Retracing participants in longitudinal studies: Trekking the timescape of fieldwork

ABSTRACT Retaining participants in longitudinal studies is important but challenging when retracing them after a substantial gap between study phases. Utilising our nine year mixed-methods longitudinal study of 54 families affected by paternal imprisonment, we qualitatively analysed our processes and experiences of retracing participants after a seven year gap in contact. Paying attention to the spatio-temporal features that are traditionally an analytical focus of qualitative longitudinal research helped to enhance our retention rates, and our fieldwork strategies can be more deeply understood as part of the ‘timescape’ of qualitative longitudinal research. In this paper, we describe the spatio-temporal contours that shaped our efforts, successes and failures, and distil the material practices that they led to during our ‘tracking treks’ through the timescape of longitudinal fieldwork.

The literature concerning participant retention in quantitatively orientated studies has coalesced into a fairly consistent set of strategies that yield high retention rates. Scott (2004) refers to this approach as 'a bag of tracking tricks' (p. 21). The word 'tricks' underplays the great practical use of such advice in the craft of longitudinal research, but such approaches can be greatly complemented by overarching frameworks to apply responsively to different circumstances, and simply deploying a greater number of 'tricks' is not related to higher retention rates (Teague et al., 2018). This lack of a straightforward linear relationship is unsurprising given the complex social setting of participantresearcher interactions. Lofland, et al. (2006, p. 121) describe the organisation of social settings as 'the intersection of one or more actors engaging in one or more activities at a particular time in a specific place'. Quantitatively orientated participant retention literature is rich in analysis of actors (participants) and activities (retention strategies), but generally has less to say about the times and places (or spaces) relevant to participant retention. However, the field of qualitative longitudinal research (QLR) engages deeply with the spatio-temporal features of participants' lives (see Thomson & McLeod, 2015), and a core principle of QLR is that it 'designs in' temporality as an analytical consideration . Some authors have extended this approach to reflect on the implications of temporality for the longitudinal research process itself (see, Neale & Flowerdew, 2003;Taylor, 2015;. We add to that body of work by describing (spatio-) temporal considerations that related particularly to participant retention in the last phase of our mixed-methods longitudinal study, which involved retracing participants after a seven-year gap in contact. We describe how the material realities of our fieldwork could be understood in relation to deeper organising principles of space and time, and how harnessing these principles helped us maintain an adequate retention rate despite the contact gap.
This paper is divided into four sections. Firstly, we describe quantitatively orientated approaches to participant retention, and suggest how spatio-temporal insights from QLR can be applied to existing findings. Secondly, we summarise our retracing methods and retention outcomes. Thirdly, we present a qualitative analysis of our fieldwork which describes the spatio-temporal themes that shaped our efforts, successes and failures. Lastly, we discuss our findings in light of Neale's (2020) conceptualisation of the timescape of longitudinal studies, and describe our distilled set of 'tricks' that were born from our increased sensitisation to the contours of the broader 'treks' we undertook through the timescape of retracing participants in longitudinal fieldwork.

Time and space as analytical concepts
In their conceptual exploration of time in qualitative longitudinal research, Neale (2020) maps its timescape on five planes. Prospective-retrospective time concerns people's orientations towards past (looking back) and future (looking forward). Intensive-extensive time concerns the framing of time in terms of its duration, tempo and rhythms. For example, constructing hours, seasons and years (Durkheim, 1995) and linear versus cyclical time trajectories. Micro-macro time distinguishes 'a close-up vision of individual lives, or a wide-angled view of social or historical processes' (Neale, 2020, p. 31). Continuous-discontinuous time describes the cross-cutting between different forms of temporality, and how people reconcile those forms in their lives. Time-space recognises that spatial dimensions significantly shape how time is experienced. Within all of these planes lies a distinction between experiencing time in a fixed or fluid manner (Adam, 1990). Fixed time flows inexorably (or in fixed repetitive cycles), and its structures are externally, socially, or institutionally imposed on individuals. Fluid time is more subjectively experienced by individuals, is contextual, relational and multi-faceted. Lastly, within the flow of time lie a series of fateful moments (Giddens, 1991) or turning points (Neale & Flowerdew, 2003). We draw upon these features to understand the timescape of our longitudinal study.
Although space is inescapably integrated with time (see, May & Thrift, 2001), we also engage with ideas of space in its own right, because of how significantly it shaped our experiences of retracing participants. Harvey (2006) and Löw (2001) both propose a tripartite conceptualisation of space as absolute, relative and relational. Absolute space refers to physical space created by objects. Relative space is the order and structures between objects and people. Relational space is an elaboration of both: it is space physically and symbolically created by interactions between people, objects and environment, which create meanings and conventions around, for example, how to behave, and who does or does not 'belong'. These elements are interlinked and reinforcing, with people acting as social agents whose actions shape space, but are also shaped by space. Globalisation and technology have de-naturalised and compressed space (Harvey, 1990) in such a way that relational spaces now exist which are not rooted in physical locations (for example, social media networks). We use this broad conception of relational space to inform our discussion.

