 |  |
| www.dg.dial.pipex.com | 643 readers since 19 Aug 2007 |
| |
| www.dg.dial.pipex.com | 643 readers since 19 Aug 2007 |
Plowden (1967) Notes on the text Volume 2 Preliminary pages Foreword and Contents
The 1964 National Survey: Appendix 3 1964 National Survey
Appendix 8 Social services and primary education
Report (full text) about Plowden |
The Plowden Report (1967)
A Report of the Central Advisory Council for Education (England) London: Her Majesty's Stationery Office 1967
Volume 2 Appendix 9
Foreword 1. The survey contained in this appendix was made by a team from the School of Education, University of Manchester, directed by Professor Stephen Wiseman, and is concerned with ten year old children in the Manchester area. 2. The note contained in the rest of this Foreword has been written jointly by the authors of this survey and of Appendix 4. 3. The attentive reader is bound to notice some striking differences between the tables of Appendix 4 and those of Appendix 9. For example, in Table 8 of the latter 'Size of class' stands out. In Table 4 of the former it does not appear at all. Since this can hardly mean that 'Size of class' is peculiarly important in Manchester some other explanation must be sought. 4. In fact the explanation is to be found in the difference between the methods used for summarising the evidence. In both cases the evidence before it is summarised consists of a large number of correlation coefficients. Thus for the analyses between schools in Appendix 4 there were 104 variables and six age and sex groups, giving 33,072 coefficients in all. For the Manchester evidence there were 87 variables, giving 3,741 coefficients. These numbers remind us of the bishops in the Gondoliers, who were in point of fact too many. They must be reduced to put the evidence into a more comprehensible form. 5. In both cases the reduction is made by replacing the original variables by composite variables which are weighted sums of the originals. But different principles are used to obtain the weights. In the factor analysis used to reduce the Manchester data the principle is to choose the weights so that the 87 original variables, which are correlated, are replaced by 87 composites which are not. At first sight this seems little improvement, but the advance lies in the fact that nearly all the variation is concentrated in the first six composites (and about three quarters of it in the first one) so that the rest may reasonably be discarded as unimportant. This is not quite the whole story. There is in fact an infinite number of sets of uncorrelated composites that contain the same information as the original variables, so that further principles of choice are needed. After the factor analysis has been done, the factors are 'rotated' by the computer so as to achieve what is known as 'simple structure': i.e. a set which maximises the number of zero or near-zero loadings in each factor. This produces a simplification of the 87 composites, so that each one now covers only a proportion of the original variables; its effect is concentrated over only part of the total field. It so happens in this analysis that nearly three quarters of the total variation in the educational tests is concentrated in the first factor. A consequence of this is that the weights (or loadings) of the other variables in the first composite (or factor) are roughly proportional to, but rather greater than, the average correlations of these variables with the educational tests, as may be seen by comparing Table 3 with Table 8 in Appendix 9. For example, for 'Size of Class' the average correlation is .32, and the loading is .36. 6. In regression analysis, which was used to summarise the evidence from the English sample, the principle is to select one variable as the criterion and reduce the others to a single composite, namely that composite which has the highest correlation with the criterion. In this process, which is described at some length in Section 4 of Appendix 3, the weights (regression coefficients) are not proportional to the simple correlations, because allowance is made for the extent to which the variables in the composite are correlated among themselves. A variable that has a substantial correlation with the criterion may have a negligibly small partial correlation, when the other variables are held constant, and if so it will have no weight in the composite. Because of this only about a dozen of the original 104 appeared in the composite for each age and sex group. 7. Each mode of analysis brings out important facts. For example in the factor analysis the substantial weight given in the first factor to 'Average size of class' brings out the fact that on the whole schools with large classes make somewhat higher scores in the tests than those with smaller classes. But 'Average size of class' has no weight in the regression composite for any age and sex group. This brings out the fact that when the other variables are held constant it is no longer the case that schools with large classes make higher scores in the tests. This applies generally. 8. There are three stages in inquiries of this kind. First the evidence must be collected. Then it must be reduced to a more manageable form. Finally it must be interpreted. The second stage consists first of the choice of the principles to be used for the reduction, and secondly of straightforward though lengthy computations. The choice of the principles for the reduction is a matter of judgement. This choice determines the nature of the computations, which then automatically give the results. The interpretation of the results is again a matter of judgement. 9. Association, though necessary, is not sufficient evidence of causality. For example, a factor loading -or a first order correlation coefficient - showing a positive relationship between size of class and school attainment does not justify the corollary that increasing class size will improve scholastic ability. A similar association with number of books in the home does not mean that if one made a present of a dozen books to a family, this would improve the child's school work. Winning the pools, and thus dramatically raising the family income, is unlikely to have any advantageous educational effects on the children, despite the correlation between income level and attainment. If the school nurse disinfests the hair of one or two children, this is irrelevant to their results on the next arithmetic test, in spite of the Manchester results on 'cleanliness'. 10. What both analyses show is that adverse factors tend to go together; and that of these adverse factors, the most important are those concerned with parental attitude (encouragement, interest and aspiration) which are clearly more important than social class, occupation or income. School variables are further down the list, and here the adverse factors - like those in the home - hang together, as might be expected. 11. It is when the interested reader looks at individual variables (such as size of class, or streaming) hoping to find support for his own beliefs or ammunition against his opponent, that the greatest care must be exercised in interpretation, and where our warnings against the assumption of causality are most needed. An appropriate approach here is the Bayesian one: to accept that the weight of the evidence needed depends upon the nature of the conclusion. If a conclusion is no more than a quantification of a general proposition acceptable to the common sense of those who know something about the matter in hand comparatively little weight is needed to support the quantification. If, on the other hand, any result appears to fly in the face of such common sense then very heavy weight is needed. Thus a result that 30 per cent of the school achievement was owing to (caused by) their parents' attitudes, and 20 per cent to their parents' material circumstances, is a mere quantification of the view, which most sensible men would accept, that part must be owing to the one and part to the other. On the other hand, a result that, other things being equal, the achievement would be improved (a statement of causality) by giving the teachers more pupils to cope with is contrary to the common sense of sensible men, and therefore needs very heavy weight and much more supporting evidence before such a jump from association to causality can be made. One naturally looks to alternative explanations of the association. Bayes' point that the posterior probability is the product of the prior probability and the likelihood is a most important one, particularly since readers of educational research are often prone to seize on isolated results and either accept them as evidence of propositions that are very hard to believe, or alternatively as evidence that the whole effort is completely worthless.
