Peer effects and peer group processes: joining the conversation on quantitative and qualitative methods.

The interest in `peer effects', which includes the effects on educational attainment of school ability or socioeconomic status composition, raises methodological issues that require careful discussion. It is argued with reference to a recent collaborative text by Australian and New Zealand authors representing quantitative and qualitative approaches that school composition effects are not necessarily peer effects and that the term `effect' cannot be restricted by a positivist definition. It is further argued that the practice of using aggregated student responses as measures of institutional properties is fundamentally unsound. This case is supported by a detailed study of attitudes to school and their association with educational progress held by two groups of secondary students. It is concluded that a realist understanding of the properties of social entities might offer the basis for a collaborative approach more productive than the conventional paradigm now dominant in this field.

Introduction

The recent emergence of `peer effects' as a distinct field of study in educational research indicates the continued vitality of the quantitative tradition. It can be argued that the standard narrative in the sociology of education, now established for a generation, emphasises the centrality of peer group processes in ensuring the overall reproduction of dominated social groups (Bourdieu, 1971; Connell, Ashenden, Kessler, & Dowsett, 1982; Walker, 1988; Willis, 1978). Social psychology, moreover, has seen a considerable resurgence of interest in peer group processes and their effects, stimulated largely by Harris (1998), and the intellectual climate might seem more than favourable to the hermeneutic approach of ethnographic research. A peer effect by that particular concept, however, is actually one revealed by statistical analyses that find some proportion of the variance in attainment to be associated with the properties of schools and classrooms as distinct from the properties of individual students. The qualitative interest in peer effects, by that concept, signifies a focus on school composition effects, classroom effects, and so on, which have been established as such by correlational research. Although securely occupied by the forces of positivism, this field of peer effects has become the site of an engagement between quantitative and qualitative researchers, the outcome of which should not be regarded as a foregone conclusion. This paper attempts to enter the fray with a contribution informed by a realist philosophy of science (Bhaskar, 1993; Bunge, 1996, 1998). It will confront two significant texts: one a joint review by local researchers drawn from the quantitative and qualitative traditions (Wilkinson et al., 2000), the other a representative product of the prevailing orthodoxy in educational research (Marks, McMillan, & Hillman, 2001). Why some students make more progress at school than others requires an elaborate theory, within which the significance of a multitude of influences, not all of them easily constructed as factors, can be recognised and weighed. Such a theory will require clarity about cause and effect, which is actually at risk in this field, and about the real properties of social entities. It will be argued, in particular, that the definition of `effect' cannot be restricted to `statistical effect', and that institutional properties are not the sum of individual properties. The substantive finding that responses to `school climate' items seem to be unrelated to students' progress at school, both at the school level and in actual social groups, raises questions for theory and methodology that qualitative and quantitative researchers may both need to consider. The investigation of qualitative and quantitative approaches is largely confined to the context of their application in the substantive field of peer effects and does not attempt a comprehensive review of the general issues at stake (LeCompte, Millroy, & Preissle, 1992). This is an argumentative work: there is no point in science unless arguments are taken seriously: this engagement attempts to follow that injunction.

A conversation on peer effects

The relationship between quantitative and qualitative methodologists is characteristically strained (Gorard, 2001; Hammersley, 1989). In the explanation of social differences in education, however, quantitative methods are essential. Without some defensible means to assess the relative importance of variables, the sociology of education is reduced to teaching what is, in effect, a `list theory' in which the significance of home, school, peers, and so on, is determined by the political considerations of advocacy research. The fact that policy makers, constrained by their office to weigh matters on their objective merits, have become heavily reliant on quantitative analyses has not gone unnoticed by the producers of educational research. It is in this context, that Marks et al. (2001) note that `policy makers and others interested in educational outcomes often ask which are the most important influences' (pp. 61-62). They point out that this is a reasonable question given the `plethora of factors' shown to `influence educational outcomes' and conclude `from a policy perspective it is important to understand which influences are weak and which are strong so that policies can be directed at influences that can make a difference'. This apparently reasonable view, however, actually supports a normal paradigm that should not go unchallenged.

