Between loss and restoration: The role of liminality in advancing theories of grief and bereavement

A recent national survey of bereaved partners found high levels of complicated grief and psychological distress, with evidence that loneliness and isolation may contribute to these outcomes. However, the mechanisms of action for this have not been explored. To advance grief theory this paper reports analysis of the survey free-text data to examine the relationship between social support and emotional responses to bereavement. Individuals bereaved of a civil partner or spouse 6 – 10 months previously were identified through death registration data. 569/1945 (29 %) completed surveys were received. Of those, 311 participants (55 %) provided responses to two free-text questions which asked about their ‘feelings since the death of their partner or spouse ’ , and ‘about the support around ’ them. Data were analysed using corpus-assisted discourse analysis and the discourse dynamics approach for figurative language. Participants described diverse emotional responses to the bereavement (e.g. sadness, anger, denial, acceptance), and the value of formal and informal bereavement support. Although many of the experiences described are accounted for in existing grief theory, some participants described a liminal experience not recognised within these theories. They felt trapped, unable to engage with loss or restoration, and unable to move forward as their planned future no longer existed. They sought out ‘communitas ’ (solidarity in experiences), but often found support from their social networks had diminished. Metaphors were used to describe this liminality, with partner grief expressed as a dark agentic force, a monster, an abyss, and as water. The findings of this study offer original insights into experiences and trajectories of bereavement, and our un-derstandings of prolonged or complicated grief. A novel model ‘Between Loss and Restoration ’ is presented to include these experiences. Recognition of the place for liminality within the spectrum of grief experiences could enhance grief literacy and improve formal and informal bereavement support provision.

A recent national survey of bereaved partners found high levels of complicated grief and psychological distress, with evidence that loneliness and isolation may contribute to these outcomes.However, the mechanisms of action for this have not been explored.To advance grief theory this paper reports analysis of the survey free-text data to examine the relationship between social support and emotional responses to bereavement.Individuals bereaved of a civil partner or spouse 6-10 months previously were identified through death registration data.569/1945 (29 %) completed surveys were received.Of those, 311 participants (55 %) provided responses to two free-text questions which asked about their 'feelings since the death of their partner or spouse', and 'about the support around' them.Data were analysed using corpus-assisted discourse analysis and the discourse dynamics approach for figurative language.Participants described diverse emotional responses to the bereavement (e.g.sadness, anger, denial, acceptance), and the value of formal and informal bereavement support.Although many of the experiences described are accounted for in existing grief theory, some participants described a liminal experience not recognised within these theories.They felt trapped, unable to engage with loss or restoration, and unable to move forward as their planned future no longer existed.They sought out 'communitas' (solidarity in experiences), but often found support from their social networks had diminished.Metaphors were used to describe this liminality, with partner grief expressed as a dark agentic force, a monster, an abyss, and as water.The findings of this study offer original insights into experiences and trajectories of bereavement, and our understandings of prolonged or complicated grief.A novel model 'Between Loss and Restoration' is presented to include these experiences.Recognition of the place for liminality within the spectrum of grief experiences could enhance grief literacy and improve formal and informal bereavement support provision.

