Concurrent and short-term longitudinal associations between peer victimization and school and recess liking during middle childhood
Abstract
Background Prior studies outside of the UK have shown that peer victimization is negatively associated with school adjustment.
Aims To examine concurrent and short-term longitudinal associations between peer victimization (physical, malicious teasing, deliberate social exclusion, and malicious gossiping) and two measures of school adjustment (school liking and recess liking), and test if these associations were moderated by year and sex.
Sample A UK sample of 429 pupils in Years 4, 5, and 6 (Grades 3, 4, and 5, respectively, in USA) participated in the Autumn/Winter (Time 1) and 189 of these provided follow-up data during the Spring/Summer (Time 2) of the same school year.
Method Peer nominations of victimization, and self-reports of school adjustment were collected in individual and small group interviews.
Results At time 2 (but not Time 1), victimization predicted concurrent school liking among year 6 pupils but not among year 4/5 pupils, and victimization predicted recess liking among all pupils. Victimization also predicted changes in School liking among boys (not girls) and among Year 6 (not Year 4/5) pupils, and victimization predicted changes in recess liking among all pupils.
Conclusions The associations between victimization and poor school adjustment found elsewhere were replicated with this British sample. The implications of these results for children's social adjustment at school were discussed.
Negotiating the transition to school, and remaining well adjusted to school in the ensuing years, are important developmental tasks for young people. However, early studies showed that up to 30% of pupils may experience school adjustment problems (Achenbach & Edelbrock, 1981; Rubin & Balow, 1978). School adjustment has been defined, operationally and conceptually, in a variety of ways in the literature (see Ladd, 1989). Ladd (1990) used this term to refer to ‘the outcome(s) of children's attempts to adapt to the demands of the school environment’ (p. 1083), and he argued that those demands are more likely to be met when children exhibit positive attitudes towards school. Policy makers, too, have recognized that negative attitudes towards school can be detrimental to motivation and achievement in school (Rushton & Larkin, 2001; Woodhead, 1996, cited in Blatchford, 1998). For these reasons, the present study set out to investigate some of the factors that could influence children's views of school. More specifically, it considers the role of peer victimization in how much they like school and breaktime (‘recess’ in the US and henceforward).
Prior studies have examined the linkages between negative peer relationships and school adjustment generally. In a well-known review, Parker and Asher (1987) concluded that children who were poorly accepted by their peers were at greater risk for early school drop-out than were children who were better accepted. Ladd and his colleagues (Buhs, Ladd, & Herald, 2006; Kochenderfer & Ladd, 1996; Ladd, Kochenderfer, & Coleman, 1997) have carried out a programme of research to examine the associations between negative peer relationships, including peer victimization, and a specific aspect of initial adjustment to grade school – school liking/avoidance. In one report (Kochenderfer & Ladd, 1996), early victimization (during the fall of a school year) was found to predict increasing school avoidance (from fall to spring of that school year), but there was no evidence that early levels of school avoidance predicted increasing victimization. In a follow-up paper with the same participants, Ladd et al. (1997) found that out of a group of four peer relationship measures (number of friends, having vs. not having a very best friend, peer acceptance, and victimization), only victimization uniquely predicted school avoidance, and moreover, it did so both concurrently and predictively (i.e. from fall to spring). These results support the view that victimization is an especially important dimension of children's peer relationships that may contribute towards early adjustment to grade school.
Subsequent studies have confirmed that peer victimization is associated with more negative attitudes towards school and school personnel among grade school pupils aged seven and above in the USA (Buhs & Ladd, 2001; Buhs et al., 2006; Nansel, Haynie, & Simons-Morton, 2003; Wei & Williams, 2004), China (Boliang & Lei, 2003), and Australia (Skues, Cunningham, & Pokharel, 2005). The only study of these issues in the UK that we could locate was carried out by Smith, Talamelli, Cowie, Naylor, and Chauhan (2004). In contrast to findings from elsewhere, they found that on a single item (‘Mostly, do you enjoy school/’) there was no difference between victims and non-victims. It is too early to say if this signifies a difference between the UK and elsewhere in patterns of associations between school liking and peer victimization. On the one hand, this is possible in the light of findings that there are some considerable national differences in the patterns of associations between victimization and measures of social adjustment (Eslea et al., 2004). On the other hand, it is possible that social desirability could have influenced the participants' responses to the single item employed by Smith et al. (2004) to measure the extent of school liking, although this seems unlikely given that numerous negative views about other aspects of school were expressed in the course of the interviews conducted. The first aim of the present study was to provide data from a UK sample that would allow firmer conclusions to be reached in this regard. Specifically, a measure of school liking/avoidance was employed whose psychometric properties could be determined. Moreover, our study builds on that of Smith et al. (2004) because it tested the hypothesis that there exists an association between peer victimization and school liking/avoidance not only concurrently, but also longitudinally. Findings indicating longitudinal associations would allow firmer conclusions to be drawn regarding the role of peer victimization in children's socio-emotional development.
