 Okay, it's good to see them. Most enthusiastic linguists here today. This is really great. And the Austronesian as Cam, well represented. One of the great pleasures of being a supervisor is when you have an opportunity to introduce one of your former students in a new role. Welcome back to several apps. As a postdoctoral researcher at Oxford, many of you know Charlotte, who graduated last year. I once spent too much time introducing her, but she's one of our spectacular products. Rave reviews on her dissertation. The Robins Prize for the Philological Society for an essay that was published in the proceedings, conference presentations, international superstar in the making. That's quite an introduction. I hope I can live up to it. So today we're going to hear about some of the work she's been doing on languages in Malaysia. And over to you, Charlotte. Okay, thank you. Firstly, as the room is very big and my voice is very quiet, if at any point I'm too low, please let me know, and I'll try and talk much louder. Okay, so today I'm going to talk about aspects of syntactic variation in Western Austronesian languages and why these aspects of variation might prompt us to move beyond classifying Austronesian languages in this part of the world as either Philippine type or Indonesian type. The aim of the talk is to demonstrate that a two-way typology is insufficient for two reasons. Firstly, because there are aspects of variation within the categories of Philippine type and Indonesian type. And secondly, because we find a series of languages that cannot be neatly classified as either one of the two classes. And I'm going to do this by looking at two phenomena that are known to differ between Philippine type and Indonesian type languages, mainly voice alternations and word order. So, given this range of variation, I'm going to argue that a more fruitful approach is to look at more fine-grained parameters of variation in order to be able to address theoretical debates and also propose historical changes that have been proposed to take place in Western Austronesian. Okay, so to start with, let's define a few key terms. So when we talk about Western Austronesian, we're not talking about a genetic subgroup of any sort. Rather, we're talking about a kind of geographical grouping of languages spoken in this part of Asia and Madagascar that share a typological characteristic, which is relatively rare cross-linguistically, namely symmetrical voice alternations. The best way, I think, to understand what a symmetrical voice alternation is, is to compare with asymmetrical alternations. Unfortunately, they don't all fit on the screen, but you can see them in your handout anyway. So voice alternations, as we know, represent different mappings between semantic arguments and syntactic functions, or various different ways that we can express notionally transitive events, perhaps given different semantic interpretations, or perhaps to fulfill different discourse functions. So in English, we have two ways of expressing transitive events. We have an active construction and a passive construction. In the active construction, it's the active semantic role that's mapped to subject, and in the passive construction, it's the undergoer. We all know this as fairly basic. In addition to this kind of alternation in which semantic role is mapped to syntactic function, there are also morphological and syntactic asymmetries. So the active is syntactically transitive. It has two core arguments, i and alue, that are both expressed as core nominal arguments. And the verb is morphologically unmarked for voice. In contrast, if we look at the passive construction, then we see that it's intransitive. It has only one core argument, and the actor now is expressed either as a prepositional byphrase or omitted altogether. And also we've got morphological marking, so we have the passive auxiliary was. So here you can see we go from transitive to intransitive and morphologically unmarked to morphologically marked, so we could think of these as being morphosyntactically asymmetrical. If we look in contrast at our western Austronesian languages, like Magyarys and Tagalog, we see again that we have alternative ways of mapping semantic roles to syntactic functions, the active voice, undergoer voice, and so on, constructions. But unlike our active passive alternations, these are morphosyntactically symmetrical. So take Magyarys, for example, which is in two. We have the active voice in which the actor is mapped a subject, and we have the undergoer voice in which the undergoer is mapped a subject, but both of these constructions take morphological marking on the verb. The root pokol takes the nasal assimilation form in the active voice, and it takes this e-prefix in the undergoer voice construction. So both of these are equally morphologically marked, and both of these look like transitive constructions. So both of them have two nominal arguments which are expressed as noun phrases rather than prepositional phrases, which would be how oblique is typically expressed in Magyarys and other languages in Indonesia. So we could think of this as being morphologically and syntactically symmetrical. And the same applies for Tagalog. Again, we can see that there are various different ways of expressing kind of the same notionally transitive event of buying. In each case, the verb is morphologically marked in a certain way. So the root bili can become bumili, binili, binilihan, and so on and so forth. And in each case, we seem to have a syntactically transitive construction that involves at least two arguments marked with core argument cases, ang and noun. So both of these alternations appear to be syntactically symmetrical, and this is kind of our key defining characteristic of Western Austronesian languages. However, as you will notice, there are several important differences, or really noticeable differences between Magyarys and Tagalog, although they both seem to be morphosyntactically symmetrical in the way that I just defined for you. First and foremost, as you can see probably more clearly on the handout than the screen, there are a much greater number of voice alternations in Tagalog. So in addition to the actor and the undergoer voice, we also see a locative voice construction in 3C, an instrumental voice in 3D, and a benefactor voice in the second 3B, which should be 3E. And there are other differences. So Magyarys is SVO word order, whereas Tagalog has verb initial order. And in Magyarys, the subject is marked by its initial position, whereas in Tagalog, the subject is marked by nominal case marking. So these and other differences have led people to talk about the alternations in 2 as being Indonesian type, and the alternations in 3 as being Philippine type. And this is our two-way system of classification. Okay, so what do these terms