Notes on Language Creation

Language creation can be a tough thing to get a hold of. There is no "How To" book for language creation. Everyone has their own opinions; everyone has good ideas. These are a few of mine (opinions, not good ideas—the latter's for you to decide).

Ergativity: The Maltese Falcon of language creation. If you'd like a linguistic definition, you can go here, but it probably won't help much. Essentially (and you should take that word with a bucketful of kosher salt), ergativity is this: In English (a nominative-accusative language), the subject of a sentence with a transitive verb and the subject of a sentence with an intransitive verb are treated alike; direct objects of transitive verbs are treated differently. In an ergative-absolutive language, the subject of an intransitive verb is treated the same as the direct object of a transitive verb; subjects of transitive verbs are treated differently. That, however, is only the very tip of the flap on top of the roof on top of the house on top of the iceberg. In fact, that definition is wholly inadequate when it comes to explaining ergativity, but many don't know why. That's fine if you're a doormat salesman; not so fine if you're a conlanger who wants to create an ergative-absolutive conlang.

In this introduction to ergativity, I'll try to explain what exactly ergativity is, and how it's manifested in natural languages, as well as how it can be used in created languages. I will be drawing on a number of resources which I'll mention throughout this introduction, and will also list at the end.

So, without further ado, I give you: Ergativity.

•1.0 Introducing Terms:

Before jumping into theory and examples, I want to make sure that we've got our terms straight.

First of all, there are the terms "nominative-accusative language/system" and "ergative-absolutive language/system". Each of these refer to a language that display either non-ergative or ergative characteristics. This does not mean that the language in question will have cases with these names. After all, English is a nominative-accusative language, but has no case (except in the pronouns, and those cases work differently than standard nominative-accusative).

With that said, the names that are given to these systems do come from somewhere. Specifically, the four words used in the system names are case names. The nominative case that identifies the subject (regardless of the valency of the verb) in nominative-accusative languages. The accusative case is a case that (usually) marks the direct object of a transitive verb in nominative-accusative languages. The absolutive case is a case that marks the subject of intransitive verbs and the direct object of transitive verbs in ergative-absolutive languages. Finally, the ergative case is the name for a case that marks the subject of a transitive verb (not necessarily the agent) in ergative-absolutive languages.

Actually, since I introduced a semantic term up above, it might be useful to go over the relevant ones. An agent is, strictly speaking, the initiator of an action. In this section, I'll be referring to the agent of a transitive verb as an A. Now, in a sentence like, "The polar bear's dancing", "the polar bear" is actually an agent—i.e. he's initiating the dancing action. I'll be referring to those types of arguments (i.e. the volitional/agentive subjects of intransitive verbs) as SA. A patient is the undergoer of an action. So, for example, in "The polar bear tapped the panda", "the panda" is the one who undergoes the tapping action. I'll be referring to these types of patients as P. Another type of patient would be "the door" in a sentence like "the door swung open". I'll be referring to these types of patients as SP. Three other semantic roles I'll be talking about are recipients (R), experiencers (E) and stimuli (ST). I'll explain these when I get to them. The prior four, though, will be important to remember as we go along.

Two processes I'll be discussing later on are passivization and antipassivization. I think it might help just to think of these as a simple valency-decreasing operation, but one typically applies to nominative-accusative languages, and the other typically applies to ergative-absolutive language. Both of these processes affect transitive verbs. The process takes the default argument and turns it into an oblique, and takes the specially marked argument and turns it into the default argument. In a nominative-accusative language, nominative is the default marking; accusative the special marking. In an ergative-absolutive language, the absolutive is the default marking; the ergative the special marking. The resulting verb is a very intransitive-like verb, in both cases. That's all this is.

Okay, those are some terms that we need to make sure we're all on the same page about. (Heh. How's that for a sentence ending with a preposition?) If you're not sure how I'm using a term later on, come back here, and it will explain.

