Patterns of Young Children's Development: An International Comparison of Development as Assessed by Who Am I? - April 2002

5.1 Age and Schooling Effects

Results on Who Am I? indicate an increase in score with both age and level of schooling. However, these two factors are so closely correlated that it is difficult to determine the effect of schooling independent of the effect of age.

One way of looking at the relative effects of age and schooling is to examine the increase in score by six-month age groups within a school level, as compared with the increase in score in the case of adjacent six-month age groups across two different levels of schooling. In this case, the age difference between the two groups is the same, but the one comparison is between groups differing only in age, while the other comparison is between groups differing in both age and schooling. For each of the achievement measures used in the study, results are reported in terms of mean standardised scores, together with the results of the tests of significance applied and the effect sizes.³

Table 9 provides a summary of the differences in mean scores on both the Copying and Symbols Scales according to relative age within a school level, as compared with differences in mean scores across school levels, where the age difference is similar to the age difference within a school level. Differences in relative age within a school level were calculated for all the samples at all school levels, while differences between adjacent age groups across two levels of schooling were calculated only for those samples that included groups at more than one school level. Differences are in all cases expressed in the form of effect sizes.

Table 9 Effect Sizes of Differences in Mean Scores for Relative Age Within and Across School Levels: Copying and Symbols Scales

Sample	Within Level	Mean Age	Across Levels	Copying		Symbols
				Within	Across	Within	Across
Australia	Mid PS*	4:11		.46		.62
Australia	SY1*	5:11	PS*/SY1*	.40	.56	.56	1.51
Australia	SY2*	6:10	SY1/SY2*	.34	.36	.30	.95
Australia	SY3*	7:70	SY2/SY3*	.16		.10
Canada (NY)	SK*	6:00		.26		.29
Canada (2000)	SK*	6:00		.24		.17
India	Grade 1	5:90		.51		.80
Hong Kong	PS1*	4:10		.01		.10
Hong Kong	PS2*	4:11	PS1*/PS2*	.31	1.67	.00	2.20
Hong Kong	PS3*	6:00	PS2*/PS3*	-.39	.65	-.47	1.44
Sweden	PS1*	5:11		.30		.35
Sweden	PS2*	6:10	PS1*/PS2*	.12	.56	.16	1.28
* PS = preschool, SY = school year, SK = senior kindergarten

From Table 9, it can be seen that differences according to relative age within a school level vary somewhat between the various sample groups, but are in general higher at the younger age levels than at the older age levels. An exception to this pattern is Hong Kong, where there is no consistent increase in score with age within a school level, and in fact a negative effect in the third year of preschool, where the younger children are performing at a slightly higher level than the older children. A possible explanation for the lack of relative age effects in the Hong Kong sample is that the schooling effect is so strong that it overcomes the age effect. However, the sample sizes in this group are too small to draw any firm conclusion from these results.

The effect sizes for differences between adjacent age levels across school levels are, in most cases, substantially higher than the effect sizes for relative age within a school level, indicating that school level has a substantial effect on performance over and above the effect that can be attributed to the age difference. These differences tend to be higher for the Symbols Scale than for the Copying Scale, with effect sizes ranging from .36 to 1.67 in the case of the Copying Scale, and from .95 to 2.2 in the case of the Symbols Scale. These effect sizes indicate substantial and educationally significant differences associated with level of schooling.

5.1.1 Gender

A comparison of scores according to gender indicated a consistent trend for girls to score higher then boys. A summary of these results, expressed in terms of effect sizes, is shown in Table 10. This pattern was consistent across all groups, and is consistent with the research evidence indicating a consistent difference in favour of girls on measures of literacy and early development.

Table 10 Effect Sizes of Differences in Mean Scores on Who Am I? by Gender

Sample	Mean Age	N	Copying Scale	Symbols Scale	Total Scale
Australia, Pre School, June	4:11	857	.34	.35	.39
Australia, First Year	5:11	1355	.29	.35	.39
Australia, Second Year	6:10	1222	.27	.29	.37
Australia, Third Year	7:80	941	.16	.28	.41
Canada (NY) Senior Kind.	6:00	687	.37	.43	.47
Canada (2000) Senior Kind.	6:00	2128	.29	.45	.44
India, Grade 1	5:90	231	.16	.27	.26
Hong Kong, Preschool	6:10	60	.41	.15	.27
Sweden, Preschool	6:60	91	.33	.22	.33
Note: Positive effect sizes indicate that the mean score of the girls is higher than the mean score of the boys.