Introduction
In England and Wales approximately 600,000 people die each year (Office for National Statistics), and around 1/5 of those (increasing from 122,000 in 2010 to 131,000 in 2019) were married or civil partnered (Office for National Statistics, 2022).People who experience a bereavement are more likely to experience poor health, including developing new health conditions, or a recurrence of a pre-existing condition, yet are less likely to access healthcare than non-bereaved people (Prigerson et al., 2001).The experience of grief for individuals varies greatly and is influenced by many factors.For some people the initial intense period of grief following a death reduces over time.For others, the intense grief persists, interrupting their ability to return to usual activities.This is described as prolonged grief disorder (World Health Organization; Prigerson et al., 2021), which is estimated to affect 10-20 % of bereaved people (Shear, 2010).Risk factors for developing prolonged grief disorder include older age, female gender and lower socioeconomic status (Kersting et al., 2011).However, the nature of the relationship has also been found to influence its intensity.Compared to the death of a parent or grandparent, death of a sibling, child or spouse is associated with higher prevalence and more intense symptoms of prolonged grief disorder (Fernández-Alcántara and Zech, 2017).While professional bereavement services are not always necessary, support from social networks is vital for bereaved individuals to aid recovery and ameliorate grief symptoms (Vanderwerker and Prigerson, 2004;Aoun et al., 2012).Recent work has called for greater attention to the relational aspects of grief, with support from social networks being one of the factors that can improve outcomes and experiences after a death (MacArthur et al., 2023).
For professionals and the public, models of grief can help to explain the emotional, social and practical responses to a bereavement, and understand the normal range of those experiences.Several contemporary models of grief have gained prominence in recent years.While earlier models focused heavily on stages or tasks of grief to be experienced or worked through (Kübler-Ross, 1969;Worden, 1996), more recent models have moved away from such a stepwise approach that could be potentially damaging, raising unreasonable expectations for the course of grief (Stroebe et al., 2017).Two more recent models have emerged which represent the Gold Standard for understanding grief experience.Firstly, the Dual Process Model (Stroebe and Schut, 1999), originally developed to understand coping with the death of a partner, describes the coexistence of grief orientation and restoration orientation, and the reality that individuals will oscillate or vacillate between these two positions as they attempt to move on with their life.Secondly, the theory of Continuing Bonds (Klass et al., 1996) recognises that individuals never move on from the relationship with the deceased, rather find a place for that relationship as they move forward.These models feature heavily in counselling, self-help and bereavement support literature, and contribute to societal grief literacy.
Post-bereavement research provides a vital lens to explore the impact of bereavement, and surveys provide important data on bereavement outcomes.Through the inclusion of free-text response boxes, research participants can explain their experiences in their own words, and reveal potential explanations for quantitative findings.Our recent national survey of bereaved partners demonstrated that a high proportion (59-66 %), of participants reached the threshold for complicated grief 6-10 months post bereavement, and reported high levels of psychological distress (Timmins et al., 2022).We found evidence to suggest that loneliness and social support may contribute to these outcomes (Timmins et al., 2022).However, the mechanisms of action for these finding have yet to been explored.The current study aimed to analyse free-text data from those surveyed individuals to explore the relationship between social support and the emotional impact of bereavement to advance theories of grief.

Research team and research paradigm
Our research team comprises clinicians (palliative care, mental health), and researchers (sociolinguist, sociologists, epidemiologists, psychologists, health services researchers) with experience of qualitative methods, palliative care and bereavement research.This study is positioned within a social constructivist paradigm which recognises the contribution of an individual's relational world, culture and society in how they interpret their experiences.

Recruitment
Full description of the sampling methods, and the quantitative analysis, are reported elsewhere (Timmins et al., 2022).The data for this study were extracted from a national cross-sectional survey of people bereaved of their partner or spouse 6-10 months previously (09/2017-01/2019).Invitations to participate were sent by the UK Office for National Statistics (ONS) to 1380 individuals who had registered the death of a different-gender spouse and 564 individuals who had registered the death of a same-gender civil partner or spouse.Survey packs included: a personalised cover letter that described the purpose of the study, the voluntary nature of participation, how to participate, opt out or withdraw, whom to contact if they had any queries, and leaflets describing bereavement support.Participants consented to the study by returning the completed paper or online survey.There was no requirement to provide additional written consent.This enabled those to participate who wished to preserve their anonymity, but individuals were asked to provide contact details if they were willing to be contacted in the future.The survey included two measures, the General Health Questionnaire-12(GHQ) (Goldberg and Williams, 1988) and the Inventory of Complicated grief-19 (ICG) (Prigerson et al., 1995), that, above a specific screening threshold, would suggest a likely psychiatric disorder or complicated grief respectively.Where a participant's responses reached either threshold (and they had provided contact details) the researchers contacted them to discuss possible sources of support.

