Psycho-emotional disablism: a differentiated process

This article presents analyses of data from a Norwegian survey. By way of descriptive statistics and linear regression analyses, the study investigates the relations between disability, psycho-emotional well-being, social participation, bullying and violence. It is demonstrated that disability is strongly associated with low psycho-emotional well-being. Relying on Carol Thomas’ concept of ‘barriers to being’, it is argued that this finding testifies to a disabling process. Her work also suggests that this process is compounded with other disabling processes such as barriers to social participation. However, the empirical findings presented in this article do not suggest such convergence. Instead, bullying and violence prove to be more important than one might imagine. It is argued that disabling processes should be conceptualized as differentiated rather than compounded, and that research on violence could provide soil for theoretical development.


Introduction
Among the myriad so-called social models of disability, some have emphasized psycho-social aspects more than others have. Among these, Carol Thomas' work is perhaps the most influential (Shakespeare 2006, 36Á45, Goodley 2010. Within a feminist materialist perspective, she repeatedly emphasizes (and relies on) 'the epistemological importance of ''experience''' (Thomas 1999, 73Á81). So doing, she also gives psycho-emotional aspects an equal ontological status as 'part and parcel of disability itself' (Thomas 1999, 66). Hence, her work revolves consistently around the argument that 'the crude equation of disability with restricted activity is a completely inadequate starting point' (Thomas 2004, 581).
To avoid this equation, she has developed the distinction between 'barriers to doing' and 'barriers to being'. The term 'barriers to doing' denotes the barriers to agency, primarily (but not limited) to physical barriers and socioeconomic exclusion. The term 'barriers to being', on the other hand, refers to barriers to 'self-esteem, personal confidence and ontological security' (Thomas 2007, 72). Thomas postulates that these forms of barriers are 'interactive and compounded' (Thomas 1999, 46, emphasis added), that is, convergent. The overall purpose of this paper is to discuss that particular postulate in the light of empirical findings from a Norwegian survey data-set. This purpose is hopefully, achieved in four stages. First, there is a deductive stage, where certain hypotheses are derived from Thomas' work (and, to a lesser extent, from the social model tradition in a wider sense). Second, there is a descriptive stage which consists of descriptive analyses of the data from Young in Norway 2002 (hereafter, YiN). Third, there is a comparative stage which investigates possible match between the descriptive findings and the 'outputs' postulated in Thomas' arguments. Fourth, there is an inductive conclusion, where a last 'convergent' perspective on disabling processes is outlined.

Framework and hypotheses
Carol Thomas is not the only one to emphasize psychological and emotional aspects of disabling processes. Within less materialist frameworks, her work is mirrored by conceptualisations such as Gregor Wolbring's work on ableism and self-identity or Fiona Kumari Campbell's work on 'internalised ableism' (Campbell 2008;Wolbring 2011;Campbell 2012). However, relevant research with disabled children and adolescents rely primarily on Thomas' work (Kelly 2005;Connors and Stalker 2007;Islam 2008;Watson 2012). Hence, her work provides the primary theoretical framework for this study.
Thomas' conceptualisation of 'barriers to doing' and 'barriers to being' belongs to a larger framework for understanding these disabling processes. More precisely, these barriers belong to a process she defines as 'disablism'. With the exception of the definiendum, which has been altered to 'disablism' from 'disability', Thomas' definition is exactly the same as that in her groundbreaking book Female forms (Thomas 1999, 60)?: Disablism is a form of social oppression involving the social imposition of restrictions of activity on people with impairments and the socially engendered undermining of their psycho-emotional well-being. (Thomas 2007, 73) In this per genus et differentiam definition, disablism is first included in a larger genus as 'a form of social oppression'. Then, two differentiae (traits that separate disablism from other forms of social oppression) are pointed out: 'barriers to doing' (restrictions of activity) and 'barriers to being' (undermining of [. . .] psychoemotional well-being).