5.1.2 Preschool Experience

Except in the case of the Indian sample, the data from these studies does not provide any direct evidence of the effect of attendance versus non-attendance in a preschool program on subsequent school achievement. In the case of the Swedish and Hong Kong samples, the children assessed were all at preschool level. In the case of the Australian sample, data on preschool attendance is available, but since preschool attendance is confounded with other factors likely to be associated with higher scores on Who Am I?, this data cannot provide evidence of the effects of preschool attendance as such. In the case of the Canadian sample, information on attendance at preschool or junior kindergarten prior to entry to senior kindergarten was not available.

The relatively higher performance of the Hong Kong sample, as compared with their age-matched peers, suggests that early exposure to a formal preschool program has an effect on the development of early copying and writing skills, as assessed by Who Am I? But, in this case, the effect may be due to the nature of the preschool program, with its strong emphasis on the early acquisition of reading and writing skills, rather than on preschooling as such, and the effect of preschooling may be less evident, or not evident at all, in a sample where a less formal and more play-centred preschool program is adopted.

In the case of the Indian sample, the data provides some evidence of the effects of a special preschool intervention program provided for children from low socio -economic backgrounds. The children who attended this program are comparable, in terms of other relevant socio-economic variables, with children who did not attend a preschool program. Comparing the performance of this group of children against that of children with no preschool experience attending government schools, it was noted that the children who had attended the preschool intervention program scored at a consistently higher level on the Who Am I? tasks, with these differences statistically significant on the Symbols Scale and the overall scale, and effect sizes of .21 on the Copying tasks, .43 on the Symbols task and .40 overall. These results indicate the positive effects of preschool attendance in the case of this sample of children.

5.2 Language and Home-Background Variables

The Canadian North York sample included children from an English-language background as well as children from immigrant backgrounds whose home language was neither English nor French. Data for this sample of children included scores on the Peabody Picture Vocabulary Test — Revised (PPVT-R), which is a measure of receptive vocabulary. Other socio-economic variables, including the education and income level of the parents, were also available for these children. This provided a basis for looking at the relative effects of language background and socio-economic status on performance on the Who Am I? tasks as compared with performance on the PPVT-R in this sample. However, in interpreting these results, it should be noted that this sample is not representative of Canadian children as a whole. It includes a higher proportion of children whose first language is neither English nor French (47 per cent as against the provincial average of 14 per cent and the national average of 10 per cent), as well as a higher proportion of one parent families and a lower than average income level (Connor, 2001).

Correlations between scores on Who Am I? and the Peabody Picture Vocabulary Test - Revised (PPVT-R) for this sample of children are shown in Table 11. The table also shows correlations between each of these tests and other background variables, including language background of the home, age of the child, parental income and parental educational level. These correlations are shown separately for children from English and from non-English-speaking backgrounds, as well as for the total sample.⁴

Table 11 Correlations Between PPVT-R, Who Am I? and Background Variables

Correlations between:	English only (N=318)	Some or no English (N=277)	Total Sample (N=595)
Total raw score: Who Am I? and PPVT-R	.29	.16	.10
PPVT-R and Language Background	-	-	.52
Who Am I? and Language Background	-	-	-.15
PPVT-R and Age	.10	.20	.12
Who Am I? and Age	.37	.49	.40
PPVT-R and Income Level	.42	.20	.46
Who Am I? and Income Level	.06	-.05	-.05
PPVT-R and Educational Level (PMK)a	.27	.23	.28
Who Am I? and Educational Level (PMK)	.36	.21	.10
PPVT-R and Educational Level (Spouse)	.10	.16	.27
Who Am I? and Educational Level (Spouse)	.04	.07	.04
a PMK indicates the primary care giver (or Person Most Knowledgeable about the child), usually the mother.

From Table 11 it can be seen that the correlation between scores on the PPVT-R and scores on Who Am I? for the total sample is only .10. This low correlation can, however, be attributed to the confounding effect of language background. When the correlation is calculated separately for children from an English-language background and for children whose home language is a language other than English, the correlation is higher, although still relatively low (.29 for children from an English-language background and .16 for children from a non-English-language background). This indicates that these two measures are tapping somewhat different skills (word knowledge versus symbolic representation).

The correlations of PPVT-R and Who Am I? scores with age indicate a substantially higher correlation with age for Who Am I? scores (.40) than for PPVT-R scores (.12), indicating that Who Am I? is more closely related to age than the PPVT-R. This is consistent with what would be expected, given that Who Am I? is designed to assess skills which are related to underlying developmental processes rather than specific learning, while vocabulary knowledge, within the limited age range included in this sample, is more likely to be related to culturally determined factors, and, particularly, exposure to a rich English-language environment. The stronger correlation between age and Who Am I? scores as compared with age and PPVT-R scores is consistent for both the English-background and the non-English-background samples. There is, however, some tendency for the correlations with age to be higher for the non-English-background group than for the English background on both Who Am I? (.49 as compared with .37) and the PPVT-R (.20 as compared with .10).

