https://orcid.org/0000-0002-6143-5265
ABSTRACT
This paper presents a new case study examining the use of police transcripts to assist the court in understanding poor-quality forensic audio admitted as evidence in criminal trials. The 1995 trial it studies was the first major Australian case to include extensive expert opinions about police transcripts provided by the prosecution. Despite the fact that experts on both sides noted serious problems with the police transcripts, the judge allowed them to assist the jury, with the expert opinions offered as (ineffectual) subsequent commentary. The legal procedures the judge used in doing this were upheld on appeal, and have been followed ever since as a model for judges admitting opinions of both police and experts. The paper demonstrates how these procedures (unintentionally) privileged the opinions of police “ad hoc” experts over those of genuine experts, enabling the erroneous transcripts to influence not only the 1995 verdict, but a 1997 appeal and a 2014 inquiry. Analysis reveals the reason for these anomalies as the fact that the procedures incorporate misconceptions about spoken language and its representation in a transcript, which, though they have been thoroughly refuted by linguistic science over many decades, remain deeply embedded in the “common knowledge” accepted by wider society – including powerful institutions such as the law. The paper ends by calling on Australian linguists to find effective ways to address the misconceptions that affect the legal handling of forensic audio, by building further on the success of other branches of forensic linguistics in seeking direct engagement with the judiciary outside the trial process. The first step in achieving this is for linguists to gain a thorough understanding of how the legal procedures for handling poor-quality forensic audio operate, both in principle and in practice. The aim of the present paper is to contribute to that understanding.
1. Introduction
Many criminal trials feature evidence in the form of spoken language captured in an audio recording, typically the product of covert surveillance conducted on behalf of police. These recordings are often of very poor quality, to the extent the court needs a transcript to understand their content. Current Australian law allows transcripts to be provided by detectives investigating the case, who are given the status of “ad hoc expert” on the grounds that they have listened to the audio many times. Linguists often find this concerning. Providing an accurate transcript of poor-quality audio requires genuine, not “ad hoc”, expertise, as well as independence from the case. Since police lack both of these, there is a high risk their transcripts could be misleading.
This concern is valid: police transcripts are inherently unreliable – not in the sense that they are inevitably inaccurate, but in the sense that it is not possible to rely on them being accurate. However, concern at this level misses a far more serious problem. The law is well aware that police transcripts are unreliable, and has instituted a series of safeguards, endorsed by the High Court of Australia in Butera v DPP (Citation1987), to protect the jury from being influenced by a misleading transcript. It is these safeguards that are the more serious problem. Developed on the basis that understanding spoken language is a matter of common knowledge, the safeguards rely on lawyers, judges and ultimately the jury checking transcripts against the audio. This takes insufficient account of well-established scientific findings that the very act of checking can “prime” listeners to hear in line with the transcript – even if it is demonstrably inaccurate (Fraser & Kinoshita, Citation2021). Far from protecting the jury, then, these safeguards tend to propagate an erroneous interpretation of powerful evidence to the whole court, including the prosecution, the defence and the judge.
The safeguards’ ineffectiveness is shown by multiple examples of demonstrably misleading transcripts having passed all the legal safeguards to “assist” jurors in their interpretation of crucial audio evidence. One such example is the 2008 trial, referred to as the “pact” case, recounted in Fraser (Citation2018). Here a police transcript passed all the safeguards, to play a significant role in securing a murder conviction and 30-year sentence. Only much later was it shown, via phonetic analysis, that the transcript was inaccurate; and, via experimental evidence, that it must nevertheless have influenced the jury (as it had influenced the whole court, including the judge). However, an application to review the conviction was rejected on the grounds that the possibility that a police transcript might be wrong is recognized and anticipated by the law, and the jury was protected via the standard safeguards, which had been followed properly by the trial judge. This ignores the strong implication that, if the safeguards, followed properly, can enable a demonstrably misleading transcript to influence the jury, they must be ineffective.
The pact case was one of several that prompted the Australian Linguistic Society to lead a 2017 “call to action”, asking the judiciary to review and reform the handling of indistinct covert recordings used as evidence in criminal trials. That call has since gained some traction (McMahon & Fraser, Citation2023), helping to raise awareness that the courts need better protection from the influence of misleading transcripts. One common suggestion, from both lawyers and linguists, is that police transcripts should be reviewed for accuracy by genuine experts. This sounds reasonable, and can help in some cases; however, it is no panacea. It still relies on lawyers, judges and juries checking transcripts against the audio – with the addition that they are now checking transcripts not just of ad hoc experts but also of genuine experts.
The present paper offers a new case study detailing just how problematic that can be. The case is the 1995 trial of David Harold Eastman, in which police transcripts of indistinct audio played a key role in securing a murder conviction and life sentence. Although old, Eastman retains relevance as the first major Australian trial to hear extended expert opinions about police transcripts from both prosecution and defence. In fact, experts on both sides raised substantial concerns about the police transcripts. Nevertheless, the trial judge, after checking them personally against the audio, endorsed the police transcripts as suitable assistance for the jury, with the experts’ opinions provided as commentary. His reasons for doing so were upheld in a 1997 appeal, making the 1995 procedures available as a precedent – and similar procedures have been followed in numerous subsequent cases involving both ad hoc and genuine expert opinions about audio evidence. For example, R v Cassar (Citation1999), widely cited for its succinct summary of how transcripts of forensic audio should be handled in court, refers to Eastman as “valuable authority”.
However, through a series of developments not related to the audio evidence (§2), the 1995 conviction was quashed in 2014, with a retrial held in 2018. During preparation for the retrial, it became clear even to the prosecution that the police transcripts used in 1995 had indeed been thoroughly inaccurate (even more so than recognized by the 1995 experts), and no police transcripts were provided to the 2018 jury (though of course they were well known to all those involved in the trial, via their review of the 1995 materials). The prosecution’s fresh transcripts (prepared by the same expert they used in 1995) retained few of the original incriminating utterances, and even those were challenged by the defence expert. These changes, along with other factors to be discussed, contributed to the 2018 jury returning a not guilty verdict, with compensation paid to the defendant for 19 years of wrongful imprisonment.
This outcome has significance beyond this specific case. Recognition that the transcripts given to the 1995 jury were misleading despite the legal safeguards having been properly followed adds weight and urgency to the call to action’s argument that the safeguards are ineffective. However, since the exclusion of the police transcripts from the 2018 retrial occurred incidentally, rather than as the result of a specific appeal, the case has brought no acknowledgement from the law of general problems with the legal handling of poor-quality audio evidence. Current cases continue to rely on the same ineffective safeguards, making further injustice inevitable.
All this indicates a need to amplify the call to action. It is really essential for the courts to recognize that, while understanding spoken language is certainly a common skill, understanding how that skill works, and what factors must be managed to ensure the jury reaches an accurate understanding of indistinct forensic audio, requires highly specialized knowledge (see detailed discussion with many references in Fraser & Loakes, Citation2020). Thus, forensic transcription should be treated by the law, not as a matter of common knowledge, but as a science.
Bringing about that change requires concerted action from multiple branches of linguistics – first, to ensure ready availability of accountable evidence-based methods for providing demonstrably reliable transcripts of poor-quality forensic audio (noting that this cannot be done purely via phonetic analysis, but requires an interdisciplinary approach – see §6); second, to address the legal misconceptions that make it hard for the courts to benefit from reliable transcripts even when they are available. The present paper aims to motivate the necessary action via detailed description of the 1995 trial, and its aftermath. Perhaps some readers will question whether our democracy is well served by the law holding Eastman as “valuable authority” for the legal handling of poor-quality forensic audio used as evidence in criminal trials.
As a brief outline of the argument: Section 2, next, provides an overview of the case. Section 3 outlines the legal safeguards, first from the point of view of the law, then from the point of view of linguistic science. Section 4 gives a detailed analysis of how the safeguards were applied in the 1995 trial, and how they enabled the misleading nature of the police transcripts to escape attention. Section 5 demonstrates how the misleading police transcripts lived on after the trial to influence subsequent reviews of the case, even those not officially examining the transcripts. Finally, there is a brief consideration of what is needed to ensure that future juries listening to poor-quality audio evidence are always and only assisted in their perception by demonstrably reliable transcripts. The paper closes by calling on the discipline of linguistics to give greater attention to finding effective ways to address legal misconceptions about speech and transcription. Before starting, I should disclose that I was the defence expert in the 2018 trial – noting that the present paper is purely about the transcripts used in the 1995 trial, in which I had no involvement.
