Wechsler Adult Intelligence Scale - Revised (WAIS-R)

The German version of the WAIS-R, the HAWIE-R, is very outdated. The documentation text on the HAWIE-R is retained here only for historical reasons and for scientific comparison purposes. Since 2006 there has been a successor to the HAWIE-R (published in 1991) in the WIE. The norms of the two tests differ.¹⁾ Since 2012 there has also been a successor to the WIE, the WAIS-IV. There should no longer be any reason to administer the old HAWIE-R rather than the newer WIE or - best of all - the WAIS-IV in the initial examination of a patient or examinee. For cases in which one wishes to compare old HAWIE-R reports with newer WIE reports, TDB2Online offers a workaround: using the equivalence tables from the work of Erzberger & Engel (2010), a conversion was programmed in TDB2Online under the “Test” outline item HAWIE-R with WIE norms. Subtest raw scores obtained with the HAWIE-R are presented in the evaluation as performance scores and age-normed scores based on the WIE norms. This makes it possible to retrospectively render old HAWIE-R results comparable with WIE results (for example, in repeated measurements of the same patient). Similar evaluation bridges to the WAIS-IV are lacking because there are no corresponding equivalence studies.

The Hamburg-Wechsler Intelligenztest für Erwachsene – Revision²⁾ is the German edition of the Wechsler Adult Intelligence Scale – Revised (WAIS-R).³⁾ The WAIS-R was published in the USA in 1981 as the third “generation” of the intelligence test battery that David Wechsler compiled in the 1930s, at that time as the “Wechsler-Bellevue Scale”, primarily for clinical purposes. The German edition was published in 1991, and a corrected manual dates from 1994.

The test manuals contain only the material necessary for administration and scoring. Historical background information on the Wechsler tests can be found in the classic handbook by Matarazzo⁴⁾, which is also available in a German translation.⁵⁾ However, this book refers to the predecessor tests of the HAWIE-R.

More recent literature on the clinical interpretation of the Wechsler intelligence test batteries for adults is currently available only in English.^{6) 7)} However, both refer predominantly to the WAIS-III, the German version of which is the WIE.

With its 11 subtests, the HAWIE-R offers a good overview of the structure of cognitive performance, especially in cognitively impaired patients. The composition of the subtests was not changed in comparison with the predecessor HAWIE⁸⁾; changes are found only at the item level within the individual subtests. However, the norms differ considerably from those of the HAWIE, which - as previously with the children's version, the HAWIK - led to various practical problems in assessment. Details have been reported by Satzger et al. 1996.⁹⁾

The HAWIE-R is divided into a Verbal part with six subtests and a Performance part with five subtests. Table 1 gives an overview of the subtests with their English names, their assignment to the test parts, the abbreviation customary in English, the rank position in test administration, and the possible raw score range.

Table 1: Subtests of the HAWIE-R

For the administration of the HAWIE-R one needs the complete test set, which, in addition to the manual and an initial supply of record forms and task booklets, also contains the test material for the practical subtests. Since 2006 the test material has no longer been commercially available.

The copyright for the HAWIE-R is held by the publisher Hans Huber in Bern.

All Wechsler intelligence test batteries are individually administered tests. The test material is varied and, through its diversity, maintains as well as possible the usually rather moderate motivation of clinical examinees; the individual administration with only one patient also contributes to this.

Administration is described well and in detail in the test manual. It is more difficult than with many other ability tests and must be learned and practiced. The verbal tests, for example, are answered freely. This is good for the examinee, but for the test administrators it means that the answer must be compared with the sample answers in the manual and scored. One must also know in which cases to ask for further explanation. It is not sufficient to read later that one should have asked a follow-up question in response to a particular answer in order to arrive at a clear score.

Administration of the complete HAWIE-R usually takes about 90 minutes in psychiatric patients. To this must be added the scoring time, which depends strongly on the experience of the administrator.