Approaches to longitudinal participant retention
One way that longitudinal retention has been studied is identifying characteristics of participants who are particularly likely to drop out, and often described as 'hard to reach'. These characteristics are mostly economic and marginalisation related; for example homelessness, low socioeconomic status, membership of an ethnic minority group, unemployment, low income, mental health and/or substance abuse problems (Badawi et al., 1999;Cotter et al., 2005;Fischer et al., 2001;Mitchell et al., 2007). However, studies in which all participants share a 'hard to reach' characteristic have still achieved retention rates of 80%+ (e.g. Coen et al., 1996;Cottler et al., 1996;Western et al., 2016). The 'hard to reach' participant is therefore relative to 'easy to reach' participants, to the retention strategies employed, and to the point in time that they are involved, as their 'hard to reach' characteristics are unlikely to be static (e.g. Williamson et al., 2014). This approach does not fully articulate the relationships between methods and participants that may lead to differential outcomes. Spatio-temporal features may contribute to the difficulty of 'reach' and connection when researchers and participants are moving in very different spaces, ill-at-ease when navigating each others' spaces, subject to different rhythms and tempos of time, and experience the flow of time differently in relation to the study. Long gaps in contact are likely to enhance these differential flows and movements, Retention literature has also considered research staff. Characteristics associated with higher retention are persistence, creativity, flexibility, comprehensiveness, sociability, credibility, good team working and cultural knowledge/matching (Armistead et al., 2004;Butler et al., 2013;Cotter et al., 2002;Estrada et al., 2014;Mitchell et al., 2007). Abshire et al. (2017) noted that researchers from studies with 80%+ retention rates reported frequent variation of their strategies. Paying attention to how space and time shape researcher decision-making may be useful for understanding retention success and failure contextually, rather than as linked to a list of desirable researcher qualities that translate smoothly into success.
In their narrative review of strategies associated with high retention rates, Ribisl et al. (1996) list: gathering comprehensive information; establishing relationships with public agencies; creating a project identity; emphasising the importance of tracking to staff; using the simplest/cheapest tracking methods first; making research convenient and rewarding for participants; expending greatest tracking efforts at initial follow-up periods; and customising tracking efforts to each individual. In their later systematic review and meta-analysis, Teague et al. (2018) noted a diversification of strategies in recent years, identifying 44 strategies new to the literature. Extending these overarching reviews are several papers that richly describe fieldwork experiences, reflecting on social factors such as power, privilege and culture (e.g. Leibrich, 1994;Miller, 2015;Sharpe, 2017). Spatio-temporal influences are discernible in these reviews. For example, successful retention seems to occur when researchers become social actors in a participant's 'space', and when relationships with participants are sustained over time. However, there are complexities: Sharpe (2017) notes that too much intrusiveness into space can replicate the feeling of surveillance that is experienced by some marginalised groups, and Miller (2015) describes the problems of 'going back' to a topic after a time period during which participants may have rewritten their personal biographies. The more recent diversification of strategies is largely driven by digital methods, which represent the 'time-space compression' of globalisation (Harvey, 1990).
These lists of strategies, while useful, can be complemented by being placed within a holistic understanding of the timescape of participant retention. In their analysis of the relationship between 'time' and 'texture' of people's social lives, Neale and Flowerdw (2003) argue that time can be used to grasp mechanisms and strategies used by individuals to manage change. We now turn towards our fieldwork experiences of retracing participants after a seven-year gap, and outline both our strategies, and how these can be understood in relation to the time and texture of the lives of participants and researchers.

