333
BEST PRACTICES AND ADMISSIBILITY OF
FORENSIC AUTHOR IDENTIFICATION
Carole E. Chaski*I. INTRODUCTION
Forensic linguistics provides answers to four categories of
inquiry in investigative and legal settings: (i) identification of
author, language, or speaker; (ii) intertextuality, or the
relationship between texts; (iii) text-typing or classification of
text types such as threats, suicide notes, or predatory chat; and
(iv) linguistic profiling to assess the author’s dialect, native
language, age, gender, and educational level. This article
discusses author identification in relation to linguistics, research,
and admissibility as evidence in U.S. courts.
Federal and states courts in the United States have
undertaken three main approaches in determining whether to
admit, partially admit, or exclude forensic authorship
identification evidence. These three approaches are forensic
computational linguistics, forensic stylistics, and stylometric
computing. Each has a distinct origin. Forensic computational
linguistics developed out of linguistic theory and computational
linguistics.^1 Forensic stylistics developed out of traditional
forensic handwriting identification.^2 The stylometric computing
approach developed out of both literary authorship identification
and machine-learning-based text classification.^3
- Institute for Linguistic Evidence; ALIAS Technology LLC, Georgetown,
DE; Ph.D., Brown University.
(^1) See Carole E. Chaski, Who Wrote It? Steps Toward a Science of
Authorship Identification, NAT’L INST. JUST. J., Sept. 1997, at 15, 18
[hereinafter Chaski, Who Wrote It?].
(^2) See GERALD R. MCMENAMIN, FORENSIC STYLISTICS 45–46 (1993).
(^3) See Moshe Koppel & Jonathan Schler, Exploiting Stylistic