2. Overview of the case
2.1 Background on the case
2.1.1 The crime, the conviction, the appeal
On 10 January 1989, Colin Winchester, Assistant Commissioner of the Australian Federal Police (AFP) was shot dead in his Canberra driveway, sparking an intense investigation by the AFP. Suspicion quickly fell on David Harold Eastman, a disgruntled public servant with a history of aggression, a grudge against the Assistant Commissioner, and no alibi for the time of the killing. He was charged with the murder in December 1992. The high-profile trial, held in 1995, was long, complex and demanding of extreme judicial patience, as proceedings were continually disrupted by appallingly disrespectful and erratic conduct on the part of Mr Eastman, who, on multiple occasions, abruptly dismissed his legal team and insisted on representing himself.
The prosecution mounted a strong circumstantial case, including evidence from psychiatrists, eyewitnesses and ear-witnesses, as well as a forensic expert who demonstrated that gunshot residue found in Mr Eastman’s car came from the murder weapon. In addition, transcripts of covert recordings featuring Mr Eastman whispering to himself in his bedroom revealed him making numerous admissions related to the murder, including the direct confession “I killed Winchester”, widely reported in the media (e.g. Campbell, Citation1995).
On 3 November 1995, the jury returned a verdict of guilty, and Mr Eastman was sentenced to life in prison. This was a controversial outcome: many felt Mr Eastman had been scapegoated, some noting his behaviour suggested mental illness (Waterford, Citation1995). It was followed by a series of appeals, notably one considered in 1997 by the full bench of the Federal Court of Australia (Eastman v The Queen [Citation1997] FCA 548). The 1997 appeal upheld the conviction, endorsing both the evidence and the trial process, and creating a legal precedent on multiple issues, including some related to the handling of indistinct forensic audio evidence. Nevertheless, concern over the fairness of the trial persisted.
2.1.2 The Martin Inquiry: Investigating persistent doubts reveals miscarriage of justice
In 2012, a “Board of Inquiry” was ordered (this is a high-level public investigation, similar to a Royal Commission). Led by Acting Justice Brian R. Martin, and culminating in the 2014 “Martin Report” (Martin, Citation2014), the inquiry was tasked with investigating a series of persistent doubts regarding the gunshot residue and other evidence admitted in the 1995 trial. These doubts (none relating directly to the audio evidence) were outlined in 19 paragraphs (38, p. 27ff; note that all point/page references are to the Martin Report unless otherwise specified).
The 447-page report (Hamer Citation2015 gives a helpful summary) dispelled most of the 19 doubts. However, those that were confirmed provided incontrovertible demonstration of serious flaws in key evidence relied upon in the 1995 trial. In particular, the gunshot residue evidence was found to have been incompetent and biased, despite having been accepted (indeed praised – see Knaus Citation2014) by the 1995 trial judge and endorsed by the 1997 appeal judges. Having taken into account a range of legal principles (1802ff, p. 436ff), Justice Martin reached the following conclusion:
1831. The applicant did not receive a fair trial according to law. He was denied a fair chance of acquittal. As a consequence, a substantial miscarriage of justice has occurred. (p. 446)
2.1.3 The Martin Report: Miscarriage caused by police mishandling of forensic evidence
Most of the flaws identified by the Martin Report were attributed to overzealous pursuit of Mr Eastman by police, specifically by Detective Superintendent Richard Ninness (Operational Commander of the investigation) and Detective Sergeant Thomas McQuillen (lead investigator). These officers appear to have displayed the kind of tunnel vision now known to be a key factor contributing to wrongful convictions (Gould et al., Citation2012).
1647. […] There is no doubt that by 1990, Mr Ninness possessed an unshakeable belief that [Mr Eastman] was the murderer. (p. 401)
2.1.4 Effects of the Martin Report: Improvements in forensic evidence – but not transcripts
Since the Martin Report, discussion of the Eastman case has been thorough and wide-ranging, and many publications, both popular (Vincent, Citation2019) and legal (Fuller, Citation2020; Hamer, Citation2015; Maxwell, Citation2019), have analyzed its findings, seeking to ensure its flaws are not repeated. In particular, the revelations about the gunshot residue expert contributed to a strong push for reform of the handling of scientific evidence in general (Edmond, Citation2014). This has led to substantial changes in the practices of law enforcement and forensic scientists (Robertson, Citation2014) – though arguably still not enough (Edmond & Martire, Citation2018).
One topic, however, has received surprisingly little critical attention, despite having played an important role in the 1995 trial. That is the audio evidence – the subject of the current paper. As mentioned, a key development brought about by the 2018 trial was recognition that the police transcripts used to assist the 1995 court in understanding the covert recordings had been thoroughly misleading – despite having been endorsed by the trial judge, the 1997 appeal judges and, tacitly, by the 2014 inquiry.
In fact, as I hope to show, the 1995 audio evidence was at least as misleading as the gunshot residue evidence. However, while the flaws in the gunshot residue evidence were (belatedly) recognized, leading to beneficial changes in the handling of this and other forms of scientific evidence, the flaws in the audio evidence have received no official recognition. The present paper hopes to help achieve, even more belatedly, a similar impact for the flaws in the audio evidence, via careful examination of exactly what went wrong in the original trial, why the failings were never properly recognized, and what needs to happen to ensure such flaws are not repeated. We begin by reviewing the Martin Report in more detail, focusing on its account of the audio evidence.
2.2 The audio evidence in the Martin Inquiry: No doubts to investigate
The Martin Report starts by quoting the facts of the 1995 case as summarized in the 1997 appeal judgment, including its account of Mr Eastman’s disruptive and disrespectful behaviour in court, and its review of the evidence presented in the trial. The latter includes an outline of the audio evidence (37, p. 25, quoted from the 1997 appeal judgment, p. 126ff). This outline explains that recordings of Mr Eastman speaking and whispering to himself in his bedroom were collected by hidden listening devices throughout 1990–1991; describes the extremely poor quality of the audio; notes that transcripts were produced by police, a prosecution expert and two defence experts (actually, three defence experts provided transcripts, but only two gave evidence in court, due to untimely changes in court processes); and provides (pp. 26–27) samples of the transcripts demonstrating that the experts expressed agreement with several key admissions transcribed by police (see §5 below). Note that, in order to avoid circulating misleading accounts of the audio evidence, quoting from the transcripts is minimized in this article.
As mentioned, none of the doubts the inquiry was asked to investigate related directly to the audio evidence. However, Paragraph 16 (1423–1720, pp. 339–417) referred to it, reviving a question raised, and rejected, during the 1997 appeal (1997, p. 143; note that all 1997 page references are to the appeal ruling). This doubt suggested that the admissions heard in Mr Eastman’s whispered monologues had been elicited by police harassment, and should therefore have been regarded as coerced confessions, and excluded from evidence in the 1995 trial. In considering this doubt, Justice Martin reviewed the audio evidence, including additional transcripts provided to the inquiry by the AFP (§5.3 below), and commented on its significance in the 1995 trial:
1425. In summing up to the jury the trial Judge directed that if the jury was satisfied that any of the recordings contained specific words which amounted to an admission of killing the deceased, such words were “direct evidence” pointing to the guilt of the applicant. His Honour added that such evidence was not part of the circumstantial evidence and that this “points up the enormous importance” of the tapes, transcripts and oral evidence. (p. 340)
1442. To a large extent the recordings were indecipherable. Significant debate occurred in the presence of the jury as to whether the prosecution interpretation of the recordings was accurate. However, these issues were before the jury and it is not part of my function to comment upon whether the prosecution view of the interpretation of the recordings, and the reliability of any statements attributed to the applicant, was well founded or otherwise. (p. 349)
Justice Martin summarized his assessment of Paragraph 16 as follows:
1663. […] I have no doubt that during the period 1989–1991 police conduct was deliberately aimed at harassing the applicant with a view to upsetting him and provoking him into reacting. (p. 407)
2.3 The audio evidence in the 1995 trial: Some retrospective doubts
2.3.1 Inherently unreliable transcripts
Mr Eastman’s flat was bugged almost continuously over a period of years, creating vast quantities of audio. Much of the content was normally clear speech, readily transcribed; and much more yields at least the gist of its content to careful listening. However, none of the utterances transcribed for these parts was directly incriminating (i.e. seeing any relevance to the murder requires specific contextual information about the prosecution case – §5.1). The explicit admissions of greatest significance in the trial were heard by police in whispered monologues of such extremely poor quality that all commentators, including the 1997 appeal judges (§4.5.2 below), have agreed they are impossible to understand without the assistance of a transcript.
