Spoken warnings or threats pervade our language behaviour: a parent admonishes a child reluctant to eat his greens, a boss chides her employees to increase productivity or face the consequences, a gang member threatens to harm another in a moment of anger or as a premeditated act of intimidation. As speech acts, threats are a common strategy we use to try to influence the behaviour of others.
Predictably, this kind of language is regularly encountered in forensic linguistic casework, but in the administration of criminal justice it is often difficult to know when an utterance has crossed the line between lawful everyday communication and language crime. Making this distinction is especially hard in cases where the speaker’s chosen words are ambiguous or neutral. Here, the ‘tone of voice’ that the alleged threatener used may be the critical factor.
However, until now the phonetic properties of this vocal setting have remained unresearched.
What does it mean to talk of ‘a threatening tone of voice’? Police and criminal courts in the UK and internationally generally seem to take no issue with the idea that this type of vocal behaviour is something we can identify unproblematically. But a lot may rest on subjective judgement: it may increase the likelihood of a guilty verdict, and the imposition of a harsher sentence. What phonetic features are we basing these decisions on? Can we isolate a set of reliable speech cues that allow the listener to distinguish ‘threats’ from ‘non-threats’, even where the words used are identical?
Previous research on verbal threats in the written domain has shown that the phenomena we investigate in this project are difficult to harness into classical theories of speech acts. We can threaten one another verbally using a variety of utterance types which look ostensibly like things other than threats: statements of fact, questions, warnings, even compliments. We may be direct: "I’m going to kill you", or indirect: "Do you like hospital food?"; and we may set out conditions: "Do X, and Y will happen" or "If you don’t do X, Y will happen". Until now, we had no clear understanding of what the phonetic characteristics of a threatening tone of voice might be.
To date, experimental data from our threat speech research have been published in the Journal of Experimental Psychology: Human Perception and Performance (James Tompkinson and Dom Watt, with Mila Mileva and Mike Burton) and in Language and Law / Linguagem e Direito (James Tompkinson and Dom Watt). The study will achieve further impact via Watt’s involvement in delivering training in forensic speech and language issues for the UK College of Policing, and through presentation of research results to academic, law enforcement and intelligence communities.
The research has been carried out by Dr Dominic Watt in collaboration with research students Sarah Kelly and James Tompkinson. We approached spoken threats from a variety of angles. We looked first at how speakers produce spoken threats, using both authentic forensic case recordings, and simulated spoken threats elicited from experimental participants. A comparison of the two datasets revealed that there is no single consistent phonetic strategy that speakers use to convey the menace or aggression associated with threats. Some speakers raised the volume, pitch or tempo of their voices, while others lowered them.
We then considered how listeners respond subjectively to recordings of utterances that had independently been rated according to perceived threat level. According to UK law, for a spoken threat to constitute a language crime it must be sufficient to cause fear or alarm on the hearer’s part. A ‘successful’ threat must therefore have two participants: a speaker who intended to cause fear or alarm, and a hearer who infers the speaker’s intention in the relevant way. Even if a recording of the offending material exists, expecting listeners to agree that a threat was being made by the speaker may not always be realistic. There is a high level of subjectivity involved, probably inevitably: the listener’s experience and attitudes are bound to play a role in how he or she perceives another’s utterances.
In order to probe this aspect of listeners’ perception more deeply, we also manipulated the speakers’ social/regional accents, and in collaboration with colleagues in the University’s Department of Psychology used facial images to examine the role the purported speaker’s appearance might play in how the audio stimuli were rated.