 Hydiateg. We are happy today to have our, well, our very own Charles Chang while I am Mae'r ddigwyddor yn Settembarau. Charles is living here at Leimon. Wrth gwrs, mae'n mynd i gyd yn gwneud y peirion. Mae'r ddigwyddor yn gwneud i'r peirion. Mae'r ddigwyddor yn gwneud i'r peirion. Darbyn i gyd yn gwneud i'r peirion. Charles carried out his PhD research at UC Berkeley on first language phonetic drift during Second Language Acquisition. I publish on many and varied topics in the world. ..on menu and varied topics and who they've been connected.. ..to the homological aspects of Mandarin, Korean and Spanish.. ..but he's now continuing to concentrate his research on the area of phonology.. ..and in multilingualism, so with L2 learning and language contact phenomena as well. And that's what he's going to talk to us today. Thank you. I'm going to talk to you about three main strands of my research that have been addressing these questions related to how people who speak more than one language, which is the majority of the world, organize the sounds of their two systems. I'll start with just very quick introduction as to some of these issues in the field of second language acquisition and how they relate to the studies that I'll be talking about. I'll walk through three different studies and they're slightly different in terms of what they're looking at and also the population. So they're either looking at change in the native language as a consequence of L2 learning, second language learning, or they're looking at the influence of the native language sound system on performance in the second language. So kind of both directions of cross linguistic influence and then I'll talk a little bit more about kind of why this is important for linguistic theory and also for practitioners of language teaching. So let's just start with some of the basic questions in the area. This is actually a very recent article from The Washington Post that speaks to an issue that's becoming actually very important in the field of language education, which is the fact that there are many people who want to learn a foreign language who already sort of know the language because they learned it when they were young and then they attritted, they lost the language as a consequence of becoming dominant in a different language of the community. And then they have arrived at the situation where their parents who were late onset learners of the dominant language of the community didn't really learn it and they have lost their native language. So they are in this situation where they can't really communicate with these people that they've lived with like their entire life. So this is just an example of an ABC and American born Chinese young man from the States who learns Shanghainese as his native language and doesn't speak it anymore. And his parents who are native Shanghainese speakers don't really speak English. So there's this communication problem and he now is trying to learn Shanghainese in school and he's a special kind of student because he has this early experience and part of what I'm interested in is what the nature of that experience is both for his Shanghainese and also for his English. So that's basically what the question is. How do native speakers of one language learn to command the sound system of a second language, which I'll refer to as L2 and kind of in the reverse direction, how does someone who is a native learner, a native speaker of some L1 first language, how does that native language system change over the lifespan as a consequence of things like learning a second language? And so what I'll try to show you in all of these studies is that these two sets of phenomena, first language perception and production and also second language perception and production in people who are bilingual can't be examined independently, but in fact there's insights that are to be gained by considering them both at the same time. So you use this sort of integrated approach to these two areas. Okay, so just some quick background. Anyone who studies L2 learning needs to know about these two things, which I'll refer to as universality versus transfer. So the first idea is that when people are learning a second language, it doesn't really matter what first language they speak. There are certain things that are common to people who are learning a target language in particular when they're adults. When they're starting to learn the language late, there are certain preferences, there are certain structures, constraints, processes that influence the learning process, and this is independent of the target language that we're talking about and also the native language. So just one quick example is that consonants tend to be easier depending upon what consonant. But consonants tend to be easier to learn at the beginnings of words or before vowels than they are at the ends of words. That's something that's been proposed to be universal across languages. So that's one set of phenomena in L2 learning. The second one is what's generally referred to as transfer. This is the idea that actually there are many very specific things that occur depending upon what the native language background is of the person who's learning this second language. So let's just take English for one example. Depending upon what your native language is, your friend's going to have a very different accent. And a lot of those pronunciation phenomena can be traced back to features of the native language background, the L1. So that's basically the idea of transfer, that your performance in the second language is highly influenced by features of your native language. To get a little bit more specific, the field of L2 phonology in particular has been influenced by... I'll try to delineate three specific principles that I'll end up contradicting. But the first one is that when you're learning a second language, there are often going to be some new sounds that don't occur in your native language. And when you're trying to predict how a second language learner is going to cope with this new sound, the way you make these predictions is to look at the two systems on a segment-by-segment basis. So the idea is that when a learner encounters some new second language sound, the way they get influence or the way the L1 transfers into the L2 is by virtue of this analogy between aliphones. So sound to sound or segment to segment. And so just to walk through one example, native Mandarin speaking learners of English, they tend to have trouble with this voicing contrast in English, but much more trouble with the contrast in final position than in initial position. So in word pairs like bit versus bid as opposed to tip versus dip. And one way to account for that according to this type of framework is to say that the native Mandarin learner of English, they're mapping this English contrast in final voicing at the ends of words based upon what sound in Mandarin is most similar to that English sound in that environment. So this is a position-specific level of analysis of individual segments. So for tip versus dip, it's not so hard because actually those sounds are very similar to a contrast that Mandarin does have, which is one of aspiration. So English T at the beginning of a word gets mapped to the native Mandarin aspirated T. And so that's actually pretty straightforward and the two sounds are very perceptually similar. So that's the nature of the explanation for why native Mandarin learners of English have this difference in performance between the two positions because it's position-specific. The second principle that pervades this literature is the idea that L1 transfer is at best neutral with respect to L2 learning. So for the most part it's going to be harmful because there are certain differences between the first language and the second language that are not going to be beneficial. They're going to interfere with the person's performance. So the second language learner, the adult learner is not a blank slate anymore and they have all these biases that they're bringing into the second language process from the first language which are sometimes irrelevant and also sometimes actually harmful. Because for example, let's take a Japanese learner of English. They tend to have trouble with this LR contrast in English and that can be explained in terms of Japanese having similar sounds but they don't contrast in Japanese. And so the Japanese learner, when they learned Japanese, learned actually to abstract away from the variation between L and R and to treat them as the same sound. So now when they go to learn English and have to now treat these as separate sounds, it's difficult because they've been trained in the native language to treat these as basically one category. Another way to explain this specifically with respect to phonetic cue is that they've learned to treat the cue to this LR contrast, the third formant or the F3 differently. It's not really contrastive in Japanese whereas it is contrastive in English. Okay, so that's the idea of negative interference. And then the third idea is that to the extent that you get any sort of flow back from the second language into the first language, this is going to be more evident when there's more second language