Linguistic Miscellany Thread

[Estav]

To be clear, I only meant to present a test for identifying unaccusative verbs, not for identifying unergative verbs or for finding out whether a verb is not unaccusative. I think it's likely that not all unaccusative verbs have past participles that are easily used in this way. I don't know exactly what other factors influence past participle use, but I think something like aspect or Aktionsart is also relevant, along with perhaps complementation requirements for some verbs.

I think the acceptability of the past participle as a nominal modifier is closely correlated to the acceptability of usage of the perfect construction (which also uses the past participle) in a relative clause:

• a train that has recently arrived
• a disease that has spread --- grammatical, but seems a bit semantically incomplete without any additional element. "A disease that has spread through the country" or "A disease that has recently spread" feel more complete, and I think ?"a recently spread disease" or ?"A disease spread through the country" sound better to me than *"a spread disease". I'm also not sure whether the potential for ambiguity with the passive participle "spread" from the transitive verb reduces the acceptability of "spread". This is the explanation I'm least sure about.
• ?*a pie that has sat on the windowsill --- "has sat" is not an invalid syntactic structure, but it cannot easily be used for pies in my dialect, and even for animate entities I think "has sat" is much less frequent than "has been sitting" (or just "sat"). I guess I can imagine something like "a pie that has sat on the windowsill for three hours", but even in this context I don't know whether "has sat" is what I would actually produce in real life, since "has been sitting" works at least as well there. In contrast, "the people who have sat down at that table" sounds better to me, and the corresponding noun phrase using the participle without have ("the people sat down at that table") sounds OK-ish to me and doesn't feel particularly foreign to my dialect.
• a figure that has mysteriously appeared

Based on Google search results, it does seem like some people use "resigned" in the way that, according to the test I presented, implies that "resign" is unaccusative in their grammar. Intransitive "resign" being unaccusative for some speakers seems plausible to me, although that usage feels a bit odd to me I think.

[Whimemsz]

I would also caution against an over-reliance on Dixon or thinking he always represents the """mainstream""" on some given point, since he has some idiosyncratic views, although this case isn't so much an instance of that, as he's just mentioning the cases where "unaccusative" and "unergative" have relevance to his model of ergativity, and he notes that the terms are also used in reference to other phenomena.

Could you elaborate on that? Since his book Ergativity by far the most comprehensive and detailed book on the subject I have found, I’ve been using it as the main source for my series on ergativity. If there are any problems with it, they would be very useful to know.

It's honestly a little hard to explain, at least for me. You kind of have to read enough of his stuff and then that of other researchers on the same topics. He just...he's an idiosyncratic guy, and is very wedded to certain concepts and models in his work, and thinks all the other ones are stupid and not worthy of his time or attention. Which isn't necessarily wrong, it just means his emphases will be different from many other researchers'. He also sometimes uses terminology differently than most others do. And it's also not to say that he's totally outside the mainstream on everything; the whole A/S/O (or A/S/P) concept for example has been very widely adopted by other functionalist typologists. (His dogmatic insistence that all languages have adjectives, or his version of "Basic Linguistic Theory" which supposedly encourages treating every language on its own terms but really is very theory-laden and necessarily fits the grammatical patterns of languages into molds just as much as any other theoretical construct, though...are less justifiable.)* I don't recall all the very specific claims he makes in Ergativity so I didn't have any particular examples from there in mind, it was a more general statement.

*Actually, so I can give one kinda amusing example of this, though it's not part of "Basic Linguistic Theory" per se. He and Aikhenvald have edited a bunch of typological overviews of this and that grammatical topic. One of them is on clause combining. His introductory chapter lays out this very specific framework for interpreting how languages combine clauses, which is definitely idiosyncratic, and he never makes it clear exactly what data have informed it (IIRC all of his examples are either from English or from one of the chapters in the volume, which were written after the first draft of his introduction, since the first draft of the introduction was circulated to the contributors as a guide). But it basically breaks down types of clause-clause relationships into these certain numbered/lettered subtypes based on their semantic relationship to one another; analysis of clause combining is then supposed to be in terms of how a given language actualizes each of these subtypes, whether it marks the main or subordinate clause, what order the clauses are in, etc. All of the contributions in the volume then slavishly follow this layout, cramming their languages into the predetermined analytical framework Dixon has given the authors, even in cases where it's clear that this is very tortured. There are two exceptions, though -- two chapters which don't follow this framework as such, but instead use a much more language-centric analysis and presentation. TOTALLY COINCIDENTALLY, the two chapters are the ones authored by Dixon and by Aikhenvald. Pretty crazy that precisely the languages they've worked on are best analyzed and presented in different terms from the cookie-cutter analysis Dixon develops, but all the languages worked on by people who needed to submit their manuscripts to Dixon and Aikhenvald for review (after having been told what analytical framework to write in) can be described in terms of this framework!

[Note that I do think Basic Linguistic Theory in one broader sense is quite justifiable, but not in the sense that in practice Dixon often uses it, which is as illustrated above: developing some generalization, then trying to fit everything into that generalization.]