Though the Martin Report does not point it out explicitly, it is evident from the names given in various sections that the audio evidence was handled by the same officers (DS McQuillen under the direction of Commander Ninness) as handled all the other evidence, including the gunshot residue evidence which Justice Martin found to be so highly compromised. As with the gunshot residue and other evidence, police had full control over the audio evidence. Investigators were responsible for monitoring, collecting, processing, enhancing, copying, editing and selecting all the audio used in the trial – undertaking much of the work themselves, while choosing, and directing the work of, experts for specific tasks. This in itself created substantial problems with the audio evidence, but those are beyond the scope of this paper. Here the focus is on the transcripts, which, of course, were prepared by police, led by DS McQuillen. It is interesting to compare the inquiry’s response to this situation to its response to the handling of the gunshot residue evidence. As we have seen, the Martin Report criticized the close collaboration between the expert and the police (1117, p. 282). When it came to transcription of the indistinct audio evidence, however, it was accepted that police were (“ad hoc”) experts in their own right (see §3.1 below).
It is well known that transcription of indistinct audio is strongly influenced by transcribers’ expectations about the content (§3.2 below). Here, DS McQuillen and his team listened to the audio in the strong belief that Mr Eastman was guilty, and, thanks to the psychiatric advice (§2.2), in genuine expectation of hearing him make admissions to that effect – and indeed their transcripts indicate they heard Mr Eastman making many direct and indirect admissions that he had killed AC Winchester.
2.3.2 Their own expert showed the police transcripts were highly misleading
For the 1995 trial, unusually at the time, police commissioned an expert to back up their transcripts. For the audio (unlike the gunshot residue) they chose a genuine expert, a respected and independent scholar of phonetic science with considerable experience in analyzing forensic audio. However, this expert was expected to work under the close direction of police (and later of prosecutors) who controlled what he analyzed, and how (§4.3.2).
Nevertheless, the expert’s transcripts showed many differences from the police transcripts – especially in relation to the utterances taken to constitute explicit admissions. As one example, where the police had “I killed Winchester”, the expert suggested, tentatively, an utterance something like “I kept watching her”. Differences like this are significant. The police version is not a predictable “mishearing” of the kind long studied by linguistic science (e.g. Garnes & Bond, Citation1980). Indeed “I killed Winchester” is one of many “statements” transcribed by police that are directly contradicted by the acoustic information in the recording. This means no transcriber could hear the utterance in this way unless they were led by contextual information to expect utterances related to killing Winchester. The fact that the officers made errors like this, then, is reason not just to correct those specific errors, but to question whether they are sufficiently independent of the case to provide a transcript at all.
In the 1995 trial, however, concerns like this were not considered. The police transcripts were not corrected for the errors noted even by their own expert, let alone by the defence experts. Indeed, the experts were not asked to correct errors, only to provide commentary for the jury to consider – as one of the safeguards intended to protect them from being influenced by potentially misleading transcripts. The next section looks briefly at the nature of those safeguards, before §4 recounts in detail how they were applied in the 1995 trial – and how they failed to protect the jury from the influence of the erroneous transcripts.
3. Background on police transcripts and legal safeguards
3.1 Legal perspective
Forensic recordings admitted into evidence are regarded as a kind of “document” (Evidence Act s 48(1)(c); Odgers, Citation2021, p. 285ff), to be understood by the jury as they do a written text or photograph. Words heard when the recording is played can be considered equivalent to a statement provided by a witness – and it is a strong legal principle that determining the meaning and significance of witness statements is a matter solely for the jury. If forensic audio is clear, and in a language known to the court, a transcript, typically produced by a professional agency, is provided as an official record for the court. Sometimes this transcript is given to the jury, after they have heard the audio, as an “aide memoire” to assist them in their deliberations. With agreement of the parties, the transcript can even be provided instead of the audio.
With poor-quality audio, it becomes necessary for the court to determine not just the meaning of the words, but what the words are. The latter is also considered, by extension, to be a matter solely for the jury – but the poor quality of the audio raises two important issues. First, it changes the role of the transcript from an optional aid to memory to a necessary aid to perception. Second, it is no longer possible for transcripts to be produced by the usual professional agencies. This is the situation in which the law allows detectives from the investigation to provide the transcript. In order to understand the legal perspective on this practice, it is useful to review its origin (see Fraser, Citation2021b, for further detail).
The 1981 New Zealand trial Menzies (see R v Menzies [Citation1982]) was the first in Australasia to admit poor-quality surveillance audio as evidence in court. When the jury found the audio impossible to understand, the prosecution sought admission of a police transcript as assistance to the jury. The judge’s initial response was an unequivocal “no”: the suggestion violated multiple legal principles. Of relevance here, he noted that the transcript represented an opinion as to the content of the audio. In a trial, this kind of opinion can be offered as assistance to the jury only by experts. Since (in the eyes of the law) creating a transcript requires no specialist expertise, there was no basis to admit it as opinion evidence. Further, police are under specific restrictions in relation to offering opinions in court, with their testimony limited (with exceptions not relevant here) to factual evidence.
However, when the judge was persuaded to listen with the transcript, he found it so helpful to his perception that he changed his mind, and did allow the transcript to assist the jury. To enable this, he overruled his own earlier decision, and conceded that producing a transcript of indistinct audio did indeed involve a small degree of expertise, which the police transcriber had gained, in relation to this particular audio (i.e. “ad hoc”), by listening to it over and over again. In providing the transcript, the judge was careful to give the jury a strong instruction to the effect that the evidence was the audio, not the transcript; determining the content of the audio (as well as its significance) was a matter solely for them; they should listen carefully, using the transcript only as assistance. This later came to be called the “aide memoire instruction” (§4.5.2).
All this was upheld on appeal (and indeed the trial judge’s evaluation of the police transcript might well have been right in this particular case). The point is that the appeal decision created a precedent allowing use of police transcripts in the burgeoning numbers of trials featuring evidence in the form of poor-quality surveillance audio. Procedures for doing so were refined via various trials, endorsed by the High Court of Australia in Butera v DPP (Citation1987), and, in the mid-1990s, codified in the Uniform Evidence Acts (Odgers, Citation2021, p. 579). As already indicated, these procedures include a number of safeguards intended to mitigate the risk that a potentially inaccurate police transcript might mislead a potentially suggestible jury.
The most important safeguard, from a legal perspective, is the judge’s “aide memoire instruction” to the jury: the evidence is the audio, not the transcript; the jury should listen carefully and, if their own hearing differs from the transcript, prefer their own hearing. This ensures that the (potentially unreliable) police transcript has no status as evidence in its own right, but is used merely as optional assistance in understanding the audio. The next most important safeguard is the fact that, in the case of a dispute, the judge can listen personally to ensure that nothing potentially misleading is given to the jury. In practice, this happens only rarely, due to the final safeguard. The defence are expected to review the transcript, taking advice, where appropriate, from the defendant (who, it is assumed, knows what was really said in the recording) – and to negotiate with the prosecution to produce an agreed version for the jury. Only if agreement is not possible does the judge listen personally – now commonly also hearing opinions of experts called by one or both parties (before Eastman, this was rare, due to transcription not being considered a field of expertise) – to ensure it is safe to leave resolution of disputed utterances to the jury, with the aide memoire instruction.
From a legal perspective, these long-established procedures are familiar and uncontroversial – to the extent that questioning them can be considered perverse. However, readers of this journal will recognize multiple problems, arising from familiar misconceptions embedded in educated common knowledge about speech and transcription.
3.2 Linguistic science perspective
From a linguistics perspective, the legal procedures just outlined raise multiple concerns (this section summarizes extended discussion in papers such as Fraser, Citation2018; Fraser & Kinoshita, Citation2021; Fraser & Loakes, Citation2020; French & Fraser, Citation2018). Most striking is the concept that police officers gain a form of expertise by listening many times. First, while listening many times is necessary for experts, it does not in itself confer relevant expertise. Second, whatever advantage police may have in understanding indistinct audio from their cases, it does not come from listening many times. If that were all that was needed, the regular transcription agencies could be paid to listen enough times to provide an independent transcript. Clearly, the reason for using police is their access to contextual information about the case, which primes them to hear words not heard by others. However, as is well known, contextual priming is a double-edged sword. While priming with reliable contextual information is necessary for reliable perception of poor-quality audio, unreliable contextual information can be highly misleading – and clearly not all contextual information available to police is fully reliable. This, combined with their lack of genuine expertise, means that police transcripts often contain errors of greater or lesser severity.
Another, less obvious but more serious, issue with the legal procedures is their reliance on the “aide memoire instruction”. The concept that jurors should use a transcript “only as assistance” assumes the transcript will either assist accurate perception or have no effect. This, however, ignores the well-known phenomenon of textual priming, through which an inaccurate or even implausible transcript can misleadingly appear to “assist” in just the same way as an accurate one does. More serious still is a concept that attracts far less attention, namely the idea that lawyers and judges administering the safeguards can evaluate a transcript simply by checking it against the audio.
