 Ossetic and dark one that are related to each other, and he does descriptive work on those languages. Gremors has also prepared an Ossetic national focus, which is online in the present visa work, and also has a bonus funded project called the documentation of the Ossetic national focus. So that's a plethora of documentation, but he also does a lot of work on the language. He was on the same day with LLG, and he's been working with Sympax and Symantec for the Ossetic national focus. Thank you for this introduction. I'm happy to be speaking here, and the talk today will be about limiting the domains of coordination and subordination, trying to provide formal definitions for both concepts, which will capture the variation of constructions that we encounter in languages of the world, and the mismatches that occur in cases where we can't really classify the constructions as clearly subordinating or clearly co-ordinating. So the notions of coordination and subordination, so traditionally they are defined as a binary distinction between two types of close linkage, and also of course linking in other areas of grammar such as the subordination of a noun-phrased verb, but mostly these concepts in what we shall discuss will mostly talk about close coordination and subordination. So subordination is a construction, traditionally believed that it's a construction where the subordinate close is located in constituent structured terms within the main clause, and is dependent on it in some way, either because it contains an overt subordination marker, such as in this case, or if it's a non-finite close, such as with the participle's converters and infinitives, and coordination is traditionally understood as a construction where we have two or more clauses which have identical status with respect to each other. None of them is embedded within the other, but rather they all form a constituent with possibly a conjunction which does not belong to any of the conjuncts, but stands sort of in between them or is attached to each of them symmetrically or is attached to one of them, but importantly it is not located within this conjunct in the structural sense, it does not mark dependency in any way. And so this means that generally this concept is an amalgamation of several ideas which is based on a general and vague, somewhat vague concept of symmetry versus asymmetry. And a good example, for example, of the traditional approach is Christel's definition in his annexing of linguistic terms, where he says that subordination is a term used in grammatical analysis to refer to the process or result of linking linguistic units so that they have different syntactic status, one being dependent upon the other and usually constituent of the other. So he says it's distinguished from coordinate linkage in that in coordinate linkage they are equivalent. Of course, in traditional grammar such terms as being equivalent or being dependent are not strictly defined, but in most cases it is quite clear what we are talking about. So we have clauses which have some sort of dependency, some clauses which don't have this dependency. And I should mention that one of the first attempts to provide clear criteria to determine to choose between subordination coordination is the classic work by Peshkovsky on Russian syntax, but generally before the second half of the 20th century there have been more or less an assumption that these notions are basic and don't need to be discussed in any way. But why do we need to discuss this and why do problems arise? There are several constructions even in well-known languages such as English which challenge the traditional binary distinction between the two clauses. So the first construction is so-called left subordinating and known from the works by Kurikova and Jakindov on this construction. So in this construction one example of which is found in one. So this is you drink one more can of beer and I'm leaving. It uses the coordinating conjunction and but with a meaning that is conditional. So in this sense it can be analyzed as being equivalent to a subordinating construction with the subordinator if. So if you drink one more can of beer I'm leaving. So in this sense the construction is not symmetrical. One of the clauses is clearly dependent on the other in the sense of being its condition. A second example is so-called pseudo coordination when a conjunction like and is used for linking a complement clause generally to a main clause like in the try and construction in English. Try and disprove my point. So essentially disprove is a complement clause dependent on the verb try even though the marker of course language is the conjunction and traditionally as well as being coordinated. A third example is so-called commutative coordination in languages like Russian and other Slavic languages and also Yiddish where preposition with is used in the same sense generally in the same sense as a conjunction which links to coordinate noun phrases. So for example in three from Russian means Peter and Vice came to me. And this verb agreement is plural which means that they denote a plural entity but the noun phrase itself clearly has a head in terms of case marking because Peter is nominative which corresponds to the subject position whereas Vice is in the instrumental case which is sub-criticalized for by the preposition with so in the sense at least of syntactic dependency this construction is sub-ordinating. Peter with Vice with Vice being in an oblique case but in the terms of semantics in terms of dependence of the verb in terms of verb agreement this construction looks just like traditional coordination. We have in many languages converged constructions where a form which is morphologically dependent which contains which lacks certain finite categories such as agreement morphemes or certain tense and aspect morphemes functions in the very same way as we find with coordinating constructions in more well-known in more familiar languages such as four from Shiri Dargwa I finished my work and Ali came just a normal sequence of clauses but the only way to express this meaning in Shiri is to use a convert clause for the first conjunct of this construction so literally it means something like me having done the work Ali came to me whereas formally the form was dependent but semantically it is indistinguishable from what we call coordination in more familiar languages. So how can we define these constructions and how can we say, okay, so how can we say that this and that construction belongs more to coordination, more to subordination can we come up with criteria which can help us distinguish them and of course the basic idea that comes up especially when we are talking about one single language is that we can just look at how canonical constructions behave which we clearly know that they are coordinating and subordinating and then try to see whether other constructions whose status is not as clear fall into one of the groups or the other group and the tests which have been proposed in the literature are for example in 5 and 6 constituencies tests on linear order and 5-6 illustrates central embedding so generally if a language generally allows embedding one clause into the other it is the subordinate clause which can be embedded and subordinate clauses cannot so 5 is an example from Kouka with a central embedding of a verbal clause in English so such techniques when coupled with a suitable interface could provide a VI person with a feedback mechanism now in 6 we try to do something similar with a coordinate construction and this fails in any possible configuration because conjuncts of coordinate constructions cannot be embedded one into the other another constraint which is well known since Ross's dissertation on constraints and variables in syntax is the coordinate structure constraint and in the general formulation it is stated that you cannot you can only extract elements you cannot extract elements from just one or two of several coordinate conjuncts of coordinate structures so in 7 we see that we can extract the direct object from the main clause in the presence of a verbal clause, a one clause who did Ali see when Russell came but we cannot do the same thing if we use a coordinating construction so we cannot say who did Ali see and Russell came because in this case we have a coordination and the interagency is extracted