Abstract
Requesting police assistance can be especially challenging in cases of domestic violence, since perpetrators may be able to overhear victims’ telephone calls. This means that callers may not be able to make direct requests for help. Simultaneously, a routine task for police call-takers is to categorize incoming calls as genuine rather than, say, accidental or nuisance. We collected and transcribed 192 audio-recorded calls to a UK police service, which included interactions between callers and call-takers as well as between national operators and local call-takers. The latter provided access to the professional parties’ pre-transfer discussion and interpretation of what kind of trouble might be occurring in silent and otherwise ambiguous calls. Using conversation analysis, we found that, as well as unambiguous requests for help (e.g. ‘I need you to come because of assault by my partner’), callers formulated apparently inapposite turns (‘hiya, you all right?’) and used non-lexical resources (e.g. breaths) to build actions which also mobilized assistance. Professional call-takers’ discussions included domestic violence-implicative interpretations (e.g. ‘I heard a woman shout’). Parties collaboratively leveraged the affordances of turn design and sequence to request and offer help without revealing to potentially overhearing parties that callers were talking to the police. Our findings have implications for understanding how actions like requesting are accomplished in social interaction, as well as for training call-takers to recognize and act on communicative ambiguities in cases of domestic violence. Data are in British English.
Introduction
The provision of emergency medical, criminal, and other life-saving assistance is fundamental to the safety of citizens. While the very purpose of healthcare and emergency service organizations is for people to call and make requests concisely and directly, nevertheless requests are conveyed in myriad ways. Requesting assistance has been a sustained focus of conversation analytic research since its inception in the work of Sacks (1967), who studied ‘the search for help’ at a suicide prevention center. Since then, we have learned a tremendous amount about the way people ask for and are offered forms of help and assistance in emergency and non-emergency domestic, social, and institutional settings; in different communication modalities such as in co-present environments or on the telephone, and in different languages (e.g. Drew and Couper-Kuhlen, 2014; Taleghani-Nikazm et al., 2020).
In this article, we investigate how callers to police ‘emergency’ (999, in the UK) and ‘non-emergency’ (101) services request help in the context of domestic violence and abuse. We start with an overview of the general concerns of conversation analytic research on requests before looking more closely at requesting in institutional settings, including emergency services. While there is limited research on calls to the police about domestic violence, it is general to all emergency calls that call-takers must sift genuine from prank, malicious, non-malicious ‘nuisance’, or non-urgent calls (e.g. Weatherall et al., 2016). Relatedly, callers’ descriptions of trouble are ‘vulnerable to suspicion and doubt’, and, therefore, that call-takers might focus on the way the trouble is reported rather than the trouble itself (Whalen and Zimmerman, 1990: 465). Given that, in cases of domestic violence, perpetrators may be physically co-located and able to overhear victims’ telephone conversations, there is a particular challenge for these callers to be treated as genuine since, as we will see, making direct requests is not always possible. That is, callers may not able to use the most obvious words and phrases to explain their reason for calling or request the help they need, which increases the risk that their call will be miscategorized. Therefore, this paper will show that, while callers do make direct (and unambiguous) requests, some exploit the machinery of turn design and sequence, including what Keevallik and Ogden (2020) refer to as ‘sounds on the margins of language’ but ‘at the heart of social interaction’ (p. 1) to convey their interactional purpose, which must then be correctly understood by call-takers.
Requests for assistance in social interaction
As Fox and Heinemann (2016: 500) explain, conversation analysts have closely described ‘the relations between linguistic form and social action: that is, how speakers design their utterances to be recognizable as doing a particular action (action formation), and how recipients recognize that action, or ascribe that action to the utterance (action ascription)’. We already know much about the social action of requesting and the way that requests are carried through different grammatical vehicles, recipient-designed by speakers to handle their own stake in and entitlement to make requests as well as contingencies like urgency or import (Curl and Drew, 2009). For instance, we do not typically find requests for urgent assistance designed as low-stakes or low entitlement (e.g. ‘I was just wondering if the fire brigade could come to put out a fire’) or the opposite (e.g. ‘It’s imperative I get a facial today’), though, of course, both are possible depending on the particulars of the situation.
Researchers have therefore identified systematic uses of grammatical formats for requesting (e.g. imperatives, declaratives, interrogatives), and Fox and Heinemann’s (2016: 501) useful summary describes ‘how a range of different formats can be employed to address a speaker’s “primary concern” with getting someone else to help with something, while their grammatical variation allows participants in interaction to at the same time address other (often multiple) ‘contingent concerns’. . .’. Furthermore, Kendrick and Drew’s (2016: 4) concept of ‘recruitment’ articulates a ‘symbiotic relationship’ between requesting and offering and the complexities of securing help. Analysts have also considered how people make requests without ‘using language’; that is, producing actions by using ‘the affordances of the visual and auditory modalities’ (Rossi, 2014: 301, see also Antaki et al., 2020). However, we know less about how people may make requests on the telephone ‘without language’ or where such resources are unavailable, such as in ‘silent’ calls to emergency services. We examine such cases in the next section, as well as in our analysis in this paper.