The Manchester Survey by Stephen Wiseman (1) 12. Our brief for this inquiry was to investigate the relationship between the educational attainment of primary school children and environmental factors, with particular reference to the environment within the school. The design of the experiment was controlled by two major considerations: (a) the experience gained in our secondary school surveys in Manchester and Salford in 1951 and 1957 (2), and (b) the necessity for a strict timetable, with the autumn of 1965 as a deadline for the completed Report. This limitation in time affected both the size of the sample, and the type of criterion measure adopted. It was clear that in the time available it would be impossible to construct specially designed measures for our purpose, and we had to rely on existing measures - measures which must, of course, be directly comparable, school with school. We decided to study ten year old children, because these children had received the full impact of primary school education. The ten year olds in 1964 had taken standardised tests at 7+, 8+, 9+ and 10+, and the results of these tests were accepted as our criterion measures. There were 12 such measures:
13. The limitations of criterion measures of this kind must be stressed. They consist of tests of the tool subjects only, and, with the exception of composition, they are all objective tests. Such measures are a necessary part of any criterion of primary school attainment, but they are far from giving a balanced picture. No account is taken of the many other activities which go on in the average primary school, the work in art and craft, in music, in science and geography and history, and in the many other pursuits which can often not be classified under the traditional 'subject' labels, but which, nevertheless, are of very great significance educationally. Our choice of criterion measures must not be taken to indicate a belief that the 3Rs are all that matter, nor should the limitations of our tests be forgotten in interpreting the results of our inquiry. 14. Our experience in earlier researches, and the fact that we were particularly interested in the environment within the school, led us to plan the research as a school based one, with the schools as the units. We were anxious, however, to support this main inquiry with a more intensive study of individual children, and we were glad to discover that we could draw on the services of the Social Survey to interview the parents of such a sub-sample. Our inquiry, therefore - and this Report - covers first, the main research comparing schools, and second, a sub-sample research comparing pupils. Sample of schools 15. In 1964 there were 176 primary schools in Manchester with children of 10+. Ninety-two of these were county schools, 40 Church of England, 43 Roman Catholic and one Jewish. It was decided to take a 25 per cent sample of these schools, stratified by school type. The two other factors which it was desirable to control in the sample were socio-economic level and school size. Our previous surveys had given us a good deal of information about the distribution of social factors over the city, and it seemed that these could best be controlled by stratifying the sample on a geographical basis. The central area of the city contains the blackest region from the point of view of housing, crime, disease and poverty. The northern extremity - hardly to be called suburban, since in this direction there is no break in the conurbation - shows a significant improvement over the central area in all social variables, while the southern wards (Didsbury, Withington and Burnage, etc.) contain the 'best' areas within the city boundary. These three divisions of the city could well form the basis of a socio-economic stratification. There remained the extreme southern section formed by the Wythenshawe wards of Northenden, Baguley, Benchill and Woodhouse Park. Our 1951 and 1957 results showed this area to be 'white' on such variables as J-index, Persons per acre and Death rate, but as 'black' as Moss Side on Birth rate, Committals to Care, Cruelty and Neglect, and Children on Probation. A high proportion of residents are from slum clearance areas, and the region is clearly an atypical one and deserving of separation from the rest of the city. As a result of considerations such as this, we stratified geographically as follows: 7 Northern wardsWhen the 176 primary schools are classified under school type and social district the picture is as shown in Table 1: Table 1 All Manchester primary schools 16. In order to control size of school we decided to draw a 50 per cent sample of schools in the first instance, to arrange the schools within each cell in the table in order of size, and to take alternate schools for final 25 per cent sample. This yielded the final sample of 44 schools, shown in Table II, which provide about 2,000 ten-year-old children: Table 2 Sample of tested schools 17. Three additional schools were chosen, one County, one CE and one RC, to form pilot schools for Mr Derrick to visit first, to try out his interviewing and recording methods before starting on the sample proper. Social survey sub-sample 18. Interview load limited the size of the sub-sample of children whose parents were to be interviewed by the Social Survey to 200: i.e. a one tenth sample. Two plans were discussed: (a) a one tenth sample of all the 2,300 children in the main sample,Plan (a) covered all the schools, but at the cost of providing, on average, only five pupils per school, and entailing a wider geographical spread for the social survey interviewers. Plan (b) concentrated the interviews, but inadequately covered the schools. And since our inquiry was to be, in the main, school based, this was a serious disadvantage. Finally, and with some degree of hesitation, we settled on the following design: (a) a randomly selected 50 per cent sample of the 44 schools, stratified as in Table 2, 
I THE SCHOOLS ANALYSIS Criterion measures 19. Our previous surveys of secondary schools were all school-based, and we had then discovered the value of employing measures of attainment which covered intensity of 'backwardness' and 'brightness' as well as the average level of ability in a school. As I have said elsewhere, 'this gives a richer and more meaningful result than using a single average score. Certainly from the teacher's point of view these two percentages mean much more than an average score, since they demonstrate so clearly the teaching problem by indicating the spread of ability. Two schools may have identical average scores, but if one has 10 per cent backward children and 10 per cent bright, while the other has one per cent and one per cent, they are very different schools to the teacher'. (3) We used the same - arbitrary - measures of 'backwardness' and 'brightness' as we had used earlier: the percentage of pupils with a standard score of 85 and below, and of 115 and above. We thus had twelve tests, with three measures from each test: a total of 36 criterion scores. School environment 20. Professor Warburton's 1951 inquiry in Salford has shown that subjective estimates of 'progressiveness', socio-economic status and quality of school buildings could be more valuable than the objective measures of attendance, class size and date of school building. This finding led us to search for ratings that might reasonably be expected to pinpoint some of the elusive components of 'school atmosphere'. As a result of preliminary meetings with the head teachers of the primary schools involved, and discussions with representatives of the LEA and with colleagues in the University, we finally decided upon 34 school environmental variables of which 16 were at least partly subjective. 21. Most of the information on the straightforward, objective variables was supplied by the head teachers in reply to questionnaires. Size of class, size of school, and percentage attendance, which had proved useful enough in Salford, were obvious first priorities. 22. So many schools are experiencing staffing difficulties these days that we decided to include some variables that might reflect particular aspects of this problem. Articles in educational journals and letters to the press constantly comment on the damage to our children's schooling caused by too many staff changes, and by the need to recruit large numbers of married women whose first loyalties may be to their own families. One reads, too, of the trials facing newly qualified young teachers, especially in 'tough' areas. There is discussion, too, about the balance of the sexes in teacher training, and the desirability of training more men than previously for primary work. Despite the fervour of many writers on these problems, very little is actually known about the influence of different categories of teachers on primary schooling. We therefore included the following variables in our survey to see if they linked up in any way with measures of academic achievement or school atmosphere: Proportion of teachers under 30 years of age, Proportion of teachers over 50 years of age, Turnover rate, Percentage of graduates on the staff, Sex of the head teacher and proportion of men on the staff. From the information supplied by the questionnaire, it was possible to work out the teacher/pupil ratio. From the heads we also learned how many pupils in each school had qualified for some form of special schooling since September 1962. We asked for the various types of special schooling to be detailed in the hope of having additional checks on any areas where physical handicap appeared unduly heavy; in practice, however, the great majority of children involved were the educationally subnormal, and the numbers were in any case low, so that it was not practicable to separate out the different categories. In addition, we obtained the number of incidents of breaking and entering school since January 1960 (strictly speaking, a 'neighbourhood' rather than a 'school ' variable) and, at the suggestion of the head teachers, the number of pupils whose mother tongue was not English. These children are causing a growing teaching problem in many of our larger cities, although perhaps it is not in its most acute form in Manchester. 23. From information supplied by the LEA, we calculated the average ratio of grammar: secondary modern places gained by pupils in each school over a period of four years. 24. The ratings on variables associated with 'school atmosphere' were made by the research officer, who spent a day in each of the 44 schools of the sample, after trying out his procedure of interviews and ratings on the three pilot schools. In each school he discussed the home backgrounds of the children with the head and with the staff as a body, but perhaps the most valuable information was gleaned during 15-20 minute visits to nearly every classroom. Naturally, the children's own class teachers tended to know more than did the heads about which children's backgrounds, physically or psychologically, were poor or atypical. Our observer then chatted to several of the children, either individually or in groups of two or three, as unobtrusively as possible, while the normal classwork continued around them. Some of these children had been pointed out by their teachers; others he chose at random. Our observer's talks with the pupils, of course, appeared to them quite informal, but many of the topics he raised were chosen to provide information, directly or indirectly, on the home backgrounds, the leisure opportunities of the neighbourhood and the school atmosphere. For instance, he frequently asked the children what they liked best out of school and where they went to play, who helped them to read, and whether they chose their neighbour in class (to provide a lead on the degree of formality in the classroom organisation). While in each school he also visited the library, the gymnasium, and the art room. 25. After each visit, he rated the school, usually on a five point scale, on the following variables: (a) First impression of the school: influenced by cleanliness, decoration and any indications of the general atmosphere;Some variables reflected school organisation and policy: (f) Recently reorganised from all-age. It seemed possible that in some recently reorganised schools, the ethos of the secondary school might tend to continue in the primary school;The other observer ratings were all planned to capture aspects of the 'school atmosphere'. (k) Appearance and sociability of the children. In this purely subjective rating, the most important factors were the attitudes towards, and verbal skills of the children in conversation with, the observer. 