Benton Visual Retention Test
Abbreviation: BVRT
Background
In 1945, Arthur Benton described in a journal article1) a simple, clinically applicable visual memory test, which was published a year later under the same title as a test by the Psychological Corporation. In this test, the examinee is to memorize simple visual designs after a brief viewing time and to draw them from memory immediately afterward. Benton developed this test as a complement to the verbal memory tests that had already existed for a long time. The procedure quickly gained international distribution. In 1955, a revised version with parallel series of designs appeared. This was the basis for the first German version2), edited by Otfried Spreen, which was published under the simplified name Benton Test and is occasionally abbreviated simply as BT. The eighth German edition is now available3). In the more recent editions, practically nothing has changed in the basic form of the test; the manuals have mainly incorporated more recent normative and validity data. However, a consolidated synthesis of the various norms has never been attempted in the manuals.
Test material
The copyright for the German adaptation of the procedure is held by Hogrefe. The manual and test material are available through the Testzentrale (test distributor) in Germany or Switzerland.
The German test kit contains three design booklets. Among other things, they contain three design series of 10 drawings each. These three design series are labeled C, D, and E. Series C is the most frequently used design series. Series D and E are normally used only for repeated measurements.
In addition to this classic test material, the German test kit contains designs that can be used for recognition (instead of drawing) of the figures, as well as extended designs with 20 drawings each. The normative situation for these extensions is clearly worse than for the original version. They are therefore not included in TDB2Online.
Test administration
The standard test instructions for administration are those designated in the manual as Instruction A. They provide for a presentation time of 10 seconds; immediately after the presentation, the drawing is to be reproduced from memory.
Since the 1950s, there have been many variants in addition to the standard instructions. In Instruction B, the presentation time is reduced to 5 seconds. Instruction C is not a memory test but only requires copying the designs. Instruction D inserts a pause of 15 seconds between presentation and reproduction. Most norming studies refer exclusively to Form C and Instruction A. The additional variants can be highly useful in clinical application (for example, to test the drawing ability of individuals independently of visual memory). However, their norms are too weak for a psychometric treatment.
Test raw scores
Scoring the Benton Test requires careful examination of the reproductions. For this, one must be familiar with the scoring guidelines and have a certain amount of experience with the test. There is one raw score each for the “Number Correct”, that is, the number of designs reproduced without error, and the “Errors”, the number of errors made while drawing (several errors can be made on one design). In TDB2Online, only these two raw scores are used.
Distributional properties of the raw scores
The processing of the Benton norms is based on the norming studies analyzed in Mitrushina et al. (2005)4), not on a literature review going beyond that. As far as could be determined from the studies reported there, no studies reported percentile distributions. Of course, assumptions about the distribution shapes can easily be made: If the expected number of correct reproductions in young adults is around 8 (with a raw score range of 0 to 10), then the raw score distribution will be strongly skewed because it hits the upper ceiling. However, we have no real data on this and can therefore only calculate with the reported means and standard deviations. A linearization of the measurement scales for Number Correct and Errors is therefore not possible.
Norming studies
To create the TDB2Online performance norms, we drew on readily accessible studies that contained data from healthy examinees, regardless of the place of investigation. A comparison across language boundaries is relatively unproblematic for a visual test such as the Benton Test. We limited the search to design series C and standard instruction A (10 seconds of presentation, reproduction immediately afterward). In addition to the German manual for the Benton Test (I had the seventh edition of 1996 available)5), the collection of normative data for standard neuropsychological procedures by Mitrushina et al., 20056) was reviewed. Table 1 shows the most important characteristics of the included studies.
Table 1: Overview of the norming studies pooled by meta-analysis
| No. | First author | Year | Age range | N | Education | Country |
|---|---|---|---|---|---|---|
| 1 | Sivan7) | 1996 | 15-69 | approx. 1,500 | n/r | USA |
| 2 | Steck8) | 1990 | 15-74 | 252 | n/r | D |
| 3 | Youngjohn9) | 1993 | 18-84 | 1,121 | 16 years | USA |
| 4 | Robertson-Tchabo10) | 1989 | 20-89 | 1,643 | well-educated | USA |
| 5 | Klonoff11) | 1965 | 80-92 | 172 | 7 years | USA |
| 6 | Prakash12) | 1992 | 15-64 | 660 | n/r | India |
| 7 | Coman13) | 2002 | 55-89 | 156 | 13 years | USA |
It is evident that many of the studies are quite old, going back to 1965. Good information on the educational level of the sample is also available only occasionally, and the reported average years of schooling vary considerably. However, one must also consider the temporal context: In rural America (Klonoff & Kennedy, 1965) of the 1960s, 7 years of schooling was entirely normal for very old examinees.
