayed recall trials between 20 and 40 minutes after the end of copying. In all studies, the delayed recall trials were preceded by immediate recall trials. Thus, no studies were considered in which the first reproduction of the figure took place only after 30 minutes.
Of the studies considered here, 5 are reviewed in Mitrushina et al. (2005)11).
Table 1: Overview of the norming studies on the ROCF
| First author | Year | Age range | N | Education | Statistics | Subtests | Country |
|---|---|---|---|---|---|---|---|
| Meyers12) | 2005 | 18–89 | 601 | 14? | Percentiles | C, IR, DR | USA |
| Fastenau13) | 1999 | 30–85 | 211 | 15 | Scaled scores | C, IR, DR | USA |
| Machulda14) | 2007 | 70–89 | 421 | ≥ 12 | Scaled scores | C | USA |
| Pena-Cas.15) | 2009 | > 49 | 332 | 6–16 | Scaled scores | C, IR, DR | Spain |
| Rivera16) | 2015 | 18–95 | 3968 | 6–16 | Percentiles | C,IR | Latin America |
| Van Gorp17) | 1990 | 57–85 | 156 | 14 | M, s | C, IR | USA |
| Boone18) | 1993 | 45–83 | 91 | 14.5 | M, s | C, IR | USA |
| Chiulli19) | 1995 | 70–93 | 153 | 15 | M, s | C, IR, DR | USA |
| Miller20) | 2003 | 25–59 | 729 | 16 | M, s | C, IR, DR | USA |
| Gallagher21) | 2007 | 16–69 | 117 | Census | M, s | C, IR, DR | Ireland |
| Ostrosky-S.22) | 1998 | 20–89 | 105 | ≥ 6 | M, s | C, IR, DR | Mexico |
Distributional properties of the raw scores
From the table of norming studies for the Rey Figure (Table 1), one can see that the first five studies listed there report percentiles or scaled scores. One should therefore think that there are enough studies from which unprocessed distributions of the raw scores for individual age groups can be reconstructed. Unfortunately, this is not so. The test manual by Meyers & Meyers23) does report, on the basis of a sufficiently large sample per age group, sufficiently finely graded T scores (and associated percentiles) between 20 and 80, but on the basis of a distribution previously generated by regression analysis (“Continuous Norming”24)). Rivera et al. proceeded in a very similar way. That leaves the studies by Fastenau et al., Machulda et al., and Pena-Casanova et al. Here, too, there are limitations: first, none of the three papers examines young adults, and second, the results are reported relatively coarsely, in the form of scaled scores between 1 and 18.
For the transformation of raw scores into performance scores, we will therefore fall back on the linear transformation in Meyers & Meyers, but want to use the three other studies to make sure that a linear transformation is justified. Figure 1 shows the conversion of raw scores into IQ standard scores for the immediate recall of the figure. In green shades, the age groups 20-24, 25-29, 45-49, and 80-89 of the Meyers & Meyers sample are plotted in the figure. As can be seen, these are almost exclusively linearly transformed by the “Continuous Norming” method. The data of the study by Fastenau et al. are shown in red, those of Pena-Casanova et al. in blue. These are in each case pure percentile-to-normal transformations, which can represent any kind of distribution. One can see that the distributions are largely linear and show hardly any floor effects and no ceiling effects. The linear conversion by Meyers & Meyers thus poses no psychometric problem.
Figure 1: Transformation curves for the conversion of raw scores into IQ standard scores in the Immediate Recall trial in the studies by Meyers & Meyers (marked M in the legend), Fastenau et al. (F), and Pena-Casanova et al. (P) for selected age groups (see text)
The back-calculated original distributions are shown in Figure 2. A slight floor effect can be seen in the oldest age group in the Pena-Casanova sample (P80-90 in the legend); otherwise these are all nicely symmetric distributions that lend themselves well to linear processing.
Figure 2: Back-calculated raw score distributions for immediate recall in the same subsamples as in Figure 1
The raw scores of the Delayed Recall trial have the same properties as those of immediate recall. The two trials correlate with r = .88 in Meyers & Meyers' sample of examinees (N = 601), and with r = .96 in a clinical sample (N = 100).
The situation is different for the raw scores when copying the figure. Here strong ceiling effects are to be expected, since most persons make only few errors when copying. The question also arises whether quantification in TDB2Online makes sense at all. Meyers & Meyers give T scores for the Copy trial only from the 1st to the 16th percentile. The percentile rank 16 corresponds in adults under 50 years to raw scores between 33 and 34. Everything better than a raw score of 33 is considered “normal” according to this scoring logic. Other norming studies do give percentiles over the entire raw score range, but not in the age group of young adults. We therefore refrain from quantification when copying the Rey Figure.