[bradrn]

I would also caution against an over-reliance on Dixon or thinking he always represents the """mainstream""" on some given point, since he has some idiosyncratic views, although this case isn't so much an instance of that, as he's just mentioning the cases where "unaccusative" and "unergative" have relevance to his model of ergativity, and he notes that the terms are also used in reference to other phenomena.

Could you elaborate on that? Since his book Ergativity by far the most comprehensive and detailed book on the subject I have found, I’ve been using it as the main source for my series on ergativity. If there are any problems with it, they would be very useful to know.

It's honestly a little hard to explain, at least for me. You kind of have to read enough of his stuff and then that of other researchers on the same topics. He just...he's an idiosyncratic guy, and is very wedded to certain concepts and models in his work, and thinks all the other ones are stupid and not worthy of his time or attention. Which isn't necessarily wrong, it just means his emphases will be different from many other researchers'. He also sometimes uses terminology differently than most others do. And it's also not to say that he's totally outside the mainstream on everything; the whole A/S/O (or A/S/P) concept for example has been very widely adopted by other functionalist typologists. (His dogmatic insistence that all languages have adjectives, or his version of "Basic Linguistic Theory" which supposedly encourages treating every language on its own terms but really is very theory-laden and necessarily fits the grammatical patterns of languages into molds just as much as any other theoretical construct, though...are less justifiable.)* I don't recall all the very specific claims he makes in Ergativity so I didn't have any particular examples from there in mind, it was a more general statement.

Thanks for the warning! I’ve actually been reading Basic Linguistic Theory, under the misapprehension that it was, well… a textbook on some areas of linguistics, rather than his description of his theory. I must admit that, to me, it doesn’t seem to have much in the way of theory, and it doesn’t appear to obviously force languages into molds like you say; rather, I get the impression that he has looked at lots of different languages and found the commonalities between them. But of course, as someone who doesn’t know a whole lot about linguistics, I probably wouldn’t be able to tell when he’s doing this; could you give an example if you can think of any?

(As for his endorsement of a universal adjective class: I actually couldn’t believe that any linguist today could seriously think that, so I had a look at what he has to say about it. In one of his and Aikhenvald’s typological overviews, devoted specifically to this subject, he describes his reasoning: all languages have some sort of difference between adjectives and verbs, often in TAM marking (he gives lots of examples, so I don’t doubt this), and thus adjectives must be a separate class to verbs. I agree that adjectives can always be identified as a distinct class, but I think that last conclusion of his is a bit odd: the differences are often so minimal that I think it would be more useful to say that adjectives are a subclass of verbs rather than being a completely distinct word class.)

*Actually, so I can give one kinda amusing example of this, though it's not part of "Basic Linguistic Theory" per se. He and Aikhenvald have edited a bunch of typological overviews of this and that grammatical topic. One of them is on clause combining. His introductory chapter lays out this very specific framework for interpreting how languages combine clauses, which is definitely idiosyncratic, and he never makes it clear exactly what data have informed it (IIRC all of his examples are either from English or from one of the chapters in the volume, which were written after the first draft of his introduction, since the first draft of the introduction was circulated to the contributors as a guide). But it basically breaks down types of clause-clause relationships into these certain numbered/lettered subtypes based on their semantic relationship to one another; analysis of clause combining is then supposed to be in terms of how a given language actualizes each of these subtypes, whether it marks the main or subordinate clause, what order the clauses are in, etc. All of the contributions in the volume then slavishly follow this layout, cramming their languages into the predetermined analytical framework Dixon has given the authors, even in cases where it's clear that this is very tortured. There are two exceptions, though -- two chapters which don't follow this framework as such, but instead use a much more language-centric analysis and presentation. TOTALLY COINCIDENTALLY, the two chapters are the ones authored by Dixon and by Aikhenvald. Pretty crazy that precisely the languages they've worked on are best analyzed and presented in different terms from the cookie-cutter analysis Dixon develops, but all the languages worked on by people who needed to submit their manuscripts to Dixon and Aikhenvald for review (after having been told what analytical framework to write in) can be described in terms of this framework!

Oh, dear. Would that be Dixon and Aikhenvald’s Complementation? I did attempt to read it, and went away thinking that it was totally useless for learning how complementation works. But some of the others in the series seem a lot better; I’m currently reading Serial Verbs, and it seems a lot more useful (maybe because Aikhenvald rather than Dixon wrote the first chapter). (Specifically, Serial Verbs doesn’t attempt to exhaustively carve up the space of all serial verb constructions in the same way that Complementation attempts to exhaustively categorise complement clauses into different categories; instead, Aikhenvald simply describes various different properties of serial verbs, such as semantic meaning, restrictions on verb choices, contiguity, wordhood etc., which may or may not coincide.)

(But are you sure you’re remembering this book correctly? If it is indeed Complementation, he didn’t used ‘numbered/lettered subtypes’, but rather used a weird division into ‘complement clauses’ and ‘complementation strategies’, where the former was divided into Potential, Activity, Fact etc. The numbered/lettered subtypes were for verbs, which he divided into PRIMARY-A, PRIMARY-B, SECONDARY-A, SECONDARY-B.)

[Note that I do think Basic Linguistic Theory in one broader sense is quite justifiable, but not in the sense that in practice Dixon often uses it, which is as illustrated above: developing some generalization, then trying to fit everything into that generalization.]

