ACTL Autumn School 2016 Anti-lexicalism in phonology and morphology Sam Steddy, UCL [email protected] 28th October 2016 Idiosyncrasy in the Lexicon Marantz (1997) notes that until c. the 1990s, different types of linguistic idiosyncrasy or irregularity were considered lexical properties. See, for example, the work of Edwin Williams on morphology, or David Pesetsky or Paul Kiparsky in phonology in the 1980s. Essentially, under lexicalist analyses, morphological idiosyncrasies are not derived in syntax, phonology, or semantics. Idiosyncratic phonology: Morphological Allomorphy (1) a. b. talk + ed speak + PAST → → (2) a. b. kid + s → child + PLURAL → talked spoke kids children 2 Idiosyncrasy in the Lexicon Today is dedicated to idiosyncrasy in phonology, but it should be noted that idiosyncrasy in meaning is an equally important topic within Distributed Morphology (Halle & Marantz 1993). Idiosyncratic meaning: Idiomaticity (3) a. b. chocolate + cake pan + cake → → (4) a. b. c. school + boy cow + boy game + boy ‘a boy who goes to school’ ??‘a boy who works with cows’ *‘a boy who goes to games’ → → → ‘a cake made of chocolate’ *‘a cake made of pans’ 3 What is Lexicalism? Without dwelling too much on the past, one way of thinking about lexicalism is that they allow some parts of a derivation to occur in the lexicon, where phonology and meaning are still ‘changeable’. LEXICON speak with PAST pan Bob the cook cake a Louise SYNTAX 4 Problems with Lexicalism On a conceptual level, lexicalism implies large scale memorisation or lexical listing of properties of morphemes or combinations of morphemes. There would be nothing wrong with this per se, but if idiosyncrasy can be derived then it is our job to work out how. • At best, some lexical approaches are open about being ’placeholder’ analyses, which suffice until more is known about the topic. • At their worst, other lexical approaches seem contradictory to principles of Universal Grammar. 5 Overview Today I’ll present two case studies which show two different ways in which phonological idiosyncrasy can be derived: i. Palatalisation in Italian verb paradigms - which is governed not by any lexical listing of the verbs a phonological rule applies to, but by phonological correspondence. ii. Suppletion in comparative adjective paradigms - otherwise termed morphologically conditioned allomorphy, which is governed by morphosyntactic structure. Though these two research domains share similar objectives, the similarity between the two has not been widely remarked on. 6 I: Phonological Correspondence Case Study: Palatalisation in Italian verbs Italian has a phonological rule that palatalises velar stops [k,g] to become affricates [tʃ, dʒ] when followed by front vowels [i,e]. (6) farmako viŋko farmatʃi vintʃi ‘medicineM.SG/PL’ ‘win1/2.SG’ For the purposes of formalism, a simple palatalisation rule is given below: (7) k,g → [+STRIDENT] / __ V+FRONT We are more interested today in where the rule applies, rather than why or how it applies. (Though on the latter point see Wilson 2006.) Note, then, that there are plenty of exceptions to this rule in Italian: (8) antiko dimentiko antiki dimentiki ‘ancientM.SG/PL’ ‘forget1/2.SG’ 8 Lexicalist-style analyses Burzio (2004), and then Krämer (2009), claim the failure to apply palatalisation is due rules like (7) being blocked in certain word forms: Eg., the feminine plural nominal suffix -e never causes palatalisation: (9) liriko liritʃi lirika lirike ‘lyricalM/F.SG’ ‘lyricalM/F.PL’ Thus, the rule is prevented from applying in feminine nouns/adjectives: (7') *k,g → [+STRIDENT] / [ __ V+FR ]N/A.FEM Many masculine nouns also fail to apply palatalisation. Krämer resorts to listing lexeme by lexeme whether the rule applies, but see Giavazzi (2010a, 2010b) for a superior stress-based analysis (which does not refer 9 to correspondence) Palatalisation in verbs Similarly, Burzio and Krämer claim that palatalisation does not apply in the first conjugation class of Italian: (10) Conj. Class 1 dimentiki, pagi ‘forget2.SG, pay2.SG’ Conj. Class 3 viŋtʃi, viŋtʃe, voldʒi, voldʒe (7'') ‘win2/3.SG, turn2/3.SG’ *k,g → [+STRIDENT] / [ __ V+FR ]CLASS 1 By the end of Krämer’s study in particular, over half the nouns and adjectives in the Italian lexicon, as well as the largest and only remaining productive verb family, have been protected from palatalisation. 