The short form of the SON-R 5.5-17

nonverbal intelligence test

Supplement to the Manual and Research Report

Peter J. Tellegen & Jacob A. Laros

February 2004

University of Groningen, University of Brasília

In this supplement additional information is given about the short form of the

SON-R 5.5-17. The SON- tests are intelligence tests for general application that can be administered without language, either in written or spoken form. The tests are therefore particularly suited for children with problems related to verbal communication and for children that were raised with another native language. The basis for applying the test, as well for the short form as for the complete form, is the “Manual and Research Report”. Hereafter, a number of relevant data from this manual will be summarized, with additional analyses in relation to the validity of the short form.

Composition of the short form

The official short form of the SON-R 5.5-17 consists of the following four subtests: Mosaics, Situations, Categories and Analogies. The subtests Patterns, Stories and Hidden Pictures of the complete form are not administered. With the exclusion of these three subtests the mean duration of test administration is reduced from 90 minutes (complete form) to 45 minutes (short form), including short breaks.

On basis of their contents the subtests of the SON-R 5.5-17 can be divided in the following four groups: abstract reasoning tests (A), concrete reasoning tests (C), spatial tests (S) and a perceptual test (P). However, principal components analysis showed that a division of the subtests in two groups, a spatial factor and a reasoning factor, emerges more clearly (see Manual and Research Report, section 6.4). In Table 1 the loadings on the first two varimax rotated components are shown for three different age groups. The subtests of the short form are the first four subtests of the table and are printed in italics.

Table 1

Loadings of the subtests on the first two components after varimax rotation (I: reasoning, II: spatial)
6.5 years / 10.5 years / 14.5 years
I / II / I / II / I / II

Mosaics

/ (S) / .3 / .9 / .4 / .8 / .4 / .8

Situations

/ (C) / .7 / .5 / .8 / .3 / .8 / .4

Categories

/ (A) / .8 / .1 / .7 / .4 / .7 / .4

Analogies

/ (A) / .6 / .5 / .6 / .6 / .6 / .7
Patterns / (S) / .3 / .9 / .3 / .9 / .3 / .9
Stories / (C) / .7 / .4 / .8 / .3 / .8 / .4
Hidden Pictures / (P) / .7 / .2 / .7 / .4 / .8 / .4

- The principal component analysis is based on correlations corrected for unreliability.

With the selection of the subtests of the short form of the SON-R 5.5-17, both the spatial aspect and the abstract and concrete reasoning aspects are represented. The subtest Analogies, the second abstract reasoning test in the short form, also has a strong loading on the spatial factor. The choice of the four subtests for the short form is based partly on considerations with regards to content and partly on considerations related to the fact that these subtests are easy to administer and that the costs are relatively low because no non-durable test materials need to be purchased. Another consideration in the choice of the four subtests was the strong reduction in the duration of the test administration.

Reliability and generalizability

The reliability of the subtests of the short form and the reliability and generalizability of the IQ score, based on the short form, are shown in Table 2 (see Manual and Research report, sections 6.3 and 7.4). The generalizability is a measure that indicates the extent to which one can generalize the test score on basis of this selection of four subtests to other random selections of four subtests from the same domain. Reliability of a test indicates the degree to which one can generalize the test score to the domain of comparable items (nevertheless, the subtests remain the same) while the generalizability refers to the much broader domain of subtests. The generalizability can be considered as core measure for the validity of a test.

Table 2

Reliability and generalizability of the short form of the SON-R 5.5-17
reliability subtests / IQ short version
Cat. / Mos. / Sit. / Ana. / Reliab. / Gen.
5.5 years / .62 / .78 / .68 / .77 / .85 / .67
6.5 years / .67 / .81 / .68 / .78 / .87 / .70
7.5 years / .71 / .83 / .68 / .78 / .89 / .73
8.5 years / .73 / .84 / .69 / .78 / .90 / .76
9.5 years / .75 / .85 / .70 / .78 / .90 / .77
10.5 years / .75 / .84 / .71 / .79 / .91 / .79
11.5 years / .75 / .84 / .72 / .79 / .91 / .80
12.5 years / .74 / .83 / .74 / .79 / .91 / .80
13.5 years / .72 / .82 / .75 / .79 / .91 / .81
14.5 years / .71 / .80 / .76 / .80 / .91 / .81
15.5 years / .70 / .79 / .76 / .80 / .91 / .82
16.5 years / .69 / .79 / .77 / .80 / .91 / .82
Mean / .71 / .82 / .72 / .79 / .90 / .77

The mean reliability of the subtests over the 11 age groups amounts to .76. The mean reliability of the IQ of the short form is .90 and the mean generalizability is .77. These indices of the short form of the SON-R 5.5-17 tend to increase with age. From 8.5 years onwards the reliability of the short form IQ is at least .90 and the generalizability at least .75. In comparison to the full version of the test the mean reliability of the short form diminishes from .93 to .90 while the mean generalizability decreases from .85 to .77.

Across the eleven groups the mean correlation between the IQ score of the full version and the IQ score of the short version is .95; the mean correlation of the IQ score of the short version with the sumscore of the three subtests that are not administered is .76.

