The Good, the Bad and the Ugly: Miscommunication in UK Police Interviews and US Police Interrogations

This is the first comparative empirical study of miscommunication in US police interrogations and UK police interviews with suspects. The research was based on an extensive real-life data consisting of 100 transcripts. The main goal was to detect when and why miscommunication occurs in these two policing contexts, whether and how it gets resolved, and what the consequences of miscommunication may be. Miscommunication arises when speech participants draw opposing inferences from the same communicative exchange. Two main sources of miscommunication were identified: a) inferential ambiguity and b) linguistic complexity. The quantitative and qualitative analyses showed that, while the types of miscommunication are shared in the two jurisdictions, their frequencies and functions differ in some respects. Namely, deliberate miscommunication is more likely to be found in the US context while in the UK miscommunication is more likely to occur inadvertently. On the other hand, linguistic complexity gives rise to miscommunication more often in the UK than in US police communication. We discuss the results with respect to the two different approaches to questioning suspects in the two locales and translate our findings into practical applications for the training of law enforcement and language professionals worldwide.


Introduction
This paper presents a study of communicative exchanges between the police and suspects in two countries, the UK and the USA. There were two main research goals: i) to determine when, why and how often miscommunication occurs in police interviews with suspects and ii) to compare and contrast the two different (and most widely adopted) approaches to suspect interviewing in the two countries with respect to the frequency and treatment of miscommunication. Miscommunication occurs when the speaker (police officer) and addressee (in this case, a suspect) fail to negotiate meaning in conversation and possibly end up with different inferences about the same communicative situation. The research combined extensive empirical insights with recent theories of communication and language processing, in order to provide practically applicable findings. The operationalisation of these findings will help achieve much-needed communicative efficiency and linguistic accuracy in police communication as well as help ensure that the legal rights of individuals are properly exercised. This interdisciplinary study falls in the domains of policing research, applied pragmatics and applied psycholinguistics, and it addresses the phenomena in language processing and language use that impact communication outcomes in policing. In order to achieve the stated goals, an extensive, original and unique database was compiled, consisting of 174 transcripts of victim, witness and suspect interviews from the UK and interrogations from the USA. 1 For our current purposes we studied 100 transcripts of suspect interviews and interrogations, half of the sample from each jurisdiction. A half of each data subset is monolingual (either English or Spanish only) and the other half is bilingual, with multiple languages other than English involved (see section Current Study for details). This is the first and only US/UK parallel database of its kind, which took 15 years to create. The data collection involved extensive fieldwork including liaising with multiple units of law enforcement and language professionals, legal translators and interpreters, in both the UK and the US. As a result, for the first time, we are able to gain more finely grained insights into how the two very different policing jurisdictions manage multilingual communication with two very different official agendas, namely eliciting confession in the US vs. eliciting best evidence in the UK. We can also get a more precise view of the scale of the problem, which was not possible in previous research based mainly on individual case studies or using observational or survey methodologies. The UK investigative interviewing and the US police interrogation are particularly relevant to analyse because both constitute policing exemplars that are adopted in numerous jurisdictions across the world. Previous UK-US comparisons in the context of policing research have focused on a variety of aspects, such as the quantity of false confessions by suspects obtained through different questioning methods in the UK vs. the US, respectively (Meissner et al. 2012). As Leahy-Harland and Bull (2016) point out there are hardly any studies that have examined suspect responses in detail, which is what this study does, from a comparative perspective. To our knowledge, there has been no prior contrastive and detailed linguistic analysis available on the handling of miscommunication in the two policing contexts. Crucially, the present focus on linguistic ambiguity and linguistic complexity as overarching common sources of miscommunication has not featured in prior comparative research.
We start with a brief review of the relevant literature about the basic distinctive features of police communication and on the key differences between UK police interviews and US interrogations. The section that follows introduces the dataset in more detail as well as the methodology and the theoretical rationale for the analyses that have been carried out. The analyses are presented in the following section, and in the last section, we give a summary of the findings, concluding remarks and suggestions for both further research and practical applications.

Language and Communication in Policing
Communication between the police and different kinds of interviewees (victims, witnesses and suspects) is quite a specific interactional context, ridden with difficulties for many reasons. For example, as Heydon (2005) observed, in addition to the participants present in the police interview (the police interviewer, the interviewee, the interpreter) there is also a "silent participant" that must be included in the exchange, namely the recording tape, which shapes interactions in an atypical way, and which makes it difficult to maintain a natural conversation flow. Furthermore, it has been shown that the exchanges in police interviews are steered in the direction of creating usable evidence for judiciary purposes, which involves adherence to particular (and to an ordinary ear, peculiar) ways of phrasing and summarising information (Haworth 2006). The power asymmetry between the interlocutors has also been studied at length, and it has been shown that the communicative manifestations of this power difference underlie certain specific adjustments that need to be made in this conversational setting with respect to turn-taking in conversation and topic control and management (Haworth 2006;Heydon 2005). For instance, interviewees are not able to start talking about a topic of their own choosing but must restrict themselves to the topic of the interviewer's question, and it is the interviewer who has the power to sanction a nonanswer by asking the same question again. Heydon (2003: 95) argues that "at the level of turn-taking, suspects are in a vastly disadvantageous position when trying to support their version of events" (see also Oxburgh et al. 2015). All these factors contribute to police interviews being atypical conversational situations, while at the same time they also involve, and encourage, some of the everyday conversational routines, including rapport-building, expression of empathy and of politeness, and the use of indirect and implied rather than explicit (direct/confrontational) formulations for the purpose of saving face (in terms of Brown and Levinson 1987). In other words, efforts are being invested towards making communicative exchanges in police contexts look like ordinary conversations, but they are far from being ordinary because the two interlocutors do not have equal power and crucially, they may not have perfectly aligned communication goals (e.g. of making sure one understands what the other meant). In fact, their goals can be completely opposite, e.g. to find out what happened (the police) vs. to hide what happened (guilty suspects). Even when the goals of the interlocutors are perfectly aligned, e.g. when innocent suspects are being interviewed, there is still a lot of potential for miscommunication, and one of the central goals of the present research was to discover how and why this happens, and how it is dealt with (or not) in concrete, real-life situations.