there. So when you are very proficient, you have lots of experience with the second language, that's when you see these marked effects of second language proficiency on first language pronunciation. That's when you kind of start sounding accented in your native language or start forgetting it, for example. Okay, so we have these three ideas, alethonic linkage, negative interference, and then this positive proficiency influence relationship. And I'll end up arguing that all of those ideas are either false or incomplete and don't accurately represent what's going on with L2 learners. So the first study I'll walk through is this study of change in Shanghainese speakers of Mandarin. So these are Shanghainese Mandarin bilinguals. One thing to note is that Shanghainese is a minority language in China. It's not the socially dominant language. And it's spoken all over the world, but principally in the area of Shanghai, which you see here at night, and you notice that it's by the water, a port city, which is actually very important because that allowed for the history of Shanghai to be shaped in a particular way, where there was a lot of immigration that happens really starting around the middle of the 1800s. For about 100 years, vast waves of people came into Shanghai from neighboring regions, in particular people who spoke other Wu languages. So Shanghainese is a Wu language. In particular, a lot of immigration came from the region of Suzhou. That is what happened for about 100 years. And then starting around 1980, the second important event in the history of Shanghai is that China implemented a national language policy where Mandarin became the standard language and was heavily promoted as the prestige language, the language of civilized people, and the language that you had to speak in school. So up to this point, it wasn't uncommon for school subjects to be taught in Shanghainese, but after this point, even in Shanghai and among native Shanghainese speakers, all subjects had to be taught in Mandarin. In fact, students were punished for speaking Shanghainese in school, like outside of class. So you had to speak Mandarin in class, you had to speak Mandarin out of class, and that was a big change from basically 1980 onwards. Those two events actually line up neatly with the particular linguistic variable I'll be talking about, which is this contrast between two mid vowels, an e and an e. So canonically, Shanghainese had this contrast between these two vowels, and they are reported as distinct in three sets of the Shanghainese lexicon, which I'll refer to by what the vowel is in the cognate set in Mandarin. So there's a set of Shanghainese words whose pronunciation in Mandarin has this vowel I, and then there's a set that has the rhyme an, and then there's a set that has the rhyme a. And they all in Shanghainese get pronounced with some sort of mid vowel, so either an a or an e. Starting around 1850, they're still distinct, it's e, e, e. So there's still a contrast between, for instance, le and le, they're different words, and they sound different. And after that you get some merger of two of these sets, and this has been attributed to the fact that there was all this immigration from neighbouring Wu language speakers who did not have this mid vowel contrast. So this is actually one level layer of contact influence. So this contrast is slowly getting lost because all these second language speakers of Shanghainese don't have this contrast in their native dialect. And then the contrast is being reported as merged right around 1980. There's no contrast now between this lexical set and this lexical set. They're all pronounced with e. And that's what's been reported in the literature. Interestingly, now this contrast seems to have been revived because this set, the Mandarin A set has now re-adopted this A pronunciation. So whereas in this stage of the language there was merger and no contrast reported, at this point the contrast is re-emerged. So basically this has been a reversal of a merger. The contrast disappeared and now it's reappeared. So this is the variable that I'll be discussing. The degree to which people show this recovered contrast, this recovered distinctiveness. Because the hypothesis, given how similar that lexical set is with the Mandarin vowel, that Mandarin has an A vowel. That the recovered contrast is coming from language contact with Mandarin. It's not just some reversal of an incomplete merger where there were residual A pronunciations kind of floating around and those got augmented. But actually what's responsible for this merger is the contact with Mandarin. Helped along by the fact that there was this national language policy. So that hypothesis led to three predictions in particular that young people who are more subject to this policy because they went through either all or most of their schooling with this Mandarin only policy, they're going to show that recovered contrast more because they're more influenced by Mandarin. The second prediction is that both young people and older people, they will show the contrast more or the merger reversal more when they are operating bilingually. So if they have to use both languages like for example in a translation task, then they'll show the recovered contrast more. And the third prediction is that the reversal, the appearance of the A pronunciation is going to be specific to this lexical set. It's not just going to show up randomly in the lexicon, but actually it's this set of words whose rhyme in the Mandarin cognates is A. They're going to show up with the A vowel and in particular it's not just going to be A, it's going to be diphthongised, it's going to actually be an A vowel. So those are the three predictions from the first hypothesis. The second hypothesis is that this is not just about segment to segment. So the analogy here, at least from the first hypothesis, is that E is being analogized to Mandarin A and that's how it's being changed into A. But the second hypothesis that we had in the study, oh I should also mention that this is all joint work I've done with my colleague Yao Yao from Hong Kong Polytechnic, who's a native Shanghainese speaker. We got interested in this actually because it was pointed out to her that she was pronouncing Shanghainese wrong by an older Shanghainese speaker. And one aspect of the pronunciation that was pointed out to her as incorrect was the fact that she was diphthongising this vowel. And so we got very interested in, I mean she's grown up her old life speaking Shanghainese, so how was she speaking the language wrong? And yeah, so that's why we got very interested in this. But the second hypothesis that we had with respect to this change was that the degree to which you're getting some sort of influence from Mandarin phonology is not strictly because this vowel is similar, that you have this natural perceptual similarity we can think about in terms of acoustics or phonological features, but there's very, there are a lot of things shared between E and A. They're quite similar vowels. That the degree to which you get this influence from A is not going to just be dependent on the occurrence of A, but also other things in the context. So the whole word, the similarity at the level of the whole word will condition the degree to which you get A influencing E. So this is now already departing from the basic thrust of the second language literature, which is basically all about segment to segment relationships. OK, so we have these two hypotheses. And just to recap, we're looking at these three lexical sets. So there's three sectors of lexicon in Shanghainese, the relevant ones that have this vowel, that differ depending upon what the rhyme is in Mandarin. In Shanghainese they're pronounced identically, or at least are reported to be at some stage of the language pronounced identically. So this is le, le, le. They're all le. And they differ in terms of what the pronunciation is in Mandarin. So the Mandarin I words have I and the Mandarin N words have N and the Mandarin A words have A. And the idea here is that it's not such a problem to represent the Mandarin and Shanghainese phonological forms of these words differently when they're not very similar to each other. But when they are similar to each other, then there's this tendency to equate the two, to link them perceptually across the languages. And this linkage, which I've shown here, is what allows for there to be this cross-linguistic influence. And in particular that's going to happen at a lexical level, not necessarily at a segment to segment level. So this is the set that we'll be watching out for in particular. OK. So what Yao did was go to Shanghai and recruited a bunch of Shanghainese speakers. And we did this in a very controlled way because Shanghai is a very big city. And so we recruited the participants in the form of parent-child pairs. And we did this to control for the fact that people move all around Shanghai. It's a big city. So we wanted to control for immigration history and where they lived in the city. So this is 24 people and it's 12 parents and their children. And that's what the contrast is. That's the two groups that are being compared. They were asked to produce the critical items that contain this