Introduction

This file looks back over some classic studies of underachievement and reviews them critically.  A critical look is crucial - these studies have become famous, a part of teacher folklore as well as teacher training.  In particular, these studies are often both better in some ways and worse in some ways than most people think.  We are going to illustrate this odd finding by looking critically (and briefly) at three studies -  Fraser, Douglas and Plowden - to give a flavour of critique

Overall, the pattern seems clear enough that working class students underachieve and do this mostly because of home circumstances and "environmental" factors which affect them, even where access has been widened.

Hit my address book here

Study 1

An early and influential classic was the work done by E Fraser (1959).  Let's look at how she went about studying the impact of home environments. Fraser was keen to do this scientifically - not to rely upon opinions or values but to measure the variables and correlate them.  This was a welcome change from mere opinions, and the study reveals the problems when we do take the issue seriously

Fraser began by subdividing various aspects of the home environment - cultural, material, emotional etc. - and then further subdividing these to get objective specific measures (e.g. she measured cultural factors by looking at parental education).   Fraser did her work, got her measures and then correlated them and came up with really precise correlations - e.g. between home background and IQ scores (r = 0.687), or between home background and school attainment (r = 0.752). (NB the closer the coefficient "r" gets to 1.00, the stronger the relation).

Let's look at her methodology more carefully.  School attainment is crucial, and she measured this by looking at the marks gained in school exams.(no SATs in those days).  There are immediate difficulties, though - do  the individual marks gained mean the same things in different classes, for example? Remember that schools were commonly 'streamed' by ability in those days. Scientific research required the construction of a standardised scale of merit from the actual marks - but how do you rate an average mark of 90% for a B-stream kid as against an average mark of 30% for an A-stream kid?  Which of those has attained more? 

Fraser simply organised a scale for the whole school by arguing that, roughly speaking, A stream students were all brighter than B-stream ones. 

This assumes that:  streaming is a rational system, that ability is simply measured in school exams, that ability is a variable which can be arranged on some universal scale - and so on. To be fair, these were common working assumptions in her day - and still may be? There was already some research showing that social judgements also played a part in practice, though.

Fraser further refined the scale by asking about the promotion of children between the streams.  So if childen in the B stream got 90% in their exams and got promoted into the A stream, then a 90% score in the B stream could be tacked on to the bottom of the A stream scores.  Similarly, if an A stream pupil was demoted if s/he got less than 30%, a score of 30% in the A stream could be seen as the grade just above 90% in the B stream.  In this ingenious (if dubious) way, all the examination scores throughout the school could be systematised.

Ten schools were studied, and these were variable too: some schools have higher standards than others. Is mediocre attainment in a high standard school better than good attainment in a low standard one?  This problem was solved fairly simply by Fraser.  She rank-ordered schools to get the whole range with high standard schools at the top.  The lowest kid in this school was ranked as just above the highest in next school down and so on.  How did she decide on the ranks of schools?  She compared the IQ scores of pupil intakes, on the assumption that the brightest students go to the "best" schools. Do you see any problems with that assumption? -- how can you NOT see any?!

Having got an overall, standardised scale with each pupil given a rank on it, we can measure effects of home environment on attainment.

Fraser used questionnaires and interview schedules to research home environment.  She also gathered impressions while on home visits which were coded into objective terms afterwards.  She spent an average of 30 minutes on each visit (!)  The coding is interesting - e.g. parents' education was assessed by awarding points for different sorts of education - 3 points for each year spent in school after the age of 14.  (5 points for "good" secondary schools).  5 points were awarded for each year of university training, and 5 for an Honours Degree. The assumptions here are interesting - e.g. people with 3 years in 6th form got 9 points and thus their education was 3 times "better" than these with 1 year.  Is a (4 years) Honours Degree 33.33% "better" than a (3 years) Ordinary Degree? 

Consider another measure of "home background" - reading habits. "Keen" readers got 5 points, frequent readers 4, occasional readers 3, rare readers 2, those who "never open a book" (sic -- based on a 30 minute visit, remember) received 1 point. What do you think? A qualitative dimension to reading was measured too - "top quality novels and some non-fiction" got 5 points, down to "low grade fiction" with 2, and "newspaper and magazines only" 1 point. Clear values affect this too, obviously.

