From mcginnis at ucalgary.ca Tue Dec 17 15:58:48 2002 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Tue, 17 Dec 2002 08:58:48 -0700 Subject: Carson Schutze: suppletion Message-ID: Greetings everyone, Things have been rather quiet on the list lately, and I'm avoiding my real work, so I thought I'd bring up a topic on which my colleagues recently provided me with some interesting data. The issue is suppletion, and the claim is the one summarized in the following passage, cut and pasted from the DM web site FAQ section. -------------- Traditionally it is often thought that there is a gradient between suppletion and other types of more phonologically regular allomorphy, and that no reasonable grounds can be given for how to divide the two or if they should be divided at all. Marantz 1997b has recently proposed that true suppletion occurs only for Vocabulary items in competition for f-morphemes, since competition occurs only for f-morphemes. An immediate consequence is that undeniably suppletive pairs like go/went or bad/worse must actually represent the spelling of f-morphemes. The class of f-morphemes is as a result considerably enriched, but since the class of f-morphemes is circumscribed by Universal Grammar, it is also predicted that true suppletion should be limited to universal syntactico-semantic categories. Moreover, given that some independent grounds might in this way divide suppletive from Readjustment-driven allomorphy, a theory of the range of possible Readjustment processes becomes more feasible. -------------- What I'm interested in is the question of whether true suppletion really is limited to f-morphemes, given some independent notion of what the latter could be. When I've heard Alec talk about this (I don't have the 1997b manuscript cited above so I can't verify), the idea of what fits in this class seems to be close to light verbs (if we're talking about verbs), e.g., something whose meaning might be represented just with basic primitives. E.g., I think 'give' would fit the bill, conceived of something like 'x CAUSE y BE AT z'. [I've no desire to start a debate about lexical decomposition here, just trying to give what I understand to be the flavor of the claim.] Below are some data that I think throw this claim into question. If you agree (or if you don't then hypothetically speaking), is it really any problem for DM to allow suppletion for "l-morphemes", i.e. any open-class vocabulary items? I gather that suppletion demands competition for insertion, and roots are not supposed to compete for insertion in general, but I've always been a little hazy on what would go wrong if roots did compete, and I'm particularly unclear on whether/why suppletion as a circumscribed kind of root competition is Bad. Is the former because we don't won't blocking among 'synonymous' roots? Is the latter not Evil, but simply an add-on that we would not have expected given the overall architecture of the theory? In other words, can't DM incorporate these facts just fine? I'm willing to bet it is statistically true that *most* suppletion is found in semantically light words, but couldn't that just be an effect of frequency, viz. those are the words you hear most often, and something needs to be heard pretty often in order for a suppletive form to be learned by successive generations? OK, the data. Here are some examples of what I think must be considered true suppletion whose meanings strike me as not fitting the expanded notion of f-morpheme in the quote above, i.e. cases where it seems we would need root suppletion. I claim no knowledge of these languages myself, but field work on some of them is ongoing at UCLA so in some cases I could find out more. I do believe the orthographies are transparent enough that we can be confident in the lack of phonological similarity, but also the need for full suppletion is claimed by those working on the languages. (These languages also have suppletion for some 'light' verbs, which I haven't bothered to type in.) San Lucas Quiavini Zapotec (some diacritics omitted) --------------------------- (1) Verbs a. plays-habitual: rgye'eht b. play-perfective: bzu'aht (the r- and b- prefixes are regular but nothing else is) c. vomits-habitual: reei'by d. vomit-perfective: guua'z (r- becoming gu- is regular) e. puts.on-habitual: ra'ahcw f. put.on-perfective: gwu'aht (e.g. 'put on a shirt') g. sees-habitual: raann h. see-perfective: mnnaa i. tells-habitual: re'ihpy j. tell-imperative: gwu'ahts k. sits.down-habitual: rbih l. sit.down-habitual-1pl: rzhu'enn (-enn is regular suffix) (2) Nouns a. clothes: laihdy b. clothes-poss: x:ahb (x:- is the regular prefix) c. field: loh nyaaa' d. field-poss: loh zhihah e. pueblo: guee'ihzh f. pueblo-poss: lahahzh: Ainu ---- (3) a. kill-sg.obj: rayke b. kill-pl.obj: ronnu Comanche -------- (4) a. fly-sg: yItsI b. fly-pl: yoti Pima ---- (5) a. arrive-sg: jivia b.: arrive-pl: dada c. kill-sg: mua d.: kill-pl: kokda -Carson From mcginnis at ucalgary.ca Tue Dec 17 21:17:46 2002 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Tue, 17 Dec 2002 14:17:46 -0700 Subject: Rolf Noyer: suppletion (reply to Carson Schutze) Message-ID: Hello DMers, as author of what Carson quoted in respect to suppletion I thought I should pop in here. I think this is an excellent question to investigate, and, in fact, I did think some about this problem a few years ago, but not much recently. It's not always clear what counts as a set of suppletive alternants. For example, are "moon" and "lun-" (as in "lunar") related derivationally? (Same thing for sun/solar, fox/vulpine, bear/ursine etc. etc.) On what grounds can one decide? The same question might be asked in connection with some of the examples Carson brought up. In trying to understand this issue I think it's important to keep in mind why (I think ...) Alec advanced this hypothesis to begin with. (Alec please correct my memory if I'm off base.) The problem of suppletion is a problem for learnability. Suppose you're the learner and you hear BLURK in some contexts and SNORP in others. Suppose that in the target language BLURK is the singular allomorph stem and SNORP is the plural allomorph stem, and the verb BLURK/SNORP means "to chop down a cactus." In the absence of negative evidence, it's going to be hard for the child to determine that what's different about BLURK and SNORP is simply their morphosyntactic distribution and not, say, the fact that they refer to different types of cactuses being cut down. The child might well conclude that BLURK means "cut down a short stubby cactus" and SNORP means "cut down a long thin cactus." If the semantic "slot" for the encyclopedia entry for BLURK/SNORP is somehow pre-given then the child has some assistance here, since then the problem of trying to exclude the extraneous aspects of the context of utterance from the essentials in the meaning of BLURK/SNORP is greatly simplified. From an historical perspective, one