The families and imprisonment research (FAIR) study
The FAIR study 1 (see, Lanskey et al., 2016Lanskey et al., , 2015Lösel et al., 2012;Markson et al., 2015;Souza et al., 2021Souza et al., , 2015 was a prospective longitudinal study of paternal imprisonment, release and resettlement for families, combining interviews and standardised measures to explore the impact of imprisonment on post-release outcomes, and identify factors underlying familial resilience to the adversity of imprisonment. Fifty-four fathers in English prisons, their 54 female (ex-)partners and 89 children were recruited in 2009. Participants were interviewed up to four months prior to the father's release from prison (Time 1), up to six months after his release (Time 2), and a third time between seven and nine years after the baseline interview (Time 3), provided they had agreed to be contacted again.
At Time 3, methods used to locate participants were: contact details provided at Time 2 (for themselves and for others that they said would be likely to know their whereabouts), scrutinising previous data for leads, directory enquiries, open electoral roll data, social media, business listings, people known to participants (e.g. relatives, friends, caseworkers), local community enquiries, prison records, newspaper articles, and general internet searching. Methods used to contact participants included: phone, letter, email, social media, and face-to-face visits. These methods were also used to contact gatekeepers who could potentially put us in touch with participants (except social media). We used several 'best practice' methods, including creating a study identity via a logo, holding regular team meetings, keeping comprehensive records, having flexible working hours, making adjustments for participant convenience, building relationships, and using reminders.
As in other studies, we faced ethical challenges created by technological changes since the last study wave. High contact insecurity meant that prior contact details yielded few successful contacts.
In keeping with the current debate on digital methods (e.g. see, Bennetts et al., 2021;Hibbin et al., 2018), we prioritised protection from harm. We rejected overly rigid notions of, for example, treating all social media data as private data, but used it as safely as possible. For example, we only messaged individuals whose profiles met multiple identification markers that matched our participants.
We also weighed the risks of harm that may come from approaching people who were likely to know the participant, against the benefits of the help gained. Our agreed protocol was to describe ourselves as researchers whom the participant had formerly been involved with, in a study of families. We reasoned that this description was broad, minimal, and unlikely to be socially compromising. We sought feedback about this method from our participants as we went along, to gauge ground-up perceptions of harm and confidentiality (discussed further in Findings), and found participants were generally happy with our approach.
Quoted figures refer to the adult participants only, as the children were always traced via their parents. We located 96.3% of participants (n = 104). Of these, three were deceased. We contacted 95.0% of contactable participants (n = 96). Of the contacted participants, 70.8% were retained (n = 68), comprising 63.0% of all the original adults in the study (n = 108), and 75.9% (n = 41) of the original 54 families (at least one member retained). This compares adequately with other studies of adults affected by the criminal justice system (e.g. Farrall et al., 2016: 52% over 15 years, Bottoms & Shapland, 2011: 78% over three years).

Analytical methods
This paper explores retention processes at Time 3 only, due to the time gap and team change since Time 2. We originally attempted to explore relationships between methods and outcomes quantitatively. However, we found that it was not meaningful to classify a single method as successful for each person, as multiple methods contributed, and our dataset indicated 54 unique pathways to contact. This seemed to reflect the increasing diversity of strategies, and suggested to us that a qualitative approach was more meaningful.
For analysis, we constructed 108 vignettes describing the retention process for each participant, which combined information from our detailed meeting/field notes. We adopted an inductive thematic approach to our analysis, progressively focussing and refining through two rounds of coding (see, Miles & Huberman, 1994). We began with the broad question: what themes characterised our fieldwork experiences, and were there patterns that were systematically related to success or failure? We first coded in a fairly open manner, for descriptions, processes and methods. We noticed quickly that success or failure were not only related to what we did, but to the totality of where, when, and how researcher-participant encounters occurred. Therefore, in the second cycle of coding we adopted a more focused strategy of identifying and describing the contexts of fieldwork, eventually identifying space and time as constructs which were relevant in the majority of the 108 cases, and identifying and refining the themes that best described their role, through team discussion.

The role of space
Three spatial factors emerged as important in the retracing process. Firstly, types of space mattered; specifically physical space (houses, streets, towns, communal areas); social space (family, friends, communities) and digital space (social/local media). Secondly, space shaped participants' visibility. Thirdly, the ease of navigation in spaces shaped researcher decision making.

