“TREAT THEM AS A HUMAN BEING”: DIGNITY IN POLICE DETENTION AND ITS IMPLICATIONS FOR ‘GOOD’ POLICE CUSTODY

Here, we examine the factors influencing whether those detained by the police feel treated with dignity. We develop a human rights-oriented conception of dignity rooted in the equal worth of human beings, encapsulated in detainees’ desire to be ‘treated like a human being’. The predictors of this are examined using multilevel modelling of survey data collected from 371 detainees in 27 custody facilities in 13 police forces in England and Wales in an Economic and Social Research Council-funded study of ‘good’ police custody. We found that perceptions of the material conditions predicted feelings of dignity, as did detainees’ reactions to being detained, their perceptions of the culture of police custody and the mechanisms used to hold the police to account. Feelings of dignity were also less likely for younger adults and for Black Asian and Minority Ethnic detainees, with these experiences being mediated by less trust in accountability mechanisms. This paper concludes by examining the implications for ‘good’ police custody.


Introduction
Police detention-where those arrested by the police are taken whilst an investigation is mounted and a decision reached about what to do next with their case-presents a number of possible indignities. For example, in police custody, inspections conducted by Her Majesty's Inspectorate of Constabularies, Fire and Rescue (HMICFRS) in 2016 and 2017 (see below), their main concern with regards detainee dignity was about the overuse of rip-proof paper suits, particularly when force was also used to remove suspects' clothing and when suspects were also left naked or partially clothed in their cells sometimes for hours at a time. This illustrates the indignities of police custody, but also the perils, in terms of psychological harm, humiliation and the failure to acknowledge suspects as citizens and as fellow human beings. These harms have the potential to stack up within and across encounters with the police, as well as eroding recognition of citizens as social and political equals (Dupré 2015;Fagan 2017: 21).
The purpose of this paper is, first, to examine how dignity can be understood in police custody. To do this, we consider the iterative process by which dignity was conceptualized in a recent mixed-method study of police custody in England and Wales. This involved both inductive and deductive forms of reasoning. Using multilevel modelling of survey data, the second aim of this paper is to explore which factors predicted whether detainees regarded their treatment as dignified. We focus on this because it has implications for 'good' police custody, the conceptualization of which is the third aim of the paper.

Debates about dignity
Debates about dignity are increasingly prominent in a number of academic spheres (McCrudden 2013). Here, we focus on law and criminology, in particular, debates in two largely separate bodies of literature, first, the police and prison legitimacy literature, in which police custody has received limited consideration, and, second, human rights-oriented discussions.
Within the police legitimacy literature, dignity has been regarded as one aspect of procedural justice-alongside citizen participation, neutral decision-making and trust in the motives of the police-which is seen as a critical antecedent of police legitimacy and cooperation by citizens with legal authorities (Mazerolle et al. 2013;Jackson et al. 2013;Jonathan-Zamir et al. 2015: 10-11). However, much of this literature relates to police-citizen encounters on the street, rather than in circumstances where citizens' liberty has been formally taken away, as is the case in police custody. Moreover, within the police legitimacy literature, dignity has tended to be seen as a route to legitimacy and citizens' cooperation with the police rather than an end itself, which is something the present paper seeks to explore. Furthermore, in these studies, dignity is rarely defined in detail. Instead, reference is made to respectful behaviour and politeness (Mastrofski et al. 1996;McCluskey et al. 1999;Sunshine and Tyler 2003;Reisig et al. 2004;Jackson et al. 2013;Mazerolle et al. 2013;Jonathan-Zamir et al. 2015;Meares et al. 2015: 10) or to limited use of force (Reisig et al. 2004;Dai et al. 2011;Jonathan-Zamir et al. 2015). However, both of these sets of indicators are broad and open to interpretation. Similar issues arise in the prisons literature. Whilst important, e.g. because it is understandable to prisoners (Crewe 2009;Liebling 2011: 182), dignity is rarely defined and is considered hard to operationalize. For example, Liebling defines dignity in terms of autonomy but also notes the difficulties of putting this concept into practice, where e.g. 'whole life sentences sweep away … principles of self-determination' (Liebling 2017: 21). Due to these limitations with the conceptualization and operationalization of dignity and its application to police custody, we turned to the human rights-oriented literature, in which dignity is seen as intrinsic to human rights (Talbott, 2014).
Here, a growing amount of attention has been paid to the meaning and origins of dignity, including in places of detention (Mavronicola 2015;2016;Finer and Wadham 2019). In terms of its origins, Dupré (2015: 24) notes how dignity began as 'an ideal shaped by hopes and fears … before being crystallised as a legal concept in an attempt to prevent mass extermination of human beings from recurring'. This culminated in the UN Universal Declaration on Human Rights (Dupré 2015: 80). Along with other great conventions on human rights, e.g. the European Convention on Human Rights (ECHR), Simon (2017) argues that such developments marked the first wave of dignity for criminal justice institutions. In this period, dignity was understood through notions of liberty (linked to Kantian ideas of autonomy and self-determination) and equality (linked to the intrinsic, equal and immutable worth of all human beings; Henry 2011; Simon 2017). As part of a second-wave of dignity, he also notes other ways of understanding dignity. Drawing on Henry (2011), he sees it as linked to public decency (which is concerned with what we might expect in a civilized society in terms of the collective respect shown for human beings). These historical perspectives show dignity to be defined through notions of autonomy, the equal worth of human beings and public decency.