might ask: why doesn't suppletion die instantly? We have to compare this to a different situation where the alternants are BLURK/BLARK. Here the inference that BLARKing is the same thing as BLURKing is easier to make in view of the phonological similarity of the alternants. The real question, to my mind, is how much help does the learner need in solving this kind of "double variable equation"? Alec's idea is the strongest (therefore most interesting) hypothesis: it says that the child needs to know that variable X is drawn from some limited (pre-given) set, whenever Y (the phonological relationship between the allomorphs) is completely unrestricted. When Y is restricted by the property of "sufficient" phonological resemblance, then X need not be drawn from any pregiven set. It is easy to imagine that the strongest form of this hypothesis does not hold, but the basic question remains the same. I think it would be useful to discuss what all this means and whether what I have said above is in fact an accurate portrayal of the motivation for the hypotheses on restricting suppletion. If I am correct that the motivation comes from some notions about learnability, then a logically feasible alternative is that the encyclopedia entries for roots whose vocabulary items are suppletive must be somehow "at core" composed of "natural concepts" the learning of which is greatly aided in some fashion by human cognitive structure. In other words, roots which have suppletive alternants ought not to be defined by highly culture-specific properties or concepts. This of course could be wrong (Carson's "pueblo" example, if correctly glossed, might be a candidate for something of this sort). The question of course ought to be an empirical one. If one could show that there was in fact a close correlation between frequency and suppletion then there might be support for the idea that suppletion is just "extreme allomorphy" which the learner is happy to acquire given sufficiently frequent exposure. But I don't know of anyone who has demonstrated that. --Rolf From mcginnis at ucalgary.ca Wed Dec 18 15:53:26 2002 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Wed, 18 Dec 2002 08:53:26 -0700 Subject: Rolf Noyer: suppletion (reply to Carson Schutze) Message-ID: I'm new on the list, so I hope I don't incidentally violate some netiquette rule by this rather long email. (1) Romance offers, I think, some evidence that either suppletion must be available on the root level or suppletion is a kind of allomorphy: French: aller 'to go' Present: 1sg: va-is 1pl: all-ons 2sg: va-s 2pl: all-ez 3sg: va- 3pl: v-ont Future: 1sg: ir-ai 1pl: ir-ons 2sg: ir-a-s 2pl: ir-ez 3sg: ir-a- 3pl: ir-ont The rest of the paradigm is based on either all- (impairfait, simple past, participles, subjunctive - the latter one based on an allophone) or on -ir (present conditional only). Imperative forms are identical to 2sg, 2pl, 1pl (as throughout French) [There are other verbs that behave in a similar way (though not quite as extremely): http://french.about.com/library/verb/bl_verb.htm Importantly, many verbs single out 1pl and 2pl] Maybe future and present conditional can be collapsed as some soft of irrealis. What about va- forms? Option 1: singular and 3pl present tense forms are special. Then what functional morpheme can competition be about if these forms are actually less specified (either [-pl] or [-person])? Option 2: all- forms are special. I have no idea what 1pl, 2pl, infinitive, simple past, etc., have in common. At any rate, it would be weird to call most of the paradigm a special case. So a priori, suppletion is root level in Romance. However, there's one more option: Option 3: there are two all-: one for the present tense, encoding [+person, +number] and the other for everything else not encoded by ir-. It's not that crazy, because other verbs do not go quite to the same extremes as aller. NB: I'm not sure I completely understand what [+person] would mean as a feature, given that [+1] and [+2] seem to be required (evidence: exclusive [1pl] and inclusive [1,2pl] plurals), but whatever. But then we get strong evidence that suppletion is an extreme form of allomorphy because 1pl and 2pl are very often special cases for *allomorphs* in French (e.g. with the verb mourir 'die'). (This point does not seem to depend directly on whether stem suppletion is available or not, I don't think) I don't know if such multiple allomorph/suppletion cases EVER fill out totally random cells of the paradigm (e.g. Nominative dual and Locative singular only). (2) Re Carson's original examples, I don't know how much it is possible to tell from pairs that are different in one feature (as it would seem to me (strictly on the basis of the given pairs) that in SLQZ, one could say that habituals are special (the case (k) may be even more special) - the prediction then being that imperfective non-habituals (progressives) pattern with perfectives. I suspect the rest of Carson's cases could also be taken to argue for suppletion of a functional morpheme. (3) Re Rolf's BLURK and SNORP: it seems to me that these are more an issue of what are possible meanings of morphemes (presumably not "to chop down a small cactus" or "to chop down a single cactus"; nor can there be a tense form meaning "on last Tuesday" or "before sunset"). So doesn't the range of meanings need to be restricted in any theory of the lexicon? On the other hand, it seems to me that the learnability issue would restrict suppletion (or any other kind of irregularity) to a finite number of items, and then it's no longer important if there are 50 or 100 of them or what they mean (within the range of allowed meanings) - as long as it's a reasonably learnable number. The standard view of Russian perfective formation claims that which prefix (of about 20) is used to form the default perfective of a given verb (perfectives are (usually) formed from imperfective stems by adding a prefix - some perfectives have a special meaning (e.g. "to run OVER", "run-ON"), but there's usually one that just makes *that* verb perfective (e.g. "run-PERF") - that's the default one) is an idiosyncratic property of that verb. If kids can learn this, evem 500 suppletive stems cannot be a problem :) (I must add here that I'm not at all sure that this standard view is correct, nonetheless I don't know of any better hypothesis.) So my conclusion would be that, yeah, some meanings are not possible as suppletive items - nor are they possible as any kind of a lexical item, and there are some generalizations to be established (e.g. "a morpheme cannot collapse the meanings of non-light verbs", or something like this - I'm not claiming this is true, just noting things better be constrained in these ways as well). It's an extremely interesting question (what's "natural"? why are some meanings good as adjectives and others aren't?), but I don't think it's special to suppletion. Finally, Carson's "pueblo" example, I think it's just "village" and it's us who call it a pueblo for our own reasons. O From mcginnis at ucalgary.ca Wed Dec 18 15:54:53 2002 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Wed, 18 Dec 2002 08:54:53 -0700 Subject: Rolf Noyer: suppletion (reply to Carson Schutze) Message-ID: Sorry, I failed to change the headers. That last message was from Ora Matushansky. Martha McGinnis Moderator, DM-List From mcginnis at ucalgary.ca Wed Dec 18 16:01:55 2002 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Wed, 18 Dec 2002 09:01:55 -0700 Subject: Heidi Harley: suppletion (reply to Carson Schutze) Message-ID: Hi all -- Re Carson's suppletion observations: Rolf and I talked about this a while ago, and he had some example of suppletion in a verb from some language that meant "dive", which we thought was an unlikely f-morpheme. More recently, I've run across suppletion in Yaqui verbs, suppleting for plural number of the subject or object (which normally doesn't trigger any change in the verb form). Most of them are pretty good candidates for f-morphemes, but some maybe not; below is the set from Dedrick and Casad (2000) and also, where distinct from Molina, Shaul, and Valenzuela (1999). I've put asterisks by the ones where the suppletion is total, involving even the first phoneme of the word. (Yaqui is Uto-Aztecan; I understand that these suppletions for number are common across the language family; I'll keep my eyes open for more. It should be easy to find 'em in Hopi, for instance.) Verb Sg. Pl (intr, supplete for # of subject) *run vuite tenne *walk weama rehte *go weye kaate arrive yepsa yaha enter kivake kiimu leave siime saka *die muuke koko *be located katek hooka sit weche watte *be sitting yehte hoote *be standing weyek ha'abwek *lie down vo'ote to'ote *sleep over vo'e to'e (tr, supplete for # of object) *kill me'aa su'aa *put yecha hoa *pick up tovokta hahau *carry toha weiya bring kivacha kiima I'm inclined to think that Carson's right about the statistically conditioned nature of suppletion merely making it coincidental that most suppletion seems to be with 'light' type elements; f-morphemes, if you will. It does seem to me that the problem of suppletive l-morphemes is a grave one. As carson outlined, the original Halle/Marantz idea with pairs like rise/raise was that some readjustment rule applies to the actual phonological string that the vocabulary item inserts and the encylopedia entry indexes. Carson asks, > >... is it really any >problem for DM to allow suppletion for "l-morphemes", i.e. any open-class >vocabulary items? I gather that suppletion demands competition for >insertion, and roots are not supposed to compete for insertion in general, >but I've always been a little hazy on what would go wrong if roots did >compete, and I'm particularly unclear on whether/why suppletion as a >circumscribed kind of root competition is Bad. Is the former because we >don't won't blocking among 'synonymous' roots? Is the latter not Evil, but >simply an add-on that we would not have expected given the overall >architecture of the theory? In other words, can't DM incorporate these facts >just fine? The way competition among l-morphemes would have to work is that there would have to be an Ur-root, say BAD (for bad/worse), which gets selected in the numeration along with the other purely morphosyntactic features. Then vocabulary insertion will insert "bad" for BAD as an elsewhere case, and "worse" for BAD as the case conditioned by comparatives. (The Ur-morpheme I mention above is like the notion of E-root proposed by Mark Volpe in a recent ms.) (If you didn't have Ur-roots, "bad" would be in competition with "good" and "smart" and "big" etc.) Now, proposing Ur-roots might not be an obviously bad thing; it's basically saying that Fodor is Right: we have all these (unanalyzable) concepts floating around, and they get hooked up to particular phonological strings. I don't have a problem with that. The problem, as Rolf and I described Alec's observation (I don't have Alec's original ms. either, darnit, so have to rely on our description of it from our Peeters volume paper), is this: "In terms of learnability, as Marantz suggests, given a space of universal conceptual primes, the child can associate two phonologically unrelated VIs with some cell in that space. But without this pre-given structure, the child has no way of determining that two phonologically unrelated alternants do not in fact denote two different sorts of objects (or predicates)." So it comes down to a question of acquisition: it's crucially the regularity across environments of the form/meaning relationship that lets you asociate phonological shape "a" with concept A. What happens if concept A is sometimes realized with phonological shape "a" and sometimes with phonological shape "b"? Shouldn't you guess that "a" and "b", although very similar, refer to different concepts? (How else can you figure out that "wheat" and "oats" are different concepts? The fact that they have different names is a big clue, it seems to me). So you'd end up with a situation where an English learner thinks that "go" refers to concept A, and "went" refers to concept B, and they happen to get used in very similar circs (like "Doberman" and "St. Bernard"). As I write this I think that maybe the solution is not so hard after all. Suppletion occurs when phonologically dissimilar strings refer to the same concept in morphosyntactically distinct contexts ([+past], [+compar], etc). All the learner would have to do is figure out that there is a 'paradigm gap' for a particular phonological string/concept mapping -- he never hears 'went' in situations where the present tense is present; he never hears 'go' in the past -- and then make the appropriate association. Certainly we know that kids are monstrous statistical machines; they keep track of everything. let's say their first pass conceptual mapping assigns 'go' to A and 'went' to B -- but that as they accumulate stats on 'go' and 'went' over time, they can notice that they are in complementary distribution w/r to nonpast/past, and that their conditions of use are otherwise extemely similar. Having the ability to noticing complementary distribution is crucial for other purposes, right? So the upshot of that idea would be that suppletion can only be conditioned by universal morphosyntactic features, kind of the definition of suppletion, I suppose. You shouldn't, then, see suppletion conditioned by participation in an idiom, e.g., or by cooccurence with 'dog' rather than 'horse'. Of course, assuming that the numeration has access to Ur-roots makes one think that the morphosyntax (e.g. vocabulary insertion of f-morphemes) *could* be conditioned by the difference between 'dog' and 'horse' -- it's rather counter to the whole point of Late Insertion. We'd have to come up with another reason why the morphosyntax doesn't pay attention to non-linguistic