only from one of the two clauses and in pair with the coordinate structure constraint came its inverse side, the possibility of it across the board extraction and unlike extraction from single clauses which is allowed with subordination is not allowed with coordination this works the other way so with coordinate constructions we can extract the same element from both of the two clauses and with subordination constructions we cannot do that so 9 illustrates that we can extract who in this case from both the direct object from both clauses who did Ali see and Russell here but we cannot do this when we have an adverbal clause and we cannot extract the conjunct from both clauses both the main and the adverbal clause so who did Ali see when Russell heard is as far as I know ungrammatically in English finally another test which I would like to point out is the possibility of cataphera so generally when in the coordinate structure only and the four correlations between the two clauses are allowed meaning that the antecedent must precede the pronominal which refers to it but in subordination if the subordination clause precedes the main clause the pronoun can stand before its antecedent so in 11 when he arrived Ali went to Russell he can't be interpreted as he being co-referent with Ali but 12 does not have the same interpretation he arrived and Ali went to Russell generally cannot mean that he equals Ali so there's a lot of these tests and the work in a particular language they work rather well to delimit the domains of two constructions coordination sense of ordinations and then we can try to think about what the rest of the unclear field between them falls into and one of the well-known proposals about this is for invariance idea that we can have a third notion between coordination and subordination which they call aptly enough core subordination so we sort of have something in between coordination and subordination and they cast this idea into an opposition between two features dependency and embedding so a coordinate construction is neither dependent nor embedded a subordinating construction is both dependent and embedded whereas core subordination in their approach is dependent but not embedded so it's sort of a detached subordinate close as it were and the main criterion that they use in their work is scope of executioner operators so in 13th from Amelie from language Amelie there is question marker 4 at the end of the sentence which takes scope which must obligatorily take scope over both clauses so peak having run out men kill them question even though peak run out is marked as subordinating the question scopes over both clauses so the sentence means did the peak run out and did the man kill it so it doesn't mean did the man kill the peak when it ran out for example which would happen if it were a subordinating construction so the question operator scopes over both clauses and does not distinguish between whether one of them is dependent or main close and in their approach coordination has optional wide scope whereas subordination has only narrow scope on the main close so that's the difference between the two constructions according to Foli and Van Weylen so this works out well for the languages in question but the idea is that the same tests at least more or less should apply across languages and should generally use the same results across languages so we should find the same three classes in all languages and we should find that these classes generally behave in the same way according to the same tests and this unfortunately is not as good as well fulfilled as we might think and Walter Zalbiko in his 2010 paper has tried to apply generally the test and coordination subordination that I used in literature to a test sample of 24 languages to try and find whether there is any prototype there is any cluster of tests that would delimit any kind of close-linkage types and it turns out that he did find evidence for certain prototype of subordination as he says but there is no evidence for a cross linguistically stable and consistent set of tests which would delimit coordination or course abordination so essentially it means that only one notion of the three notions proposed by Foli and Van Weylen makes any sense cross linguistically whereas the notion course abordination essentially is some sort of just a common, just a dumping area for anything that doesn't fit into our preconceived standards for either subordination or coordination as we generally understand didn't do it from European languages but maybe it's not the distinction itself but the method of defining the notions via tests that is float so there's another approach which takes a more lenient stance this is approached by Christian Leman which says essentially that there is a continuum between coordination and subordination it's based on several scales which mark the different degrees of dependency of certain closes on the others and so generally it is assumed that some constructions fall into one of the sides of all these scales so there are certain constructions which are dependent according to both morphological dependency and syntactic impossibility of embedding and other constructions which are not dependent and not embedded and so on but the problem with this approach is that the same criticism applied to it that is that there is no fixed set of tests and no clear procedure for situating a given construction in each of the scales there's a general understanding that there is a client from one type of close juncture to the other but there is no particular procedure that we can apply systematically and determine whether a construction is here or there in the continuum so in general this is at best a general idea that can lead us to getting new facts but this doesn't really predict anything and essentially it implies the rejection of a cross-legged distinction between coordination and subordination so it implies that languages can vary in unlimited terms with respect to these things and they just cannot possibly we cannot say that there are any limits on this variation so maybe we should abandon the notions of coordination and subordination at all there are maybe just traditional notions which do not have any place in the modern theory of grammar this is what in fact Hyman and Thompson have proposed and many other linguists follow their track especially those working at functional frameworks so for example this quote shows that subordination is not a grammatical category at all this is simply a concept in which we rationalize how western educated intuitions which renders the completely circular so we first define certain things in our languages based on some sort of tests and then we try to find the same things in other languages but there is no particular reason to believe that these notions that we find in our languages will automatically transfer to other languages but the problem with this idea that we should abandon the notions like with any other kind of this kind of radical proposal but everything is that time and again across languages we find that similar clusterizations of constructions in the general sense so we don't have the same test given the same results but we do find the tendency for so coordinating like meanings to be expressed by more symmetrical constructions to have more symmetrical properties whereas subordination like ideas are expressed by more embedded and dependent constructions which have more subordination like properties so if we want to abandon these notions as Hyman and Thompson proposed we would have to propose something else except for these notions and simply abandoning them and saying that there is unlimited variation would be a step back because in this case we would just throw away all the data that has been gathered in work on this issue so what I think is a more promising approach and in fact what the approach that I will develop attempt to develop in this talk is the multilevel approach to coordination and subordination which can been proposed in two seminal papers one of them is by Kudykov and Jacobov 1997 dedicated to left subordinating and in English which we just had an illustration of and you are sensing 2002 which is dedicated to several other constructions in languages of the world and the idea of these two papers is that the notions of coordination and subordination remain binary