The research conducted by Kanampiu, Martin, and Culbertson (2025) investigates the role of semantic features in the processing and representation of noun class systems in the Bantu language isiXhosa. This study addresses a notable gap in the literature regarding the empirical evidence for the semantic basis of noun class divisions, particularly in languages with complex grammatical gender systems. While grammatical gender is a well-studied phenomenon, the specific mechanisms by which speakers of Bantu languages utilize semantic and morphophonological cues in noun class assignment remain underexplored. This research is significant as it contributes to our understanding of how linguistic features are cognitively processed in a language family that exhibits a rich diversity of noun classes.

The methodology employed in this study is innovative, utilizing an auditory lexical decision task to assess how native isiXhosa speakers respond to various noun class prefixes paired with noun stems. The experiment involved 90 participants and included both real isiXhosa words and pseudowords that were designed to create semantic violations. Specifically, the study contrasted responses to putative semantic violation items—where a noun class prefix was mismatched with a noun stem from a different class—with syntactic violation items, which involved a clear grammatical category mismatch. This dual-violation approach allows for a nuanced examination of whether speakers treat noun classes as semantically coherent, particularly in the context of isiXhosa, where certain classes have clearer semantic associations than others.

The findings of the study reveal that participants exhibited significantly different rejection rates for semantic and syntactic violation items, supporting the hypothesis that isiXhosa speakers apply semantic rules when processing noun classes. Specifically, participants robustly rejected syntactic violations, indicating a clear understanding of grammatical structures. In contrast, responses to semantic violation items varied based on the noun class prefix used. For the NC4 prefix, which is associated with inanimate nouns, there was a high rejection rate for items that combined it with animate stems, suggesting that speakers utilize semantic coherence in their processing. However, for the NC10 prefix, which encompasses a broader and more eclectic range of nouns, rejection rates were lower, indicating that speakers may not have a clear semantic rule to apply in this case.

The broader significance of this research extends to fields such as language technology and natural language processing (NLP). Understanding how speakers of isiXhosa and similar languages process noun class information can inform the development of more sophisticated language models and machine translation systems that account for grammatical gender and noun class distinctions. Additionally, insights from this study can enhance computational linguistics applications by providing empirical data on how semantic and morphophonological cues interact in language processing. This knowledge is crucial for creating more accurate and contextually aware language technologies that can better serve speakers of Bantu languages and others with complex grammatical systems.

Source: glossa-journal.org