conceptual differences, or else stipulate that Ur-roots are inserted first, at vocabulary insertion, and then competition for suppletion happens -- essentailly making suppletion a readjustment rule too, which doesn't seem right. Hmm. thoughts, anyone? yrs musingly, hh >-- Original Message -- >Date: Tue, 17 Dec 2002 08:58:48 -0700 >Reply-To: The Distributed Morphology List > >From: Martha McGinnis >Subject: Carson Schutze: suppletion >To: DM-LIST at LISTSERV.LINGUISTLIST.ORG > > >Greetings everyone, >Things have been rather quiet on the list lately, and I'm avoiding my real >work, so I thought I'd bring up a topic on which my colleagues recently >provided me with some interesting data. The issue is suppletion, and the >claim is the one summarized in the following passage, cut and pasted from >the DM web site FAQ section. > >-------------- >Traditionally it is often thought that there is a gradient between >suppletion and other types of more phonologically regular allomorphy, and >that no reasonable grounds can be given for how to divide the two or if they >should be divided at all. Marantz 1997b has recently proposed that true >suppletion occurs only for Vocabulary items in competition for f-morphemes, >since competition occurs only for f-morphemes. An immediate consequence >is >that undeniably suppletive >pairs like go/went or bad/worse must actually represent the spelling of >f-morphemes. The class of f-morphemes is as a result considerably enriched, >but since the class of f-morphemes is circumscribed by Universal Grammar, >it >is also predicted that true suppletion should be limited to universal >syntactico-semantic categories. Moreover, given that some independent >grounds might in this way divide suppletive from Readjustment-driven >allomorphy, a theory of the range of possible Readjustment processes becomes >more feasible. >-------------- > >What I'm interested in is the question of whether true suppletion really >is >limited to f-morphemes, given some independent notion of what the latter >could be. When I've heard Alec talk about this (I don't have the 1997b >manuscript cited above so I can't verify), the idea of what fits in this >class seems to be close to light verbs (if we're talking about verbs), e.g., >something whose meaning might be represented just with basic primitives. >E.g., I think 'give' would fit the bill, conceived of something like 'x >CAUSE y BE AT z'. [I've no desire to start a debate about lexical >decomposition here, just trying to give what I understand to be the flavor >of the claim.] > >Below are some data that I think throw this claim into question. If you >agree (or if you don't then hypothetically speaking), is it really any >problem for DM to allow suppletion for "l-morphemes", i.e. any open-class >vocabulary items? I gather that suppletion demands competition for >insertion, and roots are not supposed to compete for insertion in general, >but I've always been a little hazy on what would go wrong if roots did >compete, and I'm particularly unclear on whether/why suppletion as a >circumscribed kind of root competition is Bad. Is the former because we >don't won't blocking among 'synonymous' roots? Is the latter not Evil, but >simply an add-on that we would not have expected given the overall >architecture of the theory? In other words, can't DM incorporate these facts >just fine? > >I'm willing to bet it is statistically true that *most* suppletion is found >in semantically light words, but couldn't that just be an effect of >frequency, viz. those are the words you hear most often, and something needs >to be heard pretty often in order for a suppletive form to be learned by >successive generations? > >OK, the data. Here are some examples of what I think must be considered true >suppletion whose meanings strike me as not fitting the expanded notion of >f-morpheme in the quote above, i.e. cases where it seems we would need root >suppletion. I claim no knowledge of these languages myself, but field work >on some of them is ongoing at UCLA so in some cases I could find out more. >I >do believe the orthographies are transparent enough that we can be confident >in the lack of phonological similarity, but also the need for full >suppletion is claimed by those working on the languages. > >(These languages also have suppletion for some 'light' verbs, which I >haven't bothered to type in.) > > >San Lucas Quiavini Zapotec (some diacritics omitted) >--------------------------- >(1) Verbs >a. plays-habitual: rgye'eht b. play-perfective: bzu'aht >(the r- and b- prefixes are regular but nothing else is) > >c. vomits-habitual: reei'by d. vomit-perfective: guua'z >(r- becoming gu- is regular) > >e. puts.on-habitual: ra'ahcw f. put.on-perfective: gwu'aht >(e.g. 'put on a shirt') > >g. sees-habitual: raann h. see-perfective: mnnaa > >i. tells-habitual: re'ihpy j. tell-imperative: gwu'ahts > >k. sits.down-habitual: rbih l. sit.down-habitual-1pl: rzhu'enn > (-enn is regular suffix) > >(2) Nouns >a. clothes: laihdy b. clothes-poss: x:ahb >(x:- is the regular prefix) >c. field: loh nyaaa' d. field-poss: loh zhihah >e. pueblo: guee'ihzh f. pueblo-poss: lahahzh: > > >Ainu >---- >(3) a. kill-sg.obj: rayke b. kill-pl.obj: ronnu > >Comanche >-------- >(4) a. fly-sg: yItsI b. fly-pl: yoti > >Pima >---- >(5) a. arrive-sg: jivia b.: arrive-pl: dada >c. kill-sg: mua d.: kill-pl: kokda > >-Carson --------------------------------------------------------------------- Heidi Harley Department of Linguistics Douglass 200E University of Arizona Tucson, AZ 85721 Ph: (520) 626-3554 Fax: (520) 626-9014 hharley at u.arizona.edu From mcginnis at ucalgary.ca Wed Dec 18 16:02:37 2002 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Wed, 18 Dec 2002 09:02:37 -0700 Subject: Dan Everett: suppletion (reply to Carson Schutze) Message-ID: Carson's questions and Rolf's answer were both interesting. Let me suggest an additional example to those Carson brings up. In Wari' many verbs have suppletive forms on an ergative pattern (described in the Everett & Kern 1997 Routledge Grammar). I don't have the examples off the top of my head and am typing this at home, but the process is simple. Many intransitive verbs have suppletive forms for plural subjects while many transitive verbs have suppletive forms for plural objects. The suppletion seems to run from what Carson refers to as 'true' suppletion (such qualifiers make me wince, but I will resist the temptation to say why) and 'Readjustment'. I will close with a digression that those who know me will not find surprising: statements to the effect that 'hypothesis X is more interesting because it is stronger' are a form of epistemological sloganeering with little to commend them. Dan ......................... Dan Everett Professor of Phonetics and Phonology Department of Linguistics Arts Building University of Manchester Oxford Road M13 9PL Manchester, UK dan.everett at man.ac.uk Dept. Fax and Phone: 44-161-275-3187 