Vocabulary Test WST
Wortschatztest WST, a German language vocabulary test
Authors: Schmidt & Metzler, 19921)
Caution: This is not the German language version of the Mill Hill Vocabulary Scale by Bulheller & Ibrahimovic,2) which occasionally also appears as “Wortschatztest (WST)”.
Background
Vocabulary tests measure crystallized verbal intelligence. In general, vocabulary tests are the measures that correlate most highly with the overall result within a intelligence test battery. They thus capture a relatively large share of the general factor g of intelligence. They are therefore very popular whenever the aim is to estimate an adult's intelligence relatively quickly.
Two variants can be distinguished, namely those that measure active vocabulary (one example is the Vocabulary subtest in the Wechsler batteries (e.g., the WAIS-III),3) which requires explaining a word) and those that require recognizing a word within a series of “non-words” and thus test vocabulary passively. Examples of the second type are the German language MWT-B4) and the German language Wortschatztest WST.5)
Passive vocabulary tests have the advantage that they depend less on an intact memory. Passively distinguishing whether a word seems familiar or unfamiliar places far fewer demands on memory than explaining a word in other words. In this respect, passive vocabulary tests are the methods of choice when it comes to estimating premorbid crystallized intelligence in persons with suspected dementia.
Test material
The WST6) is a vocabulary test that the examinee completes independently in writing. In each of 42 rows, one must recognize the real word among five pseudowords and underline it. Example (item from the test instructions):
Renek - Skerk - Erenk - Kern - Nerk - Lersk
To administer the test, one needs the test manual, an answer sheet, and the scoring template for evaluation.
Copyright
The WST is a copyright-protected test; the rights are held by Beltz-Test-GmbH, which now belongs to Hogrefe. It can be obtained through the Testzentrale (test distributor) in Germany or Switzerland.
Administration
Most examinees complete the test in 5 minutes; there is no time limit.
The original instructions tell examinees to respond only if they have recognized a correct word, that is, not to guess. In English translation, the original text reads:
- If you have recognized a word as correct, please cross out the word. You are not required to give a definition of the word. However, we ask you not to guess! If you do not know any of the words in a row, do not cross out anything.
This instruction is somewhat problematic for insecure patients, who are very common in a clinical context. For very intelligent examinees it hardly matters, because the guessed portion of the items is small. For less intelligent examinees, however, this instruction can lead to patients with mania or personality disorders always receiving a few more IQ points than insecure and depressed patients, because the former guess anyway and the latter adhere to the instruction. That is not really what one wants. Actually, this instruction is therefore not sensible in the clinic. However, the test was normed with this instruction, even though it is not known how many persons in the standardization sample actually adhered to it…
Test raw scores
There is only one raw score, the sum of the correctly answered items. The items of the WST are Rasch-scaled. This has two practical consequences: First, the raw score is thus a statistic that exhaustively summarizes the test result. Second, there are no scaling values for the possible raw scores 0 and 42.
Distributional properties of the raw scores
The Rasch scaling of the items entails a normalization. From the norm table in the WST manual, the raw score distribution of the standardization sample can be back-calculated; it is shown in Figure 1.
Figure 1: Back-calculated distribution of the raw scores in the WST standardization sample
The distribution is unimodal, without floor effects and without major ceiling effects, but strongly negatively skewed. From this one can infer that the sensitivity and measurement precision of the WST will be considerably higher in the lower intelligence range than in the upper; this will become clearer below.
Linearization of the measurement scale
The Rasch scaling entails a normalization of the scale.
Norming studies
To our knowledge, the only norming study is the one published in the test manual. It was conducted by the test authors on a sample of 300 female and 273 male persons, all native German speakers, aged between 16 and 90 years. They were patients in general medical and dental practices as well as hospital employees. The intention behind this selection was to remain largely independent of the variables education and intelligence. A pure random sample was not possible for the authors at the time.
The age distribution of the sample is given in the manual. In the age groups 16-19, 20-29, 30-39, 40-49, and 50-59, between 8 and 14 persons per year of age were tested; in the higher age brackets there were considerably fewer (between 60 and 69 still about three per year, between 70 and 90 only one to two). The mean age was 40 years (SD 15.5). According to the authors, the occupational education classes, which are also tabulated in the manual, are in agreement with the 1990 classification of the Federal Statistical Office.
Unfortunately, there is no breakdown of the norms by age group. The authors justify this by stating that there was no significant influence of age on the test scores. In this case, however, that is a rather formalistic argument. The correlation of the ability estimates, which are linearly related to the z scores and thus to the IQs, with age was r=.08. With 573 persons, this is just barely not significant at the 5 percent level, with a p of 0.056. Since the correlation is positive, one must suspect, as with the vocabulary tests in the WIE or HAWIE-R, a curvilinear relationship, with an increase in performance over the first three or four decades of the sample and a decline thereafter. Because the older age groups are less well represented, this is likely to result in a low positive correlation. It is a pity that, for lack of normative data, the influence of age on the WST cannot be modeled.
For the inclusion of the WST in TDB2Online, this means that, deviating from the usual procedure, the performance scores are not based on a separate group of young adults but, for lack of further differentiation, on the entire standardization sample. Age-appropriate results unfortunately cannot be calculated. We console ourselves with the fact that, for vocabulary tests, and especially for passive ones such as the WST, the influence of age is at least much smaller than for other abilities.
Overview of performance norms and age norms
Figure 2 shows the performance norms for the WST. Since there is no age differentiation in the standardization sample, the age norms coincide with the performance norms.
Figure 2: Raw scores and performance scores for the WST
Noteworthy is the very different differentiating ability of the WST in the lower and upper intelligence range: between the performance scores 70 and 85 there are 15 raw points, between 115 and 130 only four, and even between 100 and 115 only five. The WST is thus a test that measures very well and comparatively precisely in the lower half of the intelligence range, but not in the upper. This need not be a disadvantage, because the test is in any case used predominantly with elderly or very elderly people to estimate premorbid intelligence. Of clinical relevance is above all the differentiation in the lower performance range, which the test does well.