vowel, the Shanghainese e vowel, in these three lexical sets. And for the first study, there's three experiments, but for the first study, the reference level in all the models, the baseline that we'll be comparing everything to, is a pronunciation of this Mandarin I set, which is a set that we don't expect there to be any diphthongisation in. We also were interested in maybe there would be some effect of frequency of the lexical items. We entered that as a factor in the design and then also whether there would be any effect of the onset consonant. So these are other factors. The most important thing here in terms of the design is that the speakers are producing these items in two different types of tasks. And these tasks are designed to mimic distinct language modes. We had the participants read the items in the context of a meaningful sentence where there's no Mandarin involved, other than the fact that they're reading Chinese characters. So basically, we have them read a sentence and just ask them to read it in Shanghainese. All the interaction with them is in Shanghainese. There's no Mandarin produced in this experiment. So that's the first task, which is as monolingual as we can get in Shanghainese, which otherwise is problematic to elicit without use of some sort of writing form. And the second task, which is meant to be bilingual, to elicit a bilingual mode, is this translation task. And what they're doing is basically they're hearing a Mandarin word being uttered over speakers, and they're asked to immediately translate it orally into Shanghainese. So this is inherently a bilingual task. They have to have both languages activated because they're hearing Mandarin and immediately speaking in Shanghainese. Okay. And then we're looking at the acoustic correlates of vowel height and backness, F1 and F2, at the onset and the offset of the vowel to see what the quality is at both of these points and in particular whether there is movement of the vowel quality, movement of the formants because that would indicate diphthongisation. I'm not going to talk so much about the statistical analysis, but it's in a mixed effects regression model and we're entering all of these factors that I've talked about to see whether there's going to be any effect of these different predictors and then also entering in random terms for the fact that individuals are going to vary and also what family they ended up in is also going to have to some degree an effect on the way they perform. Okay. So an experiment one, what you see here is a graph of how many of the tokens they produce get diphthongised towards the E direction. So if they move up and front then we counted them as diphthongised towards E. What you see here is that the Mandarin I baseline is in the middle here because that's what we're comparing both Mandarin An and Mandarin A to. There's no significant difference between the An and I words. Those both are really not very often diphthongised whereas the Mandarin A words are frequently diphthongised by both the older and the younger groups, although the younger groups show more of an effect. In the translation experiment the pattern is similar except everything is augmented in the Mandarin A set. So there's more diphthongisation for both groups but still the younger speakers are doing it more. If we look at the quality in terms of F2 and F1 so these are vectors that are showing you the change in the formants at the onset and offset. So if you see a long line going in this direction that's showing you that it's being heavily diphthongised towards E. It's going from A all the way to E. In the reading experiment on the left you see that the Mandarin A sets which are these squares which you can't really see so I'm just going to point them out are getting much more diphthongised. That's both for the male and the female speakers and both for the older speakers in the grey and the younger speakers in the black. But the younger speakers in the black are diphthongising more. Those lines are a lot longer and if you compare the reading experiment to the translation experiment it's quite pronounced that both groups are diphthongising more when they have to operate bilingually and again the younger speakers are doing it much more than the older speakers. So I have a few sound clips here of people uttering this item thuppe which means to match and this is family 11 the mum and her daughter in the reading experiment. That's the parent. That's the daughter and I think you can hear already that the daughter diphthongise is a little bit more than the mum. And then in the translation experiment you can really hear especially the daughter diphthongising. Where the daughter is really diphthongising it's just an A vowel it's almost like the Mandarin vowel. So basically we've confirmed this suggestion in the literature that yes, there's this A vowel emerging and that is essentially reviving this contrast that didn't used to be there at least at a certain stage of the language. So yes, these Mandarin A words do seem to be more likely to be diphthongised and they're diphthongised to a greater degree than these other two lexical sets whose Mandarin rhymes are not so similar. And we're seeing this trend more in younger speakers compared to the older ones in the bilingual task compared to the monolingual task. Now we wanted to see whether this would be modulated in some way by features of the lexical context because that is the main hypothesis that this is not just a feature or a side effect of A being close to A but also other aspects of the context being similar or dissimilar. So the question is basically if you have aspects of the lexical context that are different between the Shanghainese item and its Mandarin cognate does that lessen the degree to which you get cross-singuistic influence from Mandarin? According to an alaphonic linkage account it shouldn't affect it because the only thing that's relevant is A to A relationship but in the view that I'm promoting basically this should matter relationships between words are also something that should influence the degree to which you get influence from a second language. So in experiment two we're looking at a subset of Mandarin A words that I'm going to refer to as structure mismatched because the Mandarin cognate has this additional medial what in traditional Mandarin phonologies is analyzed as a medial you could say that like in a twey word for example that way is just a vowel it's a complex vowel but the traditional analysis of Chinese phonology usually has this somewhere else in the syllable structure this is somewhere else and where we put it is not super important and I'm not going to commit to a particular analysis all that is important is that it is different this set of Mandarin A words is syllabically different in structure from the Shanghainese in comparison to the maximally matched Mandarin A words that we saw in experiment one are being compared to the regular matched Mandarin A words and also to the baseline Mandarin I in experiment three we're doing something similar but now we're looking at effects of dissimilarity in onset so if the Shanghainese word begins with a voiced onset but the Mandarin word begins with a onset of different voicing will that affect the degree to which you get this Mandarin influence so the same two sets are being compared to in experiment three one thing I want to mention before I go through these graphs is that we don't know and there was no prediction in advance of whether the Mandarin A mismatched set would show a significant amount of difference from the Mandarin regular Mandarin A set the matched set because in fact one way to think about this is that the Mandarin A set is leading this sound change and they're doing it because they're so so similar to the Mandarin the Mandarin cognates the vowels and the onset everything is identical essentially the structure mismatched and also the onset mismatched Mandarin A words they are very similar too but they're a little bit different and the prediction is that they're also going to undergo the change by virtue of the the vowel which is still similar and the prediction is that they're just going to be behind lag in the sound change the factor of the matter is that there's really no way to predict how far behind they will be so the important thing here is to look for patterns in the data that are either consistent or inconsistent with a scenario in which they are lagging so there might be some cases in which they're not any different from the matched Mandarin A set but what would be important would be if they actually show a difference from the Mandarin A set, the matched set that is only in the more strongly cross-linguistic task the translation task as opposed to the reading task so that's the relevant contrast to keep in mind but the fact of the matter is that when we look at experiment to we do see that the mismatched Mandarin A set is squarely intermediate between the matched Mandarin A set the regular Mandarin A words and the Mandarin I words when we crank up the level of potential cross-linguistic influence and make them do translation now the difference between the mismatched Mandarin A set and the matched one goes away and now they are not significantly