Physical space
At Time 3, we found that 83.9% of participants (n = 78) who were alive and not imprisoned still lived in the same area as Time 2, but only 28.4% (n = 27) lived at the same address. People moved often, but rarely far. 'Contact insecurity' was therefore only insecure when space was conceptualised as a single property, rather than a spatial area. Studying participants' local area helped us identify physical spaces where a participant might be visible, even if they had moved residence: We visited Joshua's 2 last address, but there was no answer. The property was adjacent to the high street. We knew Joshua had a history of addiction. We reasoned that he might be visible in places close by that facilitated this. Along the high street we identified a pub, three off-licences and two bookmakers. Joshua turned out to be well known in one of the off-licences and we were told that he had recently returned to prison (we did not mention prison) but his mother still lived locally.
Prior local knowledge was important for navigating physical space. Researchers could navigate more easily and better identify possible information sources. The advantage of ease in the field has been described elsewhere as important for social decision-making (Bourdieu & Wacquant, 1992). Accessibility was also important. For example, in suburban residential areas people were generally not home during the day. Such areas often required extended or repeated visits. Conversely, active communal areas were helpful for locating participants. Many who were not visible at a residence were visible in communal spaces (e.g. shops, pubs, sports areas), and communities within such spaces were often more helpful than in highly residential areas. These experiences frequently led us to evaluate likelihood of success by the characteristics of the space we visited, as well as the participant's characteristics.
Other structural and agentic factors interacted with space to shape retracing success. Sometimes, despite helpful spatial factors, structural factors made locating people more difficult: At Time 2, Mel and her son Kevin lived on a high-rise estate in a London borough. People were friendly, and there was a communal space where a group of young men were playing basketball. They knew Kevin by 'reputation' and one thought they might 'know people who knew him'. He made enquiries but concluded that everyone had lost touch with Kevin. We later found Mel and Kevin through social media. In the period between Time 2 and Time 3 they had been rehoused by the council several times to different areas of London.
On other occasions, participant agency meant that spatial factors were less relevant: Angela lived in a small town. We identified the area where she was living, but could not find an address. There were plenty of communal spaces in the form of small independent shops and cafes. People were very keen to help. A café owner spontaneously texted several friends to ask if they knew Angela. However, Angela was invisible, and when we did find her (via her ex-partner) she was very socially isolated.
There were also complex interplays between spatial characteristics, agency, and perceived researcher identities: We thought Steve might be living in a small village in an area where one of the researchers grew up. We visited the pub and made enquiries. Locals confirmed that Steve had worked there as a bouncer a year ago. The researcher was asked lots of questions and people assumed she was a police officer. Statements and ID to the contrary were met with laughter and scepticism, however the researcher's local identity was accepted. Eventually two men made phone calls to their wives, assuring us that they 'knew everybody'. They confirmed that Steve had not been seen in the area for a long time and most likely had moved on. This was in fact true. Tilly (2004) describes this kind of exchange as a 'sizing up' process, where the interaction of individuals from different social networks in the same space leads to the formation of social boundaries. Strangers are judged as outsiders or insiders depending on notions of who is (not) a member of a group in that space. In this case, the interplay between a potential 'outsider' identity as a police officer, 'insider' identity as a local, and unfamiliar identity as a researcher required a delicate negotiation, influenced by the nature of the space (a small rural pub).

Social space
It was important to build understanding of a participant's social space. Participants often remained in the same area due to family ties, and it was family members who ultimately made the participant visible by giving updated contact details. Often, participants' parents had higher contact security than the participants themselves. However, when this wasn't the case, re-establishing contact involved careful integration of social and spatial information: However, visibility in social space was not guaranteed, even if we were able to identify a social network. People in that network acted as gatekeepers and exercised agency in whether to help us. Fifty-seven percent of all participants (n = 61) were located due to help from a gatekeeper, and their presence was statistically significantly related to greater retention likelihood (χ(1) = 7.535, p = .012). The gatekeepers we encountered predominantly did offer help (88.2%). Often, they provided information after our other efforts had failed, acting as a crucial turning point towards success, particularly in spaces that were difficult to navigate: Douglas lived in London at Time 1. At Time 3 he was digitally visible in a medium sized Scottish town. However, we could not find an address. We identified his mother's address, and travelled to Scotland to contact her. She spoke to Douglas for us, arranged an interview, and helped us get to the remote village where he now lived.
Unhelpful gatekeepers were rare, but did exist, and had impact. All four unlocated participants had an unhelpful gatekeeper. When it did occur, types of unhelpfulness included agreeing to help and then becoming unresponsive (n = 5); acting as a liaison point instead of facilitating direct contact (n = 4); declining on the participant's behalf (n = 3); in the case of prison staff, procedural delays (n = 1); giving conflicting information at different times (n = 1) and withholding information due to suspicion (n = 1). This often led to dead ends: Gina lived in a suburban area at Time 2, but no longer lived there at Time 3. She and her ex-partner had separated since Time 2. We visited her parents' address, in a suburban area of another town. Her father declined on her behalf and refused to pass on a letter to her. We did not locate Gina.
This was a 'doorstep encounter': doorsteps were a very important space to navigate socially. Establishing 'doorstep credibility' worked in some complex ways. Sometimes it was evidently linked to our institutional credentials, but more often our credibility as 'people who care' was most relevant: Paulette was invisible to us through the initial methods we tried. Eventually we received a suggestion of a family friend who might help us. We visited her. We knew that Paulette's ex-partner had been a refugee from a natural disaster. The visiting researcher was very interested in this event and took time to learn more about it. The family friend turned out to be a refugee from the same disaster, and once she realised the researcher's knowledge/interest, there was a long afternoon of conversation and food. At the end, she called Paulette for us. When Paulette agreed to speak to us, the family friend was keen for the interview to be conducted with the visiting researcher specifically, not with another member of the team. She commented to Paulette, 'This one is good, she knows and understands. You should talk to this one.' In this example, space, credibility and relations within space interacted to produce a result. This encounter may have gone very differently if conducted via phone call, for example. We learned to avoid cues that risked misidentification as somebody 'official' (e.g. clothing, a notebook), as many of our participants had had difficult experiences with the state and local authorities. Relatability was helped by genuineness, humility and warmth, which placed us as somebody respectfully requesting help, and communicated an enthusiasm for what we were trying to do. This sometimes led to navigation of space that extended beyond the doorstep: We visited Robert and he asked for time to think about whether to take part, initially seeming ambivalent. Over a few weeks he exchanged a series of text messages and phone calls with one of the researchers, asking about the study but also chatting about his life. He disclosed a wish to 'give back' to people experiencing difficulty and seemed to relate to the researchers as people also wanting to do social good.

