Linguistic Profiling and Dialectology as Evidence

Dialectology rests on decades of survey work. The Survey of English Dialects (1950s-1960s), the Atlas of North American English (Labov, Ash, and Boberg, 2006), the Survey of Regional English in Germany, the Atlas Linguistique de la France, and their equivalents elsewhere have documented which phonological features occur where, and at what rate. This reference base is what allows an analyst to move from 'this speaker has a non-prevocalic /r/' to 'this feature is associated with non-rhotic British dialects and is absent from most American varieties'.

The features most useful for origin inference are those that vary systematically across geography and that are resistant to accommodation (the unconscious tendency to shift toward a listener's speech patterns). Among the most informative in English:

• Vowel quality in the TRAP and BATH sets: the distinction between Northern and Southern British English; the Northern Cities Vowel Shift in the US Great Lakes region.
• Rhoticity: whether /r/ is pronounced after a vowel ('car', 'bird'). Non-rhotic dialects characterise much of England, New Zealand, and some parts of the Eastern US; rhotic dialects dominate most of North America and Scotland.
• T-glottaling and glottal stops: common in urban British varieties, especially Cockney and its geographic spread, and in many other British regional accents.
• Consonant cluster reduction and AAVE features: systematic patterns associated with African American Vernacular English, including copula deletion and habitual 'be', which indicate community membership rather than geographic origin.

John Baugh's experiments in the 1990s and 2000s have become foundational in the forensic and sociolinguistic literature on discrimination. Baugh called landlords advertising apartments in the San Francisco Bay Area, selecting cities with varying racial compositions, using three carefully controlled voices, each representing a different accent: Standard American English (or what Baugh called 'Sounding White'), African American Vernacular English ('Sounding Black'), and Chicano English ('Sounding Latino'). The same person, the same words, the same apartments, but systematically different callback rates.

The results were consistent across cities. In predominantly White or Asian neighbourhoods in particular, callers using minority-associated accents were significantly less likely to be offered viewings. The experiments satisfied the methodological requirements for a controlled study: all variables except accent were held constant, and Baugh documented the calls with recordings. The work has been cited in US fair housing litigation and in academic analyses of voice-based discrimination across many jurisdictions.

Baugh's work also gave the field a precise vocabulary. He distinguished between legitimate dialect analysis (using documented phonological features to infer origin) and discriminatory linguistic profiling (using perceived accent as a proxy for race or ethnicity to make adverse decisions). The former is a forensic tool; the latter is a civil rights violation in jurisdictions where fair housing laws prohibit discrimination on racial grounds. Forensic linguists must be clear about which side of that line their evidence sits on.

Studies of diaspora communities in North America and Europe have documented how second-generation speakers, born in the host country, retain phonological and prosodic features associated with their heritage language community. Somali-heritage speakers in Minneapolis, Toronto, and London, for example, show consistent patterns in English prosody, consonant inventory, and vowel quality that differ systematically from both standard host-country norms and from first-generation Somali immigrant speech.

From an investigative standpoint, these patterns mean that a recording from an unidentified speaker can sometimes be placed within a diaspora community with reasonable confidence. From a discrimination standpoint, the same patterns have been documented as triggering prejudiced responses in employment, education, and criminal justice contexts.

The single most important thing to communicate about dialect-based origin inference is its probabilistic character. A speaker whose phonological profile is consistent with a Northern English origin may have been born in Newcastle, may have grown up in a Northern English-origin family in New Zealand, may have spent formative years in the North after being born elsewhere, or may have adopted Northern features through contact with Northern-dialect speakers. The features constrain the possibilities; they do not select one.

Dialect levelling complicates origin inference further. A speaker who grew up in a strong dialect community but has spent twenty years in a different region, or who has undergone speech therapy, or who consciously modifies their accent for professional contexts, will show a mixture of features that may confound place-of-origin analysis. Analysts must account for the possibility of dialect shift and should not treat a current phonological profile as a stable biographical record.

Forensic linguists may be called on to assess dialect evidence not just in English but in other languages, and in contexts involving interpretation. Two complications arise. First, the reference data for origin inference in many languages is far less developed than for English, limiting the precision of any regional inference. Second, where an interview was conducted through an interpreter, the interpreter's own dialect or register may have influenced the record in ways that affect later analysis.

A witness from Somalia may speak a regional variety that an interpreter from a different Somali region renders differently in Somali, before translating into English. Features of the original utterance, including markers of education level, emotional register, or community affiliation, may not survive the double transformation. A forensic linguist asked to analyse the English transcript must account for the distance between the original speech and the record they are examining.

• Always work from audio recordings when available, not just transcripts, because transcription collapses phonological information.
• Establish the chain of linguistic custody: who conducted the interview, in which language, with or without an interpreter, and what the interpreter's own dialect background is.
• Calibrate conclusions to the quality of the recording and the degree of language mediation: a distant, poor-quality, interpreted record supports weaker inferences than a direct, high-quality, uninterpreted one.

The International Association of Forensic Linguists and national professional bodies emphasise that dialect analysis should be grounded in systematic phonological description, referenced against documented survey data, and reported with explicit confidence intervals rather than as categorical claims. Several additional practices are recognised as best procedure in casework.

1. Blind analysis
   The analyst should examine the recording without being told the identity or demographic background of the suspect or witness, to avoid confirmation bias shaping which features they attend to and how they interpret ambiguous cases.
2. Feature documentation
   Every claimed dialectal feature must be documented with timestamps, examples, and frequency counts. An analyst who says 'the speaker has a Yorkshire accent' without a systematic feature list has provided an impression, not an analysis.
3. Reference population disclosure
   The geographic or social reference population used for comparison must be stated explicitly, along with the survey or database from which it derives, so the court can assess the match between that population and the claimed origin.
4. Probabilistic reporting
   Conclusions should be stated in probability terms that acknowledge alternative origins consistent with the observed features, not as identifications. The analyst may use the ENFSI verbal equivalence scale or an equivalent framework to communicate strength of evidence.