In judgements by the European Court of Human Rights (ECtHR), linked largely to Article 3, 1 however, the focus has been on failures to maintain dignity as revealed by the practices of actors in prison and police custody. These practices fit into three broad themes: excessive coercion and threatened coercion, poor material conditions and inadequate access to health or mental health care. 2 Of the 28 available and relevant cases considered in the content analysis used to identify these themes, 3 Bouyid v Belgium is particularly important, being seen as a 'watershed' (Finer and Wadham 2019: 160) and a 'key moment' in the development of dignity-related case law (Mavronicola 2016: 15). The case revolved around two suspects being slapped once in the face whilst in police custody. The Grand Chamber concluded that as the suspects' liberty was taken away, which made them situationally vulnerable, and owing also to additional imported vulnerabilities (one suspect was a child), and to the suspects' behaviour not justifying the slaps, the use of force inherently infringed human dignity and amounted to degrading treatment under Article 3. It was also noted that, since 1973, there has been a 'particularly strong link between the concepts of "degrading" treatment or punishment within the meaning of Article 3 of the Convention, and respect for "dignity"' (para 91). As a result, they concluded that 'any conduct by law enforcement officers vis-à-vis an individual which diminishes human dignity constitutes a violation of Article 3 of the Convention' (para 101).
Given the importance attached to dignity in this case and the potential for indignity to result in a violation of Article 3, dignity is mentioned surprisingly little in other relevant European and national level standards, strategies and policies, 4 with the exception of the National Preventative Mechanism (NPM). In England and Wales, the NPM is a multi-organizational team that includes HMICFRS, who are required to regularly inspect police custody facilities. In inspections conducted around the time that data from the present study were collected in 2016 and 2017, 5 indignity was operationalized in two main ways. First, it was operationalized in relation to the overuse of rip-proof paper suits, as described above (mentioned in 7 of the 14 reports). Second, it was operationalized in relation to privacy concerns about CCTV cameras, in which detainees could be seen on CCTV monitors whilst using in-cell toilets or during strip searches, which meant that they may have been naked/partially clothed (mentioned in 5 of the 17 reports).
Dignity is, therefore, conceptualized in the police legitimacy literature as respectful behaviour and limiting the use of force. Historically and in human rights scholarship, 1 This concerns the right not to be subjected to torture, inhuman or degrading treatment or punishment. 2 A table of cases that fall under each theme is available on request. Bouyid v Belgium illustrates the excessive coercion theme. 3 We searched the HUDOC online database, using the terms 'dignity and detention', then selecting 'Key Cases' and Article 3 judgements. This search yielded 67 cases of which 9 were duplications, 8 were not available in English and 22 were not relevant (e.g. they related to sentencing). This left 28 cases, with judgements from these cases being issued between 1999 and 2016.
4 See e.g. the standards used by the European Committee for the Prevention of Torture and Inhuman or Degrading Treatment or Punishment (CPT), the 2017 National Police Custody Strategy and the 2019 PACE Codes of Practice Code C. 5 There were 17 inspections conducted in 2016 and 2017 (though three of these inspections were published in 2018). A 'dignity' key word search showed that it was discussed in all bar three reports. In most cases, it was discussed in relation to indignities not dignified treatment.

DIGNITY IN POLICE DETENTION AND ITS IMPLICATIONS FOR 'GOOD' POLICE CUSTODY
1669 dignity has been conceptualized as autonomy, the equal worth of human beings and public decency. In more current debates, European jurisprudence and NPM inspections, the focus has been on limiting: excessive coercion and threatened coercion, poor material conditions, inadequate access to health/mental health care and inadequate levels of privacy. In subsequent sections of this paper, we examine the applicability of these understandings of dignity to quantitative data collected from detainees in police custody in England and Wales and to notions of 'good' police custody.

Good police work and good police custody
In Skinns et al. (2015), drawing on notions of 'good' police work, we examined the ways in which police custody might be made 'good'. First, it was concluded that, like police work, police custody might be made 'good' because of its capacity for building strong links with the community, in part because police custody areas are local institutions detaining primarily local populations but also because of their capacity for nurturing civic values. Second, ensuring an equitable distribution of the goods and impositions of police work is also important to 'good' police custody, due to the historically poor relationships between the police and Black Asian and Minority Ethnic (BAME) citizens in England and Wales, in which they have felt 'over-policed and under-protected' (Patel and Tyrer 2011;Phillips and Bowling 2017: 62;Macpherson 1999: 45.7). In this context, Skinns et al. (2015) concluded that the recipients of the impositions of police custody were, therefore, disproportionately young minority ethnic men (See also Newburn et al., 2004). Third, as with other forms of police work, it was concluded that whilst legitimacy is useful to understandings of 'good' police custody, legitimacy may also be harder to achieve than for citizens on the street, given the more problematic relationship that suspects typically have with the police (Choongh 1997;Skinns et al. 2015: 40-41).
