What Kids Can Tell Us About Negative Sentences in Korean

What kids can tell us about negative sentences in Korean 1. Subject-Object-Verb (SOV) language. • The verb is always last • Normal order is subject first, then object, then verb. Chelswu-ka ppang-ul mek-ess-ta. Chelswu-SUBJ bread-OBJ eat-PAST-DECL ‘Chelswu ate the bread.’ Short negation: Affirmative (3) 2. Negative particle an before the verb: Consider: Short Negation (6) (4) Languages have re-ordering rules that turn underlying structure into the structures we see. A generalization about English (SVO) is that the object of a verb follows the verb: Robin will eat Pat’s soup. But yet in some sentences we hear, this seems not to be true: Verb followed by ci, negative particle, and ‘do’: Chelswu-ka ppang-ul mek-ci anh-ass-ta. Chelswu-SUBJ bread-OBJ eat-CI NEG .do-PAST-DECL ‘Chelswu did not eat the bread.’ Long Negation (7) a. b. Whose soup will Robin eat? Pat’s soup will be eaten by Robin. We can maintain this generalization if we suppose it is true of an underlying structure and that re-ordering rules move parts of the sentence around. The plot Start: Robin will eat whose soup Start: Robin will be eaten Pat’s soup 1. Move tense (will) left of the subject 1. Move subject to the end, insert by a------l a--------------l ? 1 1 ? will Robin — eat whose soup — will be eaten Pat’s soup by Robin Some assumptions about linguistic structure First assumption: Take on faith: There is a lot of evidence for breaking sentences up this way, into hierarchically organized constituents (drawn as trees). Third assumption: • Some background assumptions and our goals • Look at the child data—in particular, errors • Reach conclusions about adult Korean we might not have otherwise. 3. did stands in for ate pizza. Second assumption: Sentences have an “underlying” structure. Chelswu-ka ppang-ul an mek-ess-ta. Chelswu-SUBJ bread-ACC NEG eat-PAST-DECL ‘Chelswu did not eat the bread.’ Long negation: and Mary did too 1 z = ate pizza Suggests: [ate pizza] is a unit, to the exclusion of Bill (or Mary). There are two ways to form a negative sentence in Korean: (2) Bill ate pizza Negation in adult Korean Korean: (1) (5) Sentences have structure. S qp NP VP 1 3 1 V NP 1 1 1 Bill ate pizza S = Sentence NP = Noun Phrase 2. Move question phrase left of tense. a--------------l ? 1 Whose soup will Robin eat — 2. Move object left of tense. a----------l ? 1 Pat’s soup will be eaten — by Robin Result:Whose soup will Robin eat? Result:Pat’s soup will be eaten by Robin. VP = Verb Phrase The verb and object form a unit. 1 2 A large part of what concerns linguists looking at sentence structure is characterizing these re-ordering rules. What are they? When are they used, and when aren’t they used? For example, Is the rule that moves tense left of the subject in Whose soup will Robin eat? the same rule as the one the moves tense left of the subject in Will you please leave?? Or in Not in a million years will I forget this week.? Is it the same rule that in French makes you ask Qui a-t-il vu? although you say Il a vu Henri.? Or the rule that puts the verb in “second position” in German? Why does the rule that moves question words left of tense work in Whose soup will Robin eat? but not in Whose soup will Sam cry after Robin eats? or in Whose soup will Robin meet the man who bought?? And so forth… The local hope is that we might find a simple way to understand the structure and rules involved in Korean negative sentences. As it stands now (what might be our best first guess), we don’t have anything much better than two rules to form negation, from which you “choose”: (Short negation) Form S-O-V-ci Neg-do (Long negation) • Those re-ordering rules—kids have to learn them. (after all, they seem to operate differently in different languages— for example, in Korean, you don’t move question words to the front) • What happens before kids learn them? A guess… (9) ‡ na an pap mek-e I NEG rice eat-DECL (‘I do not eat rice’) (C&H:34 [2;2-6]) What could the rule be (for adults) that, when not applied, would yield (9)? How about a rule that moves the object (e.g., pap ‘rice’) to the left of negation (an)? The global hope is that a relatively small number of (relatively simple) rules can be formulated to account for the sentences which are “part of” a language (and predict which sentences would not be part of the language). Form S-O-Neg-V. Where we’re going with all of this: Hypothesis: In adult Korean negation, there is a rule which moves the object noun phrase to the left of negation. a------l ? 1 (10) Chelswu-ka ppang-ul an ppang-ul mek-ess-ta. Chelswu-SUBJ bread-ACC NEG eat-PAST-DECL ‘Chelswu did not eat the bread.’ From [S-O-V]+Neg, 4. 2. Children learning Short negation What counts as a noun? Early on (2–3 years old), kids learning Korean sometimes put an before objects (according to Kim, Son, and Sung 1999, it can be anywhere from 20% to 50% of the time) The hypothesized rule moves a noun phrase to the left of negation. What counts as a “noun phrase”? Of course, ppang ‘bread’ does, as above. But there are other things as well… (8) In Korean (like in Latin, Russian, German, …) nouns can be case-marked: a. b. ‡ na an pap mek-e I NEG rice eat-DECL (‘I do not eat rice’) ‡ an Gemco ka Gemco go (‘(I) do not go to Gemco.’) (C&H:34 [2;2-6]) (11) Chelswu-ka ppang-ul mek-ess-ta. Chelswu-SUBJ bread-OBJ eat-PAST-DECL ‘Chelswu ate the bread.’ So: A noun phrase is something that can be case-marked. (C&H:35 [2;2-6]) NEG Children never hear this pattern—no adult would say these. In English, verbs can be “nominalized” (turned into nouns) by adding a suffix: Idea: This mistake kids are making can tell us something about how adults form negative sentences. (12) 3 a. b. I walk. I like walking. (cf. I like grapes.) 4 (13) a. b. c. d. I completed the task. I liked [completing the task]. [Completion of the task] is its own reward. (cf. Money is its own reward) [The [completion of the task]] filled me with excitement. Hypothesis: In adult Korean negation, there is a rule which moves the object noun phrase to the left of negation. a----------l ? 1 (18) Chelswu-ka ppang-ul mek-ci an ppang-ul mek-ci -ass-ta. Chelswu-SUBJ bread-OBJ eat-CI NEG -PAST-DECL ‘Chelswu did not eat the bread.’ In Korean, verb phrases can be nominalized too. For example, adding -ki can do this: (14) a. cul nem-ki rope jump-KI ‘rope-jumping’ b. pomul chac-ki treasure hunt-KI ‘treasure-hunting’ If this is true, we are happy linguists. We have reduced the “choice between the Long Negation rule and the Short Negation rule” from before to a single rule. The difference between Long and Short Negation then really boils down to whether or not you choose to nominalize the verb phrase—and we know that the nominalization rule exists independently (for example, it is used creating (15) above). Why do we believe these are nouns now? When you nominalize a verb phrase, it can be case-marked like a noun: (15) But we’re not quite there yet. The movement shown in (18) doesn’t explain why Long Negation has the verb ha ‘do’ after the negative marker. Chelswu-ka ppang-ul mek-ki-lul hay-ss-ta. Chelswu-SUBJ bread-OBJ eat-KI -OBJ do-PAST-DECL ‘It’s eat the bread that Chelswu did.’ 5. 4. In Long Negation, the “main verb” is ha ‘do’ (anh < ani ha). Back to “Long Negation” (19) Recall that there is another form of negation, “Long Negation”: verb-ci NEG do(In fact the structure of (15) even looks a lot like long negation (16)). (16) Chelswu-ka ppang-ul mek-ci anh-ass-ta. Chelswu-SUBJ bread-OBJ eat-CI NEG .do-PAST-DECL ‘Chelswu did not eat the bread.’ Whence ha ‘do’? Chelswu-ka ppang-ul mek-ci anh-ass-ta. Chelswu-SUBJ bread-OBJ eat-CI NEG .do-PAST-DECL ‘Chelswu did not eat the bread.’ Long Negation Long Negation But, the rule in (18) doesn’t mention ha ‘do’ anywhere. So where does ha come from? The suggestion: It comes from the same place that do comes from in English. Look: You can put a case marker after ci: (17) (That is, it will come from an independently operating process, like nominalization—then it will no longer be the “responsibility” of the movement rule and we can keep the simple version that covers both types of negation). Chelswu-ka ppang-ul mek-ci-lul anh-ass-ta. Chelswu-SUBJ bread-OBJ eat-CI-OBJ NEG .do-PAST-DECL ‘Chelswu did not eat the bread.’ So, let’s see where do comes from in English— Conclusion: The verb phrase in Long Negation is nominalized. Now: If the verb phrase in these cases is a noun, and the nominalized verb phrase appears left of the negative morpheme an, and the way we construct Short Negation is to move an object noun left of an. maybe Long Negation is using the same rule as Short Negation—only with a bigger “noun” (the whole nominalized verb phrase) 5 6 In English, there is a position for words like will (representing future tense) between the subject and the verb (20a). Nothing appears there in the past tense (20b). (20) a. b. John will leave. John left. To maintain a uniform structure (i.e. a generalization), let’s suppose that (20a) tells us that: • tense is always between the subject and the verb. Hypothesis: In adult Korean negation, there is a rule which moves the object noun phrase to the left of negation. a----------l ? 1 (23) Chelswu-ka ppang-ul mek-ci an ppang-ul mek-ci -ass-ta. Chelswu-SUBJ bread-OBJ eat-CI NEG -PAST-DECL ‘Chelswu did not eat the bread.’ That is, (20b) John left is underlyingly John PAST leave but is pronounced as John left. There’s a (pronunciation) rule that says PAST + leave ⇒ left. So we did it! We managed to account for both forms of negation in Korean with a single rule (and the help of independently needed processes, like nominalization and doinsertion). This isn’t so crazy, really. Most of us know about a similar phenomenon in French. (21) a. b. c. d. e. la maison à la maison le restaurant * à le restaurant au restaurant ‘the house[feminine]’ ‘at the house[feminine]’ ‘the restaurant [masculine]’ (‘at the restaurant [masculine]’) ‘at the restaurant [masculine]’ 6. To recap: • Kids sometimes put negative an before the object instead of before the verb. • Hypothesis: Adult Korean has a rule moving a noun phrase over an ‘NEG’. The child errors are a result of not using that rule. • Looked at this way, the other form of negation (“long negation”) can be seen as a result of the same rule that forms short negation. In French, we’d say (and are in fact taught) that à + le ⇒ au (and de + le ⇒ du, …) That’s the same kind of pronunciation rule as PAST + leave ⇒ left. Now, let’s look at the negative sentences in English: (22) a. b. Summary and regrouping John will not leave. John did not leave. Major point: Looking at the errors kids make has given us evidence for how sentences in (adult) Korean are structured. (22a) tells us that not comes after the tense “slot” (will). In (22b), the combination of PAST and leave doesn’t happen with not in the way. • We don’t get left, it stays as leave. • We get did where PAST was. 7. Looking deeper into children’s errors—a connection to “optional infinitives”? First: • How is it that children only sometimes make errors? Wouldn’t you think once they learn the rule, they’ve learned the rule? (That is, shouldn’t they always make the error, and then one day learn the rule and from that day forward never make the error?) Hypothesis: If the tense “slot” is not next to a verb (i.e. if not is in the way), do is recruited and the tense morphology is realized on do. Suppose the same thing is happening in Korean. Then, any time tense is separated from the verb, we would expect do (or, in Korean, ha) to appear next to the tense marker. Looking at the Negation rule again (repeated on the next page as (23)), we see that what it does is move the verb away from the past tense marker. And so we get ha ‘do’. 7 Presumably: Once kids get it right once, they know the rule. So, what must be happening is that sometimes they don’t use the rule. Why not? Here’s a way to think about it: By virtue of the kid’s being 2 years old, his/her cognitive capacities are somewhat limited. To help them out, in a way, they know another rule: Don’t move things. 8 So the kids know two rules Move a noun phrase over the negative morpheme and Don’t move things. The problem is that the kids can’t obey both rules at once. 8. Conclusion/reiteration • For children learning Korean, there is a common error in the formation of negation. • Viewed a certain way, this error could be a symptom of a crosslinguistically well-attested kind of error kids at this (st)age make. • Looking at negative formation in Korean this way (as involving movement), leads us to a simple analysis that subsumes both forms of negation (Long and Short) under a single negation formation rule. Faced with this problem, the kid has to choose which rule to obey. • When the kid obeys the Move a noun phrase rule, we hear adult-like negation. • When the kid obeys the Don’t move things rule, we hear child-like negation. Over time, the Don’t move things rule becomes less important, and the kid ends up always obeying the Move a noun phrase rule, but until then, the kid sometimes chooses one rule to obey and sometimes chooses the other. Appendix: Over the last 10 years or so, it has been observed that in a lot of languages, children around this same age (roughly 2 years old) go through a stage in which they will sometimes use an infinitive form of the main verb of the sentence. This is the optional infinitive stage. The child negation data I refer to comes from Cho & Hong (1988) (marked as C&H with a page number reference and the age of the child in brackets). Child negation data has also been presented by Kim (1997), Wexler & Baek (1999), Kim, Son, and Sung (1999), and C. Lee (1993), Choi & Zubin (1985), Hahn (1981), to name just a few. The character of the optional infinitive stage is very similar to the situation we have been discussing re: Korean negation. Sometimes kids make errors, sometimes they don’t. Wexler (1993), Baek (1998), Wexler & Baek (1999) present a very similar analysis of negation to the one proposed here. Wexler & Baek (1999) contains some additional arguments from case marking using new child data. See also Hagstrom (1996, 1997a, 1997b, forthcoming) for additional arguments from scope interpretation, morphology, and restrictions on movement. Whitman (1995) provides an abstractly similar analysis (involving leftward movement over an, although set in the “antisymmetry” framework of Kayne 1994 which requires the additional assumption that even surface-SOV languages are underlyingly SVO). The structure of Korean negation has been a hot topic over many years. Other analyses can be found in Ahn (1991), Cho (1975), Cho (1993), Ho & Park (1995), Kang (1998), Kim (1993), Kim (1997), J.