10 Palatalisation by correspondence The account developed hereon is based on theories of phonological correspondence between words (Benua 2000, Kager 2000), which is also known as Output-to-Output Phonology. In brief, it will be proposed that palatalisation in conjugated verbs corresponds with palatalisation in the infinitive form of the verb - but only when the relevant segments of the infinitive receive stress. (11a) Palatalisation blocked in 1st Conjugation Class: dimentik-áre (11b) diméntik-o diméntik-i ‘forgetINF/1.SG/2.SG’ Palatalisation free to apply in 3rd Conjugation Class: víntʃ-ere vínk-o víntʃ-i ‘winINF/1.SG/2.SG’ 11 Correspondence between words One of the original uses for Phonological correspondence (Benua 2000) was to account for the emergence of marked phonological structures in otherwise unexpected positions. A famous example: (American) English words cannot end in a low front V and [ɹ] in coda position, (ie. *æɹ(c)). (12) c[ɑ]r h[ɑ]rd *c[æ]r *h[æ]rd Though this is permitted in certain nicknames (13) L[æ]rry → ✔L[æ]r B[æ]rry → ✔B[æ]r 12 Mapping correspondence relations Phonological Correspondence, or derivations that take into account more than just input and output forms, is usually formalised in Optimality Theory (as based on McCarthy & Prince’s (1996) analysis of reduplication). This can be abstracted away from by mapping out the relationship between the input form, the output form, and a third base form. Correspondence with a base usually takes priority over correspondence with the input (14) Base Input L[æ]rry Output *æɹ L[æ]rry+ ? L[æ]r *æɹ 13 Stress-dependent correspondence Consider the vowels highlighted in the classic pair of examples below (in that they pre-date theories of phonological correspondence). (15) a. cònd[ɛ]nsátion b. còmp[ə]nsátion The two forms show the same stress assignment, but the vowel in (15a) resists reduction to [ə]. There are two ways of thinking about this alternation. 14 Stress-dependent correspondence Either i) There is no base form for (15b), in contrast to (15a), meaning that the derivation is regularly input-to-output (16a) cond[ɛ]nse /condens-ation/ (16b) cònd[ɛ]nsátion *comp[ɛ]nse /compens-ation/ còmp[ə]nsátion 15 Stress-dependent correspondence Or ii) There *is* a base form for (15b), but it does not stress the relevant vowel so trivially provides [ə] for correspondence (17a) cond[ɛ]nse /condens-ation/ (17b) cond[ɛ]nsation còmp[ə]nsáte /compens-ation/ comp[ə]nsation 16 Back to Italian Let’s now consider a more comprehensive derivation of the palatalisation patterns seen in Italian verb paradigms. Something I’ve neglected to show so far: the base form should be derived by regular phonology. Derivation of palatalisation and stress in the base (ie. the infinitive): (18) a. dimentik-are → dimèntikáre ‘forget’ b. vink-ere → víntʃere (cf. viŋko) ‘win / (1.SG)’ c. tak-ere → tatʃére (cf. takwi) ‘be quiet / (PAST)’ *Note (18a-c) implies that stress is mobile in infinitive forms. This is only true in the third conjugation class (b-c), where it seems sensitive to weight, resembling the Turkish stress rule (Sezer 1981). See also Davis et al (1987). Underpalatalisation in the 1st conjugation With the infinitive derived, it can now serve as the base for the remainder of the Italian verb paradigm. First, the first conjugation class. Palatalisation is blocked before the 2.sg suffix -i (19b). (19) a. dimenti[ká]re /dimentik-o/ diménti[ko] *[ki] b. dimenti[ká]re /dimentik-i/ diménti[ki] *[ki] 18 Normal palatalisation in the 3rd conjugation Next, the third conjugation class. For verbs whose infinitives do not