Five years on, we return to these dimensions of 'good' police custody and, based on a more robust set of data, we offer firmer conclusions about how it may be conceptualized, examining in particular whether it can be harnessed to notions of dignity. The contribution of this paper is, therefore, three-fold. First, we explore how dignity-a term little explored in police custody-can be conceptualized. By linking it to human rights-oriented understandings of dignity, we also provide a useful addition to discussions about police legitimacy. Second, by arguing that the dignity of citizens, including those in police custody, should be seen as an end in itself rather than a route to legitimacy and cooperation, it alters how we understand police-citizen interactions and their importance to 'good' police work. Third, we employ a novel multilevel data analysis technique for looking at mediation effects, which is little used in criminology, thereby providing scope for it to be employed in other criminological research. This mediation analysis was undertaken to further examine unexpected findings in the multilevel analysis.

Data collection
This paper is based on data collected during an ESRC-funded study entitled '"Good" police custody: theorizing the "is" and the "ought"' (hereafter referred to as the GPCS).
In Phase 1, in 2014, survey data were collected from custody managers in 40 of the 43 police forces in England and Wales. The survey focused on who works in, manages and owns police custody blocks and on size, busyness and fitness for purpose. In Phase 2 in 2014-15, qualitative data were collected through observation and interviews in four custody blocks in four police forces. In total, the research team spent 532 hours observing and conducted 97 interviews (47 with staff and 50 with detainees). The Phase 2 data were analysed thematically, picking out broad themes and sub-themes. Based on this analysis, two separate closed-question questionnaires were constructed for use with detainees (and staff) in Phase 3 of the research in 2016-17. 6 After piloting in one police force area, the detainee questionnaire, which is the focus of this paper, was used to collect data on perceptions of: (1) fairness and authority (44 items); (2) the conditions of custody (6 items); (3) reactions to being detained in police custody (8 items); (4) governance and accountability (6 items); (5) occupational culture (8 items); (6) distinctions between different types of staff (4 items); 7 (7) the purposes of police custody (3 items); (8) and risk (1 item). All of these survey items entailed Likert scale responses of strongly agree, agree, neither agree nor disagree, disagree or strongly disagree. Demographic data, as well as information about how long people had been detained and what they had been arrested for were also collected.
The questionnaires were completed by 371 detainees within 27 custody facilities in 13 police forces in England and Wales whilst they were detained in police custody. This involved a member of the research team spending approximately a week in each custody facility waiting for opportunities to invite detainees to participate in the research. This is an important set of data as studies examining the views of suspects are rare and, as far as we are aware, a large-scale criminological study collecting quantitative data from suspects in multiple sites has never been done before. 8 Forces were selected for reasons relevant to the research, such as whether they were predominantly public or private or a hybrid of the two. Within each force, we focused on the two busiest custody facilities (identified through data collected in Phase 1). As random sampling is not possible in police custody due to the nature of the suspect population and the busyness of the custody environment, the researchers aimed to invite all suitable respondents to participate in the research. This necessarily involved researcher-and police-led selection criteria to accommodate ethical and risk-based considerations. Only detainees aged 18 years and above, who had been in custody for a minimum of four hours were deemed suitable. In addition, anyone who staff considered vulnerable and thus likely to be harmed by the research or anyone who was seen as posing too great a risk to the researchers were also excluded from the research. 9 The self-selecting way in which research participants were recruited, when combined 6 The study was an 'exploratory sequential design' (Creswell and Plano Clark 2018: 84). Qualitative research is undertaken first, with findings from it being used to develop the quantitative research that follows. The quantitative research is 'grounded' in the perspectives of participants, making it more robust (Creswell and Plano Clark 2018: 84). 7 These questions focused on the discerned distinction in the roles of civilian detention officers compared to warranted police officers. Given their limited relevance to dignity, data collected on this particular theme were excluded from the analysis presented here. 8 The last large-scale study conducted in multiple police custody facilities and forces was in the late 1990s but surveyed staff not detainees (Bucke and Brown 1997). 9 Excluding vulnerable and/or agitated participants was unavoidable, in spite of the potential effect this had on the underestimation of detainee indignity, given that such feelings are likely to be more prevalent amongst these groups. To counter balance this, the research team were alert to changes in a detainees' demeanour. For example, if a detainee was initially deemed unsuitable by the police, the research team sometimes waited until they calmed down before inviting them to participate in the study.
with the researcher-and police-led selection criteria, means that the findings may not be considered generalizable. They are nonetheless an important starting point for understanding detainee experiences given the dearth of other similar studies. By and large, participants filled-in the anonymous questionnaire by themselves on tablets (using IdSurvey software) and sometimes on paper. Resource constraints meant that the team did not reach the intended target of collecting data from 1,080 detainees (30-40 per custody facility). 10 On average, the percentage of detainees completing the survey as a proportion of those invited to participate in the research was 64 per cent (ranging from 38 per cent in some custody blocks to 77 per cent in others) compared to 93 per cent for staff.

Data analysis
First, we conducted separate principal axis factor analyses, one for each of the relevant themes described above, which each represented separate sections in the questionnaire (i.e. one factor analysis for fairness and authority and another for custody conditions). Perceptions of the purposes of police custody and of risk included insufficient items for factor analysis. These were included as individual items in subsequent analysis. The purpose of factor analysis is to identify clusters of questions and the underlying factors responsible for this covariation among survey responses. We, therefore, used it to understand the structure of responses within each theme in the questionnaire and to reduce the data to a more manageable size for use in subsequent analysis. 11 The factor analysis was conducted in SPSS. In each factor analysis, we extracted factors with eigenvalues above one. Where the analysis yielded more than one factor, Oblimin rotation was used as we found correlations between the factors within each theme. Factor scores were produced by the regression method, resulting in a mean of zero. In relation to missing data, listwise deletion was used. This meant that each factor analysis was run only on cases with a full set of data. This resulted in a loss of only 10 cases from the outcome measure, and fewer (between 0 and 3) from the factor analyses that provided the predictor variables.