different although it tends to be a little bit less diphthongized and that's also the case in translation where in the reading experiment these formant trajectories for the mismatched Mandarin A set are not as big I mean they're also diphthongized as the matched Mandarin A set that difference goes away in the translation experiment where they're both very diphthongized the reading experiment is similar except some of these differences go away so the Mandarin onset mismatched Mandarin A set is intermediate but the difference here is not significant and that difference goes away basically in the translation experiment and you can see that in the formant plots too so basically we found things that are consistent with the predictions we found that the structure mismatched words they are showing less of the sound change than the regular Mandarin A words and that that only happens in the task that doesn't overtly encourage cross-linguistic influence so the reading task that happens both respect to the offset formant values and also the rates of diphthongization the pattern is similar in experiment 3 although some of these differences go away so what these results taken together seem to be suggesting is that the way people are relating these languages cross-linguistically is not being influenced only by these segment to segment relationships but also these relationships between lexical items that there's a lexical linkage not just segment to segment so that's the main conclusion to draw from these results so these are fluent bilinguals they speak Shanghainese on a daily basis they speak Mandarin on a daily basis and the idea was whether you'd also get similar multifaceted levels of cross-linguistic influence in people who are just starting out so maybe the phenomenon is actually different if you were to look at ab initio learners people who are just starting out, the novices and this set of experiments I'm going to talk about was sort of related motivated by statements in the literature on L2 speech that goes something like this where an L2 that is hardly mastered should not have much influence on L1 while an L2 which is mastered to a high degree should exert more influence so this is expressing the idea that there's this direct correlation between how good you are on the L2 and how L2 influenced you are going to be in your L1 so that raises the question of so is it the case actually that people who are just starting out are immune to this cross-linguistic influence because that's essentially what this is saying and also is it the case that you get this direct relationship between proficiency in L2 and influence from L2 and again the same question that we had in the Shanghaini study is this cross-linguistic relationship that's leading to the influence occurring on an alethon to alethon basis so those are the same questions that we saw in the Shanghaini study but here we're looking at people who are just starting to learn and these are speakers of American English they're young college graduates and they've gone to Korea to teach English part of their training is to learn Korean because they're going to be in for some of them they're going to be in very rural areas of the country and so they're undergoing this very intensive six-week course of Korean instruction it's starting from scratch and during this language program I was there at the site and I was interested in well actually I originally gathered these data as a control measure but then started seeing all these changes in their English and then that became my dissertation so what happened here was they were reading the same English items every week so the same set of items and the same procedure and I'm analyzing three main sets of measures the voice onset time in their stops this is correlated with the voicing contrast and then the fundamental frequency or F0 onset in the following vowel which I expect to be influenced on the basis of some properties in Korean that I'll be talking about in a second and then the acoustic correlates of vowel height and backness F1 and F2 and the statistical analysis is similar although I'm using a slightly different version okay so why would we think that the pronunciation of English stops would change as a result of learning Korean well one reason is that the Korean stops are quite different in terms of their phonetic properties so this is a table of VOT values of the English stops in comparison to the most similar Korean stops and what you see here and the important thing to take away is that for the voiced stops the quote unquote voice stops of English BDG those ones spelled that way in comparison to the short lag VOT stops of Korean the fortest stops there's basically no difference they're very very similar predicts then that there isn't going to be a lot of effect on the English stops because learning this very similar stop series is not going to change their pronunciation of the voiced stops the aspirated stops on the other hand they're all significantly more aspirated than the English voiceless stops which are also aspirated in at least word initial position before stress so what you see here is that consistently the Korean series of stops is more has longer VOT but the difference differs depending upon the gender and also the place of articulation and one thing I want to point out in particular is that for female talkers coronal stops are not very different between English and Korean and in particular that difference is the only difference comparing voiceless to aspirated that falls below the just noticeable difference or JND for VOT so humans have a limit on the degree of acoustic difference that they can discriminate and for VOT it's been shown that generally this falls somewhere around 23 milliseconds in the range that we're talking about and this difference is smaller than that all the other differences are longer and so that's something to keep in mind because then these different places might pattern differently so that's the picture for VOT so the prediction basically is that if you're going to be Korean influenced in your English voiceless stops you should start aspirating more and more and more F0 the reason why this is relevant is that these stop series both the four stops and the aspirated stops of Korean are laryngily marked and they've been described at nauseam in the Korean literature but the important thing is that they're both associated with this pitch accent on the following vowel they raise the voice pitch of the following vowel and given that these are the stop types that are most similar at least in terms of VOT to the English stop types what that predicts is that if you're going to be influenced in your voice pitch by learning Korean everything's going to go up your voice pitch is going to go up at least following the stops so is that what happens what you see here is both of these variables VOT on the X and F0 standardised against each person's range on the Y and the women are in black and the men are which are fewer are in grey and the others are showing you their value for that week in the language program so basically you just follow one, two, three, four, five so let's start with the females so the females are not really changing much in VOT keep in mind this is two milliseconds so that's nothing so they're not really changing in VOT but they're steadily going up in F0 so it is the case that over time their pitch is going up consistently in VOT but their F0 is not changing in the same way we can talk later maybe in the questions about why this might be but it's important to point out that all the teachers in this language program are female there are no male teachers so the males do not have a male model for L2 so yes so for this range it's generally around four hertz or so so the effect of time is significant for VOT sorry not significant for VOT but is significant for F0 and that's being driven by the women there's this gender by time interaction it's only for the women that the effect of time on F0 is significant so those are the voice stops the voice of stops are a little bit different because the women are increasing both in VOT and F0 so they're aspirating more and they're also raising their voice pitch whereas men are aspirating more 1, 2, 3, 4, 5 but again they're not showing this consistent effect on their voice pitch so time is significantly affecting VOT and F0 and there's also this gender by time interaction ok so this is completely consistent with the predictions what's interesting is that if you break the data apart by place of the stop so these are the voiceless stops to see how much more aspirated people are getting that there's a difference between the places where if you recall the alveolar stops the coronal ones they were so similar in VOT at least for the females that the difference between them was not going to be noticeable most likely they are the place of articulation that also shows the smallest increase in VOT so there's two things that are significant here the first one is that they increase in VOT at all because if they're related segment to segment that should not happen so the fact that they're going up in VOT as well suggests that there's something else going on that these learners are equating not ta to ta but voiceless