but they can be defined separately for two levels of language syntax and semantics so we can have syntactic coordination subordination semantic coordination subordination they can coincide they usually coincide but sometimes they don't and sometimes we can have mismatches and Kudykov and Jacobov discuss a mismatch where a construction is syntactically co-ordinating but semantically subordinating whereas you are sensing they describe a separate kind of mismatches where a construction isn't syntactically subordinating and semantically co-ordinating and crucially different tests apply different levels so we have a set of tests which distinguishes syntactic coordination subordination a set of tests which distinguishes semantic coordination and subordination and even though there can be conflicts between these two groups of tests there can be no conflicts within one and the same group of tests and so why do they argue that there should be distinction between syntax and semantics now Kudykov and Jacobov based on English left-over meaning and show some properties which it does not share with normal English co-ordinating answers for example there is no possibility of right note raising so while you can have something like big glue found out about and big glue put out a contact on the guy who stole some wood from the gang you can't do this in a conditional meaning so you can't say something like big glue finds out about and big glue puts out a contact on the guy who stole some wood from the gang this is impossible if you want to have a conditional reading of this construction the second effect that they use is the impossibility of gaping so you cannot gap the verb in a sentence like big glue steals one more k-radio and little glue the hubcaps you can do this in a normal co-ordinating reading but you can't do this in a conditional reading then there is the possibility of binding reflexive in the first clause from the second clause another picture of himself appears in the newspaper as Susan Sting's John will definitely go out and get a lawyer finally there is no across the board extraction from left subordinating end so you cannot say this is the thief who you just point out and we arrest on the spot in the sense that if you point him out we arrest him on the spot so from these tests we can be led to believe that left subordinating end is simply a subordinating conjunction which just happens to be corresponding co-ordinating conjunction but the problem is that it is quite exceptional for English subordinating conjunctions if it were one of them because first of all it has an exceptional position following the subordinate clause rather than preceding it it is the only of its kind in English we would assume so and it doesn't allow we are not allowed to reverse the order of the clauses and the first clause after the second we cannot say even though with normal conditional clauses we can post post the if clause after the main clause in left subordinating and we cannot do that and their solution is to say that this construction essentially has co-ordinating syntax which explains the last two properties but subordinating semantics and according to Kudykov and Jackendorf all of these four tests and some other tests that they use are all dependent on the semantics now yours and said they have a separate kind of constructions and one of the examples of the constructions they look at is a Japanese te subordination or co-ordination that's the clause chaining so Maki went to Osaka yesterday and Hiro will return from Kyoto tomorrow and in this sentence the first clause is headed by a convertible form it te and the main clause is headed by a finite form in u i ku and so according to you are sensitive this is the fact that this the first form is non-finance it cannot be used in an independent sentence it means that it is dependent on the main clause subordinate in the syntactic sense but it is co-ordinating according to them because it has no co-ordinate structure it does not obey the co-ordinate structure constraint it allows it does not allow catephora and it is not found in presupposition which according to you are sensitive is a characteristic feature of subordination so they say that this construction is on the contrary semantically co-ordinating but syntactically subordinating and so what we are synthetic proposes is a general typology of the construction that we find the languages of the world according to these two distinctions so we have co-ordination subordination which clearly align syntax and semantics and we have pseudo-coordination which is a construction which has co-ordinating syntax but subordinating semantics and we have pseudo-subordination which has subordinating syntax but co-ordinating semantics now what will I do here is first of all to attempt to apply the multilevel approach to aesthetics to the co-ordination which is a general repeating of my recent paper in journal of linguistics where I have this in more detail then I will demonstrate that we need three levels instead of two if we want to use the multilevel approach and I will also demonstrate that what critical interjection often the U.S. says that co-cementics is actually syntactic level of representation and I will formulate the findings of LFG and finally show how the same approach allows generalizing some effects similar effects in Russian so first some general info on aesthetic here you have a map of the region where it's spoken in yellow and it is spoken in the North Caucasus mostly in North Ossetia in the center of the North Caucasus between the Black and the Caspian Seas in the West it is bordered by West Caucasian and Turkic languages in the East by East Caucasian North Dagestanian languages Chechen and Ingush and finally to the South it is bordered by South Caucasian languages like Georgian so it has a very peculiar position but genetically it is not part of the same linguistic area it is an Iranian language which ultimately distants from a group of from a Ski-to-Sarmatian tribe the Alans which moved to the Caucasian to the border to the border with the Caucasian mountains in the first centuries A.D. and stayed there and mostly had contact with the local populations which has determined which has led to the substantial differences between the grammar of Ossetia and the grammars of other more than Iranian languages which can have different explanations sometimes it is explained by language contact sometimes simply by development but in general under half a million speakers it is an SOV language which completely accuses the alignment something which is not typical for this area which has mostly at least some traces of ergotivity in the languages and it has an inflection of ethnophology and agglutinating mostly innovative ethnophology now let's first take a look at the general way in which subordination and coordination are carried out in Ossetia for subordination the main strategy when I heard your voice I became happy which literally reads like something like when I heard your voice then I became happy so the subordinate clause is found at the left periphery of the main clause contains subordinator and it is obligatory to resume in the main clause by demonstrative element which is metaphorically connected to the subordinate clause and this pattern is used for not only verbal clauses but also for which boy I saw he's my friend is the way of forming relative clauses in Ossetia and also even for complement clauses which are formed in ways like I know that you came that I know so the second that is a demonstrative so this is the main way of forming subordinate clauses in Ossetia and as for coordination Ossetia utilizes the standard average European as it were way of forming coordinate clauses that is simply using a subordinate clause like in 12 so in 20 sorry I was I'm staggering like a drunkard but I'm going to you so here from the national corpus like a drunk man I'm staggering but to the I'm going now the conjunction will be in focus of our investigation today is the conjunction and and on this slide you can see that in general it is used as a purely standard like in 21 Avan's sister and Zaur's wife came to me it can combine adjective for adjective phrases like in 22 my good and beautiful wife like in 22 finally it can combine whole sentence whole clauses like in 23 Achsartak cut off the head of the giant and enters the yard of the