It is possible to see, in fact, that quantitative and qualitative researchers in this area might need each other. One set of investigators has an effect--actually a proportion of the variance in attainment associated with an aggregate variable--looking for an explanation, and the other has an observed process looking for an effect. The question for those who have discovered a statistical effect is what processes have caused it. And the question for those who have observed teaching and learning in schools with distinctive class compositions is what effect does this have. There is obviously some basis for accommodation here: if researchers in these different traditions can provide each other with the answer they are seeking, their cooperation might be successful. Wilkinson et al., as an interdisciplinary group, attempt to confront the issues that divide quantitative and qualitative researchers in this area. The principal difficulties lie in providing a definition of institutional composition effects, quantifying such effects, and identifying peer processes as their cause. These problems are quite severe, and it is instructive to follow the authors' attempts to resolve them.

Many studies of educational attainment have reported that students with specified characteristics, most significantly prior attainment (ability) and social class, will achieve more in schools where aggregate ability and social class means are high. The widespread belief in the influence of `school-mix' has the appeal of common sense, and gains at least qualified support from an extensive professional literature (Teddlie, Stringfield, & Reynolds, 2000; Thrupp, 1999). Wilkinson et al. (2000) are noticeably enthusiastic in their discussion of Bryk and Lee's (1993) study of attainment in United States Catholic schools. They observe that `this is probably the most robust and interesting statistical study to have been published because it provides strong inferential evidence of the links between school composition and school organization' (p. 92). Bryk and Lee reported that mathematics attainment was higher in Catholic schools than in other public schools, even when student intake characteristics were controlled by the available indices, thus leaving a residual school composition effect. These authors also found, however, that the effect disappeared when mathematics attainment was used as the dependent variable and the number of mathematics courses taken by students was controlled. In other words, there was no composition effect when a definitive system property of the school, a property of its structure as an educational institution and one directly relevant to the outcome, was included in the analysis. Wilkinson et al. (2000) are aware of this, and draw attention to `the methodological point that composition needs to be entered into the analyses ahead of organizational factors such as curriculum options, otherwise the compositional factors are obscured' (p. 92). The point, however, is not so much methodological as substantive. Indeed the remark that compositional factors are `obscured' unless, in this case, aggregate mathematics attainment is entered before the number of mathematics courses studied is telling. What are the `compositional factors' referred to by Wilkinson et al. that seem to disappear? It becomes obvious on reflection that there are no such factors. The statistical composition effect observed in Catholic schools is accounted for--and sufficiently explained--by the provision and take-up of mathematics courses. It is far from clear, moreover, that the greater than predicted mathematics results, or the high take up of mathematics courses, should be regarded as being the result of peer group processes. How, indeed, might such processes be recognised? All one might conclude from the data presented by Bryk and Lee (1993) is that no oppositional peer group culture with sufficient power to resist the practices sustained in Catholic schools has arisen. For Wilkinson et al. (2000), what is important, however, is the existence of a school composition effect taken as evidence of its causation by peer group processes--even if that can only be demonstrated by omitting from the analysis the school variables acknowledged as most likely to be responsible for the substantive effect.