The present study also contributes to the literature because it employs a developmental perspective that compares children of different ages. It did so because studies have found developmental discontinuities in both peer victimization (Olweus, 1993; Smith & Levan, 1995; Whitney & Smith, 1993) and school adjustment (Ladd, 1989). Thus, it is important to study the linkages between victimization and school adjustment in pupils of diverse ages. In the present study, we did so for pupils in the final 3 years of junior school. Our decision to focus on these age groups was prompted by the finding that although there is a gradual decline in victimization across the grade school years, it still occurs with relatively high frequencies among this age range (Eslea et al., 2004; Olweus, 1993). As well as allowing comparisons with data from children of various ages reported elsewhere, the design of the present study also enabled us to test for school year (‘year’ henceforward) differences within our sample, that is, if age moderates the association between victimization and school adjustment.
It is now accepted that children can victimize their peers in many different ways. For instance, Bjorkqvist, Lagerspetz, and Kaukiainen (1992) distinguished between direct physical (e.g. hitting, pushing, kicking), direct verbal (e.g. name-calling, threatening), and indirect (e.g. telling tales, spreading rumours) forms. Crick and her colleagues (Crick, 1995; Crick, Bigbee, & Howes, 1996; Crick & Grotpeter, 1995) identified relational aggression which is directed at damaging victims' peer relationships. While it has been shown that the different types of victimization often co-occur, they do not all uniquely predict psychosocial maladjustment (Cullerton-Sen & Crick, 2005; Hawker & Boulton, 1997). Consequently, the present study measured several of the commonly found measures of peer victimization.
The present study also examined the impact of victimization on pupils' attitudes towards breaktime (called ‘recess’ in US, and below). It is becoming increasingly clear that pupils' views concerning recess are an important dimension of their overall feelings about school (Blatchford, 1998; Pellegrini, 2005). On the positive side, the great majority of pupils hold positive views about recess and it is often their favourite part of the school day (Blatchford, 1998; Sutton-Smith, 1990). Blatchford (1998) asked 11 year old pupils what they liked best about recess and found that ‘break from work’, ‘playing games’, and ‘being with friends’ were the most common categories of response. Recent studies have shown that children benefit in many different ways from recess, particularly via their positive experiences with peers during unstructured play (Blatchford, Baines, & Pellegrini, 2003; Pellegrini, 2005; Pellegrini, Blatchford, & Bains, 2004; Watkins, Dwyer, & Nielsen, 2005). These findings attest to the importance of studying recess in the context of children's positive socio-emotional development.
However, it is also apparent that recess is when many undesirable peer-directed behaviours take place including bullying (Slee, 1995; Whitney & Smith, 1993), racist name-calling (Kelly & Cohn, 1988), teasing (Mooney, Creeser, & Blatchford, 1991) and fighting, and other forms of aggression (Boulton, 1993; Tapper & Boulton, 2004). Many pupils also express concerns about having to spend time on the playground where these things take place (Blatchford, 1998; Tizzard, Blatchford, Burke, Farquhar, & Plewis, 1988). Given these findings, it is somewhat surprising that few researchers have examined the impact of victimization on children's liking of recess. Among those that did, Slee (1995) found that self-identified victims were significantly more likely to report feeling unhappier and less safe in the playground. Smith et al. (2004) followed up a sample of pupils over a 2 year period and found that victims reported lower liking of recess compared to non-victims. These studies are important because they were the first to suggest an association between peer victimization and reduced recess liking. However, methodological issues mean that conclusions must remain tentative. Specifically, the results reported by Slee (1995) and Smith et al. (2004) are open to shared method effects since peer victimization and recess liking were both measured with self-reports. So far as we can tell, no study in this area has avoided shared method variance, none has employed a dedicated scale to measure pupils' liking of recess (as opposed to one or two items), and none has utilized a longitudinal design in which earlier levels of victimization were used to predict changes in liking of recess. The present study builds on extant literature by incorporating these three methodological principles. Thus, a further aim of the present study was to test the hypothesis that peer victimization is associated, concurrently and over time, with recess liking.