-H. Lee (1993), Oh (1971), Park (1992), Sohn (1995), Song (1971), Suh (1989), Yang (1976), Yoon (1994). And surely more. The currently popular hypothesis about how this happens goes the same way: • Kids know a rule that says main verbs must be finite. • But kids also know a rule that says Don’t move things. The idea is that a finite verb requires something to move (sometimes it’s the verb—for example, in German, nonfinite verbs are at the end of the clause like in Korean, but finite verbs move to the “second position”). So the kid is faced with the same dilemma—unable to obey both rules at once, he/she chooses one: • When the kid obeys the main verbs must be finite rule, we hear an adult-like sentence (in German, the verb moves to the second position). • When the kid obeys the Don’t move things rule, we hear a child-like sentence (in German, the verb is nonfinite and stays at the end of the clause) Korean doesn’t show any evidence of optional infinitives (whereas children learning English, French, German, and many other languages seem to have them at around this age). However, the negation error looks like it is a symptom of the same “problem”. This suggests that having the Don’t move things constraint is actually something that is part of being a child learning a first language—it seems to be present in many different languages, even if it has somewhat different effects in each. ☞ Details, notes, citations, connections We’ve learned something deep and interesting about the way language is learned by children. 9 Incidentally, with respect to (15), one might want to take the same view of these structures as we are proposing for negation—namely, these (“VP Focus”) constructions involve leftward movement of the nominalized VP. This idea was discussed in Hagstrom (1995), the connection between the VP Focus construction and the Long Negation construction was discussed in by Kang (1998). The idea that negation might trigger overt movement is not new. For a discussion of such movements in the case of negative polarity items, see Haegeman & Zanuttini (1991). Although my crosslinguistic research on the topic has been so far somewhat limited, I came across evidence in Nkemnji (1995) that seems to indicate that Nweh (Grasslands Bantu language spoken in Cameroon) also has movement to a position left of negation. In fact, what it looks like in Nweh is that the complement of NegP moves into an inner specifier of NegP, such that it gets sandwiched between one negative morpheme (presumed to be in the outer specifier of NegP) and the negative head. This kind of “tucking in” movement has been discussed and defended by Richards (1997). In this connection, note that the Korean movement, if it is structurally the same, seems to be movement to an outer specifier (presuming that an is in SpecNegP). Nweh and Korean differ in their basic word order (Nweh is SVO, Korean is SOV), and there might be a 10 correlation to be drawn there between word order and the target of movement into a second specifier, but the research has not been done. The view of do-support discussed in connection with (22) is an old one; originally addressed in Chomsky (1957), it has recently regained popularity through the work of, among others, Halle & Marantz (1993), Bobaljik (1994), and Lasnik (1994). The discussion in section 7 primarily describes work of Kenneth Wexler and colleagues over the past 10 years. For an overview of the optional infinitives phenomenon, see Wexler (1994, 1998), and Schütze & Wexler (1996). For details on the connection to the OI stage in Korean, see Hagstrom (forthcoming) in particular. The basic analysis (prohibiting “double movement” is a very minor variant on the “Unique Checking Constraint” proposed by Wexler 1998). The reason the noun phrase in a negative sentence has to move again, I assume, is for casechecking. An object marked (even abstractly) with accusative case is assumed to move to a higher position in the clause to “check” (validate, license) its case marking. This is not performed by the movement driven by negation, it still must happen whether an object is moved by a negation or not. Some evidence for this view from the scope of negation is reviewed in particular in Hagstrom (1997b). With respect to kids choosing to violate one constraint or another, the question of how kids choose which constraint to violate will be a central topic of tonight’s linguistics talk. Concerning the child error data: Children do not use long negation until they are somewhat older, at which point they basically have the negative formation under control. Why this might be isn’t quite clear. There is evidence that the children can use the ci morpheme (or at the very least, they can use a homophonous morpheme in other contexts). Cho & Hong (1988) give evidence that kids can understand long negation, but they won’t utter it; in a repetition task they report, kids will repeat an adult’s utterance except they will substitute short negation for long negation. Children will never place an before a subject of transitive or unergative verbs, but there are attested cases of children placing an before the subject of unaccusative verbs. This makes sense if (more or less following Hale & Keyser 1993) we take unaccusative subjects to be underlyingly in the object position, and unergative subjects to be in a more external position (the same position as occupied by transitive subjects). For more details on implications of the particular errors kids make to the details of the structural representation, see Hagstrom (1997a, forthcoming). 11