stops with aspirated stops and are equating these norms and that's how you're getting this influence on alveolars which shouldn't otherwise happen however that influence is being modulated by this segment to segment linkage as well so I don't want to say that doesn't exist because it does seem like that's having some sort of influence and tamping down the effect of the natural class level voiceless stops to aspirated stops if you look at the increase in F0 and compare the increase for the vowels following voiceless stops with the items the control items in the set that don't start with a stop so these are words like all but to start with a vowel what you see is this so this is time versus F0 so you see an increase in F0 of the vowel initials too and again this is not expected if the vowels are getting equated to other vowels because there's no F0 in elevation in Korean just because you start a word has to do with an initial stop so again this is unexpected on a segment to segment basis but makes sense if we're thinking about F0 being linked globally across language that there's some sort of shared control mechanism for a modulation of voice pitch that is being influenced globally by the fact that you have to raise your voice pitch a lot in Korean for the articulation of these stops but you notice like we saw with the alveolar stops that they're like intermediate they're not showing the same increase as the other places as the other onsets and that is seems to be indicating that there is still this influence of the stop to stop relationship that the ones that are stop initial these items are going up in F0 more than the ones that don't have the benefit of starting with that stop so that there's these two levels of influence that are going on is what you should take away from this so that stops so English speaker stops are changing rapidly what would happen with vowels so I looked at all over the literature to find some phonetic norms for English vowels of the states and the important thing to take away from this figure which is very full of things is that first there's enormous dialectal variability within the United States in terms of the organization of the vowel space those are all the colored trapezoidal like things or parallelogram like shapes and then the Korean heart shaped vowel space in black first of all contains fewer vowels and the inventory differs in two systematic ways there's fewer low vowels there's fewer high F1 vowels because it's just ah they don't have an at and an ah for example and there's also fewer front vowels there are fewer high F2 vowels because English has E E E A whereas Korean has E and A so that's the important thing to keep in mind because when we look over the systems at the way these vowel systems differ phonetically that leads to this difference where it doesn't really matter what American dialect we're talking about the Korean vowel system if we take F1 and F2 on average over all the vowels in the system is systematically lower in F1 and also systematically lower in F2 by virtue of the fact that there are fewer low vowels and also fewer front vowels and I should also point out that these differences between the systems are greater than the just noticeable difference for frequencies in the range of vowel formants so I mean this is much bigger than like 4 hertz ok so what happens to the vowels well on a segment to segment hypothesis we should get the vowels actually going all over the place because their position in this F1 F2 space is different with respect to some sort of nearby vowel neighbor and you don't get that you don't get this disparate motion of all the vowels changing positions some going down some going up some going left some going right they all go up and these are the females because they're the more numerous ones but actually the few males that I have do do the opposite and we can talk later about why that might be but this pattern seems to be some sort of global thing that's happening all the vowels are going up and this is the case regardless of whether the nearest vowel that they're closest to would have them go down versus up so like for instance here is going up even though that takes it farther away from the nearest Korean vowel which is and this is the horseshoe vowel the put vowel this vowel is also going up but that actually brings it closer to the nearest Korean vowel which is so it's not really possible to construct a coherent analysis of this where people where the learners are constructing these segment to segment relationships and consistently wanting to assimilate to the second language system basically you get this generalized raising across the system okay so there's a sum up really quick for the stops we saw a change in F0 and also in VOT in these novice learners this happens very quickly after essentially a week of instruction so it's happening quickly it's happening in assimilation to the second language and it's happening in this way that seems to both be specific to segments and also generalizing above the level of the segment whereas vowels are doing something different they seem to be changing in a way that is systemic there are changes happening to the vowels that are not occurring on a vowel to vowel basis but happening over the entire system so the conclusion basically to draw from this set of results is that novice learners do show the same type of influence on the second language as these advanced learners that have talked about a lot in literature and the way there are being influences on the basis of some perceptual linkage that is not just occurring on a segment to segment basis now when I looked at the amount of VOT lengthening that was occurring in the novice learners I was surprised at the fact that it was approaching 20 milliseconds which is a large increase in VOT and when you compare that to the results that have been found for advanced learners it's a lot bigger and I was confused as to why this might be but then I thought about the fact that the fact that they don't know Korean could help this happen because it's so new to them this new language is so I guess you can think about it as exciting perceptually they've never heard this before so could that give some cognitive boost basically to the encoding of this information maybe that's why they're actually showing this pronounced effect in order to see whether that was the case I compared those novice learners to people in the exact same program and these are people who have learned Korean before so they have some experience with the language either through heritage exposure or formal study and the question was will they show less of an effect on their English by virtue of the fact that this is not so new to them they've heard this before and that is what happens so if you look at VOT for voiceless stops they do get influenced by the fact that now they're in this immersion environment studying Korean intensively but the effect on their English VOT is not as great so you get the significant interaction between time and group where that's coming about basically because the increase in VOT is not as great for the experienced group and it's the same thing for F0 so they are getting influenced in terms of F0 raising but it's not as great as it is for the novice learners ok so basically the conclusion of that subset of studies is that as you increase the amount of experience so your proficiency also in the L2 actually what's going on is that the potential influence on your L1 goes down and this has started to be followed up on by people from Judy Crowell's lab and an alternative explanation for this is that there's also some inhibition that's going on where when you're just starting to learn a second language it's harder to do it because your native language is relatively more strong compared to this new language so you have to do more work in terms of inhibiting the first language and that aids the influence of the second language on the first language so that's one converging I don't know that's actually inconsistent with what I've proposed but that's another way to think about these results because of that it's more proficient in L2 the L1 is now not as relatively strong as it was before so there is not as much of a need to inhibit that system okay so these are novice learners and so far I've talked all about L1 and really there's two levels two directions of influence that we should be considering so last set of studies that I'll be talking about has to do with influence of L1 on L2 and I love this set of studies because people always like hearing the sound files so I'll get to those very quickly which is right here so this is a study that had to do with perception, speech perception so so far I've talked about pronunciation but here we're talking about final stops in English, final voiceless stops and they vary in terms of whether there is full pronunciation, a release or an unreleased pronunciation and these are spectrograms of the same speaker uttering the words PUP and PUT either released PUP where you see that there's a lot of information in the burst which is here in the spectrogram there's a difference in loudness there's a difference in the frequency distribution there's a difference also in the duration when you don't