For each of the aims expressed above, we also set out to investigate if sex moderates the association between victimization and school adjustment. This is important given that sex differences have emerged in prior studies of the key variables examined in the present study; girls have been found to be involved in less physical but more ‘psychological’ victimization than boys (Ahmad & Smith, 1994; Bjorkqvist et al., 1992; Crick et al., 1996; Rivers & Smith, 1994; Skues et al., 2005); girls expressed greater fear of being victimized in the future than boys (Boulton, Trueman, & Murray, submitted); boys tend to have more favourable views of recess than girls (Blatchford, 1998; Eslea et al., 2004); Simmons and Blyth (1987) have documented sex differences in school adjustment; and Smith et al. (2004) found that boys reported significantly greater liking of school than girls, although the size of the mean difference was small (2.82 vs. 2.73 on a five-point response scale). Moreover, other studies have found that patterns of associations between peer victimization and other measures of psychosocial adjustment (i.e. not school/recess liking) vary as a function of sex (Crick & Bigbee, 1998; Eslea et al., 2004; Lopez & DuBois, 2005).
In summary, the present study tested the hypotheses that there would be concurrent and longitudinal (i.e. predictive) associations between victimization and (1) school liking/avoidance and (2) recess liking, and also if any associations detected were moderated by school year or sex. It builds on previous work by overcoming reliance on self-reports of peer victimization and the associated problem of shared method variance, and by developing a dedicated scale to measure recess liking.
Method
Participants
This study collected data from 429 children that were drawn from 10 junior schools in the UK. At each school, the head teacher selected from one to three classes for participation based on their availability during periods of data collection. Parental consent was then requested for all of the children in these classes, and was obtained in 96% of cases. At Time 1 (November–January) there were 38 girls and 39 boys in Year 4 (mean age=8.8 years, US Grade 3), 94 girls and 88 boys in Year 5 (mean age=9.5 years, US Grade 4), and 86 girls and 84 boys in Year 6 (mean age=11.3 years, US Grade 5). At Time 2 (May–July of the same academic year) there were 22 girls and 22 boys in Year 4, 30 girls and 27 boys in Year 5, and 42 girls and 46 boys in Year 6 (N = 189). The smaller n at follow-up was due to several schools only allowing access to children at Time 1. Given that the two lower Years were close together in age relative to those in Year 6, the analyses of Year differences was based on a comparison between Years 4/5 combined and Year 6.
Procedure
Data were collected in individual or small group (2–4 children) interviews, using standard protocols. Pilot testing had shown that most children could not complete all of the items in a single sitting, and so at both Time 1 and Time 2 each child was interviewed on two occasions separated by a day or two. These were carried out in a quiet location away from each child's home classroom, where it was felt that the close proximity of classmates could inhibit responses to many of the socially sensitive items. Prior to the interviews, the children were reminded that their participation was voluntary, that they were not being tested, that there were no right or wrong answers, and that their responses would remain confidential.