These and other legal principles are clearly founded in deep-seated misconceptions, well known to linguistic science, that are embodied in educated common knowledge about the nature of speech, and its representation in transcripts – for example, the misconception that speech perception is a simple “bottom up” process of recognizing “sounds” and combining them into words; or the misconception that transcription is a simple transduction of words from spoken into written form, requiring little expertise beyond ability to spell. These ideas promote the misconception that the transcript is essentially equivalent to the audio it represents – a manifestation of the “written language ideology” (Eades et al., Citation2023) that creates serious problems even with clear audio (Haworth, Citation2018), and the problems only get worse with poor-quality audio (Fraser, Citation2022b).
The legal procedures governing the use of transcripts of indistinct audio evidence, then, are flawed in principle. However, they are even more flawed in practice – in ways that, since they happen “behind the scenes” of the trial, are rarely brought to the attention of linguists, even those few who do transcription casework. The next section offers insight by taking a detailed look at how the safeguards played out in practice in Mr Eastman’s 1995 trial.
4. Safeguards applied in the 1995 trial embody serious misconceptions
This section reviews the judicial reasoning applied to safeguard the 1995 jury from being misled by potentially inaccurate police transcripts (as recounted in the 1997 appeal ruling). From a legal perspective, this reasoning was and remains uncontroversial: only one aspect was queried at the appeal, and that was upheld (§4.5.1). However, from a linguistics perspective it incorporates misconceptions which make the legal safeguards ineffective – and indeed, as mentioned above, it has now been acknowledged that the police transcripts used in the 1995 trial were thoroughly unreliable, despite the safeguards having been applied in accordance with legal principles, as confirmed by the appeal ruling.
4.1 Police had listened many times
The legal concept that police achieve “expertise” as transcribers by listening many times gains little support from linguistic science (§3.2). However, the judges in Eastman, as in many other trials, seem to have placed genuine faith in it, and were at pains to emphasize that the police transcribers had fulfilled it:
the transcripts … had been produced by Sergeant McQuillen and Constable Lawson, the police officers who spent literally thousands of hours listening to the tapes as part of their duties in electronic surveillance, and in the preparation of the transcripts. (1997, p. 126, ll. 23–25)
the fact that Mr McQuillen and Ms Lawson had each spent enormous periods of time working on deciphering the tapes and swore that they independently arrived at the interpretations reflected in their transcript […]. (1997, p. 203)
This misplaced confidence in “ad hoc expertise” meant the additional safeguards discussed below were applied from an unjustified baseline assumption that the police transcripts were probably mostly right.
4.2 The accused was unable to offer a credible alternative
In Eastman, as in many cases, it was assumed that, if the police transcript was wrong, the accused could readily correct it simply by stating what he really said. Inability to remember was not an excuse, since “the recordings themselves should have provided an aid for him to refresh his memory” (1997, p. 203). On this reasoning, failure to provide a credible alternative gave implicit support for the police version:
[…] the jury would be justified in treating [the admissions in the police transcripts] as incriminating in the absence of evidence from the appellant which raised, at least as a possibility, an innocent explanation […]. (1997, p. 203)
Even if the accused does provide an alternative version, it has a second hurdle to pass: that of being found more credible than the prosecution version. In the 1995 trial, Mr Eastman did offer alternative interpretations for some utterances:
1425. In his evidence to the jury the applicant drew upon the poor quality of the recordings and submitted that many of the sections relied upon by the prosecution as confessions were indecipherable. He offered innocent interpretations for some of the more incriminating statements. (p. 342)
4.3 Expert opinion (apparently) supported the police transcripts
4.3.1 Over-emphasis on a few partial similarities between police and experts
Like the prosecution, the defence took up the option of commissioning expert transcripts, also choosing experienced, independent experts. All the expert transcripts, on both sides, showed numerous differences from the police version. Sometimes they offered alternatives as substantially different from the police version as “I kept watching her” is from “I killed Winchester”. Other times they cast doubt on apparent confessions in the police version by bracketing words to signal their uncertainty.
Unfortunately, though unsurprisingly, the experts often differed from each other in the alternatives they (tentatively) suggested. From a linguistics perspective this would be seen as further confirmation of the very poor intelligibility of the audio, drawing attention to the need for extra care in ensuring that priming and bias were managed appropriately. However, from a legal perspective it seems to have caused disregard for the many substantial differences between the police and expert versions, putting the focus on the few cases of partial agreement among all the transcripts, with the appeal judges going so far as to say that the prosecution expert had “verified” the police transcript (1997, p. 126, l.23).
Although there are many differences in the four transcriptions, the Crown claims that it can draw substantial support from the similarities. (1997, p. 126, ll. 49–51)
In the 1995 trial, however, there is an even stronger explanation for the (limited) similarity between the expert and police transcripts.
4.3.2 Failure to recognize real cause of apparent similarities between police and experts
While the experts were certainly independent in the sense that they had no personal interest in the outcome of the case, they were not independent of the police transcripts. That is because experts on both sides provided their commentary by editing the police version, as instructed by their clients.
The problem with this editing method is seen in the fact that the prosecution expert initially accepted “I killed Winchester”. Although he later recognized this transcription as inaccurate, by then it had influenced both the court and the media – which continued to publicize the alleged confession despite his change of mind. It is hard to know whether or which other parts of his transcript might have been affected, but it is notable that, as mentioned earlier, the fresh transcript he produced for the 2018 trial was far further from the police versions than his 1995 transcripts had been.
The problem was even more forcefully brought to the attention of the senior defence expert, when he found that a page of the police transcript he was editing was missing. As it took some time for the missing page to be provided, he went ahead and transcribed that section independently. When the missing page eventually arrived, he realized that his level of agreement with the police version was far lower for that section than for others. This caused him to suspect his earlier work might have been more influenced by the police version than he had realized (Butcher et al., Citation1995). Later it was discovered there was a whole tape he had not yet transcribed, and he insisted on producing an independent transcript of that. Again, his transcript showed far less agreement with the police version than the other experts’ transcripts did (see §5.1). The defence raised these discoveries in the 1995 trial. However, at the time there was little phonetics research directly backing up this expert’s experience (indeed he noted that, as a phonetician, he had been unaware till this experience of just how much a transcript could influence his own perception). His uncertainty appears to have made a poor impression – with the result that his (valid) misgivings had little impact (they are mentioned in neither the 1997 appeal ruling nor the Martin Report – see §5).
During the 1997 appeal (p. 203), the defence made a submission (not an official “ground” of appeal) to the effect that the prosecution expert’s evidence may have been “contaminated” by the editing process. However, the appeal judges dismissed this idea on the grounds that it gave insufficient recognition to the many times the police officers had listened (§4.1), or to the prosecution expert’s independence and expertise.
In closing this section, it is important to emphasize that, despite the editing methodology, all the experts, on both sides, found egregious problems with the police transcripts.
4.4 The trial judge listened personally
Since the defence challenged the police transcripts, the 1995 trial judge followed the safeguard of listening personally to ensure they were not potentially misleading to the jury. As in the Menzies appeal (§3.1), the fact that a neutral, responsible judicial listener noted no substantial errors seems to have been particularly salient to the appeal judges – indeed, in general it is very rare for a transcript that has been reviewed by a judge to be called into question (§5).
His Honour [the trial judge] listened to the tapes, and he was aware that there was disagreement between experts in phonetics who had been instructed by the Crown and the defence. The trial judge took the quality of the tapes into account as one of the factors to be considered in the exercise of his discretion whether to exclude the tapes on the ground of unfairness. His Honour said: “having listened to the tapes and read concurrently the transcripts of them I must say that with some exceptions which may well have been irrelevant it was, I thought, not difficult to comprehend with the benefit of the transcripts what was being said. In some rare cases it is possible to detect what is being said without reliance upon the transcript itself … ”. It was not suggested that this was an erroneous assessment of the tapes. The trial judge ruled that he should not exercise his discretion to exclude the evidence on the ground “that it would be dangerous to go to the jury”. (1997, p. 203, quoting 1995 trial judge)
4.5 The jury were able to listen for themselves
From a legal perspective, the ability of the jury to listen to the audio directly, and reach their own conclusion as to its contents, is the most important safeguard (§3.1) – as was emphasized explicitly by the appeal judges:
[…] in so far as the interpretation of the sounds was to be challenged by the appellant, the jury itself could listen to tapes [sic]. Unlike evidence of an out of court identification by a witness, this evidence was capable of independent evaluation by the jury by listening to the tapes itself. (1997, p. 203)
Further, it was not just exposure to erroneous transcripts that mattered. The conditions under which the jury listened to the audio made it extremely unlikely that transcript errors would be detected and corrected.