I’m not sure how this is different to what any other linguistic theoretician does. Whenever I read a paper on such-and-such a linguistic theory (which I have been doing recently in research on ergativity), it seems that they all follow a basic pattern: look at one language, construct a whole elaborate theory around what this language does, and then go from that to ‘so this theory must describe every single language of this type without exception’. If you’re very lucky, they’ll look at several languages, all of which mysteriously happen to be just those languages which fit the theory perfectly. I can recall reading exactly one paper which didn’t do this, and for that reason I ended up understanding its conclusions to a much greater extent than the other papers I’ve read. Amusingly enough, Dixon himself complained about this in Ergativity:

[Theoretical models] are sometimes suggested on the basis of data in a very limited set of languages, but are then put forth as general accounts of how all human languages operate. When unexpected data from new languages come to notice there can be a number of reactions: ignore it; reinterpret the data so that it fits the theory; revise the theory so that it does explain the data; acknowledge that the theory cannot explain the data and as a consequence abandon it.

(I think this is a big reason why I dislike syntactic theories: they all seem to be based on this sort of overgeneralisation.)

[Whimemsz]

I don't have much time atm so I'm just gonna respond to this part for now and try to get to the rest tomorrow, sorry!

Oh, dear. Would that be Dixon and Aikhenvald’s Complementation? I did attempt to read it, and went away thinking that it was totally useless for learning how complementation works. But some of the others in the series seem a lot better; I’m currently reading Serial Verbs, and it seems a lot more useful (maybe because Aikhenvald rather than Dixon wrote the first chapter). (Specifically, Serial Verbs doesn’t attempt to exhaustively carve up the space of all serial verb constructions in the same way that Complementation attempts to exhaustively categorise complement clauses into different categories; instead, Aikhenvald simply describes various different properties of serial verbs, such as semantic meaning, restrictions on verb choices, contiguity, wordhood etc., which may or may not coincide.)

(But are you sure you’re remembering this book correctly? If it is indeed Complementation, he didn’t used ‘numbered/lettered subtypes’, but rather used a weird division into ‘complement clauses’ and ‘complementation strategies’, where the former was divided into Potential, Activity, Fact etc. The numbered/lettered subtypes were for verbs, which he divided into PRIMARY-A, PRIMARY-B, SECONDARY-A, SECONDARY-B.)

No, the one I'm thinking of is The Semantics of Clause Linking.

[bradrn]

I don't have much time atm so I'm just gonna respond to this part for now and try to get to the rest tomorrow, sorry!

That’s fine.

Oh, dear. Would that be Dixon and Aikhenvald’s Complementation? I did attempt to read it, and went away thinking that it was totally useless for learning how complementation works. But some of the others in the series seem a lot better; I’m currently reading Serial Verbs, and it seems a lot more useful (maybe because Aikhenvald rather than Dixon wrote the first chapter). (Specifically, Serial Verbs doesn’t attempt to exhaustively carve up the space of all serial verb constructions in the same way that Complementation attempts to exhaustively categorise complement clauses into different categories; instead, Aikhenvald simply describes various different properties of serial verbs, such as semantic meaning, restrictions on verb choices, contiguity, wordhood etc., which may or may not coincide.)

(But are you sure you’re remembering this book correctly? If it is indeed Complementation, he didn’t used ‘numbered/lettered subtypes’, but rather used a weird division into ‘complement clauses’ and ‘complementation strategies’, where the former was divided into Potential, Activity, Fact etc. The numbered/lettered subtypes were for verbs, which he divided into PRIMARY-A, PRIMARY-B, SECONDARY-A, SECONDARY-B.)

No, the one I'm thinking of is The Semantics of Clause Linking.

Oh, right. I happen to have access to that one as well, so I’ll look through it and see if I can find what you were talking about. (I’ve already found the weird numbered/lettered subtypes.)

[bradrn]

Oh, dear. Would that be Dixon and Aikhenvald’s Complementation? I did attempt to read it, and went away thinking that it was totally useless for learning how complementation works. But some of the others in the series seem a lot better; I’m currently reading Serial Verbs, and it seems a lot more useful (maybe because Aikhenvald rather than Dixon wrote the first chapter). (Specifically, Serial Verbs doesn’t attempt to exhaustively carve up the space of all serial verb constructions in the same way that Complementation attempts to exhaustively categorise complement clauses into different categories; instead, Aikhenvald simply describes various different properties of serial verbs, such as semantic meaning, restrictions on verb choices, contiguity, wordhood etc., which may or may not coincide.)

(But are you sure you’re remembering this book correctly? If it is indeed Complementation, he didn’t used ‘numbered/lettered subtypes’, but rather used a weird division into ‘complement clauses’ and ‘complementation strategies’, where the former was divided into Potential, Activity, Fact etc. The numbered/lettered subtypes were for verbs, which he divided into PRIMARY-A, PRIMARY-B, SECONDARY-A, SECONDARY-B.)

No, the one I'm thinking of is The Semantics of Clause Linking.