Requests in emergency situations
Calls for emergency assistance are well-studied by conversation analysts. While such calls are often what Bolaños-Carpio (2017) refers to as ‘monotopical’, focused on ‘dealing with one single task at hand throughout the entire call’ (p. 9), there may be much complexity in aiding apparently obvious problems (e.g. ‘the house is on fire’; ‘I’m having a heart attack’). So, depending on the specifics of organizational systems and their protocols and technologies, callers may need to request one from a range of emergency services (police, ambulance, fire); they must convey a genuine request for help rather than being treated as a hoax or nuisance caller, and, despite calling a service designed for ‘emergencies’, still need to convey the urgent nature of their request in a compact way, maximizing comprehensibility and progressivity and minimizing misunderstanding, delay, and repair (e.g. Tracy, 1997). They may also be calling to request assistance for themselves directly, or as third parties for others, in relation to a trouble they have experienced or witnessed.
Meanwhile, call-takers – who are the gatekeepers of service provision (Zimmerman, 1984) – must discern what kind of assistance is needed and, indeed, whether requests are genuine (see Garcia and Parmer, 2011; Larsen, 2013). Much research reports that ‘[e]mergency centers all over the world receive non-emergency calls on a regular basis, a problem which impedes their ability to assist civilians who are truly in need of emergency help’ (Blushtein et al., 2020: 916). But defining what counts as ‘genuine’ is also not straightforward. Sampson (2004) distinguishes a range of call types, from unintentional ‘phantom’ wireless calls, misdials, and hang-ups to prank (e.g. Emmison and Danby, 2007), ‘exaggerated emergency’, and ‘lonely complainant’ calls (p. vii). Other research has shown that, while calls to the emergency services may ‘not be true emergencies; nevertheless, they will be calls for citizens who need help’ (O’Looney 1997, cited in Ward, 2018: 142). And, as Friedman and Albo (2017: 142) note, what constitutes ‘an actual emergency may differ between the government and citizens and even among citizens’. In their research, they found that minoritized callers may be disadvantaged when calling the police for what were perceived to be legitimate reasons and that this ‘degrades the relationships between police and these minority groups’ (p. 159; see also Raymond, 2014, on callers navigating 911 services when they do not speak the language in which service is offered, and Cromdal et al., 2008, on how children produce ‘institutionally “workable” reports’, p. 949).
Perhaps the best-known work on the empirical reality of 911 emergency calls is that of Whalen, Zimmerman and Whalen (e.g. Whalen et al., 1988; Whalen and Zimmerman, 1987; Zimmerman, 1992), who described the ‘achieved organization’ of calls to US-based 911 services and their composite sequences, from opening/identification; complaint/request, the ‘interrogative series’ of questions and answers; response/promise of assistance, and closing (Zimmerman, 1984). Kevoe-Feldman and Pomerantz (2018) note that the promise of assistance is crucial in that it ‘serves to mark the end of the sequence that was initiated with a request for assistance or service, while simultaneously functioning as a preliminary move toward call closing’ (p. 491). While this organization is demonstrably robust, there is much variation and idiosyncrasy within and across turns. Some variation is ‘designed in’ right from the start of the call, depending on the way in which the summons (i.e. the ringing telephone) is answered. For instance, Cromdal et al. (2012) showed that answering with the ‘stand-alone emergency number opening typically results in callers asking for a specific service. In contrast, opening turns that end with a direct query about the incident tend to solicit brief descriptions of the trouble’ (p. 371) – and that the latter produces more ‘accurate and rich’ first requests from callers, as well as removing the need to articulate urgency.
Bolaños-Carpio (2020) showed that requests for help when the situation is life-threatening differ from non-life-threatening cases. In the former, callers may make requests via descriptions of the problem (e.g. ‘a woman has been attacked with a machete’, p. 231); in the latter, requests are more conditional (e.g. ‘I’m calling to see if. . .’). Kent and Antaki (2020), in their analysis of calls to UK emergency (999) and non-emergency (101) call centers, further show that the call-taker’s subsequent first question (which initiates the ‘interrogative series’) has implications for the progression and outcome of the call, including ‘upcoming refusal of police action’ (p. 640). More specifically: ‘requests for the caller’s location (which are treated as indicating that police action will be taken); open-ended requests for further information (treated as neutral); and queries of the relevance of the incident or legitimacy of the caller, and reformulations of the caller’s reason for calling (both projecting upcoming refusal of police action)’.