mean? Although the terms are highly prevalent in the literature, often they are not very clearly defined. So people tend to use these terms without really saying what they mean by them. As such, it's quite difficult to know if you had language X, would you classify that as being Philippine type or Indonesian type? So most models that try and make this definition a little bit more explicit tend to focus in on structural properties that seem to cluster around the languages the Philippines like Tagalog, and cluster around the languages in Indonesia like Magyarys. So one such definition is given on the handout in page 3, and it picks up a series of structural properties, and I've sort of summarized these and some others in table 1. So Indonesian type languages have symmetrical alternations just like Philippine type languages, but they differ in the sense that in addition to this they also have a passive construction that looks very similar to English. They also have applicative suffixes. If you wanted to make an instrument or a benefactor or a locative into the subject in Indonesian you would have to use an applicative construction combined with the undergo a voice morphology. And I think this should be the other way around. So this should be a no and this should be a yes. So they don't have micro roles like instrument etc. with their own voices. In contrast, Philippine languages do. They also have in addition to voice morphology, mood marking morphology and they had the case marking that we saw in Tagalog. And I've added to this list also the differences in word order that we saw. So kind of roughly speaking when people talk about Indonesian type and Philippine type this kind of seems to be what they're talking about series of different structural properties. And so these two labels are a nice starting point to reflect important differences in the languages. But the question that I want to ask for the rest of the talk is does this two-way typology or does this kind of little table here really reflect the full extent of variation that we find in this western Australian language area? And as I've said already I'm going to argue no and I'm going to argue no on the basis of variation that we find in voice system and in word order. Okay. So I think that if this kind of model was the correct one then it kind of makes the prediction that if we have a language with a multi-voice system then it should have the kind of properties that are associated with Philippine type languages and vice versa if we have a language that has a two-voice system then maybe we expect to find the sort of Indonesian type properties if it's a good model then that's kind of the predictions that we might make from it. However what I'd like to show you is that this doesn't really hold up in fact there's a lot of variations in the types. The number of voice alternations that we find in the different systems and also in the sorts of properties that are associated with them. So although those multi-voice systems of the type that we saw in Tagalog are prevalent in Formacen and Philippine type languages and have in fact been reconstructed for proto-Austronesian if we turn to page four in the handout we'll see that actually there are a large number of languages in the Philippines and in Taiwan that have actually a reduced voice system with only three alternations so you see examples in four and five from Kavalan which is a language spoken in Taiwan and Karazan Dusan which is a language spoken in northern Borneo but you can see that there's no real pattern to which voice is reduced in these three voice systems so Kavalan has in addition to actor and under-goer voice also an instrument voice whilst Karazan Dusan has in addition to actor and under-goer a benefactor voice and Bluss kind of says that in fact from all the languages that he's looked at there really isn't any dominant pattern in terms of which voices are lost so this might be kind of some aspect of variation that we might want to account for in some way what sort of I think more problematic for this model is that there are a series of languages particularly in Borneo and Sulawesi that have either multi voice or two voice systems without necessarily having these properties split in this way so for example if you look on page five you'll see the languages Lundaya from Borneo and Tondano from Sulawesi both of these languages have more than two voice alternations so they have a multi voice system but as you can see they differ from the Tagalog examples in interesting ways so they don't have case marking and in the case of Tondano the word order is SVO so we have multi voice systems that don't necessarily have these properties that we've said are typically Philippine type on the other hand we also have languages like Sabban which is also from northern central Borneo that has a two voice system and kind of looks if you look at the example in eight very similar to the Magerees alternation but unlike other Indonesian type languages it doesn't have applicative suffixes and doesn't seem to have a true passive construction and kind of last but not least we find some languages like Tukangbasi which seems to have a mixture of both so it kind of has a multi voice system and it has case marking in its verb initial but it also has applicative suffixes so really I think that there's kind of greater range of variations than we might expect if this really were the kind of model that we wanted to go for and so our two way distinction doesn't really allow us to make any kind of particularly interesting typological predictions so what's the alternative for them? so I'm going to propose that a more interesting way of looking at voice alternations in western Austronesian is to compare the different voices in terms of their morphological, syntactic, semantic and discourse properties and I argue that this will allow us to address an important theoretical debate namely the nature of alignment in western Austronesian languages and also a proposed historical change namely an alignment shift from Ergative in the Philippine type languages to Accusative in the Indonesian type languages so just a kind of brief reminder then of what we mean by alignment if you look at the bottom of page 6 you'll see that typically when we talk about Accusative alignment what we mean is that the actor of a transitive clause is treated in the same way as the single argument of an intransitive clause by contrast Ergative alignment is normally understood as the undergoer of a transitive clause being treated in the same way as the single argument of an intransitive clause so what the reason why symmetrical voice languages have been so tricky for this kind of understanding of alignment is that as we saw already when we looked at Magyarys and Tagalog we have languages that seem to have multiple different types of