Digital space
Digital space comprised directory enquiries, the open electoral register, Companies House, social media and news articles. Of the participants that we searched for digitally, 79.3% were digitally visible (n = 65). Those who were not were slightly older (M = 42.1, SD = 9.0 years vs 36.7, SD = 8.5 years, t (80) = 2.304, p = .024), or lacked nominal visibility (i.e. being able to use their name to identify them). This was a factor for three of the four participants we failed to locate. For example: There was no trace of Alain at Time 3. His name was very common in two large Western countries (one in which he grew up), and his African country of origin. Despite having detailed information about him, the number of digital sources matching his name and details was too large to be able to distinguish him.

Navigation of digital space was relatively simple and sometimes revealed a vast amount of information:
Jade was very publicly visible due to some fundraising that had attracted attention nationally. The press coverage she received contained enough information to locate the town, area and building that she lived in. She was well-known locally, and members of the community directed us to her front door.
However, digital space was difficult to navigate ethically and we typically found ourselves questioning whether we should make use of information. Participants had given us permission to locate them again in multiple ways and made information about themselves publicly visible. However, this permission was given at a time when the digital landscape was very different. Using social media in research is still very much debated in the literature, although it is generally agreed that protection from harm should be an overarching consideration (e.g. see, Hibbin et al., 2018). To monitor these issues, we sought feedback from a selection of our participants about how we located them. Their responses mostly demonstrated unconcern about their digital visibility, and they did not find social media messages any more intrusive than letters or phone calls. However, there were occasions where we decided not to utilise digital information because we felt a participant may be distressed if we acted on it. . We found Joseph on Facebook and messaged him, but received no reply. His ex-partner volunteered to speak to him for us, but each time we followed up she was evasive, stating she hadn't spoken to him yet. We decided this may be passive non-consent. On Facebook, Joseph's profile showed only his picture. However, one day he used Facebook marketplace to sell an item, and this revealed his postcode. Although this gave us another means of contacting him, we chose not to. Navigation of digital space was, therefore, the easiest form of navigation, but the most challenging ethically. To add to this challenge, we found that we could not rely entirely on our own cultural presumptions about harm/intrusion and communication methods. For example, some participants reported ignoring post because it was associated with debt, so letters in fact triggered more stress than digital contact methods, which we initially assumed might be more intrusive. However, the meanings that people attach to forms of communication are individualised, and notions of harm need to be carefully conceptualised with this in mind.

The role of time
For participants, two temporal trajectories were relevant: their 'life trajectory', and their 'everyday' routines. For researchers, their time spent tracing participants comprised both a linear trajectory and repetitive cycles, comprising a series of repetitive actions which gradually produced results, then tailed off. Fluid perceptions of time were important, including researchers' own experience of time spent tracing participants, and participants' perceptions of researchers' time and effort.

