UC Davis

The acoustic cues used to distinguish different vowel categories from each other differ from language to language. While native speakers of a language are well attuned to the important cues needed to identify different sounds with a high degree of accuracy, language specific cues and cue weightings can cause potential problems for second language learners because of the learned L1 contrasts. German is a language that shows quantity in its vowel system with every long vowel having a corresponding short vowel. It is unclear, however, whether spectral information or duration is used as the primary cue by German listeners to disambiguate whether a vowel is short or long. Additionally, the quantity-quality debate suggests that only one of the two features is distinctive while the other one is redundant, ignoring the potential use of secondary cues. This dissertation seeks to provide a clear answer to whether quality or quantity is used as a primary cue in vowel perception in German as well as the potential use of secondary in the disambiguation of long short vowel pairs.Results show that American English listeners perceived all German vowel pairs as different native categories, with the exception of /ɛ:/-/ɛ/ and /u:/-/ʊ/, largely relying on spectral features when identifying non-native vowel sounds. Therefore German vowels show substantial spectral differences in production that are salient enough for non-native listeners to exploit. In native speech perception, results show that while duration is used as a primary cue, spectral information was used as a secondary cue to disambiguate whether a vowel was long or short. While listeners identified tokens as long less often as tokens approached short durations, they still identified those tokens containing originally long spectral information as long more often than those containing short spectral information. The same patterns were found in the second experiment, with duration being used as the primary cue in disambiguating whether a vowel was long or short. Tokens containing originally short durations were selected as long less than 50% of the time, regardless of spectral manipulation step. Additionally, German listeners used spectral movement at least partially. Results from experiment four show that while identification accuracy stayed high overall, results indicate that German listeners rely on dynamic information, with silent-onset-offset tokens having lower identification accuracies than the silent-middle tokens. This is further evidence for spectral information being used as a secondary cue in the disambiguation of German long and short vowel pairs.