Oh, right. I happen to have access to that one as well, so I’ll look through it and see if I can find what you were talking about. (I’ve already found the weird numbered/lettered subtypes.)

I’ve managed to look through The Semantics of Clause Linking a bit more thoroughly now. I can definitely see what you said about it — every chapter except Dixon’s and Aikhenvald’s goes in turn through the same sequence of clause linking types: temporal, consequence, possible consequence, addition etc., even when the language itself divides up the clause types a bit differently. (Taking a random example: the Ojibwe chapter lists ‘conjuncts’ as being used with lots of different clause linking types, and I assume that these should really have been discussed together.)

On the other hand, I don’t see anything actually wrong per se with their categorisation of clause linking types, apart from maybe some slight overanalysis. And in fact, I see this categorisation as potentially being rather useful, for the purpose of enumerating the various ways that clauses may be linked across many languages. I think this is why I quite like Dixon’s work: from a conlanging perspective, it can be very useful to have this sort of exhaustive enumeration, where I can then look and it and say something like ‘for this conlang, how will I mark each of these types, which ones will have the same markers, and which categories will be conflated?’.

[akam chinjir]

[Theoretical models] are sometimes suggested on the basis of data in a very limited set of languages, but are then put forth as general accounts of how all human languages operate. When unexpected data from new languages come to notice there can be a number of reactions: ignore it; reinterpret the data so that it fits the theory; revise the theory so that it does explain the data; acknowledge that the theory cannot explain the data and as a consequence abandon it.

(I think this is a big reason why I dislike syntactic theories: they all seem to be based on this sort of overgeneralisation.)

I don't really see why you'd think it's overgeneralisation if you interpret data in the light of past conclusions. I really don't see why you'd think it's overgeneralisation if you adjust the theory in the light of new data.

Like, as I understand it, work primarily on Bantu languages led to the idea that the extended projection of the verb can include an applicative head. Once linguists have found such a thing in a bunch of languages, understandably they wonder if it also shows up elsewhere. Some (it's controversial of course) ended up concluding that the English double object construction is actually a sort of applicative. As far as I can see this is all totally reasonable.

That reminded me that I was going to comment on a comment Whimemzs made:

the whole A/S/O (or A/S/P) concept for example has been very widely adopted by other functionalist typologists

An awful lot of liguists, including Chomskyan ones, use A/S/O (or P) as a sort of shorthand. How many agree that it's in any significant sense a primitive, and that, for example, "syntactic rules in every grammar are framed in terms of them" (Ergativity, 113)? I mean, as opposed to the many linguists who think it's something like the relational grammar distinction between initial 1s and initial 2s, or the Chomskyan distinction between external and internal arguments, or the typological distinction between actor and undergoer arguments that's more basic?

(This is actually an issue where you do find people getting tripped up by their own theory, I think. For example, the A/S/O framework doesn't provide an obvious way to distinguish unaccusative from unergative verbs, so e.g., the very robust generalisation that it's only internal arguments / initial 2s / undergoer arguments that can be incorporated is easy to miss if you're working in terms of Dixon's theory.)

[bradrn]

[Theoretical models] are sometimes suggested on the basis of data in a very limited set of languages, but are then put forth as general accounts of how all human languages operate. When unexpected data from new languages come to notice there can be a number of reactions: ignore it; reinterpret the data so that it fits the theory; revise the theory so that it does explain the data; acknowledge that the theory cannot explain the data and as a consequence abandon it.

(I think this is a big reason why I dislike syntactic theories: they all seem to be based on this sort of overgeneralisation.)

I don't really see why you'd think it's overgeneralisation if you interpret data in the light of past conclusions. I really don't see why you'd think it's overgeneralisation if you adjust the theory in the light of new data.

In those cases, it isn’t overgeneralisation. But too often it feels like people are doing option 1, ‘ignore it’. And often when they do 2, ‘reinterpret the data so that it fits the theory’, it often feels like they’re forcing the data into a theory which really isn’t suited for it.

That reminded me that I was going to comment on a comment Whimemzs made:

the whole A/S/O (or A/S/P) concept for example has been very widely adopted by other functionalist typologists

An awful lot of liguists, including Chomskyan ones, use A/S/O (or P) as a sort of shorthand. How many agree that it's in any significant sense a primitive, and that, for example, "syntactic rules in every grammar are framed in terms of them" (Ergativity, 113)?

Well, do you have any example of a syntactic rule which can’t be framed in those terms? (Not trying to criticise your argument; I’m genuinely curious in a counterexample.)

(This is actually an issue where you do find people getting tripped up by their own theory, I think. For example, the A/S/O framework doesn't provide an obvious way to distinguish unaccusative from unergative verbs, so e.g., the very robust generalisation that it's only internal arguments / initial 2s / undergoer arguments that can be incorporated is easy to miss if you're working in terms of Dixon's theory.)

I always thought that it was only O and S arguments that can be incorporated, which fits perfectly in terms of Dixon’s theory. (I vaguely remember reading this in Whimemsz’s polysynthesis thread on the old board.) How is this incorrect?