Participants: Daily routines and life trajectories
Recognising participants' everyday routines was important for identifying opportune times to make contact. The 'right' and 'wrong' times were highly related to the time of day, week and year. For example, fieldwork was often more successful in the summer, as people were more likely to be outside and sociable: We knocked at Tyler's last address but received no reply, nor did any neighbours answer their doors. We were getting back in the car when a neighbour appeared, and enquired if we needed help. He confirmed that Tyler and his family did live at the address but neither they nor any other house was likely to answer due to gang and police activity. We returned on a weekend day that was much sunnier. This time, the front door was open and there was a lot of noise and the smell of good food. We were greeted by Tyler's sister, who got Tyler to come out and speak to us. He agreed to be interviewed again and suggested meeting the next day, and conduct the interview outside in the good weather, as the house was very crowded.
We observed that people were less inclined to engage in a chat in poor weather, or in twilight hours. Fieldwork during working hours, commute times, and school pick-up were generally less successful, but this varied according to participants' individual routines. While exact circumstances were rarely possible to predict, routines could be intelligently guessed at, and strategies adapted accordingly. This ability to slot fieldwork rhythms into the temporal rhythms of people's lives is noted by Neale (2021) as important, as time is also a form of finite resource which makes people more or less available/willing to devote that resource to a study.
We also found that the intersection of the study with important life events was relevant for retention. Some participants were contacted at times of personal upheaval, which meant they were disinclined to engage and highly focused on the present. We found that keeping in touch with these participants periodically was helpful in two ways. Firstly, it allowed a continued channel of communication and recognition of when they were in a more stable position to participate. Secondly, it built relationships over time, which in itself increased engagement with the study. . When we spoke to Ashley, she volunteered to take part and booked an interview twice, but cancelled both. When we asked if she still wanted to take part, she explained that she was in the process of moving out of temporary housing, and was expecting a baby shortly. She sounded very weary. After seven months of intermittent contact, Ashley said she was in a much better place to take part. She had moved into permanent housing and the birth had gone well. She sounded much better.
However, intersection of the study with life events could negatively influence retention, when events were intimately connected to the study topic: Dean enthusiastically agreed to take part again, talking about how far he had come since his release from prison, and how proud he was of being a father to his son, who lived with him. He saw the study as a good opportunity to reflect on his achievements. An interview was booked. The day before, Dean called to cancel. He was deeply upset because his son had chosen to return to live with his mother. He asked for some time to cope with this. We tried to contact him periodically over the following months but he didn't re-engage with the study.
Ashley and Dean both had some characteristics of 'hard to reach' participant groups, and lacked others. But their characteristics were of far less relevance than their time-bound circumstances.
As well as being influenced by life events, retention was influenced by generalised orientations towards the topics of 'prison' and 'family' that had changed slowly over time; for example, fathers distancing themselves from their criminal past: Christina agreed to take part and to ask her partner Cameron to do so. Upon arrival, Cameron was present but acted in a suspicious and dismissive manner. He declined to take part. He had not been back to prison since Time 2 and the sentence was a 'family secret' kept from the children. After Christina's interview, she told us in a follow-up phone call that Cameron had changed his mind about taking part. He initially didn't want to discuss the past, but he saw her interview had been 'not that bad' and that she'd enjoyed it.
Cameron's response was typical (but somewhat more positive) of fathers who had never returned to prison. We found that we were statistically less likely to retain fathers who had avoided returning to prison (χ (1) = 6.041, p = .020). This is due in part to imprisoned fathers being easier to trace through institutional records. But additionally, families for whom prison had become a distant (but still painful) memory expressed greater reluctance to take part. Similarly, when family relationships had broken down, participants were sometimes more reluctant to engage, although this was far truer of fathers who had lost contact with their children, not just when parents had separated. When the study topics were conceptualised as part of a distant, painful past, or bleak future, retention was more difficult than when conceptualised as part of the recent past or present, or when a stable present and hopeful future eased the pain of talking about past events.

Researchers: Linear, cyclical and fluid time
The length of time tracing participants influenced retention: 51.5% of successfully retained participants, required more than two field visits. Persistently sifting large amounts of information, retrying methods, and trying creative new methods were important. The example below illustrates what this sometimes involved: Javel was digitally visible until four years previously, but initially untraceable. Weeks were spent building up a comprehensive picture of him. Contact efforts included phone calls using the phone book of his country of origin (conducted late at night, due to the time difference), contacting his former Probation Officer, visiting prior addresses, visiting churches in his last known area of residence, and contacting a running club he once regularly attended. None of these efforts led anywhere. After comprehensively reading social media posts, an individual was located who we thought might be Javel's brother. His brother was eventually located living under his wife's surname. We visited, and he offered to help, but now only had a distant relationship with Javel. However, he gave us contact details for their mother, a family friend and another brother, through whom we were eventually able to contact Javel. However, within this progress over time, there were repeated peaks and troughs in tracking activities. Often, taking a break would precede a new breakthrough in cases that previously felt hopeless, which revived motivation, and fostered persistence in the long term. Researchers' responses to the time it took to track participants also affected their activities. For example, when the tracking process felt intrinsically interesting, when researchers had a high level of autonomy over their time, and when jointly conducting fieldwork, long field periods felt less 'heavy' and more enjoyable, resulting in more persistence and creativity, and protecting against burnout.
Perceptions of the time that we took (e.g. travelling a long way) were often interpreted by participants as a signal of their value to the study. When describing their reasons for participating, they made comments such as, 'it must be really important if you did that much to get in touch with me'. Time and the accompanying effort therefore, served a social function: it was not just actual time but perception of time that mattered. The simplest/cheapest tracking methods (such as sending a letter or a social media message) seemed more efficient, but they resulted in a lot of nonresponses, whereas the more time-consuming, and effortful methods received a response more often. It may be that perceived time/effort activated reciprocal norms in a way that cheap/ efficient methods did not (for further discussion of reciprocal norms in participant retention, see, Estrada et al., 2014).