A key challenge for all emergency services call-takers is how to respond to ‘silent’ calls (Independent Office for Police Conduct, 2020). In the UK context in which the data analyzed in this paper were collected, ‘silent’ calls are those in which callers could be in immediate danger but are ‘unable to make a noise or speak’ (Independent Office for Police Conduct, 2020). In such cases, call-takers must sift genuine emergencies from nuisance calls. From their analysis of what they termed ‘active silent’ calls, Kevoe-Feldman and Sutherland (2018: 5) offered as training advice the ‘four-second rule’, which stipulates that ‘Active Silent callers need at least four seconds to respond to dispatchers’ silent call prompts’. They analyzed ‘patterns of silence’ to ‘inform dispatchers how to maximize their opportunity for identifying true Active Silent callers’ (p. 9). We return to their research later in the analysis when we consider silent calls in our own dataset.
Assistance in the case of domestic violence
We know less about domestic violence calls as a sub-category of requests for emergency (acute and non-acute) assistance. We know more about non-urgent requests to, say, domestic violence charities and helplines, through a series of studies conducted in New Zealand (Tennent, 2019; Tennent and Weatherall, 2019; Weatherall, 2020; Weatherall and Tennent, 2021). A key feature of calls to the police about domestic violence is the possibility that the ‘problem’ for which help is being sought is another human being (the perpetrator) who may be co-present or otherwise in a position to hear one side of the conversation. This leaves the caller with interactional challenges, such as calling one type of recipient (the police) without being understood to be doing so; or requesting help via different actions or without the words that connote the type of help needed. As noted earlier, the risk for the caller is to be treated as making a prank or otherwise non-genuine request, and this interactional challenge is one that is less studied in the wider conversation analytic literature on requesting.
The potential for victims of domestic violence being overheard by perpetrators while seeking help increased during the coronavirus pandemic. Immediately, reported cases of domestic violence increased globally, as victims were less able to leave their homes (Richards et al., 2021). In April 2020, the Independent Office for Police Conduct (IPOC, in England and Wales) promoted a ‘Silent Solution’ tool to assist callers during the lockdown and which ‘alerts the police to 999 mobile callers who are in need of emergency assistance but are too scared or unable to speak’ (Christie, 2020). The data that we analyze in this paper were collected both before and during the pandemic. While we do not take a comparative approach, the combined datasets provide an opportunity to identify more instances of how callers and call-takers manage potential co-presence of perpetrators. While other studies have examined participation frameworks (Goffman, 1981) in which there are (multiple) co-present and/or overhearing parties who do not speak (e.g. Richardson and Stokoe, 2014) or are non-addressed recipients who remain ‘backstage’ to a primary encounter (Stokoe and Sikveland, 2020; Whalen and Zimmerman, 2005), disguising actions as other actions adds another layer of complexity when it comes to understanding requests and, more generally, how interlocutors achieve intersubjectivity. This is the aim of the current paper.
Data and method
In the UK, members of the public experiencing domestic violence and abuse can request assistance in an emergency by calling 999 or, in less urgent circumstances, 101. Calling 999 involves (at least) three parties: two from the police service plus the caller. On dialing, callers are first connected to a national call center and asked to select a service from police, fire, or ambulance; an operator subsequently establishes the need for police assistance. We do not have this section of the overall process or workflow in our corpus, although it is recorded from the point at which the caller and national operator establish a connection. The national operator then connects the call to a local call-taker when the required service has been established. Once connected, the call-taker elicits information from the caller about what has happened, where they are, and who is involved. If the caller is disconnected, or if they are connected but apparently not responding (e.g. voices/noise can be heard in the background), the system enables the operator and call-taker to nevertheless establish the need for a police response and locate the caller geographically. This might include the ‘playback’ of a recording of the initial interaction between caller and operator prior to connecting callers to local call-takers. Therefore, the workflow includes conversations between operator and call-taker that provide insights into their immediate interpretation of, and stance toward, the calls they are hearing, including silent and otherwise ambiguous calls.
As part of a research and training collaboration, we were granted access by a UK police service to digital recordings of all incoming and outgoing calls made at call centers made as part of the routine work of the organization for training and evaluation purposes. Data processing agreements were drawn up by the police with the authors as research partners. All personal information was redacted on site at the police station and the recordings anonymized. Data were stored on encrypted systems that are specialized for forensic and legally sensitive materials at our universities, where the project likewise received ethical clearance. In the data presented, pseudonyms are used throughout. Each extract heading contains information about whether the call was made prior to or during lockdown, and whether the call was made to 999 or 101 (e.g. ‘Pri-75-999’ refers to the 75th call to 999 in the pre-pandemic dataset).