Survey
The survey contained demographic questions and validated outcome measures (see reference for full list) (Timmins et al., 2022).It also contained free-text response boxes, which offered opportunities to provide additional information or clarification on responses to the measures.There were also two questions that sought to gain more detailed information related to the bereavement experience.Responses to these two questions form the data for the current analysis.To illustrate the influence of question ordering within the survey, the first free-text question of interest was within a section entitled 'Your own current feelings around the loss'.This opened with the first eight questions of the Texas Revised Inventory of Grief (Faschingbauer, 1981) (described as a measure of 'normal grief'), which explore past behaviours related to the loss, followed by the Inventory of Complicated Grief-19 (Prigerson et al., 1995) (designed to assess indicators of pathological grief), after which the following free-text question was asked: (1) 'Please use this space to share any more information about your feelings since the loss of your partner or spouse'.
Immediately after this was a section entitled 'Social Networks and Support'.This opened with the UCLA 3-item Loneliness Scale (Hughes et al., 2004) (which assesses companionship, feeling left out, and isolation) and the Medical Outcomes Study -Social Support Survey (Moser et al., 2012) (which assesses different types of support), after which the following free-text question was asked: (2) : 'Please use this space to share any more detail about the support around you.' The decision to focus on these two specific questions together was underpinned by our quantitative survey findings, which suggested a relationship between psychological responses to bereavement and access to support (Timmins et al., 2022).

Analysis
A corpus-assisted discourse analysis was conducted (Marchi, 2010;Marchi et al., 2018).This mixed-method technique combines the deductive quantitative procedures of corpus linguistics with the inductive qualitative techniques of discourse analysis.A criticism of corpus linguistics is that, due to the size of the datasets used, it often fails to consider the broader context of the examples of interest.In contrast, discourse analysis can be criticised for its over-reliance on individual examples and lack of transparency around their selection (Hardt--Mautner and Baker, 2009).Combining these two methods allows an appraisal of the whole corpus, with a more focused analysis of key aspects within that corpus selected in a systematic manner.This approach lends itself to large survey datasets, such as this one, where a summative content analysis is too superficial for the intended purpose, and other methods such as thematic analysis would be unwieldy due to the high volume of cases, hindering the potential for interpretation and comparison.Below we describe seven stages of analysis (A-G), across three phases.Analysis was led by the lead author (KB) and refined through discussion with the research team.

Phase 1: corpus linguistic analysis
Handwritten responses within the two free-text questions were transcribed verbatim using Microsoft Word, and pseudonymised (removal of all names of people, places, institutions etc).Microsoft Word files were converted into text files and imported into AntConc V3.4.4 for analysis.The main analytical tools within corpus linguistics are keywords, colocation and concordance.The keyword function identifies the most frequent keywords within a corpus allowing for a thematic description of the corpus.The concordance function allows the researcher to view all instances of a keyword in its immediate context within the corpus (see Fig. 1 for an example), and the colocation function enables the researcher to see which words are most frequently located near the keyword (commonly up to four words either side).In the present study these functions were combined using the deductive three stage funnelling down approach (Marchi, 2010) to identify and describe the corpus content related to emotional responses to the bereavement (macro level analysis) and to guide the subsequent discourse analysis (micro level).(A) A keyword search was conducted within the corpus to identify key semantic domains.Function words (including pronouns, prepositions, determiners and auxiliary and modal verbs) were excluded, leaving only content words (nouns, verbs, adjectives and adverbs).The most frequent content words related to emotional responses were identified and reduced to their lemma forms (the form of a word that appears as an entry in a dictionary, and that is used to represent all the other possible forms).Content words that occurred less than 10 times within the corpus were excluded, as were any words that, on review, did not relate to emotional responses.(B) For each keyword a colocation analysis was conducted to explore the textual behaviours of each term, followed by (C) concordance analysis to explore them within their lexical context.Each concordance list was then extracted as a subcorpus for discourse analysis.

Phase 2: discourse analysis
After familiarisation with each sub-corpus, an inductive discourse analysis was conducted in four stages to identify patterns and common features or themes within the data.(D) The first stage sought to develop a taxonomy of conceptualisations of emotional responses to the bereavement.This phase focused on identification of literal descriptions only.This was repeated with each keyword until a full list of conceptualisations (the taxonomy) had been identified.(E) The second stage sought to identify figurative descriptions of emotional responses.The process of identifying and coding the metaphors and similes across the corpora drew on the Discourse Dynamics Approach (Cameron et al., 2009), which focuses on identifying 'vehicle terms' in the data.These are words or phrases where the meaning in context stands out due to its figurative usage in comparison to the more basic (contemporary) meaning as used in other contexts (Pragglejaz Group, 2007).The item being described (in this instance grief) is the tenor, and the vehicle is the image that carries the weight of the comparison.The corpus-assisted discourse analysis described above was then repeated for all conceptualisations of support structures within the corpus.