transitive clause that all look kind of at least more syntactically equally basic so the question becomes do we compare the intransitive clauses with the actor voice construction in which case we might say the alignment is accusative or do we compare the intransitive clauses with the undergoer voice construction in which case we'd probably say that the alignment is Ergative and I'm going to follow Craig in saying that the way we can decide which way to go and therefore what the alignment would be is to try and work out whether AV and UV which one of them is more basic basically and there are various ways that we can kind of understand what a voice or a transitive clause to be basic we can look at morphosyntactic clues like syntactic transitivity and morphological marketness but as we saw with Magyarys and Tagalog already in the case of symmetrical voice alternations these don't necessarily tell us very much so in these kind of context it falls to look at the levels of semantics and the levels of discourse and ask ourselves is one or other of these voices more semantically, prototypically transitive and is one or other of these voice constructions more prototypically transitive on a discourse level hopefully what we mean by this will become clear as we go through some examples ok so if we adopt this sort of approach then I argue that we can see some asymmetries between the voices at least in terms of semantic and discourse properties which would support the idea of a shift from Ergative to Accusative from Philippine to Indonesian type languages we can also identify a series of languages particularly in parts of Borneo that represent possible intermediate stages in this transition and therefore reinforce why it's so important not to just think of languages in Western Austronesian as belonging to one of two structural classes but to look in kind of more detail at the types of variation that we find ok so I said already that Western Asian languages have been kind of defined as being morphosyntatically symmetrical which would suggest that morphology and syntax kind of wouldn't give us any evidence at all as to whether languages are basic but it's not quite as simple as that in fact we do find that morphosyntactic asymmetries I won't have much to say about that here but I thought it was interesting anyway to note that there are some languages where we do see one or other of these voices being unmarked as opposed to the other one so for example in Pangaturan Summer the undergo of voice construction is unmarked as you can see in 10a whereas the active voice construction is marked so this is starting to look a little bit more like a prototypical anti-passive alternation in terms of the morphology and we also sometimes find syntactic asymmetries so in Kapampangan which is another Philippine type of language if you look at the cross-referencing system in 11 you'll see that if you look at the active voice in 11a then only the actor is cross-referenced in that particle yang whereas if you look at the undergo of voice construction in 11b then you'll see that actually that the particle n cross-references both the actor and the undergoer so this might suggest in some way that the UV construction is more transitive in that the cross-referencing particle has to make reference to both of these two arguments and that would then support UV as being the basic clause and hence alignment as ergative so that's Philippine type and if we look at Indonesian type languages we find kind of some syntactic evidence that might support the identification of AV in contrast as being more basic so one example would be secondary predicates in Balinese which can modify both the actor and undergoer in an AV construction as shown in 12a but only modify the undergoer in a UV construction as in 12b so that this might suggest then that AV is somehow more syntactically transitive than UV and that the alignment is accusative Riesberg looks at Balinese in quite a lot of detail and concludes that if we look at other syntactic tests apart from secondary predicates overall we would probably want to conclude that Balinese is really syntactically symmetrical and that the actor in a UV construction really is a core argument but it kind of just goes to show that maybe these morphosyntactic symmetries are a matter of degree rather than a kind of symmetrical or asymmetrical binary contrast ok so the more interesting levels for us I think in terms of comparison of symmetrical voice languages are semantics and discourse ok so when we talk about semantic transitivity what I kind of the sort of model that I'm adopting is the one suggested by Hopper and Thompson and they suggest that transitivity is not just a syntactic notion but rather can be associated with particular semantic interpretations and particular semantic properties some of which will be associated with prototypically transitive events and some of which will be associated with kind of lower transitivity or less prototypical events and they have a list of different parameters which they class as being high and low and I've kind of summarized these in table 2 according to the different semantic properties that are most commonly associated with active clauses passive clauses and anti-passive clauses so active clauses in semantic terms generally have 2 distinct participants they'll have an actor that could be seen as being highly agentif and volitional kind of willingly initiates the action they have an undergoer that's usually completely affected by the action of this event and the event is usually or usually entails punctual and telic action and these kind of semantic properties are generally associated with active clauses in semantic terms in contrast anti-passives will generally have a non-specific or unaffected undergoer and are often associated with ongoing or non-punctual action and finally passives are also kind of lower in transitivity on a semantic level and are associated with non-specific or perhaps non-agentif actors and stateif or resultative interpretations so if we kind of use these let me put them on the screen for you if we kind of use these definitions of kind of basicness in terms of semantic properties then we'll see that in Philippine type languages generally UV looks semantically like an active transitive clause whereas the AV construction looks a lot more like an anti-passive forward to Indonesian type languages we kind of find that actually AV looks like an active transitive clause and maybe in some languages UV is starting to look a lot more like a passive in semantic terms and what's kind of interesting is that we also find a series of languages that are somewhere in the middle so maybe sometimes AV has anti-passive semantics and sometimes it has active semantics so this again could be kind of a midpoint in our transition so some of the evidence for