From mcginnis at ucalgary.ca Tue Dec 17 15:58:48 2002 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Tue, 17 Dec 2002 08:58:48 -0700 Subject: Carson Schutze: suppletion Message-ID: Greetings everyone, Things have been rather quiet on the list lately, and I'm avoiding my real work, so I thought I'd bring up a topic on which my colleagues recently provided me with some interesting data. The issue is suppletion, and the claim is the one summarized in the following passage, cut and pasted from the DM web site FAQ section. -------------- Traditionally it is often thought that there is a gradient between suppletion and other types of more phonologically regular allomorphy, and that no reasonable grounds can be given for how to divide the two or if they should be divided at all. Marantz 1997b has recently proposed that true suppletion occurs only for Vocabulary items in competition for f-morphemes, since competition occurs only for f-morphemes. An immediate consequence is that undeniably suppletive pairs like go/went or bad/worse must actually represent the spelling of f-morphemes. The class of f-morphemes is as a result considerably enriched, but since the class of f-morphemes is circumscribed by Universal Grammar, it is also predicted that true suppletion should be limited to universal syntactico-semantic categories. Moreover, given that some independent grounds might in this way divide suppletive from Readjustment-driven allomorphy, a theory of the range of possible Readjustment processes becomes more feasible. -------------- What I'm interested in is the question of whether true suppletion really is limited to f-morphemes, given some independent notion of what the latter could be. When I've heard Alec talk about this (I don't have the 1997b manuscript cited above so I can't verify), the idea of what fits in this class seems to be close to light verbs (if we're talking about verbs), e.g., something whose meaning might be represented just with basic primitives. E.g., I think 'give' would fit the bill, conceived of something like 'x CAUSE y BE AT z'. [I've no desire to start a debate about lexical decomposition here, just trying to give what I understand to be the flavor of the claim.] Below are some data that I think throw this claim into question. If you agree (or if you don't then hypothetically speaking), is it really any problem for DM to allow suppletion for "l-morphemes", i.e. any open-class vocabulary items? I gather that suppletion demands competition for insertion, and roots are not supposed to compete for insertion in general, but I've always been a little hazy on what would go wrong if roots did compete, and I'm particularly unclear on whether/why suppletion as a circumscribed kind of root competition is Bad. Is the former because we don't won't blocking among 'synonymous' roots? Is the latter not Evil, but simply an add-on that we would not have expected given the overall architecture of the theory? In other words, can't DM incorporate these facts just fine? I'm willing to bet it is statistically true that *most* suppletion is found in semantically light words, but couldn't that just be an effect of frequency, viz. those are the words you hear most often, and something needs to be heard pretty often in order for a suppletive form to be learned by successive generations? OK, the data. Here are some examples of what I think must be considered true suppletion whose meanings strike me as not fitting the expanded notion of f-morpheme in the quote above, i.e. cases where it seems we would need root suppletion. I claim no knowledge of these languages myself, but field work on some of them is ongoing at UCLA so in some cases I could find out more. I do believe the orthographies are transparent enough that we can be confident in the lack of phonological similarity, but also the need for full suppletion is claimed by those working on the languages. (These languages also have suppletion for some 'light' verbs, which I haven't bothered to type in.) San Lucas Quiavini Zapotec (some diacritics omitted) --------------------------- (1) Verbs a. plays-habitual: rgye'eht b. play-perfective: bzu'aht (the r- and b- prefixes are regular but nothing else is) c. vomits-habitual: reei'by d. vomit-perfective: guua'z (r- becoming gu- is regular) e. puts.on-habitual: ra'ahcw f. put.on-perfective: gwu'aht (e.g. 'put on a shirt') g. sees-habitual: raann h. see-perfective: mnnaa i. tells-habitual: re'ihpy j. tell-imperative: gwu'ahts k. sits.down-habitual: rbih l. sit.down-habitual-1pl: rzhu'enn (-enn is regular suffix) (2) Nouns a. clothes: laihdy b. clothes-poss: x:ahb (x:- is the regular prefix) c. field: loh nyaaa' d. field-poss: loh zhihah e. pueblo: guee'ihzh f. pueblo-poss: lahahzh: Ainu ---- (3) a. kill-sg.obj: rayke b. kill-pl.obj: ronnu Comanche -------- (4) a. fly-sg: yItsI b. fly-pl: yoti Pima ---- (5) a. arrive-sg: jivia b.: arrive-pl: dada c. kill-sg: mua d.: kill-pl: kokda -Carson From mcginnis at ucalgary.ca Tue Dec 17 21:17:46 2002 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Tue, 17 Dec 2002 14:17:46 -0700 Subject: Rolf Noyer: suppletion (reply to Carson Schutze) Message-ID: Hello DMers, as author of what Carson quoted in respect to suppletion I thought I should pop in here. I think this is an excellent question to investigate, and, in fact, I did think some about this problem a few years ago, but not much recently. It's not always clear what counts as a set of suppletive alternants. For example, are "moon" and "lun-" (as in "lunar") related derivationally? (Same thing for sun/solar, fox/vulpine, bear/ursine etc. etc.) On what grounds can one decide? The same question might be asked in connection with some of the examples Carson brought up. In trying to understand this issue I think it's important to keep in mind why (I think ...) Alec advanced this hypothesis to begin with. (Alec please correct my memory if I'm off base.) The problem of suppletion is a problem for learnability. Suppose you're the learner and you hear BLURK in some contexts and SNORP in others. Suppose that in the target language BLURK is the singular allomorph stem and SNORP is the plural allomorph stem, and the verb BLURK/SNORP means "to chop down a cactus." In the absence of negative evidence, it's going to be hard for the child to determine that what's different about BLURK and SNORP is simply their morphosyntactic distribution and not, say, the fact that they refer to different types of cactuses being cut down. The child might well conclude that BLURK means "cut down a short stubby cactus" and SNORP means "cut down a long thin cactus." If the semantic "slot" for the encyclopedia entry for BLURK/SNORP is somehow pre-given then the child has some assistance here, since then the problem of trying to exclude the extraneous aspects of the context of utterance from the essentials in the meaning of BLURK/SNORP is greatly simplified. From an historical perspective, one