Phase 3: theoretical engagement
(F) During the final phase of the analysis the findings of the corpus assisted discourse analysis were further interrogated through a deductive analysis informed by contemporary theories of grief, to explore to what extent these models account for the specific experience of partner bereavement.The models included for this phase were the Dual Process  Model (Stroebe and Schut, 1999), and Continuing Bonds (Klass et al., 1996).(G) When this was completed a final stage of theoretical engagement and refinement was conducted to account for any findings that fell outside the abovementioned grief theories.We identified the concept of liminality as providing an explanatory frame for these unaccounted for findings.Liminality has its roots in anthropology and has been in existence for over a century.It describes the feeling of being 'betwixt and between' two states (Turner, 1967(Turner, , 1969)).In essence, liminality is a universal experience.Individuals will have many significant transitions within their lives punctuated with liminal periods (Thomassen et al., 2018).Liminality can affect individuals alone, social groups or societies as a whole and can last moments, periods or epochs.As well as being related to an individual's sense of self, identity and metaphorical boundaries, liminality can also be used to describe spatial dimensions and physical borders, boundaries and thresholds (Thomassen et al., 2018).Many experiences of liminality fall within what is regarded as a normal frame of existence, however the intensity of that liminal experience will be shaped by the extent to which it removes the individual from existing structures and processes.

Participants
In total 569/1945 responses to the survey were received (response rate 29 %).Of those, 311 bereaved partners (55 %) provided responses to either or both of the free-text response questions of interest, contributing to a corpus of 26,173 words.Just under 2/3 of our participants were female (197/311, 63.3 %), and the mean age was 67.4 (range 33-97).Just over half of participants (172/311, 55.3 %) were bereaved of a different gender partner or spouse, and a similar proportion described themselves as completely or mostly heterosexual (171/ 311, 55.0 %).(See Table 1 for further participant demographics).When compared to the characteristics of the survey respondents who did not provide free-text responses, there was little difference in the age of the two groups (free-text providers mean age of 67.4 years (range 33-97) compared to 67.6 years (range 40-98) for those who did not provide free-text responses).Participants providing free-text responses were however more often female (63.3 % vs 53.3 %), more likely to describe themselves as gay or bisexual (40.8 % vs 36.2 %), and more likely to have been bereaved of a same gender partner (43.3 % vs 38.1 %) than those who did not contribute free-text responses.

Alignment with existing models of grief i. Emotional Responses -Keywords
The initial keyword search for emotional responses to the bereavement identified 24 lemma forms including verbs, adjectives and abstract nouns (see Table 2).
ii. Emotional Responses -Findings Participants described in detail the spectrum of emotional responses to the death of their partner, particularly the pain they had experienced (example 1, Table 3).Sadness was frequently referred to (e.g.'emotional', 'cry easily', 'bereft', 'desolate', 'distressed'), and was characterised as 'deep', 'intense', 'constant', and 'overwhelming'.Feelings of loneliness (example 2), depression and anhedonia (example 3) were often described, particularly in relation to the loss of the primary relationship, and the irreplaceable nature of that bond.Although for some the depth of the bond remained a source of comfort and familiarity (example 4), for others for whom the bereavement had been traumatic, the nature of the bond resulted in all consuming grief.This was characterised by 'desperation', 'shock', and 'despair', feeling 'on the edge of a breakdown', and that it was 'not worth going on' with life (example 5).For others, the realisation of their partner's death left them feeling vulnerable after the bereavement and fearful of continuing alone ('low in self-esteem', 'helpless', 'anxious', 'incapable').
Feelings of anger were also common after the bereavement (example 6).This manifested in different ways ('bitter', 'jealous', 'resentment', 'irritable') with some individuals describing anger towards those in their social network, healthcare professionals, or the disease itself, and others describing a generalised anger.Negative emotions were also directed inwards in the form of guilt (example 7), with participants feeling they had not done enough for their partner, regretting care decisions they had made on their behalf, or feeling guilty for surviving.
Some participants described having reached a point of acceptance in their grief, having reconciled with the death of their partner, and being 'at peace' with the permanence of the loss.They described a renewed sense of value in their life, alongside a feeling of gratitude ('lucky', 'glad') for their health, support networks, and the time they had had with their partner (example 8).However, denial was also described by some participants but varied in how it manifested.While some felt overwhelmed by the injustice ('unfair') of their partner's death, or 'confused' or 'disoriented' by the death, for others denial presented as grief avoidance, or feeling unable to begin to engage with the memory of their partner or their grief as it was too painful to bear (example 9).