•1.1 Introducing Some Test Words:

In explaining (and hearing explanations of) ergativity, I've always found it more helpful to look at invented examples than actual examples from natural languages. I will talk about natural languages below, but most of the examples will be shown using the words listed below. The words below will be used to illustrate all examples, so that we're not switching languages from example to example, and so that it'll be easier to familiarize yourself with what exactly is going on. Or that's the plan, at least. So below are a list of words from a language that we'll call Ergato:

English	Ergato	English	Ergato
I	ko	panda	palino
you	pe	fish	tanaki
she	li	sheep	folime
to dance	talu	man	hopoko
to sleep	sapu	woman	kelina
to pet	lamu	book	kitapo
to see	fisu	wind	makipo
to give	kanu	house	paleni
and	i	General Preposition	sa
Valency Reducing Marker	-to	Oblique Marker	-k
Past Tense Marker	-ri	Recipient/Dative Case	-s
Plural Marker	-ne	Extra Case Marker	-m
Default Case Marker	—	Special Case Marker	-r

It's important to understand why the markers above do not say things like "ergative case marker", or "antipassive marker". These markers are going to be used differently in different contexts in the examples below. Thus, the "special case marker" will show up as both an accusative case marker and as an ergative case marker. Now I'll start in with the examples.

•2.0 The Pristine System:

There are a lot of conlangs out there that are, essentially, pristine systems (note: this is my term). A pristine system, when talking about language, is a system where there are no irregularities, and everything works the same way, no matter the context. This is ideal for an IAL, or a loglang. If your goal is to create a natural language, though, a pristine system is something to be avoided, because no natural language is pristine (not even Turkish). Nevertheless, a pristine system (or an attempt at a pristine system) is what many first-time conlangers aim for (most of the time unconciously). I'm now going to show you what a pristine nominative-accusative system and a pristine ergative-absolutive system looks like. I'll start with a nominative-accusative system.

•2.1 A Pristine Nominative-Accusative System:

Before I begin, I want to say that I'm assuming that a pristine system will utilize case marking, because, when it comes to conlangs, that's usually the case. There is such a thing as a pristine language that doesn't use case marking, but I'll get to those later. So now for the pristine nominative-accusative language. To test for pristineness (pristinity?), there are some general sentences you can use. You will want to test:

A sentence with an intransitive verb with a patient-like subject (SP).
A sentence with an intransitive verb with an agent-like subject (SA).
A sentence with a transitive verb with an agentive subject (A).
A sentence with a transitive verb with an experiencer subject (E).
A sentence with a ditransitive verb.

So, let's test those sentences in pristine nominative-accusative Ergato:

Kelina sapu. "The woman is sleeping."
Kelina talu. "The woman is dancing."
Kelina lamu palinor. "The woman is petting the panda."
Kelina fisu palinor. "The woman sees the panda."
Kelina kanu kitapor hopokos. "The woman's giving the book to the man."

The above is extremely indicative of a pristine nominative-accusative system. The thing that tips you off to its being a nominative-accusative system is that the subject kelina, "woman", is in the same case (the default case) in sentences (2a), (2c) and (2e). The thing that lets you know that the system is pristine is that kelina is in the same case for sentences (2a) and (2b), and also for sentences (2c) and (2d). English is not a pristine system when it comes to this criterion, though it's not because of case. Take the two translations of sentences (2c) and (2d) above and compare each to its incorrect counterpart in English below:

The woman is petting the panda.
*The woman pets the panda.
The woman sees the panda.
*The woman is seeing the panda.

Sentences (3b) and (3d) above are grammatical, but they don't mean the same thing as sentences (3a) and (3c), respectively. This is because in the present tense English is sensitive to whether the subject is an experiencer (E) or an agent (A). Instead of it being marked as a case, it's marked with the presence or absence of the auxiliary "be".

Now, it's not enough to merely test the sentences in (1) to determine whether or not the system is pristine. I'll explain more about why this is later. Suffice it to say that you should also test:

A sentence with a pronoun as the subject of a transitive verb.
A sentence with an inanimate noun as the subject of a transitive verb.
A sentence in the past tense with a transitive verb.