To identify calls to include in our collection, we used the UK-based definition of domestic violence and abuse as stated in Sections 1 and 2 of the Domestic Abuse Act (2021). The Act defines domestic abuse as any single incident or pattern of conduct where someone’s behavior toward another is abusive, and where the people involved are aged 16 or over and are, or have been, ‘personally connected’ to each other (regardless of gender or sexuality). The abuse can involve ‘physical or sexual abuse; violent or threatening behaviour; controlling or coercive behaviour; economic abuse. . .; psychological, emotional or other abuse’. The term ‘Personal connection’ applies to two people who are currently, are due to be, or have been, married, civil partners, in an intimate relationship; parents (or had a parental relationship) to the same child or are relatives (the Act gives further definitions of ‘relatives’).
Our total corpus comprised 192 calls made to ‘999’ and ‘101’ by members of the public reporting being victim to or witnesses of domestic violence and abuse both prior to (in 2019) and during the COVID-19 lockdown in England in 2020. This included multiple calls in which the first (incoming) was ‘silent’ or otherwise ambiguous and a follow-up call (outgoing) which sometimes enabled us to understand and categorize the first. As noted previously, although it is possible to do some kinds of comparative analysis between 101 and 999 call types, we examined the datasets to identify as many instances as we could of how callers and call-takers manage potential co-presence of perpetrators. We took cues from both caller and call-taker as to the nature of the call. Sometimes the calls were explicitly characterized by one or both parties as being a report of domestic violence and abuse. At other times, this was more implicit. Within the corpus most callers made either direct requests (e.g. ‘I need the police; my partner’s assaulted me’) or implicit requests via descriptions of events in the ‘reason for calling’ slot (e.g. ‘my partner’s being a nightmare. . .’; see also Bolaños-Carpio, 2020). However, we also identified a subset of ambiguous calls (e.g. only muffled background noise could be heard). In these cases, call-takers typically make a follow-up outgoing call to the number which may or may not be answered/connected. We identified 18 (15–999; 3–101) such incoming and 11 such outgoing calls (~10% of our overall dataset). These 29 calls were transcribed using the Jefferson (2004a) system (see also Hepburn, 2004) for conversation analysis and it is this subset that we worked on for the analysis below.
Analysis
We present three sections of analysis. The first shows that and how callers make unambiguous requests for police assistance using formats that have been well-described by other researchers. By unambiguous, we mean that the reason for the call is formulated across relatively short sequences with relatively little repair initiation or other features of interactional trouble. The second section presents analysis of how call-takers and operators identify possible cases of domestic violence as they analyze incoming calls. Finally, we examine the ways in which call-takers and callers collaboratively leverage apparently inapposite ‘non-request’ actions, as well as ‘silent’ calls or those containing non-lexical items, to nevertheless progress genuine requests.
1. Unambiguous requests for assistance
The first extract comes from the start of a call to 999 recorded during the pandemic. Our interest is in how a direct request for police assistance for an initially unspecified emergency is quickly formulated between the caller (CLR) and call-taker (CTR) as an ‘assault’ by an ‘ex-partner’ and a situation in which the parties to the assault are a female victim and a male perpetrator (Lee, 1984).
The call-taker answers the telephone with an institutional identification ‘Police Emerge
The caller’s request is framed using a high entitlement and high contingency format (‘I need’: Curl and Drew, 2009) and subsequently formulated as an imperative ‘to come to’ an address. In this way, the caller conveys the urgency of the request (‘I need yuh to come’) and the location which will be crucial to fulfilling the request. Notably, her request does not include an account for this need. Note also that, although the caller asks for ‘you to come’, it is clear that ‘you’ refers collectively to the police and not the call-taker herself. Larsen (2013) also found that high entitlement requests were more likely to get rapid dispatch than low entitlement formats; Kent and Antaki (2020: 17) further found that ‘[i]n the absence of incident-specific details, request entitlement alone can (and does) provide a metric to position the call-takers first substantive question along a gradient from projected acceptance to refusal of the request’. Furthermore, since the caller supplies her address as part of her request, the call-taker confirms and expands upon these details rather than ask ‘what’s your address’ in their first substantive turn, a question that Kent and Antaki refer to as ‘dispatch-implicative’.