Donbett that is a fragment of an art epics mythical characters and in all these cases you have a between the two conjuncts and you have the conjunctions staying right between each of them and historically too the conjunction goes back to Proto-Iranian Hamma it is it is a cognate to Persian Ham so it's a purely it means also in Persian and probably meant also in the Proto language so in its origin this is also purely standard coordinating conjunction with nothing suspicious about it but unfortunately or maybe fortunately it's used also in some uses which are not as typical for coordination as we might think which I call through the coordination following you are Sanseidic due to the fact that they at least seem to have subordinating semantics in spite of having coordinate syntax so 24 illustrates a construction which I call causal through the coordination and in this construction the combination of the data form of the distal demonstrative pronoun women causal meaning so the sentence in 24 I do not allow you to go to work because you are ill literally means something like I do not allow you to go to work for that and you are ill so here the causal meaning arises from the combination of the data form the demonstrative and the conjunction end the second construction which I will be looking at is so called complement through the coordination where the conjunction end is used to introduce a complement close like in 25 do you think that we will part ways like this this is a translation and literally the sentence means do you think end we will part ways like this so this is somewhat similar to the English Try End with a difference that here you combine not just heads but whole clauses with their own subjects and their own structure and importantly in the causal construction the polynomial element the relative element is not part of the same word that is the construction but it can float freely in the main clause it's categorically refers to the second clause so for example 26 if you wanted to focus the semantically subordinating clause we would put the polynomial element to the preverbal position but the conjunction will stay between the two clauses so it is because I cannot forgive that I've reached the state this is an example 26 and literally it means I have reached the states to that I these places to that have reached so here to focus the subordinating clause we just move the dative form of the pronoun into the focus position preverbally but the conjunction stays in the same place so the marker of closed linkage in the purely linear and structural sense is end it's not the dative pronoun and now let's take a look at the properties of these constructions and try to determine whether they are coordinating or subordinating the first test that I will look at is embedding and coordination gives the expected results so canonical subordinate clauses can be that is correlates can be embedded if the subordinate was preceded the correlate like in 27 Zaur that he was hungry because of that ate a whole foot chin so here the subordinate clause is inside the main clause and 28 illustrates that we cannot do the same thing with coordinate clauses just like we saw recently just for English and according to this criteria both of the subordinating constructions are coordinating meaning that under no circumstances can the subordinate the semantic subordinator secondary clause be embedded within the main clause as shown in 29 and 30 the second criterion again in relatively linear order is the position of the conjunction so generally we can expect that coordinate clauses have the conjunction between the two clauses subordinating clauses have the conjunction inside one of these clauses and the aesthetic it is perfectly because of the fact that most subordinating conjunctions in aesthetic are pre-verbal like in 31 so here the sentence means I know that Zaur has done it so yeah I know that Zaur is a liar and as you see the subordinating subordinator K can only be located before the verb of the subordinate clause it cannot be clause initial it cannot be in the second position only pre-verbal so we can clearly see that it is at least in surface terms clause internal and there is a smaller subclass of subordinators which can which have a freer position but still they are characterized by the fact that they can appear in any linear position within the subordinate clause starting from the beginning of the clause and ending in the pre-verbal position so again generally we can put the subordinator within the subordinate clause and for and this is not possible so we have to place it between the clauses I called Zaur and he came here the conjunction and has to stand after Zaur it cannot move farther to the right or to the left in fact and as far as this test is concerned again we see that and in both of the subordinate constructions behaves just like a coordinating conjunction should so it cannot be embedded in any possible way as you see in 34 and 35 so some things along him and Cheats is impossible and the same concerns the 35 the final test which is related to linear order is the fact that generally in coordination in subordination you can coordinate two clauses containing the subordinating conjunction so I know that you came and that you repent so you have two that clauses coordinated whereas with coordinating conjunction you can't do the same thing so you cannot say Zaur came but he didn't prepare for the lesson and but he was ill for example and this is the contrast between coordination subordination which is also true for aesthetic as shown in 36 and 37 so I know that Zaur has done it but that he repents is perfectly fine but the sun is shining but it is cold and but I don't don't want to go for walk is bad and according to this criterion again the two constructions are subordinating so you cannot coordinate two clauses beginning with this end now let's take a look at more abstract features and here we finally find certain mismatches between the two constructions and coordination so in 40 we see that in the coordinated construction if some external constructions such as a conditional clause assigns mood features to the sequence of coordinate clauses the mood feature must be assigned to both of them you cannot assign it to the first if you come home and go to sleep you will pass your exam well tomorrow and both come home and go to sleep have to be in the subjunctive it can be the case that the first clause in the subjunctive and in the second clause in the future both can be in the future or both can be in the subjunctive so this has to be the same feature across the coordinate structure in 41 we see that subordinating construction on the contrary there is no such requirement so if you find out that he has arrived to me so here only the verb no carries the subjunctive feature whereas the complement clause has simple past tense and according to this tense clause also the coordination is coordinating meaning that both clauses must carry the same mood feature so if you take a wife because she has money you will not be happy so here if has to assign the subjunctive mood to both clauses you cannot use present tense or the future tense in the second clause if the first clause has subjunctive whereas in complement to the coordination the assignment of mood features to the subordinate clauses perfectly independent from what is assigned by if another test that also is related to the distribution of features across clauses is canonical subordination so unfortunately we cannot directly test the coordinate structure straight in the city because there are very few non-local dependency constructions so essentially WH movement and aesthetic only occurs locally within the same clause so we cannot for example extract WH word from within the complement clause into the main clause so we cannot apply the test directly but we can do some other things and in particular one of the things which is very similar to how the coordinate structure constraint is generally tested is the fact that when we have a sequence of two correlative clauses which correspond to the same correlative pronoun in the main clause both of them a correlative pronoun sorry a subordinator if they are coordinate so 44 shows that this is not the case for subordination when I found out what Zauer has done I stopped talking to him literally what Zauer has done when I found this out