'Appearance' was reckoned only in terms of neatness and cleanliness - no school in the central area was rated low if the children had worn, but clean, clothing;26. We had planned to include a few variables about Parent-Teacher Associations since this seemed an obvious way of assessing objectively one aspect of parental interest in the school. To our surprise, not one school in the sample had such an organisation. Home and neighbourhood environment 27. In our previous Manchester surveys, social variables were calculated on a ward basis; in 1957, for instance, we allocated each child to the ward of his home address and his 'score' on each social variable was the rating of his ward. This time, however, we concerned ourselves only with the children and families represented in the schools of the sample. Naturally, this approach limited the environmental variables we could use. Birth rate, Illegitimacy, Deaths under one year, useful though they had been in 1957, had, of course, to be dropped. Fortunately, however, many of the variables that had been most closely linked with educational success in 1957 could be calculated on a school, as well as on a ward, basis, and so were used in the present research: Children taken into care, Verminous conditions, Distribution of shoes and clothing either Free, at Part-payment, or at Full cost (the last category applying to families on National Assistance who still save money as the clothing is cheaper than it would be retail). Head teachers, Medical Officers and Welfare Officers, as seemed most appropriate in each case, informed us of the numbers of children in each school to whom these variables applied. Each head teacher also gave us the number of pupils in his school receiving free meals, another variable clearly related to the economic level. 28. Independent assessments of the standard of housing in the school neighbourhood were made by the School Health Department and by the observer during his visits. The school nurses had the advantage of some knowledge of the interior state of the houses while our observer could view them only externally. These two ratings were pooled. They were very similar to the rating of 'School neighbourhood' which proved valuable in Salford. 29. To obtain some social data for individual pupils we supplied both the Medical and Welfare departments with a list of the names of the children in the sample and asked them to mark those whose homes, in their confidential opinion, were suffering from any of the following conditions: (a) Shortage of material needs: the food, clothing, living space that an average home provides;We were especially interested in the correlates of such disrupted homes since Fraser (1959) had shown this factor to be particularly detrimental to school achievement. 30. Four other pieces of individual data were related to the socio-economic level: Father's occupation (rated on the Registrar General's five point scale); whether the child's mother worked, and whether the family lived in a Corporation house. One physical variable, the child's height was also included. 31. The Manchester police, too, were most cooperative in allowing us access to their records. From these, we found out which children in the sample had already been Convicted or officially cautioned, and those cases in which another member of the family, or someone living in the same house had a police record. Difficulties arose when someone of a different surname, but with the same address as a child in the sample, had a record, especially a recent record. There was no way of distinguishing between members of the child's immediate family circle (e.g. a foster brother, a stepfather or maternal cousin), lodgers, co-tenants and previous tenants. Hence our variable Criminal record of house address probably contains a pretty heterogeneous sample. Carter and Jephcott (1954) in Radby, and research being at present carried out in Wythenshawe, have shown how there can be 'black' and 'white' patches, or even streets, lying cheek by jowl in the same district. As an additional neighbourhood measure we therefore calculated the number of families, in which at least one member had a conviction, living within 50 houses of the home address of each child in our sample (Neighbourhood crime). A full list of all variables, with their bases of classification, is given in the appendix. RESULTS A. CORRELATION ANALYSIS 32. It will be seen that our research had grown to a formidable size by this time. Including the test results (remembering that there are three separate scores for each school for each of the 12 tests) we now had 87 variables. The first job was to calculate intercorrelations for the whole matrix. This showed us which of the environmental variables had a significant association with the results of the educational measures. Perhaps the easiest way of summarising the data is to calculate the average correlation of each variable with all the 36 test variables. (This was done ignoring the signs of the correlations, since the signs for backwardness and brightness are in opposite directions.) Mean correlations 33. The highest correlation (.622), as might have been expected, was with the four year average of 11+ success, but this might reasonably be regarded as a criterion variable rather than an environmental variable. Of the remaining mean correlations, 17 are found to be over 0.3. Ten of these are with neighbourhood variables, seven with school variables. This stronger connection with the social factors in the home and neighbourhood is brought out even more clearly if we average the correlations of all the 18 social variables with the total tests. The result, .295, contrasts with a similar average for the 34 school variables, .212. And if the quasi-criterion variable of 11+ success is omitted, the average school variable correlation drops to .199. It seems as if the forces operating outside the school walls are more pervasive and more powerful than those within. The table below lists the 17 correlations over O.3: Table 3 Average correlations of particular environmental variables with the 36 test variables 34. The emphasis on measures of dirt and of poverty support the results we found in our earlier surveys of secondary school pupils. The inclusion of two crime variables emphasises the importance of the moral and psychological atmosphere of the home and neighbourhood, even at the primary school stage. 35. Our interest, however, is particularly in the school environment. The entries in the right-hand column are headed by Appearance and sociability, a rating made by our observer, and one which attempted to disregard socio-economic level. It is, perhaps, a measure which should be thought of as straddling the 'school' and 'neighbourhood' columns: no doubt the school may have some effect here, but the home is almost certainly the major influence. This variable is closely followed by Attendance (.414) - a logical result, though perhaps some might have expected a higher correlation. Another logical association is that with Children qualified for special schools (.360). 36. The third most powerful association in the school list is one of particular interest, that of Streaming (.385). The direction of the association should be noted: streamed schools have better records of attainment. An average of .385 means that some individual tests have much higher correlations with streaming than this: for example, brightness, 10+ composition, .659, brightness, 10+ arithmetic I, .604; brightness, 9+ English .597. A study of the individual correlations shows that the effect of streaming lies mainly in increasing the proportion of bright children in a school. Of the 10 highest correlations (all over 0.5) seven are with measures of brightness, and three with average score. There is only one correlation greater than 0.4 with the measures of backwardness: 10+ Arithmetic I (-.457). But there is no suggestion that streaming has an adverse effect on the children of low ability: streamed schools tend to have fewer backward children than unstreamed schools at all ages and in all tests. The 12 measures of backwardness have an average correlation of -.237 with streaming, compared with .471 for the measures of brightness. 37. The positive correlations between attainment and class size (.316) and school size (.315) indicate that large schools and large classes are associated with good attainment: a result which supports other researches in this field (e.g. Kemp, and Morris). Many schools were unstreamed merely because of small size. A one-form entry school is almost certain to be an unstreamed school. And in view of the positive correlation between school size and educational attainment, it may well be that the relationship between streaming and attainment is caused by this third factor. A correlation of .655 between streaming and school size is found, emphasising the relationship already mentioned. The size of this coefficient makes it all the more necessary to check on the possibility that our streaming/attainment association is an artefact. We therefore calculated the partial correlation between streaming and the mean results of all the tests at the 10+ age level, holding size of school constant. This analysis reduced the correlation from .464 to .326, but it is still significant, even when the effect of school size is removed. 38. The positive association between educational attainment and size of class is one unlikely to carry conviction with the practising teacher, and may well be that this is an example of the effect, already noted, that bedevils all research of this kind - an artefact of correlation arising from the association of each variable (class size, and educational attainment) with a third more pervasive variable (the socio-economic level of the school neighbourhood). The central area of the city is the area affected by slum clearance, and many of the (small) schools still existing there produce class-size figures much lower than those of the new and larger primary schools in the well-to-do suburbs. In order to investigate this possibility, we calculated some partial correlations, by which means it is possible to eliminate the effect of a third factor. We took the average correlation between all the 10+ tests and class size (.372) and calculated what this would be if the variables of parental occupation were held constant. The correlation dropped, but only slightly, to .317. We then tried again with housing standard partialled out. This produced a slightly greater fall, to .301. Finally we tried free meals, which produced a partial correlation of .296. It looks, therefore, as though there is a small, but positive, association between size of class and educational attainment. What is abundantly clear is that this variable is much less important than the quality of the home and the standard of the neighbourhood. 39. The final correlation listed in the right hand column of the tables is that of .303 for homework. Schools which give homework (at the 10+ level) tend to produce higher scores in the tests. The pressures for homework in the final school year often come from the middle class parents in the outer suburbs, and it may be that the .303 correlation is merely reflecting the more pervasive social factors. This is unlikely, however, since the r between occupation and homework is only .107. Nor does the partial correlation, holding parental occupation constant, give any support to this theory; no reduction is produced in the correlation between homework and 10+ attainment. Age trends 40. So far we have looked at the relationships revealed by a study of the mean correlations over the whole of the 36 educational measures. But these measures cover four age groups (7, 8, 9 and 10) three types of test (intelligence, English and arithmetic) and three methods of assessment (school mean score, per cent of 'bright' children, and per cent of 'backward' children). These breakdowns enable us to look a little more closely at the effect of environment on attainment. First, let us consider age: what differences appear here? The average correlations rise as the children get older, as will be seen from Table 4. The same trend is shown for both neighbourhood variables and school variables. Table 4 Average correlations for each year of age 41. Over and above the general trend shown in the table, there are interesting results from some of the individual variables. The greatest increase with age is found with streaming (.17, .39, .39, .46), with school size (.14, .35, .33, .36), with class size (.18, .32, .30, .37) and with homework (.14, .28, .39, .35). What are even more interesting are those school variables which show no increase with age, or even a decrease - such as teacher turnover (.14, .15, .06, .03), school equipment (.21, .20, .14, .13), and proportion of teachers over 50 years of age (.13, .09, .13, .06). Although these correlations are low, the absence of the prevailing age trend might be suggestive. 