4.5.1 Police version first – expert versions second and secondary
The first time the 1995 jury heard the audio, they had the police transcript in front of them – “assisting” their perception as they listened. The expert opinions were provided later, not as transcripts for the jury to follow, but as commentary for them to consider. Each expert, sequentially, spelled out each difference between their transcript and the police version, line-by-line, via tedious, confusing, question-and-answer examination and cross-examination by the barristers – with many interruptions, mix-ups regarding which version of the transcript everyone had (if any), which page they were up to, etc. This method of presentation is unlikely to reduce the overall influence of the police version followed initially.
It is worth emphasizing that this approach did not arise from disrespect for the experts, who were all praised for their independence and scholarship. However, the outcome was that the experts’ opinions were effectively ignored, except to the minor and misleadingly teleological extent that their transcripts appeared to support the police transcripts (§4.3.1).
4.5.2 The jury never listened without a transcript
The decision of the trial judge to allow the police transcript to be provided the first time the jury listened was the only substantial issue about the audio evidence raised at the 1997 appeal, as Ground 11. The defence argued that the jury should have heard the audio first without any transcript, as in the procedure endorsed by the 1987 High Court ruling in Butera.
It was submitted that the jury should have first listened to the tapes unaided by the transcript or Mr McQuillen’s evidence, and only then been given the transcripts as an aide-memoire for use thereafter in deciphering the tapes. (1997, p. 200, l. 12ff)
The poor quality of recording [sic] would have made it very difficult for the jury to have comprehended the recorded utterances if it had listened to the tapes unaided by the transcript prepared by Mr McQuillen. It seems to us that it would have been a largely pointless exercise to have required this, and one that would have been immediately followed by what in fact occurred, namely that the jury was informed of Mr McQuillen’s interpretation of the sounds, and then the tapes were played while the jury had Mr McQuillen’s transcript before it as an aid. (1997, p. 202, ll. 8–14)
It is particularly interesting to notice that the judges’ reasoning (listening to the tapes unaided would be “a largely pointless exercise”) implicitly acknowledges that the transcript contains information not available in the audio itself – without questioning the legal concept that the transcript has no status as evidence, to be established as reliable in its own right (§6.2).
4.6 The effect of the safeguards: Legal procedures, upheld on appeal, actively promoted inaccurate interpretation of crucial evidence
If the arguments in this section are valid, the procedures followed in the 1995 trial, and upheld by the 1997 appeal, not only failed to safeguard the jury from exposure to erroneous police transcripts, they actively privileged the erroneous police transcripts. This is certainly not to suggest the trial judge intended this effect. He surely expected, based on established legal concepts, confirmed by his own listening to the audio in light of advice from experts on both sides, that the police transcripts were reliable enough to assist the jury in making out the true content of the indistinct audio (with the aide memoire instruction as a final safeguard in case his own views might be incorrect). Nevertheless, it is important to recognize the practical consequences of the procedures he followed.
While it is not possible to know in detail what the jury made of the audio evidence, the considerations outlined above make it likely they would have accepted at least some parts of the erroneous police transcripts as more reliable than they really were – as indeed the trial judge himself had done openly and explicitly (§4.4). And the aide memoire instruction to “listen carefully” is unlikely to have helped: recall that even the experts did not catch all relevant errors despite their careful, repeated listening (§4.3).
Given the inevitable influence of a transcript on perception of poor-quality audio, it is worth recalling (§3.2) that police transcripts arise not from any real, if “ad hoc”, expertise on the part of investigators, but from their contextual understanding of the case. This creates a dangerous circularity. The transcripts cause listeners to feel they are hearing admissions directly “with their own ears”, as independent evidence supporting the prosecution case – with insufficient recognition of the role the prosecution case has played in creating the transcripts that “assist” their hearing.
Most importantly, this privilege for the police opinion occurred despite experts on both sides pointing out substantial errors in the police transcripts. As demonstrated above, the safeguards give more weight to the perception of lawyers, judges and juries than to the opinion of experts. And since, in Eastman as in many other trials, the perception of the lawyers, judges and juries had already been heavily influenced by the police transcript before they were even made aware of the experts’ opinions, it was nearly impossible that errors in the police transcripts could have been properly corrected during the trial. Indeed, as demonstrated in the next section, the erroneous police transcripts lived on unchallenged to influence further proceedings, including the 2014 Martin Inquiry.
Before moving on, it is worth pausing to address a possible response to this case study: “things are not that bad any more”. It might be suggested that the quality of the 1995 police transcripts was unusually poor – perhaps the “cop-killer” nature of the crime contributed to this – and Mr Eastman’s appalling behaviour in court, despite the extreme forbearance of the trial judge and others, did nothing to help his cause; anyway, the handling of audio evidence has improved since 1995. The last point is certainly true. Awareness of the priming effect of inaccurate transcripts has increased both in the law and in forensic phonetics. As a result, it may be that sole reliance on police transcripts is decreasing, with more consultation of independent experts to resolve disputed utterances. However, the overall process of the 1995 trial, in which police, lawyers and judges manage the selection of experts and the conditions under which they work, and ultimately evaluate all transcripts, including those of experts, by checking them personally against the audio, remains in place. Finally, it is worth reflecting that, unusually poor as the 1995 police transcripts might have been, they were nevertheless able to pass all the legal safeguards intended to protect the jury from misleading “assistance”.
5. Beyond the jury: How unreliable transcripts lived on after the trial
The discussion so far demonstrates that the legal safeguards, properly applied and incorporating advice from experts, were unable to prevent misleading police transcripts being provided as assistance to the 1995 jury in understanding the poor-quality covert recordings. The present section details how the inaccurate 1995 police transcripts lived on, potentially influencing later decision-makers, even those not directly concerned with the audio evidence. This makes Eastman an excellent example of the “textual travels” enabled by problematic language ideologies (Heffer et al., Citation2013) – though forensic transcription arguably raises issues not fully addressable through an “ideologies” lens (§6).
It is useful to start by recalling two essential principles of legal procedure (§3.1). First, the evidence is the audio – the transcript is merely optional assistance to the jury in understanding the evidence. Second, determining the content of the audio is a matter solely for the jury – who are urged via the aide memoire instruction to reject any part of the transcript they do not agree with. These two principles surely entail the possibility that the jury might reject all or part of any transcripts provided as assistance.
The clear implication of these strong legal principles is that no one but the jury knows whether or to what extent the transcripts were accepted, or what the audio content was ultimately determined to be. This might create the expectation that, in discussing the case, reference to transcripts used to assist the jury would be made tentatively, if at all. That expectation is violated in Eastman, as in other cases (Fraser, Citation2018).
5.1 The transcripts in the 1997 appeal
In recounting the evidence presented in the 1995 trial (§2.2), the 1997 appeal judges quoted two sets of short excerpts from the transcripts – each set citing first the version produced by the ad hoc experts and then those of the genuine experts. There is no indication of how these particular excerpts were chosen. Perhaps they were the ones the appeal judges found most salient in reaching their decision. Whatever the reason, these brief decontextualized excerpts became the de facto “official record” of the voluminous covert recordings, widely cited in discussion of the case (cf. Haworth, Citation2018). It is interesting to note that reliance on this official record divorces the transcripts from the audio, which is unavailable to readers of the appeal ruling. This is somewhat ironic, considering the foundational legal principle, of which judges carefully remind juries, that the audio is the evidence, with the transcripts provided only as assistance, which the jury is invited to reject.
However, the effect goes beyond irony when we consider just how misleading an impression the quoted excerpts give of the audio evidence as a whole. The first set shows one of the highest levels of agreement between the ad hoc and genuine experts’ transcripts in the whole corpus – bolstering the false impression that the experts largely agreed with the police transcripts (§4.3.2). However, the “admissions” represented in these excerpts are extremely indirect (§2.3.1). Seeing any relevance to the crime requires accepting prosecution speculation about the content of Mr Eastman’s internal ruminations regarding his relationship to the authorities (see Georgiou, Citation2019).
The second set includes one of the most directly incriminating admissions in the police transcript (second only to “I killed Winchester”) – but shows far less agreement between the versions. In fact, the real level of agreement is even lower than suggested by the quoted transcripts. This is one of the sections of audio that the senior defence expert transcribed without having seen the police transcript (§4.3.2). His transcript specifically rejects the incriminating phrase (and it is worth noting that recent experimental evidence (Fraser, Citation2021a) demonstrates this phrase is rarely if ever heard before it is suggested, particularly in the “enhanced” version of the audio preferred by the 1995 trial). However, this expert’s commentary is not noted in the 1997 appeal ruling, and his transcript is silently omitted from the set provided as the “official version”. The result, naturally enough, is that many commentators cite the incriminating admission in the official version, apparently not questioning why there is only one defence transcript here, when two are given for the other excerpt (e.g. Hamer’s very balanced recount of the prosecution case (Citation2015, p. 438) notes “even the defence experts [sic] accepted that Eastman had ‘probably' said [the incriminating words]”).