then I stopped talking to him so when only has to be found in the main clause not in the subordinate clause whereas in coordination you have to put the subordinator in both clauses so I know that Alan came home and went to sleep you have to repeat that Alan came home that came and then went to sleep that went to sleep that I know but you cannot use this conjunction in just one of the clauses and according to this criterion again we see that component coordination behaves just like a subordinating construction meaning that the conjunction has to be present in the semantic main clause the primary clause whereas causal coordination requires it it cannot be present in another test a final test on this in this part of the talk is right dislocation so aesthetic has a productive right dislocation construction where a polynomial and critique in the clause it assumes an NP mark by the same corresponding case at the right edge of the clause so for example 48 I saw him Zaur so he is an and critique and Zaur is a full nonphrase to the right to the main clause related to the clause that lies exactly to its left so in a subordinating construction it's possible to interrupt the main clause and the and the main nonphrase which corresponds to the critique by a subordinate clause because technically the right dislocated element still is a neighbor of the main clause because the subordinate clause is embedded within it so he strongly wants you to come Zaur here the particularizes for genitive case of the experience and as you see Zaur is found after the subordinate clause and Zaur is a preference by that by the critique but it cannot be such a construction is not possible in coordinating constructions like in 50 I saw him but he told me nothing Zaur here Zaur cannot cross the boundary of a coordinating construction it cannot enter the first conjunct by bypassing the second conjunct which is also a test which is a consequence of the coordinate structure constraint now as far as to the coordination and subordinate is concerned again we find that component to the coordination behaves like a subordinating construction so we can do this kind of jumping over the second clause whereas we cannot do this jumping over in 52 in causal to the coordination now the table here on the slide summarizes what we know so far and for the component construction the results fit rather well into the two level approach so we see that we have three criteria which are related to linear order and three criteria which are related to various distributive phenomena generally which can generally be related to the coordinate structure constraint and we see that canonical constructions rather consistently while the component to the coordinating construction has coordinating linear order properties but subordinating scope and extraction properties which refer to what Kurikovan Dzekidov described for English but the causal construction the results are somewhat surprising because it turns out to be both semantically and syntactically coordinating something which we do not expect of course in general and isn't this a strange route for causal meaning now generally we can say yes but fortunately there have been certain attempts to show that certain languages actually do have a similar example is German then so some of you may know that German has two main causal constructions while and then and they're distinguished in terms of word order and subordinate clause so then has main clause word order verb second whereas while has subordinate clause word order verb final and this also corresponds according to a recent book by Scheffler to their scope properties and where as then clauses cannot be in the scope of any semantic operators such as focus so 53 why did the cat jump because it's our mouse because can only be while the subordinate thing cause and not then and similarly in scope of equation we can only find while and not then in German and first solution is to say that basically is to use response distinction between conventional issue meaning that it is part of the truth conditions which can be negated which can be asserted and there is an additional level of information which is not directly asserted which is only implied which is cancelable and this is conventional caricature and then introduces a conventional caricature whereas while introduces an at issue that's the difference between the two conjunctions and interestingly this corresponds to a long tradition going since first work on conventional caricatures which states that co-ordinated conjunctions in general and not asserted meanings so with co-ordinated conjunctions we find the same kind of behavior just so for this German then so for example who came but he was ill you cannot have this but close in the scope of the question you can but you can do this so who came but he was ill is not a possible construction and similarly an answer to a question if a question is when did you come a felicitous answer is I came when you called me but it is not very felicitous when did you come you called me and I came because the meaning the temporal meaning which is implied by and is cannot be in the focus of the question cannot be in narrow focus or in assertion a similar idea has been proposed by the law and down low and then a group for French the causal conjunction car so French also has two conjunction and car one of them is more co-ordinated and one of them is more sub-ordinated in English 4 and they show that just like Dan and German car cannot be in the scope of various semantic operators such as negation so 57 there is only a content car and I and I but car if a boy is impossible so you cannot say this I is not happy because she got a name but because the sun is shining you cannot put the causal meaning within the scope of negation with car but you can do this with Pascal and similarly within the scope like maybe you can put the causal meaning applied by car this co-ordinating causal meaning but you cannot do the same thing with Pascal so the idea is that a car is a discourse level conjunction which introduces rhetorical relations rather than assertive predictions and again this finds its correspondence in work on the semantics of co-ordination so in Turuk in 2003 it has been argued that and also introduces discussions on the temporal order or causality between the two clauses let's keep this so the solution that the Logan Danlo proposes is cast in the framework of segmental discourse representation theory of Escher and Lascaritas so here on the slide you see the sexually proposed for the sentence Lisa is pleased because she got a name with the conjunction Pascal the subordinating conjunction and the idea is that there is just one speech act one DOS which contains two facts and these facts are linked by subordinating relation which is part of the same speech act whereas for Kahr we have two separate speech acts which are act by discourse level rhetorical relation explanation which explains rhetorical relations unlike normal predicates cannot be negated you cannot negate something which is used to build the structure of the discourse which explains the scopelessness of Kahr so maybe aesthetic because is just like what people have described as strange but this is not the case so 60 shows that this construction to the coordinating course can be freely used as an answer to a question so 60 is an example from the aesthetic national corpus why do you want to go to work at the theater and answer it because they cannot live without the stage and here because is to that end the construction which we have just been discussing another criterion which shows that they are not scopeless is the test which has been proposed for Russian impact which cleft like construction where preposing a demonstrative problem before the subordinate clause is possible but not before the coordinate clause so 61 modern photography has become banner of lashes and uninteresting and all of this because it imagines itself to be art this is possible with subordination in Russian but similarly built construction is impossible for coordination so you cannot say in Russian Marsha was occupied by preparations for this end relation and for aesthetic does have the same construction and it gives the same results so chronical subordination you can use it that was while he didn't yet see Hamerzah so here the demonstrative that is preposed to