Discussion
The FAIR Study sample possessed characteristics associated with reduced retention. Yet almost all participants were retraced, and almost two-thirds were retained after a 7-9 year contact gap. We followed best practice strategies for longitudinal retention and found these helpful. Yet analysing our processes from a spatio-temporal perspective added highly useful insights into what enhanced or constrained those strategies. Our fieldwork was shaped by relational space. The interlinked domains of physical, digital and social space shaped the likelihood of finding participants, and enabled/constrained researchers' ability to navigate within them. Participant retention activities 'flowed' in space, shaped by the physical and representational structures we encountered. Researchers had to fit smoothly into people's temporal routines to get best results. Both long and short temporal horizons related to study engagement, as did changing biographical orientations to prison and family. Researchers' fluid time experiences significantly shaped their activities. Digital space-time compression has vastly accelerated what researchers can achieve in tracking participants, but this does not translate simply into efficiency. Rather, participants' perceptions of what researchers do with their tracking time matter, as do the ethical sensitivities of navigating spatial information.

Trekking the timescape: An integrated temporal-spatial approach to participant retracing and retention
The main implication of our analysis is that understanding and harnessing a study's timescape is a useful approach to retaining participants in longitudinal studies even when they are not explicitly QLR orientated. Our study took a mixed-methods approach and had a moderately largely sample (or at least, larger than many qualitative studies). Yet a spatio-temporal approach not only enhanced our methods, it was a method in and of itself, and yielded satisfactory follow-up rates even in a so-called 'hard-to-reach' group. In the remaining discussion, we contextualise our findings within Neale's (2020) five temporal planes described at the beginning of the paper, and use them as a basis to suggest a set of pragmatic, but theoretically informed steps that longitudinal researchers can 'design in' to their retracing efforts. We view these steps as close neighbours of the 'tracking tricks' approach, but complemented by broader cognizance of the contours of the 'tracking trek' through the timescape of a study.

Prospective-retrospective time
As in other studies (e.g. Miller, 2015), we found that participants' orientations to the study topics (family and prison) were situated within their personal constructions of their biographies, and influenced their decisions to participate. Biographical orientations are ultimately something that researchers have little control over. However it does imply, as others have also suggested (Neale, 2013;, that consent can be treated as time-contingent. Participants' orientations shifted both between phases of the study, and within the most recent phase. For example if a family member suddenly became part of a painful past rather than an imagined future, or if prison abruptly returned to the present, participants often changed their minds about taking part (in either direction).

Intensive-extensive time
Fieldwork is an immersive process, and we found it gave rise to intensive temporal experiences. Within the fixed time-cycle of the project, experiences of fluid time varied dramatically. Time proceeded faster or slower depending upon experiences of 'turning points' towards success, and peaks and troughs in motivation. We found that harnessing methods that bowed to time's more personalised rhythms (for example, allowing autonomy over time, joint fieldworking, allowing breaks in tracking activity) was more effective than imposing rigid rhythms and tempos, which can be tempting within the pressurised schedule of a research study. Despite more intensive processes in the most recent study phase, the study as a whole adopted a more extensive design, which may have limited our sample maintenance.

Micro-macro time
Effective retracing required frequent zooming in and out of spatio-temporal levels. Paying attention to the cyclical and linear trajectories of people's individual lifestyles helped us to structure our efforts to contact and interview them responsively to their personal flows of time. Conversely, taking a wide-angled view of what we knew about participants' spaces and communities helped us to make informed guesses about how they may have moved through space since we last met with them, and therefore where we might find them again.

