<p>Language identification has become increasingly critical in real-world applications, particularly in forensic linguistics and natural language processing. This study employed an eye-tracking visual world paradigm to investigate the dynamic modulation of attention during lexical and language identification. Analyses of key-pressing and eye-tracking data across six conditions — speaker-dependent (1), object-dependent (2), speaker- or object-dependent (3), mismatch speaker-object (4), speaker-only (5), and object-only (6) — revealed that lexical identification takes priority over language-speaker identification in auditory speech processing. Nevertheless, a mismatch between the lexical referent and the speaker’s language triggered an early (about 600 ms in English and 320 ms in Dutch after word onset) increase in pupil size and early peaking and dropping of inspections to speakers, revealing early attentional integration of the two processes. The results also indicate that language identification is achievable even given very limited prior exposure to an unfamiliar language (when speaker-language association is useful, M = 0.67, SD = 0.38). Specifically, this is achieved through cross-linguistic mutual exclusivity and non-lexical phonetic cues, with individual variability in the use of these strategies.</p>