‘When You're Sitting in the Room with Two People One of Whom... has Bashed the Hell Out of the Other": Possibilities and Challenges in the Use of FGCs and Restorative Approaches Following Domestic Violence

Domestic violence continues to be a primary reason for referrals to state child welfare services in advanced industrialised countries. There is growing concern in many state child welfare services to develop responses to it that are both more effective and more humane. The use of restorative approaches, in particular Family Group Conferences (FGCs), has been suggested as one such response. This article draws from data gathered from an evaluation of a UK Government funded ‘Innovation Project’ part of which extended the use of FGCs in an urban local authority area which was already making extensive use of them. This paper presents and explores a typology of FGCs used in situations of domestic violence: pragmatic, resolution-focussed and restorative FGCs, developed from the evaluation data and augmented by relevant literature. The study data revealed pragmatic FGCs to be the most used, restorative the least. It is suggested that each type of FGC brings potential benefits but only restorative FGCs offer the possibility of full restoration in the traditionally understood sense. It is argued that the present mother-centric, risk-adverse, child protection systems which currently operate in many countries provide a powerful resistor to the greater implementation of this restorative way of working.

A C C E P T E D M A N U S C R I P T

Introduction
Domestic violence continues to be one of the primary reasons of referrals to state child welfare services in post-industrial countries. State social work practice in Anglophone countries has moved from a dominant perspective of minimising domestic violence to encourage family preservation in the early post-war period (Gordon, 1989) to a current separation perspective where a mother's separation from a male perpetrator is often seen as the only option, to be enforced by the state regardless of a mother's own preferences and circumstances, and often with limited support to a mother (Goodmark, 2015). There is growing concern in many state child welfare services to develop responses that are both more in City allows the data to make a unique contribution in taking forward conceptual and applied understandings of FGDM.

A C C E P T E D M A N U S C R I P T
In this paper we discuss the use of FGCs in situations of domestic violence, one of the primary foci of the Innovation funding in City. The paper firstly provides an overview of the complicated relationship between FGCs and Restorative Justice (RJ) before secondly identifying the ways in which the current operation of policy and practice around domestic violence are providing a driver for the development of alternative responses. Thirdly, we introduce and explore a three part typology of FGCs in situations of domestic violence developed from the study data (pragmatic, resolution-focussed and restorative FGCs). We conclude by arguing that while the sensitive use of any of the three types of FGC can serve as a bulwark against the undue responsibilisation of mothers for preventing domestic violence, only restorative FGCs, where perpetrators overtly take responsibility for the harm they have caused, offer restoration in the traditionally understood RJ sense. We suggest that the present mother-centric, risk-adverse, child protection systems which now operate in many countries (Goodmark, 2015) provide a powerful resistor to the greater implementation of restorative ways of working.

Conflation, confusion or useful hybridization?
FGCs developed in the late 1980s in New Zealand in discussions with the Maori community around ways of responding to the over-representation of Maori children in the New Zealand A number of different approaches fall within the RJ label including FGCs themselves, but also victim-offender mediation, community reparation panels and peace making circles. The most obvious thread that binds them is attempts to establish dialogue between victims, perpetrators and communities affected by a crime in a way that balances the, often conflicting, aims of safety, accountability, empowerment and restoration (Ptacek, 2010). As this description may suggest, the relationship between FGCs in child welfare and RJ is not straight forward: the acknowledgement of an offending harm and the attempt to restore personhood to victim and offender are less obviously applicable to child welfare than criminal justice contexts. This may explain why FGCs were not initially identified as an RJ practice. However, their use in New Zealand with young offenders became rapidly embraced as an example of RJ in action and spread rapidly to other jurisdictions outside of New Zealand under this label. The use of FGCs in matters of child welfare, including in situations of domestic violence, also spread internationally but was notably less associated with the concept of 'Restorative Justice'.
The importation of FGCs into different policy, legal, cultural contexts has further resulted in conflation, and arguably confusion, about the relationship between FGCs and RJ. In the UK, the use of FGCs was imported from New Zealand in the early 1990s. While closely following the FGC process developed in New Zealand, their use in the UK is discretionary rather than mandatory, resulting in marked geographical variation in the extent to which FGCs are used between different local regions. Additionally, though the importation of FGCs to the UK

