GRS LX 700 Language Acquisition and Linguistic Theory Week 13. Models, input, intake, attrition The Monitor Hypothesis A linguistic expression originates in the system of acquired knowledge, but prior to output a “Monitor” checks it against consciously known rules and may modify the expression before it is uttered. Learned competence (the Monitor) Acquired competence output What makes input into intake? Apperception: Recognizing the gap between what L2’er knows and what there is to know. Comprehensibility: Either the semantic meaning is determinable or the relevant structural aspects are determinable. Attention: Selecting aspects of the knowledge to be learned (from among many other possible things) for processing. Output: Forcing a structural hypothesis, elsewhere used to shape input into a form useful for intake. Input apperception Some input is apperceived, some isn’t. That which isn’t is thought of as blocked by various “filters”: Time pressure Frequency non-extremes Affective (status, motivation, attitude, …) Prior knowledge (grounding, analyzability) Salience (drawing attention) Apperception comprehension Modification of speech to learner (“foreigner talk”) Redundancy Negotiation for meaning (often, meaning is a precursor to being able to assign a syntactic representation). Note: Much of the following discussion is probably more about learning than acquisition. Cf. “L2A” of C++. Foreigner talk Like the better-known phenomenon of “baby talk”, it also turns out that people conversing with others whom they perceive to be non-native speakers (NNSs) will often use a form of “foreigner talk”—modified language forms presumably intended to simplify the utterance. Foreigner talk Slower, clearer articulation Higher frequency vocabulary, fewer idioms Providing more definitions Less elliptical More gestures Short, simple sentences Moving topics to the front of the sentence, new information to the end of the sentence More repetition, restatements. Recasting NNS’s incorrect statements Foreigner talk The ways in which this happens varies a lot—where it happens at all, there are many different ways that sentences are “simplified”. The adjustments often happen in the face of an evident lack of comprehension. Foreigner talk NNS: How have increasing food costs changed your eating habits? NS: Well, I don’t know that it’s changed them. I try to adjust. NNS: Pardon me? NS: I don’t think it’s changed my eating habits. Foreigner talk NNS: How have increasing food costs changed your eating habits? NS: Oh, rising costs… we’ve cut back on the more expensive things. Gone to cheaper foods. NNS: Pardon me? NS: We’ve gone to cheaper foods. Foreigner talk The “simplification” sometimes even sacrifices grammaticality, which is probably of dubious value both for comprehension and learning. Basil: It’s not fire; it’s only bell. NNS has an object from a grab-bag, NS is trying to guess its identity. NS: Ok, little guy! Yeah, yours! Okay! Yours is it for eat? NNS: Eat. No. Foreigner talk A: Yesterday my country change, ah, President. NS: Oh yeah? Now, is the new one a good one? A: Um? NS: Is a good President? Do you like him? No? NS: Does she speak English? C: No. NS: Nothing? C: No. NS: She doesn’t talk? Always quiet? No talk? Comprehension In general, this appears to be in service of comprehension—done in order to make linguistically less sophisticated interlocutors able to understand. Once there is understanding, we also are ready for there to be intake of the input as well. “Backchannel cues” L2’ers often foil this process by providing “backchannel cues” which indicate to the NS that communication is proceeding, comprehension has been achieved. “Smile and nod”. I’d like to buy a TV. NNS is trying to buy a TV, but accidentally called a repair shop. … Ah Sony please. We don’t work on Sonys. Or Sylvania. Sylvania? Uh huh. Oh, Sylvania, OK. That’s American made. OK. I’d like to buy a TV. All right. Portables have to be brought in. Hm hm. And there’s no way I can tell you how much it’ll cost until he looks at it. Hm hm. And it’s a $12.50 deposit. OK. And if he can fix it that applies to labor and if he can’t he keeps the $12.50 for his time and effort. Hm hm. I’d like to buy a TV. How old of a TV is it? Do you know off hand? 19 inch. How old of a TV is it? Is it a very old one or only a couple years old? Oh, so so. The only thing you can do is bring it in and let him look at it and go from there. New television please. Comprehension vs. output Comprehension can come in various ways, some of which have nothing to do with the structure. With some knowledge of the situation, and assuming speaker will make sense, be relevant, provide given and new information appropriately, be cooperative, the listener can come quite close to understanding