This definition is the theoretical driving shaft of Thomas's analyses. When examined closely, it demonstrates that her convergence postulates are far from peripheral or secondary. Instead, they testify to the importance of social structure in her work. First, the genus 'social oppression' is clearly a structural concept in Thomas's analytical vocabulary, not least in relation to disability: The concept of disability refers to the relationship of ascendancy of the non-impaired over the impaired. Disability, like patriarchy, is a form of social oppression'. (Thomas 1999, 40) Second, the differentiae are themselves structural notions. The coordinating conjunction and indicates that they are compounded or at least structurally related. As opposed to or, it suggests that these social phenomena occur together. Thomas does not define disablism by way of one differentia, or the intersection of the two, but by converging barriers (to doing and to being, respectively). While Thomas, of course, does not claim that people with impairments are always subjected to both forms of disablism, it is the convergence, or at least the mechanisms underpinning both forms, that is of interest in her theoretical work. Thomas analyses different (and not always empirically correlated) forms of disabling barriers in order to analyse an overall (structural) pattern of converging barriers.
So doing, Thomas makes at least three postulates. In this article, those postulates will be translated into hypotheses, which will then be discussed through descriptive statistical analyses: (1) Disabled respondents report less psycho-emotional well-being than their nondisabled peers.
(2) Disabled respondents report a lower degree of social participation than their non-disabled peers.
(3) Low psycho-emotional well-being converges with low social participation (at least among disabled respondents).
Thomas' strong emphasis on structure Á which leads her to postulate convergence Á demonstrates how her perspective is an expansion of the now classical UK social model tradition. While other feminist scholars have proposed that disability studies should leave this model behind (Schriempf 2001;Gabel and Peters 2004), Thomas explicitly sees her own work as a reformulation Á or even a 'rescue' Á of crucial insights in the UK social model: the radical ideas that laid the foundation for the social model of disability contain a social relational kernel that now needs to be rescued and developed. [. . .] However, I am certainly not suggesting that the social model of disability be abandoned. (Thomas 2004, 22, 33) At least according to Thomas, her postulates are not specific to her work. She argues that a 'social relational' understanding not only holds together large parts of UK disability studies, but also 'needs to form the explicit foundation for future work' (33). To the extent that the YiN findings pose a provisional challenge to Thomas' framework, this challenge is also relevant to large areas of disability studies.

Data
The study examines data from the Norwegian national youth survey YiN [Ung i Norge 2002] (Rossow and Bø 2003). The survey was carried out in a school setting Á in 73 schools in different parts of Norway Á giving a fairly high response rate (92, 3%). Although girls are slightly under-represented in the survey, compared to general demographic data, this gives an acceptable representativity. With 683 variables, and a valid n of 11,928, this is a large survey in a Norwegian context. The respondents are evenly distributed within the age span of 13Á18. While some variables are analysed without recoding, most analyses deal with indexes that are recoded into categorical variables. Disability (in a social sense) is not measured in the survey. Instead, impairment is measured by five questions: (1) Do you have an illness or condition that leads to a lot of absence from school?, (2) Do you use a hearing aid?, (3) Do you have a motor impairment, without using any technical aids?, (4) Do you use a wheelchair, a walker, or other technical aids (due to motor impairment)? and (5) Do you have another physical disability?
Rather than analysing these variables separately, which would deal with very limited numbers of cases, a variable called 'impairment status' was constructed through a four-stage process: (1) the missing values were replaced by the value 0. (2) An additive index was constructed whereby only individuals with no reported impairment status were given missing values.
(3) If respondents reported four or five impairments, which would be very unlikely, they were recoded as missing. (4) Finally, the additive index/indexes was/were recoded into a dichotomous variable: while the respondents who had reported one or more impairments were given the status 'disabled', the remaining were given the status 'non-impaired', producing a final valid n of 10,387, of which 564 (5.4%) were defined as 'disabled'.
These data are not fully commensurate with social definitions (or ontologies) of disability. Nevertheless, it is commensurable with common empirical operationalisations. First, Thomas defines disablism as 'a form of social oppression' that specifically affects 'people with impairments' (Thomas 2007, 73, emphasis added). Second, defining the disability studies tradition even broader, Dan Goodley describes its rationale as following: 'undoubtedly, societies subject people with impairments to discrimination. [. . .] Disability studies respond to these acts of political and cultural life.' (Goodley 2010, 9). Finally, the cost of discrediting YiN, with reference to its implied ontology, would be far greater than the benefit. The dismissal would, in fact, render most quality of life surveys in Europe useless to disability studies.