Measures and data reduction
Victimization
Participants were shown either head and shoulders photographs, or a class list, of all classmates, and asked to indicate which individuals were targeted by other pupils in their school for four specific victimization behaviours (see below). They were told that they could give as many or as few nominations as they wanted. Four separate items were used in an attempt to index the different manifestations of victimization – relational, indirect, physical, and verbal – identified in prior research (Bjorkqvist et al., 1992; Crick & Grotpeter, 1995; Rivers & Smith, 1994). The items were ‘Gets hit and kicked by other kids for no reason’, ‘Gets called nasty names by other kids for no reason’, ‘Gets left out of games and things for no reason’, and ‘Gets called nasty things by other kids behind their back for no reason’. For each child, the percentage of classmates that nominated her/him for each of the four subtypes was calculated. These four scores were all highly correlated (see Table 1), and maximum likelihood factor analysis of Time 1 data indicated one main factor with an Eigen value of 3.08 (3.24 at Time 2), that accounted for 77.0% of the variance (80.9% at Time 2), and a minimum factor loading of .84 (.82 at Time 2), and Cronbach's α was found to be .92 (.94 at Time 2). Hence, an overall ‘victimization’ score was computed as the average of the four subtypes, with high scores indicating that more victim nominations were received.
School liking
This was measured with 12 items from the school liking and avoidance scale devised by Ladd and Price (1987), and modified by Kochenderfer and Ladd (1996) and Ladd et al. (1997). All items were followed by three response options (‘Agree a lot’, ‘Agree a bit’, and ‘Disagree’), scored 1–3. Reverse scoring was then used so that for all items, a high score indicated greater liking of school. Maximum likelihood factor analysis indicated it was appropriate to regard the 12 items as representing a uni-dimensional scale. Specifically, at Time 1, the single dominant factor had an Eigen value of 6.0 (7.2 at Time 2), accounted for 50.3% (60.2% at Time 2) of the variance, and all items loaded above .47 (above .60 at Time 2). Internal reliability was very high (Time 1 α=.92 and Time 2 α=.95). Thus, a total score across all 12 items was computed for each participant, and this was used in all subsequent analyses. This variable was labelled ‘school liking’ henceforward.
Recess liking
This was indexed with the recess liking scale developed by the first two authors for the present study. It consists of seven items (‘How much do you look forward to playtime/’, ‘How much do you like playtime/’, ‘How happy do you usually feel at playtime/’, ‘How would you feel if you had to stay in at playtime/’, ‘How much would you like to stay indoors at playtime/’, ‘Do you think playtime should be made longer or shorter/’, and ‘How would you feel if you school stopped having playtime/’), each followed by five response options scored from 5 (very positive attitude, e.g. ‘I really look forward to playtime’) through to 1 (very negative attitude, e.g. ‘It upsets me a lot to think about playing out’), with a neutral attitude reflected by the middle response (e.g. ‘I neither look forward nor get upset thinking about playtime’). Maximum likelihood factor analysis confirmed a single factor (Eigen values=2.74 at Time 1 and 2.21 at Time 2; 39.2% of variance accounted for at Time 1 and 31.6% at Time 2; and minimum factor loading=.52 at Time 1 and .39 at Time 2). Internal reliability was high (Time 1 α=.81 and Time 2 α=.76). Thus, a recess liking score was computed as the total across the seven items, with a high score indicating greater liking of recess.
Results
Correlations among variables
The zero-order correlations among the study variables are shown in Table 1. The two school adjustment variables were significantly correlated but the coefficients were not high at either Time 1 or Time 2. Each of them exhibited significant but only moderate stability from Time 1 to Time 2. At Time 1, Victimization was not significantly correlated with either school liking or recess liking, but at Time 2, it was significantly correlated with both of them. Year was significantly correlated with Time 1 and Time 2 victimization, and with recess liking at Time 1. Sex was significantly correlated with Time 1 and Time 2 victimization and recess liking at Time 1 and Time 2. This pattern of correlations means it is appropriate for us to test the hypotheses of concurrent and longitudinal (predictive) associations between victimization and school liking and recess liking, and for Year and sex differences in those associations.