[akam chinjir]

Well, do you have any example of a syntactic rule which can’t be framed in those terms? (Not trying to criticise your argument; I’m genuinely curious in a counterexample.) [...] I always thought that it was only O and S arguments that can be incorporated, which fits perfectly in terms of Dixon’s theory. (I vaguely remember reading this in Whimemsz’s polysynthesis thread on the old board.) How is this incorrect?

There's a robust generalisation that in languages that allow incorporation of S, it's only unagentive S that can be incorporated; you miss something pretty important if you don't distinguish unaccusatives from unergatives. (I've seen Ainu claimed as an exception to this generalisation, there may be others; I so far haven't tried looking into this.) Anyway, in languages that work this way, here's a rule that can't be stated in terms of S/A/O: only internal arguments (or initial 2s, or undergoer arguments) can be incorporated.

Another (from the Perlmutter article): in at least many languages with impersonal passives, only unergative verbs can form impersonal passives.

Another: in English, only unergatives can be made transitive by adding a resultative phrase ("I ran myself ragged," that sort of thing).

Or the rule (found in many languages) that only unergatives can take cognate objects.

Or rules that target transitives with two nonagentive arguments, like biabsolutive constructions or quirky subjects in some languages.

Active/stative and fluid-S alignments seem like they should also be on this list, to be honest.

[chris_notts]

I’ve managed to look through The Semantics of Clause Linking a bit more thoroughly now. I can definitely see what you said about it — every chapter except Dixon’s and Aikhenvald’s goes in turn through the same sequence of clause linking types: temporal, consequence, possible consequence, addition etc., even when the language itself divides up the clause types a bit differently. (Taking a random example: the Ojibwe chapter lists ‘conjuncts’ as being used with lots of different clause linking types, and I assume that these should really have been discussed together.)

On the other hand, I don’t see anything actually wrong per se with their categorisation of clause linking types, apart from maybe some slight overanalysis. And in fact, I see this categorisation as potentially being rather useful, for the purpose of enumerating the various ways that clauses may be linked across many languages. I think this is why I quite like Dixon’s work: from a conlanging perspective, it can be very useful to have this sort of exhaustive enumeration, where I can then look and it and say something like ‘for this conlang, how will I mark each of these types, which ones will have the same markers, and which categories will be conflated?’.

I know what Whimemsz means about Dixon's opinionated (and sometimes evidence free) writing style. Another example is Adjective Classes: A Cross-Linguistic Typology where Dixon effectively adopts the position of ultimate splitter. Any difference in morpho-syntactic behaviour, no matter how small, of a set of morphemes with adjective-like meanings, no matter how small the set, justifies claiming that a language has a separate class of adjectives, and as a consequence he makes the strong claim that all languages have a separate set of adjectives. On the other hand, he arbitrarily adopts the position of ultimate lumper in other areas with just as little justification.

On the other hand, I do like the series of books that he and Aikhenvald mostly edited because they bring together chapters on typologically diverse languages by functional domain (adjectives, subordination, clause linking, serial verbs, ...), which I think have more value for conlangers than typology books which just give scattered examples from different languages. You need a bit more of an in-depth description of individual languages to get a feel for how the system works as a whole, and reading individual language grammars, while also valuable, has the disadvantage that every author has their own theoretical framework and areas of interest so often it is harder to compare than a book with an editor bringing at least a little consistency. Dixon should also be mentioned for his magnum opus Australian Languages: Their Nature and Development. On the other hand, I was not particularly impressed by his series Basic Linguistic Theory: while I don't disagree with his general descriptive framework that much, there's a lot of opinionated waffle in there and it seems to me that the entire series could be condensed to a single book without losing much.

I'd actually like recommendations for other books which bring together chapters on a single functional area by language experts. I'm aware of other series of this type by Oxford and Cambridge university presses, e.g. the recent Oxford "handbook" series. Some examples:

The Oxford Handbook of Polysynthesis
The Oxford Handbook of Modality and Mood

Many of which have a section with specific language chapters. There are also of course various regional language surveys, but these differ in whether they devote chapters to languages or not. The Amazonian Languages (Dixon again) does, whereas others such as Foley's The Papuan Languages of New Guinea doesn't really (although it's nevertheless amazing for other reasons). Routledge's Language Family Series also generally has chapters by language, but the quality is variable to bad in my experience. Most of the chapters just go through the motions, with very cursory surveys of high-level features and random tables listing some inflectional forms, pronouns etc.

[bradrn]

Well, do you have any example of a syntactic rule which can’t be framed in those terms? (Not trying to criticise your argument; I’m genuinely curious in a counterexample.) [...] I always thought that it was only O and S arguments that can be incorporated, which fits perfectly in terms of Dixon’s theory. (I vaguely remember reading this in Whimemsz’s polysynthesis thread on the old board.) How is this incorrect?

There's a robust generalisation that in languages that allow incorporation of S, it's only unagentive S that can be incorporated; you miss something pretty important if you don't distinguish unaccusatives from unergatives. (I've seen Ainu claimed as an exception to this generalisation, there may be others; I so far haven't tried looking into this.) Anyway, in languages that work this way, here's a rule that can't be stated in terms of S/A/O: only internal arguments (or initial 2s, or undergoer arguments) can be incorporated.

Another (from the Perlmutter article): in at least many languages with impersonal passives, only unergative verbs can form impersonal passives.