Thus, at line 04, the call-taker confirms the caller’s address by repeating it; this is re-confirmed in third position by the caller (line 05) just as the call-taker asks for the caller’s postcode. After a gap (line 08), the call-taker confirms the postcode and continues to ask ‘an’ what’s happened’ (line 09). As in Cromdal et al.’s (2008: 934) data, the turn ‘projects an answer, or more to the point, a formulation of the caller’s acute reason for the call (i.e. the accountable “emergency”)’. The caller answers this wh-question with one word: ‘assault’, which, in Gricean terms, provides the maximum amount of relevant information in the most efficient way: it categorizes the event (the ‘parties to an assault’ are a victim and a perpetrator, see Lee, 1984); the event is relevant to the service; it conveys entitlement to be granted assistance. However, the fact that the caller’s response contains no agent (e.g. ‘my ex-partner has assaulted me’) is treated as an omission by the call-taker who asks, after a delay in line 11, ‘from who sorry’ (line 12). The caller then provides the information that it was ‘my ex-partner’. Like ‘assault’, the call-taker confirms the information by repeating it (line 14) and then asks, ‘is he still the:re’. This is an orientation both to immediate and ongoing threat as well as the possibility of the perpetrator overhearing (or interrupting) the current call. The call-taker also assumes that the caller’s ex-partner is male, which is confirmed in the repair initiator at the start of line 15 (‘He-’). We leave the extract as the call-taker attempts to confirm the caller’s current precise location.
Extract 2 is an example of the way requests may be achieved via other actions such as describing situations or events in the ‘reason for calling’ slot (Bolaños-Carpio, 2020; Hofstetter and Stokoe, 2015). The extract starts just after the national operator has transferred the call to the local call-taker.
In response to the call-taker’s question, ‘What’s your emergency’, the caller does not produce a direct, high entitlement request of the kind we observed in Extract 1, despite also being a call to 999 (though note that this call-taker’s first turn is much longer than in Extract 1, containing a ‘go ahead’, confirmation of connection and identification, and then a wh-question). Rather, the caller formulates the circumstances she is currently experiencing in the ‘reason for calling’ slot. However, the call-taker treats her description as a tacit request for assistance by producing a sequence-closing third response (‘okay’; line 10) before launching a second turn construction(al) unit (TCU) to progress the call with a ‘dispatch-implicative’ question ‘
Extracts 1 and 2 therefore include the same basic components: establishing what and has happened and where. In Extract 1, the caller provides her address as part of a high entitlement direct request and the call-taker follows up by confirming the address details and asking what is happening. In Extract 2, the caller describes what is happening and the call-taker follows up by asking for the caller’s address. In both cases, the nature of the caller’s complaint is built through formulations of aggressive behavior (‘assault’; ‘pushed me around’) which is attributed to a male partner/ex-partner who, also in both cases, remains inside the location where events are taking place. In both cases, there is no particular ambiguity as to what is occurring, and police assistance is dispatched.
In the next section, we consider cases involving operators and call-takers, as well as call-takers and callers, and the resulting interactional layers that support progress toward requesting and offering assistance.
2. Identifying possible cases: ‘I heard a woman shout’
The next two extracts exemplify how the professional parties – operators and call-takers –interpret what they hear in recordings immediately as callers are connected to police. In Extract 3a, recorded prior to the pandemic, the operator (OPR) is transferring a caller to the local call-taker. This call is ‘live’. That is, while the operator and call-taker speak, the caller is apparently speaking in the background. At the start of the extract, the operator formulates these background voices as ‘raised’.
At lines 07–09, the operator informs the call-taker that, while there has been no request for assistance, they can hear ‘raised voices’ in the background – represented at lines 05 and 10. Between lines 12 and 28, the call-taker attempts to interact with the caller but receives no lexical responses; note, however, there are audible rustling noises and outbreaths that are placed sequentially in ways that may be designed to be heard as attempting to communicate. These turns are ‘semantically underspecified tokens’ (Keevallik and Ogden, 2020: 12) that, while designed to be ‘treated by participants as meaningful’, are interestingly different in our data because they are designed to be opaque in meaning by one potential (overhearing) recipient but clear to another (the police call-taker).
We rejoin the interaction 20 lines later following a discussion between call-taker and operator about the caller’s location. Note that, at the end of this extract, ‘cleared’ is the term that police use to indicate that the call has ended or been disconnected.
At line 53, after the call-taker has informed the operator they can disconnect from the call, the call-taker again attempts to interact with the caller. Between lines 55 and 68, there are a series of non-lexical turns apparently taken by the caller (at least, they are occurring in an audibly channel separate to the call-taker’s) comprised of sharp out-breaths, clicking, and blowing sounds. Compared to the in- and out-breaths that are a well-observed feature of turn design and turn-taking, these sounds are potentially produced as distinct components. In our dataset, we found cases of accidental dials in which it later turned out there was no need for police assistance. However, systems are in place to support callers who are unable to speak. For instance, the Metropolitan Police website instructs callers who are in danger to ‘try to speak to the operator, even by whispering’ or that they may ‘be asked to cough or tap the keys’ on their phone. As discussed earlier, this also means that call-takers and operators must distinguish between genuinely accidental and action and action-oriented key presses. One way that call-takers do this work is to ask yes/no questions (e.g. lines 53, 70) and enable the participation of callers who may be overheard and unable to provide semantically meaningful turns.