iii. Support Structures -Keywords
The initial keyword search for terms related to support structures in bereavement identified 62 lemma forms including verbs, adjectives, adverbs and nouns (see Table 4).

iv. Support Structures -Findings
Participants provided detailed descriptions of their social networks  and sources of formal and informal support.Individuals described the value of frequent contact with friends and family, and established connections with networks and groups, which provided a vital distraction from grief (Table 5, example 1).For older participants, having people nearby who could help with practical matters was important, particularly as many within their social network were also aging and less able to provide help.For those with no close family, individuals valued having someone they could call upon in an emergency.Strategies to manage their own grief included keeping busy and filling their time with new activities, which for some fed into the self-perception of the need to 'get on with it' (example 2).Despite this, evenings and weekends were challenging, when isolation and loneliness were more pronounced, due to time spent alone at home.This forced the participant to oscillate between moving forward with life and returning to grief.
Participants described varied levels of need for, and access to, formal  support groups and counselling.A value of formal bereavement support was the chance to share experiences with others who would understand what they were going through, without further burdening friends and family who were also grieving (example 3).Conversely, others found that exposure to people's grief and hearing about their problems was unhelpful to them (example 4).Those who had accessed one-to-one counselling generally found it to be useful due to the opportunity to share openly and honestly their experiences.However, for some Fig. 2. Between loss and restoration: Liminality as an interstice between loss and restoration in experiences of grief.
K. Bristowe et al. participants, waiting times for formal bereavement support were prohibitive, or precluded access when it would have been most valuable.
Although not everyone felt they needed formal bereavement support, there was great variability in terms of what was offered.For those who did not receive any offers of support, this compounded their feelings of loneliness (example 5).

Moving beyond contemporary models of grief: the relationship between emotional responses and support structures in co-creating liminality for bereaved partners
The experiences described above (emotional responses and support structures) are accounted for by contemporary theories of grief (see Fig. 2).However, a number of our participants described experiences within their responses that fell outside of those recognised within the models.Some described feeling emotionally removed from their grief, something that is not as well recognised within grief theory beyond the context of trauma.Rather than an active denial or avoidance of the grief, this emotional removal was presented as an 'emptiness', feeling 'numb' and 'apathetic'.Participants described feeling 'flat' and 'desensitised' (Table 6, example 1), with a sense of being 'abandoned' and yet 'not being anywhere'.Rather than describing oscillating between loss and restoration orientation, as described in the Dual Process Model (Stroebe and Schut, 1999), they felt trapped in a liminal position, in an interstice (a narrow channel), between the two (example 2).This sense of liminality and interruption in their life was further elaborated by some participants in relation to being unable, or powerless to move forward or transition from the liminal state, because the future they had planned with their partner no longer existed (example 3).They could not move towards restoration orientation because their 'normal' could not be restored (example 4).This was also associated with a loss of identity.In their relationship they had felt valued, loved and needed, but now they were left with a lack of purpose and motivation.
A key concept related to liminality is that of 'communitas'.Participants within our study described the importance of solidarity (communitas) and sharing experiences with others in the same situation or liminal position (example 5 and 6).However, many also talked about the importance of being able to express their emotions and be honest with those in their broader social network, to invoke communitas, and to feel that their friends and family wanted to hear what they had to say.Refusal of family members to talk about the deceased partner, or expectations that the bereaved person should have returned to normal by now, resulted in further intrenchment of the liminal position, withdrawal and isolation (examples 7 and 8).The feeling of liminality was also exacerbated by seeing others return to normal life, contributing to feelings of being left behind (example 9).Lastly participants also described more spatial dimensions of the liminal experience, avoiding people and places that had been important to them pre-bereavement due to the distress it caused, resulting in further withdrawal from social networks (example 10).This liminal state, living between loss and restoration, is depicted in figure two.