stress the relevant segment, palatalisation applies normally before 1.sg suffix -o and the 2.sg suffix –i: (20) a. vín[tʃe]re /vink-o/ b. vín[ko] *[ki] vín[tʃe]re /vink-i/ víntʃi *[ki] 19 Overpalatalisation in the 3rd conjugation Something perhaps unexpected is how palatalisation can overapply. This happens with third conjugation verbs whose infinitives *do* stress a relevant segment. This forces the affricate to be maintained even when the stem takes a suffixes that should not trigger the rule (21a). (21) a. ta[tʃé]re /tak-o/ b. tá[tʃo] *[ki] ta[tʃé]re /tak-i/ tá[tʃi] *[ki] 20 Can lexicalism be abandoned? Despite a reduction in its use, the analysis is not ‘lexicalism-free’: recall that stress assignment in the infinitive had to be sensitive to conjugation class. Looking at the language as a whole, though, some reference to lexical properties must be made to integrate the analysis of verbs presented here with a quite different phonology in the language’s nouns and adjectives. 21 Categorial phonology Giavazzi (2010) on palatalisation in Italian nouns and adjectives (22) k,g → [+STRIDENT] / σ.σ.__ V+HIGH,+FRONT *k,g → [+STRIDENT] / σ.__ V+HIGH,+FRONT (23) líriko antíko líritʃi antíki ‘lyricalM.SG/PL’ ‘ancientM.SG/PL’ These rules won’t work for the phonology of verbs as seen earlier consider the derivation of [víntʃere]. However, a noun/adjective vs. verbal divide in a language’s phonology is a simpler classification that that proposed by Käger. However, do consider how this split seems to be quite common crosslinguistically. (If you are interested see Smith (2011) on Categorial Phonologies.) 22 II: Morphologically Conditioned Allomorphy Suppletion Suppletion is the substitution of one phonologically unrelated root for another, as conditioned by morphological structure: (24) ADJ COMPARATIVE SUPERLATIVE a. b. long good long-er bett-er long-est be-st *c. good *good-er *good-est Traditional grammars refer to suppletive phenomena as irregular verbs, adjectives, plurals, etc. This is also termed Morphologically Conditioned Allomorphy, as opposed to regular Phonologically Conditioned Allomorphy. 24 Suppletion and allomorphy Despite bearing similarities, suppletion is not the same as regular phonological allomorphy: (25) Phonological allomorphy of the plural a. b. (26) /z/ → z /z/ → s / C[+voc] __ / C[-voc] __ eg. dog → dogs eg. cat → cats Morphological allomorphy (ie. suppletion) of the plural a. b. c. PL PL PL → ø → -(r)en → /z/ / ]N __ / ]N __ / ]N __ where N = sheep, foot, ... where N = child, ox, ... elsewhere, ie. (2) 25 More suppletion Suppletion is very common cross-linguistically, and isn’t particularly restricted to any one area of morphology: (27) a. b. c. d. Comparative Past Plural Plural long good long-er bett-er *good-er walk go walk-ed wen-t *go-ed girl person girl-s people-ø #person-s lošadj čelovek lošad-i ljud-i ‘hors(es) ‘person/PL’ Russian 26 Suppletion rules Suppletion is modelled by having ABSTRACT morphemes (otherwise known as √roots) spelled out in different ways depending on their morphological context: (28) (29) a. b. GOOD a. b. GO GOOD GO → → be(tt)good / __ ] CMPR → → wengo / __ ] PAST These spellout rules are also termed Rules of Exponence. 27 Bobaljik’s Containment Hypothesis Bobaljik (2012) shows that suppletion is conditioned by morpho-syntactic structure. Developing an understanding of the locality conditions on suppletion, Bobaljik and others (eg. Moskal 2015) elaborate on the finer details of certain morphological structures. Chief among these is that superlative adjectives are always built on comparatives, which in turn are built on simple adjectives: (30) a. b. *c. 28 Czech In some environments, this is entirely transparent, but in others it is not: (31) a. b. (32) ADJ COMPARATIVE SUPERLATIVE mlad-y špatn-y mlad-ši hor-ši nej-mlad-ši nej-hor-ši a. YOUNG → mlad- b. c. BAD → → horšpatn BAD ‘young’ ‘bad’ / __ ] CMPR 29 Suppletion patterns: ABB This pattern generalises to all of the languages in Bobaljik’s