the meaning without having any kind of syntactic analysis for it. If learning the structure of the target language is considered to be the ultimate goal, this kind of comprehension may be unhelpful. Comprehension vs. output No such crutches are available for production, however. If you’re going to say something in the target language, you’ll need to choose a syntax. Output viewed this way could be a way of creating grammatical knowledge (not just using pre-existing knowledge)—forcing an analysis where there was not one before. Output and negative evidence Of course, output will give the L2’er practice, allow for the “automation” of certain things allowing attention to shift elsewhere. Additionally, output provides an opportunity for negative evidence, correction from the outside. Negotiating for meaning Very often a NS-NNS (or NNS-NNS) conversation will involve a fair amount of negotiating for meaning—where understanding has not happened, the conversation takes a detour to repair the problem. Negotiating for meaning (S) Had to declare—declare?—her ingress. (J) English? No. English no (laugh)… ingress, her ingress. Ingless? Ingress. Yes. I N G R E S S more or less. Ingless. Yes. If for example, if you, when you work you had an ingress, you know? Uh huh an ingless? Yes.. Negotiating for meaning Uh huh OK Yes, if for example, your homna, husband works, when finish, when end the month his job, his boss pay—mm— him something. Aaaah. And your family have some ingress. Yes, ah, ok ok. More or less ok? And in this institution take care of all ingress of the company and review the accounts. Ok I got, I see. Ok. My father work there, but now he is old. Pre-empting negotiation In the category of “foreigner talk” we might also include these… Lots of comprehension/confirmation checks and clarification requests: I was born in Nagasaki. Do you know Nagasaki? And your family have some ingress…more or less ok? (When can you go to visit me?) Visit? (…research). Research, I don’t know the meaning. Questions often come with suggested responses When do you take the break? At ten-thirty? Healthy miscommunication A failure to communicate can serve to focus attention on areas where the NNS’s grammar is non-native-like, prompting negotiation for meaning and providing possible intake data. Michael Long’s “Interaction Hypothesis” is that this kind of negotiation for meaning and resulting attention is necessary for advancement toward the grammar of the target grammar—in part because it connects input, existing knowledge, selective attention, and output in productively trying to solve a current language deficiency. Still, feedback isn’t everything Ideally, a learner produces an ungrammatical sentences, gets negative feedback indicating that there is a problem (those involved in the conversation negotiate for meaning), focusing attention on the problem area, and the learner takes input bearing on this as intake, incorporating it into his/her grammar. Still, feedback isn’t everything Problem is, such evidence is not very consistent—it might be helpful when it happens, but it’s hard to be sure when it is happening. First: Not all incorrect forms get corrected (e.g., if the hearer understood). Second: Errors leading to misunderstanding might not be revealed until quite a bit later, if at all… Still, feedback isn’t everything NS: When I get to Paris, I’m going to sleep for one whole day. I’m so tired. NNS: What? NS: I’m going to sleep for one whole day. NNS: One hour a day? NS: Yes. NNS: Why? NS: Because I’m so tired. …enduring silence… Still, feedback isn’t everything Moreover, how is this useful feedback? NS: Did you fly to Singapore yesterday? NNS: Did I flied here yesterday? NS: Pardon? NNS: Did I flied here yesterday? How does this help fix the problem? Still, feedback isn’t everything Sometimes it works… NNS: There is a library. NS: A what? NNS: A place where you put books. NS: A bookshelf? NNS: He pass his house NS: Sorry? NNS: He passed, he passed, ah, his sign. Still, feedback isn’t everything Feedback (“negative evidence”) is just too inconsistent to be reliable—to really be the whole story about how people learn a second language. Interaction does seem to help, though, for whatever reason… Mackey 1999 Looked at question formation in ESL speakers. Tasks (designed to spark questions) Story completion Picture sequencing Discovering the order of a picture story Picture differences Working out a story by asking questions Identifying the differences between similar pictures Picture drawing Describing or drawing a picture. Mackey 1999: Procedure Wk 1 1 Test / Day treatment 1 2 Pretest Treat 1 1 3 Treat 2 1 4 Treat 3 1 2 3 5 5 5 Posttest 1 Posttest 1 Posttest 1 Activity Pic diff Story compl, Pic seq, Pic draw Story compl, Pic seq, Pic draw Story compl, Pic seq, Pic draw Pic diff Pic diff Pic diff Examples 3 1 each 1 each 1 each 3 3 3 Measure of development using question formation 2: SVO? 3: Fronting Wh/Do Have you drawn the cat? 5: Do/Aux-second (wh-front, inversion) What the cat doing in your picture? 