24.962 Advanced phonology 4 Apr 2005 More on paradigm regularization (1) What we saw last week • Phonological processes often apply non-normally within inflectional paradigms (overappli-cation, also underapplication?) – When a phonological process fails to apply as expected, is it because something was learned wrong, or because of extra paradigmatic pressure? – Some clear cases of paradigmatic pressure (Spanish stress, underapplication of flapping in militaristic, etc. • OO-F analysis: normal application somewhere in the paradigm, overapplication elsewhere by IDENT-OO – Parallel to reduplication: normal application in one copy, overapplication by IDENT-BR � IDENT-IO – IDENT among all paradigm members, or to privileged base? (Suggestive evidence that privileged base is needed, at least in some cases; is egalitarian OP-IDENT also crucial in some cases?) Goals today: • See how these factors play out in the analysis of a particular phenomenon: double retraction of stress in Russian gen. pl. • More broadly: step back and consider why such effects occur (2) Reminder: yer alternations in Russian noun paradigms (Warning: data from Zalizniak dictionary—archaisms may abound) • Yer vowel shows up when no suffix (yer in suffix) ‘piston’ sg. pl. orsh’en’ p´ nom. p´orshn’i orshn’a p´ej gen. p´orshn’´ orshn’u p´ dat. p´orshn’am orsh’en’ p´ acc. p´orshn’i orshn’em p´ instr. p´orshn’ami orshn’e p´ loc. p´orshn’ax • SYLSTRUC � *yer: forces yer to vocalize in nom/acc sg (*[porshn’]) • In affixed forms, [shn’] is possible onset, so yer does not need to vocalize (*yer � *COMPLEX) (3) Russian has numerous stress patterns (unpredictable): • Stress on root (ote´l’ ∼ ote´l’a ∼ ote´l’i ‘hotel-nom.sg./acc.sg/nom.pl’) – Also p´orsh’en’ ‘piston’ in (1); likewise plav’en’ ‘flux’, d’egot’ ‘tar’ ozhd’ ∼ dozhd’ ´• Stress on affix (where possible—retracts when suffix is yer; d´a ∼ dozhd’´ı ‘rain’) – Examples with yer alternations: ‘day’ sg. pl. ‘fire’ sg. pl. en’ dn’´ı nom. og´ nom. d’´on’ ogn’´ı gen. dn’a´dn’´ej ej gen. ogn’a´ogn’´u dn’´u ogn’´dat. dn’´am dat. ogn’´am en’ dn’´ı acc. og´acc. d’´on’ ogn’´ı em dn’´em ogn’´ instr. dn’´ami instr. ogn’´ami loc. dn’e´dn’´axax loc. ogn’e´ogn’´∗ Likewise p’en’ ‘stump’, lom´ot’ ‘slice’, pleten’ ‘hedge’24.962—4 Apr 2005 p. 2 Stress on suffix in singular, root in plural (kol’es´o∼ kol’es´a∼ kol’´esa ‘wheel’) • ‘chisel sg. pl. nom. gen. dat. acc. instr. loc. dolot´o dolota´dolot´ 1 dolot´o dolot´om dolote´dol´ota dol´ot dol´otam dol´ota dol´otami dol´otax – Stress in plural is on final syllable of root – Likewise gn’ezdo´‘nest’, zh’erlo´‘mouth’, zv’eno´‘link’, puto´, etc. • Stress on root in singular, suffix in plural (j´akor’ ∼ j´akor’a ∼ jakor’´a‘anchor’) – We’ll ignore this pattern here (4) Let’s assume, with Kenstowicz, that stress is a lexical property • Roots want to be stressed or stressless; F (stress) penalizes a root that is inappropr iately stressed • Roots that have different stress in plural are marked [+retraction] (a RETRACTION constraint is violated if a [+retraction] root has suffix stress in the plural) /dolot-a/−str,+retract RETRACTION IDENT(str) a. dolot´a *! ☞ b. dol ´ota * (5) The data of interest: double retraction ‘trade’ sg. pl. ‘string’ sg. pl. o rem´ o vol´nom. remesl´esla nom. volokn´okna gen. remesla´rem´ okonesel gen. volokna´vol´u rem´u vol´dat. remesl´eslam dat. volokn´oknam acc. remesl´esla acc. volokn´oknao rem´ o vol´om rem´ om vol´instr. remesl´eslami instr. volokn´oknami loc. remeslje´rem´oknaxeslax loc. voloknje´vol´• Likewise dupl´o ‘hollow’, b’edr´o ‘hip’, brjevn´o ‘log’, v’esl´o ‘oar’, p’atn´o ‘blemish’, etc. (6) Deriving double retraction with a paradigmatic constraint • Not predicted straightforwardly by RETRACTION constraint /remesEl-∅/−str,+retract RETRACTION IDENT(str) a. remes´el * b. rem´esel * c. r´emesel * – Kenstowicz marks extra F violations as stress moves left in the root (favors remes´el) – If we assume that roots simply want to be stressed or stressless, then all have equal vio-lations (no winner, rather than wrong winner) • Paradigmatic pressure gives rem´el/r´esel the edge over remes´emesel (cand. a � cand. b); but unfortunately, if evaluated a la McCarthy, it incorrectly prefers uniform [rem´esl-] paradigm!24.962—4 Apr 2005 p. 3 /remesEl-o/, /remesEl-a/, /remesEl-u/, /remesEl-o/, /remesEl-om/, /remesEl-je/, /remesEl-a/, /remesEl-E/, /remesEl-am/, /remesEl-a/, /remesEl-ami/, /remesEl-ax/ [−str,+retract] OP-Ident(str) RETRACTION IDENT(str) a. remesl ´o, remesl´a, remesl ´u, remesl ´o, remesl´om, remeslj´e, rem´esla, rem´esel, rem´eslam, rem´esla, rem´eslami, rem´eslax 12 × 6 = 72* 6* b. remesl ´o, remesl´a, remesl ´u, remesl ´o, remesl´om, remeslj´e, rem´esla, remes´el, rem´eslam, rem´esla, rem´eslami, rem´eslax (6 × 6) + (5 × 7) + 11 = 82* 6* c. remesl ´o, remesl´a, remesl ´u, remesl ´o, remesl´om, remeslj´e, remesl´a, remes´el, remesl´am, remesl´a, remesl´ami, remesl´ax 22* 5*! * ☞ d. rem´eslo, rem´esla, rem´eslu, rem´eslo, rem´eslom, rem´eslje, rem´esla, rem´esel, rem´eslam, rem´esla, rem´eslami, rem´eslax 0 12*! If evaluated “all or nothing” (all match or there’s disagreement), or “count the allomorphs”, same problem: All must match *Allomorphy a. * (pl. different) * (pl.) b. * (pl. different, gen. pl. yet different) ** (pl., gen. pl.) c. * (gen. pl. different) * (gen. pl) ☞ d. (all same) (all same) – It seems that sg. and pl. must act as separate paradigms, or plural will cause retraction to overapply everywhere (singular as well as gen. pl.) (7) Is a paradigmatic constraint really needed here? • We stated retraction as: don’t be stressed on the affix in the plural • Maybe we can eliminate *[remes´el] by reformulating it: don’t have final stress in plural ☞ Now the problem is rem´ ami, avoiding final stress and matching