have that information and the stop is unreleased PUP you can still hear the difference and the way you hear the difference or the thing you're queuing into is this difference in the transition in the final part of the vowel so here the F2 frequency is showing a different trajectory here it's relatively flat in PUP and in PUT it goes up quite steeply so this is how without the benefit of the information in the burst one can tell that someone is saying PUT as opposed to PUP so the question here is how native language experience will affect the way people do this in English and before I get to those results it's important to point out that American English which is the target language here shows this disparity between frequency or how often something occurs in the language and the status of a variant in the language so what I'll refer to is ethnicity but for final voice of stops they are very frequently not released at all places of articulation but despite that the released versions are still recognized as canonical and have more influence in priming tasks for example and so this is really the variant that we could reasonably posit is somewhere in the underlying representation where unreleased T is very clearly the more frequent in conversational American English but still American English speakers you put them in a priming experiment and you ask them to identify dog as a real word they're going to do that quicker if they hear cat as opposed to cat even though cat is the more frequent pronunciation so there's this disconnect between disparity and canonicity for release in American English this is different for other languages where frequency and canonicity coincide is in fact the unreleased variant is the only way you can pronounce these sounds so this is the case in Korean where final stops final voice of stops must be pronounced as unreleased in fact if you release it it sounds like you're saying like an additional syllable so this is also important because final stop place contrast has a heavy functional load so there are many minimal triplets like pat pat pat that only contrast in the place of articulation of the final stop so that leads to the question of if you are a Korean speaker and you're learning English will that help you in telling apart these final stops in English because you were trained so much in Korean at using that information in the final vowel to tell apart these final place contrasts so there's three groups in this study and so the question is whether these same unreleased stops in American English will be better perceived by the people who should be the old standard, the native listeners or people who are not native in some way but have some experience from another language and the groups here are the native English group who have no experience with Korean or any other language that has unreleased final stops like Cantonese or Thai for example they stack up with the other groups in the following way so for all three of these groups both the native English, the native Korean late learners of English so these are mostly people who came to the states in their 20s or grad school or something like this and then the heritage Korean group which is a group of Korean Americans who either were born in the states or came to the states very early and essentially are native English speakers they've been educated all in English and they have this early experience with hearing Korean too so for all of these groups the unreleased stops are at least frequent in their input for two of these groups the native Korean group unreleased version is also the ideal version for some language the canonical version and two of these groups the native Korean group and the heritage Korean group has early exposure to Korean but only the native English group and the heritage Korean group has early experience with English so the native Korean group came late to the states so that's how these groups stack up against each other and if we were to take this information into consideration then if we believe that being ideal in your language might give you some sort of perceptual advantage on perceiving that structure then the prediction is that native English listeners are actually not going to be the best at doing this even though English is their native language the Korean speakers will be better because they've received more training essentially in how to do this and I looked at this in two conditions one with non-words so without this information and then one with lexical information so the first experiment has to do with identification and the stimuli are being spoken by these two Maryland English speakers and the groups are as I've shown you here and they're pretty similar in age they're all in their 20s but again the Korean group they're generally grad students so they're a little bit older and one thing I want to point out is that this heritage Korean group they are native English speakers they've grown up in the States but they have this additional experience with Korean other thing to point out is that they're not native Korean speakers technically Korean is their first language but their Korean is not native like both in terms of how I mean I also tried speaking with them in Korean afterwards after the experiment and also had them rate themselves for example and their Korean is not native like no one rates themselves as native like on these measures ok so this is what they're listening to nonce words that are bysylabic and they're constructed to be specifically English like so for that purpose the first CVC is rwz so rw is very Englishy it also doesn't occur in Korean z is another segment that doesn't occur in Korean and then the second vowel and the consonant vary in the following way where the second vowel contains either a monopthong i-u-a or a dynamic vowel nucleus a-o-i-r where those nuclei do not occur in Korean and also objectively should be harder because they inherently have formant movement so the task of trying to tell what the formant movement is for the vowel versus what's coming from the consonant is going to be harder for these dynamic nuclei and the consonant varies between either t-t-k or unreleased or nothing so we could have and there's a variation in stress so we could have items like rwz-i-t versus rwz-i-t versus rwz-i-t so these are the types of things they're listening to and to make it even harder for the Korean speakers I embedded these in a sentence and had them basically identify the last sound of the final word in the sentence so they're hearing something like this so I'll play it one more time so you can try to guess what the correct answer is now the word is so the correct answer is k now the word is rwz-arc so this is a speeded task they're being asked to do this as quickly as possible to say p-t-k sounds like p, sounds like t, sounds like k or sounds like something else and the analysis is similar to what I've done before in the previous studies and in this identification task this is what you get so the native Koreans are in black the heritage Koreans the intermediate group is in gray and then the native English group is in white on the stops the final stops on the left the native Korean group is better than the native English group although that difference is only marginally significant the heritage Korean group is significantly better than the native English group for the rwz-i type stimuli everyone is good people are performing a ceiling it's not that hard to say that rwz-i doesn't end in a p-t-r-k so that's expected importantly if you look at the same data but just look at the judgments on the stimuli that had a-o-i-r it's the same so that advantage of having Korean experience extends to when you're looking at vowel environments that do not occur in Korean so this is some sort of general knowledge about how to extract information from the final part of a vowel because they never would have gotten that experience from having this form and transition for A in Korean that's not something they could just be pretending like they're hearing in Korean if you look at the reaction times what's important about that is a speed accuracy trade-off type of thing because in fact these groups and particularly the heritage Korean group tends to be responding faster so it's not the case that they're doing better just because they're taking longer and thinking about it more to enter their response so at least without lexical information there seems to be this benefit of knowing Korean maybe that was a little unfair because people use lexical information to speak to perception to extract the lexical information to the native English listeners and maybe they'll do a lot better then because now they have native lexical knowledge that they can use to narrow down the set of candidate parses of some ambiguous signal so this is an AX discrimination task and the stimuli here are these monoslavic English words that were balanced for spoken frequency using the corpus of contemporary American English so these are stop pairs that differ in the place of the final stop so like lip versus lick and also pairs that differ in whether there is a stop like peak versus P and so the task is as quickly as you can say whether the two different people are saying the same word or