Previous research on police communication has revealed many reasons that create barriers to understanding. One of the features of the use of language in policing that was deemed problematic on a number of occasions is the complexity of words, constructions and sentences. Gibbons (1990: 234-235) mentions that some long utterances by the police contain as many as 9 constituents and 6 prepositional phrases (see Data Analysis and Discussion section for examples and further discussion). It is not only the semantically complex words or legal terms that create difficulties in understanding: familiar words used in specialist contexts, such as "approach" or "relationship" (Gibbons 2003 andFilipović 2021 for details), can also be hard to interpret. Filipović (2019a) has explained and exemplified the ways in which overly long turns by the police, including multiple relative and conditional clauses within a long main sentence, are particularly hard to process, and also very hard to keep in memory and to translate (see also Filipović 2019b). Previous studies have also shown that some fundamental pieces of police communication, such as the Miranda Rights in the US, are difficult to understand even for native English speakers and for fluent speakers of English as a second language (Ainsworth 2008(Ainsworth , 2010Pavlenko 2017). For instance, many have no understanding of what "waive rights" means exactly in this context, and the situation gets even more complex if these difficult meanings need to be communicated in translation, for those speakers with zero or limited proficiency in English. An additional problem that arises then is the use of unprofessional interpreters in US policing. Interpreting in US police contexts is often done by bilingual police officers, sometimes with very limited proficiency in the required other language (Berk-Seligson 2009Filipović and Abad Vergara 2018). There is a high chance in such interviews that a suspect may agree with something without having fully understood what was being said and some confessions under such circumstances have later been challenged or proven to be invalid (Berk-Seligson 2009). In the UK only professional registered interpreters are used in police interviews, which goes a long way towards ensuring higher quality of interpreting and, crucially, eliminates the serious ethical issues related to the dual role of the officer-interpreter that plague the US system (see Filipović and Abad Vergara 2018;Filipović and Hijazo-Gascón 2018;Hijazo-Gascón 2019;Filipović 2019a, b;in press, accepted). We shall see in this paper how these multiple factors contribute to the miscommunications that occur in both the UK and the US context.

UK Investigative Interview vs. US Interrogation
In spite of the substantial amount of previous research on topics of police communication and police interpreting in both locales, there has been no comparative research that included a) concrete and substantial authentic data from both jurisdictions with a comparative focus on the linguistic means used for the two different purposes, respectively (confession elicitation in the USA vs. best evidence elicitation in the UK) and b) a critical assessment of communication outcomes in the two law enforcement contexts that reveal good vs. bad practice (advance notice: both are in evidence in both places). As briefly indicated in the Introduction, the UK interview and US police interrogation are particularly relevant for contrastive purposes because a) they involve very different approaches to interviewing and b) they are the two most widely adopted communication models in policing across the world. Shuy (1998:13) offers a very clear distinction between interviewers and interrogators. Interviewers probe, enquire, suggest, uncover rather than cross-examine, challenge, demand or trap, which is what interrogators do. The US interrogations are characterised by an adversarial approach based on the Reid Technique 2 and primarily aimed at eliciting confession. In contrast, the UK approach to interviewing suspects has its foundation in the PEACE approach 3 and includes some of the strategies developed within the Cognitive Interview 4 (Geiselman 2012). It outlaws oppressive techniques by favouring open questions and banning the use of deception strategies such as presentation of false evidence. Its aim is to obtain the highest quality testimony for investigative and evidentiary purposes rather than confessions without corroboration (Milne and Bull 1999). However, an earlier study by Stephenson and Moston (1994) 5 found out that most of the UK investigative interviewers who were surveyed (80%) still believed that the main purpose of the interview with suspects was the confession. Things have moved on since then and improvements have been made through PEACE-focused training (Clarke and Milne 2001), but the success of the implementation of the recommended improvements in practice has been questioned and has not always been fully satisfactory (Dando et al.2009). It has to be pointed out here that obtaining good-quality information, as opposed to obtaining a confession, remains particularly challenging for the interviewer (Oxburgh et al. 2015). Namely, he or she must mitigate the substantial institutional power held over the interviewee and create an environment that is conducive to the elicitation of evidence unimpeded by the fact that this is done by the very representative of the institutional power-the interviewer (see also Vanderhallen and Vervaeke 2014). As Oxburgh et al. (2015) explain, the interviewer dominates over the account given in an interview anyway, even when not intended, because of the nature of this specific interactional context that actively reinforces the interviewers' dominance. When you add insistence on confession to this mix, as is done in the US interrogations, it seems that the risk of obtaining inadequate evidence such as a false confession is bound to become higher. As shown in Meissner et al. (2012) review, false confessions are more frequent in the US than the UK and the communication method is likely to have something to do with that, namely the US use of coercive questioning focused on admission of guilt (sometimes at all costs; we shall see examples in Data Analysis and Discussion section). It has been documented that the use of coercion and manipulation and similar psychologically unsafe practices has in some cases led to serious miscarriages of justice around the world (see Gudjonsson, 2003;Kassin and Gudjonsson 2004, for a review).
