Relative importance of stress correlates in native listeners’ identification of Spanish lexical stress produced by monolingual and bilingual speakers

• Acoustic cues to word-level stress in speech perception do not always align with those in speech production. • Listeners exploit various stress correlates in the speech signal and assign different weights to them, which is context-specific. • When there is a pitch accent, native listeners mainly attend to f0-related cues but, in the absence of a pitch accent, intensity plays a stronger role. • Clustering based on stress correlates is not consistent with the predetermined monolingual-heritage-L2 group division. • While clustering in some contexts (i.e., nuclear positions and unaccented contexts) could be explained by speakers’ language learning experience, this factor did not seem to play a definitive role in the clustering in other contexts (i.e., prenuclear contexts). Spanish has many minimal stress pairs, and lexical stress in Spanish is marked primarily via suprasegmental cues. Thus, sensitivity to suprasegmental information is crucial for spoken-word identification in Spanish. Using stimuli produced by speakers of Mexican Spanish with varying language learning experience (i.e., monolingual speakers, heritage speakers, L2 learners), this study examines native listeners’ identification of Spanish lexical stress under enhanced variability in phonetic cues. Our data demonstrate that listeners exploit various stress correlates in the speech signal and assign different weights to them, which is context-specific; when there is a pitch accent, native listeners mainly attend to f0-related cues, whereas in the absence of a pitch accent, intensity plays a stronger role. Our data also show that clustering based on stress correlates is not consistent with the predetermined monolingual-heritage-L2 group division, which indicates that language learning experience alone is not sufficient to explain how Spanish speakers mark stress. This study highlights the importance of incorporating variable speech data in speech perception research and adopting a data-driven, individual-centered approach to speaker grouping in cross-sectional studies.