Psycho-emotional well-being is not specified to particular (external) objects or specific action contexts. In YiN, self-perception is measured by a well-established psychological research instrument: an abbreviated Norwegian version of the American psychologist Susan Harter's Self-perception Profile for Adolescents (SSPA) (Wichstraum 1995). This 25-item variable instrument has fairly high data validity, as the missing values' percentages vary between 0.9 and 3.1%. Five aspects of self-perception are each measured by these five variables: Global self-worth, school performance, close friendship, attractiveness and physical appearance.
Beside domain-specific indexes, two general well-being indexes were constructed. On the first index, each category represents a quartile along a simple additive index of the SPPA variables. The second index, on the other hand, is constructed to reflect that the original variables are categorical rather than numerical: In the SPPA, each question asks the respondents to choose between four possible replies, indicating very high, high, low or very low well-being. Hence, the second index distributes the individuals according to their average choice among the four alternatives.
Social participation is measured by 18 questions in YiN, where respondents are asked if they took part in this or that activity last week, and if so how many times. From these variables, three additive indexes were constructed to measure different aspects: the breadth of social participation (the number of interaction arenas), the frequency of free time participation (the number of interactions), and finally the frequency of unorganized participation. Perhaps a bit surprisingly, reliability analysis shows that these indexes are clearly inter-reliable (Cronbach's a 00.859). Hence, it is difficult, and hardly fruitful, to explore them separately.
In separate reliability analyses, measuring Cronbach's alpha for each 'pair' of variables the index measuring the number of participation arenas was the greatest contributor to the internal consistency. Hence, this index was renamed social participation, and employed in the further analyses.
Physical violence is measured in two ways in YiN, both very explicit in nature. First, four variables measure violence by severity (from threats of violence to violence causing injury that requires medical treatment). Second, nine questions measure violence by perpetrator. All these variables were combined into a severity index through a two-step procedure: (1) If respondents had experienced several types of violence (in terms of severity), they were given the value that corresponds to their most severe experience. (2) This violence index also entails the category unknown severity, which includes violence towards respondents who reported violence perpetrators without reporting the severity of the violence. Similarly, it also includes several categories of violence with unknown perpetrator, although these have an n of 0.
Bullying is measured by an explicit question in YiN: 'Are you bullied at school (or on the school route)?'. The respondents are asked to categorize the bullying by frequency, with the categories ranging from less than every week to every (or almost every) day.

Method
Thomas Á as well as many other scholars researching psycho-emotional disablism Á seems to employ causal models. She explicitly conceptualizes psycho-emotional disablism in terms of 'enactment' and 'effects' (Thomas 2007, 152). With regard to violence against disabled people, Dan Goodley and Katherine Runswick-Cole similarly argue that 'disabled people experience violence because of contemporary society's deeply held contradictory discourses about dis /ability' (Goodley and Runswick-Cole 2011, 3, italics added).
While this calls for causal modelling, the statistical analyses presented in this article are purely descriptive. At least since Hume, it has been clear that causality can never be modelled without 'an object precedent and contiguous to another' (Hume 1975, 170). Empirical investigations of causality require diachronic data. The YiN consists solely of synchronic data, and cannot meet that requirement.
Hence, this study consists mainly of bivariate descriptive analyses. However, linear regression analyses are also carried out, not to raise issues of causality, but to strengthen the descriptive analyses. The methodological design can be visualized as a four-stage trajectory ( Figure 1).
The overall purpose of this trajectory is to ensure that the discussion is based on robust findings. First, the second stage (chi-square tests) limits this basis to findings that are statistically significant at 0.1% level. Second, linear regressions are carried out, in order to locate any important background variables. Third, any such findings are used to refine the analysis. In this final stage, 'original' categorical variables are replaced by 'intersectional' ones, in order to give broader descriptions.