Another risk that seems to be higher when employing an accusatory approach is that of losing the interlocutor altogether. A recent case study of officer attitude in the UK police context (Musolff 2019) has shown that even though it is more time-consuming, the non-aggressive approach that does not explicitly exert power or involve accusations and direct confrontation about the apparent inconsistencies in the suspect's account ultimately leads to better evidence, because a talking suspect is still providing very useful information for evidentiary purposes. A more accusatory approach used in one of the UK police interviews, which was in some respects reminiscent of the US method, was shown to lead to the suspect fully shutting down as soon as the accusations get heated and any chance of obtaining any information whatsoever is permanently lost (ibid.). In more recent years police training in the UK has clearly insisted on the search for truth and evidence elicitation in a nonconfrontational manner (see Pablos-Ortega 2019 and Pounds 2019 for examples and details), though as Musolff (2019) shows (and see also Filipović 2019a) confrontational and accusatory strategies, laced with presumptions about a suspect's guilt, are still present on occasion in UK police interviewing (though significantly less often than in US interrogations, where they are present in every suspect interview; see Data Analysis and Discussion section).

Theory
This study uses psycholinguistic insights about how meaning is expressed, processed and understood to explain the implications of specific instances of language processing and use for legally relevant communicative outcomes. One of the key insights in the study of human communication has been that context allows hearers to derive an appropriate interpretation of what the speaker has said since words and sentences can mean different things in different contexts-they are underspecified and able to prompt different inferences. These inferences were termed implicatures, by the philosopher Paul Grice (1957Grice ( , 1975, who argued that understanding the meaning of an utterance means understanding the speaker's intentions behind that utterance. There is now a large and growing body of empirical research that supports the view of communication as negotiation in interaction: what the utterance means needs to be negotiated and interactively co-constructed among speakers in conversation (e.g. Elder and Haugh 2018; Elder 2019). Thus, the argument is that meanings conveyed are not just tied to the communicative intentions of the speaker, but they also involve hearers' interpretations as an integral part of the process of co-creating meaning. In other words, speakers and hearers are jointly involved in the processes of negotiating meanings that need to be mutually agreed for communication to be successful. In essence, this view still reflects cooperation as the defining feature of communication (Grice 1957(Grice , 1989, which assumes that both the speaker and the addressee have common goals: to understand each other and be understood by each other. However, the cooperation can be apparent and not real because different personal or institutional motivations can lead interlocutors to have misaligned communication goals, for example to reveal the truth (the police) vs. to prevent the revelation of truth (guilty suspects or bogus witnesses). This results in ignoring and disregarding unwanted yet possible or obvious interpretations of what is said, or purposefully drawing the less likely interpretation from speakers' statements. And this is how miscommunication occurs, inadvertently or deliberately. It is possible that either one or both sides fail to detect that a miscommunication has taken place because they are firmly attached to their preferred interpretations and not aware of the different understanding of their interlocutors. It is also possible that one or both interlocutors detect the miscommunication but decide not to negotiate the meaning further and instead to let the miscommunication persist. We shall see instances of both in our data and assess their impact on communicative outcomes (see also Berk-Seligson 2011 on negotiation of meaning in police interrogations).

Data
The dataset used in the current study consists of 50 US transcripts (dated between 2000 and 2010) and 50 UK transcripts (dated between 2005 and 2015). The average length of the transcribed data is 47 pages per document. The content of each transcript is also variable, from singular "no comment" answers throughout the interview to rich, long narratives and detailed responses to questions. The US transcripts contain bilingual Spanish-English and monolingual Spanish conversations, while the UK transcripts contain monolingual interviews conducted solely in English and bilingual exchanges in English and another language (Lithuanian, Portuguese, Polish or Russian). All of the UK interviews are accompanied by an audio or video recording (from which only audio content was considered), while only 5 of the US cases had audio CDs available. It is important to point out that the UK and the US transcripts differ in one other important respect (addressed in detail in Data Analysis and Discussion section). The US transcripts are all bilingual and verbatim (i.e. presenting everything that was said in both languages by all the parties involved), while all UK transcripts are monolingual (English-only) and non-verbatim (i.e. not everything gets transcribed). Therefore, for the analysis of the material from the UK it was essential to refer both to the original tape recording of the interview and to the transcribed data. The US tapes were only randomly checked to ascertain that the transcripts were indeed verbatim. The conversation topics overlap to a great extent, though overall the US cases contain a higher number of exchanges with individuals suspected for more serious crimes (murder, incestuous sexual assault, rape, armed assault), while in the UK the suspected offences are more mixed and include both serious and less serious ones (e.g. possession of child pornography, sexual assault, robbery, domestic violence, supermarket theft and car insurance fraud). The transcripts from the two jurisdictions are illustrated in Figs. 1 & 2, respectively:

Methodology
The methodology used in this study includes mainly qualitative analysis of linguistic forms and on their meanings in concrete contexts of use, though some relevant and quite telling quantitative observations are also made (e.g. the general frequency of miscommunication occurrences in the two locales as well as the relative frequency of different types). The central goal is to identify how often and why the two main sources of miscommunication identified in prior research (e.g. Filipović 2019a, Berk-Seligson 2009, ambiguity and complexity, lead to problems in police communication and in evidence gathering in the two jurisdictions in question. Ambiguity is defined here as a multiplicity of meanings that can be derived from a single linguistic unit (in this case, an utterance), and complexity is defined as a multiplicity of structures within a single communicative turn (see details in the next section). This study is substantially different from many of the previous studies of police communication, which were mainly rooted in conversation analysis (CA) and focused primarily on interactional aspects in this context (e.g. conversation management) or on specific formulations of police questions (e.g. open vs. closed). By placing the focus onto detecting miscommunications and the different sources for it, as well as their consequences, we will be in a position to explain why and how often these problems occur, as well as suggest what can be done to mitigate, avoid or prevent their occurrence.