42. When we turn to the home and neighbourhood variables, those showing a steeper rise than the average are free clothing (.22, .34, .41, .43), per cent of verminous children (.34, .49, .55, .49), and material needs (.34, .40, .51, .48). Thus the effects of poverty and dirt appear to be cumulative. Those variables showing no rise with age are - curiously enough - disrupted homes (.26, .10, .20, .18) and crime - house address (.14, .08, .14, .07). 43. The general tendency for the correlations to rise with age merits a closer examination. There is some doubt as to whether this represents a true picture. It must be remembered that these correlations are between the test results of children who were 10 years of age when the environmental variables were measured, so that the 10+ test results and the other measures are coincident in time. But the 7+ test results are those of these same children three years ago, the 8+ two years ago, and so on. It would not be surprising, therefore, if these correlations have been subject to some degree of shrinkage. If so, the possibility exists that the level of correlation between attainment and environment does not rise with age. It may even fall: our results cannot be used to disprove such a hypothesis. A comparison of our results from primary school with those from secondary schools in 1957, studying pupils of 14+, suggests that this might well be the case. The correlations of attainment with environmental variables all tend to be higher at 10+ than at 14+. Only two of the social variables were identical in the two surveys: per cent of verminous children and per cent in receipt of free shoes and clothing. The average correlations of the former with attainment are .483 at 10+, .383 at 14+. For the latter the figures are .237 and .173. The fall from 10+ to 14+ is apparent for all three types of test, with the change for arithmetic being greater than for English, the smallest change being with the intelligence test. 44. The suggestion that the impact of environmental factors on attainment gets progressively weaker as we go up the age range is supported by other evidence. Furneaux's study of university selection (4) finds that 'the proportion of those wishing to have a university education who are actually able to apply is much the same for pupils in all occupational groups', and again, 'among pupils who have already reached the upper sixth form, differences of home background are only associated to a very small degree with the strength of the desire for a university education'. Much the same was found by the Robbins Committee. Their Table 7 (Appendix I, Robbins Report) shows that 'within each ability group at 11+, there is no significant difference in performance [at A level] between children from the different [social] classes who stay on'. Types of test 45. Let us now look at average correlations for the three types of test. Table 5 shows these for all variables, and also separately for home and neighbourhood variables and school variables. Table 5 Average correlation for each type of test 46. Overall, the English tests show a higher mean correlation with the environmental variables than do the intelligence tests or the arithmetic tests. When we consider 'home' and 'school' variables separately, however, it is seen that intelligence has a stronger relation with the 'home' variables than the two measures of attainment, and this is in marked contrast to the results for the 'school' variables. The differences shown in the last two lines of Table 5 may be considered small, but it should be remembered that these are average coefficients (taking due regard of sign) and are based on a considerable number of single correlations. The figure of .190, for example (the mean correlation of IQ with school variables) is derived from 306 separate correlations, while the entry .215 for English is based on 408 coefficients. The mean correlation for IQ with home and neighbourhood variables may be regarded as the least stable of these figures, but even this is derived from 162 separate coefficients, while the 'school' and 'arithmetic' figure is based on no fewer than 510. 47. It is interesting to speculate on the possible reasons for the different emphasis found for the measures of intelligence. Why should lQ be more closely connected with the home and neighbourhood environment? A possible explanation lies in the stronger genetic element in the results from such tests, and the correlation between the intelligence of children and that of their parents. The factors in the environment which affect parents (and which are affected by them) are likely to show more association with children's IQ (correlated with parental IQ) than are those factors which affect children only (i.e. most of the school factors). In other words, the adverse social environment tends to contain more parents with low IQs, whose children tend to have low IQs. When we look at the individual correlations, there is some support for this. The strongest gradient on the IQ, English, Arithmetic line is found for poverty and crime: Neighbourhood crime: .401, .275, .266; Free meals: .544, .491, .410, Family crime: .456, .388, .325. A reversal of this trend is found with the highly individual variable of mother working: .197, .262, .253. Turning to the school variables, the strongest gradient here - in the opposite direction, of course - is found with Streaming: .293, .468, .351, class space: .245, .315, .252 and class size: .272, .363, .305; while a reverse effect is found with Appearance and sociability: .480, .432, .380 - surely a variable which is closely connected with parental care and upbringing. Backwardness and brightness 48. We might look now at the differences revealed between average correlations for backwardness and brightness. It will be remembered that some of the most interesting results from our secondary school survey were related to differences between these measures. Table 6 shows the average correlations. Table 6 Average correlations: Backwardness and Brightness 49. It will be seen that the environment appears to have more effect on brightness than on backwardness, and that this differential is rather more marked with the school variables than with the home and neighbourhood variables. There are, however, some interesting exceptions to this general trend. Neighbourhood crime correlates .326 with backwardness, only .243 with brightness; Family crime shows the same trend, .394 and ,315. Among the school variables five can be noticed which correlate more highly with backwardness than brightness: Appearance and sociability, .424, .383; Progressiveness (LEA rating), .233, .157; Teachers under 30, .185, .084; Teachers over 50, .108, .021; and Corporal Punishment, .100, .000. Only four variables however (excluding 11+ success) show mean correlations greater than .4 with the measures of backwardness: cleanliness (.443), appearance and sociability (.424), verminous children: .419) and free meals (.415). In contrast with this there are 10 correlations over .4 with brightness:
50. The strong connection with poverty and dirt is clear, as it was in our 1957 survey. 51. It should be emphasised that these results have been obtained from an investigation of schools and not of individual children. Nevertheless, the association between brightness and environmental factors is a particularly important one. The consistency of the findings from three separate surveys in Manchester, together with supporting evidence from Burt and Fraser, point to the virtual certainty that an adverse environment has its greatest educational effects on children of above average ability. If we can discover how to counteract such effects - even partially - then the educational profit will be very large. B FACTOR ANALYSIS 52. So far we have been looking at correlation coefficients. We have tried, by means of averaging, to see what general trends are discernible. In this research, with 87 variables, there are 3,741 individual correlations to be related to each other. This calls for the drawing of many millions of comparison between pairs of coefficients. Moreover, as we have already become aware in the foregoing section, the degree of correlation between two variables A and B is by no means the only information contained in a correlation table, since A and B may also agree in that they both correlate highly with certain other variables, C, D, E and F, moderately highly with G, H, I and J, not at all with K, L, M and N, and negatively with O, P, Q and R. There are untold millions of complex comparisons to be made, a task beyond human competence. This can be performed, however, by the process of factor analysis, which replaces scores on the 88 variables, which are partly dependent on one another, which are of relatively small statistical importance (variance), and which have low powers of implication and prediction, by scores on a small number of factors, which are completely independent and which have far greater predictive and implicative powers than any of the individual variables. Although other methods of analysis can be used (and will be used in the later phases of this research) such as the comparison of the mean scores of certain subgroups (analysis of variance), such techniques are much weaker for explanatory purposes, since the investigator himself decides on the crucial 'factors' and no tests of the relative weight (as opposed to the 'significance') of variance factors can be obtained. Nevertheless, in following up results for the more interesting individual variables, the technique of variance analysis and regression analysis should be of considerable value. Its application to scores for the major factors revealed in the factor analysis is an obvious example for its application. To begin with, however, an ordering of such a complex (and relatively unexplored) field by factorial techniques is essential. 53. The Atlas computer produced for us a Principle Components analysis, followed by a Varimax rotation of the first 10 factors. A scrutiny of this showed that six of the 10 factors covered 93.6 per cent of the total variance on the educational tests: the following analysis deals only with these six factors. Table 7 Percentage variance on the 12 tests contributed by 6 factors 54. As will be seen from Table 7, the first factor is overwhelming in its importance. It covers 72 per cent of the total educational variance, leaving only 22 per cent for the remaining five factors. From a practical point of view, therefore, its effect is all important. From the standpoint of a research worker, however, the patterns revealed in the other five factors are of particular interest, and we must certainly not ignore them. In the remaining sections of this report, however, it must be kept in mind that factor I carries more than three times the weight of all the rest put together. 