The correlation between PPVT-R score and income level indicates a strong correlation both overall (.46) and for the English language group (.42). This correlation is rather lower for the non-English language group (.20). However, there is no relationship between scores on Who Am I? and income level for the total group (-.05), or for either the English background group (.06) or the non-English background group (-.05) considered separately. This suggests that the skills assessed by Who Am I? are less affected by economic variables than the skills assessed by the PPVT-R. This again indicates that these two measures are tapping somewhat different skills, and a measure of copying and writing skills is not a substitute for a measure of vocabulary knowledge. Each is tapping different aspects of development, both of which are important in children's development and readiness for learning.

The correlations with educational level indicate a tendency for scores on both the PPVT-R and Who Am I? to correlate more highly with the educational level of the primary care giver, usually the mother, than with the educational level of the spouse of the primary care giver, usually the father. This tendency is consistent for children from both English-speaking and non-English-speaking backgrounds.

5.2.1 Differential Effects of Language Background on Language Tests and Who Am I?

The correlations shown in Table 11 also indicate a strong positive correlation between language background and scores on the PPVT-R (.52) in the North York sample. In this sample, there was, however, a small negative correlation between scores on Who Am I? and language background (-.15), indicating that children from a non-English-language background scored higher on Who Am I? than children from an English language background.

The Australian sample also included a small proportion of children from a non-English-speaking background (4 to 5 per cent of the total sample). An analysis of the data from this sample indicated a similar tendency for children from non-English speaking backgrounds to score higher on Who Am I? than children from English-speaking backgrounds, although the English-speaking group scored higher on the language measures included in the Australian study (the Literacy Baseline test and Reading Progress Tests 1 and 2.⁵) A summary of this analysis is shown in Table 12, where the differences in mean score by language background on Who Am I?, and on the various language measures administered, are expressed in terms of effect sizes.

From Table 12, it can be seen that while the children from an English-speaking background scored higher on the language-based measures in both the Canadian and the Australian samples, the children from a non-English-speaking background scored higher on Who Am I?. This tendency for children from a non-English-speaking background to score higher on Who Am I? is, therefore, consistent across these two samples. While the reasons for this are not immediately obvious, these findings add support to the use of Who Am I? as a measure of developmental level which is relatively independent of specific verbal knowledge.

Table 12 Effect Sizes of Differences in Mean Scores on Who Am I? and Language Measures by Language Background: Canadian North York and Australian Samples

	Canada	Australia, by Year of Schooling
	North York (N=595)	Preschool (N=513)	First Year (N=876)	Second Year (N=821)	Third Year (N=659)
PPVT-R	.99
Literacy Baseline test			.18	.45
Reading Progress Test 1					.40
Copying Scale	-.33	-.46	-.22	-.02	-.30
Symbols Scale	-.33	-.34	-.11	.09	.09
Total Who Am I?	-.39	.-38	-.18	.00	-.17
Note: A positive effect size indicates that the English-background group scored higher and a negative effect size indicates that the non-English background group scored higher.

• ³Effect sizes provide a measure of the difference in mean score between two groups expressed in terms of standard deviation units. They are, therefore, comparable across different studies regardless of the actual unit of measurement, and are commonly used to compare results across different studies, as, for example, in meta-analyses. Effect sizes are calculated by subtracting the mean raw score of the first group (the control or reference group) from the mean raw score of the second group (the experimental or comparison group), and dividing this difference by the standard deviation of the control (or reference) group, or by the standard deviation of the total sample (Cohen, 1969). Positive effect sizes, therefore, indicate higher scores for the experimental or comparison group, while negative effect sizes indicate higher scores for the control or reference group. In the case of the comparisons based on relative age, positive effect sizes indicate that the scores of the older age group are higher than the scores of the younger age group, while negative effect sizes indicate that the scores of the younger age group are higher than the scores of the older age group. Following Cohen (1969), an effect size of .20 is interpreted as a small effect, an effect size of .50 is interpreted as a moderate effect, and an effect size of .80 is interpreted as a large effect.
• ⁴Since this sample was drawn from an English-speaking area, the sample did not include any children whose first language was French. For this reason, the comparison according to language background is based on English-speaking versus non-English speaking groups.
• ⁵The Literacy Baseline and Reading Progress Tests 1 and 2 are part of a British series of tests designed to assess reading skills at primary level (see Vincent, Crumpler, and de la Mare, 1996). The Literacy Baseline is designed for administration at the beginning of the first year of school. This test covers pre-reading and early reading skills, including phonological awareness, concepts of print, knowledge of letter names and sounds, recognition of words through matching of picture to word, word to picture and sentence to picture and spelling (six simple words). Reading Progress Tests 1 and 2 are group tests of reading comprehension designed for children at the end of their first and second years of school.