This study began as a study of the prognostic value of certain typical objective tests in selection for secondary education in an African context,but, from the sheer interest of the problems, developed into something larger and more speculative, so that, in retrospect, the title of the thesis appears most inadequate.
The application of the tests prodüced results in line with those obtained in other parts of Africa in showing a serious depression of raw scores. The tests were examined for reliability. The reliability was found to be below that quoted in the test manuals. Part of the reduction may be due to the effects of speed and of function fluctuation, but the effect of limited range of performance is a more serious factor. It was estimated that, in the case of a Moray House Test of intelligence, if the range of raw scores were similar to that obtaining in the United Kingdom, comparable reliabilities would be obtained. The Frimar; Mental Abilities Tests, which are individually much shorter, are consequently less reliable than the foray TTouse Tests. With the actual reliabilities obtaining in the _-population tested, the value of these tests for differential prediction appears limited.
Differential studies in relation to sex and to the occupational status of children's parents were carried out. It seems unlikely that there are any significant differences in performance between boys and girls except in the case of spatial perception. The remarkably low scores on the space test may be related to deficiencies in those experiences, which, in the case of more fortunate children, are alleged to aid the development of spatial perceptual ability, but, since these deficiencies are the same, qualitatively, for boys and for girls, there is much scope for empirical investigation here. Rather remarkably no notable significant differences in ability between different occupational groups were found; such differences as there were seemed sporadic and did not favour "higher" occupational groups. Because the group used in the investigation was already a selected group, investigation on a much wider scale was urged. It seemed possible to the writer, however, that the effect reflected the uniformity of primary schooling, common to all occupational status groups, and uniformity of parental attitude, and that these were more important than economic status.
Prior to investigating the validity of the tests as prognostic instruments, a more recently developed selection procedure, the "multiple cutting -score method" was contrasted with the more usual multiple regression procedure. An empirical comparison of the two procedures was made using data acquired during the course of the main investigation, which confirmed, in the educational field, the results obtained by other workers in the vocational field, that the two methods had approximately the same predictive value, but that the multiple cutting -score method was less time consum- ing. The writer felt, however, that the multiple cutting - score method was more erratic, and that, provided a method of selection of variables, such as that due to Linhart, was used in the multiple regression approach, the saving of time by the multiple cutting -score procedure was perhaps not as noteworthy as other studies had suggested. It was felt that this might be a purely personal preference, and that in cases where the necessary degree of statistical sophistication was not available, the multiple cutting -score method had considerable value.
The method of selecting variables due to Linhart, in combination with a back -solution method for finding the values of the regression coefficients based on the "square root method ", was employed in studying the prognostic value of the tests, a criterion based on a combination of teachers' marks being employed. In one study correlations between the tests and the criterion were worked out for each Form group and were combined using Fisher's "z "; evidence was adduced to show that the correlations were such as would be obtained from a series of samples from the same larger group. In another study the correlations were combined using a formula producmng the result which would have been obtained had the criterion scores been scaled on the Sierra Leone Common Entrance Examination scores as a common variable. In the study first mentioned a battery composed of Moray House Testa of Intelligence and English together with the Common Entrance Examination was selected and gave a multiple correlation of . 0.594; a battery of Moray House Tests of Intelligence, Arithmetic and English, gave a not significantly different multiple correlation of 0.586, but could have been administer- ed in less than half the time occupied by the first -mentioned battery, and in less time than the Common Entrance Examination alone which correlated only 0.434 with the criterion. The
best single predictor was the Moray House Test of Intelligence. The Primary Mental Abilities battery was considered on its own since none of its tests was selected from the total battery by the Linhart method. The tests Vw, Rf, and N were selected as predictor variables with a multiple R of 0.453. This is little better than the correlation between the criterion and the Common Entrance Examination, but it is a quite noteworthy result when it is realised that these three .M.A. tests can be administered in 2I minutes as against I80 minutes for the Common Entrance Examination. In the second study, in which, effectively the criterion was scaled against the Common Entrance Examination, the Moray House Tests of Intelligence, Arithmetic, and English were again selected; the multiple correlation improved to the value 0.650, which, corrected for shrinkage, becomes 0.646.
Again making use of the Common Entrance Examination results, known both for the regression sample and for the general child population of secondary school selection age, correlations corrected for selection were computed and the analysis repeated. The same three variables, M.H.T., 14 T.H.A., and M.H.E. were selected as predictor variables and there was an interesting indication that the P.M.A. picture test, V, might be of value as a suppressant. Without thus using Vp the multiple correlation rose to 0.729; with the use of the suppressant it rose to 0.743. This phase of the study was completed by a cross validation study using a group of boys and girls deliberately excluded from the regression sample for this purpose. Using the regression equations each child's criterion score was "predicted" and the correlation with the actual scores found. in the case of the girls there was actually a small chance gain in predictive efficiency; in the case of the boys a loss. For boys and girls together the correlation between predicted scores and actual scores was 0.620, which, compared with the "shrunken" multiple correlation coefficient of 0.646 is seen to differ by less than the standard error of the multiple R.
The multiple correlations obtained, whilst inferior to many obtained in the United Kingdom in recent studies, are not dissimilar to those found by Emmett (1945) in the Yorkshire investigation. The comparison is felt to be a reasonable one, for in Emr ett's investigation the discrim- inatory efficiency of the tests was stated to be low; although the Moray House Tests used in the present study were later ones, having under United T,ingdon conditions much greater discriminatory efficiency than those used in the Yorkshire investigation, in the Freetown child popul- ation their discriminatory efficiency was not high because of the relatively general poor performance. it is con- sidered that the results obtained provide useful empirical evidence of the value of objective tests under the condit- ions described, and it is suggested that with tests of a similar nature, but designed for more efficient discrim- ination under those conditions, improved predictive eff- iciency would be obtained. Even with the tests actually used, tests of a kind which, it is sometimes argued, are inappropriate for an African child population, better results were achieved than with the well conducted but much lengthier Common Entrance Examination. The fact that the verbal group test of intelligence proved to be the best single predictor emphasises its particular value in selection for secondary education under conditions in which primary education is not all that it ought to be.
The errors made by a group of boys in performance on the verbal group test of intelligence were studied using an interview technique and a classificatory scheme due to Lawrence (1957). It was found that errors arising from "inadequate formulation of the task" were more frequent than any other kind, with errors arising from insufficient knowledge the next most frequent. Selz's theory of error was found of most value in attempting to understand the origin of the errors made. The overwhelming importance of "inadequate formulation of the task" in causing error, combined with speculative theories such as that of Hebb (1949) showing how experience interacts with innate ability in producing "operational intelligence" led to a less pessimistic view of the depressed raw scores actually found, and this, combined with personal knowledge of the state of primary education in Sierra Leone, led to the suggestion that an experiment in providing primary education of the highest quality, with psychological control and extended research into the problem solving behaviour of African children, might produce results which would lead to much needed improvements in Primary Education in Sierra Leone.