So, let's test those quickly in pristine nominative-accusative Ergato:

Li lamu palinor. "She's petting the panda."
Kitapo lamu palinor. "The book's petting the panda."
Kelina lamuri palinor. "The woman petted the panda."

Now, with sentence (5b), you're going to have to use your imagination. So let's say a woman has a very clean panda that she doesn't want people petting with their hands (because hands have germs). So, not wanting to offend her (or her panda), you pick up a book and kind of stroke the panda with it. Suddenly, the woman asks, "What are you doing?" You reply, "I'm petting your panda." "With your filthy hands?!" she screams. You reassure her, "No, no. The book's petting the panda." Far-fetched, but it will serve our purposes.

Anyway, the point is that nothing has changed with respect to case marking. The subject of the sentence still gets default marking, and the object still gets special marking.

Based on all this evidence, you can determine that the system is a nominative-accusative system, and that it's pristine. That is, the subject of the sentence will always get default marking, no matter what the tense is, or what kind of verb it is, what tense, animacy, etc. It's hardcore nominative-accusative. And that means that you can safely label the -r suffix as being an accusative marker.

Now that we've determined what kind of system we have, let's look at the valency-reducing mechanism. This will only apply to verbs that have at least two arguments: A subject and object (however they're marked, casewise). So we can ignore intransitive verbs for now. So let's look at a couple sentences:

Kelina lamu palinor. "The woman's petting the panda."
Palino lamuto (kelinak). "The panda's being petted (by the woman)."
Kelina kanu kitapor hopokos. "The woman's giving a book to the man."
Kitapo kanuto hopokos (kelinak). "The book's being given to the man (by the woman)."

So, a few things to notice. The first and most obvious thing to notice is that what was the object in the transitive sentence (marked with -r) is now the subject in the passivized sentence (now given default marking). Second, the verb is marked with -to, to let you know the passivization process has occurred. Third, the actual subject of the sentence has been made superfluous. That is, just as you can say "The panda's being petted", so can you say Palino lamuto in this version of Ergato. Expressing the actual subject is optional. Finally, with respect to that optional subject, notice that if you do express it, it's no longer in subjective case (default marking/nominative), but in an oblique case. This is the case for just about every language that has a passive. What will change is what that oblique case is. So, in English we just have a prepositional phrase headed by "by". In Turkish, you have something similar, only with a postposition. The point is that the noun will be marked in some totally different way, and will be treated a different way by the syntax.

Well, that's about it for pristine nominative-accusative Ergato. So, on to pristine ergative-absolutive Ergato!

•2.2 A Pristine Ergative-Absolutive System:

This should go a lot faster. In section 2.1, I wanted to explain why we were doing a lot of the things we were doing. Now that you know, though, we can right to the examples. So, here are our initial batch of test sentences:

Kelina sapu. "The woman is sleeping."
Kelina talu. "The woman is dancing."
Palino lamu kelinar. "The woman is petting the panda."
Palino fisu kelinar. "The woman sees the panda."
Kitapo hopokos kanu kelinar. "The woman's giving the book to the man."

Immediately, something should jump out at you as being radically different. Aside from the case marking, the subject is appearing in totally different places! This is because this system is pristine. A truly pristine system would line up cases on the same side of the verb, no matter what. So the equivalent to the pristine nominative-accusative system is an ergative-absolutive system where the absolutive case (now the default marked case) always comes before the verb, the ergative case (now the -r case) always comes after the verb, regardless of whether it's the subject of the sentence or not. A good many first-time ergative languages are not pristine, but usually it's unconcious, because, since English is a nominative-accusative language with no case marking, it seems natural to always put the subject on the same side of the verb. That's not the way a pristine ergative-absolutive system would work, though.

Now that we've hurdled that...hurdle, we can talk about the other differences. Most notably, the subject of the sentence is being marked differently depending on whether it's in a sentence with a transitive verb or a sentence with an intransitive verb. Notice, though, that this system isn't sensitive to the status of the subject. So in an intransitive sentence, the subject is marked with the absolutive, regardless of whether it's an SA or an SP. Similarly, in a transitive sentence, the subject is marked with the ergative, regardless of whether it's an A or an E.

