GRS LX 700 Language Acquisition and Linguistic Theory Week 9. Second Language Acquisition: introduction Scientific study of language What constitutes one’s knowledge of language? How is that knowledge acquired? Looking at adult native languages, we’ve found that language is very complex (see LX 522, 523, for example) Looking at kids, we’ve found that kids seem to learn this complicated system with surprisingly little help from the environment. L1 acquisition We posited a genetic predisposition for language, something which guides the kinds of languages kids learn (Universal Grammar): Kids learn fast Kids end up with systems that are more complicated than the input data justifies (they can judge ungrammatical sentences in the same way as other native speakers). Kids don’t fail to learn language despite differences in environment, and without getting or making use of negative evidence. Kids seem to go through similar stages, across kids, across languages. But what about L2 acquisition? Adults seem to have a harder time learning language than kids do learning their first language (there may be a “critical period”). Adult second language learners rarely reach a native-speaker-like level of competence. Adult second language learners already know a language. Adult second language learners are often given negative evidence (“you don’t say it that way”) when taught in a classroom. L2A seems very different from L1A. Is L2A like learning to play chess? Like learning calculus? Do we just learn the rules of the language and apply them (sometimes forgetting some of the rules, never quite learning all of them, etc.)? It’s very tempting to think that’s true. Scientific study of language What constitutes one’s knowledge of language? How is that knowledge acquired? We can still study these questions in L2A as well and try to determine the answers, whether they are related to the answers we got for L1A or not. And perhaps surprisingly, they might be. L2 competence Learners of a second language have some kind of (systematic) linguistic knowledge. They have retained their L1 knowledge, and they have knowledge of a sort which approximates (perhaps poorly) the knowledge held by a native speaker of the learner’s L2. This knowledge is often referred to as an interlanguage grammar—not L1, not L2, but something different (…and to what extent this knowledge might be related to or influenced by L1 or L2 is yet to be determined). A real-world example, Japanese case-marker omission Adult knowledge is complicated, relies on the Empty Category Principle, which says that an empty category (including a dropped Case marker) must be properly governed. The ECP is taken to be responsible for the face that in Japanese you can drop a Case marker in object position but you cannot drop a Case marker in subject position. Kanno 1996 John ga sono hon o yonda. nom that book acc read ‘John read that book.’ John ga sono hon _ yonda. nom that book Ø read ‘John read that book.’ * John _ sono hon o yonda. Ø that book acc read ‘John read that book.’ Kanno 1996 English speakers (learning Japanese) know the ECP, because they know: Who did you say Ø t left? *Who did you say that t left? But this is a very different context of use from the use in Case marker drop. The question is: Do English speakers respect the ECP in their interlanguage grammar (toward Japanese)? A broader way to ask the question: Is the interlanguage grammar constrained by UG? …but a flawed premise It’s really hard to test this kind of thing. Do English speakers actually know the ECP? Or do they just know that *Who did you say that left? ? The usual theoretical claim: ECP holds of Language, including English; *Who did you say that left? is a consequence of the more general constraint. If English speakers know the ECP, they know the ECP. Thus, if they obey the ECP in L2 Japanese, this is probably transfer of knowledge from the L1, not some kind of direct intervention by UG. Or at least we can’t tell the difference, contrary to the premise of Kanno’s experiment. …but a flawed premise Kanno’s conclusions tacitly rely on this assumption: ECP is a property of the LAD, you can only learn an L1 that obeys the ECP. The L1 you learn if you learn English, however, is simply a set of context-specific rules that applies to *Who did you say that left? but has nothing to say about Case marker drop, since there are no Case markers in English. …but the