this is in the handout for example if we look at Sabuano we'll see that in 13 we're on page 9 now the undergo of voices associated with punctual action which remember we said was a high transitivity parameter whereas the active voices associated with non-punctual interpretations and moreover we may well know it's very common in Philippine languages for the AV to be used only when the undergoer is non-specific indefinite or non-presuppositional and again you can see some examples from Tagalog that show it's actually ungrammatical for us to find a definite undergoer in an AV construction and it's also at least strange if not completely ungrammatical for us to use an active voice construction in a context where the undergoer would be inherently affected so if we wanted to say something like Juan killed a dog we would have to use the undergoer voice construction I don't know why we would want to say that but if we did then we would have to use the undergoer voice construction rather than the active voice because that would be semantically odd so these kind of tend to suggest maybe that UV is basic in semantic terms and AV is less pretty typically transitive in Indonesian type languages in contrast we don't find this kind of restriction against definite undergoers in the AV clauses in fact if we looked at Balinese in 16 we would see that we can get active voice constructions when we have an event that inherently affects the undergoer like kicking the dog I feel like the dog is treated very badly in all of these linguistic examples but also you can see the dog can be modified by a definite suffix so here it's clearly possible to get a definite undergoer in the AV construction and therefore AV starts to look a lot more like an active type clause with these sorts of active semantics and there have been corpus studies to suggest that in some Indonesian type languages as I said the UV construction starts to look a lot more like a passive in that it tends to be used with non-specific actors or perhaps with no actor altogether and sometimes even has kind of stative interpretations so at least in some languages we start to see the kind of opposite pattern where we look like we have accusative alignment what's worth noting is that actually in Indonesian type languages more so than in Philippine type languages there's quite a lot of variations so there are some Indonesian type languages or at least languages with those sets of properties that we discussed where the active voice construction is still associated with some anti passive like properties such as ongoing or non punctual action and there are other languages where the UV construction is also still associated with some more active characteristics like dynamic action and so on so it depends a little bit on which language you're looking for and actually whilst this variation is kind of messy for our previous model and we wanted to call them more Indonesian type it's not such a problem if we think of these languages as somehow being on a scale or representing lots of different points in a transition from more prototypically ergative to more prototypically accusative so I think in actual fact it's quite nice that we find this variation because it seems to support the idea that these languages could be undergoing some sort of alignment shift at least at the level of semantics and we see this even more clearly if we look at languages in Borneo like Moro Nene where the active voice so in Moro Nene the undergo voice construction always appears like an active transitive clause it always has these high semantic properties but the active voice construction can sometimes appear like an anti passive and sometimes appear like the active ok so this might suggest that we want to look at languages in sort of terms of some sort of shift like I said really where we start with languages that kind of look ergative in semantic terms where the UV is associated with high transitivity and the AV is always associated with low transitivity and then we move through stages where we get different degrees of symmetricality towards the kind of stage at the end where we have an accusative alignment system with UV being associated now with low transitivity like a passive and AV being associated with high transitivity like an active ok so if we look then at differences in the discourse levels we'll find kind of more or less the same sorts of patterns and now we're thinking about discourse in terms of Givon's discourse topicality and he says that in a typical active clause in discourse terms we're going to have an actor that's highly topical and an undergoer that's also quite topical but less topical than the actor this is kind of what we would expect from our basic transitive clause and if instead we're dealing with something that's functionally a bit more like a passive or an anti-passive then we might expect only one of these arguments to be really high in topicality so in a passive we would expect the undergoer to be the most topical and the actor to be kind of not very topical and in an anti-passive it's the other way around so we would expect the actor to be topical and the undergoer to be less so and if we kind of take that as our basic understanding of discourse properties of voice alternations then again we see the same sort of patterns whereby in Philippine type languages it's UV that seems to have this sort of discourse structure and AV looks a bit like an anti-passive whereas in Indonesian type languages AV is the one that looks like the active clause and UV starts to look a bit more like a passive and then we find some languages that could represent possible intermediate stages like collaborative where both AV and UV seem to have or seem to be able to convey events in which both actor and undergoer are topical and you can see this if you look in table four which is a study that looks at the topicality of the different arguments and quantifies this using various metrics proposed by Guy Vaughn and then they're scaled and averaged and presented in the table so we can talk about how this is done in more detail if you want later on but the important kind of point to take from this I think is that if you look in Sabuano the UV construction looks kind of more prototypical or more basic then the AV construction the actor is highly topical and the undergoer less so whereas in the AV construction the actor is kind of still more topical than the undergoer but less so overall in Indonesian we kind of get something that looks a little bit different so in AV the actor is topical and the undergoer is less so but that kind of looks a bit more like an active clause and UV is starting to look a bit like a passive in these terms at least because the undergoer is more topical than the actor and then what's sort of interesting is if you look at Calabit in the middle of those