The inconvenient truth about sound change: Lessons from modern variation “Assumptions” of the comparative method 1) Sound change is exceptionless 2) Language split is sudden 3) After the split, there is no contact among daughter languages 4) The proto-language is uniform, without variation (Campbell, 1999) 1 Variation in the proto-language The seeds of sound change don’t fall far from the tree - on-going variation? 2 thorn <þ> and edh <ð> Old English: <þ / ð> (interchangeable allographs) [+voice] / [+voice]__[+voice] [-voice] / elsewhere = complementary distribution based on voicing 3 thorn <þ> and edh <ð> Old Icelandic: <ð> [+voice] / [+voice]__[+voice] , / _# <þ> [-voice] / elsewhere = complementary distribution based on voicing 4 Historical Development Grimm’s PIE *T > þ *D > t *Dh > ð d / # __ (yes, this is all very simplified) 5 Historical Development Grimm’s PIE *T > Verner’s þ (after unstressed syllable) ð *D > t *Dh > ð d / # __ (yes, this is all very simplified) 6 Historical Development Proto-Gmc OHG þ (initially) <th>[þ]? thenkan (medially& finally) <d> [d] bruoder ð <t> [t] muoter d <t> [t] tohter t <z> [ts] z ît 7 Historical Development OHG NHG <th> [þ]? <d> [d] denken <d> [d] <d> [d] Bruder <t> [t] <t> [t] Tochter <z> [ts] <z> [ts] Zeit 8 Historical Development Gmc OE < ð > [ð ] <d> [d]? <þ> [þ] <þ> [þ/ð] broþar <t> [t] <t> [t] tide <d> [d] <d> [d] daughter moder 9 Theoretical considerations WG strengthening incompatible with Proto-Gmc allophonic voicing OE OHG ð d moder t muoter þ þ broþar d bruoder Proto-Gmc (from Grimm’s and Verner’s) 10 Theoretical considerations •Lack of orthographic distinction for voiced/voiceless <þ> in runes •Parallel lack of distinction in Modern English •Strengthening before voicing assimilation 10 Theoretical considerations Old Icelandic: <ð> [+voice] / [+voice]__[+voice] , / _# <þ> [-voice] / elsewhere = complementary distribution based on voicing 11 Theoretical considerations Gothic: <d>* > [ð] / [+voice]__[+voice] , [d] / elsewhere <þ>** = [þ] [-voice] * complementary distribution based on manner of articulation (sonority) ** no allophones posited 12 Modern Variation American English as spoken in Columbus, OH Frequent assimilatory voicing in dental fricatives Based on surrounding voiced sounds Interplay of stress and prosodic position 13 The study: Strategies for maintaining a voiced/voiceless distinction: •Contrast in aspiration (in stops) or duration (in fricatives) •Contrast in manner of articulation •Contrast in place of articulation 14 Variation 5000 0 0 0.5777 Time (s) “the” Speaker 2, voiceless, emphatic, preceded by pause 15 Variation 5000 0 0 0.121 Time (s) “þar” Icelandic speaker, voiceless, emphatic, preceded by pause 16 Variation th voice 0 0.121 Time (s) “þar” Icelandic speaker, voiceless, emphatic, preceded by pause 17 Variation 5000 0 0 0.159 Time (s) Icelandic: “horfa þau yfir heiðavötnin bláu” Completely voiced <þ> 18 Variation 5000 0 0 0.1907 Time (s) “those” Speaker 2, voiced and nasalized, preceded by velar nasal 19 Variation 5000 0 0 0.1214 Time (s) að: Icelandic, near approximant, completely attached, unstressed, intervocalic 20 Re-syllabification 8000 6000 no glottal pulse 4000 2000 0 math 121.9 122 122.1 time (seconds) and 122.2 <th> attaches to following syllable 21 aspiration / devoicing 8000 6000 4000 2000 0 I 240.3 think 240.4 time (seconds) voice loses intensity, vowel begins devoicing 22 Other variation in this corpus, relevant to Gmc Gemination and lengthening West Gmc Strengthening Change in place Edh as an approximant Intervocalic voicing or weakening Word initial and word final devoicing Assimilation of eth following /n/ Devoicing of sonorants following thorn 23 Conclusions? Continuous variation is possible. Historical linguistics needs to step into the 21st century: •Phonetic variation does not cease to exist. •Language is constantly in a state of flux. •Analogy is a good start. 24 Conclusions: Continuous Variation •Variation inherited from the Proto-language •Grimm’s and Verner’s Laws also messy •Phonetic variation in outcome •Regularized by generalization and analogy 25 Thank you The study: Measurements 8000 6000 frication begin end 4000 2000 0 every 35.3 35.4 time (seconds) begin end voice bar The study: Measurements 8000 6000 frication 4000 begin end 2000 0 think 0 0.05 time (seconds) end begin voice bar 0.1 Presence of “voice bar” regular waveform voice bar irregular waveform no voice bar 1. <th> duration (x) and voice bar (y) sorted by phonetic environment -presence or absence of adjacent voiceless segments Sp 1 100% +voice QuickTime™ and a TIFF (Uncompressed) decompressor are needed to see this picture. 50% +voice 1. <th> duration (x) and voice bar (y) sorted by phoneme Sp 1 QuickTime™ and a TIFF (Uncompressed) decompressor are needed to see this picture.