Let's quickly look at our other test sentences:

Palino lamu lir. "She's petting the panda."
Palino lamu kitapor. "The book's petting the panda."
Palino lamuri kelinar. "The woman petted the panda."

As you can see, there's no change in case marking, or in the placement of the subject.

Now on to antipassives. Antipassives seem to really confuse a lot of folks, and I think it's because, to a nominative-accusative speaker, there doesn't seem to exist a conceivable reason to ever use an antipassive. The usual example from English used to try to explain antipassives is the verb "eat". So, you can say "I ate breakfast", or you can say "I ate". Thus, the object is kind of superfluous. This, however, is not the same thing, and that's not why antipassives are used. I'll do my best to explain here.

To begin with, let's actually see some antipassive sentences. Here goes:

Palino lamu kelinar. "The woman is petting the panda."
Kelina lamuto (palinok). "The woman is petting (and what she's petting is the panda)."
Kitapo hopokos kanu kelinar. "The woman's giving the book to the man."
Kelina hopokos kanuto (kitapok). "The woman is giving to the man (and what she's giving is a book)."

I used those convoluted translations in (9b) and (9d) to try to show how the optional phrase in an antipassive feels to the speaker. It really is extra, unnecessary information.

Anyway, notice what happened. If the absolutive is the default, unmarked case, and the ergative is the special, marked case, what an antipassive did was got rid of the special case. Thus, you might say that there's less mental work involved when it comes to case in antipassives (maybe). Also, an antipassive allows you to focus on one aspect of the action, in this case, the performer of the action. Finally, think about why we use passives in English most of the time. If you think about it, the usual reason to use a passive is if you want to conjoin things in discourse. So, let's say we're talking about an accident where one car is at fault (i.e. it hit the other one). I might say, "I saw the car that was hit". I probably would never say, "I saw the car that the car at fault hit it" (that's probably not even grammatical). The second sentence is how you'd have to say it, though, if there were no passive. Why? Because when two sentences are conjoined in English, the subjects go together. So, if you say, "The Toyota hit the Honda and skidded", the car that skidded has to be the Toyota, and could never be the Honda. The same kind of thing happens in ergative-absolutive languages, but instead of the subject being carried over, it's the absolutive argument. Maybe an example will help explain:

Palino lamuri kelinar i [palino] talu. "The woman petted the panda and [the panda] danced."
*Palino lamuri kelinar i [kelinar] talu. "*The woman petted the panda and [the woman] danced."

That is, in my opinion, probably the reason why valency-reduction systems exist. If you don't have them, everything you say becomes extremely roundabout. For example: "Yesterday, there was an accident that I saw. A Toyota came and smacked a Honda and the Honda skidded along the street. Later on, I saw the car, such that the Toyota hit it. The Toyota had banged it up pretty badly. The Toyota made it such that its trunk wouldn't close, and also made it such that one couldn't see out of its rear window." If you allow for valency-reduction (in this case, passivization), the whole thing becomes much shorter and easier to understand. In this way, antipassivization is no different from passivization. Think of it as a kind of luxury. After all, not all languages have valency-reduction systems. You best thank your lucky stars that your language does! (Or, well, that the language you're reading right now does.)

•3.0 Syntactic Ergativity:

You know, I think it'd be easier to explain syntactic ergativity before going on to split-ergativity. So I'll do that. I'm going to explain how pristine syntactic nominative-accusative and ergative-absolutive languages work, because, basically, it's identical to what's above, but without the case-marking.

•3.1 A Pristine Syntactic Nominative-Accusative System:

English is just about a pristine syntactic nominative-accusative system. Almost. Its sensitivity to experiencer verbs in the present and its pronouns are the only thing standing in the way. Close, though.