In Extract 3b, the call-taker, by using a summons (lines 70 and 76), attempts to initiate interaction with the caller. Note the extended production of each ‘hello’ with ‘querying’ intonation. Lines 72–75 and 78–79 are turns taken between colleagues in the control room. Our interest is, again, in the call-taker’s report to the operator at line 74 – that they have ‘them’ (the subjects of the call) back on the phone and that they ‘heard a woman shout’. This, following the ‘raised voices’ referred to in the previous extract, builds a category-implicative description of an activity – an argument – which partitions into participants – the ‘parties to an argument’ (Lee, 1984; see also Kameo and Whalen, 2015). Note again the use of the indexical ‘a’ to formulate one of those parties as a woman in categorial not specific terms, thus categorizing the overall event as an argument in which at least one of the parties is a woman.
The case presented in Extract 3 continues over two further recordings. For completeness, we present sections of these calls to show that the ambiguity of the case is never resolved. Extract 3c contains the start and end of a third incoming call.
((Five minutes elapse))
In this third conversation, the caller, ‘Rita’, speaks to the call-taker. As part of her first substantive turn, the caller says that she is ‘ass
The caller’s turn is delivered with many components of upset including audible aspiration and ‘wobbly voice’ (Hepburn, 2004). It is apparently difficult for the call-taker to fully comprehend what Rita is saying. For instance, note the overlapping and unclear talk at lines 09–12 during which the call-taker says. ‘>Sorry< slow d
‘Where are you’. The call continues for 5 minutes, during which much confusion ensues, and the call ends with the caller hanging up the phone and the conversation returning to the operator and call-taker.
In the final call, the call-taker makes an outgoing call to Rita in which she claims to have lied and invented her previous allegations.
((58 lines omitted during which the call-taker pursues clarity of the situation))
While the call-taker does not readily accept the caller’s claims to have lied about her previous report of assault, the outcome of the call is that police do not attend the scene. But the ‘truth’ of the situation does not get resolved. In the UK, many complaints of domestic violence are withdrawn, creating further barriers to justice for victims (e.g. Johnson Astills, 2017). This may be such a case. But from the call-taker’s perspective, the ambiguity remains.
The final case in this section comes from a call recorded during the pandemic. The call-taker and operator play and discuss a caller’s key-press turns (lines 18, 20, 30, 33) while background voices are audible (lines 11–16, 22–28, 31).
Before playback starts, the operator formulates a possible reason for the call: ‘I think there was a request’, and the possible category of event: ‘there was a male shouting in the background’ (lines 03–05). The operator also orients to their careful interpretation of what they heard rather than rushing to decide (‘I thought it was the tee vee to start off with’, see Jefferson, 2004b).
In addition to the key-press turns, at line 27 the caller apparently utters, in a whisper, ‘ºº(Get) Policeºº’, though our transcription conveys uncertainty about the word ‘get’. What was said, and the certainty of hearing something, is discussed in the final turns in the extract between the operator and call-taker (lines 40 and 51). At line 44, the call-taker aligns with the operator about the nature of the interaction they both just heard, though is also careful to avoid certainty (‘Yeah, okay. Hmm. Does sound like’,). Note that the operator’s initial formulation of ‘a male shouting’ (lines 03–4) is progressed by the call-taker with ‘she whispered something’ and not ‘a female whispered something’. The subsequent reference connects the pre- and post-call segments (and the parties) together.
We return later to this case (as Extract 7) to see how it evolves as a genuine request for assistance. In the final section, we examine cases in which the parties collaboratively produce genuine requests through practices other than asking for assistance.