two lines both AV and UV seem to have the discourse properties that we expect from an active, ergative clause so again we seem to have something where we've got something that looks kind of discourse ergative something that kind of looks discourse symmetrical and something that looks kind of discourse accusative and I think it's sort of worth pointing out that actually if we look just in terms of the structural properties that Calabit has we might be tempted to class Calabit as being Philippine type like Tagalog and many of the other languages that we saw but that would then miss some of these kind of important differences between Calabit and Cebuano that we see when we analyse kind of semantic and discourse differences as well so that was an awful lot of data and models and so on but the point that I really hope was clear from all of this is that distinguishing between the two typological groups Philippine type and Indonesian type doesn't really capture the surface level morphosyntactic variation and definitely doesn't really capture the more fundamental differences in alignment that we see if we kind of compare the different voice constructions on all of these levels so a better approach is to do just that than to look at how differences in morphology, syntax, semantics and discourse might have implications for how prototypically transitive the different voices are this approach lends support to the idea that western Australian languages have undergone a shift in alignment from ergative to accusative and provide some evidence of possible intermediate stages and it also suggests that we might want to consider alignment in general as being kind of more of a scale from prototypically ergative we might expect morphology, syntax, semantics and discourse all to point to one of the constructions as being basic to prototypically accusative where Ditto we might expect all of these levels to support AV as being basic via a series of kind of possible intermediate stages where the different levels of structure will reveal different levels of symmetry ok so now if I still have time and I haven't put everyone to sleep then I wanted to show you that if we look at word order we kind of are forced to come to the same conclusion really that Philippine type and Indonesian type or a two-way system isn't really sufficient to capture the kind of variation that we find and I'll show this by showing that there's important variation among both verb initial and SVO languages in terms of how flexible the word order is there are also seemingly languages where the basic word order differs depending on which voice construction we're looking at ok so firstly although it is kind of true that most Philippine type languages are verb initial just calling them verb initial or kind of calling them more Philippine type misses quite an important distinction between languages that are rigidly VOS and languages that allow a flexible VOS so sorry VOS VSO order like Tagalog so an example of a fixed word order language would be SEDIC in 19 and Tagalog is a flexible language in 20 so why this is kind of a non-trivial distinction is because it seems to correlate with certain other word order patterns as discussed by Aldridge namely the position in which adjuncts can be questioned so where we have fixed VOS word order languages like SEDIC and also Malagasy I believe we can only question adjuncts in C2 so they are always questioned following the verb however in Tagalog and other alternating languages it seems that it's possible to question adjuncts initially so they can be somehow extracted or appear in the initial position or however we want to analyse this theoretically so this has led to different possible theoretical accounts at least in Aldridge's work and there's therefore I think we can assume that there's potentially not a non-trivial distinction between them so there's more variation in verb initial languages than is accounted for if we just call the more verb initial is basically the point that I'm making we also get variation in Indonesian type languages we get variation in terms of how flexible languages are so some are really fixed and some allow almost all possible combinations of subject verb and object depending on things like information structure and we find that in Indonesian type languages sometimes the word order choice is affected by the voice construction so for example in language although Majuese has SVO regardless of which voice construction we're looking at in a language like Indonesian as you can see in 23 if you have an undergo voice construction which involves a first or second person agent then you tend to have a different construction in which you get the order subject that would be undergoer active verb or subject object verb as shown in 20 and sometimes you find that the possible orders vary depending on the voice construction so in Balinese for example you can get SVO and verb initial orders but you can only find VSO order in the active voice I don't really have a very good explanation for why you find these things I just thought it was worth noticing that SVO actually incorporates calling these all languages all SVO incorporates a kind of range of different possibilities that we find and what I think is quite interesting is that we find a series of languages particularly in Borneo like West Coast Bajau and also Calabit where the basic word order seems to differ according to the voice construction so in all of these languages it seems to be the case that in AV our active voice construction the basic word order is SVO whereas in UV the basic word order is verb initial it could be that this is if we looked at different genres it tends to be that most studies look at narratives but that kind of seems to be the pattern that's being reported in the literature and certainly was the pattern that I also found when I looked at it in Calabit narratives as summarised in table 5 that's quite interesting and we would kind of miss this fact I think if we again tried to say that Calabit was Philippine type and therefore were somehow not accounting for the fact that the basic word order appears to change depending on the voice construction how much time do I have left so again I want to propose to you an alternative possible way of looking at variation in word order as opposed to just saying this group of languages is verb initial and this group of languages is SVO and I want to propose that the way to do it is to look at variation in the orders that we find the possible orders and the degree of word order flexibility and also to look at the kind of choices that affect which word order we choose so not just to say this language is SVO but to say in this context we might use this construction and so on and I think this is an interesting way of looking at word order in Western Austronesian because it allows