Time-space
We found fieldwork to be as much latitudinal as longitudinal. Navigating the physical and social features of space was a core feature of retracing and retention. Devoting time to understanding the topography of the field yielded enhanced retention rates in ways that more distant, formless methods did not. We found that time stretched movement through space. Family reconfigurations, housing problems, job changes and other lifecourse events meant that people were rarely static in space, although we use the term 'stretched' because people often remained anchored to a central point in some way. Navigation of space was also indivisible from time: either smooth navigation saved time spent searching for leads, or time facilitated or impeded accessibility of spaces, and relations formed within them. Relationships were affected by space and time, because both influenced how researchers were perceived by participants and gatekeepers. The notion of sustaining relationships is a core tenet of sample maintenance in QLR (Neale, 2021), and while we did not build that into the time lapse between Times 2 and 3, the two year period of Time 3 involved deep engagement in sustaining relationships. A noticeable feature of our navigation of digital space was its more abstract and somehow empty quality. Although it was at times an efficient way to trace people, it was unrooted in (although a helpful precursor for) the spatio-temporally embodied relationships that eased us towards successful retracing and retention.

Continuous-discontinuous time
A powerful point of synchronicity was participants' perception of researcher's time as effort. If a researcher put a significant amount of time into finding and re-contacting a participant and the participant perceived it to be a significant amount of time, this framed time as a powerful social signifier of value, which aided retention. However, this may be a discontinuity from more linear 'researcher time' that can place limitations on expending too much time on too few individuals. The greatest potential for discontinuity is the cross-cut between 'researcher time' and 'participant time'. By paying attention to participant biographies, cycles and trajectories, we were able to bring our working methods more in sync with participants' personal time.

Trekking the timescape in longitudinal fieldwork: Practical strategies
We are not claiming that the following practical suggestions are new to the literature, particularly not to QLR researchers (see, Neale, 2021 for a more comprehensive primer on QLR methods). Rather, they are a distillation of the practices that we came to value and rely on as part of a broader theoretical appreciation of the timescape of our particular study. They indicate what spatiotemporal features we became sensitised to, and how we harnessed them. If treated in isolation as a 'bag of tricks' they are likely to be of less use than if they are harmonised as part of part of the longer and wider trek through longitudinal fieldwork, and as such we offer them as indicative rather than definitive.
Space. At initial interview, with consent, gather information about spaces participants occupy (past and present) and their social networks. Collect comprehensive contact details for participant and related others, but communal and familial information are more enduring than individual contact details. Identify nicknames, name variants, name changes and aliases. At subsequent interviews, review participant files for spatial and personal information. Combine them by using (for example) directory enquiries, the electoral register together and visual mapping tools. Assume people are unlikely to move very far unless information suggests otherwise. Plan fieldwork carefully: do not only visit specific addresses but identify communal spaces and how best to navigate them. Pre-existing local knowledge is helpful. Invest as much effort in locating/contacting members of a participant's social network as is invested in the participant themselves. If possible, make contact with participants and gatekeepers in a way that minimises social distance. Face-to-face, or at least voice contact is preferable. Cultivate genuine interest in participants and their lives, and develop ease with fieldwork in a variety of spaces. Use digital spaces comprehensively but use a situated ethical approach that considers the likelihood of harm for each person. Seeking feedback from participants can act as a helpful ethical compass.
Time. Pay attention to participants' biographies, including where they may be in their life relative to the topic studied. Try new methods, but also the same methods at a different time. Treat consent as dynamic, but without pressurising. Build relationships with participants, but not too intrusively. Acknowledge that researchers' experience of fluid time will vary. A management strategy that promotes basic human needs and autonomy over time may be more effective than one focused on targets and fixed time schedules. Plan fieldwork according to daily/weekly/seasonal rhythms. These will vary by participant, but generally, summer may be the best time for intensive fieldwork.

Limitations of our analysis
Our findings are limited to families who experienced paternal imprisonment. Spatio-temporal factors may play out differently in other samples. We acknowledge that our sample is smaller than some quantitative longitudinal studies, and as such, our approach may become less practical as sample size increases. Our analysis was largely retrospective, which meant we may have missed recording some space/time related information until we began to pay more attention to these features.
Dr. Souza has held a number of teaching positions, most recently as a Lecturer on the Forensic and Legal Psychology program at Marymount University in the U.S. She also has prior clinical work experience as a Mental Health Screener in a correctional center. Dr. Souza's areas of experience and published work focuses on the courts and sentencing, offender management (esp. risk-needs assessment and program evaluation), offender reentry and resettlement (with a particular focus on families), juvenile justice, and community/restorative justice initiatives.