hypothesis of modern theoretical syntax is that ECP is a property of the language knowledge, playing a role in generating/judging/comprehending English. Kanno 1996 Let’s look at the experiment anyway, though… Kanno tested 26 college students in Japanese II on case particle drop. In an effort to ensure that the students, if successful, will have “gone beyond the input”, she examined what the students would have been exposed to by the textbook up to the point where they took the test, to see if they were taught when not to drop the case markers. What the Japanese II students saw… 41 cases of object case-marker drop, like: Enpitsu Ø kudasai ? pencil give ‘Can you give me a pencil?’ 8 cases of subject case-marker drop, in the exceptional case when it is allowed (with a final emphatic particle—these don’t violate the ECP): John Ø sono hon o yonda yo. John that book acc read part ‘John (indeed) read the book.’ (I think) What the Japanese II students saw… Certain verbs have nominative case on their objects, and case can be dropped on those objects too… John ga kankokugo (ga) dekimasu. John nom Korean nom can-do ‘John can speak Korean.’ 69 of 110 such verbs in the book had the object case marker dropped. What the Japanese II students saw… Japanese allows arguments to be omitted (somewhat like Italian pro drop), so there were many cases with just one argument (the object) with no case marker: Kami Ø irimasu ka? paper need Q ‘Do you need paper? / Is paper necessary?’ What the Japanese II students saw… Worst of all, the topic marker can be dropped, which looks a lot like a subject marker being dropped. Tanaka-san (wa) itsu kaimasita ka? top when bought Q ‘When did Tanaka buy it?’ ‘As for Tanaka, when did he buy it?’ What the Japanese II students saw… “ga [nom] might be deleted, but with a reduction of the emphasis and focus conveyed by its inclusion.” (No hint that sometimes—even usually—it is not allowed) “If o [acc] is deleted, [the object] would simply lose a bit of its emphasis and focus. On the other hand, the addition of o would give added emphasis and focus.” The poor Japanese II students… There’s pretty much no way they could have reached the right generalization based on what they were provided. Nom can be dropped from object position Top can be dropped from subject position Nom subject can be dropped with a particle Explicit instruction was only about emphasis. But did they anyway? The experiment To test this, the sentences used wh-words. Wh-words in general do not allow topic marking, so if the particle is dropped from a subject wh-word, it could not have been a topic drop. subject wa wh-phrase Ø verb Q? *subject Ø wh-phrase acc verb Q? pro wh-phrase Ø verb Q? *wh-phrase Ø pro verb Q? Kanno’s missing controls Here’s why wh-subjects were tested: Subject marker ga cannot be dropped in L1J. Topic marker wa can be dropped in L1J. Wh-phrases do not allow wa in L1J. Hence, a wh-subject with no marker, for L1J, would have an illicitly dropped subject marker. The first two points follow from the ECP and the general knowledge that things can be dropped. But do the students know that wh-phrases do not allow wa? If not, we’re back to square one. She did address this objection by citing a difference between 1-argument and 2-argument test items with respect to whether ga is droppable for test subjects who seemed to be counterexamples. Kanno’s missing controls Subjects were given controls to test their naturalness rating of dropped case markers in general. But the crucial contrast has to do with the naturalness of overt vs. dropped case markers on wh-words. Yet no naturalness measure of an overt case or topic marker on a wh-phrase was obtained. So, we’re left comparing overt case markers on non-wh-words with dropped case markers on wh-words. Kanno’s results 3 2.8 2.6 2.4 2.2 NP wa NP — NP — NP o 2 pro NP — 1.8 NP — pro 1.6 1.4 1.2 1 Students NSs “UG” in L2A Kanno’s conclusion L2 learners of Japanese have nevertheless (statistically significantly, as a group) gotten the rule about dropping subject case markers, despite the lack of evidence from the textbook, the instructor, or even English. That is, they appear to know the ECP. This shows that L2 learners are able to bring their knowledge of the ECP from L1 to bear on L2A. (But it doesn’t reach Kanno’s conclusion that UG as opposed to L1 is constraining L2A.) Statistically significantly, as a group… The other thing that