55. A table giving the full Varimax analysis is given in the Appendix. Table 8 (5) summarises the major loadings on the six factors listed above, and also shows, at the bottom of the table, the way the total 'educational' variance is distributed over age, type of test, and level of performance. One of the interesting things about Table 8 is the absence of some 12 of the environmental variables: absent because they have no sizeable loadings on the six factors which are associated with educational attainment. The absence of some of these may be viewed with some degree of equanimity: variable 6, for example, Mother tongue not English. The maximum incidence of such children in our 44 schools (see Appendix - List of Variables) was 13.6 per cent: a clear indication that this is, as yet, far from a major problem in this city. And the factor analysis shows that this level of incidence has little or no degree of association with level of attainment. It is more surprising, however, to find that 22 Quality of head teacher and 23 Quality of staff have insignificant loadings on the 'attainment' factors. A study of the full Varimax analysis shows these variables with heavy loadings in Factor II: but this contributes only 1.35 per cent to the test variance. Our two measures of Progressiveness (24 and 25) are also missing - and here we have a contrast with our secondary school survey, where Professor Warburton found a significant association between Progressiveness and both reading and arithmetic. The loading of .26 on Factor I is in the right direction, but is disappointingly low. Table 8 Summary of factor loadings Factor I 56. Let us now look more closely at the major attainment factor. A glance at Table 8 reveals that the major loadings are concentrated among the Home variables, with the School and the Neighbourhood much less involved. Indeed, if we calculate the amount of 'environmental' variance contributed by each of these categories to Factor I (making appropriate adjustments to take into account the fact that there are different numbers of individual variables in each category) we find that School effects contribute 18 per cent, Neighbourhood effects 20 per cent, and Home effects no less than 62 per cent. This is a finding of the greatest significance, and underlines with considerable emphasis the problem facing the primary school. The adverse forces in the home are, at this stage of education, the overwhelming ones, and the lower loadings (and the fewer loadings) on the school variables suggest that in the present circumstances we are doing little to counteract them by forces within the school. It is surprising to find that none of the teacher variables have loadings over .3 on this factor: the major effect seems to lie in the direction of organisation and policy, with streaming and homework as two of the outstanding variables. Size of class, too, bears out the association already noted in our correlational analysis, with schools with large classes tending to have better standards of attainment than those with small classes. 57. When we focus our attention on the crucial sector of Home, we get again the close connection between attainment and both poverty and maternal care. This is seen very clearly if we list the major loadings in order of size:
58. These figures, together with the likelihood (already noted) that the early years bear the heaviest impact of adverse environmental forces, suggest two obvious lines of attack in our efforts to rescue the underprivileged and socially handicapped child from the worst effects of his adverse environment. First there is the provision of nursery schools, so that some educational counter-measures may be applied before the age of five. And such provision is seen not so much to counteract the poor physical environment of the home as to begin the process of adaptation to the school and the teacher, and with the hope that it might affect the child's attitude towards education. The continued appearance of the maternal care factor suggests that the underlying mechanism may lie in the formation of what McDougall called the 'self-regarding sentiment', with all that this implies for the formation of attitudes towards education and towards authority. If this is so, therapeutic action before the age of five is essential. 59. The second line of attack is to support and encourage a recent new development in the field of teacher training. A few colleges of education have begun to offer courses which include some element of training in social work. The results of this research underline the desirability of increasing the number of such courses. The aim should not be that of trying to produce fully-trained social workers, but rather the training of a proportion of teachers with some knowledge of (and first-hand experience in) adverse environments, with some appreciation of the problems and techniques of social work, and with the attitudes necessary to achieve effective cooperation with social workers and social agencies. If local authorities possessed a teaching force which included, say five per cent of teachers trained in this way, they could expect them to become familiar with the home backgrounds of the children in the most difficult areas, and to engage in liaison work between home and school. The most difficult cases would, of course, be beyond them, but part of their function would be the important one of early referral to more expert and more capable agencies. It may be argued, with some justification, that as yet we do not know enough about the adverse forces and their method of operation to be able to mount such training courses with any degree of certainty about the validity of method and content. But this is being too timid and too conservative. The development of more courses of this kind will assist in the development of new knowledge and new insights, and if they tend to show differences in approach and in organisation this may be all to the good at this developmental stage. Minor factors 60. As we have emphasised, Factor I is by far the most important, carrying nearly three quarters of the total variance on the educational measures. The other five factors are, however, of considerable theoretical interest, and, as we shall see, the effect of some of them to particular directions is far from negligible. 61. Let us look first at the partitioning of the 'educational' variance shown at the foot of Table 8. Factor VII is clearly one which is important for Brightness, and Factor III supports this in a smaller degree. Between them they account for 20 per cent of the Brightness variance. Factors V and X (and IX to some extent) have their greatest effect at the 7+ and 8+ ages, with Factor V concentrating very heavily on Arithmetic. 62. The age pattern is of particular interest. Quite clearly the pattern of abilities changes radically between the age of 7 and 10. Notice how only 52 per cent of the 7+ variance is covered by the major Factor I. Environmental forces appear to be much more complicated at the earlier ages. 63. The summary at the foot of Table 8 shows the way the educational variance is spread over the six factors. We can do an equivalent analysis on the environmental variance. It is a little more complicated because of the difference in the number of individual variables in the categories School, Neighbourhood and Home. But if we adjust for this, and if we take the total environmental variance on these six factors only as being 100 per cent, we arrive at Table 9. A comparison of this with the breakdown of educational variance shown in Table 8 is most interesting. Notice how Factor I, with its lion's share of educational variance, is now overshadowed by Factor III. The other four factors are all about the same level of weighting, nine per cent or 10 per cent each. The side totals show roughly equal contributions by school, neighbourhood and home (but notice that that the school is lowest of the three) although individual factors show very obvious differences here. Table 9 Percentage allocation of environmental variance among six factors 64. If we consider the top line of the table, we may analyse a little further, by breaking down the variance on the School variables into the three divisions already employed in the upper part of Table 8. This breakdown is shown in Table 10. Table 10 Percentage allocation of variance on the School Variables among six factors 65. We see from the right hand totals that the physical aspects of the school environment have a much greater importance than either the teachers or the organisation and policy, and that this effect is concentrated very heavily in Factor III. Factor X is the one which is almost entirely concerned with teachers - and this, it will be remembered, contributes only two per cent of the variance on the educational tests, although it is significant that its effects are concentrated on the younger children. Factors associated with age 66. Table 8 shows us that Factors V, IX and X have much stronger effects at 7+ and 8+ than at the upper end of the primary school. Factor V has its heaviest loading in 17 teacher turnover rate. Indeed, this is the only factor out of the 10 which shows any connection with this variable. And it is logically sensible that high rates of teacher turnover are particularly deleterious to the younger children (and particularly important for attainment in arithmetic). It is, however, a little surprising that the ten year olds seem almost completely unaffected by this. Other 'school' loadings on Factor V are for 38 School equipment, 36 Age of building (both in the 'wrong' direction, curiously enough), 37 Quality of building and 16 proportion of Graduate Teachers. This last is particularly interesting, since it is supported by Factor X (also 'young children' factor) and is in contrast to Factor I where a lowish loading on this variable is negative. Why should graduate teachers in primary schools have a positive effect at the lower end of the age range, and a negative effect at the upper? Many people would believe that, if any such differential existed, it would be in the reverse direction. The presence of married women graduates, with young children of their own, may have a bearing on this. 67. Factor X is almost entirely a 'teacher' factor, and from this it will be seen that attainment at the 7+ and 8+ levels is associated with older rather than younger teachers, with women teachers rather than with men teachers, and that the heaviest loading of all is for the proportion of married women teachers with children of their own. It is heartening to find reasons additional to mere teacher shortage for this return to the profession. The third factor associated with young children (Factor IX) has only one loading of particular note, that of 39, Classroom space. It is interesting to find this associated positively with test attainment. Factors associated with 'brightness' 68. Factors VII and III show particular associations with high scores on the tests. A study of the major loadings show that newer schools, larger schools and larger classes tend to have higher proportions of bright children, while streaming and the existence of homework are also positive factors. Neighbourhood variables involved include housing standard and play areas, with oddly perverse signs for crime (house address) and school breaking and entering. Of the 'home' variables the only sizeable loadings are free meals (-.45) and mother working (-.45). SUMMARY 69. This report has been concerned with the statistics related to 44 schools, their environment, and the environment of their children. The unit of analysis has been the school, and not the individual child, and it is necessary to keep this in mind when we survey the results of our labours. 70. We have analysed the intercorrelations of 87 separate variables, 36 of them scores on tests, at four stages, and 51 of them measures of the children's environment, school, neighbourhood and home. Our analysis has first been done by means of average correlations, and then, more comprehensively, by factor analysis. 71. The most important of our findings, perhaps, is the demonstration that the major forces associated with educational attainment are to be found within the home circumstances of the children. These 'home' variables have, pro rata, nearly twice the weight of 'neighbourhood' and 'school' variables put together. That this may - at least partially - be caused by our choice of variables, and that a different selection might upset this order of influence, is a valid argument. But a scrutiny of the categories of measurement would suggest that our results might underestimate the power of the home. We endeavoured to cover all aspects of the school environment in our survey, taking account of not only those matters which are capable of objective measurement such as school size, pupil/teacher ratio, and the age distribution of teachers, but also of those aspects of school life which are matters of judgement rather than matters of counting: variables such as progressiveness, social atmosphere, quality of building, of equipment and of staff. Estimates of such matters are subject to error and are far from the level of reliability that can be attached to measures such as the age of the school building, or the proportion of men teachers on the staff. The factor analysis demonstrates, however, that the element of unreliability was not so great as to swamp the element of valid judgement. Our school variables, therefore, are many and varied, covering children, teachers, equipment and structure and it is difficult to believe that we have omitted any set of variables that would show a significantly stronger relationship with attainment. Our measures of the home and of the neighbourhood, on the other hand, are relatively few in number and limited in scope. It is not impossible that a more comprehensive coverage would have the effect of increasing the average effect of 'home' as against 'school'. 