Universal and language-specific sublexical cues in speech perception: A novel electroencephalography-lesion approach

The phonological structure of speech supports the highly automatic mapping of sound to meaning. While it is uncontroversial that phonotactic knowledge acts upon lexical access, it is unclear at what stage these combinatorial rules, governing phonological well-formedness in a given language, shape speech comprehension. Moreover few studies have investigated the neuronal network affording this important step in speech comprehension. Therefore we asked 70 participants—half of whom suffered from a chronic left hemispheric lesion—to listen to 252 different monosyllabic pseudowords. The material models universal preferences of phonotactic well-formedness by including naturally spoken pseudowords and digitally reversed exemplars. The latter partially violate phonological structure of all human speech and are rich in universally dispreferred phoneme sequences while preserving basic auditory parameters. Language-specific constraints were modelled in that half of the naturally spoken pseudowords complied with the phonotactics of the native language of the monolingual participants (German) while the other half did not. To ensure universal well-formedness and naturalness, the latter stimuli comply with Slovak phonotactics and all stimuli were produced by an early bilingual speaker. To maximally attenuate lexico-semantic influences, transparent pseudowords were avoided and participants had to detect immediate repetitions, a task orthogonal to the contrasts of interest. The results show that phonological ‘well-formedness’ modulates implicit processing of speech at different levels: universally dispreferred phonological structure elicits early, medium and late latency differences in the evoked potential. On the contrary, the language-specific phonotactic contrast selectively modulates a medium latency component of the event-related potentials around 400 ms. Using a novel event-related potential–lesion approach allowed us to furthermore supply first evidence that implicit processing of these different phonotactic levels relies on partially separable brain areas in the left hemisphere: contrasting forward to reversed speech the approach delineated an area comprising supramarginal and angular gyri. Conversely, the contrast between legal versus illegal phonotactics consistently projected to anterior and middle portions of the middle temporal and superior temporal gyri. Our data support the notion that phonological structure acts on different stages of phonologically and lexically driven steps of speech comprehension. In the context of previous work we propose context-dependent sensitivity to different levels of phonotactic well-formedness