The first-person possessive (1POS) is known to be realised in different ways: [maɪ], [mi], [mə], [mɑ] in the North-East of England (Moelders to appear). While previous studies have investigated the influences of social factors on the proportional use of the different realisations, less attention has been paid to variation in 1POS’s phonetic realisations. Specifically, there is a question of whether the variant [mi] is a phonetic realisation of the 1POS (Childs 2013) or “an extension of the object form” of possessive me (Anderwald 2004:177). A growing body of work shows that patterns across levels of linguistic representation are probabilistic, with systematic variation encoded at the level of the lemma (e.g., Drager 2011). Here, we investigate two questions: (1) what is the phonetic profile of the variants of 1POS?; and (2) to what extent do lemma-specific realizations of 1POS [mi] differ from the first-person accusative object (1OBJ) [mi]. To address these questions, auditory coding of ~1500  tokens is combined with acoustic analysis of F1 and F2 over the vowel trajectory. Measurements were taken along the vowel trajectory at 10-percent increments. Preliminary results suggest that the phonetic realization of 1POS variants do show the range of attested variants (i.e., [mai, mi, mə, mɑ]). In addition, differences arise between [mi] when fulfilling the role of 1POS and 1OBJ. The onset of 1OBJ [mi] is longer than 1POS [mi]. These findings lend weight to models that suggest words/lemmas are stored with rich acoustic information (e.g., Pierrehumbert 2002).