might ask: why doesn't suppletion die instantly? We have to compare this to a different situation where the alternants are BLURK/BLARK. Here the inference that BLARKing is the same thing as BLURKing is easier to make in view of the phonological similarity of the alternants. The real question, to my mind, is how much help does the learner need in solving this kind of "double variable equation"? Alec's idea is the strongest (therefore most interesting) hypothesis: it says that the child needs to know that variable X is drawn from some limited (pre-given) set, whenever Y (the phonological relationship between the allomorphs) is completely unrestricted. When Y is restricted by the property of "sufficient" phonological resemblance, then X need not be drawn from any pregiven set. It is easy to imagine that the strongest form of this hypothesis does not hold, but the basic question remains the same. I think it would be useful to discuss what all this means and whether what I have said above is in fact an accurate portrayal of the motivation for the hypotheses on restricting suppletion. If I am correct that the motivation comes from some notions about learnability, then a logically feasible alternative is that the encyclopedia entries for roots whose vocabulary items are suppletive must be somehow "at core" composed of "natural concepts" the learning of which is greatly aided in some fashion by human cognitive structure. In other words, roots which have suppletive alternants ought not to be defined by highly culture-specific properties or concepts. This of course could be wrong (Carson's "pueblo" example, if correctly glossed, might be a candidate for something of this sort). The question of course ought to be an empirical one. If one could show that there was in fact a close correlation between frequency and suppletion then there might be support for the idea that suppletion is just "extreme allomorphy" which the learner is happy to acquire given sufficiently frequent exposure. But I don't know of anyone who has demonstrated that. --Rolf From mcginnis at ucalgary.ca Wed Dec 18 15:53:26 2002 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Wed, 18 Dec 2002 08:53:26 -0700 Subject: Rolf Noyer: suppletion (reply to Carson Schutze) Message-ID: I'm new on the list, so I hope I don't incidentally violate some netiquette rule by this rather long email. (1) Romance offers, I think, some evidence that either suppletion must be available on the root level or suppletion is a kind of allomorphy: French: aller 'to go' Present: 1sg: va-is 1pl: all-ons 2sg: va-s 2pl: all-ez 3sg: va- 3pl: v-ont Future: 1sg: ir-ai 1pl: ir-ons 2sg: ir-a-s 2pl: ir-ez 3sg: ir-a- 3pl: ir-ont The rest of the paradigm is based on either all- (impairfait, simple past, participles, subjunctive - the latter one based on an allophone) or on -ir (present conditional only). Imperative forms are identical to 2sg, 2pl, 1pl (as throughout French) [There are other verbs that behave in a similar way (though not quite as extremely): http://french.about.com/library/verb/bl_verb.htm Importantly, many verbs single out 1pl and 2pl] Maybe future and present conditional can be collapsed as some soft of irrealis. What about va- forms? Option 1: singular and 3pl present tense forms are special. Then what functional morpheme can competition be about if these forms are actually less specified (either [-pl] or [-person])? Option 2: all- forms are special. I have no idea what 1pl, 2pl, infinitive, simple past, etc., have in common. At any rate, it would be weird to call most of the paradigm a special case. So a priori, suppletion is root level in Romance. However, there's one more option: Option 3: there are two all-: one for the present tense, encoding [+person, +number] and the other for everything else not encoded by ir-. It's not that crazy, because other verbs do not go quite to the same extremes as aller. NB: I'm not sure I completely understand what [+person] would mean as a feature, given that [+1] and [+2] seem to be required (evidence: exclusive [1pl] and inclusive [1,2pl] plurals), but whatever. But then we get strong evidence that suppletion is an extreme form of allomorphy because 1pl and 2pl are very often special cases for *allomorphs* in French (e.g. with the verb mourir 'die'). (This point does not seem to depend directly on whether stem suppletion is available or not, I don't think) I don't know if such multiple allomorph/suppletion cases EVER fill out totally random cells of the paradigm (e.g. Nominative dual and Locative singular only). (2) Re Carson's original examples, I don't know how much it is possible to tell from pairs that are different in one feature (as it would seem to me (strictly on the basis of the given pairs) that in SLQZ, one could say that habituals are special (the case (k) may be even more special) - the prediction then being that imperfective non-habituals (progressives) pattern with perfectives. I suspect the rest of Carson's cases could also be taken to argue for suppletion of a functional morpheme. (3) Re Rolf's BLURK and SNORP: it seems to me that these are more an issue of what are possible meanings of morphemes (presumably not "to chop down a small cactus" or "to chop down a single cactus"; nor can there be a tense form meaning "on last Tuesday" or "before sunset"). So doesn't the range of meanings need to be restricted in any theory of the lexicon? On the other hand, it seems to me that the learnability issue would restrict suppletion (or any other kind of irregularity) to a finite number of items, and then it's no longer important if there are 50 or 100 of them or what they mean (within the range of allowed meanings) - as long as it's a reasonably learnable number. The standard view of Russian perfective formation claims that which prefix (of about 20) is used to form the default perfective of a given verb (perfectives are (usually) formed from imperfective stems by adding a prefix - some perfectives have a special meaning (e.g. "to run OVER", "run-ON"), but there's usually one that just makes *that* verb perfective (e.g. "run-PERF") - that's the default one) is an idiosyncratic property of that verb. If kids can learn this, evem 500 suppletive stems cannot be a problem :) (I must add here that I'm not at all sure that this standard view is correct, nonetheless I don't know of any better hypothesis.) So my conclusion would be that, yeah, some meanings are not possible as suppletive items - nor are they possible as any kind of a lexical item, and there are some generalizations to be established (e.g. "a morpheme cannot collapse the meanings of non-light verbs", or something like this - I'm not claiming this is true, just noting things better be constrained in these ways as well). It's an extremely interesting question (what's "natural"? why are some meanings good as adjectives and others aren't?), but I don't think it's special to suppletion. Finally, Carson's "pueblo" example, I think it's just "village" and it's us who call it a pueblo for our own reasons. O From mcginnis at ucalgary.ca Wed Dec 18 15:54:53 2002 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Wed, 18 Dec 2002 08:54:53 -0700 Subject: Rolf Noyer: suppletion (reply to Carson Schutze) Message-ID: Sorry, I failed to change the headers. That last message was from Ora Matushansky. Martha McGinnis Moderator, DM-List From mcginnis at ucalgary.ca Wed Dec 18 16:01:55 2002 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Wed, 18 Dec 2002 09:01:55 -0700 Subject: Heidi Harley: suppletion (reply to Carson Schutze) Message-ID: Hi all -- Re Carson's suppletion observations: Rolf and I talked about this a while ago, and he had some example of suppletion in a verb from some language that meant "dive", which we thought was an unlikely f-morpheme. More recently, I've run across suppletion in Yaqui verbs, suppleting for plural number of the subject or object (which normally doesn't trigger any change in the verb form). Most of them are pretty good candidates for f-morphemes, but some maybe not; below is the set from Dedrick and Casad (2000) and also, where distinct from Molina, Shaul, and Valenzuela (1999). I've put asterisks by the ones where the suppletion is total, involving even the first phoneme of the word. (Yaqui is Uto-Aztecan; I understand that these suppletions for number are common across the language family; I'll keep my eyes open for more. It should be easy to find 'em in Hopi, for instance.) Verb Sg. Pl (intr, supplete for # of subject) *run vuite tenne *walk weama rehte *go weye kaate arrive yepsa yaha enter kivake kiimu leave siime saka *die muuke koko *be located katek hooka sit weche watte *be sitting yehte hoote *be standing weyek ha'abwek *lie down vo'ote to'ote *sleep over vo'e to'e (tr, supplete for # of object) *kill me'aa su'aa *put yecha hoa *pick up tovokta hahau *carry toha weiya bring kivacha kiima I'm inclined to think that Carson's right about the statistically conditioned nature of suppletion merely making it coincidental that most suppletion seems to be with 'light' type elements; f-morphemes, if you will. It does seem to me that the problem of suppletive l-morphemes is a grave one. As carson outlined, the original Halle/Marantz idea with pairs like rise/raise was that some readjustment rule applies to the actual phonological string that the vocabulary item inserts and the encylopedia entry indexes. Carson asks, > >... is it really any >problem for DM to allow suppletion for "l-morphemes", i.e. any open-class >vocabulary items? I gather that suppletion demands competition for >insertion, and roots are not supposed to compete for insertion in general, >but I've always been a little hazy on what would go wrong if roots did >compete, and I'm particularly unclear on whether/why suppletion as a >circumscribed kind of root competition is Bad. Is the former because we >don't won't blocking among 'synonymous' roots? Is the latter not Evil, but >simply an add-on that we would not have expected given the overall >architecture of the theory? In other words, can't DM incorporate these facts >just fine? The way competition among l-morphemes would have to work is that there would have to be an Ur-root, say BAD (for bad/worse), which gets selected in the numeration along with the other purely morphosyntactic features. Then vocabulary insertion will insert "bad" for BAD as an elsewhere case, and "worse" for BAD as the case conditioned by comparatives. (The Ur-morpheme I mention above is like the notion of E-root proposed by Mark Volpe in a recent ms.) (If you didn't have Ur-roots, "bad" would be in competition with "good" and "smart" and "big" etc.) Now, proposing Ur-roots might not be an obviously bad thing; it's basically saying that Fodor is Right: we have all these (unanalyzable) concepts floating around, and they get hooked up to particular phonological strings. I don't have a problem with that. The problem, as Rolf and I described Alec's observation (I don't have Alec's original ms. either, darnit, so have to rely on our description of it from our Peeters volume paper), is this: "In terms of learnability, as Marantz suggests, given a space of universal conceptual primes, the child can associate two phonologically unrelated VIs with some cell in that space. But without this pre-given structure, the child has no way of determining that two phonologically unrelated alternants do not in fact denote two different sorts of objects (or predicates)." So it comes down to a question of acquisition: it's crucially the regularity across environments of the form/meaning relationship that lets you asociate phonological shape "a" with concept A. What happens if concept A is sometimes realized with phonological shape "a" and sometimes with phonological shape "b"? Shouldn't you guess that "a" and "b", although very similar, refer to different concepts? (How else can you figure out that "wheat" and "oats" are different concepts? The fact that they have different names is a big clue, it seems to me). So you'd end up with a situation where an English learner thinks that "go" refers to concept A, and "went" refers to concept B, and they happen to get used in very similar circs (like "Doberman" and "St. Bernard"). As I write this I think that maybe the solution is not so hard after all. Suppletion occurs when phonologically dissimilar strings refer to the same concept in morphosyntactically distinct contexts ([+past], [+compar], etc). All the learner would have to do is figure out that there is a 'paradigm gap' for a particular phonological string/concept mapping -- he never hears 'went' in situations where the present tense is present; he never hears 'go' in the past -- and then make the appropriate association. Certainly we know that kids are monstrous statistical machines; they keep track of everything. let's say their first pass conceptual mapping assigns 'go' to A and 'went' to B -- but that as they accumulate stats on 'go' and 'went' over time, they can notice that they are in complementary distribution w/r to nonpast/past, and that their conditions of use are otherwise extemely similar. Having the ability to noticing complementary distribution is crucial for other purposes, right? So the upshot of that idea would be that suppletion can only be conditioned by universal morphosyntactic features, kind of the definition of suppletion, I suppose. You shouldn't, then, see suppletion conditioned by participation in an idiom, e.g., or by cooccurence with 'dog' rather than 'horse'. Of course, assuming that the numeration has access to Ur-roots makes one think that the morphosyntax (e.g. vocabulary insertion of f-morphemes) *could* be conditioned by the difference between 'dog' and 'horse' -- it's rather counter to the whole point of Late Insertion. We'd have to come up with another reason why the morphosyntax doesn't pay attention to non-linguistic