The origin of variation in Norwegian retroflexion
This dissertation deals with a phonological process in Norwegian called ‘retroflexion’. When a word ending in the tap / -r / is followed by a word beginning with an alveolar coronal / t d n s /, the / -r / deletes, and the alveolar surfaces as a retroflex coronal [ t˛ d˛ n˛ s˛ ]. The study presented here does two things:
One, this dissertation documents the variation that exists in Norwegian retroflexion with the use of production experiments and judgment experiments. It finds that retroflexion is obligatory for / t d n /, but optional for / s / (chapter 1). It finds that retroflexion is applied to / s / more often when the following segment is a consonant than when it is a vowel, and more often when that consonant is a / k / than when it is a / t / (chapter 2). Finally, it looks only at words in / sV- / and finds that for words that already exist in the language retroflexion is more common when the words has many frequent phonological neighbors, whereas for novel words retroflexion is less common when the word has many frequent neighbors (chapter 4).
Two, this dissertation accounts for this variation. For the different alveolar onsets we have production data for, / t- d- n- sk- st- sV- /, it finds that the larger the perceptual distance is between the alveolar and the retroflex, the less likely retroflexion is to occur. It is proposed that this correlation between perceptual properties and phonology has arisen through mechanisms of word categorization (chapter 3). For the dichotomy between existing and novel words in / sV- /, the crucial difference between them is that only existing words have a history. It is suggested that the transmission history of existing words plays a major role in how such words are produced by speakers. Since novel words have no transmission history, they are unaffected by such properties (chapter 4).