The data analysis was carried out in two phases. Phase one included the initial reading and annotation of the transcripts with the purpose of locating and counting the points at which there is a breakdown in the exchange. Most of the US transcripts were in an electronic form (40), and they were annotated using NVivo 12. There were 3 straightforward annotation categories: ambiguity, complexity, ambiguity + complexity. The remaining 10 US transcripts were in hard copy, as were all of the UK transcripts. They were manually annotated and all annotations were counted together and analysed jointly. The breakdowns in communication were detected by noticing the explicit signalling of a communication problem (e.g. "Sorry? I don't understand.") or by noticing the inadequacy of the response to a question (i.e. an off-topic or back-channelling response) by the addressee. Phase two involved a detailed qualitative analysis of the reasons for miscommunication in each case as well as determining whether and how it was resolved. This also included checking the interpreting quality in bilingual interviews in order to establish whether an inadequate negotiation of cross-linguistic and cross-cultural contrast in translation was causing, or adding to, the specific instances of miscommunication (which we do not report on at present, but see Filipović 2021, in press, accepted, for specific details in this regard). We have to point out here that we distinguish between the two main sources of miscommunication, ambiguity and complexity, even though there is some overlap between them. For example, some complex turns by police officers also contain statements that may be considered ambiguous and that can be interpreted in multiple ways. However, our key argument behind the distinction drawn in the analyses is that words or sentences need not be semantically and syntactically complex in order to be ambiguous in context. It seems that ambiguity can occur without complexity (Examples (1) and (2)), and complexity need not create ambiguity (e.g. as in the example (11)), and therefore we feel there is a justification in drawing this distinction in order to understand better why miscommunication arises.

Data Analysis and Discussion
The data analysis revealed very interesting trends, some of which are shared by both law enforcement contexts, and some that are not. Before we begin with the discussion of the similarities and differences detected, we have to report on the exclusion of certain transcripts from the analyses that had to be made. There were 7 interviews with just "no comment" responses in the UK, and in the US set there was 1 invocation of the Miranda Rights. Apart from these, the remaining 92 transcripts (or 92% of the dataset) did contain at least one instance of miscommunication and 74 transcripts (or 74% of the dataset) contained multiple miscommunication points (5 times or more). Some transcripts contained significantly more instances than some others, and this was mainly due to the long and complex monologues by the police officers in both locales (see further below in this section). Overall, we have strong confirmation that miscommunication is quite frequent in police interviews and interrogations, which is not unexpected. Miscommunication, both purposeful and accidental, is also very frequent in everyday language, and sometimes even welcome (as mentioned before, e.g. as a face-saving strategy, if there is a desire to remain non-committal or as a back-up for denial in case of a potential challenge). Directness and insistence on clarification is often avoided in communication in general because indirectness may be culturally or situationally preferred (see Terkourafi 2014) and because directness by speakers and insistence on straight answers by addressees denies the option to save face. Therefore, it is no surprise to find numerous instances of possible mismatched inferences between two interlocutors in police discourse. However, unresolved miscommunication in this context, unlike the everyday conversations we have, can lead to potentially serious consequences. In the following two subsections we focus on the main sources of miscommunication, the reasons for them and consequences for communicative outcomes. We shall contrast examples of accidental vs. purposeful miscommunication, and also illustrate the different ways in which they are dealt with, or not.

Ambiguity: The Consequences of Mismatched Inferences
We adopt a very broad operational definition of "ambiguity" here, because our focus is on all instances of words, constructions, sentences and conversational uses of these that are open to more than one interpretation in a given communicative situation and that can, as a result, create miscommunication. 6 Among the different kinds of ambiguity that can exist (for individual words, whole constructions, noun phrases referring to entities, etc.) some are much less relevant in the present context and do not appear to be a frequent source of miscommunication. For example, the lexical verb "drop" is ambiguous in English and has two possible interpretations, according to whether the action was intentional or unintentional. Consider for instance the sentence "He dropped the girl on the stairs". Was it done on purpose or not? English allows for both possibilities to be inferred. Resolving this ambiguity can be crucially important in some criminal cases, especially when the speaker of a language like English does not make clear which meaning is intended and the translator in another language (such as Spanish, which does not have this ambiguity) must decide whether to use the intentional or the non-intentional verb or construction when interpreting (see Filipović 2007Filipović , 2013Filipović , 2021. Within the present context our focus will be primarily on what we can call inferential ambiguity, namely the kind that is created interactively by speech participants at the utterance level and that involves implicatures in the sense of Grice (1975). This happens when two interlocutors derive two different inferences from the same exchange. Inferential ambiguity, resulting from two different interpretations of the same conversational exchange, appears to be the most frequent source of miscommunication in our data. Context is needed in order to establish which of the possible inferences is the stronger one and more likely to be derived, but even context does not always guarantee that both sides will settle for the same inference (see also Current Study: Theory). Here is an example from the data that illustrates how this type of miscommunication may arise due to the impossibility of deciding firmly which of the possible inferences was drawn by each speaker and whether the same inference has been agreed on: (1) Police officer: I know that you are a Sureño, 7 it is no big secret to me. Suspect: Well, it is your job. This brief exchange took place between a US police officer and a suspect in serious crime (serious assault with a deadly weapon). The police officer produces an accusatory statement and is expecting a confirmation from the suspect. The suspect does not either confirm or deny the accusation but does provide a response, which by inference, indirectly, could be taken to mean that he is indeed guilty of gang membership. However, even though this may be the most likely inference (because had the suspect been innocent, he would have denied the accusation more fervently or at least explicitly), the suspect did not actually commit to the content of the officer's statement. We see later in the interview that the police officer continues with the interrogation as if the confession about gang membership has been given (see Filipović 2021 for more details on this and other examples of inadvertent confessions). Such implied rather than confirmed verbal commitment can easily be cancelled (e.g. as in: "Well it is your job, but you did not do your job well -I am not a Sureño"). Similar examples appear in the UK data, and one of them is given in the following example (from a domestic violence case): (2) Police officer: Did you threaten to kill her? Suspect: Her brother threatened to kill me.