Another: in English, only unergatives can be made transitive by adding a resultative phrase ("I ran myself ragged," that sort of thing).

Or the rule (found in many languages) that only unergatives can take cognate objects.

Or rules that target transitives with two nonagentive arguments, like biabsolutive constructions or quirky subjects in some languages.

Active/stative and fluid-S alignments seem like they should also be on this list, to be honest.

Thanks akam chinjir! I’m not sure I understand most of this, mainly because I still don’t really understand unergative/unaccusative verbs (I’ll have to reread the previous discussion, since I wasn’t really following it all that closely), but I definitely agree that active/stative alignments are a good example.

I’ve managed to look through The Semantics of Clause Linking a bit more thoroughly now. I can definitely see what you said about it — every chapter except Dixon’s and Aikhenvald’s goes in turn through the same sequence of clause linking types: temporal, consequence, possible consequence, addition etc., even when the language itself divides up the clause types a bit differently. (Taking a random example: the Ojibwe chapter lists ‘conjuncts’ as being used with lots of different clause linking types, and I assume that these should really have been discussed together.)

On the other hand, I don’t see anything actually wrong per se with their categorisation of clause linking types, apart from maybe some slight overanalysis. And in fact, I see this categorisation as potentially being rather useful, for the purpose of enumerating the various ways that clauses may be linked across many languages. I think this is why I quite like Dixon’s work: from a conlanging perspective, it can be very useful to have this sort of exhaustive enumeration, where I can then look and it and say something like ‘for this conlang, how will I mark each of these types, which ones will have the same markers, and which categories will be conflated?’.

I know what Whimemsz means about Dixon's opinionated (and sometimes evidence free) writing style. Another example is Adjective Classes: A Cross-Linguistic Typology where Dixon effectively adopts the position of ultimate splitter. Any difference in morpho-syntactic behaviour, no matter how small, of a set of morphemes with adjective-like meanings, no matter how small the set, justifies claiming that a language has a separate class of adjectives, and as a consequence he makes the strong claim that all languages have a separate set of adjectives. On the other hand, he arbitrarily adopts the position of ultimate lumper in other areas with just as little justification.

I was just reading that book actually, in order to reply to Whimemsz’s comment above. (See my comment above starting with ‘As for his endorsement of a universal adjective class…’.) I actually think he makes some really good points in there: for instance, he notes that most languages allow adjectives but not verbs in comparative constructions, a point which I think is often forgotten. It’s certainly convinced me that adjectives are distinct from verbs. But I think it would have been more accurate to say that adjectives always make up a distinct subclass of verbs, rather than saying that adjectives are a completely different word class to verbs (as Dixon concludes).

On the other hand, I do like the series of books that he and Aikhenvald mostly edited because they bring together chapters on typologically diverse languages by functional domain (adjectives, subordination, clause linking, serial verbs, ...), which I think have more value for conlangers than typology books which just give scattered examples from different languages. You need a bit more of an in-depth description of individual languages to get a feel for how the system works as a whole, and reading individual language grammars, while also valuable, has the disadvantage that every author has their own theoretical framework and areas of interest so often it is harder to compare than a book with an editor bringing at least a little consistency.

That series varies in quality — I’m reading Serial Verbs now, and it’s a truly excellent resource, but I didn’t like Complementation much. But I agree that the overall framework (typological description, then case studies) is good; it’s just their execution of it which varies. (And see Whimemsz’s comment on the weird editing in The Semantics of Clause Linking, which should be enough to make anyone suspicious.)

On the other hand, I was not particularly impressed by his series Basic Linguistic Theory: while I don't disagree with his general descriptive framework that much, there's a lot of opinionated waffle in there and it seems to me that the entire series could be condensed to a single book without losing much.

Pity… it’s the only book I’ve found so far which has in-depth, typological descriptions of many different areas of linguistics. For instance, I’ve been wanting to learn about the various types of demonstratives for a while, and the relevant section in Basic Linguistic Theory is the only resource I’ve seen. (There’s something to be said for being able to just go to a known place to get information, rather than having to search through the literature for some resource that may or may not exist.) Do you have any other recommendations for (preferably freely-available) textbooks like that? (And I do consider this sort of book to be a textbook.)

I'd actually like recommendations for other books which bring together chapters on a single functional area by language experts.

I would love some too — I’m always looking for such books to expand my linguistic knowledge. (I’ve even been thinking of making a thread to collect these sort of recommendations.) The only comparable resource that I know of and which wasn’t written by Dixon is Palmer’s Mood and Modality, which you can get from The Stack. (It probably isn’t quite what you’re looking for — it’s just a book on mood and modality, without any language-specific chapters — but it’s nonetheless an extremely useful resource.)

I'm aware of other series of this type by Oxford and Cambridge university presses, e.g. the recent Oxford "handbook" series. Some examples:

The Oxford Handbook of Polysynthesis
The Oxford Handbook of Modality and Mood

Many of which have a section with specific language chapters.

From a recommendation by akam chinjir (I think), I’ve been reading through The Oxford Handbook of Ergativity. I’ve been finding it remarkably useless compared to the other resources on ergativity I know of — lots of theoretical speculation of varying degrees of baselessness, and very few articles on typology (or indeed any other area which would be useful for conlanging).