The argument that only controlled statistical analysis can provide the demonstration required to observe an effect, made by Wilkinson et al. and Marks et al. (2001), has dramatic consequences that should be faced squarely. What might be called the standard narrative in the sociology of education maintains that social differences in educational attainment are generated largely by processes of cultural production by working-class students who resist schooling with a repertoire of practices drawn from their class experience. This is evidently a theory in which peer group influences are of paramount importance, but the critical ethnography on which it is based is indifferent, if not hostile, to the methods of quantitative analysis. The explanation of `how working-class kids get working-class jobs' fails for Wilkinson et al. because the processes described by Willis (1978) are not linked to the asserted effect, the overall failure of working-class students in the educational system, in the only way these authors, despite their varied methodological commitments, are prepared to recognise. Willis's study remains descriptive, in their terms, unless it is demonstrated by quantitative analysis that students who act like the `lads' are more likely than those who do not, the `ear'oles', to get working-class jobs when prior ability is controlled. Moreover, if the lads as a group were less scholastically able than others who succeeded, then the entire thesis is unsound. There is only Willis's word for it that they were not. As far as Wilkinson et al. (2000) are concerned, all the brilliance of such studies is wasted in so far as it is not demonstrated that ability, and perhaps genetic ability at that, is not responsible for the causal effect explained. This is what finally separates so-called quantitative and qualitative researchers. It is basically an argument about how relationships of cause and effect are to be taken as demonstrated. The distinction is not really between different ways of collecting data, or even about how it is to be analysed, but about how associations between observations can be detected and incorporated in causal explanations. There is obviously no possibility of cause and effect being established when there is deemed to be no effect to explain unless all relevant experimental or statistical controls have been applied in the methods used to obtain it.

In the end, despite their attempt at engagement and rapprochement, Wilkinson et al. allow the quantitative definition of `peer effects' by objective statistical analysis to dominate their text. The irony of all this is considerable, for such effects, which are actually composition effects, may not be caused by peer group processes at all. Marks et al. (2001), for example, found no socioeconomic status (SES) composition effect. Furthermore the effect due to aggregate ability was less than one percentile point and observed in only a handful of schools. It is likely that so-called composition effects are often caused by inadequate controls on school intake characteristics and are, therefore, effects of school selection. The assumption that there is no effective selection for individual schools on the basis of high aspirations and favourable dispositions towards education is unfounded and likely to be mistaken. The quantitative paradigm may, therefore, have come close to eliminating peer effects altogether, notwithstanding the deepening realisation in social science that peer group influences are more fundamental in the process of socialisation than allowed in the standard narrative of family-based practices.

The unreal power of the aggregate

The standard quantitative view that only controlled statistical analysis (or experiment) can demonstrate the existence of an effect to be explained has further consequences for the study of peer effects. Marks et al. (2001) argue, for example, that a causal effect may only be attributed to the school when an aggregated variable is associated with achievement after its individual effects have been taken into account:

   The logic of these analyses is that, for a contextual effect to be
   established, significant contextual effects must be identified net of
   individual-level factors that influence student performance. (p. 53)

This logic, of course, is actually a technical requirement and likely to be rejected by qualitative researchers. The argument that composition effects (whether at the level of the school or classroom) are caused by aggregate properties of the institution, implies that the aggregate properties of individuals can be regarded as properties of institutions. This assumption is highly problematic. Marks et al., for example, routinely devise scales from students' individual responses to questionnaire items and employ the mean values as an indicator of school climate, or some such concept. The argument has a rough plausibility. If it happens that a substantial proportion of students at a school are discontented, and if their attainments are lower than those in other schools, the relationship between those states of affairs may well be causal. That does not, however, make acceptable the argument that institutional properties are generated by the aggregated properties of individuals in this manner. If the morale of an army might be regarded as a real property of the collective--and this is at least arguable--then it has its existence in the dispositions of the soldiers. It is certainly their willingness to obey orders and act as members of a unit that creates an emergent property, a functioning social entity with properties appropriate to its nature, but such dispositions do not become, in some directly additive way, a property of the institution.

Marks et al., however, argue that properties of schools (referred to as school characteristics) may be assessed by aggregating student-level variables, those with the largest effects being self-concept of ability, school climate, and parents' aspiration for post-school study. Basically these are student perceptions of teachers' pedagogic and disciplinary practices and whether they can be regarded as an indication of real properties of the school should be supported with an adequate argument. It would be more satisfactory, to say the least, to construct better-specified properties, with a broader empirical foundation. These aggregated variables, it should be noted, make no contribution to tertiary education scores once their individual effect has been controlled (Marks et al., Table 24, Model 7). It is important to understand that there are no composition effects observed with any variable constructed from the aggregated responses of students to `Quality of School Life' items. Therefore what we have are correlations, of a moderate size, between a rather large number of variables--Table 25 lists 14--almost all of which are actually indicators of student dispositions. The difficulties with this approach to the study of institutional properties are considerable. It is quite indefensible to treat aggregate variables of this kind as real properties of institutions. Even if all theoretical arguments are put aside, there remains the substantive point that we have no evidence that students acquire these dispositions at school.

