This article considers the variation in the metres of the ‘ὥσπερ ξένοι’ epigrams, collected in the Database of Byzantine Book Epigrams (DBBE). In its canonical form, these epigrams follow a dodecasyllabic metrical pattern. The seemingly unmetrical decasyllabic and decatetrasyllabic variants are explained from a cognitive-linguistic perspective as the pairing of different cola – 5+5 and 7+7 instead of the usual 7+5 or 5+7. From this perspective, cola can be equated with the cognitive ‘idea’ or ‘intonation units’ (IUs) used in ordinary speech.