3. How call-takers support callers’ genuine requests: ‘Are you able to talk freely?’
Extract 5 is an outbound call made by a police call-taker to a female caller. This call follows a previous inbound call from the current caller’s son, who reported threatening and violent behavior from his mother’s new husband. It is worth mentioning that, in the preceding inbound call, the son has admitted a history of making hoax calls to 999 and the call-taker explicitly questioned the truthfulness of his report. However, we focus now on the conversation between the mother and the police, and especially on the call-taker’s question ‘Are you able to talk fr
Whatever the ‘truth’ of this situation, we can see that, in response to the call-taker’s questions, the caller denies that an ‘incident’ has occurred (e.g. line 18) and challenges the call-taker’s right to ask about it (e.g. line 29). However, our interest is in both the form and function of the call-taker’s turns, starting with lines 08–10. After a greeting, apology for calling, and both an institutional and personal introduction, the call-taker asks a question using a yes/no interrogative format: ‘Is ↑everythin
In response, the caller starts by challenging the presupposition of the question: ‘Y
After another long delay, the caller replaces the call-taker’s category ‘partner’ with ‘husband’. Also note the use of the reference ‘he’ (‘- I d
In Extract 5, the caller does not respond in yes/no terms to the yes/no questions asked by the call-taker; that is, they do not take the opportunity to convey something tacitly. This contrasts with the next case. Extract 6 comes from another outbound call. The police call-taker has previously received a third-party report about the caller from the caller’s brother. Referring to his sister, the caller’s brother has stated that she is ‘very scared if she talks to police then her husband would beat her up’. The police call-taker has told the brother that they will telephone her ‘cos I don’t want to send a police officer round while he’s there. . . cos obviously we need to see her when she can speak openly to us’. Like Extract 5, the call-taker attempts to establish whether the caller can talk freely, but the trajectory in this call is quite different.
Like Extract 5, the call-taker starts with an account for calling (cf. ‘s
Asking yes/no questions across a series of turns not only enables the participation of a caller who is, as the call-taker formulates at lines 26–27, ‘r
This final extract brings together observations made throughout the paper: the operator and call-taker’s tentative characterization of a ‘silent’ call as one of potential domestic violence and the subsequent call with the caller in which both parties collaboratively enable a tacit request for assistance. Unlike previous extracts, in this case the caller is not ‘free to speak’, and thus the call-taker launches the ‘interrogative series’ (Zimmerman, 1992) of questions (e.g. 07, 12, 15, 18, 22, 26, 33) about what is occurring at the caller’s house, including establishing whether there are children or weapons in the house, and whether anyone is hurt. Each of the call-taker’s questions enable one-word responses from the caller such that she can convey to the call-taker the nature of the threat.
However, also note the way that the caller responds to the first turn from the call-taker, in which they identify themselves as from the police and that the caller had dialed 999. At line 05, the caller returns the greeting (‘hiya’) and initiates a ‘how-are-you’. Both components routinely appear at the start of a conversation between friends or people who already know each other (see Stokoe, 2018). Here, though, the caller’s turn is positioned in overlap with the call-taker’s turn. By saying ‘hiya y’all right’ in this position at the start of the call, and in overlap, precision timed to appear when return greetings and ‘how are yous’ are due, the caller uses the sequential and structural norms of interaction to begin to convey to the call-taker that her request is genuine but that she is not ‘free to speak’. In other words, the basic structural organization of a different kind of interaction is leveraged by the caller in contrast to the basic structural organization of emergency calls (Kevoe-Feldman and Pomerantz, 2018) in ways that nevertheless enables the call-taker to launch the interrogative series of questions and elicit information relevant to progressing her request.
The same pattern, in which callers position turns with precision in overlap with call-takers to convey to whoever might be overhearing their conversation that they are talking to someone other than the police, was reported in Stokoe (2018), see also Albert et al., 2019). She showed that in a 911 call, the caller asks, ‘I’d like to order a pizza for delivery’. When the dispatcher begins to state that she is calling an ‘emergency line’ and challenge the fit between request and service, the caller places a turn in overlap to say, ‘Yeh.h uh: large with half pepperoni half mushroom?’ The practices shown in Extract 7, like the pizza call, begin to show us how callers and call-takers collaboratively work to progress requests as genuine. Extract 7 also shows us how the caller leverages sequential and, crucially, categorial resources. The caller designs turns at talk that are, to the potentially overhearing perpetrator, hearably category-bound to the category ‘friend’ rather than ‘caller-to-the-police’, thus comprising a kind of simulation (see Sharrock and Watson, 1985).
Discussion and conclusions
The aim of this paper was to examine how callers to police emergency services request assistance in cases of domestic abuse and violence. As well as describing relatively unambiguous requests, the interactional design and trajectory of which fitted previous research on emergency requests, we also identified a subset of calls in which the reason for calling was unclear. This meant that the communicative task for emergency call-takers involved not only responding to clear requests for help but also attempting to understand and progress those where what was occurring was opaque.
Calling the police in cases of domestic violence and abuse therefore involves interactional challenges for both parties. For callers, who are often victims, the physical environment of domestic space means that perpetrators may be able to overhear their telephone conversations. Consequently, callers may not have available to them the basic resources of action, words, and phrases that work in aggregate to make direct requests for assistance. Instead, callers deployed ‘semantically underspecified tokens’ (Keevallik and Ogden, 2020: 11) such as in- and out-breaths, as well as key presses on their telephone, to convey meaning to their recipient. However, while ‘clearly an interactional asset’ (Keevallik and Ogden, 2020: 11), callers’ non-lexical turns in the context of requesting emergency assistance have a double affordance since they are designed to be opaque (or convey something other) to one potential (overhearing) recipient but clear in their action-orientation to another (the police call-taker).