Nevertheless Marks et al. provide a list of so-called `school characteristics' that are simply aggregated individual values presented as measures of classroom and school properties. As all of these variables are inter-correlated, and all are associated with school differences in tertiary entrance score in a pattern difficult to summarise or interpret, the authors use a technical process to generate a simplified model. The implications of this are worth close attention. For example, to eliminate `classroom climate' and retain `school climate', accepting the order generated by `backward entry' in a regression analysis, conveys the inescapable suggestion that the institutional structures, individual dispositions, and social practices, that constitute the real properties of the school, can be conceptualised and measured in this manner. In this case, the variables are actually constructed from responses to the same three items (`students eager to learn', `students make good progress', and `students were well-behaved'), the values of which were then used, first, as a scale of classroom climate and second, in an aggregated form as a scale of school climate. It is hardly surprising that these variables effectively cancel each other out when they are used in the same statistical model as measures of individual and institutional properties.

Nominalism, another Humean legacy, constructs a social world in which the constant flow of activity at different sites is splintered into fragments, each of which may be treated as a valid object of measurement and treated as distinct, when there is, in fact, no adequate reason to support this procedure. The demonstrated intercorrelations between these `school properties' (which are created, of course, by the fact that children who agree with items such as, `I like to do extra work' and `I like to ask questions in class' tend also to agree with items such as `the work we do is interesting' and `most students are well-behaved') are noted, but substantively ignored because of the imperative to reduce and order the list of variables in the belief that the research will thus be of more practical use to policy makers. Yet, without well-developed heuristic skills, policy makers are all too likely to be misled by, for example, the suggestion that school climate is more important than classroom climate. Indeed, should that somehow be translated into the idea that school processes (controlled by the actions of principals?) are more important than classroom processes (controlled by teachers?), then the misunderstanding would be profound. And if this is not what policy makers should conclude from these findings, then what should they conclude? It seems risky, in fact, to base any sensible policy on a mechanism of ordering the list--a sort of survival of the fittest by backwards elimination--which, despite its apparent technical rigour, is almost meaningless. The declared end of this procedure, which is to provide policy makers with a useful idea of the factors that make a difference to educational attainment, so far from justifying the practice is a good reason to abandon it.

An illustrative case

The Progress at School data may be used to provide an illustration of what may be achieved by a realist approach to the study of peer groups and educational achievement (Nash & Harker, 1998). The longitudinal Progress at School project involved 5393 students attending 37 New Zealand secondary schools. A number of variables on individual students were obtained, including SES, prior attainment, and examination results at three different levels. Information was also obtained on some school characteristics. The project employed a version of the self-report Quality of School Life (QSL) instrument (Williams & Batten, 1981) used by Marks et al. (2001). It comprised 30 items of the kind, `teachers treat me fairly', `I know how to cope with the work', `I get tired of trying', `people think a lot of me', and so on, and including questions about future aspirations. The analysis presented here uses a `liking for school' factor highly loaded on the items `teachers are fair and just' (0.74), `teachers treat me fairly in class' (0.72), `teachers help me to do my best' (0.69), `teachers listen to what I say' (0.67), `teachers give me the marks I deserve' (0.47), and `I find that learning is a lot of fun' (0.44). Agreement with these items might be taken as evidence that students are, at least, content with the way they are treated at school. Marks et al. report that items of this kind are significantly associated with variance in educational achievement. The Progress at School research generates somewhat different findings.