A C C E P T E D M A N U S C R I P T
reflected some of the social justice and rights based principles which accompanied their introduction in New Zealand, their use has also sometimes been more pragmatically framed as a mechanism for reducing the use of state care for children and, even, as a mechanism by which to ration services to families (Morris and Connolly, 2012).
Ironically, though the use of FGCs in situations of domestic violence is one area of child welfare which lends itself more obviously to the formulation of harm and restoration common to RJ approaches, the desirability of restorative approaches in this area has been heavily contested for other reasons. RJ practices were initially developed to respond to stranger crime rather than intimate partner violence, and in some jurisdictions their use in situations of domestic violence is specifically outlawed (Ptacek, 2010). It has been argued that RJ approaches do not take sufficient account of the particular nature of intimate partner violence as a repeat offence, often involving overt and subtle forms of coercion and control that are embedded in a relationship, and which are targeted on a specific individual who has had a long-term connection to the perpetrator (Stubbs, 2002). It has been questioned whether restorative approaches can ensure sufficient safeguards for women and children; whether perpetrators can be held properly accountable during, and after, a restorative meeting; and whether the dynamics of wider community relationships necessarily provide a proper foundation for addressing family violence through restorative fora (Stubbs, 2010 in order to 'mediate between "victim" and "perpetrator" ' (p.108) and those FGCs that are focused on safety planning for women and children. Their focus is on safety-orientated FGCs which, they argue, can also be reparative and therefore partly restorative, by confronting a perpetrator with the impact his actions have had on a family. We suggest a different emphasis that the focus of restorative FGCs should be, rather than mediation, the recognition of the harm done by the domestic violence and the attempt to address this harm and restore relationships. We would also highlight that there is empirical evidence that the sensitive use of restorative FGCs in cases of domestic violence can also foreground safety issues, including those related to power and gender power dynamics (Burford & Pennell, 1998;Pennell & Burford, 2000, 2002. Restorative approaches are currently being used to address complex harms including only in situations of intimate partner violence, but also child to parent violence, sexual violence and child sexual abuse (Ptacek, 2010). They offer the potential for victims' voices and experiences to be heard in a way that formalised court processes do not; for appropriate redress for the harm done to be given to victims; and, for perpetrators to take responsibility for their violence in a way that facilitates their reintegration into the community and reduces the likelihood of recidivism (Braithwaite and Strang, 2002).

Domestic violence and child protection: practice developments, contemporary questions
The most powerful international driver for change in domestic violence policy and practice is the impact of current policy and practice on women. State acknowledgement of the impact of the harm domestic violence causes can be viewed as a positive result of feminist-led activism, primarily located in non-state agencies, which campaigned for recognition of domestic violence as a public harm. The criminalisation of perpetrators can also provide the redress and protection some survivors of domestic violence seek. However, the state co-option of feminist concerns about domestic violence has also considerably contributed to increased

A C C E P T E D M A N U S C R I P T
state surveillance and regulation of many marginalised women who have been victimised through domestic violence (Goodmark, 2009 Such responses also reinforce underlying structural inequalities. The impact of essentialist context-insensitive laws, policies and practices inevitably weigh most heavily on marginalised families -among them poor, non-white, immigrant and refugee and LGBT families -whose lives are more likely to be the target of public surveillance, disapprobation and regulation (Coker and MacQuoid, 2015). In the USA it has been noted that separationfocussed state responses to domestic violence in poorer families, coupled with a carceral approach to male perpetrators of domestic violence, contribute to both hyper-surveillance and hyper-incarceration of marginalised communities, feeding inequality, poverty and disrupted communities and family connections. In turn these conditions make domestic violence more likely (ibid.) The experience of a number of women subject to domestic violence is also that the criminal justice system is more focussed on securing a perpetrator's conviction than in meeting their needs for the harm done to them to be acknowledged and addressed (Pennell et al., 2013).