Figurative expressions in the construction of liminality for bereaved partners
Several metaphorical vehicles were used to express the experiences of partner bereavement, which contribute to its construction as a liminal state.

i. Grief as a Monster
One of the most prominent vehicles used was that of grief as a monster or enemy.Indeed, the embodiment of liminality as a monster has been previously recognised.Culturally, when an abnormal experience is encountered, and it is not possible to identify it within one's own realm of normality, then what is being faced is a monstrous phenomenon (Giesen et al., 2018).Within the current study this was attended to in many ways, including grief as an agentic malign entity, or force, with many descriptions depicting the assault of grief upon the person.One such example was grief's ability to 'consume' individuals (see Table 7, example 1).The sense of grief as a sinister entity, or something to fear, was further underpinned by the descriptions of grief as unreal, like it is  'happening to someone else', like a 'dream you cannot wake up from' (example 2), leaving one 'vulnerable and scared to be outside after dark'.The metaphorical vehicle of grief as an enemy was further supported through the descriptions of how individuals viewed themselves since the bereavement.Participants described the injuries endured from grief leaving them less than whole; having lost part of themselves, experiencing a 'half life', and 'crumbling inside'.Participants also described the battle against grief as an enemy in physical terms.Being 'pushed down', having to 'push it down' and needing to 'reinvent oneself from the ground up' reinforced the sense of grief as an agentic entity to battle against and recover from (example 3).While having to 'push through' the grief suggests the ability of grief to trap or contain an individual.The representation of the liminal state of partner grief as an agentic being was also attended to in descriptions of the removal or shift of agency for the bereaved partner.Sometimes this was expressed in terms of vehicles of movement, with participants talking about being 'stuck', and unable to move, and describing their life as 'standing still' while others continued to live (example 4).For others this was expressed as moving, but not being in control of that movement, with descriptions of 'going through the motions' and the 'rollercoaster of emotions'.While some did express a sense of agency when 'letting go' of their partner, others described the powerlessness of being 'left alone'.
ii. Grief as an abyss Partner grief was also conceptualised as an abyss or black hole.This has parallels with the previous descriptions of liminality as a monster and something that 'hides its true identity and intentions behind a façade that cannot be trusted' (p64) (Giesen et al., 2018).In these data this was constructed with descriptions of a dark featureless space associated with 'disorientation' and 'helplessness'.Participants described the 'emptiness', 'hole', or 'void' that grief created.These vehicles contributed particularly to the description of the liminal position with participants being 'lost', 'disoriented', 'in limbo', or 'not being anywhere' (example 5).Participants also talked of the precarious nature of bereavement, forcing them to exist alongside the 'black hole' and the inherent risks this posed to their wellbeing.Grief was described variously as a 'burning hole', like being 'on a cliff edge', or at the 'edge of breakdown' (example 6).As well as grief being described itself as a void or empty vessel, participants also elaborated on this by describing the need to move their grief and memories to another space to protect them from the emotional impact (example 7).
iii.Grief as Water Another important figurative vehicle was that of grief as a liquid or water.As with the conceptualisations of grief as a monster or abyss, this conveyed a sense of the loss of agency, participants not being in control of their grief, and the risk this posed to the individual.This vehicle was used in the descriptions of the emotional impact of grief, with the pain characterised as 'deep', 'overwhelming', and as coming in 'waves' (example 8).Participants furthered descriptions of grief as water, fearing the risk of 'going under' (example 9).The metaphorical vehicle of water was also used to describe the sense of loss or abandonment characterised by grief, with support 'drifting away' from them over time.This description of grief as water, again has parallels with previous descriptions of liminality as a monster, characterised by deception, with a treacherous nature and representing danger beneath the surface (Giesen et al., 2018).