It will be convenient to begin by testing the association between school climate and progress at school. The standard method is to correlate the slope from the regression of the prior ability (intake) scores with Year 10 marks in the national School Certificate examination with the aggregated `liking for school' factor. The correlation is not significant and it must be concluded that there is no overall association between school climate, measured in this way, and progress at school. The reasons for this are worth discussing. It should be noted immediately that perceptions of school are actually more favourable in working-class than in middle-class schools. This may seem surprising, but at least three reasons for this apparently negative finding, which has some positive implications, may be offered. First, there is the tendency of Pacific students (and to a lesser extent Maori students) to record favourable responses (which is probably the expression of a specific cultural reluctance to criticise adults in positions of authority); second, many middle-class students, particularly bright students of European origin, have become critical consumers of education and have no inhibition about expressing their opinions and, finally, it is just possible that multicultural, working-class, coeducational schools are more tolerant of the varied styles of self-presentation and modes of behaviour adopted by students. Many schools of this type, indeed, have little practical alternative but to be accommodating about non-essential matters, including the uniform code, given the need to concentrate on the priority matter of enforcing the non-violence policy.

The overall lack of association between favourable perceptions of school, social class, and attainment does not, however, mean that these dispositions of students, particularly when combined with high aspirations and positive academic self-concepts, have no consequences for progress at school. On the contrary, students who possess such an overall pattern of dispositions are significantly more likely to demonstrate relative educational progress even when ability and social class are controlled. Moreover these frames of mind are distinctly associated with actual peer groups, but nothing is as simple as it seems, as the following analysis will demonstrate. A careful study was conducted in one multicultural school in the Progress at School sample, and the argument will be illustrated by reference to those data. Knowledge about friendship patterns was obtained from sociometric questionnaires and from observations of interactions in and out of school. The focus will be on just two groups of working-class students, boys and gifts, mostly of Pacific and Maori origin, who entered secondary school with similar below average Year 9 mean ability test scores (the z-score means were: boys'--0.56 and gifts'--4).33). What is more, and it is the key point, these two groups responded in the same way to the `liking for school' items: there is no significant difference between the mean values of the two groups. Their perceptions of how they were treated at school were basically favourable and more or less identical. However, despite all this, the relative progress at school demonstrated by the members of these groups was very different. The gifts' group gained an average of 0.65, and the boys' lost--0.83. This is a difference of 1.48 standard deviations achieved in three or four years. One of the seven boys did not even attempt the School Certificate examination. The boys and gifts left school with different qualifications and entered, by and large, different occupations. Where the differences between these groups in prior ability, social class, and `liking for school' were trivial, the differences in their final attainment were substantial and had definite consequences for their future. This finding is offered as evidence of a peer effect in the most realist sense.

Although these boys and gifts recognised that they brought different attributes to their school, and understood that their aspirations and levels of motivation were related to their labour market expectations, the boys held little or no hostility to school, and were willing to accommodate those one or two members of the group who did attempt to gain educational qualifications. Most boys gave the impression of being as indifferent to school as it is possible to imagine: they spent little if any time talking about schoolwork, they read only as much as they were required to do, and basically did as little as possible. The gifts, on the other hand, did take more interest in their schoolwork, and were aware that the office and retail jobs they aspired to required at least a satisfactory level of literacy.

How can the mode of adaptation to school of these students be described? The groups had, in fact, a rather loose form of organisation. They had no names. The students themselves might refer to `our group', but that is about as far as it went, and others referred to them as `the Maori gifts', `Viv's crowd', and so on. The gifts' group, in particular, was not so much a group as a set of pairs linked together: this was particularly so in the case of Marisse and Yolanda, and Viv and Isabella. But the girls did associate together, they exchanged information, protected each other's reputations, respected each other's relationships with boyfriends, and when they went out at the weekend, there were generally two or three of them to keep each other company. The boys behaved in similar ways. On fine days, they played informal games of league at break times and they often associated in out-of-school activities. It is, in any case, always a little dubious to talk about groups as if the individuals who comprised them were interchangeable. Tom and Ed were related and good friends; in his first year at secondary school, Hone had associated with a European boy in another group and maintained that relationship to some extent; Wesley was accepted and, when not on his own, he was with this group, but he had a justified reputation as a bully, and was not much liked. Our observations carried out in Year 12 indicated, notwithstanding their Year 10 average mode responses to the Quality of School life items, that Hone and Wesley, in particular, were far from satisfied with their experience at school.