different words so the experiment goes something like this le so same or different le yeah different lake versus lay and here is another pair different they are different cat versus cap yeah so what you're going to see as the dependent variable on the graph is D prime which is a model of how sensitive people are to some sort of change in the signal and what you see is consistent with what you saw in the identification experiment so there is this tendency the difference between the native Korean group and the native English group is not significant but the heritage Korean group shows a non-significant tendency to outperform the native English group on the stop zero pairs the native Korean group is better than the native English group and the heritage Korean group is better still so they are the best of all and they are also better than the native Korean group if you look at the reaction times to the same situation the heritage Korean group is actually faster they tend to be faster than everyone so this is not a speed effect ok so basically what we have here is seemingly at least some beneficial effect of having some different language experience that is not the target language so both the groups of Korean speakers the native ones and the heritage ones seem to be better than the native English ones at perceiving English at perceiving the stop contrast to English and this is not due to priming for example because I didn't address the heritage Korean speakers in Korean so I spoke to them all in English up until after the experiment so this is interesting because if you look at some other findings in the literature in particular those of Sun Young Lee Alice who worked on heritage speakers too of Korean it seems like there is this kind of best case scenario in terms of early perceptual experience where in her case she was finding that the heritage Korean speakers that she was looking at they were patterning like un-English speakers native English speakers are perceiving some sort of English specific contrast in syllable structure and in this case what I'm seeing is that if the Korean experience has a potential benefit then you do get transfer from Korean so she found no transfer from Korean because that transfer would have been negative would have been detrimental and what I'm finding is that if the transfer happens to have a potential benefit then you do get the transfer so that taken together this suggests a kind of best case scenario in terms of transfer from the native language also bilingualism is in the news a lot lately because of these cognitive benefits that have been identified in terms of delaying Alzheimer's and so on and what I would add is basically that knowing another language is beneficial because it gives you more resources to tackle some sort of task so these native Korean and heritage Korean listeners they've learned that there's a lot of information in the final part of the vowel maybe we can use that to our advantage and they've learned that in a way that the people who have heard English their entire lives haven't gotten from that experience okay so just returning to those three points that we had at the beginning it doesn't seem like the way people are equating structures across languages is segmental specifically going on at the lexical phonetic level as well as the segmental level things below the level of the segment or natural class level and linkages also at the level of a system like we saw for vowels the way L1 transfers going to affect what you do in the L2 is not going to be necessarily negative but actually it'll depend specifically on the properties we're talking about and how they match up with each other and with respect to this flow back to L1 this seems actually to be the reverse of what has been assumed in the literature in fact when you gain word experience it decreases the amount of influence that you're going to get cross-singuistically okay so this approach the reason why I've taken it is because I think that looking at the way two languages are interacting within a mind gives us something in terms of insight that you can't get without looking at a cross-singuistic situation so in particular for instance looking at historical sound change if you want to analyze patterns of lexical diffusion for example there are certain patterns of sound change that might be explained in the exact way that I've talked about with Shanghainese where there's some specific cross-singuistic similarity that is motivating the change and that's where it starts and that leads to a specific pattern of diffusion across the lexicon that's something that you can't really explain in terms of or if it's a contact induced change you need to look at both languages and how they line up with each other in order to make some sort of prediction the other thing that I would point out is that for example for vowels and change in vowels you need to look at a cross-singuistic situation in order to understand that at some level speakers are representing knowledge about overall height or overall feature of the vowel system which you can't get unless the language happens to have some sort of phenomenon that is making reference to height over all vowels of the system whereas here we've seen that effect come out by virtue of this influence or change in the overall height of the system across languages so something that is a systemic feature might really be better investigated by looking at cross-singuistic phenomena ok and just practically people who are doing research I think need to really carefully define who they are looking at because when you look at linguists who are working monolingually often actually you find that the population they purport to investigate is not actually the population that they recruited so for example if you look at studies of Swedish for example they are not specifically about Swedish bilinguals yet a lot of these studies look at Swedish bilinguals and it's not reasonable to conclude that that study is going to generalize to monolingual Swedish speakers living in Sweden because they're not monolingual Swedish speakers living in Sweden they're a different population and you know I've only talked about phonetics but really this is the case at any level of language many other studies have shown that in aspects of morphosyntax, aspects of cognitive representation and also gesture that people are influenced by this other language knowledge that they have and just in terms of just language study I mean if you are if your goal is to get native like input that's not likely to happen if you're not living in the native speech community by virtue of the fact that your teacher unless they're just really sheltering themselves from the dominant language of the community is also likely to be influenced by the dominant language of the community so I mean basically the implication is that if you are learning from an imported language instructor it's not that they're not going to teach you well but that input is going to be different than the input you would get in the native language environment okay just a quick few words about where to go from here one thing that I'm very interested in is the degree to which having acquired this second language experience seems to have at some level restructured a representation of your native language how long does that restructuring last because there's virtually nothing in the literature that is looked at effects or influence of distal or distant second language experience that was acquired a long time ago so you learn some language and you go back to your native country basically don't speak that language for a while do you go back to returning like do you return to a language like pronunciation for example is not a question that has been addressed really to degree that I've looked at it with this set of learners the pattern seems to be no you don't go back to baseline you stay a little bit different but this is something that I think needs to be looked at systematically and then with respect to the heritage advantage the one thing I want to point out is that it is not always the case that heritage speakers are native like in the heritage language or the dominant language and so what are the set of factors that determine whether they extract the maximum benefit from the heritage language I think is a major research program and I think is really promising in terms of questions for these sorts of cross-singuistic issues that's it one thing I want to mention is that there are many many studies here and they benefited from the input of lots of people and also funding from lots of institutions and discussion with lots of labs around the states and also the world so thanks to them and to you thank you very much Charles for some very some very interesting findings I'm sure there have been some questions you have at the time for these guys any questions? you mentioned that in short the effect on L1 of experience because it's less than by love is it maybe because they're already raised their voice no so that's a good question if you look at this plot is this right? yes so actually the difference in week 1 between the novice and the experienced learners is not significantly different and that's also the case for F0 so it's not the