Threatening Language: Taxonomy and Analysis

In ordinary usage, 'threat' covers a wide range. It includes formal written demands, angry voicemails, social media posts, graffiti, and ambiguous remarks overheard in a corridor. The law in most jurisdictions requires, at minimum, that the recipient felt reasonably threatened, and many statutes also require the sender to have intended to place them in fear. Linguistics starts in a different place. It asks: what is this text doing at the level of language structure and function?

The speech act framework developed by Austin and Searle in the 1960s gives linguists the vocabulary. A threat is an illocutionary act, a communicative move that does something beyond just describing a state of affairs. The propositional content names a future harm. The illocutionary force is the speaker's expressed commitment to bringing about (or tolerating, or failing to prevent) that harm. The perlocutionary effect is what happens to the recipient: fear, compliance, changed behaviour.

The practical consequence is that forensic linguists testify about the structure and features of a text, not about the sender's dangerousness. What did the words mean? What registers did they draw on? Were there markers of specificity, knowledge, or rehearsal that distinguish instrumental planning from expressive venting? These are the questions that fall within the linguist's competence.

The single most practically important classification in threatening-language analysis is whether a threat is conditional. In a conditional threat, the harm is linked to a trigger: something the sender wants, something the recipient might do, or an event that might occur. In an unconditional threat, no such anchor exists. The harm is presented as inevitable or at the sender's sole discretion.

The legal implications follow directly. Extortion and blackmail typically require a demand: do this or I will cause harm (or withhold a benefit). That structure is conditional. A pure unconditional threat, 'I am going to hurt you', is more likely to be charged as a criminal threat or as part of a stalking or harassment pattern, where the focus is on placing someone in fear rather than extracting compliance.

The analyst should also be alert to what might be called false conditional framing: a text that presents the harm as conditional as a rhetorical device but where the condition is set so that it is already met or impossible to avoid. 'If you ever talk to anyone about this again, I will destroy you' from someone who has already been told about a report that was filed is functioning more like an unconditional threat in practice, because the triggering event is effectively certain.

A direct threat names the agent, the target, and the harm explicitly: 'I will kill you if you testify.' An indirect threat achieves the same communicative effect through implication, metaphor, reference, or reported hypotheticals. 'It would be a shame if something happened to your family' is a threat structured as an expression of concern. 'People who talk don't last long around here' is a generalised statement that functions, in context, as a specific warning.

Indirect threats are analytically harder to handle in court. The defence will argue the text is innocent on its face. The prosecution will argue that in context, no reasonable person could have read it any other way. The linguist's role is to make the pragmatic reasoning explicit: show the path from the literal meaning to the implied threat, identify the contextual features that close off innocent readings, and explain why ordinary language norms make the threatening interpretation the most natural one. Grice's cooperative maxims, relevance theory, and politeness theory all provide frameworks for doing this rigorously.

The FBI's National Center for the Analysis of Violent Crime has published threat-assessment guidance that draws a key distinction: the pathway to violence is a behavioural process, not a single communicative event. This means a threat communicated in writing is best understood as one data point in a sequence that may include surveillance behaviour, weapon acquisition, planning disclosures, and escalating communications. The text itself does not determine risk level; the whole pattern does.

1. Text characterisation
   The linguist describes the text's structural features: conditional or unconditional, direct or indirect, explicit or implied harm, specificity of target and method, any markers of prior knowledge or planning, register and emotional tone.
2. Contextual anchoring
   The text is placed in its communicative context: relationship between sender and recipient, prior contact history, whether the communication followed a triggering event such as a dispute, dismissal, or relationship breakdown.
3. Behavioural integration
   The text is considered alongside behavioural information gathered by threat assessors, law enforcement, or HR: surveillance, weapons access, social isolation, and recent life stressors. This step is outside the linguist's expertise and is performed by the assessment team.
4. Risk communication
   Findings are communicated in a form that allows decision-makers to act. The linguist contributes to the language dimension and is clear about what the text can and cannot support as a conclusion.

WAVR21 (White and Meloy, first published 2007; second edition 2010; third edition 2016) provides 21 factors organised across five domains: motivation, communication, intent and planning, means and access, and inhibitors. A forensic linguist primarily contributes to the communication domain but may also illuminate intent markers visible in the text. The key discipline is staying within the linguistic evidence and not claiming expertise about the sender's psychological state unless qualified to do so.

The volume of threatening language received by organisations, public figures, and law enforcement is enormous. Most of it is never acted upon. The practical challenge is not identifying threatening language but separating the small fraction from which a genuine threat emerges from the large volume of venting, frustration, and attention-seeking that uses threatening vocabulary without genuine intent. Linguists call the latter expressive anger.

• Specificity of planning: texts that include specific locations, times, methods, or knowledge about the target's routines are qualitatively different from general expressions of hostility.
• Prior access and knowledge: references to information the sender should not publicly know (home address, daily schedule, family members' names) suggest surveillance that goes beyond text.
• Escalation pattern: a series of communications that increase in specificity, intensity, or proximity over time differs from a single outburst.
• Absence of demand: some threatening texts that lack any demand or condition are harder to attribute to an instrumental goal and may reflect a different psychological register, which the behavioural team needs to assess.
• Register coherence: expressive anger often mixes registers (casual to intense to casual), shifts addressee mid-text, or uses hyperbole patterns recognisable in ordinary argument. Instrumental threats may be more consistently focused.

Every meaningful speech act is produced in a context, and threatening language is more context-dependent than almost any other speech act type. 'I'll kill you' shouted at a referee by a frustrated player after a bad call is processed differently from the same words sent in writing to an ex-partner at 3 a.m. The propositional content is identical. The communicative function is very different. A forensic linguist who submits a text-only analysis without addressing context is producing an incomplete report.

Context encompasses the relationship between sender and recipient (strangers vs. intimate partner, employee vs. employer, student vs. teacher), the medium and channel (anonymous letter vs. traceable email vs. public social media post), the history of prior contact (first communication vs. ongoing pattern), any triggering event (disciplinary action, restraining order, public criticism), and the platform's ambient communicative norms (a community known for hyperbolic language operates differently from a formal workplace email).

The methodological obligation is transparency. An analyst who does not have full contextual information should say so clearly and qualify every conclusion accordingly. A report that makes confident claims about intent while acknowledging an absence of background information is epistemically irresponsible, however linguistically sophisticated. Courts and investigators need to know what the analysis rests on, not just what it concludes.