The outcome measure for this paper is dignity linked to feelings of equal worth, henceforth referred to as equal worth. This was the first extracted factor in the factor analysis of the 44 items for the fairness and authority theme and accounted for 32 per cent of the variance; details of this factor are provided in Table 1. The factor analysis of the fairness and authority theme yielded an additional seven factors with eigenvalues above one. These and the amount of variance they accounted for were: obligation to obey (7 per cent); humanity of detention officers compared to police officers (3 per cent); routinization of treatment and processes (3 per cent); trust in staff (3 per cent); keeping detainees informed (3 per cent); importance of outcomes to detainees (3 per cent) and compliance based on inducement (3 per cent). These further seven factors are not the subject of this paper and are not included in Table 1. They were not used as predictors in subsequent modelling because the degree to which factors are allowed 10 See Greene and Skinns (2018) for a discussion of the complexities of the data collection process. 11 We conducted exploratory rather than confirmatory factor analyses within the individual themes because, although the themes to be included in the survey were decided in advance, on the basis of the findings from Phase 2, we did not know in advance how many latent constructs there would be within each theme.
to correlate within Oblimin rotation is determined by the value of delta, by default set to 0 in SPSS, which ensures that high correlation between factors is not allowed (Field 2018: 794). Hence, the degree of relationship between factors from the same principal axis factor analysis is somewhat artificial. Factors derived from the separate factor analyses of the other themes in the survey, as set out above, were used as predictors in the subsequent multilevel modelling and details of these are provided in Table 1.
The second stage of the analysis involved multilevel modelling to examine the extent to which equal worth was predicted by factors from the separate factor analyses of the other themes in the survey, together with demographic and other individual-level variables, as well as custody suite level variables from the custody manager's survey conducted in Phase 2. Multilevel modelling was used because of the data's hierarchical structure. Clustered data violates the assumption of independent observation necessary for ordinary single-level regression analysis. Consequently standard errors (SEs) are underestimated if ordinary regression analysis is used with clustered data (Rasbash et al. 2016: 3). This leads to biased significance tests and increased likelihood of concluding that there is a significant relationship when there is not. Multilevel models are equipped to deal with clustered data. Multilevel modelling was performed using MLwiN version 3 (Charlton et al. 2019) with detainees at Level 1 and custody facilities at Level 2.
Third, we used mediation and moderation analysis within a multilevel model. A moderator variable affects the direction and/or strength of the relationship between a predictor variable and an outcome measure (Baron andKenny 1986: 1174). In other words, the relationship between the predictor and the outcome measure changes as the value of the moderator changes. Moderation is indicated by a significant interaction between the predictor and the moderator in relation to the outcome measure.
As for mediation, a variable is considered 'to function as a mediator to the extent that it accounts for the relation between the predictor and the criterion … mediators speak to how or why such effects occur' (Baron andKenny 1986: 1176). Mediation analysis is about the mechanism by which the predictor influences the criterion or outcome measure (Hayes 2018: 78). Mediators help to explain how the relationship works, i.e. 'the process or mechanism by which one variable affects another ' (MacKinnon et al. 2007: 594).
Estimation of mediation within multilevel models is more complex than in singlelevel models because of the data structure. With clustered data, scores within a cluster are not independent of one another. In this situation 'mediational analysis suffers from the same difficulty as other traditional analytic techniques' (Krull and MacKinnon 2001: 250). In this paper, indirect effects were tested using Rockwood's Multilevel Mediation (MLMed) macro in SPSS (Rockwood 2017;Rockwood and Hayes 2017). The macro performs all necessary data management, including computing group means for lower-level predictor variables and decomposes the indirect effect into within-group (individual-level) and between-group (group-level) effects. 12 MLMed also allows for three mediators to be considered simultaneously. Inclusion of multiple mediators in an integrated model allows for formal comparison of the indirect effects to determine which is stronger (Hayes 2018: 148-9). MLMed constructs Monte Carlo confidence intervals (CIs). Our models are all 1-1-1 models which means the predictor, mediator and outcome are all at Level 1. Restricted maximum likelihood estimation was used.  Mediation analysis was conducted to investigate some unexpected findings in the multivariate multilevel model regarding the relationships between the outcome measure equal worth and the predictor variables age, ethnic group and perceptions of autonomy. Bivariate multilevel models suggested that younger and BAME detainees were less likely to feel that they were treated with dignity, which is what we expected based on existing literature (see above). However, when other predictors were incorporated into the model, ethnic group and age were no longer significant predictors, suggesting mediation. In the bivariate analyses, as expected, autonomy had a highly significant positive relationship with equal worth. However, after the inclusion of other variables, the coefficient for autonomy became negative. This suggested that the positive effect of autonomy was being mediated through other variables (see the Results section below).
The moderation analysis was motivated by observations from the data that prompted an interest in exploring whether autonomy might moderate the impact of poor material conditions on feelings of equal worth, which would have important practical implications. Prior analysis, with autonomy as the outcome measure, indicated that, while autonomy was strongly related to material conditions, autonomy also depended on other factors, such as not seeing treatment as routinized. A scatterplot of conditions and autonomy (Figure 1) indicated that people may feel a lack of autonomy even in good conditions. At the same time, a scatterplot of conditions and equal worth ( Figure 2)  were strongly predictive of equal worth, where conditions were poor, a number of people rated equal worth quite highly, despite the poor conditions.