A C C E P T E D M A N U S C R I P T
The two-sided nature of policy and practice developments can also be seen in the UK. The Adoption and Children Act (2002) in England and Wales provided welcome recognition of the harms that children can suffer as a result of witnessing domestic abuse between adults and, indeed, the co-occurrence often of abuse to women and children (Rogers and Parkinson, 2018). However, as in the USA, women and families, can be subject to the double jeopardy of victimisation by an abusive partner and child protection intervention due to also viewing her victimisation as failure to protect her children from domestic violence (Stanley and Humphreys, 2017). While the harm to children of experiencing domestic violence is rightly a concern for services, the pressure for mothers to separate from abusive partners can be combined with the responsibilising of mothers for managing and enforcing that separation, with limited support from the state agencies mandating that separation. Such responses often reflect a failure to fully recognise that domestic violence and coercive behaviours frequently on women, or on men with few approaches that work with all family members simultaneously around domestic violence (Hester, 2011). While there is evidence of some innovation, many services are currently only able to be focus on the provision of safe haven

A C C E P T E D M A N U S C R I P T
and immediate care for women and children at risk of harm, with funding for ongoing support services and community based provision limited. These services have been particularly vulnerable to spending cuts, with evidence of that the poorest localities in the UK have been seen the greatest decreases in funding and meaning that accessibility of services between areas is uneven, with the areas of greatest need the least able to provide needed supports (Author publication, 2018).

Methods
In City, the FGC service had existed since 2008 and before the Innovation funding provided a service to families referred for an FGC by the child and family social work teams in City due to any care and protection concerns, including domestic violence. Internal requirements in City mean social workers have to refer families for an FGC when there is sufficient concern to merit the calling of an Initial Child Protection Conference (ICPC)a child protection meeting called in the UK system when there are serious care and protection concerns -

A C C E P T E D M A N U S C R I P T
in state care, and there was also a specific focus on reducing the number of repeat referrals to services due to domestic violence (Author publication, 2017).
The FGC service evaluation was mixed-method and multi-modal, taking place over an eight month period. Ethical approval was granted via City's ethical governance arrangements and by University X. All those taking part did so through a process of informed consent and all data was stored securely and anonymised. In the data presented pseudonyms are used and some case details have been changed to protect family and professional identities. The methods adopted included:  An analysis of administrative data held by the service for all children and families referred to the service in the 2014 and 2015 years;  15 days of practice observation in the FGC teams over three months;  The development of ten case studies of families who were tracked over the observation period;  Focus groups with different groups of co-ordinators convened at three points over the course of the study;  Semi-structured interviews (n=39) and questionnaires (n=66) with FGC co-ordinators and managers;  And, structured telephone interviews with adult family members in families who had been offered an FGC by the FGC service. Respondents were mainly birth parents but in some cases were kinship carers for children in their family network (n=36). These families included those referred due to concerns about domestic violence, but were not exclusively so.

A C C E P T E D M A N U S C R I P T
The typology of FGCs presented below was a key finding from the evaluation, primarily developed from the qualitative strands of the study. These qualitative data consisted of open text questionnaire responses from FGC co-ordinators, semi-structured interviews with professionals and parents, field notes from observations and repeat informal interviews with FGC co-ordinators during the observation period.
Initial data around domestic violence was gathered from August to December 2015 by interviews with the four FGC managers, focus groups with co-ordinators from across the four Issues around practice with domestic violence came up both from questions specifically asked around it in interviews and focus groups, given the Innovation Project funding aims, but were also raised by participants over and above this. In the vast majority of domestic violence cases discussed and observed the violence was primarily male to female physical intimate partner violence, with a serious assault bringing the family to the attention of police