survey. The comparative, as the more local affix to the adjective, causes suppletion which is maintained in the superlative (33) a. b. c. d. e. f. ADJ COMPARATIVE SUPERLATIVE good bad špatn-y hea asko guat bett-er worse hor-ši pare-m gehi-ago pez-ar be-st wor-st nej-hor-ši par-im gehi-en pez-ar-ste ‘bad’ ‘good’ ‘many’ ‘good’ English English Czech Estonian Basque Cimbrian 30 Suppletion patterns: ABC The ABC pattern, though comparatively rare, also exists: (34) ADJ a. bon-us b. da c. maìth COMPARATIVE SUPERLATIVE mel-ior gwell ferr opt-imus gor-au dech ‘good’ ‘good’ ‘good’ Latin Welsh Old Irish 31 Suppletion patterns The following paradigm exhausts the possible patterns: (35) ADJ CMPR SPRL a. b. c. A A A A B B A B C big - bigger - biggest good - better - best bonus - melior- optimus *d. *e. A A B A A B *good - better - goodest *good - gooder - best This is based on >300 languages, c. 70 of which have comp. suppletion. From these, he identifies 116 cognate triples (111 ABB, 5 ABC). eg. good - better - best; hea - parem, par-im, etc. = 1 triple 32 Deriving ABB and ABC Given Bobaljik’s comparative and superlative structures, these two patterns are derived as a product of how many spellout rules an adjective has and how specific the rules are: (36) English a. b. (37) GOOD GOOD → → be(tt)good → → → optmelbon- / __ ] CMPR Latin a. b. c. GOOD GOOD GOOD / __ ] CMPR ] SPRL / __ ] CMPR 33 Comparative deletion NB: In languages like English, the comparative must be deleted in the context of the superlative: (38) a. *b. long - long-er - long-est long - long-er - *long-er-est This can be captured with a specific ‘null’ rule for the comparative: (39) b. CMPR → ø- / __ ] SPRL 34 Deriving *ABA If the superlative did not contain the comparative, we would expect to find ABA patterns, contrary to fact. (40) Without CMPR, suppletion context (a) would not be met, thus ‘good-est’ is wrongly predicted, via (b): a. b. GOOD GOOD → → be(tt)good / __ ] CMPR However, AAB is claimed to exist, but not for comparatives. I’ll save this data for the tutorial. 35 Anti-homophony If the superlative could be directly merged with the adjective, it would be theoretically possible to propose a spellout rule for superlative contexts that that mimicked comparative contexts: (41) *GOOD → be(tt)- / __ ] SPRL But, Bobaljik appeals to principles of antihomophony: (41) is disallowed as /be(tt)/ is the exponent of a different rule. This similarly rules out deriving an ABA pattern in the way that the ABC pattern is derived: (42) *GOOD → good / __ ] CMPR ] SPRL 36 Comparative syntax All the data shown so far are synthetic comparatives, ie. comparative forms that form one ‘word’. By zooming out, some further points can be made about how suppletion is conditioned. Analytic comparatives, where comparative structures form more than one ‘word’, show more evidence for Bobaljik’s containment hypothesis: (43) ADJ a. grand-e b. psil-ós COMPARATIVE SUPERLATIVE più grand-e pjo psil-ós il più grand-e o pjo psil-ós ‘big’ ‘tall’ Italian Greek 37 Comparative syntax Bobaljik considers comparative structures to be complex heads derived from an articulated syntactic structure (via downward head movement): (44) a. b. The intuition to the derivation is that analytic comparatives are derived when this movement does not take place. Why derive comparative structure? The derivation in (44) allows several additional facts to be understood. First, briefly, it disallows the undesired derivation seen earlier in which an adjective is directly affixed by a superlative. Second, the derivation in (44) also explains what Bobaljik terms the Synthetic Superlative Generalisation: (45) No language has synthetic superlatives (X-est) without also having synthetic comparatives (X-er) ADJ COMPARATIVE SUPERLATIVE *long more long long-est 39 The Root Suppletion Generalisation (RSG) Thirdly - and most importantly for studying suppletion - is that it explains another of Bobaljik’s observations, the Root Suppletion