is surprisingly often overlooked is that the hypothesis is not about groups, it is about learners. Yet in many studies, results are reported solely in terms of the group. And all this really tells us is that probably some in the group conform to the hypothesis. What about the performance of individual subjects? Statistically significantly, as a group… To look at the performance by subject, Kanno classified each subject in terms of their preferences (whether that particular subject generally rated ga drop or o drop more favorably). And, while not 100-0, the subjects overwhelmingly preferred o drop. So the conclusion (that the L2’ers are obeying the ECP) appears to be safe. All I really needed to know I learned in UG “The linkage of concept and sound can be acquired on minimal evidence, so variation [among languages] here is not surprising. However, the possible sounds are narrowly constrained, and the concepts may be virtually fixed. It is hard to imagine otherwise, given the rate of lexical acquisition, which is about a word an hour from ages two to eight, with lexical items typically acquired on a single exposure, in highly ambiguous circumstances, but understood in delicate and extraordinary complexity that goes vastly beyond what is recorded in the most comprehensive dictionary, which, like the most comprehensive traditional grammar, merely gives hints that suffice for people who basically know the answers, largely innately.” Chomsky (2000, New Horizons in the Study of Language and Mind), p. 120. Influence of UG in some form is probably inevitable… Like in L1A, the input is almost certainly degenerate, and the negative evidence there might be isn’t enough to make the subtle complexities of language learnable, and for negative evidence (in the form of correction) to be of any use, L2 learners have to make errors, yet for these subtle complexities, the learners don’t seem to make the crucial errors that would be required to learn them. Kanno’s experiment (among others) shows that L2 learners seem to “go beyond the evidence.” How is UG “used” in L2A? What is UG really? Probably the simplest view of it is that UG constrains the kinds of languages we can learn. For the moment, assume we’re talking about L1A. UG says: You can’t learn a language that lacks the ECP. You can’t learn a language that doesn’t respect constraints on movement out of an How is UG “used” in L2A? UG shaped your L1, we take that to be essentially beyond dispute in some form… but when you learn L2, you still know L1. So, perhaps: UG constrains how you learn L2 (directly, like it constrained your L1) Or, perhaps: Your L1 constrains how you learn L2 (indirectly, UG constrains L1, L1 constrains L2) Or, perhaps: Nothing language-related constrains how you learn L2—it’s like learning chess. How is UG “used” in L2A? A lot of the early (198*) studies tried to classify their hypotheses about the involvement of UG in L2A in terms of “access”. Full Access—UG constrains L2A. (Partial Access—UG constrains L2A partly.) No/Indirect Access—UG is not involved in L2A (except insofar as it constrains L1) An independent question— what role does L1 play in L2A? Full Transfer—the properties (parameters) of L1 are taken as the “starting point” in L2A. Partial Transfer—some of the parameters of L1 are taken as the “starting point” in L2A, while some others start in an independent setting. No Transfer—the parameter settings of L1 do not affect L2A. Access hypotheses The model these early hypotheses work with is essentially that UG provides a blueprint or a template for languages, which is used to create a concrete instantiation of a language. •Principles L1A •Parm 1: — (A, B) •Parm 2: — (A, B, C) •… UG •Active Principles •Parm 1: A •Parm 2: B •… L1 Access hypotheses Once L1 has been instantiated, the template might become unavailable. In this case, the only available information about what languages are like is what’s instantiated in L1. •Principles •Parm 1: — (A, B) •Parm 2: — (A, B, C) •… UG •Active Principles •Parm 1: A •Parm 2: B •… L1 Access hypotheses The No/Indirect access hypothesis supposes that the principles and parameters of L1 (but not the information in UG) are available in forming an instantiation of L2. •Principles •Parm 1: — (A, B) •Parm 2: — (A, B, C) •… UG •Active Principles •Parm 1: A •Parm 2: B •… L1 Access hypotheses The full access hypothesis supposes that the template is still available