Main findings
This paper presents an analysis of the relationship between emotional responses to partner bereavement and social support.At 6-10 months post-bereavement, many participants continued to describe powerful and distressing emotional responses.While some had reached a point of acceptance and recovery, for many the experience was still characterised by a preoccupation with the deceased and an intense and unrelenting yearning, pain, anger, denial, loneliness and guilt.These emotional responses are well recognised in contemporary models of grief (Stroebe and Schut, 1999;Klass et al., 1996).They are also recognised in descriptions of prolonged grief disorder (Prigerson et al., 2021), which is characterised by persistent symptoms of grief, and functional impairment, despite the passage of time.However, some participants in our study described a different type of prolonged experience characterised by liminality, which is not recognised in contemporary grief models.Rather than an intense yearning they described a numbness, emptiness and a sense of 'not being anywhere'.Instead of a preoccupation with the deceased, they described being absorbed by the 'hole' or 'void' left behind.Whereas prolonged grief disorder is characterised, as the name would suggest, by a protracted period of intense grief, these participants described a prolonged liminal period, being beyond the intense grief but unable to access recovery.They felt trapped, unable to engage with either the loss or restoration, and unable to move forward as the planned future with their partner no longer existed.Drawing on the findings from our data, these differing trajectories of grief are plotted in Fig. 3.
The concept of liminality has been used to explain experiences of illness and death (MacDonald et al., 2021;Ghosh and Bk, 2022), particularly in cancer (Marshall et al., 2019;Little et al., 1998), however its relevance to bereavement is less well recognised outside of the context of sudden death (Carlsson et al., 2022).In relation to the present study, it has particular utility in understanding the mechanism of action for this type of prolonged experience, with participants describing 'not being anywhere' or, as defined by Turner, 'betwixt and between' (Turner, 1969).These data provide evidence of a liminal bereavement experience across multiple planes.Firstly, participants described being unable to engage with either grief or recovery, and feeling trapped and lost between them.Secondly, they described being unable to move forwards, because a central part of their identity and the future they had expected with their partner could no longer exist.Thirdly, participants described spatial liminality (Thomassen et al., 2018), feeling unable to return to people and places they visited pre-bereavement, resulting in further isolation and withdrawal.Fourthly, our participants described the importance of 'communitas' and sharing their experiences and emotions with others.When this was not encouraged or reciprocated individuals chose to conceal their emotions to avoid distressing others.This resulted in further entrenchment of the liminal experience, social withdrawal and isolation.The figurative expressions utilised by participants further supported the conceptualisation of the liminal state.Metaphorical vehicles of a monster, an abyss and water collectively contributed to the characterisation of the liminal bereavement experience as a dark agentic force that wounds, overwhelms and traps the individual leaving them powerless, lost and alone.
The findings of this study offer original and important insights into the experience of partner bereavement, and have implications for contemporary models of grief.The description of the liminal bereavement experience adds new interpretations to the work we have conducted with LGBT + bereaved partners, and the Acceptance Disclosure Model of bereavement (Bristowe et al., 2016(Bristowe et al., , 2023)).The model posits that experiences in bereavement are shaped by relational aspects -the extent to which you feel able to disclose your bereavement to those around you, and whether that is accepted and acknowledged by those individuals.Participants in the present study described having to conceal their emotions, and hide their grief, in order to engage with others.They described feeling that their grief was no longer acknowledged or accepted, and they were left isolated as support drifted away (see Fig. 2).Applying the concept of liminality to the Acceptance Disclosure Model may provide important nuanced understanding to marginalised bereavement experiences.
The findings of this study also have implications for outdated but still widespread models of grief and bereavement, which may fail to represent the experiences of some individuals who have been bereaved of a primary relationship.The characteristics of the liminal bereavement experience are not recognised in models which suggest that in grief one will reach a stage or recovery or adjustment, or that highlight the importance of acceptance, adjusting to the new reality and moving on from the bereavement (Kübler-Ross, 1969;Worden, 1996).The implicit assumption that there is one right way to grieve, may contribute to a lack of understanding and grief literacy.These assumptions are unhelpful for those seeking to support others, and set up unrealistic and potentially damaging expectations for bereaved individuals themselves (Stroebe et al., 2017;Cruse Bereavement Support).
This work also advances understandings of bereavement at the intersection between two other seminal grief theories: the Dual Process Model (Stroebe and Schut, 1999) and Continuing Bonds (Klass et al., 1996).Our model (see Fig. 2), Between Loss and Restoration, takes forward the Dual Process Model by suggesting a third state (liminality), between grief orientation and restoration orientation, which an individual may occupy.Our work also adds important interpretations to the Continuing Bonds model of grief (Klass et al., 1996).In the context of the loss of a primary relationship, the depth of the relationship may preclude recovery, because of the centrality of the relationship to one's identity.The nature of the loss interrupts an individual's ability to engage in grief or recovery, and is characterised by a loss of self, identity, future and purpose.Although our data draws on experiences of bereaved partners or spouses only, these findings have relevance for the loss of other significant relationships too (such as a sibling or child) where the death dramatically changes the course of the surviving individual's identity and future.Lastly, our findings provide important new insights into the prolonged or complicated grief experience, which may not necessarily present as intense grief, but as a liminal state.Inclusion and recognition of liminality within the spectrum of bereavement experiences is vital.Through its inclusion in theory, and professional and public facing discourse around grief and bereavement, this experience of liminality can be normalised and better understood.This could in turn improve formal and informal approaches to offering bereavement support and societal grief literacy, and target support and resources to individuals experiencing liminality in their grief.
The exploration of figurative expressions of partner bereavement in the present study also represents an important contribution.Although use of metaphor has been explored extensively in cancer experiences (Semino et al., 2017), and in specific cases of bereavement (such as the death of a child (Hooghe et al., 2012)), their use in other significant bereavements has received limited attention.In the present study figurative language was central to the construction of the liminal bereavement experience.As such, paying attention to the use of figurative language may offer important insights for friends, family and professionals alike into how individuals are managing in their bereavement.