There are two possible inferences based on what the suspect said: a) "I did it because I was threatened myself", and b) "I was threatened but I did not do it". At this point the line of questioning shifted to the topic of the alleged victim's brother and the threat he made to the suspect, which meant that the suspect successfully evaded the answer about the alleged threat he himself had made. Miscommunication in the two cases above means that it remains unclear whether a successful meaning negotiation took place whereby both sides agreed on the same interpretation. The police officer may think they have gotten the confirmation about the alleged offences by the suspects, when in fact they did not and any claim to this effect can easily be denied by both suspects interviewed in excerpts (1) and (2).
This kind of miscommunication caused by inferentially ambiguous, inexact exchanges where the negotiation of meaning has not been fully accomplished between the two interlocutors is found in every analysed transcript at least once. In over half of the UK and over two-thirds of the US transcripts (27 and 39 out of 50 transcript files, respectively) it occurs multiple times (at more than 5 points in the conversation). This finding is not surprising because as we observed earlier, many different kinds of conversational exchanges, personal and professional, are typically replete with indirect responses, vague and implied meanings for a variety of reasons. However, in police communication directness should be favoured because what matters most is 6 The general view of ambiguity that we present in this applied linguistics context does not distinguish between ambiguity and vagueness, even though this is a well-defined distinction in semantics (see e.g. Kempson 1977 andJaszczolt 2002), since this is not relevant for our purposes. We also extend the term "ambiguity" to include "implicatures" in the sense of Grice (1975) that arise when participants in the same conversational exchange derive different inferences from it, and we refer to such cases as "inferential ambiguity". 7 Sureños are one of the feuding gangs in San Francisco (USA), who were competing for dominance with Norteños. The two groups reportedly agreed on truce in 2018: (https:// www. mercu rynews. com/ 2008/ 02/ 14/ rodri guez-former-norte os-sureos-unite-in-cease-fireminis try/). finding out the facts. At the same time, linguistic means with functions other than fact-finding are also necessary in order to establish a communicative relationship, such as words and phrases that help create rapport between the interlocutors. Sometimes, it may even be advisable to put directness and pushy negotiation of meaning on hold and let some instances of evasiveness and potential miscommunication be left unchallenged in order to keep the conversation going (e.g. see Musolff 2019) because ultimately, this would lead to more content and in all likelihood more evidence. Finally, as Musolff (2019) points out, the way in which instances of apparent miscommunication are addressed can hold the key to communicative success. Namely, an aggressive and accusatory approach to the resolution of apparent instances of miscommunication can lead to a complete shutdown on the part of the suspect, while letting it go can have the benefit of keeping the suspect talking and eventually providing enough evidence about guilt or innocence. A potential danger of letting unresolved miscommunications persist, as previous research has shown, is that they can lead to serious misinterpretation of events, sometimes to the point where denial is being interpreted as confession (e.g. see Filipović 2007Filipović , 2019aFilipović , 2021. Thus, the main point here is that not each and every instance of ambiguity should be challenged-this is not a viable option anyway because miscommunication is unavoidable and omnipresent, as we explained. What police officers can do, however, is resort to subtle decision-making about which instances of ambiguity should be challenged and resolved, or prevented if possible. The suspects should also be told that they are allowed to ask for clarification if something is ambiguous or unclear in the officer's question, especially in interpreter-assisted interviews where cross-linguistic transformation adds to the problem (Filipović 2007(Filipović , 2019a(Filipović , 2021. Here is an example of a case where a clarification was necessary and it was properly sought by the interrogator, which was fundamental for avoiding a potential confusion as to the agency of the suspect in a stabbing incident and for securing the suspect's statement on record. The suspect's line "he tripped" was rightfully not accepted as an answer to the investigator's question "did you stab him?", because that answer creates inferential ambiguity (between "I did not stab him because he tripped and stabbed himself" or "He tripped and I stabbed him"-it becomes clearer in the last line below and also even more so later in the interrogation that the latter was indeed the case): (3) Police officer: Okay, what kind of screwdriver did you have on you, then? Suspect: Long, about like this. Police officer: Uh-huh, what color was it? Suspect: Red and blue.
Police officer: Hm, and did you stab that guy? Suspect: Man! Police officer: You didn't stab him? Suspect: He tripped. Police officer: Uh-huh, when you swung at him with the … the screwdriver did you stab him or didn't you stab him? Suspect: I threw it at him and it hit him back here, it didn't, didn't, didn't, didn't, didn't hit him in the, anywhere else, it didn't go in him. Police officer: Okay, and when you said, "I threw it at him", did you try to stab him or did you throw it at him, what, what is it? Suspect: Well, just to, to hit him like this [gesture]! In (4) we have the opposite situation, namely when the much-needed clarification was not sought. In this example the miscommunication occurs because the suspect was not able to understand what was expected from him (i.e. to clearly say yes or no to the Miranda Rights questions) and this miscommunication instance was then extended further by the continuation of the interrogation even though the suspect clearly did not respond as required. The suspect's answer that was a question, "What do you want me to talk about?", could be interpreted interactionally as either a yes (meaning "Ok, start, what is this about?") or a no (meaning "I first want to see what this is about before I accept to talk"). It is the duty of the police officer to elicit a distinct yes or no response upon the reading of the Miranda Rights, and this officer does not do so but chooses to interpret the answer below as "yes" and promptly launched into further questioning: (4) Police officer: So, having in mind these rights, do you want to talk to us? Suspect: What do you want me to talk about? Police officer: Well, about what happened tonight. Suspect: Well, this.. Police officer: Go, go ahead. Continue.