A C C E P T E D M A N U S C R I P T
and children's services, or escalating concern if they were already involved. Such violence was the primary concern in three out of the ten case study families. In one further case study family a pre-birth assessment was being conducted relating to the severe domestic violence the mother had experienced from a previous partner, which had resulted in her older children living outside her care. The early development of the typology was inductive and arose after analysis of the observation field notes and initial interview , focus group and open-text questionnaire data. From this analysis, pragmatic and resolution-focussed FGCs were identified. However, interviews with FGC co-ordinators and managers also suggested that FGCs were happening that did not fit these two types, but which we had not observed. The description of r estorative FGCs was therefore initially developed inductively from interview and focus group data but was further supplemented by engagement with relevant wider empirical literature, particularly the work of Burford and Pennell (1998) and Pennell and Burford (2000;2002). The typology was presented, alongside other key findings, to a member checking event involving the whole FGC Service in City in October 2016, after which it was refined based on the feedback given by the workforce.
The statistical data presented in this paper comes from two sources. Firstly, FGC coordinators completed, at two points in time six months apart near the start and end of the evaluation (T1 and T2), a self-evaluation questionnaire consisting of on 18 item list connected to their practice. They rated their confidence in each item on a scale of 1 (least confident) to 10 (most confident). This was administered twice to track whether there was any change in co-ordinator self-assessed confidence and knowledge in the period of the evaluation. The mean co-ordinator scores for the 32 co-ordinator questionnaires for which we had responses at both time points were sorted into Excel and comparisons for each of the 18 items were made using a paired T test to consider statistical significance. Secondly, basic administrative data was kept on all families referred to the FGC service and was primarily

A C C E P T E D M A N U S C R I P T
analysed to provide some basic descriptive and outcome data on families who received an FGC. We do not present outcome data here, rather the descriptive data were cross-tabulated and re-analysed using R to provide additional insight into the involvement of fathers in FGCs in City using Chi square tests for statistical significance. This was undertaken in order to help better understand the apparent relative rarity of resolution-focussed and restorative FGCs.
For both sets of statistical data, critical values for significance were set as p < .05. As the administrative database did not record referral reasons to the FGC service we were unable to separate cases where domestic violence was the referral issue for the analysis. However, the observation work in the FGC teams and previous data on service referrals in City suggested that around a third of referrals to both the FGC service and to child and family social work teams were ones where domestic violence was a primary concern.

Describing the typology
We explain pragmatic and r esolution-focussed FGCs with brief descriptions from two of the case study families -the most accessible way of illustrating the differences in the FGCs described. As we did not have opportunity to directly observe a restorative FGC, the description of this type is supplemented by one co-ordinator's description of their work, supplemented by engagement with previous literature in this area. We recognise that FGCs are unlikely to fit neatly into only one of the practice types outlined. However, our data suggest that an FGC tends to gravitate towards one of the three types over the course of a family's engagement. This gravitation will be influenced by the extent of paternal, as well as maternal family engagement, and the extent to which the harm caused by the domestic violence, and its underlying causes, are explicitly addressed in the framing purpose of the FGC.

Pragmatic FGCs
The

Resolution-focussed FGCs
This type of FGC focused on the resolution of disagreements or practical issues relating to the care of the children involving the mother and father, the maintenance of children's links with wider family networks, and discussion of family arrangements involving maternal and paternal networks. This type of FGC involved some representation of paternal as well as maternal networks. The FGCs observed were post-separation after domestic violence, but with overt intentions for fathers and/or paternal networks to maintain a role in children's lives. These FGCs were forward-looking rather than seeking to provide redress for past harms caused by domestic violence. Children's welfare was foregrounded within the aim of resolving family disputes, and children could be the drivers for this type of FGC taking place through their desire for contact with a non-resident father. Social workers' awareness of the challenges mothers faced in facilitating contact plans could also be raised through their own participation in this type of FGC.

Restorative FGCs
This is the most challenging and contested type of FGC in the area of domestic violence and was also the least prevalent in City. Restorative FGCs are premised on the desirability of engaging paternal as well as maternal networks, including the perpetrator of the domestic violence, in the preparation of the FGC, and, where safe, the FGC itself. The FGC will focus on acknowledging the harm caused by the perpetrator's violence, providing redress on the part of the perpetrator to those affected by it and putting in place a plan to support both the well-being of the mother and children and the perpetrator's behavioural change. We draw here on Ristock and Pennell's concept of 'links and interruptions' praxis as illustrated in the work of Pennell and Burford (2000). The FGC seeks to fire the mechanism of galvanising greater networks of support and professional support for core family members ('links'). In the preparation work there will be also be extensive engagement and information sharing with the perpetrator and their own networks around the nature of their family violence, its harmful effects and discussion around the prevention of future violence ('interruptions'). This type of FGC also draws on the concept of reintegrative shaming (Braithwaite, 1989). The perpetrator's actions are condemned, but with a view to reintegration rather than exclusion.