Parental interest was crucial - Fraser rated this by asking primary school teachers to assess it for parents, using a 5-point scale and the returns were then normalised - i.e. fitted into a normal curve. This was done for measurement reasons, enabling Fraser to use some powerful statistics (but see below). It meant that in practice, parents were divided into batches - top 5%, next 20%, middle 50%, lower 20%, lowest 5% - and only 5% of the parents could get 5 points, and only 5% zero points etc.    NB There is an important technical point here - Fraser found a strong correlation between attainment and parental interest, but this is not surprising because both are nicely standardised.   A correlation measures the extent to which variables vary together - attainment varies "normally" and so does parental interest.
 

Study 2

J.W.B.Douglas studied a panel of children born in 1948 and followed them through their careers. The Home and the School (1964) studies the panel at the stage of going to selective secondary school (an important stage in those days).  Allegedly, selection procedures used by LEAs focused solely on 'ability to benefit' - but there was a strong suspicion that social class was involved.  Douglas investigates this possibility (and examines other factors such as the availability of places at selective secondary schools).

The nub of the analysis appears in the famous table, (p.155 of the Panther edition).  Douglas tries to show the effects of both "ability" (as measured on a standard test, specially constructed for his study by the NFER - who design SATs today), and "social class" (he used a definition of his own,  based on occupation and levels of  parental education), and the relation between them.

I reproduce the table here as a (pretty grotty) graphic. If you are printing this out, you may wish to print it separately. Click here for electronic version

NB test scores are NOT standard IQ scores but arise from a special NFER test

The table reveals a classic methodological strategy to compare the effects of two variables at the same time on a set of results. Correlations enable you to do the same thing, but I like diagrams. As with all such devices, you have to know how to read them though. Think of this table as having rows and columns. Children are organised into three bands according to their ability (the rows in the table).   They are also divided into four classes (the columns) for each band, giving 12 cells in all. 

It can help, sometimes, to concentrate on one row or column at a time, and blank off the others (literally if you like, by placing a card over the ones you are ignoring for now). If you look at the third row first, you will see that, for those in the high ability band (scoring 61 or over in the test), about the same percentage did indeed get to selective secondary school.  Class seems to have little effect with these children.   However, with children of lesser ability, parental class does have an effect:  the percentage going to selective schools in the middle and lower bands (second and first rows) drops, in general, as you move across the table from column to column.  Here, ability seems to have little effect: remarkably, for example, 40.1% of upper middle class students in the lowest band were still getting into selective secondary schools!. The table overall shows an inter-relationship between ability and class, and two implications arise:
 

1. Substantively, the old tripartite system did reward  those with high ability regardless of their parental  social class, and was thus much more "open to  talent" than is sometimes believed nowadays  (confirmed by the results of the Oxford studies too --  click link).
2. Douglas's method, of comparing two variables at once, laid the foundations for much more  sophisticated manipulations of data in later  studies.   We can now examine complex inter- relations with class, race, gender and ability, for example.

You might like to look back over both studies and ask yourself similar questions:

1. What is problematic about using teachers' definitions of an element like "parental interest"? Are teachers in a position to be able to assess parental interest? How might teachers misinterpret the contacts they have with parents?

2. What alternative measures are available to researchers of items like "parental interest" or "home environment"?

3. What are the pros and cons of trying to measure issues like "home environment" scientifically? Is the only alternative to rely upon teachers' or politicians' impressions - and would these be any less open to critique?

The Plowden Report (CACE 1967) offers another exercise in a well-worn field, but in an even more sophisticated way. A huge sample of 107 schools and 3000 students was drawn (in the National Survey), and other surveys were commissioned - eg a survey of HMIs. The analysis too was specially sophisticated for its day: the Report was keen to sort out the relative influence of all the known factors correlated with achievement. This is still practically important today. We knew at the time, for example, that family size is connected with attainment, and that family attitudes to education are as well - but which is the more important factor, as far as policy is concerned? Should the Government spend its money promoting birth control or campaigns to change attitudes? At the level of teacher activity, Plowden tried to settle the clash of opinions among teachers at the time - do working class children underachieve because they are poor or because they are "culturally deprived"?

The findings can be summarised quite briefly:
 

1. All the many variables at work can be reduced to three main clusters of variables - "parental attitudes", "home circumstances", and " state of the school" However, not all the variance was explained by these three factors - a  table on p.33 of Vol.2 admits that 61% of the variation in achievement between infants in the same school was unexplained (much less unexplained variance exists for other categories). Like all research of this kind, Plowden only investigated known factors - see below. Clearly, a level of unexplained variance like this indicates that other factors were at work which had not been investigated.