the subordinate clause and the meaning is the focalization of the while relation that was only while he didn't yet see Hamerzah and 64 and this and I haven't learned your lesson this is not possible it's not aesthetic and cause also the coordination turns out to be subordination according to this criterion so it is perfectly possible to say something like 65 this is because sacrifice all his life and creation for the sake of his people so here the demonstrative that is preposed to the semantically coordinating unless we just define semantic coordination we're using a limited set of tests but in this case the use of the term semantic to describe this relations seems to be rather circular so we first introduce a set of tests say that they are semantic then we look at the construction look at this test it's semantic but it doesn't correlate neither with the meaning of this construction nor with other criteria which are semantic proper and of which we know independently that they are semantic so if we don't want to abandon the level approach we need to assume that there are three levels and three sets of tests corresponding to them one of them is for linear order this syntax that's what Kulikova and Jackendorf originally proposed the second level is for the scope of semantic operators such as negation and epistemic modality and so on which is not has not been discussed in prior work but which can be safely assumed to be semantic and finally the level which has been discussed by Kulikova and Jackendorf for the coordinate structure constraints, distribution of mood features constraints on an alpha and so on somehow we don't know where to put it it's some third level which we don't know what it can be and what might that be and I think that a good answer will be from the architecture of lexical functional grammar a theory which most of you probably know about which assumes exactly at least three levels for the representation of syntax and semantics so syntax and this theory and functional structure constituent structure encodes C structure encodes purely hierarchical and linear information about words in the sentence illustrated on the left via normal constituent structure tree whereas the F structure functional structure encodes grammatical dependencies grammatical relations and it is on F structure that things like constraints on anaphora distribution of features the coordinate structure constraint and so on the constraints on extraction are formulated in particular coordination of G is described in the following way so you have at C structure a symmetrical construction where you have one category like in this case IP above then you have two children who are also IPs and the conjunction between them and at the level of F structure you have a set of two F structures so Alan has seen Stanislav and Brasul has seen Ali you have a set of two in the publications Alan has seen Stanislav and Brasul has seen Ali and important the F structure even though it's quite close to semantics and originally in early work in RLG it was sort of people try to do work in to do semantics in F structure it is not semantics and there is a separate level for semantics that is not pure level but a deductive system which allows so the syntax supplies as premises for glue proofs for semantic derivations to the rules of the lambda calculus and importantly the resulting so there is a separate level where we have truth conditional meanings of the sentence which is distinct from F structure so in this case we have formula of the first order logic but we can have also other kinds of semantic representations in so far as they are compositional in so far that they can be formulated using the lambda calculus for example DRT so this is not crucial what's crucial is that the same as people assume and work on compositional formal semantics so as I have shown here LFG has this view of coordination and subordination but the modular architecture of LFG means that we can define actually this concept separately for each level so we don't have to assume that for example C structure is like this and it must correspond to a set it can also correspond to subordinating F structure and conversely coordinating F structure can correspond to a more subordination like C structure so we can have a separate definition for each level at C structure we can say that the distinction between power taxes and hyper-taxes that is there is a for coordination we have the same category label above in the mother node and the same category labels in the children node possibly with the exception of conjunctions whereas in subordination the category information is only inherited from one of the two child nodes it is similar to this teaching is one of two or more sister nodes whose category information percolate to the mother node and in this subordination it only percolates from one of the children nodes so on F structure we can define the difference as a relation between some F structure keeping a mathematical function of another F structure like on the left or the two F structures being members of the same set and on the semantic level we can adopt for example the discourse definition which I just showed for subordination we have one speech act and some sort of semantic relation between the two facts or propositions or events or whatever kind of abstract objects you like or on the for coordination we have a discourse rhetorical relation between two speech acts and the syntactic tests and critical injecting and you as I say that belong to C structure and most of the semantic tests belong to F structure which explains why we have a causal construction with certain coordinating properties which they would call semantic and how this is not a pure stipulation but this is in fact quite uncontroversial in the literature so across the board extraction and the scope of more features are explained in the analogy in the same way it is the consequence of the rules of feature resolution of sets so if you apply a feature to a set and if its feature is distributive anything from just one of the conjuncts because in this case there will be different values for the different conjuncts and you cannot if you assign a mood feature to a set similarly it has to distribute over all of the conjuncts of this set and the coding structure constraint is a bit more complicated but generally also it is explained by the same rule and the only property semantic tests are tests on the scope of external operators focus ability negation questioning of the linking relation what I have just and other languages and in this approach of static constructions will fall will cluster in the following way so the complement see the coordination will be coordinating and C structure and subordinating and both F structure and semantic whereas the causal construction will be coordinating and C structure too but coordinating and F structure whereas semantics will also be subordinating just like with complementation and this illustrates the analysis so for C structure we assume a symmetrical structure like shown on the slide the same structure we assume for the complement construction so the complement causes in syntax in C structure terms merely coordinated with the main clause I think and if cheated me they seem to be the same level but on F structure for the causal construction we assume that there are members of the set so I came because you called me there is just a set of two situations I came and you called me then in the first which refers to the second clause and which includes the causal relation whereas for complement clauses we just have a normal complementation F structure which corresponds to F subordination and on the semantic level for both constructions we have quite a controversial subordination meaning that there are just relations within one speech act and importantly what I would like to stress is that is the difference between this approach that is for example this causal coordination approach and the other approaches which are based on tests here we do not assume the tests are the primitives of our analysis the primitives on our analysis are abstract structures which the theory supplies us with the framework then we formulate hypothesis based on these structures which we have available and then we see what in this particular language can follow from this what it predicts in this particular language