Students' friendship groups exist for the fundamental purpose of offering mutual support, and they tolerate considerable differences in personality, ability, and aspiration. Students' liking for school is, for many groups, not the most salient characteristic in their formation. The most alienated students, those who respond negatively to almost all items, did not seem to constitute actual friendship groups--this is not a set of dispositions on which friendship is based--but were more or less integrated with others and found among all social classes. The salient features in shaping friendships are, in this order, gender, ethnic origin, social class (of origin and anticipated destination), and demonstrated academic ability. Their characteristic mode of adaptation to school, which can often be recognised as classed trajectories, is not determined only by the social origin of students but, particularly in the case of those in the upper school, by their anticipated destination. By the age of 16, many students have more than begun the process of accumulating the symbolic and social capital that will shape their individual trajectory. The fact that students from working-class families, particularly when they are academically talented, are more often than not able to develop high ambitions and form a determination to invest in the school as a route to tertiary education and a professional occupation, does not signal the irrelevance of class ideology in the educational system. What is interesting, certainly, is the relative uncoupling of origins and destinations, and the degree of autonomy students seem to possess to adopt discourses and practices defined by the latter rather than the former.

One of the most important finding of the Progress at School project was that students with high aspirations, positive academic self-concepts, and favourable dispositions towards schooling make more relative progress at school than those with exactly contrary frames of mind. Marks et al. (2001) conclude, in similar fashion, that `effective schools are those which have environments that are conducive to learning, and where students feel they are performing well and have high educational expectations' (p. 61). However, whereas Marks et al. regard these individual characteristics as properties of the school, the Progress at School argument is more cautious. If there is a social entity to which a climate can be attributed, then it is the peer group not the school. Moreover, if the two groups of students studied here had attended different schools in the company of others like themselves, a highly significant `school effect' would have been observed. There is actually no evidence that students develop their characteristic dispositions of mind, which are often expressions of distinctive modes of adaptation to school mediated by extra-institutional pressures, as a result of school practices. In fact, the mode of adaptation is not so much to school, but to gendered ways of life more or less established within the class and ethnic communities from which they come and within which their trajectories generally ensure they will remain. The Bourdieusian influenced family resource framework that informs the Progress at School research recognises the effective resources of income, literate cultural capital, and social connections, as the material and symbolic basis for supportive educational practices within the home and, of course, within the school itself. The extent to which the family, the school, and peer groups are all involved in the shaping of students' adoption of practices that directly and indirectly affect their educational progress, requires a complex investigation. The assumption that practices--homework is a good example--can be located within a specific institutional site is an error with considerable potential to mislead. Whether homework is set, whether it is actually completed, and with what degree of conscientiousness, is a student practice, and students' actions in this regard are shaped by influences not necessarily under the control of the school. The fact that groups of students with similar attitudes towards school--at least as indicated by their responses to questionnaires--make different educational progress is actually no more encouraging for the standard positions in qualitative research than it is for quantitative research.

Conclusion

The Progress at School project held several conversations about peer influence with the students in our observational schedule. One of the most interesting exchanges took place between two working-class boys who, although close friends and members of the same informal group, maintained quite different aspirations throughout their secondary schooling. Warren completing his schooling and entered university to study environmental sciences, whereas Dion, who according to his earlier achievements was actually the more able, dropped out to work in retail sales. A field worker once asked them, `Do you get this peer pressure thing?'.

   Warren: It's more the friends type thing. The way I see it the peer group
   is more than your friends. The people like your friends, good friends, not
   your family as such but other people who are close to you, the ones who
   really know you, they're the ones that can hurt you the most. That's what
   I've found the people that are closest to you get to hurt you the most, the
   ones that can have the most influence over you. If it's just peers, someone
   just says, `Let's do something', Oh no! But if it's people close to you,
   yeah.