The studies in detail:
- The German manual reports the data of several studies, albeit very briefly and not very transparently. In the first two tables of the chapter on the norms, “expected values” are given for Correct responses (Table 4.1, page 66) and Errors (Table 4.2, page 67) for three and four age groups, respectively, and six and eight groups, respectively, with different (estimated or measured) intellectual ability. They come from three investigations: one by Benton himself with “over 600” examinees, first published in the third English edition of the test (1963)14), a second by Arenberg (1978)15) with 769 examinees, and a third by Randall et al. (1988)16) with 120 examinees. The data of these three investigations were combined by an unspecified method and led to norm tables 4.1 and 4.2. Standard deviations are not reported in the tables. Of the three original investigations underlying the tables, only Randall et al. (1988) reported standard deviations, but these are not specified by age. Without standard deviations, the data cannot be used for norming purposes. As with some other individual neuropsychological tests, the norms in the manual are frequently stratified by intellectual ability. The norms of TDB2 and TDB2Online always refer to samples that are as representative of the population as possible (more on this in the TDB2Online documentation). For this reason as well, the norms in Tables 4.1 and 4.2 cannot be used further. The next two tables of the manual (Tables 4.3 and 4.4) list the results of two further investigations:
- Steck et al. reported data of a German investigation in 199017). For the norming, the data of 252 healthy examinees aged between 15 and 74 could be used. Steck et al. (1990) used all three design sets one after the other in random order. The data were therefore simply divided into thirds.
- Youngjohn et al. (1993)18) reported norms separately for three groups with different intellectual performance. For TDB2Online, the values were averaged across these groups with weighting.
- The next four investigations are not in the manual but are listed in Mitrushina et al. (2005). Robertson-Tchabo & Arenberg published data of the Baltimore Longitudinal Study of Aging with 1,643 examinees in 198919), stratified by age, education, and sex. The data are available only for Errors. For TDB2, they were averaged with weighting across sex and education within each age group.
- Klonoff & Kennedy investigated 172 persons between 80 and 92 years of age with the Benton Test in 196520).
- Data from an Indian investigation (Prakash & Bhogle, 1992)21) are available for 660 persons between 15 and 64 years of age, but only for Number Correct, not for Errors.
- Coman et al. reported data of 156 persons aged between 55 and 89 in 200222). These data are available only for Number Correct.
Mitrushina et al. (2005) list further publications with data from the Baltimore Study23). However, these appear to be subsamples of the extensive publication by Robertson-Tchabo, which is why we omitted them here.
With the help of a meta-analytic summary, these norms were integrated and thus made relatively independent of the peculiarities of the individual studies.
Meta-analytic summary of the norms
Figure 1 shows the age courses of the Number Correct for the five studies that provided data for this. The thicker line represents the weighted means of the five studies, smoothed with a third-order moving average. In addition, the values from Table 4.1 of the manual (IQ range 95-109) are also shown.
Figure 1: Age course of the Number Correct
It is surprising how closely the values of the five studies from three countries lie together. The values from Table 4.1 of the manual also lie within a reasonable range, considering the coarse age structure (only 3 age groups).
Figure 2 shows the corresponding data for the Errors. Here, only four studies are available for the meta-analytic integration.
Figure 2: Age course of the Errors
Here too, the values of the four studies lie very close together. The manual data of Table 4.2 (middle IQ range) also fit in without difficulty.
In terms of content, the expected age-related decline in performance in visual memory appears relatively early across the lifespan, somewhat more clearly in the Errors than in the Number Correct.
Overview of performance norms and age norms
Figures 3 and 4 show, in composite graphics, the relationship between raw scores and performance scores as well as the (average) influence of age on the number of correct responses (Figure 3) and on the number of errors (Figure 4) in the BVRT. In both figures, the vertical marks indicate the performance scores from 40 to 145. The raw scores are plotted in their correct positions 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 represent the percentile ranks 2.5, 16, 50, 84, and 97.5. One such line in each case, namely the one corresponding to the examinee's age group, is drawn (in a different form) in the TDB2Online profile sheet to aid in the individual interpretation of the test results. Because the performance scores are calculated from the results of the 20-24-year-old young adults, 4 of the 5 markers of the 20-24 age group enclose the performance scores 70, 85, 100, 115. Because the test ceiling of the 10 designs of the Benton Test is too low to measure very high performance, the 5th marker is not at 130. The maximum number of correct responses, that is, a raw score of 10, corresponds to a performance score of 122. The minimum number of Errors, that is, a raw score of 0, corresponds to a performance score of 121.
Figure 3: Performance scores for the Number Correct on the Benton Test with raw scores and age norm limits (see text)
Figure 4: Performance scores for the Errors on the Benton Test with raw scores and age norm limits (see text)
Both parameters of the Benton Test react very sensitively across the age course. The age-associated decline in performance from young adulthood to old age amounts to roughly two standard deviations.
Discussion
The Benton Test is - given some experience in administration and scoring - an economical test for assessing visual memory in clinical settings. It may have lost some of its earlier popularity, which was partly based on unrealistic expectations regarding sensitivity to all possible neuropsychological impairments. However, it is still a good test when an assessment of visual memory is to be carried out, whether as a complement to verbal memory or independently of it.
Similar to the AVLT, the psychometric properties of the test are only moderate. The test ceiling is quite low, so very good performance cannot be measured. The two parameters correlate very highly with each other. Especially for the errors, a very strong dependence of the variance on the mean (across age) must be noted, a consequence of the lack of linearization of the scale.