We have to highlight here that it is no coincidence that the US data contain fewer invocations of the Miranda Rights (1 instance) than the Right to Silence invocations based on the UK equivalent, the Police Caution (7 instances). There is only one suspect in the US data who exercises his right to remain silent and a number of others try to, but do not succeed, along similar lines to the one illustrated in example (4) above (see Filipović, accepted, for more examples and discussion). It seems that miscommunication of the Miranda Rights and the expectations about the responses are seized upon on many occasions in the US portion of the dataset, whereby suspects' responses are purposefully manipulated in order to proceed to interrogation (ibid.). These multiple cases of improper Mirandisation in the US data point to the lack of appropriate negotiation of meaning, driven by the institutional agendas, such as confession elicitation at all costs. In the UK, the communication of Police Caution is not completely unproblematic, especially the middle part of it, which tends to be the most difficult to comprehend due to the complexity of that portion. However, the UK police officers are stringently trained to explicitly seek understanding from the interviewees about the meaning of each portion of the Caution (Filipović 2021).
Overall, it can be said that the abuse of miscommunication is an issue in police interviews and interrogations, though this is significantly more often the case in the US than in the UK data where it occurs only sporadically. In addition, it is crucial to highlight an important UK-US difference with regard to exploitation of miscommunication: it is mainly the police officers who are engaging in resorting to deliberate miscommunication in the US (see the next section), while in the UK it is not the police officers but rather the uncooperative suspects who are doing so. In the following example from the UK portion of the data the suspect is apparently exploiting the ambiguity of the word "where". The police officer wants to know whether there had been any exchange of bodily fluids between the suspect and the alleged victim, and the suspect refuses to provide the requested information for quite a long time, 8 as we see in the exchange below (see also

Complexity: How (Not) to Waste a Speaking Turn
In this section we focus on the syntactic and semantic complexity of police officers' speaking turns in conversations because it is the second most common source of miscommunication in our dataset. It is also an area in which conversational repairs are the most difficult to achieve. For instance, if a word used by a police officer or a suspect is not understood, this could be easily remedied (e.g. by providing definitions and checking understanding of the key terms; Filipović 2021). Complex syntax, however, poses a greater challenge and has been identified as a source of processing difficulty and communication problems in previous work. 9 For example, students' comprehension of examination instruction was significantly improved when complexity was removed from the formulation of the questions (Abedi and Lord 2001). More specifically, in police communication research, Berk-Seligson (2009) found that syntactic complexity, such as the use of embedded clauses, or multiple clauses per sentence, is typically found at the beginning of the police interrogation (in her US case studies). She argues that even a native speaker of English who had no more than a high school education would have difficulty processing these structures. We can agree entirely with this statement with the addition that in our dataset we found complexity of this kind in multiple places throughout both the UK interviews and the US interrogations. 10 Previous literature has shown that inappropriate or unproductive questions can lead to distorted responses (Milne and Bull 1999) or even false confessions (Gudjonsson 2003). Complex questions, negative questions and statement questions have all been identified as inappropriate (Milne and Bull 1999;Shawyer and Walsh 2007;Griffiths and Milne 2006;Oxburgh et al. 2010). Interestingly, statement questions are one of the most frequent types of question asked in the context of UK policing (Leahy-Harland and Bull 2016; Filipović 2019a). Leahy-Harland and Bull (2016) argue that statement questions per se were not considered problematic in their research of suspect responses and suggest that a study of statement questions merits a more detailed qualitative analysis, which is what we give below. They further point out that negative questions in their analysis were associated with an increased likelihood of suspects denying the offence. And even though negative questions were not frequent in their study, the fact that they exist may be a cause for concern that should be addressed in police training of interviewers. Filipović (2019a) found that negative questions can persist in specific interview styles of some police officers in the UK, who use them significantly more than most officers, and during all phases of the interview. In spite of them being less frequent, and not so widely spread across many interviews, negative questions are certainly worth mentioning in the context of miscommunication.
We analyse the key types of questions that lead to miscommunication under one heading of complexity because it is their syntactic and semantic complexity that is the main root of the problem, as we exemplify and explain further below. Crucially, the different complex questions perform different functions in the UK vs. the US law enforcement communication. Our goal was not to count all of the complex questions, but to identify those that created miscommunication and explain why this happens and what the consequences are. We found that problematic complex questions occur in almost a third of our UK transcripts (30%), and in the US portion they are found in one sixth (12%) of the transcripts. We start by exemplifying the different types of complex questions that we identified in our dataset (see also Filipović 2019a for further discussion). These include multiple referent introduction (example 6 and also (8a)), negative interrogatives (example 7), complex statement questions (examples (8a) and (8b)): (6) Police officer: You mention ringing his cousin -were he and the girl in that club with you? Suspect: X stayed with me all night towards the end close to the incident time when we were arrested; that was the one time we parted. Police Officer: Was his cousin and the girl in the club as well? Suspect: Yes they were there, they entered that place. (7) Police officer: Did you not find that strange that he wanted his notebook? Suspect: He's explained that he's fallen out with whoever he was renting from, whether that was his sister or his landlord, I didn't understand, and then he was going to his friend's […].
(8a) Police officer: In the first pub there is some CCTV of X's cousin and the girl and it also shows another girl who had her phone stolen. The victim can be seen holding her phone and she then puts it away. She then approaches the bar in the pub and is followed by X's cousin and the girl? Suspect: I dunno.
(8b) Police officer: So X told me that at night you had gone into her bedroom and lay next to her in her bed. Then you put her hand on his penis and it felt wet. From what I understand X then got up from the bed, so did you, and then you gave her some biscuits and milk for breakfast. Suspect: It was a normal thing to give her at breakfast.
All the above examples come from the UK portion of the dataset, and we also exemplify and discuss some US examples further below. Example (6) illustrates multiple referent introduction ("he", "his cousin") delays the processing and impedes the understanding of the content, as evidenced in the suspect's response, which seems off-topic. The second question by the police officer, which is simpler and much clearer, results in a straightforward answer. Thus, the initial, complex question was a waste of time and in fact introduced a risk that not answering it properly could be interpreted as a resistance strategy and create negative impact on the rapport (see Filipović 2019a).