and the test that we use here in fact it means that we have to devise separate tests for each of the language and the fact that the same test do not apply in the same way in all languages is not in any way falsified this idea so for example for aesthetic most of the tests that I used here are not something which is usually used but something which is just came up for a particular language based on the fact how these things should behave given the formal framework so that's a way to see what are the cluster they don't here we don't necessarily have to have the same test apply and to illustrate somehow that this approach is not at hope but also can provide insights in other consensus in other language I would like to focus on a second case study shortly but this is often rough labels for their translation and the first thing that immediately comes up when we look at them is that only Pada Mushta and Taka Kalo epistemic uses like he's probably not sleeping because there's light in his window so here the because clause does not mark the cause of the main clause in the logical sense it marks the fact that we infer that change concerned according to the test which can be related to f-structure all of these three constructions are subordinating so all of them do not allow across the board extraction for example on 67 what did Peter throw away because vice broke this is not possible for either of the three conjunctions and similarly possible with any of the conjunctions but semantically they do not align as neatly so for example Taka seems to be scrupulous it cannot be in the scope of this demonstrative that constructions will be focused so you cannot say as for this but this as it has been raining this is not possible with Taka so as far as more itself art an example which we already had above and also it is possible with Pascorko so it was very funny for me but this is since I know many of those about whom the story is concerned so here we have Eto Pascorko this is possible at least according to Corpus data and my intuition confirms this too so yeah let's keep this but so this means but this does not at all correlate with C structure because both Taka and Pascorko one of them is coordinating and the other is subordinating can be embedded in the main clause like in 73 and 74 show it can be pre-post so you can say Vasya came for Peter Coltham Tanka Peter Pascorko Vasya on pre-show you can pre-post the subordinate clause you can embed it Vasya Taka Pascorko but it has been long remarked in literature in Russian that you cannot do the same things with Potamushka even though it's a more widespread and apparently more subordinating in the semantic sense conjunction so 75a shows that pre-posting of Potamushka clauses is decidedly weird so Potamushka Peter Pascorko Vasya on pre-show maybe on in a parenthetical context in this approach which I'm proposing here we have a level of C structure on level F structure we have consistent subordination on level of semantics we have Taka which is coordinating while the other conjunctions are subordinating Potamushka has some discourse users which can be classified as coordination which I didn't stop on but you can see it in one of the slides in the handout and as far as C structure is concerned it does not in any way correlate with the first sort of say that Potamushka is always coordinated regardless of its meaning whereas Taka and Pascorko are subordinating even though Taka is semantically coordinating to show this at least for Russian this approach proves fruitful and allows us to at least it doesn't the Russian data confirms that the tests fall into three categories they can be mismatching but they can be in general tests which are belong to one level match with each other so this at least for generally has to be investigated more further in this approach with three levels and LFG allows us to do this in a very straightforward way but of course this sounds like this joke from Spengram so if someone says 28 is very likely universal constraint it means that I know for sure that it works for English French and certain lot of them is direct so I've made a similar claim here so basically it knows for English and Russian therefore it's sort of universal you're promising this direction and currently it explains more than the other approaches and I hope that it can in the future work for other languages and explain the data too thank you very much thank you very much lots of exciting data ideas and I'm curious I'm not quite sure whether I can say that right but I'm curious curious about cases where it looks like you have a correlation between the feature and the linear order which you wouldn't expect if you dissociate them so much and there's two cases I think of one is you have similar examples like the partial agreement with NP coordination so where if you have John and Mary came you typically get full plural agreement if the conjunction but in many languages you get partial agreement if the conjunction you follow so there came John and Mary in Arabic you get partial agreement but with the first conjunction there's the exception but typically across the board it will be linear order based but the reflex will be feature based and the the second example is right now raising where if you have in English it's best with negative polarity I washed but then didn't dry any of the dishes and the NP is licensed by the second conjunction mm-hmm so if you turn around I didn't I didn't wash but then I dry all the dishes if you don't get the NP I my preferred example in Germany was where it's cased if you see I saw I saw I saw but then didn't help the girl on the road then the girl on the road will be dated because she wants to date object whereas if you I helped but didn't see you know that it will be accusative because c is accusative so in both cases what you have is there's a particular feature like in the case of the agreement which looks like F structure I guess but then but then the determination of that is clearly linked to linear order to the C structure and I always wonder and I always thought that LFG would trouble some person about this situation but maybe that's yeah thank you it's an interesting question and in fact this unbalanced coordination there's a lot of literature and it's generally a problem for LFG but there's been a proposed analysis of it by Rachel Nordling and Luisa Svedler I think and what they propose is that you have a mapping so you haven't a flat structure at C structure then you have a set at F structure but the set is ordered according to in relation to each other so there's a relation called F precedence which maps F structures to C structures and all of the nodes which correspond to this F structure in C structure have to precede the nodes that correspond to another F structure so based on this feature they define edge features so they define features which only are taken from the left most or the right most conjunct and in this way they resolve this problem of anyway although of course and the important thing is that these features are all features of normal coordinate constructions they are not features of some sort of mismatching things so it seems that these features should be thought of as basic to how LFG handles sets and C structure in general and not this kind of more complicated mismatch things but at least the current proposals are based on F precedence and they seem to work generally for the F structure some of mimics C structure yes yes so there's no you can't just draw a very clear line between the two and there are other cases like this for example a junction if you just look at C structure you can not always delineate a junction from coordination if you are joined an IP to another IP you have IP above two IPs below you just can't say what is coordination subordination based on C structure also that's kind of you that's an idealization but it allows us to first have to separate the things look at what we can do separately and then maybe look at possible correlations on no correlations so that's a nice way to do it I think yeah thanks you talked about coordinating causal constructions but if you apply your diagnosis it's a very frastic causal constructions it's a multi-posal so I got my son to watch the car go to market buy some very free so we have a string a recursive string with the embedded clause mm-hmm how does that figure in the