Concepts: dignity rooted in equal worth
In keeping with the mixed-methods approach of the study, data analysis and interpretation, e.g. in relation to key concepts such as dignity, were informed by inductive and deductive forms of reasoning. In Phase 2 of the GPCS, participants talked about the importance of staff recognizing suspects as fellow human beings, i.e. as no different from them, as innocent until proven otherwise and, therefore, worthy of their help and  respect. When asked about staff-detainee relationships, one custody officer said, e.g.
'a]ll I know is this bloke here, he is a human being and that's what we treat him like', whilst a detainee said, 'Well, [they] treat you like a human being'. We, therefore, asked questions based on this theme in the Phase 3 survey but without, at this point, connecting these ideas to notions of dignity, suggesting an inductive approach. This need to recognize detainees as fellow human beings was subsequently explored in more detail when factor analysing the Phase 3 data. At this point, an initial review of the literature indicated that these ideas accorded with human rights-oriented conceptualizations of dignity, as discussed above, rooted in feelings of equal worth, suggesting more of a deductive approach. This refers to the idea that human beings are imbued with dignity because they are human, irrespective of their status or rank. This is a universal and permanent feature of human beings. This type of dignity is also embedded in human relations, being guided by the idea of reciprocity, namely that 'all humans owe respect to, and deserve respect from, each other as beings of equal worth' (Henry 2011: 202-3).
The first extracted factor in the factor analysis of the fairness and authority theme, the outcome measure in this paper, seemed to encapsulate human rights-oriented definitions of dignity linked to equal worth. This factor encompassed 13 different survey items, 12 of which corresponded to detainees' need to be treated like a human being, equal in worth to anyone else (see Table 1). It included e.g. 'I am treated as a human being in this custody suite'.
Taking all of this together, we expected to find that: (1) Dignified treatment-potentially linked to feelings of equal worth-would be integral to 'good' police custody (as suggested by Phase 2).
(2) Material conditions may be significantly implicated in the detainee experience, including potentially to experiences of equal worth (as suggested by ECtHR judgements). (3) Younger and BAME detainees may be less likely to experience their treatment as dignified (as suggested by existing research).

Results From the Multilevel Analysis: What Predicted Dignified Treatment?
Before examining predictors of equal worth, we first established whether there were differences between custody facilities. We compared an empty (or null) two-level model in which the intercept was allowed to vary across facilities with a single-level model that constrained the intercept to be non-varying across facilities. A likelihood ratio test comparing the deviance of the two models indicated that there were significant differences between facilities in the extent to which detainees felt treated with equal worth (reduction in −2 × log likelihood = 5.149, p = 0.012 13 ). As there were significant differences between facilities, the effects of adding predictors were tested in a two-level random intercept model. 14 13 The difference has a chi-square distribution with 1 degree of freedom. Significance should be determined as a one-tailed test (Snijders and Bosker 2012: 98).
14 As there were only 13 police forces, this was considered an insufficient number to make inferences about differences between forces. However, in order to check whether the inclusion of police force as a level would make a difference to the significance of variables at lower levels and what was included in the models, we also repeated the procedures with three-level random intercept models. We found that the variables included in the final model and the indicated levels of significance provided in Tables 2 and 3 are the same for a two-level or a three-level model. We began by examining bivariate relationships between each predictor variable (see Table 2) and equal worth. The predictor variables were added to the two-level model individually, one at a time, to establish the strength of bivariate relationships between each of the predictors and the outcome measure. Significance of these fixed effects at this stage, and in the subsequent multivariate multilevel analysis, was tested by calculating the ratio of the slope estimate to the estimate of the SE. The ratio indicated a significant relationship at p ≤ 0.05 if the absolute value is 1.96 or greater (two tailed).
At the next stage, we included in a multivariate two-level random intercept model all variables which, in the above bivariate analyses, had a significance level of p < 0.25 (two tailed). Use of a lower p value in the screening process, such as 0.05 may fail to identify important variables (Hosmer and Lemeshow 2000: 95). We next successively dropped from the model any variable which had a p value greater than 0.05, one at a time, starting with the least significant predictor. We also included in the model an interaction term for conditions and autonomy (for the reasons set out above). Outcome measure is equal worth, high score indicates favourable attitude. In this and Table 3, all predictor variables from factor analyses scored so that high score indicates favourable attitude; hence, high score for conditions means conditions good, while high score for derision means staff do NOT treat with derision. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 (two tailed).
Variables were dropped until all the predictors left in the model had a p value of 0.05 or less or were involved in the interaction effect. We then tried adding to the model variables that had been dropped at an earlier stage or that had a p value of 0.25 or greater in the bivariate analyses. This was done in order to ascertain whether any of these variables were significantly related to the outcome measure, equal worth, when other variables were controlled for.