A C C E P T E D M A N U S C R I P T
The perpetrator's acknowledgement of the harm they have caused, their acceptance of the need to undertake reparative action to address the harm, and the victim's acceptance of the reparation offered as a suitable redress underpin the reintegration of the perpetrator. In cases of domestic violence, this reintegration can also be a means of restoring personhood to fathers thereby facilitating them to play an ongoing positive role in their children's lives Though the mother wasn't afraid of her partner quite a few family members were and they were very anxious he was going to kick off. And I just felt he really, rea lly needed to be there.
He was living in the home, the children were living in the home, his issues just sort of needed to be confronted and dealt with so I did quite a lot of work talking to aunties, uncles and cousins and I felt he could sit through and behave himself with enough pr eparation and I just kept toing and froing and giving people the option that you can leave at any time, and the afraid people did come and he [the father] didn't do anything, and they arrived at a good family plan (Co-ordinator 11, Focus group i )

Exploring the typology
Each type of FGC outlined has potential merits and challenges. Stubbs (2002)  These meetings are likely to be tense, but can allow family issues that would otherwise not have been openly discussed to be planned for, with arrangements explicitly agreed. At their most effective Resolution-focussed FGCs can support the restoration of relationships between a mother and father that allow them to discuss and manage practical child care issues and the safe management of non-resident fathers' contact with their child or childrencontact which may well have otherwise occurred in an unstructured manner, without parameters. They are also limited in the extent of restoration they can offer due to the lack of explicit focus on the

A C C E P T E D M A N U S C R I P T
past harm caused by a father's prior violence. There is also a possibility that such forwardlooking meetings could minimise past domestic violenceone of the enduring concerns around the past use of family 'mediation' following domestic violence (Ptacek, 2010 Restorative FGCs offer the possibility of achieving fuller restoration through focussing on the harm caused by domestic violence and addressing its underlying causes. FGCs cannot be the sole mechanism of such change, and not all violent men will be capable of, or willing to, engage in an FGC process in a manner necessary for it to be restorativein this respect it is relevant to note that City had also commissioned wider service development focussed on supporting men to address their violent behaviour. However, restorative FGCs could play a substantial role. Many men who have been violent will remain in contact with their children,

Exploring reasons behind the rareness of restorative FGCs
We interrogated the data to better understand the reasons which might militate against the greater use of restorative, and to a lesser extent r esolution-focussed, FGCs in city and found three potential interlocking influences: co-ordinator practice confidence around domestic violence; maternal-network centric FGC practice; and, separation-focussed social work practice.

Co-ordinator practice confidence working with family violence
FGC co-ordinator practice was generally highly-skilled, empathic and committed to supporting family strengths (Author publication, 2017). Underlying this, co-ordinators appeared very confident in their practice. The self-evaluation questionnaire data showed coordinators started off with high levels of confidence in their job role at the start of the evaluation (T1, mean overall confidence rating, 7.6) and remained so close to the end of it (T2, mean overall confidence rating, 7.81). However, co-ordinators were less confident working with families where violence was an issue (see Table 1).