conceptual differences, or else stipulate that Ur-roots are inserted first, at vocabulary insertion, and then competition for suppletion happens -- essentailly making suppletion a readjustment rule too, which doesn't seem right. Hmm. thoughts, anyone? yrs musingly, hh >-- Original Message -- >Date: Tue, 17 Dec 2002 08:58:48 -0700 >Reply-To: The Distributed Morphology List > >From: Martha McGinnis >Subject: Carson Schutze: suppletion >To: DM-LIST at LISTSERV.LINGUISTLIST.ORG > > >Greetings everyone, >Things have been rather quiet on the list lately, and I'm avoiding my real >work, so I thought I'd bring up a topic on which my colleagues recently >provided me with some interesting data. The issue is suppletion, and the >claim is the one summarized in the following passage, cut and pasted from >the DM web site FAQ section. > >-------------- >Traditionally it is often thought that there is a gradient between >suppletion and other types of more phonologically regular allomorphy, and >that no reasonable grounds can be given for how to divide the two or if they >should be divided at all. Marantz 1997b has recently proposed that true >suppletion occurs only for Vocabulary items in competition for f-morphemes, >since competition occurs only for f-morphemes. An immediate consequence >is >that undeniably suppletive >pairs like go/went or bad/worse must actually represent the spelling of >f-morphemes. The class of f-morphemes is as a result considerably enriched, >but since the class of f-morphemes is circumscribed by Universal Grammar, >it >is also predicted that true suppletion should be limited to universal >syntactico-semantic categories. Moreover, given that some independent >grounds might in this way divide suppletive from Readjustment-driven >allomorphy, a theory of the range of possible Readjustment processes becomes >more feasible. >-------------- > >What I'm interested in is the question of whether true suppletion really >is >limited to f-morphemes, given some independent notion of what the latter >could be. When I've heard Alec talk about this (I don't have the 1997b >manuscript cited above so I can't verify), the idea of what fits in this >class seems to be close to light verbs (if we're talking about verbs), e.g., >something whose meaning might be represented just with basic primitives. >E.g., I think 'give' would fit the bill, conceived of something like 'x >CAUSE y BE AT z'. [I've no desire to start a debate about lexical >decomposition here, just trying to give what I understand to be the flavor >of the claim.] > >Below are some data that I think throw this claim into question. If you >agree (or if you don't then hypothetically speaking), is it really any >problem for DM to allow suppletion for "l-morphemes", i.e. any open-class >vocabulary items? I gather that suppletion demands competition for >insertion, and roots are not supposed to compete for insertion in general, >but I've always been a little hazy on what would go wrong if roots did >compete, and I'm particularly unclear on whether/why suppletion as a >circumscribed kind of root competition is Bad. Is the former because we >don't won't blocking among 'synonymous' roots? Is the latter not Evil, but >simply an add-on that we would not have expected given the overall >architecture of the theory? In other words, can't DM incorporate these facts >just fine? > >I'm willing to bet it is statistically true that *most* suppletion is found >in semantically light words, but couldn't that just be an effect of >frequency, viz. those are the words you hear most often, and something needs >to be heard pretty often in order for a suppletive form to be learned by >successive generations? > >OK, the data. Here are some examples of what I think must be considered true >suppletion whose meanings strike me as not fitting the expanded notion of >f-morpheme in the quote above, i.e. cases where it seems we would need root >suppletion. I claim no knowledge of these languages myself, but field work >on some of them is ongoing at UCLA so in some cases I could find out more. >I >do believe the orthographies are transparent enough that we can be confident >in the lack of phonological similarity, but also the need for full >suppletion is claimed by those working on the languages. > >(These languages also have suppletion for some 'light' verbs, which I >haven't bothered to type in.) > > >San Lucas Quiavini Zapotec (some diacritics omitted) >--------------------------- >(1) Verbs >a. plays-habitual: rgye'eht b. play-perfective: bzu'aht >(the r- and b- prefixes are regular but nothing else is) > >c. vomits-habitual: reei'by d. vomit-perfective: guua'z >(r- becoming gu- is regular) > >e. puts.on-habitual: ra'ahcw f. put.on-perfective: gwu'aht >(e.g. 'put on a shirt') > >g. sees-habitual: raann h. see-perfective: mnnaa > >i. tells-habitual: re'ihpy j. tell-imperative: gwu'ahts > >k. sits.down-habitual: rbih l. sit.down-habitual-1pl: rzhu'enn > (-enn is regular suffix) > >(2) Nouns >a. clothes: laihdy b. clothes-poss: x:ahb >(x:- is the regular prefix) >c. field: loh nyaaa' d. field-poss: loh zhihah >e. pueblo: guee'ihzh f. pueblo-poss: lahahzh: > > >Ainu >---- >(3) a. kill-sg.obj: rayke b. kill-pl.obj: ronnu > >Comanche >-------- >(4) a. fly-sg: yItsI b. fly-pl: yoti > >Pima >---- >(5) a. arrive-sg: jivia b.: arrive-pl: dada >c. kill-sg: mua d.: kill-pl: kokda > >-Carson --------------------------------------------------------------------- Heidi Harley Department of Linguistics Douglass 200E University of Arizona Tucson, AZ 85721 Ph: (520) 626-3554 Fax: (520) 626-9014 hharley at u.arizona.edu From mcginnis at ucalgary.ca Wed Dec 18 16:02:37 2002 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Wed, 18 Dec 2002 09:02:37 -0700 Subject: Dan Everett: suppletion (reply to Carson Schutze) Message-ID: Carson's questions and Rolf's answer were both interesting. Let me suggest an additional example to those Carson brings up. In Wari' many verbs have suppletive forms on an ergative pattern (described in the Everett & Kern 1997 Routledge Grammar). I don't have the examples off the top of my head and am typing this at home, but the process is simple. Many intransitive verbs have suppletive forms for plural subjects while many transitive verbs have suppletive forms for plural objects. The suppletion seems to run from what Carson refers to as 'true' suppletion (such qualifiers make me wince, but I will resist the temptation to say why) and 'Readjustment'. I will close with a digression that those who know me will not find surprising: statements to the effect that 'hypothesis X is more interesting because it is stronger' are a form of epistemological sloganeering with little to commend them. Dan ......................... Dan Everett Professor of Phonetics and Phonology Department of Linguistics Arts Building University of Manchester Oxford Road M13 9PL Manchester, UK dan.everett at man.ac.uk Dept. Fax and Phone: 44-161-275-3187