As a final stage, we investigated introducing a random coefficient for each of the variables in the model, one at a time. However, each did not improve the model fit, as assessed by likelihood ratio tests. This meant that the relationship between the outcome and each of the predictor variables is similar for different custody facilities. As the addition of random coefficients did not improve the fit of the model, we adopted the more parsimonious random intercept-only model. Details of the final model are presented in the final column of Table 3. Table 3 also provides details of a model with only ethnic group and age as predictors and models which incorporate the mediators of ethnic group and age, as described in the section on mediation effects below. It can be seen from the final model that, in order of significance, feelings of equal worth were more likely if detainees: • trusted in accountability mechanisms used to monitor police custody and address police complaints; • sensed a culture of camaraderie between detainees and staff (in which staff overtly shared a joke with detainees to help them cope better); • felt they 'had something' in terms of access to material goods that met their basic needs (e.g. reading and writing materials or limited personal possessions); • sensed a culture of decency rather than of derision (staff overtly laughing at detainees) and of suspicion (treating detainees as if they were liars); • perceived the material conditions more favourably (e.g. whether the custody block felt not underground and well maintained); • saw police custody as being mainly about looking after detainees' welfare and • were held in facilities that custody managers judged 'fit for purpose '. 15 Age, ethnicity and feelings of autonomy had a complicated relationship with equal worth. We consider these next.

Age and ethnic group
In bivariate multilevel models, as expected, age and ethnic group (white compared to BAME) significantly predicted perceptions of equal worth. Younger adults and people from BAME groups viewed their treatment as a person of equal worth less favourably than older people and those who were white. However, when other predictors were incorporated into the model, ethnic group and age were no longer significant predictors of perceived equal worth.
In relation to ethnic group, incorporating feelings of having nothing and insecurity, trust in accountability mechanisms and culture of derision into the model made ethnic group no longer a significant predictor of equal worth. Each of these variables was, in turn, significantly predicted by ethnic group. BAME respondents were significantly more likely to report feelings of having nothing and insecurity and that there was a culture of derision (and thus a lack of decency) and significantly less likely to trust in accountability mechanisms. These three variables all potentially mediated the relationship between ethnic group and equal worth. Analysis using Rockwood's MLMed macro indicated that there were significant within-group indirect effects of ethnic group on equal worth through trust in accountability mechanisms (b = −0.157, SE = 0.077, 95 per cent CI = [−0.011, −0.313]) and through feelings of having nothing and insecurity (b = −0.116, SE = 0.059, 95 per cent CI = [−0.002, −0.238]) but not through perceptions of a culture of derision/lack of decency. 16 Within a given custody area, BAME detainees had less trust in accountability mechanisms and were more likely to feel that they had 15 This finding should be treated with caution as there may be problems with the estimates of Level 2 variables given only 27 custody facilities were included in the analysis. With 30 groups, SEs are estimated as 15 per cent too small (Maas and Hox 2005: 90). This would make custody managers' judgement of facilities' fitness for purpose insignificant. 16 None of the between-group indirect effects were significant. This is to be expected; significant between-group indirect effects would imply that an increase in the proportion of BAME respondents within a facility would decrease facility-aggregated perceptions of dignity by way of facility-aggregated changes in the mediator. We would not expect individual-level demographic variables to have this effect. nothing and to feel insecure and this translated into less favourable views regarding their sense of equal worth.
A similar approach was taken with age. It showed that, within a given custody area, younger detainees had less trust in accountability mechanisms (b = 0.009, SE = 0.002, 95 per cent CI = [0.013, 0.004]), were less likely to view police custody as being mainly about looking after detainees' welfare (b = 0.004, SE = 0.001, 95 per cent CI = [0.006, 0.002]) and were more likely to feel that there was a culture of derision/lack of decency (b = 0.006, SE = 0.002, 95 per cent CI = [0.010,0.002]), each of which led to less favourable views of equal worth.

Autonomy
There was a significant interaction between material conditions and autonomy (see Table 3). The relationship between material conditions and equal worth was moderated by perceived autonomy. Although, on the whole, better judgements of material conditions were associated with more favourable views about equal worth, where autonomy was considered to be high, material conditions had a weaker relationship with equal worth than where autonomy was limited. Where autonomy was considered to be limited, the relationship between material conditions and equal worth was considerably stronger.
Separate multivariate multilevel analysis with autonomy as the outcome measure indicated that autonomy was dependent on material conditions; detainees felt that they had more autonomy where material conditions were judged favourably. However, autonomy also depended on other factors, such as not seeing treatment as routinized. The moderating effect of autonomy on the impact of material conditions in relation to equal worth suggested that, if staff behaved in ways that increased perceived autonomy, 17 they may be able to offset the negative effects of poor material conditions on feelings of equal worth.
Mediation analysis indicated that there was a significant within-group indirect effect of autonomy on equal worth through more positive feelings in relation to having nothing and insecurity (b = 0.281,SE = 0.041,95 per cent CI = [0.363,0.205]). Within a given custody area, greater feelings of autonomy led to greater feelings of security and that detention met basic needs, leading to greater feelings of equal worth. There was also a significant between-groups indirect effect (b = 0.267,SE = 0.154,95 per cent CI = [0.603,0.017]). An increase in facility-aggregated autonomy resulted in increased facility-aggregated feelings of equal worth by way of facility-aggregated feelings that detention contributed to a sense of security and that basic needs were met. In other words, those within custody blocks, where detainees felt that they had more autonomy, were less likely to feel insecure and as if they had nothing and this led to feelings of increased equal worth.
