Back to How Tests EvolovedReliability and
Validity
Reliability
Copyright 1998 by Rawley Silver, reprinted with permission
from its original publication.
No portion of this work may be copied without written consent by Rawley Silver.
The following are excerpts from
Silver, R. A. (1998). Updating the Silver Drawing Test and Draw A Story manuals; New studies and summaries of previous research. FL: Ablin Press.
To determine scorer reliability, three studies were undertaken. In one, seven judges blindly scored SDT responses by six children. The correlation coefficients were .93 in Predictive Drawing, .91 in Drawing from Observation, and .96 in Drawing from Imagination, indicating strong reliability.
In another study, five judges blindly scored responses to the Drawing from Imagination Task by 15 children, adolescents, and adults selected at random. Using the statistic Intraclass Correlation (ICC), the correlation coefficients were .94 for Emotional Content and .74 for Self-Images, again showing reliability.
In the third study, six judges rated pre-test post test scores of eleven learning-impaired children in the study by Silver & Lavin (1977) summarized later on. The reliability coefficient was .852 for ability to form groups (select and combine) and .944 for spatial orientation. The obtained coefficients indicated that the six judges had a similar frame of reference and displayed a high degree of agreement in scoring the tests.
To determine test-retest reliability, Moser (1980) administered the SDT twice to 12 learning disabled adolescents after an interval of one month. All coefficients were significant at the .05 level (.80 in Predictive Drawing, .84 in Drawing from Observation, and .56 in Drawing from Imagination).
To determine whether training is needed in order to administer and score the SDT, three studies were conducted. None of the teachers or mental health professionals who administered and scored the SDT were trained to use it; all relied on the test manual guidelines.
In the first study, an art therapist administered the test and scored responses by 20 fourth-graders and 16 tenth-graders. Their test booklets were then scored blindly by the author (who had not met the art therapist). All correlation coefficients were significant, five at the .01 level (ranging between .65 and .86), and one at the .05 level (.45 in Drawing from Observation).
In the second study, another registered art therapist administered the test to nine mentally handicapped, hospitalized adults, then scored responses which were again scored blindly by the author. All correlation coefficients were significant at the .01 level (Predictive Drawing, .99; Drawing from Observation, .89; Drawing from Imagination, .91; total scores, .96).
In the third study, a classroom teacher, a psychotherapist, a graduate student in art therapy, and two art therapists scored test performances by ten children. None of the scorers had met the author. Significant correlations were found at the .01 level (r = .66). These findings seem to indicate that training for teachers and mental health professionals is not needed.
page 12-13.
To determine the validity of the SDT, its relationships to ten traditional tests have been examined.
Otis Lennen School Ability Test, Metropolitan Reading Achievement Test, SRA Math Achievement Test, Iowa Test of Basic Skills, WISC Performance IQ, and Canadian Cognitive Abilities Test
Although significant correlations were found between each of these tests and various SDT subtests, the reliability coefficients were low, ranging between -.15 (SRA Math) and .55 (Iowa). It should be noted that these tests are weighted with verbal items which do not appear in the SDT, and the SDT may tap cognitive skills not measured by these tests. Although all are measures of intelligence, they use different assessment techniques and emphasize language and visuo-spatial skills to different extents.
page 13.
Continue to
Art as a Tool for Adjustment