   Dion: Mind you, the peer pressure is not too bad in our group. Some of the
   people I'd hang around with from Thomas Street and all that. I remember we
   just used to go down to the local school at night and they'd have a few and
   they'd pass a joint around. They'd say, `Do you want one?' and I'd say,
   `No' and that was it. A lot of peer pressure would be, `Oh, go on, why
   not?', you know? But I didn't want it and I just said no.

   Warren: The people that teach you think it's people, peer group pressure is
   people making you say `yes' and `no'. It's not people saying, `Yes do it'
   when you get down to it. Like they're handing out joints, you just say no
   and it goes on to the next person. Peer group pressure is more of the
   people who are close to you and influence you. I think peer pressure is
   with all the people who are closest to you, and saying `yes', not people
   telling you `do this do this'. `Cus I mean with our friends if they say,
   `do this' you can say `no'. But they influence you into different sort of
   things.

   Dion: It's just like the desire to be accepted, eh? It's not like, `Oh, go
   on, go on.' `No no no.' But everyone's doing it and they're your friends.
   It's not so much that you don't fit in if you don't do it. It's just that
   you want to do what they're doing, they're your good friends, and you
   think, `oh, well, what the hell?' But it's not that they're saying, `Do
   this do this'. It's not physical. They don't say anything, it's just inside
   your own head, and you see everyone else doing it and you think, `I wanna
   be in that group', and it just happens.

The conversation between these boys is worth some attention. Students bring their desires to the groups they form, and these are shaped and mediated by social interaction, but peer pressure is rarely the most appropriate concept with which to grasp the reality of identity formation or the constitution of subjectivity.

Social differences in educational attainment are properly at the heart of all research in the sociology of education. The recognition of peer effects arises from models of learning in which discrete causal influences can be identified and measured. The term `peer effects' should be recognised as a theoretical concept within a particular scientific paradigm. The isolation of peer effects from family and school effects is made within an approach to statistical modelling with unmistakable neo-positivist foundations (Riordan, 1997). Some of the central problems of this approach have been discussed. The indirect approach to the measurement of peer influences is often unavoidable, given the nature of bureaucratic databases, but the full implications of that should be recognised: the observed effect may not be a genuine effect at all but an artefact, and even when it is, it may not be a genuine property of an actual group. Schools have system properties distinct to their constitution as educational institutions, which are all real emergent social properties, but social class composition is not a system property of the same kind. The concern of a realist social science, in fact, ought to be not with peer effects, which is to put the cart before the horse, but with the ways in which student peer groups have a causal influence on educational performance. The question for realism is this: In what fashion is educational attainment directly or indirectly affected by the practices of other students? In that sense, somewhat needless to say, the influences are likely to be extensive, and its answer requires a comprehensive discussion. There should be no doubt that school students influence each other's behaviour in a multitude of ways. Students have been recognised as active producers of varied cultural practices with a characteristic orientation to schooling. An extensive literature on modes of adaptation to school, generated within the symbolic interactionist sociological perspective influential in the 1970s, has been supplemented more recently by broadly based post-structuralist theories in which students achieve their subjectivity through investment in discursive positions available in the social environment. Although it will not be easy to forge a secure link between quantitative and qualitative research into peer effects, as the failure of the attempt by Wilkinson et al. (2000) indicates, the effort must be made. The linkage problem, as it might be called, is most unlikely to be solved within the theoretical and methodological constraints imposed by the dominant partner, and it will take a sophisticated appreciation of the conceptual foundations of social science to make genuine progress.