Concurrent analyses: Does victimization predict concurrent school adjustment/
Data from Time 1 and Time 2 were used to test for concurrent associations between victimization and school adjustment, and for sex and Year differences in those associations. In each case, two hierarchical multiple regression analyses were computed, one in which school liking served as the criterion, and the other in which recess liking served as the criterion. The procedure described by Aiken and West (1991) guided these analyses (and the longitudinal analyses, see below), including dummy coding of categorical predictors (sex and Year, with the latter based on Years 4 and 5 collapsed and compared with Year 6), centring of continuous predictors so they had a mean of zero (victimization, school liking, and recess liking), testing interaction/moderation with product terms (e.g. sex × victimization), using a hierarchical order of entry of predictors in which lower order terms are entered prior to higher order interaction terms, and where significant interaction/moderation effects are interpreted on the basis of follow-up regressions for each of the levels of the relevant categorical predictor. In each regression, sex, Year, and victimization were entered on step 1, and sex × victimization and Year × victimization product terms entered at Step 2. The rationale for this design was that sex and Year were controlled at Step 1 because they were significantly correlated with victimization (see Table 1) and because they had to be entered into the regression model prior to testing higher order interaction terms that involved them, and this enabled us to examine if victimization was associated with the two dependent variables independent of its association with sex and Year. The Step 2 product terms allowed us to test if sex and Year moderated the association between victimization and the two dependent variables. Table 2 summarizes the results.
At Time 1, victimization was not able to account for a significant proportion of the variance in school liking or recess liking. Additionally, neither the sex × victimization nor the Year × victimization interactions was significant. In contrast, at Time 2, victimization significantly predicted school liking (after controlling for the effects of sex and Year), such that high victimization was associated with low school liking. This was qualified by a significant Year × victimization interaction such that among Year 6 pupils (β=−0.39, p<.001, adjusted R2=.14), but not Years 4/5 pupils (β=−0.05, ns), high victimization was associated with low school liking. Additionally, victimization was significantly associated with recess liking (after controlling for the effects of sex and Year), such that high victimization was associated with low recess liking. Neither sex nor Year moderated this relationship. These results are summarized in Table 2.
Longitudinal analyses: Does earlier victimization predict changes in school adjustment/
Two hierarchical multiple regression analyses were computed to test the hypothesis that victimization predicts changes school adjustment, and to test for Year and sex differences in those associations. In one model, Time 2 school liking served as the criterion, and in the other Time 2 recess liking served as the criterion. The Time 1 school adjustment measure that corresponded with the dependent variable was entered on Step 1 (i.e. if Time 2 school liking was the dependent variable, Time 1 school liking was entered on Step 1). This allowed us to test for changes in the measures of school adjustment. Sex, Year, and Time 1 victimization were also entered on step 1 for the same reasons outlined in the previous paragraph. Similarly, the sex × Time 1 victimization and the Year × Time 1 victimization product terms were entered at Step 2, for the reasons outlined in the previous paragraph. The results are summarized in Table 3.
Time 1 victimization significantly predicted changes in school liking such that high victimization was associated with negative changes in school liking. This main effect, though, was qualified by sex and by Year, such that the association was evident among boys (β=−0.32, p<.001, adjusted R2 change=.10), but not girls (β=0.02, ns), and among Year 6 pupils (β=−0.37, p<.001, R2 change=.14), but not Years 4/5 pupils (β=−0.02, ns). Additionally, victimization was associated with negative changes in recess liking, but neither sex nor Year moderated this association.
Discussion
The present study has shown that the negative association between peer victimization and liking of school found in kindergarten pupils and in grade school pupils outside of the UK (see Introduction), is also evident in junior school pupils in the UK but with certain caveats. Our findings revealed that this association was evident at Time 2 but not Time 1, and was moderated by year. Additionally, the present study provides the most detailed test to date of associations between victimization and liking of recess. All of these issues will be discussed below.
School liking
There was no evidence of a significant association between victimization and liking of school at Time 1. Such null findings are in accord with those of the only other study with British pupils which failed to find any evidence that victims disliked school more than non-victims (Smith et al., 2004). In contrast, at Time 2, victimization was negatively associated with liking of school. Moreover, the longitudinal analysis provides our strongest evidence to support the hypothesis that victimization contributes to reduced liking of school, a finding in accord with those reported in the literature in samples from outside of the UK (Boliang & Lei, 2003; Buhs & Ladd, 2001; Buhs et al., 2006; Nansel et al., 2003; Skues et al., 2005; Wei & Williams, 2004).