Another type of a difficult-to-process question detected in police interviews involves negative questions such as the one in example (7). Such questions are harder to process than the regular interrogatives and they carry an inherent bias -a presumption that something was done that should not have been done, or vice versa. Additionally, both yes and no answers can be given as a confirmation or a denial, as in "yes, I did not think it strange" vs. "yes, exactly, I did think that strange" or "no, I did not think it strange" vs. "no, I actually did think it strange". As we can see, this complexity of possibilities is mind-boggling at the best of times, and we often cannot be sure if we have answered a negative question in the way we intended even under normal circumstances, when chatting with friends, let alone when under pressure in police questioning. The suspect in (7) starts explaining why he thinks that the back door was used for the entry, but in effect, it sounds like he was avoiding a straightforward answer, which need not have been the case. It is the police officer in this case that is using a confrontational, presumptive and accusatory communication style by formulating this (and multiple other) questions with a negative interrogative.
Examples (8a) and (8b) illustrate another kind of problematic atypical phrasing, which is the use of complex statements as questions, very common in police discourse and more pronounced in the UK context where these examples come from. As Oxburgh et al. (2015) point out, it is important to focus on the function of a question rather than just the form in order to determine what the question wants to achieve. Statements often function as confirmation-seeking questions, but if they consist of multiple clauses or sentences, and multiple logical or temporal operators (such as "if" or "when"), problems arise in language processing. Not all statement questions are problematic. Simple statements questions, as in e.g., "She unbuttoned her pants?", are very frequent in UK police interviews and US police interrogations, but they do not cause miscommunication because they are simple. The complex statement questions, on the other hand, are very problematic because they make it difficult for the suspect to process them and respond to them, and they can elicit responses that can easily be interpreted as lack of cooperation. For example, in (8a) the suspect just answered "I dunno", while the expected response would be to either confirm or deny the description of this event, or provide more detail, if wanting to be cooperative. Importantly, as observed before (Filipović 2019a), statement questions can significantly delay the elicitation of evidence because their exact function or the format of an expected response is not clear to the interviewees. Furthermore, what we also see in (8a) is multiple vague and unclear references to specific objects or events (using "it" or "that" without specifying what those refer to precisely) and such questions fail to perform the main function of elicit more information or confirmation. In (8b) we also see a very complex question where the serious accusations in the beginning of the question are not addressed and only the last portion, and the most irrelevant one in this context, is picked up in the suspect's answer. The reason for this could be that the suspect is avoiding the serious accusations but still wants to look cooperative and appear to be answering the questions. The fact that he decides to talk about the appropriateness of cereals for breakfast instead of accusations of molestation of a child may even seem like a form of mockery. However, we also have to bear in mind that loaded questions such as this may often end up in triggering a response related only to the last portion of the question due to the so-called recency effect. This phenomenon captures the feature of human memory to recall the most recent items on list the best (see Murdock 1962;Bjork and Whitten 1974). Whether this specific case of miscommunication is genuine or manipulated on the part of the suspect, we cannot say (unlike in some other cases where it is obviously the latter; see below). What we can say is that the response in (8b) is certainly odd and apparently uncooperative, and it delays and impedes the elicitation of evidence.
What is important to highlight here is that in most cases these complex statements that are supposed to function as questions do not appear to be used to threaten or provoke in the UK context (in contrast to the US context, see further below); rather, the reason why they are used is to probe for confirmation or, denial, or for more detail, which can be used as evidence. It is only on some rare occasions (in two transcripts) that UK police officers slip into the adversarial/ accusatory mode characteristic of the US police interrogations and not endorsed in the UK investigative interviewing model. An example from our UK data illustrating this is given in (9): (9) Q: Is that true? I have given you the opportunity -if it is not, when we view it numerous times in the past here people tell me things like that and we go through it all and it doesn't ever appear like they tell us. A: I didn't do nothing wrong.
It is interesting that complexity was less of a problem for communication in the US context, though it is not unexpected if we take into account the different type of discourse that the US interrogation is compared to the UK police interview. We found that the syntactically complex structures such as those in examples (6) and (7) are much rarer and only occur occasionally. This is not very surprising: police questions in the US context are short, succinct and fired in quick succession, as exemplified in the example (10): (10) Police officer: How many, how many times did you touch her breasts? Interpreter: He said that he was joking around, playing is probably the way I touched her. Police officer: She says seven times. Is that accurate? Interpreter: He said, yeah, probably, yeah, but [Inaudible/overlapping voices] I didn't have bad intention. 11 Nevertheless, in 12% of the US transcripts (6 out of 50 files) we still have instances of the complexity of the kind exemplified in (8a) and (8b). In contrast to the UK data, the function of these complex questions is completely different. It is aimed solely at persuading the suspect to confess. A police officer may speak at length about the different possible outcomes, using multiple conditional sentences, multiple clauses within sentences, combined with accusatory or threatening statements and imperative sentences requiring explicit confirmation, as in the examples (11) and (12), originally conducted in Spanish only (here cited in the English translation from the transcript). In some cases, the police officer is doing most of the talking (in one case over 90% percent of the time(!)). The suspect in example (12) barely manages an occasional short turn or an expression of apparent concurrence ("um-hum") throughout the interrogation.