I'm just going to list the sentences. Note that when I say syntactically nominative-accusative or ergative-absolutive, it means that relations are determined by word order. So here's pristine syntactic nominative-accusative Ergato:

Kelina sapu. "The woman is sleeping."
Kelina talu. "The woman is dancing."
Kelina lamu palino. "The woman is petting the panda."
Kelina fisu palino. "The woman sees the panda."
Kelina kanu kitapo hopoko. "The woman's giving the book to the man."

In the examples above, the object comes after the verb, and the subject before, in all cases. In the case of an indirect object, it's put after the direct object (remember: this is a pristine system. If the direct object is going to come after the verb, it should always come directly after the verb). Aside from sentence (11e), this should look a lot like English. Now for the next set:

Li lamu palino. "She's petting the panda."
Kitapo lamu palino. "The book's petting the panda."
Kelina lamuri palino. "The woman petted the panda."

Again, not different from English. If this were a purely syntactic language (i.e. isolational), you might expect the past tense suffix to be a past tense word, but that really doesn't have any bearing on what we're doing now. So, now for the last set:

Kelina lamu palino. "The woman's petting the panda."
Palino lamuto (sa kelina). "The panda's being petted (by the woman)."
Kelina kanu kitapo hopoko. "The woman's giving a book to the man."
Kitapo kanuto hopoko (sa kelina). "The book's being given to the man (by the woman)."

In these examples, the preposition is used to indicate the demoted subject, just like English "by". Notice that the demoted subject comes after the indirect object (which now sits next to the verb) in (13d).

Well, that really does it for pristine syntactic nominative-accusative Ergato. The important thing to notice is that what is what is wholly dependent upon word order. We'll see more of the same with pristine syntactic ergative-absolutive Ergato below.

•3.2 A Pristine Syntactic Ergative-Absolutive System:

Now we can see the flip-side of the pristine syntactic coin. Here's the first set of examples:

Kelina sapu. "The woman is sleeping."
Kelina talu. "The woman is dancing."
Palino lamu kelina. "The woman is petting the panda."
Palino fisu kelina. "The woman sees the panda."
Kitapo hopoko kanu kelina. "The woman's giving the book to the man."
Kelinane sapune. "The women are sleeping."
Palino lamu kelinane. "The women are petting the panda."
Palinone lamune kelina. "The woman is petting the pandas."

Here the absolutive argument always comes sentence-initially, and the ergative argument always comes directly after the verb. Also, you should know that the placement of arguments (i.e. where the absolutive argument goes, where the verb goes, etc.) is totally arbitrary. As long as those places are honored no matter what happens, the system is considered pristine. Also, if you take a look at the last three sentences, note that the plural agreement marker -ne is placed on the verb when the absolutive argument is plural, and at no other time. You can think of this as subject agreement, except that rather than S and A constituting your subject, S and P constitute your subject. Now let's look at our secondary examples:

Palino lamu li. "She's petting the panda."
Palino lamu kitapo. "The book's petting the panda."
Palino lamuri kelina. "The woman petted the panda."

Again, these extra facets don't affect the position of the arguments in the sentence. Now for our antipassive examples:

Palino lamu kelina. "The woman is petting the panda."
Kelina lamuto (sa palino). "The woman is petting (and what she's petting is the panda)."
Kitapo hopoko kanu kelina. "The woman's giving the book to the man."
Kelina hopoko kanuto (sa kitapo). "The woman is giving to the man (and what she's giving is a book)."

Here again, in these examples, the absolutive and ergative arguments are switching places, and the demoted absolutive argument (the old one) is optionally expressed as a PP headed by our all-purpose preposition sa.

And that's how a syntactically ergative language works. Rather than looking at case marking, you look at word order, and how the different arguments show up in different types of sentences. Admittedly, it's probably easier to see this kind of thing when there's case marking, but not all languages mark case overtly. Plus, a syntactically ergative conlang would be a real rarity; quite unique.

Now it's time for the tough stuff.