It is hard to over-estimate the influence that seeing “official” excerpts like these can have on readers, especially on those who know and trust the legal safeguards intended to ensure that transcripts provided to the jury are not misleading. No matter how many reminders are given that the audio was indistinct, the well-established “continued influence effect” (e.g. Ecker et al., Citation2010) makes it unlikely that even the most responsible of readers could avoid being lastingly affected by them (see in particular Wistrich et al., Citation2005).
5.2 The transcripts in the Martin Report
Recall that Justice Martin’s 19-paragraph brief did not request him to look into the audio evidence, beyond considering the Paragraph 16 doubt (§2.2). However, the audio did play a role in the conclusion the Martin Inquiry finally reached (§2.1.2). It is listed as part of the evidence taken into account in determining what verdict the jury might have reached if the flawed evidence had been omitted from the 1995 trial:
1793. Statements by the applicant covertly recorded when he was talking to himself in his residence were capable of being viewed as admissions of guilt of the murder. The quality of the recordings was, at times, particularly poor, but it was open to the jury to accept that the applicant made statements which the prosecution relied upon as admissions to killing the deceased. (p. 434)
Here, again, then, we see the police transcripts living on beyond the trial. Again, there is no opportunity for readers to review the audio (the real evidence, according to the law). This time, however, there are no expert versions of the transcripts for readers to compare with the police versions – and no indication that almost every “statement” in the AFP summary was opposed to some degree, not just by the defence but by the prosecution’s own expert. Worse still, in copying the “statements” out of the police transcripts, numerous editorial errors were made. For one example, two “statements” attributed to separate tapes are actually alternative transcripts of the same section of audio, one by the police and one by the prosecution expert (testament to how different these often are). In other examples, punctuation marks – including the brackets used by the experts to signal their uncertainty (Section 4.3.1) – are omitted or altered.
Please note there is no suggestion here that anyone involved in the Martin Inquiry intended the “summary” to mislead. All the above observations are readily explained by misconceptions about the nature of transcripts (§§3.2, 6.2). Nevertheless, it is impossible to avoid the observation that the inquiry gained a deeply misleading impression of Mr Eastman’s whispered confessions. It is necessary, then, to consider what effect this might have had on Justice Martin’s conclusions.
5.3 What made Justice Martin “fairly certain” of Mr Eastman’s guilt?
It is notable that Justice Martin’s major reason for concluding that the 1995 conviction should be quashed (§2.1.2) was procedural unfairness uncovered by the inquiry (1801, p. 436) – as opposed to a strong sense that the 1995 guilty verdict might have been factually wrong.
1836. I am fairly certain that the applicant is guilty of the murder of the deceased, but a nagging doubt remains. (p. 447)
1800. If an accurate complexion is placed upon the evidence of [an earwitness] and the strong opinions of [the gunshot residue expert] are ignored, and if regard is had to the police harassment, nevertheless a strong circumstantial case remains if the recorded statements are included. (p. 436, emphasis added)
Questions like these, of course, are impossible to answer. The point of raising them is not to suggest one answer or another. Indeed, in a very real sense the answers no longer matter, as Mr Eastman has now been fully compensated for his wrongful conviction. The point of raising them, rather, is to highlight a deeper question underlying them: why was the 2014 inquiry ever exposed to inaccurate and misleading transcripts of crucial audio evidence in the first place? Without these, there would be no need to wonder what influence they might or might not have had on its findings. And this raises an even more fundamental question – why did the inquiry’s brief contain no doubts about the audio evidence among its 19 paragraphs? The incompetent ad hoc expert transcripts were at least as misleading about the audio evidence as the incompetent expert report was about the gunshot residue evidence – which had also been before the 1995 jury, having also been admitted in full compliance with legal procedures, as upheld by the 1997 appeal. Yet the fate of the two types of evidence could not be more different. The gunshot residue evidence was subject to diligent critique by Mr Eastman’s defence team over a period of years, culminating in a well-supported doubt for the Martin Inquiry to investigate. Thus it was eventually found to be severely flawed – contributing not only to this specific conviction being quashed, but to concerted efforts to reform the handling in general of scientific evidence in criminal trials (§2.1.4).
The audio evidence, by contrast, was subjected to no critique after the failure of the half-hearted submission to the 1997 appeal (§4.3.2), and no doubts about it were submitted to the inquiry (§2.2). So the problems it had caused for the trial were neither revealed nor rectified, and there was no call to reform the practice of allowing transcripts by ad hoc experts, with ineffective safeguards, to “assist” the jury in understanding poor-quality forensic audio. Far from it: the Martin Inquiry’s lack of concern about the audio evidence was taken as further endorsement of the standard legal procedures for admitting audio evidence, and the model developed in Eastman for incorporating expert opinion into the trial remains in use to this day, enabling demonstrably inaccurate transcripts to “assist” juries in understanding forensic audio, even in the face of justified expert critique.
One response linguists might have to this situation is a sense of despair. However, there is a more optimistic option. The first step is to avoid blaming the law, which is after all operating on the basis of misconceptions that are widespread in society, even among the highly educated. The next step is to recognize that linguists, at least in Australia, have the ability, and thus arguably the responsibility, to challenge misconceptions about language and speech that threaten the right to a fair trial in our democracy – as has been done effectively in other branches of forensic linguistics (Eades et al., Citation2023). Preparing an effective challenge requires careful analysis of the exact nature of the misconceptions – and Eastman’s juxtaposition of gunshot residue and audio evidence offers a useful lens.
5.4 Why was the audio evidence treated differently from the gunshot residue evidence?
Gunshot residue evidence, like most other scientific evidence, is opaque to the court. Understanding it is self-evidently beyond the reach of common knowledge, leaving no choice but for the jury to be assisted by expert opinion in the form of a report explaining relevant scientific findings. The law strongly prefers evidence that jurors can observe and weigh for themselves, according to the mores of society. Scientific evidence forces the court to rely on the opinion of a third party – and rules of admissibility are designed to mitigate the risk that the third-party opinion might be misleading. Since judges cannot directly analyze evidence like gunshot residue, these rules require them to assess the scientific standing of the methods used by the expert, and the expert’s accreditation in their use. At the time of the Eastman trial, the courts were still in the grip of serious misconceptions about scientific method (Cordner et al., Citation2020; Julian et al., Citation2021), contributing to the gunshot residue evidence being admitted despite its flaws. That situation is now improving, thanks to increasing scientific literacy among the judiciary (Maxwell, Citation2019), generated by a concerted response from the scientific community, in Australia and internationally, to cases like Eastman. While it is not clear that judicial education holds all the answers, the recent exoneration of convicted baby-killer Kathleen Folbigg offers an indication of the value of this united approach (Chin, Citationin press).
Poor-quality covert recordings are just as opaque to the court as gunshot residue. They too require the assistance of expert opinion, in the form of a transcript. However, there are two key differences. First, a transcript can give listeners an immediate and compelling personal experience of being able to hear the words in the transcript “with their own ears”. To the law, this seems like a good thing, enabling the jury to evaluate the audio evidence directly, rather than relying on the opinion of a third party. The problem is, as we have seen, that the jury does in fact rely heavily, though unwittingly, on the opinion of a third party (the transcriber). If that opinion is wrong, jurors’ perception will very likely be misled – potentially contributing to misleading interpretation of other evidence (cf. Fraser, Citation2018).
The second difference is that, while lawyers and judges often have low confidence in their own scientific knowledge, they typically have high confidence that their own educated common knowledge enables them to evaluate transcripts by listening personally. As we have seen, however, lawyers and judges are not immune to being misled by an inaccurate transcript, even against the advice of well-qualified experts. This creates the risk they might allow the jury to be assisted by an inaccurate transcript. If the transcriber is a police officer, the jury will be assisted by the inaccurate opinion of a police officer – the very prospect that caused the judge in Menzies initially to give an unequivocal “no” to the suggestion of using a police transcript (§3.1). Worse still, subsequent reviewers are unlikely to question the perception of a judge, particularly after it has resulted in a legal decision and become the “official version” of the audio evidence (§5.1).
These differences shed light on the question of why no doubts about the audio evidence were submitted to the Martin Inquiry. They also help elucidate the underlying cause of problems with audio evidence – a valuable precursor to finding effective solutions.