Keywords

educational environment

educational sociology

peer groups

qualitative research

statistical analysis

student attitudes

Table 1 Characteristics of two groups of Year 12 students

                    School
                 Certificate   Relative shift
                 English mean  Year 9-Year 11   Occupation at age 19

Group 1 (boys)
Tom                        27           -0.65  doorman
Ed                         34           -0.50  factory
Hone                       38           -2.10  petrol station
Wesley                     52           -0.01  unemployed
Arnaud                     33           -0.33  unknown
Bruce                 not sat               *   factory
Kahu                       25           -1.39  unemployed

Group 2 (girls)
Viv                        70            1.30  factory office
Laura                      38           -0.12  retail sales
Isabella                   55            1.00  clerical
Arohia                     50            0.51  polytechnic (design)
Moana                      58            0.79  polytechnic (business)
Lily                       60           -0.08  transferred school
Becky                      38           -0.62  kitchen worker
Jade                       56            1.53  reception
Marrise                    50           -0.17  library assistant
Yolanda                    49            0.71  clerical
Ali                        59            1.76  bank clerk

Note: School Certificate was usually taken in Year 11. The SC English
mean is 53, with a standard deviation of 15.5. Relative shift between
Year 9 and Year 11 is given by the residual from the regression of SC
English on an ability score derived from tests administered at the
beginning of Year 9. Destination was recorded from a follow-up study
carried out when these young people were 19 years old.

References

Bhaskar, R. (1993). Dialectic: The pulse of freedom. London: Verso.

Bourdieu, P. (1974). The school as a conservative force: Scholastic and cultural inequalities (J. C. Whitehouse Trans.). In J. Eggleston (Ed.), Contemporary research in the sociology of education (pp. 32-46). London: Methuen.

Bryk, A. S., Lee, V. E., & Holland, P. P. (1993). Catholic schools and the common good. Cambridge, MA: Harvard University Press.

Bunge, M. (1996). Finding philosophy in social science. New Haven, CT: Yale University Press.

Bunge, M. (1998). Social science under debate: A philosophical perspective. Toronto: University of Toronto Press.

Connell, R. W., Ashenden, D. J., Kessler, S., & Dowsett, G. W. (1982). Making the difference. Sydney: Unwin.

Gorard, S. (2000). Quantitative methods in educational research: The role of numbers made easy. London: Continuum.

Hammersley, M. (1989). The dilemma of qualitative research: Herbert Blumer and the Chicago tradition. [London: Routledge].

Harris, J. R. (1998). The nurture assumption: Why children turn out the way they do. New York: Free Press.

LeCompte, M. D., Millroy, W. L., & Preissle, J. (1992). (Eds.). The handbook of qualitative research in education. San Diego, CA: Academic Press.

Marks, G. N., McMillan, J., & Hillman, K. (2001). Tertiary entrance performance: The role of student background and school factors (Longitudinal surveys of Australian youth, Report No. 22). Melbourne: ACER.

Nash, R. & Harker, R. K. (1998). Making progress: Adding value in secondary education. Palmerston North: ERDC Press.

Riordan, C. (1997). Equality and achievement: An introduction to the sociology of education. New York: Longman.

Teddlie C., Stringfield, S., & Reynolds, D. (2000). Context issues within school effectiveness research. In C. Teddlie & D. Reynolds (Eds.), The international handbook of school effectiveness research (pp. 160-185). London: Falmer Press.

Thrupp, M. (1999). Schools making a difference: Let's be realistic! School mix, school effectiveness and the limits of reform. Buckingham: Open University Press.

Walker, J. C. (1988). Louts and legends: Male youth culture in an inner city school. Sydney: Allen & Unwin.

Wilkinson, I. A. G., Hattie, J. A., Parr, J. M., Townsend, M. A. R. (with Fung, I. & Ussher, C.), Thrupp, M., Lauder, H., & Robinson, T. (2000). Influences of peer effects on learning outcomes: A review of the literature: Report to the Ministry of Education. Auckland: UniServices Ltd, University of Auckland.

Williams, T. & Batten, M. (1981). The quality of school life (ACER Research Monograph No. 12). Hawthorn, Victoria: ACER.

Willis, P. (1978). Learning to labour: How working class kids get working class jobs. Farnborough, Berks: Saxon House.

Roy Nash is a sociologist of education at Massey University College of Education, New Zealand, Private Bag 11 222, Palmerston North, New Zealand. Email: [e-mail] (or:[e-mail])