Validity

For a number of analyses that are discussed in the Manual and Research Report of the SON-R 5.5-17 it was verified whether the relationship of the IQ score of the short form with external variables differ from the relationship of the IQ score of the full version (see Manual and Research Report, chapter 8). The somewhat lower reliability and generalizability might lead to lower validity indices of the short form. This should be expressed in lower correlations with relevant criteria.

Table 3

Comparison of the relationship of IQ short version and IQ full version with external criteria
External variable / N / measure / IQ short / IQ full / diff.
occupational level parents (6 levels) / 1.289 / eta / .29 / .30 / -.012
native country parents (3 groups) / 1.350 / eta / .16 / .16 / .001
report marks primary school
language / 696 / r / .38 / .38 / .003
reading / 693 / r / .29 / .31 / -.018
writing / 675 / r / .23 / .22 / .007
arithmetic / 696 / r / .43 / .43 / .001
history / geography / 423 / r / .39 / .39 / .003
secondary school (5 types) / 450 / eta / .57 / .59 / -.024

From the data in Table 3 it becomes clear that the relationship with the external criteria practically does not diminish with the short version of the test. Presumably, the slight loss of reliability and generalizability is compensated because in the total score abstract reasoning gets more emphasis.

Interpretation of test scores

The standardized test scores are based on a model that describes the distribution of the test scores as a function of age. In this way, with the use of the computer program that is delivered with the test, standardized scores can be calculated based on the exact age of the person that is being tested. All indices that are calculated for the full version of the test are also reported when one applies the short form. This is true, amongst others, to the 80% probability interval, the distinction between Specific IQ and Generalized IQ, the reference age, and the statistical test of the null-hypothesis that the subtests scores are the same. The consequences of administering four instead of seven subtests are reflected in a broader probability interval of the IQ score. This probability interval is automatically printed on the front page of the scoring form and can therefore easily be included in the interpretation of the test results.

Stepwise testing

In the Manual and Research Report of the SON-R 5.5-17 it is recommended to administer the complete version of the test when the test results can have important implications for the tested person. Based on the observed very small decrease in validity, the use of the short form seems to be justified in many cases. This is especially so when in the process of decision taking other test results, observations and school results etc. are also considered. With the administration of the short form, however one needs to realize that only the IQ score can be interpreted and that there are no possibilities to diagnose differences in spatial ability and reasoning ability.

When it is desired to use the IQ score that is based on a complete administration of the SON-test, one might also consider to administer the test in a stepwise fashion. In the first instance one administers the short form and evaluates if it is desirable in view of the results and the goals of the application of the test, to administer the remaining three subtests. Is this the case, subsequently the IQ score is calculated for the full version and decisions are based on that score.

Example

Suppose that for a referral to a special type of education it is necessary that “the” IQ score of a student is lower than 90. Let us also suppose that the Standard IQ of the student is 95 with a 80% probability interval (of the Generalized IQ) of 88 to 104. Based on this interval a reasonable possibility exists that “the” IQ is lower than 90. After administration of the three subtests of the full version that are not included in the short form (Patterns, Stories and Hidden Pictures), the Standard IQ appears to be 92. This value is above the limit of 90 and the student in question is not being referred to a special type of education. This procedure would imply that the full version of the test would only be administered to students with a Standardized IQ on the short form between 80 and 97. After the administration of the four subtests of the short form, one can easily calculate the Standardized IQ with the computer program, and if so desired one can complete the administration of the full form in the same session.

Review by the COTAN

In 2003, the short form of the SON-R 5.5-17 was reviewed by the COTAN, the commission of Test Evaluation of the Institute of Psychologists in the Netherlands. The categories for evaluation are insufficient, sufficient and good. The short form was evaluated as good on all the seven criteria of test quality. In Table 4 the evaluation of the short form is presented together with the evaluations of the complete form of the SON-R 2.5-7 (Tellegen, Winkel, Wijnberg-Williams & Laros, 1998) and the SON-R 5.5-17.

Table 4

Evaluation of the SON-tests by the COTAN

======

SON-R 2.5-7 SON-R 5.5-7

Complete test Compl. test Short form

------

Basics of the construction of the testgoodgoodgood

Quality of the test materialsgoodgoodgood

Quality of the manualgoodgoodgood

Normsgoodgoodgood

Reliabilitygoodgoodgood

Construct validitygoodgoodgood

Criterion validitygoodgoodgood

======

References

Laros, J.A., & Tellegen, P.J. (1991). Construction and validation of the SON-R 5.5-17, the Snijders-Oomen non-verbal intelligence test. Groningen: Wolters-Noordhoff.

Snijders, J.Th., Tellegen, P.J., & Laros, J.A. (1989). Manual and Research Report of the SON-R 5.5-17. Groningen: Wolters-Noordhoff.

Tellegen, P.J., Winkel, M., Wijnberg-Williams, B., & Laros, J.A. (1998). Manual and Research Report of the SON-R 2.5-7. Lisse: Swets & Zeitlinger.