First Investigation .sx In the endeavour to sort out some of the intricacies of this problem the Foundation carried out two small investigations .sx A number of primary schools assisted in these studies and their help and co-operation is gratefully acknowledged .sx The first investigation was carried out in a Junior school in which the children were streamed by age- that is , they were grouped in classes according to age in the contributing Infant school and the Junior Headmaster accepted these groupings and maintained them throughout his school .sx In order to obtain a measure of performance , all the children in each of the four years of the Junior school were given , in the Spring Term 1959 , the Foundation's Sentence Reading Test 1 .sx The median raw and standardised scores together with mean ages and age ranges of the four year groups are given in Table =1 .sx Class 5 was an exception to the rule of 'streaming' by age since it contained all the poor readers from the 2nd year group .sx The standardised scores ( mean 100 , S.D. 15 ) contain an age allowance determined from the sample on which the test was standardised .sx The median standardised score for each year group was then used to determine what would be the expected median raw score for each class within the year group .sx This expected median raw score was then compared with the observed median raw score .sx If the classes had been grouped by age only- that is , within each class there was a full range of ability- then the observed and expected median raw scores should be approximately the same .sx Similarly , of course , the median observed standardised score for each class within a year group should be approximately equal to that for the year group as a whole .sx The results are given in Table =2 .sx In order to complete the analysis the children from Class 5 were distributed into Classes 3 and 4 according to their age .sx It will be seen from Table =2 that there was not a great deal of discrepancy between expected and observed scores for Classes 1 and 2 in the first year group .sx The observed scores for Class 3 , however , are well below expectation , while those for Class 4 are well above .sx The difficulty with Class 5 may have contributed to this discrepancy , but the same phenomenon occurs with the three classes in the 3rd year group and also with the two classes in the 4th year group .sx However , it will be seen that the age 'streaming' is by no means exact- there is considerable overlap in the age ranges for each class in a year group .sx This was due to the presence of a few children only and therefore the results for the 3rd year group were re-calculated and the analysis confined to those children falling within a five month group in each class .sx The results are given in Table =3 .sx It is still clear from the figures in Table =3 that Class 8 is doing considerably better than their average age indicates and Class 6 not as well .sx A satisfactory explanation for these differences could not be found in any differential treatment of the various classes within a year group .sx That is to say , all classes within a year group in the school were treated as 'parallel' and a careful balance of the teaching strength was always maintained .sx Indeed the only plausible explanation seemed to be one associated with the length of schooling .sx Information was obtained on this variable for most of the children in the 3rd year group and this showed that for the most part Class 6 had received 13 terms previous schooling and Class 8 , 15 terms .sx It must be made clear that this investigation was carried out in only one school and a relatively small number of children were concerned .sx However , the results showed that in a school where children were 'streamed' by age , older children scored higher on the average on a reading test than might have been expected according to their age , while younger children did not perform as well .sx Furthermore , the findings were in the direction expected from the hypothesis that an extra term or two of schooling would result in improved performance .sx Second Investigation .sx Following an article in an earlier issue of this Journal on the Effects of Streaming a number of primary school head teachers interested in this topic wrote to the Foundation expressing their willingness to co-operate in any relevant research .sx Although the problem of length of schooling was distinct from that of streaming , these schools together with others , readily agreed to participate in a further investigation which , it was hoped , would provide more definite evidence .sx Six Junior schools in all were asked to administer Sentence Reading Test 1 to all children in their four year groups , and to obtain for each child a record of the number of completed terms previous schooling up to the time of testing .sx A total of 1,604 children were tested and the relationship examined between reading performance and the number of terms previous schooling .sx Before considering any effect due to the age of the children , it was found that , within schools , there was a highly significant regression effect of raw scores on the reading test , on the number of terms previous schooling .sx Using this regression the expected mean reading scores for given lengths of schooling may be calculated and the results of this are given in Table =4 .sx Quite clearly reading test performance is affected by the length of previous schooling .sx After the effects of the age of the children has been eliminated , however , the regression of test score on length of schooling becomes statistically non-significant ( taking the usual 5% significance level) .sx The predicted reading scores for different number of terms schooling after the elimination of the age effect are given in Table =5 .sx A comparison of Tables =4 and =5 shows that the change in score that may be expected for an increase of one term's schooling has decreased from approximately 1.46 to 0.50 points of score , when age is eliminated .sx This second figure is no longer large enough to be described statistically as significantly different from zero .sx The regression analysis was also carried out on the results for each school separately .sx It was found that the regression of reading score on length of schooling remained statistically significant , after the elimination of age , in only one of the six schools .sx This was a large three-stream school which , for the most part , 'streamed' by age within each year group .sx The regression for this school was found to be significant at the 5% level , while the results for the remaining five schools agreed with those obtained on the total .sx These results indicate that while length of schooling is obviously related to reading performance ( r = 0.539 ) there is such a close correspondence between length of schooling and age ( r = 0.979 ) that , to allow for age when considering test score also allows for differences in length of schooling .sx This analysis seemed conclusive enough ; there remained the possibility , however , that any residual effect from the length of schooling might only be apparent during the early years of the Junior school and that by carrying out analyses over all four years the effect was masked .sx These regression analyses were repeated , therefore , covering the first two and the last two years of the Junior school separately .sx Two schools were omitted from these calculations , since their data were incomplete .sx The first repeat analysis was carried out on the first and second years of the Junior school .sx Before the ages of the children were considered , it was again found that , within schools , there was a highly significant regression of raw reading score on the number of terms previous schooling .sx However , when the effect of the ages of the children was removed , it was found that no statistically significant regression remained .sx This regression analysis was also carried out on the first two years of each school separately , but the same result was found for each .sx Thus it is clear that variations in length of schooling have no residual effect on reading performance during the first two years of the Junior school once the effects of differences in age are allowed for .sx For the third and fourth years the regression of reading score on the number of terms previous schooling was also highly significant before age was considered .sx When the effects of age were eliminated , again the regression was no longer statistically significant .sx The repeat of the analysis for each school separately , however , gave results for two of the schools which agreed with those for the totals , but for the other two schools the number of terms previous schooling still had a significant effect on reading scores , even after the elimination of age .sx It will be recalled that most of the schools participating in the investigations were interested in the problem of streaming and in fact only one of the four schools in the last analyses was streamed by ability .sx In this school there was no residual effect of length of schooling .sx One other school practiced no streaming at all .sx It was a large school and except for one 'fast' class in both the third and fourth years , in all other classes the children were grouped at random .sx This school also showed no residual effects of length of schooling .sx In the other two schools , however , streaming , not by ability but by age , was practiced [SIC] and in both these schools the residual effect of length of previous schooling on reading performance was significant after the elimination of age .sx It is interesting to note that this result agrees with the first investigation reported above , in which the school concerned also streamed by age .sx Discussion of Results .sx The number of schools participating in this investigation was fairly small and since the results seem to depend upon the type of streaming practiced [SIC] in each school the findings must be reviewed with caution .sx There appears to be evidence , however , to suggest that under the usual circumstances pertaining in most schools , if due allowance is made for the age of each child when tested , then due compensation will also be given for any differences that might exist in length of previous schooling .sx Where the practice of grouping children into classes according to their age is adopted , the evidence from both investigations reported here suggests that even after due allowance is made for the age of the children , their reading performance still varies according to the length of previous schooling they have received , although this residual effect appears only to be noticeable during the latter part of the Junior school .sx The explanation of these results is not easy to find .sx The fact that this residual effect only appears in the latter part of the Junior school makes the hypothesis that it is due purely to length of previous schooling questionable .sx If the latter is to have an effect on test scores after age has been considered , then it would surely be more noticeable with younger children , that is , in the early years of the Junior school .sx It will be observed , however , that under these circumstances of 'streaming' by age , the older children in a year group , who happen also to be those who have received a longer schooling , perform better than is expected of children of their age , while the younger children in the year group perform below the expectation for their age .sx The possibility must be considered , therefore , that the differential performance effect is not due to any differences in length of schooling , but to the fact that the children are 'streamed' by age .sx Some evidence has already been obtained that one of the effects of streaming is to increase the 'spread' of test performance .sx That is , under the circumstances of ability streaming , more high scores and also more low scores are produced than would be the case were the children not streamed .sx The suggestion here is that since older children of a year group will be doing more advanced work than the younger one , simply by reason of their age , an older stream will give the appearance of being 'better' than a younger stream .sx