72. The aspects of home background that have the strongest associations with attainment on the tests seem to be those measures of maternal care which we have previously found so powerful in their association with the attainment of children in secondary schools. Such variables, together with those associated with material need, make up most of the 'home' effect. But dirt and crime seem more important than poverty in this context. 73. The second fundamental result of our researches, and one of the utmost significance, is the demonstration that the effects of environment seem to press more heavily on the brightest of the children. The adverse elements in home, neighbourhood and school conditions are shown to reduce the proportion of able children in the schools. This finding, which has some support from other inquiries, is of the greatest significance for educationists, for politicians, and for society as a whole. When we think of the problem of material and cultural deprivation, we see it as a problem affecting the 'submerged tenth' the slum dwellers, the poverty-stricken. We tend to assume that it affects only the tail end of the ability range as well as the tail end of the income range. Both of these views are wrong, and the second is even more radically wrong than the first. Educational deprivation is not mainly the effect of poverty: parental attitude and maternal care are more important than the level of material needs. The child from a home with an income of £20 per week may be more at risk than one from a much poorer home. The assumption that educational deprivation breeds educational backwardness is true but misleading. What is more in accord with the facts is the dictum that educational deprivation prevents the flowering of latent abilities, and that the higher the potential, the more potent and the more catastrophic is its effects. 74. It seems possible that environmental effects are strongest at the lower end of the school, and tend to fall off in their impact on educational attainment as the children get older. We cannot be categorical about this, since the design of this research is incapable of demonstrating it unequivocally. But our results provide some support for such an hypothesis. This would suggest that the provision of nursery schools ought to be seen as a way of counteracting adverse home conditions at the earliest possible moment - and counteracting adverse intellectual and attitudinal factors rather than the merely physical. This is the point at which the battle should first be joined: and the battle for the bright as well as for the dull. 75. When we survey the results of our investigations of the school environment we find the surprising result that organisational and physical aspects have stronger associations with the criterion tests than have the teacher variables and those associated with school atmosphere. The major factor, covering nearly three quarters of the total test variance, contains only five 'school' variables with loadings over 0.3 - and of these, attendance can clearly be regarded as one at least partially (and perhaps mainly) mediated by the home and the parents. A second one - the proportion of children qualified for special school - can hardly be regarded as a measure over which the primary school has much control: it is mainly a measure of one aspect of the quality of the school's intake. These leave us with only three significant variables; streaming, homework and size of class. 76. Little emphasis ought to be placed on the positive relationship between streamed schools and test performance, for two reasons. First, we are dealing with a relatively small sample of schools - 44 only, of which 28 are one-form entry. An unstreamed school in this survey is not unstreamed because of a policy decision on the part of the head teacher, but because it cannot be anything else! It is true that seven of these 28 were categorised as 'streamed' by our observer, on the grounds that the age groups were mixed and the children were grouped on ability, but there was no evidence that the remaining 21 were possessed of a 'non-streaming ethos'. Secondly, it is necessary to emphasise again the limitations of our criterion measures. The tests given at 10+ were those used by the authority for secondary school selection: those at the earlier ages were of a similar type. These objective measures of the tool subjects are those which might be expected to benefit most from the formal methods of teaching which tend to accompany streaming. But this should not blind us to the possibility that other kinds of assessment might produce the reverse effect. 'Progressive' methods and 'activity' methods are designed to do more than produce skill and competence in the three Rs: the advocates of non-streaming tend to be those who argue for less formal teaching and a wider curriculum. The fact that we perforce had to use objective tests of English, Arithmetic and Verbal Intelligence as our criteria means that we have left untapped wide areas of primary school achievement, and have loaded the scales against more progressive methods of teaching and organisation. It is unfortunate that the controversy about streaming has become a major politico-educational issue. It is an extremely complex matter to investigate, and it would be folly to suggest that the results from 44 schools in a single authority can throw any but the most feeble gleam of light on this problem. It is suggested that the only fact of any significance in our results may well be the demonstration that, even using objectives tests of limited coverage, streaming does not produce adverse effects at the lower end of the ability range. As we said earlier, streamed schools tend to have fewer backward children than unstreamed schools at all ages and in all tests. 77. The positive association of homework with attainment must also be related to the nature of our criterion measures. We have shown, by using the technique of partial correlation, that this is unlikely to be the result of parental pressures from the middle-class homes: it would seem more likely to be a school reaction to the pressures of 11+ selection. 78. The factor loading of .36 with size of class brings this research into line with many others which have shown a similar relationship. Our data show that this is partly - but only partly - a function of school size and quality of neighbourhood (in a city where many of the smallest primary schools are in the worst areas, awaiting demolition). 79. The less important factors thrown up in the factor analysis provide some interesting results, particularly for the younger children of 7+ and 8+. Teacher turnover rate is shown to be particularly important here, as well as the presence of older teachers, of women teachers, and of married women teachers with children. 80. All in all, it is clear that our schools analysis has raised a number of questions rather than produced solutions to problems. This is to be expected in a field which has been so neglected in research. The limitations of correlational analysis are many, and it must be remembered that the figures we have presented here are measures of association: they are not (necessarily) indications of causality. And the special nature of a research using schools as units must be remembered: correlations between schools cannot be used as substitutes for - or assumed to be equal to - correlations between individual pupils. The second stage of our inquiry becomes important, therefore, not only for the intrinsic value of its results, but for the relationship shown (if any) between the school analysis and the pupil analysis.
II THE PUPIL ANALYSIS 81. The sub-sample of individual pupils for this part of the inquiry was drawn from 22 schools. From each of the sampled schools 10 children were chosen, using a table of random numbers. By the same method, two children were chosen as reserves. The parents of 191 children in the original sample were interviewed. Three children had left the school, and in 12 further cases no contact could be established, either because the family had moved (sometimes because of slum clearance), or were on holiday, or because the parents were reported by neighbours to be always out, or (in one case only) because the address given was non-existent. Thirteen parents refused to be interviewed. Three of these 'had no time to spare', and a fourth was ill. Of the rest, only four refused outright, the other six made appointments but later broke them. In addition, one interview was abandoned because of the Pakistani parents' lack of English. Nineteen of the non-respondents were replaced by reserves, so that there were in all 210 interviews. 82. Further wastage occurred because of lack of complete data on the criterion tests. If a child in the sample had missed only one or, at the most, two of the twelve tests used as criterion measures, or if the school specifically stated that he had not taken one or more of the 10+ (selection) tests for reasons of backwardness (and provided that he had taken all the earlier tests), the missing scores were inserted by calculating the child's average score for the appropriate type of test (English, Arithmetic, Intelligence). For 24 children the omissions were more frequent and they were excluded from the analysis. This report is thus concerned with the results from 186 children and their parents. Variables 83. The criterion variables were, of course, the results of the 12 tests of intelligence, English and arithmetic given at 7+, 8+, 9+ and 10+. There were also a number of the home and school variables used in the schools analysis that were applicable to individual children: Head teachers' rating, Material needs inadequate, Cleanliness of home, Disrupted homes, Family crime, Neighbourhood crime. 84. In selecting those responses to the COI questionnaire which should be quantified and used in the analysis, we had the benefit of a factor analysis of the questionnaire items. Using this as a guide, a first selection was made of 53 items. Some of these were closely related, e.g. Type of house and Council house; Talks with teacher and Whether discussed child's progress with teacher; Complained at school and Dissatisfaction with teachers. These were included in the correlation matrix, and exclusion decided upon in the light of the size of the correlation coefficients. This scrutiny led to the dropping of 13 variables, most because of overlap (as above) but some because of lack of correlation with the criterion variables. For example, Sex of child had an average correlation of .041 with the educational tests, Availability of children's clubs .034, Mother working -.051, No garden -.091, Husband lenient with child -.062, Parents' belief in corporal punishment .038, Age child started school .048. 85. Since we had obtained interesting results on backwardness and brightness in the school analysis, we were anxious to include similar measures for the individual study. This is not so simple, however. For a school, the percentage of children scoring over 115 or under 85 is a statistic which produces a graduated scale of measurement over a sample of schools: for an individual child his score on any particular test either is or is not outside one or other of these arbitrary limits. There are a number of ways in which the association between brightness or backwardness and social factors may be assessed in a sample of children: we finally decided to use a simple dichotomy (bright, not bright; backward, not backward) as a scoring device for each child on each test. Since the correlations between these measures themselves, and between them and the actual test score will inevitably be high, it was judged safest to perform three separate factor analyses: the first using actual test scores as criterion variables, the second using the bright: not bright dichotomies as the criteria, the third using backward: not backward. For each analysis we thus had 52 variables, 12 criterion measures and 40 'family' measures. The results of the three factor analyses are treated separately below. The annex contains a list of the variables involved, together with a description of their bases and a note of distribution or range. A. TEST SCORE ANALYSIS Average correlations 86. Table 11 shows, in the upper half, all the variables having an average correlation with the 12 tests of over .2. The lower half of the table includes a selection of the remaining variables: it is not comprehensive, and it must not be assumed that because a variable is not included here its average correlation is lower than that of the bottom entry. Table 11 Average correlations 87. The highest correlation of all is given by 62 Preferred age of leaving, a six-point scale ranging from 'under 15 if possible' up to '18+'. This is perhaps the sharpest and quickest way of indicating parental attitude towards education. The superiority of the IQ correlation (.422) over that of English (.409) and Arithmetic (.392) tends to support the view put forward in the earlier part of this Report of the importance of the genetic element. The second variable in the list (70 Homework given) is the response to the question 'Have the teachers at ... school given the child any work to do at home?' This should not be regarded as necessarily a school variable. A negative response may indicate parental ignorance and lack of interest - or, indeed, the child's refusal to conform. 