A possible way to reconcile these discrepant findings rests on the dynamic nature of peer relationships within a cohort of children. Specifically, it is possible that within a class, victimization between classmates is more unstable at the outset of the school year compared to later because the children are getting to know new members of their class. Indeed, most of the classes that participated in our study were made up of pupils that had belonged to different classes in the previous year. Such a suggestion is consistent with the evidence and arguments presented by Eslea et al. (2004) that the relationship between school bullying and its various risk factors should become stronger with age as roles within the peer group stabilize. However, our explanation is undermined by Kochenderfer and Ladd's (1996) finding that even within the first school term of kindergarten, the concurrent association between victimization and school liking/avoidance was significant. Developmental changes may be important here. For instance, it is possible that there is more instability in victimization in ‘newly formed’ classes of grade school pupils than in kindergarten classes.
Taking all of the above into consideration, what can be said with more certainty is that an inverse association between victimization and school liking is likely but not inevitable. More data are needed to examine factors that may mediate or moderate such an association including the familiarity of pupils with one another. The present study did test the possibility that sex and Year could serve as moderators in this regard, and these will be discussed in turn.
So far as the latter is concerned, our findings show that concurrently (at Time 2), the relationship between victimization and liking of school was evident for Year 6 pupils but not for Year 4/5 pupils. Moreover, earlier victimization predicted negative changes in liking of school among year 6, but not year 4/5, pupils. Here it is necessary to concede that our results do not lend themselves to straightforward conclusions that imply that one age group is more vulnerable to the negative effects of victimization than another. When they are considered alongside reports of a significant association between victimization and school maladjustment among children aged 13 years (Nansel et al., 2003), it would seem appropriate to accept that such an association is possible although not inevitable across the middle years of childhood. We think it is useful to highlight this for those at the upper end of this age range given that (1) the association was most evident in the oldest pupils in our sample and (2) victimization tends to decline in frequency across the school years (Olweus, 1993; Whitney & Smith, 1993). We think it is not unreasonable to suggest that older victims may be even more stigmatized by their peers than younger victims, or may be more aware that they are ‘different’ from the majority of their non-victimized age-mates. These things in turn could contribute to greater disliking of school among the older victims relative to the younger victims. Our findings and suggestions imply that it would be unwise to think that the general age decline in victimization means that victimization becomes a less serious problem at the upper range of middle childhood (i.e. 11–13 years) compared to the earlier years. While it may involve fewer individuals, the effects on those that are subjected to peer harassment may be just as pronounced, or even more so, in the light of our longitudinal findings and those from Time 2.
Sex was found to moderate the predictive relationship between earlier victimization and negative changes in liking of school in the present study, the relationship was evident among boys but not girls, but it did not do so concurrently. These findings add to the literature which sometimes has (Boulton et al., submitted; Crick & Bigbee, 1998; Eslea et al., 2004; Lopez & DuBois, 2005), and sometimes has not (Boulton et al., submitted; Kochenderfer & Ladd, 1996; Wei & Williams, 2004), found that sex moderated patterns of association between victimization and other diverse measures of psychosocial adjustment. While it would be premature to conclude that victimization is more damaging to boys' than girls' psychosocial adjustment, including liking of school, the finding from elsewhere that boys tend to be more reluctant than girls to seek social support when they are bullied (Boulton, 2005; Naylor, Cowie, & del Ray, 2001) provides ideas for why boys may be particularly vulnerable to such outcomes.
Recess liking
A unique contribution of the present study stems from its detailed consideration of the associations between recess liking and peer victimization. As was the case with liking of school, it was only at Time 2, but not Time 1, that victimization was negatively associated with recess liking (see above for possible explanation). Stronger evidence for this relationship came from our finding that earlier victimization predicted negative changes in recess liking. In contrast to some of the findings for liking of school, there was no evidence that sex or school year moderated any of these significant associations.
These findings indicating negative associations between victimization and recess liking are not surprising given that the playground is a common location for various forms of peer victimization and antisocial behaviour (see Introduction). However, given that recess is enjoyed by the vast majority of pupils at this age, and is often seen as the most enjoyable part of the school day, it should be a cause for concern among educators. It would not be unreasonable to suggest that playground victimization contributes towards a decline in recess liking, which in turn could precede growing negative attitudes to school in general, which in their turn may contribute to truancy and school refusal (see below). Future studies are clearly warranted to investigate these developmental possibilities.