The 11 subtests of the HAWIE-R yield raw points of different origin. For the Verbal subtests Information (IN), Digit Span (DS), Vocabulary (VC), Comprehension (CO) and Similarities (SI) and for the Performance subtests Picture Completion (PC) and Digit Symbol Coding (DSC), the raw points represent a sum of the correct solutions, whereby for CO and SI all answers are scored with one or two points according to their degree of appropriateness and completeness. In contrast to this (and also in contrast to most other tests), the raw points for the Verbal subtest Arithmetic (AR) and for the Performance subtests Picture Arrangement (PA), Block Design (BD) and Object Assembly (OA) are composed, for each item, of up to nine basic points for a correct solution and up to three additional time points for the speed of the solution. This means that the raw score distributions of these subtests - especially in the lower range - are not occupied at all at some points and jump sharply at others. The maxima of the raw scores vary between the subtests from 17 (PC) to 93 (DSC); the minima are uniformly 0.

Because of the differing numbers of points, the raw scores have been converted into so-called scaled scores by means of tables since the first generation of the Wechsler intelligence tests. The methods for converting raw scores to scaled scores have changed over the generations of Wechsler tests and are described in detail in the corresponding section on the WAIS-III. In the HAWIE-R, as in the American original (WAIS-R), scaled scores are scaled such that for each subtest the scaled score distribution of young adults between 20 and 34 years has a mean of 10 and a standard deviation of 3. This method leads to a good definition of the scaled scores, entirely in the spirit of the performance scores in TDB2Online: they characterize the performance of young healthy examinees at the peak of cognitive performance. The age range of the normative sample does not have any influence on this. Age-related deficits are found on both sides. Adolescents around the age of 16 achieve, for example, on average only 8 scaled scores on the Vocabulary test, whereas in the subtests of the Performance part the differences from adults are small. Examinees aged seventy achieve in some subtests of the Performance part only a mean scaled score of four to six. The latter in particular impairs the possible scaling of poorer performance in older persons because of the limited range of the scaled scores.

For the HAWIE-R there are practically only the normative data reported in the manual. This was still quite different for the HAWIE. In the course of the test's history of use, normative data were reported for various groups of persons that markedly changed and supplemented the original ones.

In the manual of the HAWIE-R, in addition to the usual table for converting raw scores to scaled scores based on the 20-34-year-olds, there are also tables with the age-specific scaled scores for each subtest. These tables contain the results of a linearizing percentile-to-normal transformation of the raw scores onto the scaled score scale (10;3) for each age group and each subtest individually. They served as the basis for the calculation of norms in TDB2Online.

The tables in the manual give the age-specific raw score-scaled score relationships only as integers. On the one hand, it happens that several raw scores occupy the same scaled score. On the other hand, for some scaled scores there is no raw score equivalent. The hard steps of this conversion were first smoothed by a moving average. The order of the moving average depended on the granularity of the raw score scale. At the ends of the distribution the data were extrapolated so that the transformation equation could be retained. Subsequently, a moving average of the third order was also used to align the data between the age groups. Subsequently, the data for the two age groups 20-24 and 25-34 were averaged. These means formed the basis for the calculation of the performance scores for each subtest.

In principle, the percentile-to-normal transformation leads to a linearization of the raw scores and thus to a normal distribution. At the time of the inclusion of the HAWIE-R in TDB2 this was not examined further.

For the WIE, where the starting situation is essentially the same as for the HAWIE-R, individual distributions were examined later (see the corresponding section in the detailed test documentation for the WIE). Mostly good distributional properties were found, but sometimes also more substantial ceiling or floor effects. In principle, this could also be calculated and presented for the HAWIE-R. In view of the obsolescence of the HAWIE-R, however, this should no longer be necessary.

A linearization of the raw scores is carried out implicitly via the scaled score transformation, namely via those for the age groups 20-24 and 25-34, which are used in combination for the calculation of the performance scores.