6. Finding an effective solution requires identifying the underlying cause of the problem
6.1 The problem with police transcripts is not just police transcripts
The police transcripts used in the 1995 trial were misleading in much the same way as the report on the gunshot residue – both being opinions of incompetent and biased “experts”. However, the errors in the gunshot evidence report were eventually corrected (albeit via a process that took 19 years) when the Martin Inquiry accepted the opinions of genuine experts, and acknowledged flaws in the legal handling of scientific evidence. With the audio evidence, the court had the opinions of genuine experts right from the start – but the misleading effect of police transcripts has still not been fully acknowledged.
This shows that the key problem with audio evidence is not the fact that police provide potentially misleading transcripts. A far more important problem is the fact that all transcripts are evaluated by lawyers and judges, in the belief that their own educated common knowledge enables them to safeguard the jury (§3.1). The reason, as discussed at length, is that educated common knowledge embodies the misconception that careful responsible listeners hear what is “really there”, avoiding the influence of an inaccurate transcript. In order to solve the problems created by use of police transcripts, then, it is not enough just to provide better transcripts. As Eastman shows, even if these are factually more accurate, they may not be accepted by the court. It is really essential to prevent the court being influenced by potentially inaccurate transcripts (whether these are produced by police or anyone else – the reason to focus on police transcripts is not that they are uniquely bad, but that the law privileges them in the ways we have discussed).
The solution, then, is to ensure that forensic audio used as evidence in criminal courts is always and only accompanied by a demonstrably reliable transcript, from before the start of the trial process. The key to achieving this is for forensic transcription to be treated by the law, not as a matter of common knowledge, but as a science – but not just in the sense that prosecution and defence can call individual experts to offer opinions about disputed transcripts. That leaves the whole process in the hands of lawyers and judges, who are given responsibility for deciding when an expert opinion is needed, choosing the expert, managing the conditions under which they work, and evaluating the opinion they provide – all with the aim of satisfying criteria determined more by legal precedent than by scientific principle.
Forensic transcription should be treated as a science (specifically, as a branch of the human and linguistic sciences) in the sense that the discipline has responsibility for developing and managing accountable, evidence-based methods for producing demonstrably reliable transcripts of poor-quality forensic audio – exactly as other reputable forensic sciences have responsibility for developing and managing evidence-based methods that produce reliable results in their own fields. Importantly, on this scenario the judge would determine the admissibility of a transcript, not by evaluating it personally against the audio, but by evaluating the scientific standing of the method used to produce it. Again, this is exactly what happens with other reputable forensic sciences: lawyers and judges do not personally examine gunshot residue, fingerprints, blood samples, cloth fibres, and so on.
Achieving recognition that forensic transcription is a science involves changing long-established legal procedures. This is clearly a matter for the law. However, linguists can contribute via two parallel actions. The first is to ensure ready availability of accountable evidence-based methods for producing reliable transcripts of poor-quality forensic audio. It is often assumed – by lawyers, forensic scientists and even some linguists – that all that requires is acoustic analysis by a speech scientist, or even an engineer. However, while acoustic analysis can be helpful, on its own it cannot guarantee a reliable transcript (see detailed discussion in Fraser, Citation2022a).
Forensic transcription is a complex topic requiring specialized knowledge developed by multiple branches of linguistic science. Importantly, though, while linguists commonly create transcripts of many different kinds, forensic transcription is uniquely challenging in its purpose and constraints (Fraser, Citation2022b). To see this, consider how few linguists are experienced in transcribing extremely poor-quality audio of unknown content under the influence of potentially misleading contextual information – let alone of providing an objective evaluation of such a transcript in the absence of external information about what was really said.
Providing reliable methods for forensic transcription, then, requires repurposing and extending, rather than merely applying, existing scientific knowledge – a project currently being undertaken by the Research Hub for Language in Forensic Evidence. Importantly, however, no matter how successful this endeavour is, it cannot be enough on its own. That is why the second development is so crucial: the discipline of linguistics must find effective ways to address the misconceptions that underlie current legal procedures for providing transcripts to assist juries in understanding poor-quality forensic audio.
6.2 Addressing legal misconceptions about forensic transcription
It is natural for linguists to dismiss the misconceptions of common knowledge as simply “wrong”, and concentrate on providing accurate information. However, laying accurate information on top of misconceptions can backfire. As one example, explanations of priming are frequently misunderstood as bias. This can cause lawyers and judges to take on the task of ensuring that juries and experts are not “biased”, for example by withholding transcripts from juries, or contextual information from experts. This misses the point that reliable perception of poor-quality audio actually requires priming – the problem is not priming, but priming with unreliable information. More importantly, it avoids the real issue: the fact that lawyers and judges themselves can be influenced by unreliable contextual and textual priming (§3.2).
Addressing misconceptions effectively requires careful analysis to uncover their underlying false beliefs, and find ways to disrupt those. For one example, the call to action arose from recognition that misconceptions about “ad hoc” experts were underlain by a deeper belief that responsible listeners, especially judges, can evaluate transcripts by listening personally. Once identified, this belief is easy to disrupt, simply by demonstrating that judges (like anyone else) can be unwittingly misled by an inaccurate transcript. Of course, it is essential for demonstrations like this to be provided in a respectful and constructive manner, focusing on how linguistic science can help the law achieve its own aim, of ensuring juries are provided with accurate, probative evidence capable of assisting them to return appropriate verdicts.
Another useful approach to disrupting the misconceptions underlying problematic legal procedures is to show how they generate internal contradictions. Several have been noted throughout this paper. Perhaps the most obvious is the concept that police ad hoc expertise comes from listening many times: if that were all it took there would be no reason to use police, specifically (§3.2). Another example is the quoting of transcripts admitted in court despite the strong legal principle that acceptance of the transcript is a matter solely for the jury (§5). Bringing attention to this contradiction can help demonstrate that it is really lawyers and judges who accept the transcript, with no one really expecting the jury to reach a substantially different understanding of the audio. So, while the phrase “before the jury” might seem to imply it is open to the jury to reject the transcript, a better interpretation is that the transcript has been cleared by the judge as appropriate assistance for the jury. A final example here is the concept that a transcript has no standing as evidence – in contradiction of the fact it is clearly intended to influence the jury’s perception (§4.5.2). Again, the explanation for this contradiction (which allows police transcribers to escape the scrutiny to which genuine experts are subjected) is that a transcript, once it has passed the safeguards, is expected to have only beneficial influence (this is the misconception underlying the otherwise surprising lack of scrutiny for the AFP summary of the Eastman transcripts accepted by the Martin Inquiry – §5.2).
7. Conclusion
The discovery that the courts use police transcripts, on the grounds that transcribing indistinct forensic audio requires no expertise beyond listening many times, is confronting to Australian linguists, who quickly saw the clear risk to justice, and readily supported the 2017 call to action (§1). However, while the problems raised by the call to action have been acknowledged within the law, a good solution has not yet been implemented. This is because the natural response, even from linguists (including my former self), is to recommend that transcripts by ad hoc experts be reviewed by genuine experts. However, while this can help sometimes, Eastman and other cases show that merely making expert opinions available to the courts is not enough. Indeed, it arguably makes things worse, requiring in-court debates that can be ineffectual despite their very high cost.
The point is that police transcripts are not themselves the key problem. Their use is merely one symptom of a deeper problem, namely the fact that legal procedures for using forensic audio as evidence are designed and managed on the basis of flawed “common knowledge” and unhelpful ideologies of language. Perhaps the present paper can be seen, then, as a call to action on the discipline of linguistics itself to find more effective ways of challenging misconceptions about spoken language and its representation in transcripts that are embodied in educated common knowledge, particularly as these affect powerful institutions including, but not limited to, the law. Achieving this needs concerted action across multiple branches of linguistics and forensic linguistics, and a specialized approach to science communication, that seeks not to add new information to existing common knowledge, but to identify and disrupt deep-seated false beliefs that plague our society more broadly (cf. Burridge, Citation2022). This work can fruitfully join forces with other branches of forensic linguistics. One question to explore, for example, is how misconceptions such as those discussed here interact with the language ideologies identified as the cause of problems with the use of transcripts in other areas of the law (e.g. Smith-Khan, Citation2022). Importantly, however, use of language as forensic evidence in criminal trials also needs insight from forensic science, to ensure that expert opinions are evaluated and expressed appropriately (Julian et al., Citation2021).
Whatever, the path forward, the first step is for linguists to gain detailed insight into how legal processes for handling transcripts of forensic audio in criminal trials operate behind the scenes. I hope the present paper has contributed to that insight.
Acknowledgements
This article has been several years in the writing and has benefited from discussion with many individuals, including: Alex Bowen, Kate Burridge, Andy Butcher, Stephen Cordner, David Deterding, Diana Eades, Gary Edmond, Peter French, George Georgiou, Peter Gray, Simon Moodie, Jeremy Robson, Roger Shuy, Laura Smith-Khan and Michael Stanten. Since I have not always taken their advice, it is more true than usual that remaining failings are the responsibility of the author. Thanks are also due to two very helpful anonymous reviewers.