us to address another proposed historical change which is namely the re-analysis of a kind of topicalisation construction to the basic order of grammatical functions so the idea is you can get SVO as a kind of marked construction in Philippine type languages but when you get this order it will indicate I guess the topic by the topicalisation idea but what it seems to me is that it really represents particularly newsworthy information so that might be focus or that might be a new topic or contrast or something like that so it seems to be that in Philippine type languages this construction or this order is restricted to those contexts where the subject really has a particular information structure role whereas in Indonesian type languages SVO is just the basic order of our grammatical functions and what's kind of interesting if you sort of look at these types of factors it seems that in languages like Calabit where the word order is affected by the voice construction SVO kind seems to be sort of a topicalisation construction in these UV contexts that look a bit more Philippine type whereas SVO seems to be the basic word order in AV context where it looks a bit more Indonesian type so perhaps we can just have a little look at some of the data that supports this and then we'll conclude all of this very heavy data for you okay so if you look in 25 you can see an example from CDIC which remember was our kind of what I said fixed order meaning to say that you can't find VSO in CDIC you can get a kind of SVO construction which is possibly also to be analysed as a type of class in context where that S represents particularly newsworthy information so for example in 25B if we said who drank the wine then we can get Pawan this kind of nominal particle and then drink wine so that would be a kind of SVO order but there Pawan really represents the focus information in the sense that it's the answer to the question and therefore the kind of newsworthy thing so this is really these types of order is really restricted to that context in Indonesian type languages as we saw already SVO is just the basic word order and we see that in terms of discourse frequency in various other factors and actually in Indonesian it kind of goes the other way so when we get verb initial orders these have a kind of marked pragmatic function so perhaps they mark predicate focus or sort of contrastive reading on the predicate or something like that as shown in the example 26 which comes from SVO can call passive Jakarta in Indonesian so that's kind of two ends of the spectrum if you like when we look at collab it we can see that for UV which remember the basic order in UV was verb initial in UV if we have a verb initial order this tends to be context where both actor and undergoer are topical and SVO is reserved for context in which the undergoer really has a kind of newsworthy role to play so an example of that is in 27 where again the undergoer represents the answer to the question so it's the kind of focus information in that context and in 28 you see an example of VOS where if you read through it all you'll see that both so the kind of example sentence of VOS is 28 D and it's kind of verb actor is N and then undergoer is N and both of the actor the N which is he and the undergoer fruit have been mentioned in the previous discourse so they're kind of both topical there particularly newsworthy in the sort of important information in terms of progressing the story in this case is the action and what he does to the fruit and in contrast if we look at AV constructions then SVO is by far the most frequent word order and it doesn't necessarily seem to correlate with actors that are particularly newsworthy like we saw in CDIC and I think you could see this if you looked at the example 29 where we have a question how did he get the fruit down there if there's three baskets so this is the question how did he get the fruit down and the answer is how do I say it he dropped the fruit to the ground with SVO word order he dropped fruit and I think you could say that the actor isn't particularly newsworthy here because the answer to the question would be how so actually the sort of newsworthy information in this context is probably provided by the verb rather than the actor and the undergoer which are both topical as they're given in the kind of context of the question so that's just one example but in general SVO doesn't seem to correlate particularly with an actor that's particularly newsworthy like in CDIC and actually what we find in the AV construction is that SVOS that seems to have this marked pragmatic function like in Indonesian so the example that you want to look at here is F 30 F where it says eat fruit only they and the whole discourse is going on about it's collected using a pair story and it's talking about two boys who've kind of the man has watched the two boys walking past with his fruit and he's thinking oh did they steal my fruit? No they're only eating the fruit so it's kind of a contrast if reading then on the on the predicate there so I think what you can kind of see from this is that the UV construction in collab it appears to be much more Philippine type in terms of it's word order there the VOS construction seems to be basic and SVO is reserved for sort of pragmatically marked constructions where the under go is particularly newsworthy whereas in AV what you seem to find is a situation that's much more Indonesian type in the sense that SVO is the basic order and it's the verb initial order that's kind of pragmatically marked in this case okay so to summarise then I think this suggests that the reanalysis of SVO as the basic word order might actually begin with AV rather than being a kind of across the board thing and the voice may be therefore important in this historical change as well it's possible that SVO is preferred in AV because this is an order that fits with universal word order tendencies to put A to put the act to first and B to keep the verb and the under go of the predicate together whereas if we had SVO in the under go of voice that would actually be violating both of these tendencies because it would put the under go of first and split the predicate and perhaps the changes are related in the sense that the change from antipassive to active increases the discourse frequency and that kind of leads to a change or reanalysis from the pragmatically marked order to the basic order of grammatical functions perhaps it's the other way round perhaps reanalyzing active verb under go as kind of basic word order somehow leads to verb and under go are being considered a unit and that therefore triggers the reanalysis of antipassive to active I don't really know but in any way in any case I think the point is that these kind of changes seem to be related and looking at both of the phenomena in more