In sum, feelings of equal worth were significantly more likely if detainees: • were older and white, though this was mediated through trust in accountability mechanisms and the feeling that there was a culture of derision/lack of decency and 17 Autonomy may be enhanced through small acts, such as not making detainees ask for toilet paper each time they need to use the toilet.
• felt that they had a degree of autonomy, with perceived autonomy also moderating the impact of poor material conditions.

Discussion
The purpose of this paper was to conceptualize dignity in police detention and then to examine what predicted this, as a way of understanding 'good' police custody and 'good' police work. As expected, first, detainees' sense of equal worth was found to be integral to the detainee experience and, as discussed below, to 'good' police custody. Second, material conditions were significant to detainee experiences, in particular to feelings of equal worth. This builds on academic research that sees them as critical to prisoner experiences and to prison legitimacy, for example (Sparks et al. 1996;Jewkes 2018: 308). It also supports the importance attached to material conditions in ECtHR jurisprudence discussed above, where they are one of the indicators used to assess Article 3 violations (and by implication the diminishing of human dignity, in light of Bouyid v Belgium). Third, as expected, younger and BAME detainees were less likely to feel treated as a person of equal worth. For BAME detainees, this was mediated by a lesser degree of trust in accountability mechanisms and also increased feelings of having nothing and feeling insecure. For younger detainees, it was mediated by a lesser degree of trust in accountability mechanisms by the greater likelihood that they regarded police custody as having a culture of derision/lack of decency and as a place with an insufficient focus on the welfare of detainees. In the context of post-colonial migration and 'over-policing and under-protecting' black communities, as noted above, the results presented here suggest that the lack of trust in the police custody complaints process may be another aspect of police work, in which BAME citizens feel under-protected. Collectively, this may contribute to an overall lack of trust in 'the system' (Lammy 2017: 6), which has particularly damaging effects for BAME citizens during their encounters with criminal justice actors, potentially also shaping their experiences of citizenship (Lerman and Weaver 2014: 28). This paper has three sets of implications, first, about the conceptualization of dignity, second, about its importance and, third, about 'good' police custody and 'good' police work. Though this paper has focused primarily on dignity rooted in feelings of equal worth, three of the predictor variables of equal worth might in fact also be seen as other dimensions of dignity. Within the human rights-oriented literature, autonomy (the importance of which is noted in the present research) is seen as a key dimension of dignity, as is public decency (Henry 2011). Decency refers in the present research to whether detainees felt laughed at, as this is not something we would wish to see happen to a friend or family member and is also offensive to the individual concerned and to collective notions of expected behaviour. Within ECtHR judgements, material conditions in places of detention have been clearly recognized as a dimension of dignity. Given the significant relationship between these three potential forms of dignity and dignity rooted in feelings of equal worth, this suggests that, as noted by Henry (2011), dignity may be comprised of inter-locking strands. 'This approach views dignity not as a concept, but rather as a series of meanings that share a Wittgensteinian family resemblance. The types of dignity … are unique but like siblings, they have overlapping characteristics' (Henry 2011: 188).
Another way of thinking about this, conceptually speaking, is to see equal worth, autonomy and public decency as over-arching dimensions of dignity, perhaps 'siblings' using Henry's (2011) terminology (see Figure 3). Within equal worth, notions of respect that come from the police legitimacy literature and the desire to be treated as a human being as discussed in the present paper might be seen as 'children' of the equal worth 'parent'. After all, part of the reason why being treated with respect means so much to people is because it conveys to them that they are a person in the fullest sense and are being treated and valued in the same way as others. Within the autonomy strand of dignity, 'child' sub-strands of this might include limiting helplessness and dependence on staff, identified in the present research, but also limiting coercion (seen in ECtHR judgements) and encouraging citizen participation in decision-making (seen in the police legitimacy literature as a dimension of procedural justice). All of these strands have some kind of impact on the extent to which detainees feel in control and/ or not dependent on staff. Finally, within the public decency strand of dignity, 'child' sub-strands include laughing with, not at, detainees (seen in the present research), high-quality material conditions (seen in ECtHR judgements and the present research) and adequate access to health care (seen in ECtHR judgements), all of which one would expect to see in civilized societies. This melding and recalibration of the conceptualization of dignity in legal and criminological debates hopefully brings greater clarity to a hitherto ambiguous concept (McCrudden 2008). It also suggests that some aspects of procedural justice in the police legitimacy may in fact be about different dimensions of dignity.
Whilst it may suffer from conceptual ambiguities in academic debates, dignity has tangible qualities for those in places of detention (Liebling 2011), which makes its continued presence in criminological debates important. As shown in the present research by the concept of 'treat me like a human being', meanings can be readily ascribed to dignity, from the ground up, based on inmates' lived experiences of detention. The value of dignity, therefore, lies in its capacity to transform the lives of individuals, particularly in situationally and structurally unequal contexts. The police wield considerable power over suspects in police custody as part of a broader context of growing social inequalities (Skinns 2019: 201). Recognising suspects as of equal worth has the potential to disarm and disrupt the effects of these inequalities. Dignity's value also lies in its capacity to transform criminal justice organizations. For example, in causing staff to consider their relationships with detainees and whether detainees deserve to be afforded dignity, it may encourage greater reflexivity, such as about the moral and ethical aspects of their work, in ways that are notably absent in other comparable detention settings, such as immigration removal centres, in which staff may be similarly little invested in those temporarily in their care (Bosworth 2019). Though some may see it as too 'flabby or squishy', it is no more so than other values in criminal justice processes, such as equality or liberty. Taken together, this suggests that dignity can motivate and inspire citizens and criminal justice actors to instigate change (Simon 2017: 276). This includes, so we argue below, when harnessed to notions of 'good' police custody, suggesting it may also have problem-solving qualities (Dupré 2015: 16).
