Comparison of the Revised and Original Versions of the Bayley Scales of Infant Development

Differences between Mental Scale scores on the revised Bayley Scales of Infant Development(BSID-II) and the original test (BSID) were evaluated. Both versions ofthe measure were administered to 32 subjects, ranging in age from 3 to 28 months, in counterbalanced order. Inter-test correlations indicated that the two versions of the test were significantly correlated (r = .76, p < 0.05). Significant differences between scores on the BSID-II (M = 105.97) and the original BSID (M = 124.38) were observed (t(31) = 8.41,p < 0.05). Children below the age of 12 months displayed slightly larger discrepancies than children above 12 months of age. This discrepancy did, however, appear to increase significantly with the developmental levels of the children, with subjects scoring above one standard deviation on the BSID displaying larger discrepancies than those at or below the normal range. These findings are important for psychologists, early interventionists and other professionals involved in the assessment of infants and toddlers as they interpret scores on the new Bayley Scales. This information is especially pertinent for children whose performance is evaluated longitudinally, since they will likely score lower on the new test than they did on the original Bayley Scales. The lower scores may lead to subsequent increases in the number of children who qualify for early intervention and special education services. Therefore, these findings are also important for those involved in the development and evaluation of intervention programs for young children. The findings are considered in terms of their applicability across cultures. The Bayley Scales of Infant Development (BSID; Bayley, 1969) are well established as one of the most useful instruments available in the assessment of infants (Burns et al., 1992). The Bayley Scales of Infant Development, Second Edition (BSID-II; Bayley, 1993) were introduced to the psychological community in 1993 as a revision of the original Bayley Scales. The aim of the revision was to improve the quality and utility of the BSID, while preserving the original content and intent of the test. Although much of the original content of the BSID was maintained, new items were added to improve the content coverage and ease and accuracy of administration. The original Bayley Scales have proven to be a sound instrument and have been frequently utilized in research studies with a variety of clinical samples, including children with motor and speech delays (Dilworth and French, 1990), HIV infection and prenatal drug exposure (Mellins et al., 1994) and children considered to be biologically high-risk (Aylward et al., 1995; Crowe et al., 1987; Gerken et al., 1994; Siegel et al., 1995). Numerous authors have provided empirical evidence for the validity and reliability of the BSID (Cook et al., 1989; Snyder and Sheehan, 1992), thus establishing the psychometric integrity of the scales. The need for revisions and updated normative data prompted the development of the BSID-II. Empirical evidence for the need to revise the BSID was demonstrated by Campbell and colleagues (1986). In their study of 436 twelve-month-old infants of low socioeconomic status these researchers found that on average, subjects scored 7 to 14 points above the normative sample mean Mental Development Index score of 100. The elevated scores found in their sample led these researchers to conclude that the norms for the original Bayley were outdated and were no longer reflective of the current population of children. According to Bayley (1993), there were seven major goals in the revision of the BSID: to update the normative data, to extend the age range, to improve content coverage, to update the stimulus materials, to improve the psychometric quality and clinical utility of the Scales and to preserve the basic qualities of the original Bayley Scales. To accomplish these goals, Bayley surveyed BSID users to identify strengths and weaknesses of the original scales. Among the strengths were the flexible order of administration of test items, the interesting nature of the tasks and the incorporation of the behaviour rating scale. The weaknesses identified were the layout of the kit, the lack of clarity in the directions and the poor durability of the materials. In revamping the test, Bayley attempted to improve on these weak areas while maintaining the strengths. Approximately three-quarters of the original Mental Scale items were kept in the BSID-II. Although the BSID-II has been in use since 1993, the existing literature on this measure is limited. The majority of available information on the psychometric properties of the revision is found in the test manual (Bayley, 1993). Evaluation of the reliability of the Mental Scale with the standardization sample revealed reliability coefficients ranging from 0.78 at 10 months to 0.93 at 27 months. These coefficients, calculated with coefficient alpha and based on sample sizes of 100 for each age group, are consistent with those obtained on the original BSID and indicate high reliability. The test-retest stability of the Mental Scale was assessed in a study of 175 children in four age groups drawn from the standardization sample. Intervals of 1 to 16 days took place between testing. To