detail kind of reveals the sort of things that you know the interconnections between these different changes that we find and definitely I think reinforces the idea that a two way classification just doesn't cut it as a way of capturing the sort of variation okay so concluding then the two way typology of Philippine type of Philippine type versus Indonesian type I hope that I've shown you that it's inadequate as a means of capturing the full extent of variation in Austronesian for two reasons firstly because languages with typical Philippine type and Indonesian type properties are actually subject to quite a significant amount of internal variation and secondly because we find a number of languages particularly in transitional regions in Borneo and Sulawesi that differ in a kind of non superficial manner from both Philippine type and Indonesian type languages and therefore are not neatly categorized in either one of these two classes so a better approach to categorization would be to look at the different parameters of variation and draw from a wide range of languages rather than simply assigning languages to one of these two groups. This would allow us to explore the interrelationships between things like word order, information structure and voice and contribute to a better understanding of the historical changes that have taken place and also allow us to address theoretical debates like the alignment debate in a more typologically informed manner. Thank you Randomly well so I mainly looked at languages in Borneo so most of the languages that seem to have these mixed properties that I identified were in that area and I think we would probably if we go back to the map we might expect that to be kind of a reflection of the situation as it really is because the kind of more conservative languages are thought to, I mean Austronesian is thought to have originated in Taiwan and the more conservative languages are in the Philippines and in Taiwan and the more innovative languages are in Indonesia so you might expect to find transitional languages in the areas that are kind of geographically between them but of course since I only really looked at languages in Borneo it's completely possible that we would find other languages that have developed in fact it's probably quite likely that we would find other languages that would you know, differ in certain ways from this prototypical Philippine type or prototypical Indonesian and other parts of the Western Austronesian group Yeah If I can just comment, Lauren I think we don't have the micro level information that would help with that I mean I've published an article showing in SASA different dialects of the same language of different alignment systems so I think we really need much more detail about fine grained analysis I actually remember years ago I was getting very angry about people talking about Philippine type because everyone assuming that all the Philippine languages were like Tagalog absolutely, this is Tagalog hegemony and we should deny this there's a specialist of languages, other languages in the Philippines he was very much aware of the variation between Sebuano Ilocano, Kapapangan and Tagalog and I mean taking this sort of saying okay this two-way typology is no good and what we need to do is this but I don't have a very good answer of how these might all fit together I mean I don't know if this really solves the problem and as you say the next step really would be to collect much more data on these sorts of aspects of variation from a kind of wider range of languages so that you could really see what kind of possibilities are possible so in collabat for example morphosyntactically it looks more or less symmetrical semantically the UV construction is normally associated with kind of high transitivity and the AV construction is normally associated with anti-passive like characteristics but can be also an active I mean can have active a bit like the more in any data that we saw and then on the discourse level we saw that it kind of looked symmetrical with AV so that you know depending on which level you look at with this type of approach you get a slightly different answer as to which one is more basic so you might expect that if you looked at a wider range of languages you could find different types of asymmetries on different levels depending on where they fit in this kind of change I think a similar story could be told about Benton where people always talked about symmetrical and asymmetrical and it looks as a work show so that's actually much more complex Phil do you have a valid solid genetic mapping for these areas if so what kind of match do you get between genetic affiliation and some of the phenomena you talk about I think the answer is no but that isn't a very good no I don't have a gas yeah I think that in terms of subgrouping people have kind of worked out that all of these languages in sort of the darker orange are Western Malay Polynesian that kind of are not quite sure what the subgroups people disagree on what the subgroups would be beyond that and because a lot of the languages just haven't been described in much detail I think probably that I think people think that there would be a genetic subgroup that would include quite a lot of the languages in the Philippines and they might tend to be more sort of prototypically discourse semantic ergative and I think a lot of the Indonesian type languages probably belong to a different subgroup but as for the languages in Borneo that's no one really knows these transitional properties are these mainly from North Borneo languages and other ones that are in Borneo with the A&KM as well it's also seen a bit are you as well I don't know I think people think that those languages are more kind of similar to Indonesian type but I think like Sabban they don't really have all of these so they don't seem to have passives and applicatives like we find in Indonesian type languages they just don't have the multi-voice system and I think a voice in I'm not sure about Badiu but in some of the languages like Kenya and Qayyan that's something quite different altogether question for you another difference you didn't talk about between the so-called two-way topology is that in Philippine languages the morphology actually messes up aspect and voice I mean this is usually passed over and people don't talk about it the imperfective you get certain patterns and the perfective you get different patterns that's not there in Indonesian languages what's the story? why? is it connected to anything you say here because Hopper and Thompson talk about aspect as being relevant for transit too so I don't know have you built that into your story so far? yeah no it almost certainly is relevant exactly how I don't know I know that for example with Togalog some people analyse this in infix which I