Linearization of the measurement scale
For the two trials “Immediate Recall” and “Delayed Recall”, a simple linear transformation suffices; for the “Copy” trial, quantification is dispensed with.
Meta-analytic summary of the norms
There are no great differences in the quality of the samples listed in Table 1. None of the samples is really a random sample. All are incidental samples, often also composed of different subsamples. Only in two of the samples is a comparison with the census carried out at all. One of these is the Irish sample of Gallagher & Burke (2007)25), which, with 117 persons in the age range between 16 and 69, is however too small. The other is that of Meyers & Meyers (1995)26). Here, however, there is the problem that the census-matched sample is not divided by age and the sample divided by age does not correspond to the census.
In this respect it seems most sensible to include all samples that are suitable for developed countries in the derivation of norms. The unsuitable one is the large study by Rivera et al. (2015)27) in 11 Latin American countries. As emerges from the associated paper by Guardia-Olmos et al. (2015)28), the durations of education in 10 of the 11 countries (exception: Peru) were so far below those of all other studies that inclusion would have made no sense.
For both the immediate and the delayed recall, the raw scores can be converted linearly to standard scores. Figure 3 shows the mean raw scores of the immediate recall of the figure for all age groups for which the respective samples had values. In these and the following figures, “M&M” stands for Meyers & Meyers; the remaining abbreviations are self-explanatory. The paper by Miller (2003)29) appears three times in the figures because the values were available only for three different years of education (under 16 years, 16 years and more than 16 years).
Figure 3: Age course of the means in immediate recall
The thick black line denotes the mean across all studies with values for an age group, weighted by the respective sample size. This line was subsequently smoothed with a moving average of order 3, hence the name of the line, “M smoothed”.
Figure 4 shows the standard deviations per age group and study in the same way as before. Again, “SD smoothed” denotes the weighted mean standard deviation.
Figure 4: Age course of the standard deviations in immediate recall
Considering the two figures together, it becomes clear that the norming studies for the Rey Figure are astonishingly close to one another compared with other test procedures. With the given mean standard deviations of 5 to 6 across the entire age span, practically all studies lie within a range of -1σ to +1σ. This certainly lends weight to the norms.
Figures 5 and 6 show the same for the delayed recall of the Rey Figure.
Figure 5: Age course of the means in delayed recall
Figure 6: Age course of the standard deviations in delayed recall
As already discussed above, the two parameters correlate highly with each other. In most examinees the raw scores in both trials are very similar. This need not be the case at all, however, in patients with storage or retrieval disorders.
Overview of performance norms and age norms
The figures in this section summarize the performance norms and age norms of the two recall trials of the Rey Figure as they appear in TDB2Online.
Figure 7 shows, in a composite graph, the relationship between raw scores and performance scores as well as the (mean) influence of age on performance in immediate recall. On this and all following figures, the vertical lines mark the performance scores from 40 to 145. The raw scores are plotted to scale on the black horizontal line. The alternately red and blue lines contain the norm limits for all age groups in the overview. The five markers on each line stand for the percentile ranks 2.5, 16, 50, 84 and 97.5. One such line, namely the one corresponding to the age group of the examinee, is drawn (in a different form) in the TDB2Online profile sheet to help with the individual interpretation of the test results. Because the performance scores are calculated from the results of the 20-24-year-old young adults, the 5 markers of the age group 20-24 lie exactly on the performance scores 70, 85, 100, 115 and 130.
Figure 7: Performance scores in immediate recall of the figure (see text)
The age-associated decline in performance from young adulthood to old age amounts to almost three standard deviations. The recall trials of the Rey Figure thus belong to the most age-sensitive neuropsychological test procedures.
Figure 8 shows the same graph for delayed recall. Again, there are hardly any differences from Figure 7.
Figure 8: Performance scores in delayed recall of the figure (see text)
Discussion
The Rey Figure has been used for almost 80 years in neuropsychological assessment. Test administrators generally like it, and examinees mostly cooperate well too, once the initial shock after the reproduction instruction has been overcome. There are many clinical studies on the value of the test.
For psychometric reasons, only the two reproduction trials are quantified in TDB2Online. For the copy trial, there was no reasonable percentile distribution in the norming studies. This need not be a disadvantage. The exact score when copying is in any case less important for clinical interpretation than the way in which someone organizes the figure in perceiving and drawing it. It is up to the report writer to communicate the result of the copy trial verbally in the report.