(11) Police officer: Having each one of these rights that I've, uh, just explained to you, do you want to talk to me? Suspect: [Inaudible]. Police officer: Well, it's a decision, uh, if you want to talk to me that's fine, if you don't want to talk to me that's fine too, it is your right, but I have to, the law says that I have to read you your rights that you have because a lot of ti … uh, times, people who are not from the United States come and they don't understand the rights and that is part of, uh, our job, that you, the, the people who are here with us understand the law, it's, uh, the part that, it's a thing that I have to do and here is … my … letter … … … Have you had problems with the girl? Suspect: No. (12) Police officer: … we're looking at the person that murdered her. And, and by telling me that …… you know nothing about it …… it takes it away …… from maybe it's an accident thing … … a much more serious type of murder …… and that's where we're going. Suspect: Um-hum.
The suspect in example (11) is obviously hesitating to confirm that he would talk to the police-this is reflecting in muffled inaudible speech. The third turn is crucial here (Elder and Haugh 2018; Elder 2019): we see the police officer not wanting to lose the opportunity to question the suspect, and he fires a barrage of words in order to persuade him to continue the dialogue while not eliciting the information about the suspect's understanding of the caution. We can assume that the Miranda Rights were not fully understood by the suspect and this failure by the police officer to communicate them properly, and the consequent miscommunication, seems to be on purpose (see Ainsworth 2008Ainsworth , 2010. In the example (12) the suspect's only response is "um-hum" for almost the whole duration of a lengthy interview, during which the officer was implying that there could have been some extenuating circumstances that could mitigate the severity of the alleged crime. 12 It is safe to assume that the suspects in examples (11) and (12) could not have processed everything said by their respective interogators or the implications of the officers' statements. The miscommunication in example (12) also stems from the possibility that the officer is taking the concurrence as acceptance of guilt and his explanation for it, while the suspect is perhaps just being generally agreeable and simply uttering "um-hum" as a back-channeling response. Even the legal representative of the suspect who went through the transcript later ( Fig. 3) was not sure what exactly was being implied. In fact, the legal representative marked a point (in hand-writing) in the transcript (Fig. 3) where a possible illegal promise of leniency ("if you want to tell me what happened, the thing will end quickly") appears to have been made). This case also illustrates a previously identified phenomenon in police interrogations known as confessing through concurring by both Spanish speaking suspects in the US (Berk-Seligson 2009; Filipović 2021) and Aboriginal suspects in the Australian context (Eades 1994(Eades , 2008. The concurring is usually motivated by the suspect's cultural norms of communication or by the external institutional pressure and power inequality. Ross and Mirowsky (1984) explain that acquiescence among US Hispanics functions as a self-preservation strategy of individuals who are relatively powerless in society. It is a deferential, submissive and non-resistant response, all in order to present a 'good face' and merit acceptance. The suspect in this case concurs from the beginning to the end of the interrogation.

Conclusions and Future Directions
Miscommunication takes place when the negotiation of meaning is unsuccessful. We established that there were two main sources for it, namely inferential ambiguity and the linguistic (syntactic and semantic) complexity of multiple utterances in a single speaking turn. By adopting a contrastive approach to the UK interviews vs. the US interrogations we discovered a number of interesting similarities and important differences. Inferential ambiguity is the most frequent source of problems in police communication overall, and part of the reason behind it lies in the ways everyday conversation works, namely preferring to give indirect and vague responses rather than direct and explicit ones. We saw that, if multiple interpretations are left to linger on in conversation, we may miss opportunities to understand what happened and why and put apparently important communicative evidence at risk of a later denial. However, challenging suspects at every turn and pressing them to clarify any potential miscommunication is not an option: it can antagonise the suspect and prevent the officer from creating rapport. The police officer wants the communication with the suspect to look like a normal conversational exchange even though both sides know it is not. This is why some instances of ambiguity need to be left unchallenged, and based on the research presented here, the devil lies in the detail: when to challenge and when not to challenge. The conclusion that transpires based on our findings is that a challenge is necessary when the questions related to key features of the case (e.g. confirmation of gang membership or admission of threatening somebody in examples (1) and (2), respectively).
With regard to the second most common source of miscommunication, complexity, we discovered that that there are more varied sources of complexity in the UK context than in the US context and that complexity is a cause of miscommunication more often in the UK police interviews than in the US police interrogations. As Yeschke (1993: 71) pointed out, complex (overloaded) questions are improper in contexts where the veracity of an answer counts a great deal. As we saw, it is not clear what the question even was in some of the very complex speaking turns by police officers. A solution to the complexity problem lies in the use of simpler statements and questions, and favouring the use of explicit interrogatives, whether open or closed type, as necessary. For instance, we can reformulate the example in (6) to make it easier to process by turning it into two questions, without lengthening the officer's turn significantly (e.g. "I will now ask something about the cousin of your friend -were that cousin and his girlfriend at the bar?"). A similar transformation can be done in the case of other complex questions cited in this study.
We have also discovered an important difference in how inferential ambiguity is dealt with in the UK vs. the US police communication. In the UK contexts it seems to be unintended, whereas in the US contexts it can be either unintended or intended, and in the latter case, with an aim to trick suspects into accepting to be interviewed without proper Mirandisation or coercing them into confessing. Similarly, with regard to complexity, we see the different rationales for its occurrence in the two locales, respectively. It appears in the UK police interviews in an effort to elicit either new information or confirmation for the already obtained information, and seldom to purposefully overload or confuse the suspect. In the US interrogations analysed here complexity seems to have an almost exclusive function to distract or put pressure on suspects and make them talk and confess. This is in line with the very different kinds of ethos in policing: the UK best evidence elicitation vs. the US confession elicitation.
Finally, we can conclude that even though miscommunication in police contexts cannot be completely prevented and avoided it can be dealt with effectively through avoidance of complexity and by calibrating insistence on ambiguity resolution. This can be taught through training, as is done within the author's previous and current applied research projects [www.tacit.org.uk]. It is our hope that some of the insights here will be studied further and implemented in practice in order to help improve the efficiency and accuracy of police communication, thus saving time and costs for law enforcement while also ensuring the best possible quality of public service. included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.