[akam chinjir]

Thanks akam chinjir! I’m not sure I understand most of this, mainly because I still don’t really understand unergative/unaccusative verbs (I’ll have to reread the previous discussion, since I wasn’t really following it all that closely), but I definitely agree that active/stative alignments are a good example.

To be honest, I think you probably understand the substance of the distinction well enough, it's just that the terminology is awful (I agree with Ser about that!). I mean, in an active/stative system, it's the unergative verbs whose subject takes ergative case. That's the opposite of mnemonic! But it also means that if you understand active/stative, you probably understand unergative/unaccusative, just not under those names.

[Qwynegold]

Sigh, English is a mess. Even native speakers disagree on what's grammatical.

*The resigned CEO started a new company.

This was very surprising for me. How come this sentence is ungrammatical?

[Qwynegold]

To be clear, I only meant to present a test for identifying unaccusative verbs, not for identifying unergative verbs or for finding out whether a verb is not unaccusative. I think it's likely that not all unaccusative verbs have past participles that are easily used in this way.

Okay, so are you saying that if something can be past participled it's an unaccusative verb, but if it can't be, then we can't tell one way or the other?

[Richard W]

Sigh, English is a mess. Even native speakers disagree on what's grammatical.

Isn't that true of all natural languages with native speakers? After all, it's patently obvious (and hardly surprising) that many native speakers fail to learn their native language correctly. (Evolving a language often includes emulating the effects of this process.) Learning it well enough is good enough.

*The resigned CEO started a new company.

This was very surprising for me. How come this sentence is ungrammatical?

I think Zompist was thinking that 'resigned' means 'who had resigned', as opposed to one 'who was resigned to having to start a new company'.

[Qwynegold]

Ah, I've been meaning to read that for a long time now. I'll just post my replies on this thread first, or otherwise I'll forget what my thoughts were.

I agree that it’s probably a good idea to reply here first. As my ergativity thread, unfortunately it’s not done yet, so at the moment you can only read about half of it. But I’m working on it (admittedly very slowly…), and hopefully the next part should be posted soon!

Now I've gotten that far in that thread. I have what you call "S=A verbs" and "S=O verbs". I think.

What ‘different kinds of verbs’ are those? If you give a description, there may well be some well-known and less ambiguous terms that you could use for them.

Let's see. I'll just copy what I've written on FrathWiki. I will not post a link to the whole article, because there's a lot of obsolete stuff there that I haven't deleted yet.

Transitivity and voice
Each verb in Proto-Kunnu-lūjungo is inherently either transitive and intransitive. Intransitive verbs can further be divided into two types: Those that take an agent argument and those that take a patient argument. Hūjupk'a (run) is an example of an intransitive verb that takes an agent argument, the person who runs. Hyuk'yoya (freeze) is an example of an intransitive verb that takes a patient argument, the thing that turns into ice, not the person who is freezing something. The table below shows which kind of verbs can take which voice and transitivity suffixes, and what cases their arguments will have. The number in parenthesis corresponds to the example sentences in the following sections.

Code: Select all

                   INTR                       TR                  PASS                       CAUS
Intransitive verb  agent-ABS verb-INTR (1)    -                   verb-PASS (5)              causer-ABS causee-ALL
taking A                                                                                     verb-CAUS (7)

Intransitive verb  patient-ABS verb-INTR (2)  -                   -                          agent-ERG patient-ALL
taking P                                                                                     verb-CAUS (8)

Transitive verb    agent-ABS verb-INTR (3)    agent-ERG patient-  patient-ABS verb-PASS (6)  causer-ERG causee-ALL 
                                              ABS verb-TR (4)                                patient-ABS verb-CAUS (9)

Intransitives
The intransitive marker can be used on all kinds of verbs. The absolutive case is used on the sole argument of the intransitive verb.

Code: Select all

1) 	Kutsong-a 	myosyak'ya-tto-ng.
	dog-ABS 	jump-INTR-PRES
	The dog jumps.

Code: Select all

2) 	Sw-a 	satwa-tto-ng.
	wood-ABS 	burn-INTR-PRES
	The wood burns.

The intransitive marker can be used on inherently transitive verbs for the purpose of removing the object from the sentence. The reason for doing this can be because it is obvious from context what the object is, so it does not need to be mentioned.

Code: Select all

3) 	Pot'-a 	p'yowochya-tto-ng.
	1SG-ABS 	eat-INTR-PRES
	I eat.

Transitives
Transitive verbs are zero-marked. The "marker" may not be used on inherently intransitive verbs. If one wants to introduce another argument into a sentence with an intransitive verb, the new argument needs to have some non-core case.

Code: Select all

4) 	Pot'-ak' 	ubōtt-a 	p'yowochya-∅-ng.
	1SG-ERG 	apple-ABS 	eat-TR-PRES
	I eat an apple.

Passives
The passive voice removes the agent from the sentence. It can optionally be reinserted into the sentence with the use of the genitive case. The structure of a passive sentence is patient-ABS (agent-GEN) auxiliary-TNS main-verb-PST.PTCP. (Note that the passive voice is expressed by preceding the main verb with the auxiliary verb k'ūdi and following it with the past participle suffix -ttūk.)