A C C E P T E D M A N U S C R I P T
The T1 mean score for Q.9 'Working with families in conflict (including violence)' (6.84, SD, 1.273) was the second lowest of the 18 items; and though the T2 mean score had increased (7.44, SD 1.107), the increase was not statistically significant and the item was still the joint second lowest scored item at T2. This item was also only one of two items where any co-ordinator ranked their confidence as low as a '2' at either time point. Interview and focus group data suggested this area of practice was a considerable cause of some anxiety for some co-ordinators. The co-ordinator workforce were keenly aware of the reservations that

A C C E P T E D M A N U S C R I P T
'My own anxieties about involving Dads where there has been domestic abuse in parental relationship.' (Open text questionnaire response on negatives of current job role, T2, coordinator 33)

Maternal-network centric FGC practice
The second potential factor was the relative lack of paternal network involvement in FGCs.
In 2015, the FGC service had started to record the attendance of maternal networks (any maternal family member), the attendance of external paternal networks (any paternal family other than the father), the attendance of the father, and the engagement of the father in FGC preparation work. We analysed the available data for those families for whom these indicators were recorded in 2015, looking only at those families where children were living with either or both of their birth parents. The Chi square tests for statistical significance revealed that that there was no statistically significant association between maternal family attendance at FGCs and children's living arrangements: a very high proportion of families had at least one adult from the mother's side attend the FGC, regardless ( Table 2,  2 = .514, df = 2, p > .05). By contrast there was a statistically significant association between the attendance of extended paternal family and the living arrangements for the children (Table 3, Fathers' own attendance (Table 4,  2 = 29.8, df = 2, p < .001) and engagement in FGC preparatory work (Table 5,  2 = 20.787, df = 2, p < .001) were also statistically associated

A C C E P T E D M A N U S C R I P T
with a child's living arrangements: children living with their father or with both parents, had a higher proportion of fathers attending FGCs than children who were living only with their mother (85.7 per cent compared to 42.5 per cent). Similarly, children living with either their father or both parents had a significantly higher proportion of fathers being involved in preparatory work for an FGC (90 per cent and 91.8 per cent respectively, compared to those children living with their mother (58.2 per cent). While we were unable to identify the proportion of these families for whom domestic violence was a primary concern, it is the case that the maternal-network centric nature of FGC attendance would militate against the use of both resolution-focussed and restorative

Tables 2 -5
FGCs when domestic violence was a concern. The statistical results also do not indicate why fathers and paternal were less involved and whether this was due to mother-centric practice or fathers' refusal to engage. However, the qualitative data provided some illustration of how these factors could combine. Mothers tended to be the first point of contact for FGC coordinators as they were overwhelmingly the primary carers for children. In interviews and focus group discussions co-ordinators discussed how they worked hard to involve men and we observed co-ordinators attempting to engage, fathers, step-fathers and male partners living in the family home. However, we also saw that some men did not respond to requests to meet with co-ordinators. In such instances, co-ordinators then had to make pragmatic decisions about proceeding to FGC with maternal networks alone, or trying to further engage fathers.
Particularly when fathers were not living in the family home it appeared the former was more likely as one co-ordinator illustrated in their description of practice:

A C C E P T E D M A N U S C R I P T
We also saw how decisions to proceed to an FGC with only maternal networks could be more likely in cases involving domestic violence because mothers were offered an effective veto on who else should be invited. In the Jones-Smith family case, the co-ordinator sensitively reflected on a decision to proceed with an FGC with the maternal network a lone, because pushing for the involvement of the extended paternal family involvement might jeopardise the prospect of any FGC happening: Mrs Jones has agreed to her family being involved but father's family not currently included. The absence of his family does limit the meeting in the co-ordinator's view but the co-ordinator sees this as an initia l meeting and they can work towards wider participation over time -she doesn't want to rush at this and lose the progress made.
(Fieldwork note following practice observation with Jones-Smith family).

The influence of a separation perspective
The final factor was a separation perspective evident in some children and families practice in City. Co-ordinators had considerable experience of social workers seeking to exclude men from an FGC whom they felt were a risk without giving them opportunity to try and engage the men around their violence: Have you ever excluded a man from an FGC due to domestic violence? (Resea rcher,

A C C E P T E D M A N U S C R I P T
This scenario was illustrated in the Jones-Smith case where the social worker had deemed the fa ther to be too 'high risk' to be engaged in a discussion about an FGC, due to his domestic violence and attitudes towards his violence. However, as noted, such an assessment was inconsistent with the social worker's delegation of the responsibility for managing any future potential contact between Mr Smith and the children to Mrs Jones and her family.
The existence of a separation perspective was also suggested by the reported reception of the new FGC team in City: Discussion with co-ordinators in New Team, the introduction of the domestic violence FGC service seems to have led to some negativity amongst other professionals in