2. On the whole, "parental attitudes" made the largest contribution to variations in the educational performance of the children,  whether that was measured within schools or between schools. The differences in attitudes were very wide, much wider than differences in the state of schools, for example - hence one practical implication that change will be brought about more quickly by focusing on parents' attitudes than on changing schools.

3. The three factors were separate from each other to some extent - especially "parental attitudes" and "home circumstances" - this meant, for example that home circumstances (the material factors like type of housing, occupation, income etc) did not explain attitudes. About 75% of the variation in attitudes remained unaffected by home circumstances, to be precise. This is an optimistic finding if you are interested in changing those attitudes, of course -- people can change their attitudes without major changes in their home circumstances (just as well, since home circumstances were seen as very difficult to change without major expenditure).

Williamson and Byrne (in Swift et al 1973) argue there is an underlying model at work in this research (as in the grotty graphic which follows -- click for electronic version)
 
 

Williamson and Byrne in Swift et al p.54

This is what Williamson and Byrne call a 'total system of educational deprivation' -- families, schools, culture and class all interact. This is the underlying view that has 'guided educational researchers [in the classic phase -- and still does?]'. The point of research is to see which particular variables are the most significant -- as we have seen, parental interest was an early favourite (an aspect of cultural values and social class in the lower 'Family and Community' box). NB 'deferred gratification' was popular before that - - this is the willingness to sacrifice immediate pleasures in order to benefit in the long term: working class folk were light on this cultural value, it was often said (and still is -- these assumptions lay behind the early doubts about really unqualified applicants to the Open University in 1970 -- see file?)

Again, though, we must look at the methodology that produced these findings - which have since become almost Holy Writ in some Education Departments, and which spawned a major policy effort in the establishment of Educational Priority Areas, and less directly, the community school lobby, and much of the "working ideologies" of "progressive" teachers (Plowden endorsed 'progressive teaching' on all sorts of unsubstantiated grounds in fact -- as a pioneering and skeptical study was to point out -- Bennett (1976)).

The methodological debates are "difficult", and so it might be best to read a good introduction first - eg Swift et al (1973) (Part 2). The National Survey is at the heart of the matter (Appendices 3-7 in Vol.2 of the Report). The Survey gathered data on pupil achievements, for example, by collecting (a) scores on a special test of reading and comprehension (b) the pupil's rank order in class, based on teachers' judgements. The factors of "home circumstances" and "parental attitudes" were based on measures of social class, income levels, the physical conditions of homes and their environments, and parental ambitions for their children, contacts made with schools, parental interest in and knowledge of school organisations etc. ("State of school",  for that matter, is also interesting, and measures here consisted largely  of levels of resources, type of building etc - very few of the social and cultural factors like 'organisational climate' mentioned by later studies - see below)

What happens then is much number munching of this data.  The first step is to perform a factor analysis  to establish whether answers given on all these different instruments are patterned in any significant way - eg to see if those parents who said that teachers were "easy to contact" also said that teachers were "pleased to see them" etc. General patterns  did emerge of "interested" and "non-interested" groups of parents, and these patterns can then be compared with  data on social class - eg are the interested parents mostly middle class... and so on. 

Variables are sorted like this using a series of multiple correlations and techniques which in effect involves taking each variable in turn and seeing if it is significantly related to the variance in attainment you are interested in.  This is really a mathematical version of the tabular method of Douglas - social class is related to achievement, then IQ is introduced, then it is possible to add in regional differences then parental interest (two tables) until you end up with variables that make no difference whatsoever to achievement - like the eye colour of children (a fictitious example). By this time, you would be into three- and four-dimensional tables , which would be pretty hard to draw or understand -- so we use statistics instead.  So the Report offers very thorough sophisticated analysis to test interaction effects - no wonder there were confident findings!!

However there are snags even here - one is the familiar one of actually measuring factors and definitions. The definition of class is especially important. It was seen primarily in terms of cultural attitudes and variables that somehow impact on an individual's motivation and ability (see a near-contemporary critique by Bernstein in Peters et al 1969). 

However, class is also a measure of power, power to force LEA's to provide good resources in schools, power to influence school policy etc., but this dimension was never investigated by Plowden. It is developed in Swift et al., where Williamson and Byrne (W & B) suggest that the way to proceed is to begin with an explicit theory which suggests a causal path between variables (rather than relying on statistical methods to establish patterns "empirically"). The path in W & B looks like this (another  graphic)

 
A version of the 'attainment-resources paradigm'

From Williamson and Byrne (Swift et al p.72).
 