Code: Select all

6) 	Ubōtt-a 	(pot'ya-t) 	k'ūd-ö 	p'yowo-ttūk.
	apple-ABS 	(1SG-GEN) 	PASS.AUX-PAST 	eat-PST.PTCP
	The apple got eaten (by me).

There are a few intransitive verbs, of the kind that take an agent argument, that can be passivized. In that case the verb becomes zero-valent. These verbs usually describe forces that are not in anyone's control.

Code: Select all

5) 	Rūjungo-du 	kurwa 	k'ūdi-ng 	sūnnat'-tūk.
	mountain-ADE 	hard.ADV 	PASS.AUX-PRS 	blow-PST.PTCP
	The wind is blowing hard at the mountain.

Causatives
The causative voice is used for describing what someone makes someone else do. The person who is made to do something is marked with the allative case. If the verb is intransitive, the causer gets absolutive case.

Code: Select all

7) 	Yazokk-a 	rallūlla-do 	kyot'sō-k'ya-ng.
	mother-ABS 	baby-ALL 	bathe-CAUS-PRS
	The mother bathes the baby.

But if the verb is transitive, then the causer gets ergative case and the direct object gets absolutive case. The order of arguments is causer causee DO verb. *The causer, causee and direct object are all optional, but at least one of the three must be present in a given sentence.

Code: Select all

9) 	Pot'-ak' 	myat-to 	tallūd-a 	kwe-k'ye-ng
	1SG-ERG 	3SG-ALL 	song-ABS 	hear-CAUS-PRES
	I make him hear a song.

When used on intransitive verbs that take patient arguments, the causative voice simply turns the verb transitive. The agent will then have ergative case, but the patient will still have allative.

Code: Select all

8) 	Pot'-ak' 	rōppo-do 	k'ūdi 	hyuk'yo-k'ya
	1SG-ERG 	water-ALL 	AUX.FUT 	freeze-CAUS
	I will freeze the water.

Now, as this is still very much a WIP, I'm not sure yet whether passive voice should be banned on "intransitive verbs taking P" or not. It makes sense not to allow it, because the passive voice deletes A, but these verbs don't even have an A to begin with. But that doesn't mean the morpheme can't be used, it would just have some other meaning in that case.

*The causer, causee and direct object are all optional, but at least one of the three must be present in a given sentence.

I'm not sure about this part. This is something I wrote a really long time ago, and I don't know if it still holds true or not.

I'm also not sure whether I will use the indicated noun cases in sentences of type (8). I've changes my mind several times about this.

I'll try to think of which type each of the verbs in my dictionary should be. (These lists will likely change a lot.)

"A-intransitives"
begin
work
get drunk
run
fish
go, traverse
walk
bathe
come
notice
jump
go
stand
row
fly
play
dance
cheat
fan
sit
contemplate
blow
sleep

"P-intransitives"
freeze (turn into ice)
freeze (feel cold)
die
bend
wake up
spoil
grow
float
be hungry
germinate
live
decay
fit
sound
burn
become numb
be

[Pabappa]

i would say all of the ungrammatical senses can be repaired by adding a word that does not change its part of speech. e.g.
"the since-resigned CEO", "the run-down dog", "the widespread disease" ... though the second one would still be *semantically* odd since we usually use that word for inanimate objects. "the pie sat" is arguable because my replacement would be "set", which is not transparently the same verb and hasnt been for thousands of years.

[Richard W]

"the pie sat" is arguable because my replacement would be "set", which is not transparently the same verb and hasnt been for thousands of years.

Are we requiring a transitive meaning for sit in this sentence? As I think Linguoboy implied, the verb is ambitransitive in British English. I may be missing the point, but "The people sat at the top table were not amused" is perfectly grammatical in my idiolect. Is that grammatical in US English? In British English it is ambiguous as to what induced them to sit at the top table - was their sitting there their choice, or pre-ordained?

[zompist]

*The resigned CEO started a new company.

This was very surprising for me. How come this sentence is ungrammatical?

I think Zompist was thinking that 'resigned' means 'who had resigned', as opposed to one 'who was resigned to having to start a new company'.

Right.

As to why the sentence is wrong, I'd have to do research! In general intransitive (and non-ambitransitive) verbs don't have passive participles that can be used as modifiers (*the slept woman, *the died man, *the come immigrants, *the existed creatures, *the happened events). But the current discussion is about possible classes of exceptions, and I'd really want to check a bunch of verbs rather than assume that a particular theory is correct. Plus, the observation about adverbs worries me; I'd want to check a bunch of variations to see what works.

Another complication is that participles easily become adjectives. E.g. Pabappa's "widespread" is not a participle any more— you couldn't say e.g. "*her legs are widespread" as a variation of "her legs are spread wide." I'd also say "run-down" is lexical... "run down" now means "run over", not "exhaust".

[zompist]

"The people sat at the top table were not amused" is perfectly grammatical in my idiolect. Is that grammatical in US English?

Not for me; it'd have to be "The people sitting..."