In short, the design is a fairly direct reflex of Thomas' perspective. Her postulates Á in particular, the postulate that different disabling processes correspond to each other Á are confronted with quantitative analyses which looks specifically for bivariate correspondences.

Barriers to being: impairment and psycho-emotional well-being
A vast array of research with adolescents with impairments (Singh and Ghai 2009), including clinical research (Hogan, McLellan, and Bauman 2000;Jemtå et al. 2005), gives reason to expect associations between impairment and low psycho-emotional well-being. The data from YiN confirming this hypothesis is confirmed in many ways. In particular, disability is strongly associated with distress and emotional pain, such as loneliness (Table 1).
However, being is not primarily about specific emotions (such as loneliness) but rather about perceptions of one's own self. To explore this, impairment status is cross tabulated with the two SPPA indexes ( Table 2). Both of these cross tabulations show a broad negative association of low psycho-emotional well-being with impairment.
The SPPA is grounded in 'an increasing consensus that self-evaluation is domainspecific' (Wichstraum 1995, 100Á101). Hence, two clarifying research questions are necessary: (1) which of these domains is most closely related to disability? and (2)  which of these domains are most important to the negative self-perceptions of the disabled respondents? With regard to the first question Á which measures the degree of disproportion between the disabled and the non-disabled respondents Á the effects of psychoemotional disablism seem broad and pervasive: Out of the five domains, close friendship is the only domain without a statistically significant relation to disability. While the association with social attractivity is weaker than the others (chi-square value 32, 2), the other chi-square values are almost identical (around 60). If we look at the SPPA variables separately, three of them stand out: I don't like my life, I have trouble finding the right answers in school, and I am disappointed in myself (chi-square values of 93, 87 and 78, respectively).
With regard to the second question, the relations between disability and each domain were analysed separately. In these analyses, perceptions of one's own body is the domain most closely associated with impairment (chi-square value 064.5). In particular, the variable I wish my body was different stands out, as 3% (about 61) of the impaired respondents report this self-perception. Interestingly, this differs from negative self-perceptions among the non-impaired respondents, where the domains of social attractivity and global self-perception stand out.
The importance of the body in 'barriers to being' is also confirmed by another issue. First, impairment is strongly associated with self-inflicted harm; not only does the chi-square test show statistical significance, it also produces a chi-square value as high as 163, 2. Furthermore, this association seems robust and somewhat independent of psycho-emotional well-being in general: Even when the analysis is restricted to respondents with low psycho-emotional well-being, or respondents who experience a high degree of distress or loneliness, this association remains statistically significant.
Before discussing these descriptive findings, it is important to reduce the risk of type I errors. Hence, an additional linear regression analysis was performed. This analysis includes three background variables: gender, immigrant background, and sexual orientation. These variables are all focal points of the rising discourse on 'intersectionality' in disability studies (McRuer 2006;Goodley 2012). Of these three, only gender seems to contribute substantially to R 2 (Table 3). Hence, the cross tabulation was redone, replacing the impairment variable with an intersectional impairment/gender variable (Table 4).
Although gender is clearly important to psycho-emotional well-being, the association between low well-being and impairment remains. Furthermore, the association between gender and psycho-emotional well-being becomes statistically insignificant if the sample is restricted to impaired respondents. Although the process underlying 'barriers to being' (what Thomas calls psycho-emotional disablism) is clearly gendered, its focal point is disability rather than intersectional positions or categories.
In addition to the importance of gender, the linear regression also suggested that social class is important. There is also a statistically significant association, in the sample as a whole, between psycho-emotional well-being and social class. Hence, the cross tabulation was also redone with an intersectional variable regarding disability and social class (Table 5).
This cross tabulation suggests that social class is important in at least three senses: (1) Disabling processes (in Norway) takes place in a society where low-class status is independently associated with low psycho-emotional well-being. (2) The negative association between disability and low psycho-emotional well-being seems to pervade the class structure: The association between disability and low psychoemotional well-being is significantly within all five groups along the 'original' social class variable. (3) Low-class status seems to enhance disabling processes: The association is far stronger within that sub-sample.