City. Comments include those suggesting 'Mum' should be leaving 'Dad' rather than
having an FGC (Fieldwork notes following practice observation day one, New Team) Allied with the two factors discussed above, a separation perspective set a context for an FGC which meant violent fathers were unlikely to be engaged, making restorative FGCs unfeasible, as well as resolution-focussed FGCs less likely.

Limitations
Not all of the data, most notably the administrative dataset around maternal and paternal network involvement, exclusively involve families who have experienced domestic violence and we do not therefore know to what extent maternal centric representation at FGCs was influenced by the fact that domestic violence was present, or not, as result. It is possible to say that, in terms of child welfare practice in City more broadly, these data do strongly suggest maternal network centred representation at FGCs. Whether this was down to paternal refusal to engage, practitioner choices or maternal preferences for not involving fathers is not possible to determine clearly from the data, but qualitative insights did suggest all three factors could combine. Secondly, while the administrative dataset did have some basic

A C C E P T E D M A N U S C R I P T
outcome data, we have not presented these here as it is not possible to disaggregate them for families where domestic violence is an issue and the focus of this paper is to explore the processes whereby FGCs are offered to families in situation of domestic violence. Finally, it should be noted that the typology is presented tentatively as a way of conceptualising FGC practice with families following domestic violence, but is by no means suggested as a definitive model.

Conclusion
There is growing evidence from a number of countries of the way in which separationinfluenced policy and practice following domestic violence, underpinned by an absence of recognising diversity in women's experience and wishes, are undermining victimised women's welfare (Goodmark, 2015). This article has illustrated that practice responses, even in a restorative-minded forward-looking state agency, may rub up against the constraints of mother-centric practice, men's own resistance to engaging with child welfare services and a separation perspective which currently dominates the response to domestic violence both in the UK and internationally.
A pre-condition for a violent perpetrator's involvement in a restorative process is their willingness to take responsibility for the harm they have caused through their domestic violence. Some men's unwillingness to do so will prevent restorative FGCs taking place.
However, in City, the possibility of r estorative FGCs is currently constrained by a stage before this. Fathers can be more reluctant to engage with FGC co-ordinators and fathers and paternal networks are also less persistently engaged by FGC co-ordinators; at the same time the involvement of fathers who have perpetrated violence is restricted by routine social work practicesin particular context insensitive responses which determine that families must separate and that violent men are too risky to be engaged. Such responses tend to outsource the management of violent men's future behaviour to families, largely unsupported. If more Feminist criticisms of restorative approaches in situations of domestic violence have focussed on three main questions: the safety of women and children, holding offenders sufficiently accountable and giving sufficient focus to victim experience and needs (Ptacek, 2010). These criticisms are usually made by those with many years of experience in engaging with domestic violence, and deserve careful consideration. However, it is also important to recognise that there is evidence that FGCs can be used restoratively in situations of domestic violence to reduce violence while foregrounding safety concerns for women and children.
This evidence highlights that while models of practice are conceptually important, the content of practice within the model is also crucially important. Extensive engagement and preparation with families, supports for families, the provision of preventative domestic violence services and, ultimately, beginning to unpick the structural and power inequalities underlying domestic violence, are all necessary ingredients to making restorative approaches work well.

A C C E P T E D M A N U S C R I P T
There is no declaration of interest for any of the authors which requires to be made

Funding
The evaluation on which this paper is based was funded by the Department of Education, UK as The Children's Social Care Innovation Programme.

A C C E P T E D M A N U S C R I P T
Highlights  A typology of Family Group Conferences (FGCs) after domestic violence is proposed.
 The typology consists of pragmatic, resolution-focussed and restorative FGCs.
 The study data revealed pragmatic FGCs to be the most used, restorative the least.
 Child protection systems currently inhibit greater use of restorative FGCs.