to instantiate L2 the same way L1 was instantiated. •Active Principles •Parm 1: B •Parm 2: A •… L2A •Principles L1A •Parm 1: — (A, B) •Parm 2: — (A, B, C) •… UG •Active Principles •Parm 1: A •Parm 2: B •… L2 L1 Access hypotheses A partial access hypothesis supposes that certain parts of the template are no longer available (fixed in the L1 settings) but other parts can still be used to instantiate L2. •Active Principles •Parm 1: A •Parm 2: C •… L2A •Principles L1A •Parm 1: — (A, B) •Parm 2: — (A, B, C) •… UG •Active Principles •Parm 1: A •Parm 2: B •… L2 L1 Distinguishing between access hypotheses The no access hypothesis takes L2A to be a general learning process, not constrained by properties of UG. As such, we do not expect the IL of second language learners to conform to the specifications of UG. We expect that the IL would be free to exhibit properties unlike any natural language (L1). So we look for “wildness” in the IL grammar of second language learners—for indications of grammar which would not qualify as an L1. Distinguishing between access hypotheses The full access hypothesis, on the other hand, predicts that IL grammars of second language learners, even while not actually the grammar of the target language, will still conform to the restrictions UG places on natural languages. It will operate under the same principles, and it will have parameters which are set to a setting which is possible in natural language. Distinguishing between access hypotheses The partial access hypothesis is the least well-defined. It places itself somewhere between full access and no access. We might see that a second language learner’s IL shows evidence of parameter settings different from the L1 (or not, depending on which parts of UG we are hypothesizing L2A access to). We might see evidence of principles not used in L1 but provided for in UG. The partial access hypothesis is basically the fallback position, the compromise we need to make if the facts don’t fit into one of the other hypotheses. Access/Transfer The basic hypotheses out there to explore and evaluate (not including retreats to partial transfer and/or access): Full transfer/No access: L2 knowledge is fundamentally different from L1 knowledge, based on L1 knowledge plus conversion rules. Full transfer/Full access: L2A is as flexible as L1A, with L1 as the starting point. L1 and L2 “distance” should affect ease/course of acquisition. No transfer/Full access: L2A is as flexible as L1A, and the learner’s L1 should not have an effect. A note about UG In this context, UG is probably best thought of as defining a “shape” that language knowledge can take. Parameters define ways in which stored knowledge can conform to the “shape” of UG. The LAD is a system which analyzes the PLD and sets the parameters. LAD UG PLD Binding Theory Subjacency Principles and Parameters So two languages which differ with respect to one parameter setting might be represented kind of like this. Language A Language B Modeling human language capacity Many of the discussions about “UG access” seem to confound these two aspects of the human language capacity. But it could be that the LAD becomes impaired/inactive after L1A while the “shape of language knowledge” is still available for L2A. (It could also be that the “shape” of UG is completely determined by LAD—that’s the interpretation that I called “confounded.” It’s not internally inconsistent, but it isn’t a necessary—or even usual—interpretation of UG) LAD UG PLD Binding Theory Subjacency In favor of no access… The well-known “critical period” effects seem to point toward a view like no access; adult L2A is much less uniform, typically not fully successful, and appears to involve much more conscious effort. Proponents argue that their observations about differences in the course and end result of L2A (vs. L1A) indicate that principles of UG are not being obeyed (for example, learners positing rules that appeal to linear order, rather than structure, contra Structure Dependency). Keep in mind that these arguments are really arguments about LAD and not about UG as the “shape of language knowledge” though. For example… Meisel (1997) looked at L1A and L2A of negation in German, French, and Basque. In L1A in the three languages, negation appears to go through similar stages. First, it is placed externally (generally initially, sometimes finally), unlike in the adult language