Stability Reliability of the Griffiths Scales of Child Development (3rd Edition) ()
1. Introduction
The reliability of test scores is an important factor in test selection (Foxcroft & Roodt, 2013). Stability reliability is an overarching concept that refers to the general stability of test scores, regardless of time or administration conditions (de Vos et al., 2014). Test-retest reliability, as an indicator of the stability of observed scores over time, is important for tests of child development as they may be used more than once on a specific child (Singh et al., 2016; Foxcroft & Roodt, 2013; Bayley, 2006). Inter-rater reliability relates to who administers or scores a test, and whether the administration and scoring instructions of a test are sufficiently robust to counteract the effect of different test administrators (Aldridge, Dovey, & Wade, 2017; Neuman, 2011). The Griffiths Scales is a test of child development and has a long history dating back to 1954 (Griffiths, 1954). A search of previous studies found the most recent mention of test-retest reliability to be high (.82) for a previous version of the Griffiths (Huntley, 1996). No research publications about the test-retest reliability of the latest 2016 edition of the Griffiths Scales were found.
2. The Griffiths Scales of Child Development
The third edition of the Griffiths Scales of Child Development (Griffiths III) was published in 2016, following an extensive revision and builds on its predecessors as an individually administered measure of child development (Stroud et al., 2016). The first edition of the Griffiths Scales, the Baby Scales covering the first two years of a child’s life, was published by Dr. Ruth Griffiths in 1954 (Stroud et al., 2016) in the United Kingdom (UK). The Griffiths Scales have been embraced by users of many countries across the globe as a developmental test for children (Sharp et al., 2018; Tso et al., 2018; Stroud et al., 2016; Cirelli, Bickle Graz, & Tolsa, 2015).
Through a reverse-reengineering process, the original theoretical understanding was found to remain a valid framework for assessing children (Stroud et al., 2016). To meet the needs of test users, however, the theory was enhanced with current assessment needs, such as executive functioning within a neurodevelopmental framework, including different forms of memory that impact general child development. The test is used in different settings which reflect its diverse user base of psychologists, paediatricians, and certain allied health professionals. The Griffiths III assesses development across the five subscales of Foundations of Learning, Language and Communication, Eye and Hand Coordination, Personal-Social-Emotional, and Gross Motor, as well as providing a General Development score. Each subscale can be administered on its own, but test administrators are advised to administer all five subscales to obtain a rounded picture of a child’s abilities (Stroud et al., 2016). The age range of the test is from birth to six years.
As Griffiths III is a developmental test, items are placed in order of ascending difficulty. A child will perform items on a subscale until reaching a ceiling. At this stage, the administration of that subscale concludes, and the assessment continues with the next subscale. This format allows for multiple assessments of a child over time, as a child should face new, more difficult, test items with each administration, that are commensurate with their level of development. Test users can therefore use a single measure across multiple years of a young child’s development, which creates a stable baseline whilst allowing practitioners to map a child’s development. It further assists in identifying relative areas of strength and weakness for a child that allows for the creation of focussed development practices that are tailored to the specific child. The Griffiths III represents a particularly extensive update from its predecessors, so similar test-retest reliability figures cannot be assumed. The A merican Educational Research Association (2014) directs that each test revision seeks to improve on the validity and reliability of a test, thereby necessitating this investigation. Furthermore, given its use over time for a specific child, it is important to investigate the test’s test-retest and interrater reliability.
From a test-retest reliability perspective, the challenge of developmental tests is that children are expected to obtain a slightly higher raw score over time, to correspond with their chronological age and developmental trajectory (Azari et al., 2017). This upward movement in raw scores will affect some test-retest reliability scores, such as dependent-sample t-tests, thereby affecting their accuracy. T-tests are useful indicators of stability of average scores within samples, but due to the movement of such scores for developmental tests, this statistical technique tends to be absent for such tests (e.g. Azari et al., 2017; Bayley, 2006).
Additionally, if a child is assessed by different test users from time to time, it is also important to consider inter-rater reliability. In doing this confidence is created in the longitudinal profile of a child’s development that may reflect the work of different test administrators over time.
Therefore, the aim of this study was to investigate the stability reliability of the Griffiths III as a function of test-retest reliability and inter-rater reliability.
3. Methods
A quantitative research method was employed to investigate the aim of the study. Approval from a National Health Service (NHS) Research Ethics Committee was not required in the UK or Republic of Ireland (ROI), as the study built on the standardisation project of the Griffiths III.
The interval between the two test sessions was two to four weeks. This interval is similar to studies for other developmental tests, such as the Bayley Scales of Infant and Toddler Development—Third Edition (Azari et al., 2017; Bayley, 2006), and the Ages and Stages Questionnaire (Singh et al., 2016).
3.1. Sample
The sample consisted of 53 children in the UK and ROI. The inclusion criterion for this study was for participants to have an uneventful medical and developmental history. The reason for this is that the study represented the first published research on the stability reliability of the Griffiths III, and therefore it sought to establish a baseline for future studies. Parents were required to complete a medical and developmental history questionnaire on their children, and this information was used to screen participants for the present study. The final sample consisted of 28 (53%) girls and 25 boys (47%). At pretest, ages ranged from three to 63 months, with a mean of 33 months (SD = 17.9). Thirty-two children were tested and retested by the same administrator, and for 21 children each test session was conducted by a different test administrator.