Nevertheless, the findings most of all demonstrate the pervasiveness of 'barriers to being'. Although this process clearly seems to be gendered and class-sensitive, this disabling process cuts across these axes. Hence, these findings support Carol Thomas' emphasis on social structure.

Impairment, psycho-emotional well-being, and social participation
In Thomas's account, psycho-emotional disablism is structurally intertwined with 'barriers to doing': There is an interacting and compounding relationship between psycho-emotional disablism and the imposition of restrictions on activity in employment, education and other social arenas. (Thomas 2007, 72) While 'doing' can carry very different connotations among disability scholars, Thomas's work is explicitly a 'reformulation of the pioneering UPIAS definition of disability' (Thomas 2007, 73), where disability is defined as 'something imposed on top of our impairments by the way we are unnecessarily isolated and excluded from full participation in society' (UPIAS 1976, 20, emphasis added). Hence, one should expect a strong association between psycho-emotional wellbeing and social participation, at least among the disabled respondents. However, a simple cross tabulation of social participation and disability variable does not confirm that expectation (Table 6).
As this finding could be a type II error, social participation was added to the linear regression analysis above (as predictor), while the less important background variables were removed (Table 7). This analysis suggests that social participation is important to psycho-emotional well-being. Hence, social participation and disability was cross tabulated once more, this time restricting the sample to respondents with low psycho-emotional well-being (Table 8).
In the YiN, there is clearly a statistically significant association between psychoemotional well-being and social participation. This association even has a chi-square value as high as 155, far stronger than many associations discussed in this article. Nevertheless, this interplay between psycho-emotional well-being Á which is far from surprising Á sheds very little light on the specific and disproportionate association between impairment and low psycho-emotional well-being. While 'doing' and 'being' is generally intertwined, the barriers do not seem to converge. Hence, Thomas' key postulate is not confirmed.
This lack of confirmation does not mean, however, that there is no interplay between different forms of barriers in the respondents' lives. It is not unlikely that adolescents with impairments have greater experience with 'barriers to doing', even if they may not be too present in their current daily life. Furthermore, the 'doing' indicator (the social participation index) could also be flawed. Finally, it is also very likely that 'barriers to doing', even if overcome, can be hurtful and exhausting 'barriers to being' in themselves.

Impairment, bullying, and violence
Thomas' analysis of 'barriers to being' does more than link this process to 'barriers to doing'. Relying on Iris Marion Young, she also places this process in a 'schema' of 'five faces of oppression' which she claims can encompass 'every facet of disablism' that she discusses (Thomas 2007, 74Á75): exploitation, marginalization, powerlessness, cultural imperialism and violence. The analyses above suggested that the body Á and, in particular, the harming of one's own body Á is important in 'barriers to being'. Hence, it is useful to include violence and aggression in our analysis. This is also supported by recent developments in disability studies, where one increasingly emphasizes violence and aggression. In a recent systematic literature review , it is demonstrated that violence against disabled people are more prevalent than violence against non-disabled people. Similar reviews have also found strong associations between disability and bullying across different impairment groups (Horner-Johnson and Drum 2006; Shikako- Thomas et al. 2008, 164Á165).
Hence, violence and bullying were included in the linear regression (Table 9). Since the coefficients demonstrate clearly that both contribute significantly to R 2 , violence and bullying are cross tabulated with the intersectional impairment/gender variable in two final descriptive analyses (Tables 10 and 11).  These cross tabulations show strong and disturbing associations with impairment. These become even more disturbing if we restrict our focus to severe violence and bullying on a weekly or daily basis. Among both boys and girls, the disabled respondents are approximately twice as likely as non-disabled are to experience severe violence, and approximately three times as likely to be bullied at least once a week.  One could perhaps imagine that the patterns of increased violence prevalence corresponded to 'barriers to doing', hence re-confirming the convergence postulated by Thomas. The descriptive analyses Á of disability on the one hand and bullying and violence on the other Á have therefore been redone with four sub-samples separated by degree of social participation.
Interestingly, the associations did not prove statistically significant in all four subsamples. When recoded into dichotomous variables by severity, however, the association with bullying proved statistically significant in all four sub-samples. With the exception of the sub-sample with very low social participation, so did the association between disability and violence.