Strengths and limitations
Incorporating free-text data within surveys offers many benefitsits inclusion complements the findings of, and addresses the implicit limitations of, the quantitative data.A strength of this paper is the focus on the free-text data alone using a novel method for this type of dataset, which has allowed for a detailed and rigorous examinationmaximising the use of the data provided by participants is an ethical imperative for all research.Limitations of the study include that those who provide free text responses to questions may represent a particular subset of participants.There were slight differences in the demographic characteristics of those who provided free-text responses compared to those who did not.However there may have been other differences, for example those with higher levels of literacy, with English as a first language, or more motivation to expand on their responses, limiting the transferability of the findings.In addition, all participants were in a legally recognised relationship (civil partner or spouse), therefore it could be argued that they may be afforded more benefits in their bereavement experience and access to support than others.As such, the experiences of those in more marginalised positions may not be represented here.At the same time, as 43.4 % of our participants were bereaved of a same gender partner, we may have included more participants in marginalised positions than would be found in the general population.

Conclusions
The findings of this study offer significant and original insights into experiences and trajectories of bereavement, particularly in the context of partner loss, and our understandings of prolonged or complicated grief.Although useful in describing many experiences of bereavement, contemporary models of grief fall short of accounting for the liminality some individuals experience after the loss of a primary relationship.Including liminal experiences into public and professional discourse of grief and bereavement, could help to normalise these experiences and enhance grief literacy.This could in turn improve approaches to offering formal and informal bereavement support, and target support and resources to individuals experiencing liminality in their grief.

K
.Bristowe et al.

Table 2
Taxonomy of emotional responses.

Table 4
Taxonomy of support.

Table 6
Liminality in partner bereavementexample quotes.
1'I live in the moment when in company, in the past when on my own and do not look forward, my feelings seem 'flat', desensitised.'(Participant F4086) 2 'I have two children and two grand-daughters who are very loving and supportive but my life still seems empty without my husband.I often feel as if I am going through the necessary motions of life.' (Participant E5937) 3 'Ever since losing my husband and soulmate I have lost, robbed of the life we should have been sharing together, as he was about to semi retire.As we had only been married for 15yrs and together 16, we should have had lots to look forward to, holidays, days out and going out on our motorbike to see the sights.He 'Feel life very lonely and not worth going on.Have great friends and family but feel isolation, alone and a burden.Worse now six months on, in first three months you are busy sorting things but then all stops, people go back to normal life, but not for me.Always did things together, now feel vulnerable and not needed.'(Participant D1419) 10 'Everywhere I go and everything I do evokes a memory of times and experiences shared.This can have the effect of keeping me away from places and people that I know will upset me.' (Participant F7786)

Table 7
Figurative Expressions of Liminal Partner Bereavement -Example quotes.
and cannot look at them.It is too heart breaking.I don't want to remember things yet.I avoid thinking of her or imagining her face or conversations.' (Participant C2713) Abyss 8 'The pain of her loss comes in waves which are overwhelming.'(Participant D4327) Water 9 'I know he is dead.Is not coming back and I can't let him down by going under.I can't, and don't want to, erase 57 years but I had been with him all my adult life and I miss him.'(Participant E7022).Water K. Bristowe et al.