When several subtests are summed or averaged and the result of these operations is then re-normed as a new measured value, a new metric arises that is no longer compatible with that of the subtests. This state of affairs was explained in the general documentation in the section Multiple standardization at different levels. The HAWIE-R served there as a demonstration example, which is why reference is made here to that section. We saw there that an examinee who achieves four scaled scores in every subtest does not obtain in the intelligence quotients the IQ of 70 that one would expect with a comparable metric (two standard deviations below the mean of his or her age group), but clearly lower values, namely 57 for the Verbal and Performance IQ and 48 for the Full Scale IQ. The change in the metric depends on the correlation between the subtests and the number of aggregated subtests. In the HAWIE-R, the changes in the metric turn out to be particularly extreme because of the low subtest intercorrelations in the normative sample. This led, in the change from the WAIS to the HAWIE-R, to practical problems with repeated measurements, as had been noticed in exactly the same way, but better documented, in the change from the HAWIK to the HAWIK-R^{10) 11)}

As set out and justified in the general part of the documentation, TDB2Online does not evaluate or further consider the separately standardized intelligence quotients. Instead, for the areas to be summarized we simply report their mean subtest performance scores or their age-dependent standard scores, in exactly the same metric as for the subtests.

The description of the norming study is kept brief in the manual of the HAWIE-R. The sampling followed the specifications of the American licensor and largely corresponded to what was done in the USA for the WAIS-R. Nine age ranges between 16 and 74 years were formed. In the age groups 20-24 and 25-34, 150 men and 150 women each were examined, in all other age ranges 100 men and 100 women each. Within the age ranges, in the sense of a quota sample, a distribution across the three school education groups Hauptschule, Realschule and Gymnasium was aimed for and maintained according to a key corresponding to the 1986 census. In the older groups, therefore, far more Hauptschule graduates were represented than in the younger ones. The survey was conducted “throughout the Federal Republic” (at that time West Germany), without a more detailed description of further sociodemographic variables.

The adequacy of the sample obtained is difficult to judge. There are no other similar studies with the HAWIE-R that could serve as a control. There is only indirect evidence from studies in which the same examinees or patients were given two comparable intelligence tests, namely once the HAWIE and the HAWIE-R (Satzger et al., 1996)¹²⁾ and another time the WIE and the HAWIE-R (Erzberger & Engel, 2010)¹³⁾. The main purpose of both studies was to make quite practical statements about the norm equivalence of the respective procedures examined. As a side effect, however, statements can also be made about the differences in the distribution shapes of the standardization studies of the two procedures.

Figure 1: Distribution of the Full Scale IQs of the 88 persons from the study by Satzger et al. (1996) in the HAWIE

Figure 2: Distribution of the Full Scale IQs of the 88 persons from the study by Satzger et al. (1996) in the HAWIE-R

In the HAWIE-R (Figure 2), the study by Satzger et al. (1996) showed a substantially wider spread of the intelligence quotients than in the HAWIE (Figure 1).

Figure 3: Scatter plot of the HAWIE and HAWIE-R Full Scale IQs of the 88 persons from the study by Satzger et al. (1996)

The deviations in distribution shape correspond to the anomalous curve course in the scatter plot in Figure 3. The non-linear deviations from the diagonal in the bivariate distribution can readily be attributed to the different shape of the unimodal distributions.

On the basis of these data it cannot now be said which of the two norming studies (HAWIE or HAWIE-R) was “better” or “more random”. The only evidence suggesting that it was rather the HAWIE-R normative sample that turned out somewhat anomalous (leptokurtic or steep-peaked) is that in the second study mentioned above, comparing the WIE and the HAWIE-R, the (clearly) wider distribution in the HAWIE-R compared with the WIE was likewise found (data in Erzberger & Engel, 2010).¹⁴⁾

These data, however, also include the scaling effects for the global measure Full Scale IQ, which are particularly strong in the HAWIE-R with its low subtest intercorrelations and therefore likewise produce such effects. As long as the adequacy of a normative sample is not shown more reliably by other data than is generally the case, we are at best dependent on such indirect inferences, which unfortunately do not bring much clarity.

These could still be presented here. In view of the heterogeneity of the subtests included, however, this makes no sense. These global measures can be interpreted only if the individual measures are fairly similar.