Hence, social participation does shape the disabling processes, just as they are gendered and linked to social class. Nevertheless, both psycho-emotional disablism and different forms of aggression Á in this study: bullying and violence Á seem to cut across those dimensions. Hence, the interpretation of 'barriers to being' should perhaps put less emphasis on 'barriers to doing', and more on the doing of others.

Close relations: the ambiguous importance of friendship
The domain-specific SPPA analyses gave at least one finding that may be surprising: While four out of five domains were related to impairment through statistically significant associations, self-perception with regard to close friendship were not (Table 12).
One can wonder: Are the disabling mechanisms sometimes overcome, contrasted or at least mitigated by relational intimacy? To answer this question, the previous analyses were redone with a certain sub-sample: the quartile that scored highest on an SPPA index regarding close friendship. Furthermore, the general psychoemotional well-being index was replaced with an index pertaining to all domains except close friendship. Vis-a-vis the previous findings, those additional findings differ from the previous in three ways: (1) the association between impairment and psycho-emotional wellbeing is barely statistically significant, and the relations to the parallel 'average reply' index are not.
(2) The association between impairment and violence loses its statistical significance among the boys, but the association with bullying does not. (3) While the association between bullying loses its significance among the girls, the association with violence is even stronger than in the sample as a whole.
On the one hand, these additional subgroup analyses demonstrate that close friendship relations seem to engender resilience. On the other hand, this mechanism is limited. Furthermore, impairment is associated with violence (in the sample as a whole) even if we look specifically at violence from friends. While this last association is not statistically significant at the chosen level of significance, the link between disability and violence still seems to cut across the dimension of close friendship.

Discussion: from compounded to diversified disabling processes
This study has made five key empirical findings: (1) impairment is associated with low psycho-emotional well-being, (2) there is no significant association between impairment and social participation and (3) variances in social participation cannot shed light on the association between psycho-emotional well-being and impairment. Emphasizing the afflicting implied in Thomas's description of psycho-emotional disablism, it was demonstrated (4) that impairment is significantly associated with violence and bullying.
Finding (1) confirms Thomas's analysis of barriers, and finding (4) is at least very compatible with her perspectives. Findings (2) and (3), on the other hand, challenge her convergence postulates. This demonstrates a need Á regardless if one employs structural explanations of disabling processes or not Á to take the diversified forms of these processes into account.
For this purpose, Thomas' understanding of disabling processes seems to overemphasize convergence. On one level, this is surprising, as her definition of disabling processes as 'a form of social oppression' does open up for understandings that take lack of convergence into account. The indefinite article 'a' Á as well as the genus 'form of social oppression' Á implies that there are other forms of social oppression that also can shape the lives of people with impairments. She also emphasizes relationships (or intersections) between different forms of oppression: 'the forms and impacts of disablism are invariably refracted in some way through the prism of gender locations and gender relations: being disabled is a gendered experience' (Thomas 2007, 70). While Thomas acknowledges that disabling processes are somewhat separated empirically, her adjectival use of past tense transitive verbs such as 'refracted' and 'gendered' suggests that these processes are in some way 'basically' or 'originally' homogenous. However, these transitive expressions also suggest that the causes of 'transitions' Á in this case: 'gender locations and gender relations' Á are separate from the processes they refract. Disability (and/or disablism) remains conceptualized as a fundamentally singular social phenomenon, marked by a definite grammatical article: 'the relationship of ascendancy of the non-impaired over the impaired' (Thomas 1999, 40, my emphasis).
In contrast, the YiN findings arguably give rise to a suggestion which is already present (but nascent) in Thomas' work: To the extent that disabling processes are ontologically intertwined, they are no more intertwined than any other social process. This has at least one important implication: If it is justified to describe modern societies (and their processes) as differentiated, then disabling processes should also be described as differentiated.
The question on how this description can be carried out must be discussed elsewhere. Similarly, the theoretical notion of modern societies as 'diversified' or 'differentiated' Á which is conceptualized in many different ways in both social and cultural research Á must be explicated and delineated elsewhere. The argument of this empirical study is only that the findings from YiN, alongside other findings, suggest a conceptual move in that direction.

