24.962 Advanced phonology 27 April, 2005 Introduction to exceptions (1) A pervasive phenomenon, that has cropped up in virtually every area we have discussed this semester: exceptions • Sometimes omitted from discussion—e.g., our discussion of English nicknames – Usual pattern: maximal syllable + [i] (Bar.ba.ra → Barbie, Car.la → Carly, Shel.don Sheldie, A.man.da Mandie; but Ag.nes → Aggie, *Agnie; Cath.ryn → Cathy, *Cathry) → → – But nicknames sometimes preserve less: Mil.dred Millie, Wan.da Wannie → → “Wanda, affectionately called ‘Wannie’ by family and friends, . . . ”1 (Wandie and Mildie also attested) – And sometimes involve other exceptional changes Sa.rah → Sally, Sadie Shet.land sheepdogs, ponies → Shelties (Shetlies also occasionally found, according to Google; *Shetties is apparently strictly found on German/Dutch pages, though Scot-land terriers Scotties is regular in English) → • Sometimes mentioned but set aside—e.g., Russian single vs. double retraction ‘ring’ sg. pl. ‘trade’ sg. pl. o k´ o rem´nom. kol’ts´ol’tsa nom. remesl´esla gen. kol’tsa´kol’´ eselets gen. remesla´rem´u k´u rem´dat. kol’ts´ol’tsam dat. remesl´eslam acc. kol’ts´ol’tsa acc. remesl´eslao k´ o rem´om k´ om rem´instr. kol’ts´ol’tsami instr. remesl´eslami loc. kol’tsje´k´ eslaxol’tsax loc. remeslje´rem´– Paradigm-based analysis can explain double retraction as columnar stress (rem´esel in-stead of *remes´el), but doesn’t explain why double retraction fails to apply in words like kol’´ets ‘rings-GEN.PL.’ Often introduced to cast a shadow of doubt • – Polish o-raising; numerous exceptions correlates with low productivity, casts doubt on the synchronic status of the alternation • Just a few of the cases we’ve discussed, giving incomplete discussion of exceptions: – English nicknames; Lakhota reduplication; Tagalog pseudoreduplication; all cyclic pro-cesses in English; Latin rhotacism; Russian yers and stress; Spanish stress; Polish o-raising; Rendaku; the rui part of Sanskrit ruki rule; etc. etc. etc. (2) Why it is sometimes important for phonology to proceed this way • Innocent simplification: the overall pattern is interesting and important, and existence of exceptions does not change the fact that we need to capture the overall pattern somehow – Existence of Shelty doesn’t change the fact that C2 in rising sonority sequence is usually not preserved (with or without metathesis); e.g., Kaplan → Kappie, *Kaplie, *Kalpie • In some cases, exceptions are the result of loanwords disrupting a previously “clean” system; by excluding them, we are simply examining the previous stage of the language (an utterly valid object of analysis) – E.g., Lyman’s Law and rendaku: hold only in “older” components of Japanese lexicon 1We would need to check and make sure there is no Dutch influence here; I’ve noticed that other similar cases, like Sannie for Sandra, occur mostly with Dutch web pages (or writers)—e.g.,“My name is Sandra, but people call me Sannie, feel free to do the same.”24.962—27 April, 2005 p. 2 (3) Why it is sometimes dangerous for phonology to proceed in this way • Must provide evidence that cleaner stage of language truly existed in hypothesized form, or that speakers discover generalizations as claimed, in spite of exceptions • Without such evidence, it’s not clear that we are analyzing either the language or the speaker (4) Why it is interesting to look into the exceptions: systematic patterns and subregularities Sometimes the pattern is seemingly quite arbitrary • Russian single vs. double retraction: – Sg. ∼ pl. stress alternations are not all that common in the language, but among nouns with stress retraction in entire plural, most show double retraction (more nouns like rem´ets)esel that kol’´– Nouns that have single retraction: ➢ Neuter kol’co, jajco, fem. ovca A possibly related fact: there are many nouns ending in -c with suffixal stress through-out the paradigm—except the gen. pl. ‘dust’ sg. pl. nom. pul’c’a pul’c´ 1j gen. pul’c1j pul’´ec dat. pul’ce´pul’c´am acc. pul’c´ u pul’c´ 1j oj(u) pul’c´instr. pul’c´ami loc. pul’ce´pul’c´ax ➢ Feminine: , svin’ja, sjem’ja, skam’ja, sjestra, sjer’ga Not clear to me why these should behave differently, simply note in passing that · they seem sort of like a cohesive set ➢ (Also some variable words, like gumno, gum´en ∼ g ´umen) Sometimes the pattern has solid phonological motivation (example in a minute) (5) Why it is necessary to look into the exceptions: data about exceptions can lead to better analysis • Yiddish voicing assimilation (Katz 1987, Lombardi 1999) Regressive assimilation in obstruent+obstruent clusters: (schematically) vOg ‘weight’+SOl ‘scale’ vOkSOl→bak ‘cheek’ +beIn ‘bone’ bagbeIn ‘cheekbone’ →More accurately: assimilation is quite variable – Across compound/word boundaries, it is optional (and perhaps phonetically gradient/partial) – Within (monomorphemic) words, it has numerous exceptions C1 C2 Pattern Example [+voi] [−voi] Assim. No assim. /plugte/ /kodshe/ [plukte] [kodshe] ‘dispute’ ‘Holy of . . . ’ [−voi] [+voi] Assim. No assim. /hekdesh/ /makdim/ [hegdesh] [makdim] ‘poorhouse’ ‘ahead’ – 0%20%40%60%80%100%[–voi][+voi][+voi][–voi] AssimNot AssimAn interesting statistical asymmetry: ➢ Reanalysis: Yiddish is like Mekkan Arabic (discussed last time), but with some exceptions Regressive devoicing (/DT/ → [TT]) but not voicing (/TD/ � *[DD])24.962—27 April, 2005 p. 3 Properties of exceptions (6) A useful typology of exceptions: (Kenstowicz and Kisseberth 1977, chap. 2.3) Negative input exceptions: Positive input exceptions: Segment is in proper environment to un-dergo rule, but exceptionally fails to do so Segment should not undergo rule, somehow does anyway but Negative environment exceptions: Positive environment exceptions: Segment creates the environment to condi- Segment should not provide correct envi-tion a rule, but it does not actually do so ronment, but triggers rule anyway (7) Negative input exceptions: • Word should contain input A to undergo rule A → B / C D, but exceptionally fails to • Example: exceptions to final