In this paper, we explore speaker recognition as a process that is based partlyon analysis of phonetic and linguistic features and partly on pattern recognition(Gestalt perception). Central to this paper is the challenge to incorporate thedual nature of speaker recognition in forensic casework. To meet this challenge,we propose the so-called “blind grouping” method. Blind grouping does notrequire verbal-analytic terminology, but requires the expert to compare anonymisedfragments and arrange them into groups based on same-speaker anddifferent-speaker judgements. It allows the forensic expert to use any strategy toreach a result, including pattern recognition and feature analysis. This methodis proposed as a supplement to the traditional auditory-acoustic method.