Possible limitations
While the findings in this study poses challenges to the work of Thomas and others, these challenges are (of course) only provisional. The lives of disabled adolescents are socially ordered or constituted, even if the socially constituted result is not as orderly. Furthermore, this empirical investigation (using synchronic quantitative data) cannot shed light on individual trajectories. Different forms of barriers can, while not occurring simultaneously, nevertheless be interwoven within a life-course, and hence contribute to the 'same' 'barriers to being'.
It is clearly possible to interpret this study's findings of lacking convergence within an unmodified structure-emphasizing perspective. On the one hand, the findings may be unrepresentative. One could argue, for instance, that the pervasiveness of 'barriers to doing' is well documented through systematic literature reviews (Shikako- Thomas et al. 2008;Bult et al. 2011), as is the pervasiveness of 'barriers to being' (Shields et al. 2006). It could be argued that the lacking convergence in YiN Á although it is interesting Á poses a very limited challenge to perspectives that are backed by much larger bodies of research.
On the other hand, the findings could lead to the conceptualisation of a deep structural level. Referring to the YiN findings on the role of social class, one might argue that psycho-emotional disablism, alongside other forms of disabling processes, is embedded in the deep structure we call industrial capitalism. This possibility is, perhaps, already implied in Thomas' 'materialist feminist orientation' (Thomas 2007, 50), but also in the work of non-UK scholars such as Lennard Davis. His groundbreaking book Enforcing Normalcy (1995) centres on the historiographical claim that 'we have to see disability as intricately linked to capitalism and imperialism' (Davis 1995, 85). Furthermore, the notion of a deep structural level can also be conceptualized in less materialist ways, leaning, for instance, Bill Hughes' argument that disabling processes belong to the 'generic construction/treatment of the stranger in modernity' (Hughes 1999, 157).
However, these interpretations would render the YiN findings almost analytically unproductive. In the first case, the analytical postulates would be neither substantially weakened nor modified. In the second case, the postulates would be modified, but only to the benefit of a new postulate with exactly the same implications: To replace a notion of convergence (process A is convergent with process B) with a notion of fundamental coherence (process A and process B are constituted and maintained by structure C) has very limited bearing on research conclusion. Given the empirical robustness of the findings, both interpretations seem clearly unreasonable.
More importantly, these choices would be performative contradictions. It is little doubt that the theoretical paradigms of disability studies claim to be informed by empirical research. Thomas, for instance, underlines that 'my data source is empirical research on disabled women's life experiences' (Thomas 2001, 245). She has also argued that the UK disability studies are characterized by 'empirically based ideas' (Thomas 2008, 15). Hence, her work suggests that new empirical findings should form the basis of new theoretical developments.

Conclusion: needs for further research
Although the limitations call for cautious interpretations of the findings, the finding most of all call for further empirical research. First, they demonstrate the need for life-course research. The crucial find in this article Á the lack of synchronic convergence between different forms of barriers Á does not rule out the possibility of diachronic convergence. While that possibility cannot be investigated in datasets such as YiN, it would be very fruitful to collect longitudinal data which would make that possible.
Second, the findings suggest an increased emphasis on the study of disabling actions. In the WHO systematic reviews Á of research on adults with disabilities and children/adolescents with disabilities, respectively Jones et al. 2012) Á risk factors is a critical concern. However, both discussions and metaanalyses emphasize risk factors in victims. Interestingly, risk factors in perpetrators are only discussed with reference to perpetrators who are themselves disabled. This tendency is contrary to general research on violence, which has generated a large body of research on risk factors for adolescent perpetration of violence.
Third, it is perhaps time to look at priorities in research design. While disability studies in the last decades have tried to design research, which could investigate the severity and pervasiveness of disabling processes Á for very good reasons Á it is perhaps now more pertinent to design specific research on specific aspects (such as violence). This may have huge costs: for instance, qualitative research on embodied experience would be unable to describe the complexities of the participants' lifeworlds, while quantitative research on barriers would be less useful as grounds for immediate policy action. Nevertheless, this need for modified priorities seems unavoidable.