88. What is very significant is the presence in the 'top seven' of the four variables dealing with reading, with average correlations with all tests ranging from .272 to .341, and an overall average of .312. Compare this with the bottom line of Table 2: Father's occupation .098. Although there is some slight association with social class, what matters is the degree of literacy within the home, and the attitude of parents towards books and towards school. These correlations emphasise the existence of many 'good' homes in the working class, and many 'bad' homes in the middle class. The differences between the three kinds of test are also interesting for the reading variables. The child's reading understandably shows the highest correlation with the English tests: if we average the correlations of the other three reading variables we get IQ .325, English .307 and Arithmetic .283. It is not unreasonable to assume that the standard of literacy in the home, and the interest the parents show in reading, are - like their readiness to keep the child at school beyond the statutory leaving age - correlated with the parents' level of intelligence. Indeed, one would expect a strong association here. It is not surprising therefore, to find a high correlation of these variables with the level of intelligence of the child. As in the schools analysis, variables which are parent-determined, or which are likely to be affected by parental action or attitude, tend to show higher correlations with the child's IQ than with the measure of attainment. Age trends 89. There is a general tendency, as in the schools analysis, for coefficients to rise with age; but, as we pointed out, this may not be significant. There are indications, however, of variables which seem to have stronger associations with the attainment of the younger children. The easiest way to pick these out is to run the eye down the 7+ column, noticing any figures which breaks the order of size. Four variables stand out in this respect: Disrupted home, Neighbourhood crime, Mother plays with children in the evening and - rather oddly - Father's occupation. Disrupted home and Mother's play are logically enough the kinds of factors very likely to have major impact on the youngest children but it is puzzling to see why Neighbourhood crime should show this tendency when Family crime does not. The result for Father's occupation seems to have little relation to economic factors, since Material needs inadequate shows no similar trend (7+, -.126; 8+, -.167; 9+, -.173; 10+, -.179). Factor analysis 90. A principal components analysis was carried out followed by a Varimax rotation of 17 factors. When the variance contributed by the educational tests to these factors was calculated, it was found that 94.2 per cent of this variance lay in the first factor: none of the other factors contributed more than 0.7 per cent. We have, therefore, a simple and uncomplicated solution. Table 12 Major loadings on Factor I - test score analysis 91. The major loadings on the only (educationally) significant factor are given on Table 12. It will be seen that - as we would expect from a mono-factor solution - it bears a close resemblance to the average correlation data in Table 11. The heaviest loadings are for those variables indicating parental attitude towards education and the literacy of the home. Good attainment is achieved by pupils with parents who want them to stay on at school beyond 16, who prefer a grammar school education, who are members of a public library, who possess books themselves, and who tend to read both fiction and non-fiction. Poor attainment is shown by children whose parents are worried about teaching methods or the child's school progress, who are critical of the child's teachers, who have visited the school to discuss the child's behavioural problems, and who believe that teachers should try to get children on faster in their work. The psychological aspects of the home (e.g. disruption, child difficult to control) tend to be more important than the physical environment (e.g. bedroom deficiency, no fixed bath, cleanliness, type of house). It is, perhaps, surprising to find a relatively low loading on 55 Age parents completed education: this, perhaps, does not argue a great deal of success for the educational system. Those of us who chafe at what often seems the impossibility of achieving radical change quickly - and this is undoubtedly the case for the problem of environmental handicap - sometimes try to comfort ourselves with the belief that, with succeeding generations, education itself will provide the remedy. We get small comfort from this analysis. Education beyond the statutory leaving age seems, in this sample, to have comparatively little effect on attitude towards the education of their own children. It correlates only .210 with Preferred age of leaving and .123 with Grammar school preferred. Even with the literacy variables the correlations are low: Parent's reading .215, Membership of library .208 and Books in home .129. On the other hand, our sample of parents is probably less than fully representative here. The distribution of ratings on variables 55 shows that, of 186 families, in 132 both mother and father left school at 14 (or earlier) and in 28 further cases one parent left at 14 and one at 15. This perhaps may serve to underline one of the problems of interpretation of inquiries such as this: the possibility that certain results reflect the circumstances of a particular region, and cannot be generalised to the country as a whole. The hazards of applying results obtained in large cities to country towns or villages are obvious: there may be equal hazards in generalising from the North to the South, in view of what we know of the social and educational gradient existing in Britain. 92. Age trends and test trends are shown at the bottom of Table 12, from which it can be seen that this factor has its strongest loadings on the educational measures at the age of 9+, and that its connection with general ability is clearly stronger than with English or Arithmetic. B. BRIGHTNESS ANALYSIS 93. The computer programme identified 17 factors from the principal components analysis for Varimax rotation. Of the total educational variance contributed by these factors, 76 per cent came from Factor I. The next highest fraction was from Factor 12, but since this was only 5.1 per cent, its influence is of little account. We are therefore left with another single factor analysis. The major loadings on this first factor are given in Table 13. Table 13 Major loadings on Factor I - brightness analysis 94. There are considerable similarities in this pattern and the one for the score analysis. The four measures of home literacy are again found among the top seven variables, for example. But the primacy of 62 Preferred age of leaving yields place to 61 Grammar school preference. This is perhaps a sharper measure of parental attitude for the above-average child than the wider-ranging 62. What is of most interest perhaps is the collection of variables which appear in this table but which were not included in the 21 measures having the strongest loadings in the score analysis: 68 Husband's interest, 75 Changes of school, 45 Material needs inadequate, and 63 Talks with head or class teacher. In contrast to these, there are five variables included in the previous list which find no place here: 40 Discussed problem behaviour at school, 55 Age parents completed education, 47 Type of house, 48 Bedroom deficiency, and 49 No fixed bath. It seems, therefore, that although poverty might exercise an inhibiting effect on the emergence of brightness, the purely physical aspects of the home environment are relatively unimportant. The interest of the child's father in school work, however, is markedly associated with brightness, as well as evidence of parental interest as shown by discussions with the school staff - provided, however, that these are sharply distinguished from those visits made to discuss unsatisfactory progress. The low loading on 55 Age parents completed education remarked on in the discussion of the score analysis is paralleled by an even lower loading on the brightness factor (.061 as compared with .129). 95. The bottom of Table 13 shows the age trends and test trends for the factor. As in the score analysis, 9+ shows the highest proportion of variance, but is not markedly higher than 10+. The major difference lies in the much lower contribution from the 7+ tests in the brightness analysis. This is made even clearer if the figures are expressed as ratios. Taking the 9+ variance as 100 in each case, the 7+, 8+ and 10+ contributions are, for the test score analysis, respectively 84, 96 and 97. For brightness, they are 45, 83 and 95. This indicates, perhaps, more radical changes in level of attainment over the years for those children at the upper end of the ability range, thus emphasising again the greater environmental risk for the bright child. 96. For test score, IQ contributed more to Factor I than either English or Arithmetic. Expressed as ratios, the variances are 100: 94: 94. The picture for brightness is different, with English showing up strongly, and Arithmetic much lower: 100: 105: 76. Other investigators have stressed the importance of verbal ability and its susceptibility to environmental influences. These results suggest that they may be rather more important at the upper end of the ability range than with the average child. C. BACKWARDNESS ANALYSIS 97. The backwardness analysis was a replication of that done for brightness. The computer rotated 18 factors to a Varimax solution. The first factor covered 73 per cent of the test variance, but on this occasion a second factor proved to have a sizeable weighting: Factor 16, contributing 11.7 per cent of the variance on the educational tests. It has only one sixth of the power of Factor I, but its influence is obviously not negligible. The first factor 98. Loadings of .100 and over on this factor are shown in Table 14. A fuller comparison of this table with the corresponding tables for the other analysis is made later, but certain obvious points may be noted. First, the four 'home literacy' variables again appear in the top half of the table, but with slightly less importance. The average rank order of these four measures are four for brightness, five for score, and six for backwardness. 70 Homework given goes to the top here, in contrast to its rank of 3rd for score and 10th for brightness. Of the direct parental attitude measures, 62 Preferred age of leaving parallels the score analysis as being the most important, 61 Grammar school preference coming 17th here as compared with first for the brightness analysis. Table 14 Major loadings on Factor I - backwardness analysis Only two variables appear in this table and not in the other two: 67 Child talks about school work and 66 Mother not too busy to help with school work. Thus it seems that parental attitude towards the child, and support for him in his educational difficulties, are of some importance in the prevention of backwardness. But note that this falls well below home literacy and attitude towards education in the order of importance. Factor 16 Table 15 Major loadings on Factor 16 - backwardness analysis 99. A glance at the bottom section of Table 15 makes it clear that this factor reflects associations only at 7+, and only with Arithmetic. Notice, however, that the average 7+ variance of .251 is commensurate with the 7+ variance of .285 on Factor I: for these young children, both factors are of equal importance. The variables represented in Table 15 tend to support the last comment made about the first factor, and to underline the importance of parental attitude to the child. Variable 53 has a loading more than twice as great as any other measure, and is clearly the 'marker variable' for this factor. It plays no part in the constellation of variables making up the first factors in our three analyses: the loadings on 53 Mother spends time with the child in the evenings are .120 (score), .046 (brightness) and .073 (backwardness). We now see, however, that this aspect of maternal attitude and behaviour has some relevance to backwardness, but only with reference to the youngest age group. 