In practical terms, we would suggest that attempts to help pupils maintain good attitudes towards recess, or improve them where appropriate, augment rather than replace schools' efforts to tackle the problem of bullying. Several promising interventions have emerged in recent years including training of supervisory staff to deal effectively with playground aggression (Boulton, 1996), and psycho-educational programmes (Frey et al., 2005; Hutchings, Owen, & Gwyn, 2004).
The present study had both strengths and weaknesses. On the positive side, the sample size was relatively large and was drawn from many different schools. Moreover, we overcame a reliance on self-report measures of victimization, and so it was not possible that significant associations between victimization and (self-report) school adjustment could have been inflated by shared method variance. This was a potential problem in some prior comparable studies (Ladd et al., 1997; Kochenderfer & Ladd, 1996; Smith et al., 2004).
Another strength was that our measure of victimization included four of the commonly observed forms. However, the strong correlations among these subtypes in our sample precluded an investigation of the association between each of them and school/recess liking. This would be a valuable contribution to the literature since subtypes do not all uniquely or equally predict psychosocial maladjustment (Cullerton-Sen & Crick, 2005; Hawker & Boulton, 1997). Moreover, teachers and pupils are less likely to view some forms of victimization, such as name-calling and deliberate social exclusion, in a negative way compared with physical victimization (Boulton, 1997; Boulton & Hawker, 1997), even though studies have shown that these non-physical assaults may be associated with at least as much distress as physical assaults (Crick & Grotpeter, 1995, 1996; Cullerton-Sen & Crick, 2005; Hawker & Boulton, 1997). If researchers were able to demonstrate that non-physical forms of victimization contribute to school adjustment problems in their own right (i.e. independently of physical victimization), then teachers may be more likely to implement intervention programmes to tackle these types of victimization and not just the ones they think are important (such as hitting and kicking).
Selection bias is another potential weakness of our study since head teachers chose the actual classes who participated in it. However, as was noted above they did so on the basis of simple availability during periods of data collection, and prior to being made aware of the specific aims of the study. Hence, it is unlikely that they were biased to selecting classes that they thought were especially unproblematic in terms of peer victimization and disliking of school/recess.
The analytical approach adopted in the present study also requires critical evaluation. The multiple regression approach employed here did allow the hypotheses of the study to be tested and, indeed, could be regarded as the ‘method of choice’ in studies of this type up until relatively recently. However, structural equation modelling (SEM) can now claim the latter title because it has a number of important advantages over multiple regression (Schumacker & Lomax, 2004). Not least among these is the fact that it can account for measurement error. Even though the measures of the present study had good reliability, controlling for measurement error would enable a better test of associations between peer victimization and school/recess liking. Furthermore, SEM allows for the testing of more complex models in the same analysis. For instance, whereas separate regression models were required for school liking and recess liking, these can be considered together in a future study that employed SEM. This latter point is relevant to theory-building as well since we currently lack a comprehensive account of how multiple predictors (including subtypes of peer victimization) impact upon children's overall school adjustment (including school and recess liking). Researchers concerned to rectify this state of affairs would do well to use SEM in future studies.
Implications of victimization – school/recess liking linkages for children's school careers
The implications of low liking of school/recess are not trivial. The results of the present study together with those reported elsewhere indicate that victimization is often related concurrently and longitudinally with low school/recess liking (see Introduction). Smith et al. (2004) reported that stable victims had higher school absenteeism rates than stable non-victims. Strikingly, they found that one third of stable victims admitted that fear of bullying had led them to truant from school. Together, these findings provide indirect support for an unfolding pattern of events that involve victimization, low school/recess liking, and on to absenteeism. This is an issue worthy of future studies.
In summary, the present study provides the most direct and detailed evidence to date that victimization is negatively associated with school and/or recess liking both concurrently and longitudinally in UK pupils. We argue that these findings, and those collected elsewhere, need to be considered alongside those from investigations which have revealed that victimization is also associated with other adjustment problems such as low self-esteem, depression, and anxiety (Hawker & Boulton, 2000). Importantly, such factors (i.e. disliking of school and adjustment problems) have been shown to predict early high school drop-out. Clearly, continuing efforts to tackle the problem of peer victimization in all its manifestations must remain a high priority if we are to ensure children adjust well to school and receive the maximum benefits of doing so.