No(t) I go home I go home no(t). Then, it appears sentence-internally, in an appropriate position with respect to the tensed verb for the target language (differs by language). L1A: Once children show evidence of knowing how to use finite verbs, they seem to have no particular trouble with the syntax of negation in the target language. Meisel (1997) For L2A, the consensus opinion from previous studies seems to be that second language learners, regardless of target and first languages seem to go through pretty much invariant stages with respect to negation. First, preverbal or initial negation. Then, more target-like internal negation. Sounds like the L1A sequences; this made people eager to try to apply the same explanations. However, almost all of these studies used English as the target language, and in fact some studies seemed to have “missed” the first stage. Bottom line: that “consensus opinion” is pretty suspect. Meisel (1997) Closer investigation reveals that not all second language learners go through an “initial negation” stage, even if the L1 has preverbal negation. And, unlike in L1A, where there is an initial negation stage, it does not seem to disappear at the same time as the control of finite verbs. Whereas “initial negation” in L1A is usually sentence-initial (before the subject), “initial negation” in L2A is often preverbal (but after the subject). That is to say, people were not careful with what they were willing to call “initial negation” when they were jumping on the L2Asequence-is-like-L1A-sequence bandwagon. Meisel suggests that initial negation is actually a characteristic of a certain kind of learner, a reflection of a strategy that (some) people use in L2A. Meisel (1997) Rather than observing structure-dependent negation placement based on [±finite], the results tend to suggest strategies based on linear order (i.e. put negation after the verb). Meisel concludes that any UG involvement in L2A is much less clear given these differences between L1A and L2A. Concerning this argument Notice that this is primarily an argument about sequence of acquisition. Roughly, the idea is: Because the sequence of L1A and L2A do not match, and assuming L1A is driven by UG, L2A can’t be also driven by UG. In short, this seems to be an argument about whether the (L1) LAD is involved in L2A. It doesn’t really fully reach the question of whether UG constrains L2A. To show the UG does not constrain L2A, we should be looking for IL grammars that are UG-illicit (regardless of how they are arrived at). This question will recur through much upcoming discussion… To argue for full access… Threads of argumentation: Second language learners obey certain universal principles which (appear to) work differently in the TL than in the learners’ L1. (E.g., Kanno 1996 discussed earlier…) Second language learners’ IL knowledge show evidence of a parameter setting different from their L1, indicating that the parametric options are still available In favor of full access… A simple example of thread two, from Flynn (1996), is L2A between Japanese and English. Japanese and English differ in their setting of the “head parameter”, which indicates whether the object comes before the verb (Japanese, SOV, head-final) or after the verb (English, SVO, headinitial). L2 J-->E learners appear to very quickly set this IL parameter correctly, suggesting that they know that both head-initial and head-final are possible settings for this parameter, although their L1 parameter is committed to head-final. In favor of full access… Flynn on Subjacency (thread one): In Japanese, wh-words are not “moved” to the beginning of a wh-question; Japanese is a “whin-situ” language. Its wh-words appear in the same position that the trace “appears” in English. Assumption: Subjacency is concerned only with displacement of wh-words. It is a principle which says that a wh-word cannot be displaced out of certain kinds of islands (conjunctions, embedded questions, complex noun phrases, …). In favor of full access… Thus, Subjacency does not seem to rule out any wh-questions in Japanese. It is possible to ask questions like: ‘You met the man that gave what to Mary?’ Cf. *Whati did you meet the man that gave ti to Mary? Flynn takes this to mean that Subjacency is essentially “inactive” in Japanese. It does not play a role in wh-question formation in Japanese. In favor of full access… Supposing that Subjacency is not active in Japanese, Flynn considers L2A of English by Japanese speakers. Would these second language learners nevertheless obey Subjacency in English? Do they still have access to this principle provided by UG even though it is not used in their L1? Flynn’s experiments seem to indicate that Japanese speakers learning L2 English do obey Subjacency, and concludes that they must therefore still have access to UG during L2A. Getting at the “IL grammar” What do the L2 learners know? *Productions: We don’t have a great deal of success learning about the structure of linguistic knowledge in the native speaker domain by looking just at productions. Things aren’t different for L2 learners. No information on what is ungrammatical—at best, information on what is dispreferred/avoided. Performance errors happen, but that doesn’t indicate a lack of competence. Grammaticality judgments One way of testing people’s (whole) competence is to ask them to rate sentences in their second language. Who did you say that bought John dinner? 1-bad 2-a little weird 3-natural I wonder what will John wear tomorrow. 1-bad 2-a little weird 3-natural GJ tasks aren’t perfect, though… As in any experiment, you may have biases… Some people are hesitant to take an extreme position, may never rate a sentence 1 or 3. Some people may rate the sentences based on how much sense it makes, rather than on the syntactic structure. And it’s hard to correct for that, because if you ask someone what’s wrong with What did you laugh after John bought for Sue? (or how to correct it), even native speakers won’t be able to say. GJ tasks But we have the same trouble with kids too… We can try to employ the same kinds of tricks with adults… acting out a sentence identifying which picture best depicts the subject matter of the sentence judging whether a sentence is true or false of a scene. answering an ambiguous question to see wh-word scope. … Locating the source of the errors Suppose that an adult L2 learner of E. rates What did you laugh after John bought for Sue? as natural. Does that mean they don’t know Subjacency? Well, not necessarily. They may also not understand how to make complex clauses, adverbial clauses, etc. Like with kids and quantifiers Principle B, one can only really say that people know or don’t know a principle of UG once they have the appropriate structures to apply them to. “How involved is UG in L2A?” Very (UG constrains IL) vs. not (L1 constrains IL) To figure out which is right, we need to look at UG constraints or parameters which are not used in the learner’s L1. If there is something that holds in all languages, say, the q-criterion, showing that L2 learners respect the q-criterion doesn’t tell us whether that is because UG required it or because their L1 does. Two possible things to look at Parameter settings which vary between L1 and L2… English: Bounding nodes for Subjacency are DP and IP. Italian/French: Bounding nodes for Subjacency are DP and CP. Universal principles which are inapplicable in L1 but apply in L2… The ECP as used to control case marker drop in Japanese “Universal principles inapplicable in L1?” As our theories of syntax develop, finding such things becomes harder and harder, since the goal of theoretical syntax is in general to say “All languages are really the same except for some very surface-y phenomena.” wh-movement Circa 1981, English moved its wh-words, Japanese didn’t, so Subjacency wasn’t relevant for Japanese. However, since then, the proposals have changed—all languages move their wh-words to SpecCP, just some do it after SS. Evidence has appeared which shows that under the right conditions, Japanese does respect Subjacency. Thus: Looking at whether Japanese speakers learning English respect Subjacency or not still hasn’t necessarily gotten away from L1. Atrophied options? The L2A literature has historically taken a fairly old, conservative view of UG. It tends to assume that UG provides options from which languages choose, and that something that a language doesn’t choose might become unavailable as a choice later. That is, the underlying assumption seems to be that English speakers don’t know the ECP, really. That Japanese speakers don’t know Subjacency. Modern syntacticians just don’t think of syntax as working this way anymore. This kind of view is a little bit closer to what people still mostly believe about certain aspects of