case that it's they're not going up as much because they've already gone up a lot I guess a follow up to that while other people formulate questions is that you would think that that would suggest that the question I refer to at the end with respect to if after you've been removed from the second language experience do you go back to sounding monolingual that would suggest the answer to that question is yes because these people who learned Korean at some point in time in the past after one week of instruction they're not yet different from the people who have never learned Korean before so it seems like if we kind of think about what might have happened in the past after they learned Korean in their first year undergraduate course or whatever they went back to being English like after they stopped using Korean what I think actually is going on is that that does happen but really what the important variable is is alternations and experience so they have gotten the benefit of essentially a break from Korean so they've learned Korean at some point in time and then they've lived in the States and have been operating mostly in English and that has essentially allowed the first language English system to sort of recover and go back to whatever it used to be and then now when they're encountering second language experience it's not as influential yeah but I mean I think the degree to which it actually does go back to baseline is still an open question because really we're just guessing as to what happened in the past and we don't really know yeah oh yeah so the male learners in this program they do not increase their F0 and the most likely explanation for this that I think is behind this is the fact that there is this cross gender situation with them so they don't have the right gender model in terms of their target language and so in that respect I think that they don't have the same motivation to accommodate to this higher F0 because maybe this higher F0 is higher because it's coming from a different gender the women are all women so when they have this model of a female Korean teacher talking to them they don't have the same excuse for not accommodating to that F0 because it's the same gender so the fact that this Korean speaker is using a higher F0 means that there's something significant about the higher F0 if you're a male learner and you're receiving instruction from a female Korean teacher you don't really know or it's harder to know whether the higher F0 in Korean that they're hearing is due to the fact that Korean has higher F0 or to the fact that because she's a woman she has higher F0 I think that might be interfering with their processing basically the way they're interpreting this information of higher F0 the other thing is that just generally in the sociolinguistic literature it's been pointed out that it doesn't just modulate F0 in a way that is different from females so it may be the case that males just on balance are more averse to increasing F0 range or increasing F0 in general so that's a possible other explanation yes I was wondering about one of the issues that for when you think about the cognitive system how the structure is a label of density on the different levels that take place from primary literature that when you get to a certain level semantics slows you down while phonetics and phonologies speeds you up so one of the questions I was wondering about is when you get to a certain kind of level of efficiency how this then which transfer over because one of the things that for example Mario Oberg's work shows is the question is at which point in time when you have acquired a particular language does a semantic reorganisation have come at a potential level astounding, right? what you can see in a bench reference motion to just the word and so this then speeds down into the semantic representation that then again speeds down into reorganisation of a phonological scenario right, right I think that's a very interesting point and you know I actually I don't I don't know of a good way yet to test or I guess operationalise how well developed people's semantic representations of Korean are yet but I wouldn't be surprised if it were the case that part of the reason why they seem that when the experienced learners are going through the same language program that they are being less influenced is that at least some of the words in their second language lexicon have more opportunity to be separate because they have a more robust semantic representation or more robust, I mean I think that's very likely to be the case, yeah you have an interaction with neighbourhood density on a phonological but also on a phonological level that is also interacting so that's another difficult thing to find out if you want to take part because that's not a part I mean this relates to the first part of the study, the lay study where one of the issues of course you know frequency is one thing but frequency is kind of war there's more complex interactions the density does and one thing I want to mention that also complicates this is that we have not considered tone at all which is actually one component or proposed to be one component of neighbourhood density in Chinese languages because it's a feature of a word and the reason why we haven't considered it is because it is just too difficult Shanghainese has so many tones and patterns and to try to get equivalence between the Mandarin and the Shanghainese forms at least a phonetic tone level you don't have any items left after you eliminate for all of those differences but yeah I think it's an excellent point the effect of neighbourhood density is surely going to have some sort of influence on people's production we've abstracted away from that here definitely so what's your next project going to be how are you for the next one so my next the paper I'm writing now has to do with the data that I collected from these novice learners actually also the experienced ones after they had been in Korea for a year but had stopped speaking Korean so the question is whether the people who studied Korean for those six weeks and then went to their placement and taught English but were in an English environment in terms of requiring them to speak Korean they had a homestay that spoke English so they didn't have to speak Korean they went to school and all the co-teachers spoke English so they didn't have to speak Korean these people who basically stopped speaking Korean after this language program but they're still in the foreign language environment they're still hearing Korean around them do they go back to baseline and the answer is no they don't go back to baseline they do go back down in that direction but about 52 weeks after this initial set of measurements when they do the same task they're still significantly different and I think it's very interesting because they are not actively studying the language anymore one way that you could interpret these results is that there's some special thing that happens during learning you're attending to the linguistic input more conscientiously or you're just devoting more cognitive resources to that and maybe that's part of the reason why they're so influenced by it they're just so actively engaged with the L2 at this point well okay if we remove that element does that eliminate this L2 influence and I think the explanation for the fact that they stay shifted is kind of the phenomenon where if you're like sitting in a cafe and you're trying to ignore the people sitting next to you it's much harder to do that when they speak a language that you know and so there is this difference in processing of ambient language once you know that language so even though they're not really speaking Korean with people they still hear Korean and they know it enough now that they can't block out that exposure in the way that they used to and I think that's part of what is helping promote this shifted English because essentially Korean is staying active they can't turn it off anymore they know too much they know too much but they never even studied it actually I haven't collected data from that group so I only have data from this group that has some critical mass of Korean knowledge that might help make it hard for them to tune out this ambient exposure not very good so Korean is what's categorized as a category for language which is the most difficult for native English speakers to learn and they only learned it for six weeks if you want to become high elementary level actually one plus on the interagency language around table scale you need to study Korean for 64 weeks which is over twice as long as it would take to get that same level of proficiency in a romance language like Spanish or French so basically at least in terms of time Korean is more than twice as hard as these romance languages so they didn't study it for very long so reasonably it can be assumed that they weren't very good and just from personal experience just seeing them struggle when they're ordering in the community their command of the language is rudimentary so it's not really the case that they are being influenced because they're just so good it's definitely not questions not we can retire to the IOE so let's thank Charles again thanks so much