Furthermore, callers designed their turns at talk to build actions such that anyone overhearing would not interpret them as category-bound to a call to ‘the police’ but to ‘a friend’ (‘Y’all right’). For these communicative resources to be correctly understood as vehicles for requesting assistance, callers placed them at (in)apposite positions in the unfolding sequence, such as where a response would be due if one was, say, at the start of a conversation with a friend. For call-takers, the challenge is to identify and progress genuine requests, even though many calls to police are accidental, non-urgent, nuisance, or malicious. As Garcia and Parmer (2011: 297) note, since ‘[p]olice work is one setting in which there is a greater degree of doubt and skepticism than is common in everyday life’, emergency call-takers must play an ‘information game’ in which the parties try ‘to analyze the information others give or give off in order to determine what is genuine, false, sincere, or cynical’ (p. 298, see also Kameo and Whalen, 2015). Thus, both parties collaboratively deployed the affordances of turn design, grammar, category, and sequence to request and offer emergency help without revealing to potentially overhearing perpetrators that callers were talking to police. We saw how the well-described ‘interrogative series’ of questions was adapted for recipients who may be overheard, including dispensing with the reason for call or replacing that reason with something fitted (or category-bound) to another activity or setting.
Our analysis provides insight into the professional parties’ (the local call-takers and national operators) own analysis and interpretation of the evidence they have: fragments of turns, sequences, background noises and muffled voices. One feature that stood out was the formulation of some of these noises and voices as domestic violence-implicative situations (e.g. ‘there was a male shouting in the background’; ‘I heard a woman shout’). These fragments of interaction, as well as indexical references to the parties and events themselves, were treated by call-takers and operators as recognizable as gendered, heteronormative, domestic abuse or violence. Similar formulations can be found in police interviews with suspects (e.g. Stokoe and Edwards, 2008) who have been arrested for assaulting the police officers who arrive at a scene following the kinds of 999 calls examined in this paper. For example, when asked for his account, one suspect claims that the attending police officer ‘decided to arrest me . . . because he heard me and my girlfriend arguing’. The interviewing officer states that the police have attended an address ‘and they’ve heard some screams from a female from within the flat. So obviously at that point they’ve got some concerns for that person’. Thus, (over)heard fragments of loud, perceptually ‘male’ and ‘female’ voices, co-located in a particular kind of dwelling, constitute category-based descriptions of potential cases of domestic violence.
Our findings have a series of implications. In addition to contributing to conversation analytic research on requesting (emergency) assistance and providing direct access to the empirical reality of accessing help in cases of domestic violence, our research also has implications for ongoing discussions about conversation analysis as a method and, in particular, the way ‘context’ is understood when analyzing and interpreting social interaction. In their analysis of the 911 call described above, in which a caller makes a request for pizza and the call-taker sets up an interrogative series to enable her to disclose and seek assistance for domestic violence and imminent threat, Stokoe et al. (2021) discuss the notion that it is not ‘context’ that enables the call-taker to hear a request for pizza as a genuine request. That is, just because someone calls the police does not mean that their call will be treated as genuine. As discussed earlier, the volume of nuisance and other inapposite calls (Kevoe-Feldman and Sutherland, 2018) means that the successful identification of genuine requests is not because of ‘context’. Goodwin and Heritage (1990) articulate the problem of ‘bucket’ theories of context, “in which the situation of action is treated as anterior to – as ‘enfolding’ and determining – the action that takes place within it. Rather. . ., just as a social setting determines the sense of a current action so, in turn, that action will redetermine (by sustaining, modifying, updating, or transforming) the sense of the current context” (p. 286).
A substantial proportion of genuine requests for emergency assistance are not made using what we may think of as the obvious words and grammar – they are (sometimes very) indirect. However, police call-takers are routinely able to identify them as genuine, nevertheless, and distinguish them from nuisance calls. Thus, our paper shows that the basics of sequence, grammar, category, and turn-taking are not just the focus of empirical work of conversation analysts themselves – they comprise members’ methods for interacting in situations in which the stakes are high. If people are able to request police assistance by conveying to a co-present party that they are talking to a friend, while conveying to a police call-taker that their life may be in danger, then we see the power of mundane talk-in-interaction to achieve extraordinary things.
Footnotes
Acknowledgements
We are grateful to the reviewers for their very constructive engagement with and feedback on a previous version of this article. Any errors remain, of course, our own. We are also grateful to the police service with whom we collaborated on this project and for their engagement with the research findings.
Author’s note
Emma Richardson is now affiliated with Loughborough University, UK.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.