phonology, though. More in a bit… Parameters The bottom line is: it’s going to be hard to make a convincing case that you’ve got a principle of UG which is not known (utilized) by an L1 speaker. Perhaps, if you are lucky, you might find something plausible now, but advances in syntactic theory will do everything they can to undermine your position. However, languages do differ in the values of the parameters (e.g., Subjacency bounding nodes). Thread two is the way to go. Parameters We can also look at aspects of parameter setting in L2A. Part transfer (what settings get adopted as part of the initial state of the the second language learner’s interlanguage grammar?), part accessibility/involvement of UG (can second language learners “reset” these parameters? If so, the lists of options provided by UG are still available—that is, UG is available/involved). Phonological parameters Describing adult native-speaker phonological grammars requires abstract structures not unlike the structures required for syntax. Just like for syntax, differences between languages can be characterized in terms of phonological parameters. And in this domain, there might actually be something like “phonological options not used in the L1” to talk about. Some basic concepts There is a fairly well-defined set of possible sounds that languages make use of. Languages differ in which of these sounds play a role in the language. For example, some languages have a sound like the English v, some don’t. The “unit of sound” is the segment. Some basic concepts There are a lot of different things that go into determining a segment. Place of articulation Voicing Manner of articulation Aspiration Tenseness/laxness of tongue … Segment distinctions /p/ vs. /t/ vs. /k/ : place of articulation /p/ vs. /b/, /t/ vs. /d/: ±voice /t/ vs. /s/: ±continuant /s/ vs. /sh/: ±distributive /e/ vs. /i/: ±high /e/ vs. /a/: ±back /a/ vs. /o/: ±round … Some basic concepts Languages differ in what they “pay attention to” when differentiating segments from one another. English does not distinguish aspirated and non-aspirated consonants. The p in pit is aspirated (ph), the p in spit. They “sound the same” to speakers of English. Other languages distinguish p from ph—so pit and phit could be different words, with different meanings. The distinguishable segments in a language are the phonemes of the language. One parameter of variation between languages is their phonemic inventory. R vs. L An oft-used example of this is the distinction between r vs. l in English and the lack of said distinction in languages like Chinese, Japanese, and Korean. In Korean, for example, both segments are used, but it is phonologically conditioned— between vowels, it is r and elsewhere it is l. You don’t get to choose which one you use in a given context. So there’s no distinction. (Sunhi-lul = […-rul]) L1A and contrasts Little kids start out being able to distinguish contrasts between all possible segments, but quickly zoom in on the contrasts in their environment, losing the contrasts… Looks like a critical period, difficult for UG in L2A… From Werker (1994): Hindi-English ba ~ da H adults Fine E infants (6-8mo) Fine E kids (up to 4) Fine E adults Fine d h a ~ th a Fine Fine Bad Poor Ta ~ ta Fine Fine Bad Bad L2A and UG We can ask many of the same questions we asked about syntax, but of phonology. Learners have an interlanguage grammar of phonology as well. Is this grammar primarily a product of transfer? Can parameters be set for the target language values? Do interlanguage phonologies act like real languages (constrained by UG)? Here, it it rather obvious just from our anecdotal experience with the world that transfer plays a big role and parameters are hard to set (to a value different from the L1’s value). Phonological interference If L1’ers lose the ability to hear a contrast not in the L1, there is a strong possibility that the L1 phonology filters the L2 input. L2’ers may not be getting the same data as L1’ers. Even if the LAD were still working, it would be getting different data. If you don’t perceive the contrast, you won’t acquire the contrast. Phonological features Phonologists over the years have come up with a system of (universal) features that differentiate between sounds. /p/ vs. /b/ differ in [+voice]. /p/ vs. /f/ differ in [+continuant]. … What L1’ers seem to be doing is determining which features contrast in the language. If