4: Inversion (auxiliaries not do) Your cat is black? Why have you left home? 6: Uninverted in embedded clauses Can you tell me where the cat is? Mackey 1999: subject groups Interactors Interactor unreadies Watch interactionally modified input, answer comprehension questions afterwards. Scripteds Natural interaction, but low development measure Observers Natural interaction Premodified (scripted) input, leaving very little room for communication breakdown. Controls Mackey 1999 Scripted (premodified) group NS: and now under it draw a pear. A pear is a fruit. It is like an apple. The color is green. Draw the pear under the book. Can you draw it? NNS: Ok ok I got it. Look like apple (draws) NS: Good. Now on the right of the pear draw an umbrella. Mackey 1999 Interactor groups NS: Underneath it is a pear, it’s green NNS: What is it a bear? NS: A pear, pears are fruit, it’s a fruit, juicy like an apple NNS: Ok pear, fruit like Japanese fruit nashi very delicious. You saw this in Japan? Have you eat one? NS: Yeah I did but a nashi is round yeah? Pears are round on the bottom, narrow on top. Have you eaten one here in Australia? NNS: Yes thank you. I had a pear in my lunch (time) not…juicy? (draws) Like this? Mackey 1999 Looking at whether different groups moved up a developmental stage. Must produce at least 2 higher-level questions in 2 of the 3 post-tests to have “moved up a stage” Private ESL school in Australia, about 6 months in residence overall. 27 classified as lower-intermediate, 7 classified as beginners (the interactor unreadies), average 1.7 mo. in residence. Mackey 1999: % people moved up a stage 100 86 90 % of participants who increased 80 70 71 57 60 50 40 30 16 20 14 10 0 Interactor (5/7) Control (1/7) Interactor Unready (6/7) Scripted (1/6) Observer (4/7) Interactors significantly more likely to move up 50 % who increased 45 40 35 33 30 25 17 20 15 10 5 0 Interactors (both types ) Non-interactors (control, scripted, obs erver) Mackey 1999: Increase in stage 4 & 5 questions in posttests 8 7 6 5 Interactor Control Observers Scripted Unreadies 4 3 2 1 0 T1 T2 T3 Mackey 1999 Mackey claims that her study shows that the interactors have a significant advantage over the non-interactors, based on the previous graph (production of highstage questions). Yet—should advancing stage be the real goal? In that, the observers also benefited. Mackey 1999 Delayed benefit— Interestingly in the latest posttest (2 weeks after the treatment), the numbers of highstage questions had continued to grow. Suggests perhaps that this had focused attention on areas that needed work, but the grammatical changes were not implemented immediately. Input, interaction… UG? UG hasn’t played a very big role in the discussion of the importance of interaction, converting input into intake, negotiating for meaning. How can we connect them? Parameters, triggers Recall that one of the crucial features of parameters is that (ideally) each parameter setting has a cluster of effects. It’s not just that the verb appears before adverbs—it is that the verb moves into the tense position, which means it appears before adverbs and before negation. Coming before adverbs and coming before negation are a cluster of properties tied to the single verbraising parameter. Parameters, triggers In order to set a parameter in the way which matches the setting reflected by the language in the environment, the learner needs to look for consequences of a particular setting. Designated bits of data which can serve as unambiguous indicators of one parameter setting over another are sometimes called triggers. Parameters, triggers So, for example, the L1’er’s task is to examine the input for instances of these triggers and use them to set the parameter to the correct value. Some of the consequences of any given parameter setting might be fairly obscure, not likely to show up in frequent (or easily analyzed) ambient speech data accessible to the kid. This might make it hard to set one’s parameters—but for the clustering property. Parameters, triggers Indications that the verb moves: Indications that the verb doesn’t move: Do-support (The verb does not usually move) Indications that null subjects are allowed: Seeing verbs before negation Seeing verbs before adverbs Null subjects are observed. Postverbal subjects are allowed. Indications that null subjects are not allowed: Expletive subjects are observed (it’s raining). Parameters, triggers If triggers are what setting parameters is all about, then the