Discovering nominal constructions through behavioral profiles: the case of Akan

This study examines nominal constructions in Asante Twi, a variety of Akan (Niger-Congo, Kwa), using the Behavioral Profiles (BP) approach. It investigates how constructional identification can be operationalized empirically, and what taxonomy of noun phrase (NP) patterns emerges from corpus-based distributional evidence. These questions are addressed by analyzing 3002 NPs drawn from a 200,000–word spoken corpus, and by using exploratory statistical techniques, such as cluster and multiple correspondence analyses. Three groupings of recurrent NP patterns were identified through the cluster analysis: (1) [Poss N] and [Poss.N], (2) [N MODs] and [MODs N MODs], and (3) [N] and [N MOD]. Distributional evidence indicated that these three groupings represent three macro-constructions: possessive , complex-modification, and simple-elaborating constructions , respectively. Within these broad schemas, six sub-constructions were identified, corresponding to core functions such as determination, quantification, possession, and categorization. We argue for a mid-level constructional architecture for the Akan NP family, as most constructional schemas essentially overlapped in their distributions of usage features. Methodologically, this study illustrates how multivariate statistical techniques can enhance the objectivity and descriptive adequacy of construction identification. Theoretically, it contributes to usage-based construction grammar by providing quantitative support for a network-based model of the nominal constructional domain.