the language doesn’t distinguish voiced from voiceless consonants, L1’ers come to ignore [±voice]. Phonological features, filtering Brown (2000): Presented pairs of nonwords to speakers of Japanese, Korean, Mandarin. Japanese and Korean speakers didn’t perceive the l ~ r contrast, Mandarin speakers did, although none of the languages has an l ~ r contrast. However, Mandarin does have other segments which differ in [+coronal] ([r]), so Mandarin speakers do need to distinguish [±coronal] elsewhere. Phonological features, filtering Han (1992). Japanese distinguishes geminate from nongeminate stops (consonant length; k vs. kk, e.g., black owl vs. black cat). English doesn’t (*kkat vs. kat). English speakers of Japanese (even highly proficient otherwise) either missed this contrast altogether or produced long consonants that were not native-like (too short). An interesting idea (courtesy of Carol Neidle) If you were to learn French, you would be taught conjugations of regular and irregular verbs. Regular -er verbs have a pattern that looks like this: Infinitive: donner ‘give’ 1sg je donne 1pl nous donnons 2sg tu donnes 2pl vous donnez 3sg il donne 3pl ils donnent Some French “irregulars” Infinitive: donner ‘give’ 1sg je donne 1pl 2sg tu donnes 2pl 3sg il donne 3pl nous donnons vous donnez ils donnent Another class of verbs including acheter ‘buy’ is classified as irregular, because the vowel quality changes through the paradigm. Infinitive: ceder ‘yield’ 1sg je cède 2sg tu cèdes 2pl 3sg il cède 3pl 1pl nous cédons vous cédez ils cèdent Some French “irregulars” Infinitive: donner ‘give’ 1sg je donne 1pl 2sg tu donnes 2pl 3sg il donne 3pl nous donnons vous donnez ils donnent The way it’s usually taught, you just have to memorize that in the nous and vous form you have “é” and in the others you have “è”. Infinitive: ceder ‘yield’ 1sg je cède 2sg tu cèdes 2pl 3sg il cède 3pl 1pl nous cédons vous cédez ils cèdent Some French “irregulars” However, the pattern makes perfect phonological sense in French—if you have a closed syllable (CVC), you get è, otherwise you get é. [sd] (cède) [se.de] (cédez) So why is this considered irregular? Because in English, you think of the sounds in cédez as [sed.de], due to the rules of English phonology. Infinitive: ceder ‘yield’ viewed from English 1sg je cèd(e) 1pl nous céd.dõ(ns) 2sg tu cèd(es) 2pl vous céd.de(z) 3sg il cèd(e) 3pl ils cèd(ent) Some French “irregulars” Because in English, you think of the sounds in cédez as [sed.de], due to the rules of English phonology. Since in all of these cases, English phonology would have closed syllables, there’s no generalization to be drawn— sometimes closed syllables have é and sometimes they have è. What could we do? Infinitive: ceder ‘yield’ 1sg je cède [sed] 1pl 2sg tu cèdes [sed] 2pl 3sg il cède [sed] 3pl nous cédons vous cédez ils cèdent [sed.dõ] [sed.de] [sed] Some French “irregulars” If people are really “built for language” and are able to pick up language implicitly, then if people are provided with the right linguistic data, they will more or less automatically learn the generalization. Problem is: The English filter on the French data is obscuring the pattern, and hiding the generalization. Infinitive: ceder ‘yield’ 1sg je cède [sed] 1pl 2sg tu cèdes [sed] 2pl 3sg il cède [sed] 3pl nous cédons vous cédez ils cèdent [sed.dõ] [sed.de] [sed] Some French “irregulars” Something to try: Provide people with the right data, see if they pick up the pronunciation. Perhaps: exaggerate syllabification (draw attention to it). Perhaps try to instill this aspect of the phonology first? Et voilà. Chances are good that this will make these “irregulars” as easy to learn as regulars! Does it work? I have no idea. Infinitive: ceder ‘yield’ 1sg je cède “sed” 1pl 2sg tu cèdes “sed” 2pl 3sg il cède “sed” 3pl nous cédons vous cédez ils cèdent “se—dõ” “se—de” “sed” Where we are We’re concerned with discovering to what extent linguistic theory (=theories of UG) bears on questions of L2A, with an eye toward the question: To what extent is knowledge of an L2 like knowledge of an L1? Do they conform to universal principles? (ECP, Subjacency) No? UG is not constraining L2. Yes? Consistent with UG constraining L2, but not evidence for it. Do they have a parameter setting different from the L1 (and all of the consequences following therefrom)? Yes? UG is constraining L2. No? Inconclusive for the general case.