3.2. Data Analysis
To accommodate the expected increase in raw scores associated with child development over time, the researchers used the original standardisation sample of 426 to calculate the expected increase. This was calculated to a score of .83 to .86 per thirty days. Post-test raw scores were adjusted for each child, depending on the length of time between the pre- and post-test, and then rounded. This means that for children that were tested two weeks apart, the adjustment would not have affected post-test raw scores, whilst children who were tested three or four weeks apart had post-test raw scores reduced by one mark.
The stability of the Griffiths III scores was investigated in two ways. The first was a Pearson product-moment correlation coefficient on the pre- and post-test raw scores. To determine the significance of differences between the same and different administrators, chi-square analyses were performed, with Cramér’s V for practical significance. The second set of analysis was dependent-sample t-tests to compare raw score differences.
4. Results
The descriptive statistics of scores are presented in Table 1 below. The mean scores were similar per subscale for the pretest and post-test.
For test-retest reliability, there were very high correlations for the total sample ranging from .969 to .991 between the two sets of test scores (see Table 2). The coefficients were similarly very high for the same test administrator (.967 to .989) and different test administrators (.969 to .996). Of all the scales, the difference in correlations appeared the widest for the Gross Motor subscale, but this was not statistically significant.
Dependent sample t-test was used to investigate the significance of differences in pre- and post-test raw scores for all administrators (see Table 3). The only statistically significant difference was found for the Gross Motor subscale, where a small effect of .38 was observed.
For inter-rater reliability, dependent-sample t-tests were performed for the
![]()
Table 1. Descriptive statistics for the Griffiths III pre- and post-test.
![]()
Table 2. Griffiths III test-retest correlation coefficients.
*Test Administrators.
![]()
Table 3. Differences in Griffiths III pre- and post-test scores for all administrators.
![]()
Table 4. Differences in Griffiths III pre- and post-test scores for same administrators.
same and different administrators on pre- and post-test raw scores. No statistically significant difference was found when children had been tested and retested by the same administrator (see Table 4).
The only source of statistically significant difference was when different administrators tested children, but this difference was restricted to the Gross Motor subscale, which in turn affected General Development as an averaged score across the five individual subscales (see Table 5). These differences had a medium effect size (.66 and .69).
5. Discussion
The findings of the present study point to very high test-retest reliability of the
![]()
Table 5. Differences in Griffiths III pre- and post-test scores for different administrators.
Griffiths III. The reliability figures in this study were higher than the 0.82 reported by Huntley (1996), for a previous edition of the Scales. This points to an improvement in the reliability of the Scales with the publication of the Griffiths III.
The only source of statistically significant difference appeared to be for different test administrators. The impact of reliability across different test administrators was localised to the Gross Motor Subscale, which in turn affected the General Development raw score. The Gross Motor Subscale draws on activities such as bilateral coordination, postural control, balance, power and strength, and motor sequencing (Stroud et al., 2016). Gross motor subscales are important for tests of general development to evaluate the soundness of the development within the central nervous system, and to screen for potential developmental problems (Jaščenoka, 2018; Leisman, 2016; Anderson et al., 2013). Research has also explored the relationship between motor and language development, which points to the importance of gross motor functioning in such areas as the education and learning of children (DiDonato Brumbach & Goffman, 2014; Wang et al., 2014). Referring to gross motor assessment Malina (2004) states that
“variation in performance between testing periods probably reflects normal variation in growth (changes in body size and proportions), neuromuscular maturation, opportunity for practice, motivation to perform in the test situation, and perhaps the adults administering the tests and cooperation of young children” (p. 57).
Variability in gross motor scores is, therefore, to be expected and is more reflective of the domain than of the assessment. The observed a statistically significant difference in mean raw scores between different test administrators on gross motor performance is also consistent with the reasons provided by other authors, such as intra-individual differences (Adolph, Cole, & Vereijken, 2015; Eldred & Darrah, 2010).
6. Limitations and Recommendations
One limitation of the study is the relatively small sample size of 53 children. As the Griffiths Scales are used in several other English language speaking countries such as Australia and New Zealand (Sharp et al., 2018), it is recommended that the test-retest reliability be investigated with samples from other countries. A further limitation is that, although the sample was appropriate for a normed developmental test and consisted of children with an uneventful medical and developmental history, the Griffiths III is most often used to confirm a deficit in the child’s development, either generally or in a specific area (Cirelli, Bickle Graz, & Tolsa, 2015). So, it is recommended that the study be replicated with children that have been diagnosed with developmental delays.
The findings of this study, therefore, point to stability within the Griffiths III, as explored through test-retest and inter-rater reliability. The Griffiths III would therefore be a useful test to map the developmental trajectory of children.
Acknowledgements
Our gratitude to Dr. Danie Venter from the Nelson Mandela University for his assistance in the data analysis.