interaction stuff is probably about making the triggers more salient. Unfortunately, it is difficult to interpret existing “input enhancement” type studies in these terms because they measured different things—we don’t know what triggers were present, what effect making triggers (vs. nontriggers?) salient had. Parameters, triggers If language acquisition (first or second) were just about finding triggers to set the parameters, why is it so hard then? Why is negotiation, etc. important (to L2A anyway)? This suggests that the triggers are in the “incomprehensible” input, that needs to be elaborated on in order to be used as intake (and thus to set the parameter). Ungrammatical FT Incidentally, the parameters approach makes “ungrammatical foreigner talk” even more problematic. Consider: In foreigner talk… The pronoun it is pervasively omitted. Auxiliary do is regularly omitted. Subjects are left out. What if those were triggers? Modularity (Schwartz 1999) Fodor (1983) proposed that the mind comes in “modules”: Domain specificity Information encapsulation Mandatory operation Speed Limited accessibility to consciousness Shallow outputs Regularity of development Fixed neural architecture Characteristic patterns of breakdown Modularity Vision is always good to compare to language; it has a similar level of complexity, and it has many properties that linguists often attribute to language. There could be a language module, since there is fairly uncontroversially a vision module. Vision Optical illusions— you can’t help but see them. Vision Optical illusions— you can’t help but see them. (movies by Yaer Weiss & Edward Adelson) Modularity The points about these visual illusions are: The processing involved in vision is quite complex. It is also completely unconscious. Learning that the lines are the same length or that the checkers are the same color doesn’t help. Language knowledge of the sort we’re interested in may well have these same properties; Krashen’s learned/acquired distinction might be right. A model of L2A? L in g ui sti c inp ut Appe rce ived inpu t LAD P rior lingu ist ic know ledge Co m p rehended inpu t In take Perhaps comically complex, but to some extent justified. L in g ui sti c inp ut F re quency , p ri or know ledge , affe ct, att en ti on, pro c es si ng St or a ge Hypo the si s te sti ng a ga ins t cu rre nt g ramm ar Ignorag e G ramm ar m od ific ation G ramm ar str eng then ing UG In tegra tion D isc ou rse pl an ning C ult ural k n o w led ge L ea rned Int er lan gu ag e g ra m m a r Acqu ir ed L1 U tt erance p lann ing Mod e (or al /w ritt en ), sit ua ti on , pe rs ona lity , m on it or ing , p roce ssing L in g ui sti c o u tput Integrated model? This is the part considered to represent L1A. PLD is processed by the LAD, filtered by UG, and implemented as acquired (UGcompliant, parametric) knowledge. PLD LAD UG L1 acquired Integrated model? PLD LAD If something suffers from passing the critical period, it would probably be the LAD, cutting off this avenue of acquisition. UG L1 acquired PLD affective etc. apperceived prior linguistic knowledge ? comprehensible input Integrated model? To the extent that you can still get acquired knowledge in the IL, it has to be in an appropriate form, filtered by UG. First filtered by attention, etc., and prior knowledge. UG IL acquired L1 PLD affective etc. apperceived prior linguistic knowledge ? comprehensible input Integrated model? The L1 plays an important role in defining the acquired knowledge in the IL grammar; perhaps the starting point, perhaps the ending point too. UG IL acquired L1 PLD affective etc. Integrated model? The learning part of L2A follows a general learning pattern: hypothesis testing, affecting the stored knowledge. apperceived prior linguistic knowledge comprehended input hypothesis testing intake integration storage modification/strengthening IL learned Discourse planning Cultural knowledge Integrated model? Creating output is fairly complex. Discourse planning concerns intent Cultural knowledge concerns norms in expressing this intent. Utterance planning sets up the structure. IL learned acquired Utterance planning Mode, situation, personality, monitoring, processing linguistic output Discourse planning Cultural knowledge Integrated model? Utterance structure is driven by the acquired knowledge. Filters on the planned utterance involve the monitor and other outputrelevant factors. IL learned acquired Utterance planning Mode, situation, personality, monitoring, processing linguistic output Integrated model? Does practice help convert learned knowledge into acquired knowledge? Perhaps, if the output can count as comprehensible input… UG IL learned acquired linguistic output Prisms and vision The visual system can adjust itself (after the critical period), even if it does seem for the most part “hardwired”. Fitting someone with prisms that change the angle of incoming visual stimuli causes initial visual confusion, but they “get used to it” and soon don’t even notice anything unusual. Perhaps the phenomenon of L1 attrition is something like this… and language attrition It is a very common phenomenon that, having learned an L2 and having become quite proficient, one will still “forget” how to use it after a period of non-use. While very common, it’s not very surprising—it’s like calculus. If L2 is a skill like calculus, we’d expect this. L1 attrition Much more surprising is the fact that sometimes under the influence of a dominant L2, skill in the L1 seems to go. Consider the UG/parameter model; a kid’s LAD faced with PLD, automatically sets the parameters in his/her head to match those exhibited by the linguistic input. L1 is effortless, fast, uniformly successful… biologically driven, not learning in the normal sense of learning a skill. So how could it suffer attrition? What are you left with? UG in L2A We’ve looked at the questions concerning whether when learning a second language, one can adapt the “parameter settings” in the new knowledge to the target settings (where they differ from the L1 settings), but this is even more dramatic—it would seem to actually be altering the L1 settings. It behooves us to look carefullier at this; do attrited speakers seem to have changed parameter settings? EnglishHebrew kid in Israel He doesn’t know to who it belongs. He’s thinking about with what they can play. (cf. He doesn’t know who it belongs to.) (cf. He’s thinking about what they can play with.) Hebrew doesn’t allow preposition stranding—the constraint against prepositions stranding seems to have been “back copied” HebrewEnglish Ma at midaberet al? The reverse situation is reported to hold as well; here, a Hebrew sentence (ungrammatical to monolinguals) involving preposition stranding. ItalianEnglish Italian is a “null subject” language that allows the subject to be dropped in most cases where in English we’d use a pronoun (Possible to use a pronoun in Italian, but it conveys something pragmatic: contrastive focus or change in topic) English is a “non-null-subject” language that does not allow the subject to be dropped out, pronouns are required (even sometimes “meaningless” like it or there). Not required that a pronoun signal a change in topic. Italian, null subjects Q: Perchè Maria è uscite? ‘Why did M leave?’ A1: Lei ha deciso di fare una passeggiata. A2: Ha deciso di fare une passenggiata. ‘She decided to take a walk.’ Monolingual Italian speaker would say A2, but English-immersed native Italian speaker will optionally produce (and accept) A1. (Sorace 2000) Greek too… (null subjects) Q: Jati vjike i Maria? ‘Why did Maria go out?’ A1: Afti apofasise na pai mia volta. A2: Apofasise na pai mia volta. ‘She decided to take a walk’ Reverse errors unattested Q: Perchè Maria è uscite? ‘Why did Maria leave?’ A: *Perchè Ø è venuto a prederla. ‘Because (Gianni) came to pick her up.’ That is, they don’t forget how to use null subjects so much as they broaden the contexts in which they can use overt pronouns. Postverbal subjects Q: Chi ha starnutito? ‘Who sneezed?’ A1: Gianni ha starnutito. A2: Ha starnutito Gianni. Native speakers would say A2 because of the narrow focus; attrited speakers will produce/allow A1 as well. L1 attrition It seems that the acceptability of overt pronouns (in the L1 “attriters”) broadens compared to their L1, the acceptability of null pronouns becomes more restricted. Pronouns in a null subject language are marked—they are restricted to particular discourse contexts ([+topic shift], according to Sorace). What seems to happen is that the pronouns revert to the unmarked case ([±topic shift] like in English). L1 attrition Same goes for postverbal subjects—it is a marked option for languages, and the L1 seems to be retreating to the unmarked. Like with pronouns, it seems to be not a question of grammaticality but a question of felicity. L1 attrition Certain areas of the L1 grammar are more susceptible to this kind of attrition then others. Sorace notes that the observed cases of attrition of this sort seem to be the ones involved with discourse and pragmatics, not with fundamental grammatical settings. (The attrited Italian is still a null-subject language, for example—null subjects are still possible and used only in places where null subjects should be allowed). L1 attrition So, we’re left with a not-entirelyinconsistent view of the world. Parameter settings in L1 appear to be safe, but the discourse-pragmatic constraints seem to be somehow susceptible to high exposure to conflicting constraints in other languages.