stupidly didn't actually lay out for you in the morphological gloss but anyway the undergoer voice in Togalog is some people think that this is marked through an in infix into the root so we took the root Billy and we made it Binilli so some people treat that as being a combination of undergoer voice plus perfective aspect and other people argue that actually that is just the perfective aspect marker and the undergoer voice is kind of unmarked morphologically and the kind of main argument or one of the arguments for that is if you look at the other voices apart from active voice so in 3CG and the second B that should be E then all of them have an in infix in them so kind of maybe you would get a slightly I don't know a simpler analysis if you would treat just the in as marking perfective aspect and then undergoer voice would just be unmarked so I think some people I think category in fact analysed Togalog as having kind of a split ergative system because he said that the undergoer voice is unmarked in the perfective context whereas the active voice seems to be unmarked if you look at Irealis or whatever the opposite of perfective would be because these transitional languages you were talking about like Calabat and so on they don't mark perfective and imperfective in the voice Calabat does so Calabat has something that probably comes from the in infix but has a whole load of different alimorphs so it's sometimes difficult to work out how it relates to in but anyway this seems to mark both undergoer voice and perfective aspect and you get a separate form that will mark undergoer voice and Irealis but that's not used very productively anymore I don't know if it's possible that something that was just an aspect marker over time becomes analysed as a voice marker and then loses its aspect or function but the nasal voice prefix that marks the active voice in Calabat so it's unclear to me whether it has some I think it probably doesn't have this basically it looks like in Calabat the active voice construction is starting to look a lot like what we find in Indonesian type languages in all sorts of properties and the undergoer voice kind of maintains some of this what we know from the Philippines I don't know in Calabat that's a good question I don't know I don't know I have to look I think that might be another parameter that along with the aspect or one might be I don't know I don't know I don't know along with the aspect or one might be other parameters add to the mix can I ask you about the topology of the voice system a bit more because I'm curious the reason I'm asking is because in version construction and Bantu look similar to the version construction of the direct object but in effect it's not really I'm curious whether there is a distinction between these different voice systems in terms of the predicates they combine with in terms of the information structure they can do. Later on in your talk, I think that's a question of the language-as-law of the voice system. You focus on the active voice and the undergoer voice. But the other three voices sort of don't play a lot of part. Is that because the language is not active or is that because they don't play part of these wider... Or maybe because I just didn't look at them. So in general, I think that these other constructions are much less frequent than the first two. So I can't say really much apart from about Calabit, but Calabit has three voices. It has the active, the undergoer, and then an instrumental voice. And the instrumental voice occurs very infrequently in sort of naturalistic texts. And usually it occurs in... Because one of the reasons you might want to alternate between which argument is mapped to subject is because you can only form a relative clause on subjects and there are a whole load of properties that only apply to subjects in these languages. So usually the instrumental voice is used when you want to make a relative... In a relative clause on an instrument. That tends to be when it... So it's quite a marked construction anyway. And I think within the relative clause, because it's a subordinate clause, there are certain constraints on the word order possibilities and things like that. So I mean, I didn't really look at them in terms of word order and other things like that. Could be that you might find something interesting if you did. Remind me the question. I feel like I didn't answer it. I was curious whether the benefactors slightly... They odd one out. And that's just because of my background. Of course, the other four are quite frequent and the properties are quite similar. I guess in texts they're less frequent, but cross-languistic and cross-bonded. They are frequent, but the benefactors... Oh right, you mean in terms of these sorts of inversions, yeah? And just that either within a particular language or comparatively, whether they come at a slightly different... Yeah, it could be. So collab it doesn't have a benefactor voice, so... Yeah, no, I don't know. They do... I mean, so they seem to be kind of di-transitive. I mean, I was calling these symmetrically equal and what I probably really meant by that was that the act of voice and the under-girl voice kind of looked symmetrical. These other voices appear to... They appear a little bit like di-transitive constructions in that you seem to have then three arguments that are marked with these core cases or however it works in the particular language. And what's interesting, at least for collab it, is that you don't really find di-transitive constructions in other cases. So if you used the... What's a good example of this? You would have to say, I gave something to someone. You couldn't be able to say, I... You wouldn't be able to say, I gave someone something with three kind of core arguments. The only case when you get that is with the instrumental voice construction. And there are applicatives? It's not a different matter. Well, so some people treat these... If you kind of adopt a structurally ergative analysis of these languages, which many people do, then they treat locative voice, instrumental voice and benefactor voice as being kind of applicatives. What's interesting is that there are applicatives that only apply in the under-girl voice. They can't be used in active voice constructions. Whereas the applicatives that we find in Indonesian and that kind of look a lot more like applicatives in Bantu, for example, you can get them in both with active voice and under-girl voice. So they're a kind of little bit applicative-y, but at least how I analyse them, they map this kind of instrumental locative or whatever to subject rather than to a direct object, which is, I guess, what you would think of as an applicative. So they're sort of half applicatives, maybe. Interesting, yes. I think we should thank Charlotte for deconstructing two ways. Thank you.