Disclosure statement
No potential conflict of interest was reported by the author.
Data availability statement
This paper is based on the referenced resources available in the public domain.
Correction Statement
This article has been corrected with minor changes. These changes do not impact the academic content of the article.
Additional information
Notes on contributors
Helen Fraser
Helen Fraser studied linguistics, specializing in phonetics, at Macquarie University and the University of Edinburgh, then worked for many years at the University of New England, NSW, where her theoretical and practical interests revolved around the representation of spoken language – both internally, in the minds of language-users, and externally, in writing and transcripts. Casework experience, starting in the 1990s, led her to research specifically on forensic transcription. In 2017, she instigated a call to action, led by the Australian Linguistic Society and endorsed by three other national organizations, asking the judiciary to review and reform problematic legal practices in relation to transcription of poor-quality audio used as forensic evidence in criminal trials. Following more than a decade as an independent researcher and consultant, in 2020 she became the inaugural Director of the Research Hub for Language in Forensic Evidence at the University of Melbourne.
References
- Burridge, K. (2022). Truthiness and language: Popular perception and fall-out. In R. Dixon, & I. Storch (Eds.), The Art of language (pp. 17–41). Brill. https://doi.org/10.1163/9789004510395_003
- Butcher, A., McClelland, E., & Mills, C. (1995, December 7–9). Buggers in evidence: Covert recordings, transcription and auto-suggestion. 10th Annual Australian language & speech conference, Adelaide, Australia.
- Campbell, R. (1995). Eastman is guilty. The Canberra Times, 4(November), 46–47.
- Chin, J. (in press). What can law take from science after the pardon of Kathleen Folbigg? Law Society Journal (NSW), (August), https://doi.org/10.2139/ssrn.4556224
- Cordner, S., Bruenisholz, E., Catoggio, D., Chadwick, P., Champion, J., Davey, A., Kogios, R., Williams, M., & Woodford, N. (2020). The Uniform Evidence Act and Australian judges’ ability to assess properly the validity and reliability of expert evidence. Australian Journal of Forensic Sciences, 52(3), 243–245. https://doi.org/10.1080/00450618.2020.1741762
- Eades, D., Fraser, H., & Heydon, G. (2023). Forensic linguistics in Australia: Origins, progress and prospects. Cambridge University Press.
- Ecker, U., Lewandowsky, S., & Tang, D. (2010). Explicit warnings reduce but do not eliminate the continued influence of misinformation. Memory & Cognition, 38(8), 1087–1100. https://doi.org/10.3758/mc.38.8.1087
- Edmond, G. (2014). The “science” of miscarriages of justice. UNSW Law Journal, 37(1), 376–406.
- Edmond, G., & Martire, K. (2018). Antipodean forensics: A comment on ANZFSS’s response to PCAST. Australian Journal of Forensic Sciences, 50(2), 140–151. https://doi.org/10.1080/00450618.2017.1340520
- Fraser, H. (2018). Forensic transcription: How confident false beliefs about language and speech threaten the right to a fair trial in Australia. Australian Journal of Linguistics, 38(4), 1–21. https://doi.org/10.1080/07268602.2018.1510760
- Fraser, H. (2021a). Enhancing and priming at a voir dire: Can we be sure the judge reached the right conclusion? Australian Journal of Forensic Sciences, 53(4), 224–251. https://doi.org/10.1080/00450618.2019.1695939
- Fraser, H. (2021b). The development of legal procedures for using a transcript to assist the jury in understanding indistinct covert recordings used as evidence in Australian criminal trials: A history in three key cases. Language and Law / Linguagem e Direito, 8(1), 59–75. https://doi.org/10.21747/21833745/lanlaw/8_1a4
- Fraser, H. (2022a). Forensic transcription: Legal and scientific perspectives. In C. Bernardasci, D. Dipino, D. Garassino, E. Pellegrino, S. Negrinelli, & S. Schmid (Eds.), Speaker individuality in phonetics and speech sciences: Speech technology and forensic applications (pp. 19–32). Milano.
- Fraser, H. (2022b). A framework for deciding how to create and evaluate transcripts for forensic and other purposes. Frontiers in Communication, 7(Capturing Talk), 898410. https://doi.org/10.3389/fcomm.2022.898410
- Fraser, H., & Kinoshita, Y. (2021). Injustice arising from the unnoticed power of priming: How lawyers and even judges can be misled by unreliable transcripts of indistinct forensic audio. Criminal Law Journal, 45(3), 142–152.
- Fraser, H., & Loakes, D. (2020). Acoustic injustice: The experience of listening to indistinct covert recordings presented as evidence in court. Law Text Culture, 24(1), 405–429.
- French, P., & Fraser, H. (2018). Why “ad hoc experts” should not provide transcripts of indistinct forensic audio, and a proposal for a better approach. Criminal Law Journal, 42(5), 298–302.
- Fuller, J. (2020). The David Eastman case: The use of inquiries to investigate miscarriages of justice in Australia. Alternative Law Journal, 45(1), 60–65. https://doi.org/10.1177/1037969X19886348
- Garnes, S., & Bond, Z. (1980). A slip of the ear: A snip of the ear? A snip of the year? In V. A. Fromkin (Ed.), Errors in linguistic performance (pp. 231–239). Academic Press.
- Georgiou, G. (2019). “From a nagging doubt”. Criminal lawyers association of the northern territory conference. https://clant.org.au/wp-content/uploads/Georgiou-Paper.pdf
- Gould, J., Carrano, J., Leo, R., & Young, J. (2012). Predicting erroneous convictions: A social science approach to miscarriages of justice. National Institute of Justice.
- Hamer, D. (2015). The Eastman case: Implications for an Australian criminal cases review commission. Flinders Law Journal, 17, 433–469.
- Haworth, K. (2018). Tapes, transcripts and trials. International Journal of Evidence and Proof, 22(4), 428–450. https://doi.org/10.1177/1365712718798656
- Heffer, C., Rock, F., & Conley, J. (2013). Legal–lay communication: Textual travels in the Law. Oxford University Press.
- Herald Sun. (2019). Shades of Lawyer X with solicitor taped dobbing in client over Colin Winchester murder. The Herald Sun. 4 January.
- Inman, M. (2018). Retrial of David Eastman cost about $6.5 million. Canberra Times. 22 November.
- Julian, R., Howes, L., & White, R. (2021). Critical forensic studies. Routledge. https://doi.org/10.4324/9780429505782
- Knaus, C. (2014). Retired detective Richard Ninness fears he could be target if David Eastman freed from prison. The Canberra Times. 21 August.
- Martin, B. (2014). Inquiry into the Conviction of David Harold Eastman for the Murder of Colin Stanley Winchester. Report of the Board of Inquiry, 29 May.
- Maxwell, C. (2019). Preventing miscarriages of justice: The reliability of forensic evidence and the role of the trial judge as gatekeeper. Australian Law Journal, 93(8), 642–654.
- McMahon, M., & Fraser, H. (2023). Transcription of indistinct forensic audio: Time for reform. Law Institute of Victoria Journal, (August), 20–23.
- Odgers, S. (2021). Uniform evidence law (16th ed.). Lawbook Co.
- Robertson, J. (2014). Push for royal commission into forensic procedures. Australian Journal of Forensic Sciences, 46(4), 365–367. https://doi.org/10.1080/00450618.2014.937457
- Smith-Khan, L. (2022). Incorporating sociolinguistic perspectives in Australian refugee credibility assessments: The case of CRL18. Journal of International Migration and Integration, https://doi.org/10.1007/s12134-022-00937-2
- Vincent, S. (2019). The retrial of David Eastman. The Monthly, January.
- Waterford, J. (1995). Public’s doubts to continue over trial outcome. The Canberra Times, 4 November.
- Wistrich, A., Guthrie, C., & Rachlinski, J. (2005). Can judges ignore inadmissible information? The difficulty of deliberately disregarding. University of Pennsylvania Law Review, 153(4), 1251–1345. https://doi.org/10.2307/4150614
Cases and legislation cited
- Butera v DPP (VIC) (1987) HCA 58
- R v Cassar (1999) NSWSC 436, Paragraph 7
- Eastman v The Queen (1997) FCA 548; (1997) 158 ALR 107
- Eastman v Director of Public Prosecutions [No 2] (2014) ACTSCFC 2
- Eastman v The Australian Capital Territory (2019) ACTSC 280
- R v Menzies (1982) NZCA 19
- Evidence Act (1995) (Cth.)