formula very much in causative human causative like I made I had I made the son watch the car yeah well I didn't look at causative it's not the case that I made for example can the I mean they're classified as main clause with the causative and embedded subordinate clause that was something how I think it would be semantics at least subordination of one event to another sub event maybe the causation event and the main one because you would I think the scope of semantic operatives can only for maybe he made the son watch the car for example it maybe can scope over watch not about made or something like this I mean I can't say for sure now but I think the scope of the semantic operatives would be only on the main predicate because there's maybe something yeah thank you causative causative thank you general general question not really linguistic probably to your last comment sort of presenting the stuff to the typology crowd I mean to the typologists the reason I'm asking is that what you're saying here it seems to be very much again the traditional typology however it seems to be very much in alignment with so-called canonical typology which is very popular well not very popular but here in the UK it is the sort of game popularity of such people I think you can sort of develop this very interesting picture with a canonical typology just wondering how many I hope the data from other languages not yet so much because there's not much data that I mean this requires very detailed analysis of language data well I do obviously but I also sort of encouraging everyone to try to apply the same thing to languages and try to look where it works because you have to really look and before formulating this you can't just apply a battery of tests you have to look first how language works in general that's the problem but yes this is of course in total disagreement not very optimistic so you like the general direction of thinking but because this is not comparative concept according to this yeah this very restricted definition of the term of course I'm curious about the because which it's a conjunction and at what at a date yeah it's a date of pronoun in the date of form and the conjunction and the date of pronoun sort of categorically refers to the second clause it can so it's like because of that so how can I yeah so you have an element an nominal element in the main clause in the semantic the main clause which semantically refers to the second clause and the conjunction stands between them but these date of elements can go into the focus position focusing and you can replace this date of by a we've seen that it's this date has its students the whole second situation what it's sort of like the correlatives which I just didn't say it at the beginning but it's the second the order is first the main clause it contains demonstrative and then it's categorically refers to the second clause not so what do you have a composition analysis of that in the journal for essentially what it does is that and in the causal construction in my analysis just I didn't give it here because it's this grue semantics thing and only contributes the conjunction of the two situations and the date of form is enforced so the construction enforces the date of to be there and sets the its intercedent syntactically to be the second clause and also it assigns to it the date of meaning so it's composition in the sense that it's in the grue semantics but it's associated with the construction not with the the date of itself because the date of itself on its own generally does not mark course in aesthetic it's construction specific here you see the reason I'm asking is and hard because of additive focus marking because in many languages you have additive focus markers which are end for a problem so what the English company also has too so John who went to the cinema often comes out as John I went to the cinema and it is so he abused the John but in terms of binding it doesn't work out and so that's another thing where it's not clear what the different parts do so maybe construction analysis actually would also work out how do you do you have additive focus marking anywhere additive you mean also it just a particle no real it's a shame it's very simple yeah hmm well aesthetic has correlatives I have looked at them generally just different dissertation correlatives aesthetic so in general correlatives are subordinating because they in this language so they can be but there's an interesting construction which can turn a normal correlative into a sort of a construction like John whom so he is my friend and John who so so you can post post the subordinate close place and between them and use the subordinator in the second close so you can sort of turn any subordinating construction into this kind of to the coordination a similar kind of thing I haven't yet explored them in much detail but I'm just pointing out that in aesthetic apparently also relativization certain kinds of relativization behaves in a similar way and also temporal closes because any correlative they have the same structure you can see something like I then came and when you called me meaning I came when you called me so you have some to the coordinating relativization even though the normal correlatives they are subordinating because they can be embedded they have no they don't fall into the scope of various operators they do not have to have the same mood and so on but they can be sort of lifted up in relation to relative causes which are not correlative not correlative but in general well it depends on the type of correlative causes I would say mostly subordinating would you expect to find it? yes I would I mean certainly I would but especially since you often have probably for example correlatives I am a good example because they often develop from paratexics and you would expect to have semantic or F-structure probably yeah with correlatives and possibly also with things which develop from adverbal causes like non-restrictive correlatives often correspond relative causes often correspond to verbal constructions I think like when you it's not clear whether you can classify them as non-restrictive causes maybe languages just don't have non-restrictive but you could expect that such things eventually develop into normal relative causes and in this case that's due to grammaticalization in fact I think it's simply because sort of the syntax the F-structure first you have the semantic shift to subordinating then you have F-structure shift to subordinating structure and then C-structure sort of sometimes lags behind like here so the F-structure has already become for the complement construction subordinating but the C-structure is still flat so we can expect that at some point it will also become subordinating maybe can we take another look at relative cause on the screen I'm afraid I don't have examples ready at hand maybe online in the corpus but yeah I can write on the board what's the subordinator in the relative cause? Interrogative based subordinator Interrogative based they are derived from relative cause relative elements yeah can you get the demonstrative demonstrative in the main cause it's a resumptive so you have something like and I just like this John Whom so he is my friend that's the normal way of forming relative causes in the language so the one who John so is my friend but technically it's John Whom so he is my friend and you can have an internal head you can say John which boy so he is my friend which means the boy who John so is my friend sorry the subordinator it's Whom yeah okay so it's always a WH yeah and WH you can't get the demonstrative no only here so this he is a demonstrator nice and you sort of have a division of labor so he marks the position in the main cause and Whom is internal marks the position of the relative cause so this can be redone into he is my friend and John Whom so something like can that element to the the other part of the do anything else in the language it's a normal it's it's still demonstrative and it's used also just for as a polynomial used after the communication you said that no for communication you would use either also a relative or you would use this convert construction saying you use like Turkey and other languages of North Eurasia I think it's everywhere so you say I will come saying he said meaning he said that he will come