Generalisation. (46) Root suppletion is limited to synthetic comparatives (X-er): a. b. *c. ADJ CMPR SPRL A A A B-er more A *more B B-est most A *most B 40 The RSG This can be seen in Modern Greek, which has a relatively free alternation between synthetic vs. analytic exponence. (47) a. b. *c. ADJ COMPARATIVE SUPERLATIVE kak-ós kak-ós kak-ós cheiró-ter-os pjo kak-ós *pjo cheiró-s o cheiró-ter-os o pjo kak-ós *o pjo cheiró-s ‘bad’ 41 Locality in the RSG Thus, suppletion appears to occur when two morphemes are in the same ‘word’. To this end, Bobaljik proposes the RSG be formalised as follows: (48) The Root Suppletion Generalisation β may condition α in (a), but not (b) a. α … ]X∘ … β b. α … ]XP … β In other words, suppletion can be conditioned over a head node (Xº) but not over a phrasal node (XP). 42 Locality in the RSG The Greek data can thus be accounted for via the derivation in (49) with the rules in (27). (49) (50) a. b. a. b. BAD c. d. CMPR BAD CMPR → → cheirókak- / __ ] CMPR ] → → -tero pjo / ]aº __ ] 43 The RSG in English English also has both synthetic and analytic comparatives. (51) a. b. ADJ SYNTHETIC ANALYTIC long good long-er bett-er more long more good But, the language has a complication that continues to defy explanation. Adjectives longer than ~2 syllables cannot form synthetic comparatives (52) a. b. ADJ SYNTHETIC ANALYTIC clever intelligent clever-er *intelligent-er more clever more intelligent This leads Bobaljik to claim head movement, which derives the structure that allows suppletion, is influenced - albeit indirectly - by phonology. References Benua, Laura. 2000. Phonological relations between words. New York/London: Garland, Outstanding Dissertations in Linguistics. Bobaljik, Jonathan David. 2012. Universals in comparative morphology. Cambridge, MA: MIT Press. Burzio, Luigi. 2004. Paradigmatic and Syntagmatic Relations in Italian Verbal Inflection. In Contemporary Approaches to Romance Linguistics: Selected Proceedings of the 33rd Linguistic Symposium on Romance Languages, 17–44. Amsterdam: John Benjamins. Davis, Stuart, Linda Manganaro, and Donna Jo Napoli. 1987. Stress on Second Conjugation Infinitives in Italian. Italica 64: 477–98. Giavazzi, Maria. 2010a. The phonetics of metrical prominence and its consequences for segmental phonology. PhD Dissertation, MIT. Giavazzi, Maria. 2010b. Review of The phonology of Italian, Martin Krämer. Phonology 27: 332–341. Halle, Morris, and Alec Marantz. 1993. Distributed Morphology and the Pieces of Inflection. In The View from Building 20: Essays in Honor of Sylvain Bromberger, ed. Ken Hale and Samuel Jay Keyser, 111–176. Cambridge, MA: MIT Press. Kager, René. 2000. Surface Structure of Metrical Structure in Optimality Theory. In The Derivational Residue in Phonology, ed. Ben Hermans and Marc van Oostendorp, 207– 245. Amsterdam: John Benjamins. References Krämer, Martin. 2009. The phonology of Italian. Oxford linguistics. Oxford/New York: Oxford University Press. Marantz, Alec. 1997. No escape from syntax: Don’t try morphological analysis in the privacy of your own lexicon. University of Pennsylvania Working Papers in Linguistics 4: 201– 225. McCarthy, John, and Alan Prince. 1995. Faithfulness and Reduplicative Identity. In University of Massachusetts Occasional Papers in Linguistics 18: Papers in Optimality Theory, ed. Jill Beckman, Laura Walsh Dickey, and Suzanne Urbanczyk, 249–384. Amherst, MA: GLSA. Moskal, Beata. 2015. Limits on Allomorphy : A Case Study in Nominal Suppletion. Linguistic Inquiry 46. Sezer, Engin. 1981. On non-final stress in Turkish. Journal of Turkish Studies 5: 61–69. Wilson, Colin. 2006. Learning phonology with substantive bias: an experimental and computational study of velar palatalization. Cognitive science 30: 945–982. Smith, Jennifer L. 2011. Category-specific effects. In The Blackwell Companion to Phonology, ed. Marc van Oostendorp, Colin Ewen, Beth Hume, and Keren Rice, 2439– 2463. Malden, MA: Wiley-Blackwell.
© Copyright 2026 Paperzz