However, dignity is not a magic silver bullet. Encouraging police practitioners to frame police custody through the lens of dignity may require cultural and attitudinal shifts, which may be hard to put into practice. Focusing on dignity may e.g. require custody staff to employ more proportionate responses to risk, such as when making decisions about the removal of clothing, which many staff find difficult, working as they do in what they perceive to be a climate of risk (Skinns 2019: 98-99). Dignity also has the potential to be a ' fig leaf', which may obscure poor practices, if only lip service is paid to it. That is, in some circumstances, dignity may have a hollow and ideological quality to it (Harvey 2005: 14), particularly where practices are pain inducing, coercive and totalitarian in nature (Skinns et al. 2017). Focusing on dignity may also eclipse challenges to the institution of police detention and the wider criminal justice process, and consideration of the circumstances in which detention is necessary at all, particularly for vulnerable suspects.
On balance, though, so long as these caveats are born in mind in all aspects of police custody work, from research to policymaking to practice, and actors in these settings are reflexive in operationalizing dignity, there is more to be gained than lost by encouraging a greater focus on dignity in police detention. As indicated by the many potential indignities of police detention and the accumulated harms they cause to detainees' sense of self-worth and their sense of belonging (Fagan 2017), and also given the findings from the present study, dignity is an essential aspect of the detainee experience. It is, therefore, highly salient to notions of 'good' police custody and, indeed, to good police work.
In order to build strong links with communities, the police should promote the value of dignity in police custody, in particular linked to feelings of equal worth and the need to be 'treated like a human being'. Up to 700,000 people are arrested and detained in police custody each year in England and Wales (Home Office 2018: 9), meaning it represents a prime opportunity for police forces to build relationships with their local citizenry. Emphasising dignity would also convey some of the wider purposes of police custody beyond the contribution made to criminal justice processes. Greater attention should, therefore, be paid to it in domestic police custody policies, strategies and codes of practices at the national and police force level, including the PACE Codes of Practice, the National Police Custody Strategy and Authorised Professional Practice. Given the growing attention paid to dignity in ECtHR judgements, in which violations of someone's dignity has been seen as tantamount to a violation of Article 3, changes to these policies, codes of practices and strategies might also help to prevent such ECHR violations.
The findings from the present paper also shed light on other aspects of good police custody noted at the beginning. In terms of the just dimensions of 'good' police work, the findings suggest that the goods and impositions of police custody are not yet evenly distributed. Younger adults and BAME detainees were less likely to regard themselves as having been treated as of equal worth. Therefore, as predicted in Skinns et al. (2015), police custody is not yet equally 'good' for all detainees. Therefore, thinking of police custody in terms of whether it is ' just' is a salient way of conceptualizing 'good' police custody. This is particularly so whilst police forces strive for an equitable distribution of the goods and impositions of police custody, e.g. by examining strip and search practices through scrutiny panels involving citizens.
As for the legitimacy of police custody, further analysis is required before its relevance can be ruled in or out as a means of conceptualizing 'good' police custody, which trumps all others. Were legitimacy to be further examined, a more human rightsfocused understanding of dignity would need to be incorporated into the analysis, emphasizing equal worth and the need to be treated like a human being.
This paper also reveals a further aspect of 'good' police custody, which has not previously been considered, though it has been given growing attention in prison studies and ECtHR rulings (Moran et al. 2017;Jewkes 2018). These are the material conditions of custody. Such conditions are critical to 'good' police custody given their representational nature-they convey to detainees something about how they are perceived by the police authorities, by the state and by wider society (Sparks et al. 1996: 308)-and given their importance to detainee dignity linked to equal worth.
To conclude then, dignity rooted in the equal worth of detainees and derived from the human rights literature, is central to the conceptualization of 'good' police custody, as are the material conditions in which detainees are held. Dignity is a lens to explore the building of links with communities and notions of ' just' police work, which have been previously regarded as elements of 'good' policing. Dignity linked to equal worth should, therefore, be embedded in encounters between staff and detainees, as well as in the language and cultures of police work, starting with police strategies, policies and codes of practice. Given the harms caused by indignities of different kinds, this type of dignity should be seen as an end in itself, not merely as a route to legitimacy and citizen cooperation with the police. Maintaining dignity in police custody will prevent these harms from accumulating within and across police-citizen encounters. As this form of dignity is also potentially relevant to other kinds of police-citizen encounters (Bowers 2014;Fagan 2017), it could be used to conceptualize 'good' police work not just 'good' police custody, alongside other considerations, such as legality (Bottoms and Tankebe 2012;Murphy and Cherney 2012;Meares et al. 2015;Huq et al. 2017). A human rightsinfused understanding of dignity also has a potency to it, particularly when its origins in post-war 'never again' ideals and its role in advancing human rights are acknowledged (Dupré 2015: 198) and also when it is aligned with human rights, such as to a fair trial, liberty, life etc. Together, these considerations suggest that dignity in police custody and indeed in police work is an important end in itself.

Funding
This research was supported by the Economic and Social Research Council (ES/ J023434/1).