Sieves and Herrings: The Phonology of Swedish Vowel and Consonant Length

In Standard Central Swedish (henceforth 'Swedish'), every stressed syllable has either a long vowel (e.g. [siːl] 'sieve') or a long consonant (e.g. [sɪlː] 'herring'); there are no syllables with both, e.g. *[siːlː] and no syllables with neither, e.g. *[sɪl]. So long vowels and long consonants are in complementary distribution. Now, which is phonemic, vowel or consonant length? Most scholars believe that consonant length is phonemic, rather than vowel length: /sil/ 'sieve' becomes [siːl] through a process of vowel lengthening, and /silː/ 'herring' becomes [sɪlː] by just changing the quality of the vowel. In this presentation, I report on Andersson (in prep.), where I argue the exact opposite. Words like 'sieve' are /siːl/ underlyingly (as in the surface form), while words like 'herring' are underlyingly /sɪl/ with consonant lengthening giving [sɪlː]. I argue that the data used in the literature actually favour phonemic vowel length, contra Eliasson (1985) and Riad (2013). I also introduce data never previously discussed showing that minimal pairs for vowels exist, thereby supporting phonemic vowel length. Like Riad (2013), I am not aware of any minimal pairs for consonant length. I also argue that phonemic vowel length accounts better for vowel quality facts presented in Löfstedt (1992) and historical Swedish vowel changes. My own investigations in Andersson (in prep.) also find support for phonemic vowel length in synchronic phonological processes, the vowel harmony innovated by Swedish-speaking children and the intuitions of native speakers of Swedish.