determine if reliability varied as a function of age, the 1 and 12 month old children and the 24 and 36 month old children were grouped together. Results indicated that test scores were highly stable over time and across age groups (r = 0.83 for 1 and 12 months; r = 0.91 for 24and 36-months). Inter-scorer agreement was evaluated by correlating the scores produced by an evaluator and a rater who simultaneously recorded scores in a sample of 51 children. Inter-scorer reliability was found to be high (r = 0.96). Comparison studies of the BSID-II are also described in the manual. To establish the relationship between the BSID-II and the original BSID, both versions of the scales were administered in counterbalanced order to 200 children ranging in age from 1 to 42 months (M = 15.51, SD = 10.13) (Bayley, 1993). A moderate correlation (r = 0.62) was found between the MDI scores on the two tests, as well as a mean difference of 12 points, with the BSID-II scores being lower. These findings indicate substantial overlap in the content ofthe original and revised versions of the Bayley Scales, although Nellis and Gridley (1994) emphasize that this moderate correlation reflects the notable differences between the two tests and highlights the importance of avoiding generalizing too much about the BSID-II from the BSID. This issue will be addressed further in relation to the findings of the current study. Strong correlations between MDI scores and the General Cognitive Index Score on the McCarthy Scales of Children's Abilities (r = 0.79) and the Full Scale IQ of the Wechsler Preschool and Primary Scale of Intelligence Revised (r = 0.72) provide additional evidence for the concurrent validity of the BSID-II. The results of a comparison study conducted by Goldstein and colleauges (1995) provided support for the clinical validity of the revised Bayley Scales. In this study, both versions of the test were administered to 49 high-risk, premature infants ranging in age from 12 to 22 months. Findings indicated that the mean MDI score was within 4-5 points of the score reported in the BSID-II manual with a similar clinical population. This study also reported very high correlations between MDI's on the BSID and BSID-II, with scores on the new test approximately 8 points below those on the old test. These findings suggest that the constructs measured by the two tests are very similar. Although the Bayley Scales were standardized on American children, the measure has been used in studies cross-culturally. Whereas the normative data for the BSID may not be applicable to infants from other cultures, developments in other parts of the world indicate that modifications to the original measure may result in appropriate use of the BSID in other cultures. A Dutch translation of the BSID was developed and administered to 1283 Dutch infants, resulting in the development of Dutch norms for the Mental and Motor scales, as well as for the Infant Behavior Record (van der Meulen and Smrkovsky, 1985). In India, the original Bayley Scales were used in the development of an infant scale of mental development for children from 1 to 30 months, which was found to effectively discriminate between normal and delayed infants (Jayashankarappa, 1986). The original version of the Bayley Scales was used in a study of undernourished Kenyan infants which assessed early predictors of cognitive development (Whaley et al., 1998). The modifications made to the BSID made it more appropriate for the culture and testing conditions. A recent study conducted in Australia compared both versions of the Bayley Scales in a sample of 97 healthy infants ranging in age from 18 to 27 months, revealing 4 to 35 point gains in MDI scores (Tasbihsazan et al., 1997). Thus, it becomes evident that the Bayley Scales can be used appropriately in other cultures. The comparability of the two versions of the test should be studied in the American population before the new version is used in other cultures, since the test was normed on this population. Once this has been demonstrated, new studies should emerge which compare the old and new versions of the measure in different populations. As with any revised test, it is necessary to examine the relationship between the present and previous versions for both empirical and clinical reasons. The purpose of the present study was to compare Mental Scale scores on the BSID with corresponding scores on the BSID-II among healthy, nonreferred infants. Research by Flynn (1984; 1987) suggests that tests which have been in use for some time tend to produce inflated scores and that with revisions of such tests, mean differences increase with the length of time between standardizations. Bracken (1988) describes how scores on an original and revised version of a test are significantly different, with the magnitude of the difference being dependent on the norming dates. Nellis and Gridley (1994) explain that declines in scores can be expected when a revision of a measure is published many years after the original version. The BSID-II manual explains that these declines should be expected, based on changes in the content, scoring criteria, item directions and floor and ceiling properties (Bayley, 1993). With this information in mind, it was expected that the present sample would demonstrate a decline in scores from the BSID to the BSID-II.