100. The second variable, 43 Father's occupation, also appears for the first time as a significant variable. Its loadings on the three first factors are -.077 (score), -.104 (brightness) and .078 (backwardness). The perverse sign in the last entry is probably of no significance, since the loading is not significantly different from zero. The uniformly low loadings in this variable, so often used by sociologists, re-emphasises the importance of intra-class differences in background and parental attitude as compared with inter-class differences. 101. Of the other measures included in Table 15, 65 and 67 tend to support the maternal attitude aspect of this variable, while the deleterious effect on the young pupil of changes of school is again drawn to our notice. That this factor is concerned almost entirely with backwardness in arithmetic will not escape notice: the effect of changes of school is very likely to show itself in performance in such a logical subject as this. D. THE THREE ANALYSES COMPARED 102. We have here three separate analyses, each yielding a single factor contributing three quarters or more to the total test variance. What are the similarities and differences in the factor patterns? Does this investigation of 186 children show differences between backwardness and brightness similar to those yielded by the investigation of 44 schools? Similarities 103. Let us begin by considering the common elements in all three analyses. In order to avoid making what might be misleading comparisons between the actual size of loadings, we shall pay most attention to rank order of importance of variables. If we take - arbitrarily - the 'top ten' loadings in each of the first factors, what are the variables which appear in all three analyses? The variables are these, the major common elements: 56 Membership of library104. Here we have a clear indication of the importance of home literacy (56-59) and parental attitude towards education (62, 72). The homework variable, as suggested earlier in this report, is perhaps to be regarded as a blend of school policy, parental interest and child's attitude. 105. Two other variables appear in the 'top ten' in two of the analyses only. 61 Grammar school preference is important in the brightness and the score analyses, but much less so for backwardness. This is not an illogical result: for the average and above-average pupil this is a good measure of parental attitude, but is obviously less so for the backward child, when ambition is tempered with realism. Variable 50, Disrupted home, appears in the top half of the tables for backwardness and score, but it is less potent for brightness. Differences 106. As a simple way of measuring differences, and to ensure that none was overlooked, the first factor loadings of all variables were ranked in order of numerical size for all three analyses. Rank differences were then found for each variable. Those with differences of 10 ranks or more (5) were then listed. These are given in Table 16. Table 16 Variables with 10 or more ranks difference between analyses 107. The top half of the table shows those measures whose effect is stronger for brightness, the bottom half those which affect backwardness most. This brings out very clearly the difference already noted, that the factors affecting backwardness tend to be those denoting maternal attitude to the child, while for brightness the parents' attitude to education seems the more important. The backward child is a pupil beset with problems. Support, affection and understanding at home help him to surmount these. The potentially bright child, on the other hand, needs not so much help in overcoming difficulties as encouragement to take full advantage of the opportunities offered. The parents' view of schooling, the value they put on education, are the factors that are likely to affect the progress of the able child. 108. Nevertheless, there still remain some puzzles in Table 16. Who would have hypothesised that 75 Changes of school would have the strongest effect on brightness, and the least on backwardness? It is true that the additional Factor 16 brings this in as a variable associated with backwardness at 7+, but even here its loading is low. This is a differential we find unable to understand. Similarly, the inadequacy of material needs, as a variable having a stronger association with brightness than with backwardness, is, at our present level of insight, inexplicable. Finally, 55 Age parents completed education adds a third element to our puzzlement. We have already commented on the pessimistic results for this particular variable, and it will be noted that its highest rank is 19th. But why should even this tenuous measure of association be confined to backwardness? A rank of 29 for brightness (corresponding to a loading of only .061) is, in our view, a wholly unexpected result. SUMMARY 109. The schools analysis underlined the importance of the home background as a determinant of the educational progress and attainment of primary school children. This second phase of our inquiry, in which we look more closely at 186 individual pupils and their parents, begins to unravel some of the strands in this backcloth. It supports the schools analysis in finding that economic level and social class are much less important than aspects of parental attitude, attitude to education, and attitude to books and reading. A high wage packet and a middle-class home does not guarantee a favourable background for educational progress, and literate homes with good parental attitude to school may be found in the slums as well as in the suburbs. Family literacy, as evinced by measures of reading and library membership, proves to be a highly significant measure in all three analyses of individual pupils. Parental attitude to education is an equally strong indicator of educational attainment. As a single measure of this, Preferred age of leaving is pre-eminent for the average and the backward child; Grammar school preference supersedes it for the bright child. 110. The factor of parental care which has manifested itself in all our school-based surveys begins to be more clearly delineated in the pupil analysis. Maternal attitude to the child and to his educational problems is shown as an important factor in the backwardness analysis. The problems associated with educational failure are likely to be eased with maternal support in a stable home, and so permit better school progress. On the other hand, the variable in this context which has most association with brightness is the father's interest in school work, which seems to act as a spur to (or a release of) ambition. A father's support and encouragement might well counteract the more adverse attitudinal forces of the neighbourhood and the peer group. The role of the mother defending and protecting the weak child and that of the father stimulating and encouraging the strong emerges surprisingly clearly from our analysis: but one suspects that the situation may not be quite so simple as it looks. 111. The pupil analysis is surprising in that it fails to support the emphasis given in the wider survey to factors of cleanliness and crime. The commanding position of these in the schools analysis is not paralleled in the analysis of individual pupils. This may be an accident of sampling - 186 pupils as against 2,000, and drawn from only 22 schools - or it may be that we need a more thorough investigation of the 'cleanliness' variable for individual homes. Only 23 out of 186 homes were rated by the Medical and Welfare departments as falling below an 'adequate' level: such a dichotomy is a very crude measure. Similarly only 17 of our families had any criminal record. CONCLUSION 112. Our survey was necessarily a limited one, limited in time and in coverage. We have already commented on the narrowness of our criterion measures, and on the smallness of the sample of 44 schools drawn from only one, urban, region. Nevertheless, some significant results have emerged, as well as a score or more of questions which must lead on to further research. For our part, we regard two of our findings as being of the first importance: that environmental forces bear most heavily on the brightest of our children; and that factors in the home are overwhelmingly more powerful than those of the neighbourhood and the school - and of these, factors of parental attitude to education, to the school and to books are of far greater significance than social class and occupational level. There is strong reason to believe that adverse forces of this nature operate from the earliest ages, and we see the provision of nursery schools as an important way of attempting to counteract them as early as possible. It seems patently obvious to us, too, that much more attention to environmental effects on education needs to be given in the training of our teachers, and we see the necessity to encourage, by every means, the institution of special courses within colleges of education. These should provide, for those students who so wish, a training in educational sociology and social work which will enable them to view these problems with a wider perspective, recognising that experiences outside the classroom are important not only for children but for teachers. In addition to such options in initial training courses, what is also needed are training courses for experienced teachers, so that they may be equipped to cooperate with the officers of the other social agencies within local authorities. Their task would be, not only to recognise cases of deprivation and to refer them to the appropriate experts, but also to bring home and school, parents and teachers, into closer relationship, to their mutual understanding and benefit. Such teachers are as necessary as remedial teachers, careers masters, or guidance counsellors, and would - like them - spend only part of their time within the classroom. The existence of a small corps of such teachers - social workers within an authority (and particularly the large urban authorities) would be of the greatest benefit in increasing our knowledge of the complex interactions between environment and educational attainment, and in developing new strategies in our efforts to provide the conditions for the full flowering of every child's potentiality.
ANNEXES I List of variables (school analysis) II Varimax Factor Analysis (school analysis) III List of variables (pupil analysis) IV Loadings on significant factors (pupil analysis) [Note Annex IV, as shown above, is included in the list of Annexes printed on page 384 but only Annexes I - III follow on pages 385 - 400: Annex IV does not appear in the printed edition.]
Footnotes (1) I must acknowledge my indebtedness to Professor Frank Warburton, who was concerned with the project from the start; to Mr Tom Derrick, the Research Officer, who was responsible for collecting the data; to Miss Anne Sutherland and Mr Brian Start, who monitored the computer work; and to the officers of the Manchester LEA, in particular Mr Elliot, the Chief Education Officer, and Dr Laybourne, his deputy, for their willing and wholehearted cooperation. But most of all my gratitude must go to the head teachers and teachers in the sampled schools for their interest, their patience and their invaluable help. Without their full support this inquiry would have been impossible. (2) Reported in Wiseman, 1964, Education and Environment, Manchester University Press. (3) Education and Environment, p. 78. (4) The Chosen Few (1961) (5) Out of 38 environmental variables: 37 Head teacher's rating of pupil, and 38 Result of 11+ selection tests were omitted - as they are in the table of loadings given previously - as belonging to the criterion category rather than the environmental. 
|