The idea is to use the data to test the strength of each of the connections in the circuit, as it were. As the diagram  indicates, the strongest connection between social class and attainment lies through policy, resources and provision.

2. The definition of "school factors" offers problems here - choosing kids as units gives you nice individual data about homes etc., but it is very difficult to assess the individual impact of school organisation, so we have a basic problem in comparison straight away.

3. The Report only chose to measure what seemed to be important, as argued above - the Plowden researchers focused on homes because they were influenced by the likes of Fraser and Douglas.  Conspicuously left out were teachers attitudes as a variable (and later, aspects of ethos and other elements of "the school effect" - see Reynolds 1976, or Smith and Tomlinson 1990) because no one knew of the significance of these until after the Plowden Report.  The data are only as good as the questions you ask!  Even this rather sophisticated research can end as rather self-confirming, proving what it suspected all along, merely replicating existing theories rather than rigorously testing them (and thinking out alternatives). This self-confirming tendency is exacerbated by a haste to be directly relevant and immediately practical, and by an insufficient attention to how "theories" are smuggled into procedures like factor analysis or defining terms empirically.

4. There have been many critiques of Plowden since, focused on policy or practice. These are beyond the scope of this file, but see Halsey on the disappointing impact of initiatives in EPAs (1969), or the controversies about the implicit progressivism in Plowden in Bennett (1976), and Galton et al (1980).

To return to the main theme, even in highly sophisticated analysis like Plowden, definitions are very important.  Researchers only research things that seem important to them, they have all sorts of implicit theories about what is significant and ignore other factors which might explain findings even better.  Now professional researchers are aware of this and are suitably cautious - but the users of research are not.  Users (including teachers) often assume that research has proved links exist in a casual way whereas really only correlations have been established between some known variables, and there are always possibilities for new previously unknown variables emerging.

Exercise 2

Comparing your understanding now with your views before reading this file...
1. Are you more or less confident in the findings of pieces like Douglas or Plowden? Do the methodological problems mentioned here invalidate the research?
2. When teachers make judgements about the impact of social class on attainment are they as cautious and as open about the problems with the evidence and the conclusions as the researchers? Are there any problems with these teachers' accounts that might invalidate them?
3. How might the relationship between attainment and home background or social class be researched these days? What have we learned from these classic studies?
 
 

Plowden serves to introduce a discussion of background ideology and views about classes, stratification, education and social mobility.  There is an optimistic view in Plowden that class differences are irrational - traditional only, with no real reason for class differences these days.  And schools can rearrange society.  All you have to do is to positively discriminate - to make up for the past neglect of working class areas and put them on the same footing as "average" schools - to make it a "fair contest" so that talented kids of all classes will succeed and go on to fill best jobs - and this notion of fair contest underlies much policy as other files show (see, for example IN&ED1 and IN&ED2).
 

It seems a strange idea really - to help working class kids compete, to raise them to middle class standards as it were.   What of alternative socialist strategies to make schools change, to make them cater for the needs of working class kids and not the other way about?  Research and argument since Plowden features much more pessimism  about whether inequality is not inherent in our society whether it can ever be easily removed or made "fairer", whether schools themselves  contribute to inequality, whether the rules of the contest are really rigged so as to express the culture and characteristics of a ruling class which still exists etc  ( see file on Bourdieu, perhaps)- lots of these questions and points are still worth raising 

References
Bernstein, B & Davies, B (1969) "Some Sociological Comments on Plowden" in Peters, R (ed) Perspectives on Plowden, London, Routledge and Kegan Paul.
Bennett, N  (1976) Teaching Styles and Pupil Progress, London, Open Books
C.A.C.E. (1967)  Children and Their Primary Schools (the Plowden Report), H.M.S.O.
Douglas, J (1967) The Home and the School...,London,  Panther
Fraser, E (1959) Home Environment and the School,  London, London University Press.
Galton, M et al (1980) Inside the Primary Classroom, London, Routledge and Kegan Paul
Halsey, A (1969) Educational Priority, report of a research project Vol. 1, D.E.S.
Reynolds, D (1976) "The Delinquent School" in Hammersley M & Woods P (eds) The Process of Schooling, Bletchley, Open University Press.
Smith M & Tomlinson S (1990) The School Effect... , London, Policy Studies Institute, 1989
Swift D et al (1973) Education Economy and Politics Case Studies Parts 1 & 2 (E352 Block 5),  Bletchley, the Open University Press.