•4.0 Split-Sensitivity:

I'm calling this section "split-sensitivity" because all languages show split-sensitivity to something to some degree. I've already shown an example from English. Even though it's nominative-accusative, it's sensitive to experiencer verbs in certain situations, but not in others (e.g. in the past tense). Split-sensitivity is a blanket term for any language that shows one kind of pattern in one place, and a different kind of pattern in a different place. That's all. The thing that characterizes these languages is: (a) What is split (case marking, for example); and (b) where the split occurs. We'll now delve into split-sensitivity.

•4.1 Tense-Based Split-Ergativity:

One of the most common types of ergativity is ergativity that's split based on tense. Hindi and Georgian both display this kind of ergativity. The most common way to split it is so that in the present tense (or nonpast), the language displays a nominative-accusative system, and in the past tense, the language displays an ergative-absolutive system. So let's focus on that kind of split and see what our test sentences look like:

Kelina sapu. "The woman is sleeping."
Kelina talu. "The woman is dancing."
Kelina lamu palinor. "The woman is petting the panda."
Kelina fisu palinor. "The woman sees the panda."
Kelina kanu kitapor hopokos. "The woman's giving the book to the man."

All these sentences are in the present tense, so, unsurprisingly, they look just like the sentences in (1). Now here's where the difference lies:

Li lamu palinor. "She's petting the panda."
Kitapo lamu palinor. "The book's petting the panda."
Palino lamuri kelinar. "The woman petted the panda."

Now let me stop right here to explain some things. What you see above is what you'd expect if you were melding two pristine systems (i.e. where the word order and case marking are just like those in the pristine ergative-absolutive version of Ergato). This is not usually the case, though. First off, it's much more likely that the subject of the sentence would be in the same place. Thus:

Kelinar lamuri palino. "The woman petted the panda."

Second, though it would be economical to use the same case marker to mark the accusative and ergative, the ergative languages I know of (I'm thinking of Georgian in particular) don't. Instead, what you'd see is something like this:

Kelina lamu palinor. "The woman's petting the panda."
Kelinam lamuri palino. "The woman petted the panda."

In effect, what you have is three case markers. One case marker (the default marker) marks the nominative in the present and the absolutive in the past. Another, the special marker -r, marks the accusative in the present. Then you have a third, the extra case marker -m, which marks the ergative in the past. This is exactly the type of system that Georgian has (give or take the lack of an accusative marker that's distinct from the dative, and the inappropriate use of the word "tense").

As you might expect, the valency-reduction mechanism works differently in the present and past. However, here there are further wrinkles. This is how one might imagine the system would work:

Palino lamuto (kelinak). "The panda's being petted (by the woman)."
Kelina lamurito (palinok). "The woman was petting (and what she was petting is the panda)."

That would be a nice way for it to work. And maybe there are some that do. However, there are theories about the evolution of some ergative-absolutive systems that suggest that ergativity in the past tense arose from present tense passive constructions. So what you might get would look something like this:

Kelina lamu palinor. "The woman's petting the panda." (Present Tense, Active)
Kelinak lamuto palino. "The woman petted the panda." (Past Tense, Active)
Palino ke lamu (sa kelina). "The panda's being petted (by the woman)." (Present Tense, Passive)
Palino ke lamuto (sa kelina). "The panda was being petted (by the woman)." (Past Tense, Passive)

So remember what those markers mean. The first sentence is standard issue. The second sentence, however, might look like a passive. According to some theories (I've heard this about Hindi, but it is just a theory), what happened was that the passive was used so often that it became the past tense, and so the valence-reducing marker -to now functions as (and, well, is) the past tense marker. But since it was a passive, the subject is marked with the oblique case (that's what the -k is). And, of course, in a standard passive, the promoted object is marked with the subjective case. When this construction becomes the normal past tense, though, the word order falls in line (subject first; object last), and so you get what looks like an ergative-absolutive system only in the past tense. Then what I wanted to show with sentence (22c) is that some new construction would arise to fulfill the role of the present tense passive. So, ke in that example would be some kind of auxiliary, and the reintroduced subject would be reintroduced by a "by" phrase, like English, rather than being expressed with the oblique (now ergative) case marker. Then, in the past tense...who knows? (22d) is my guess as to what could happen to create an antipassive. It might be advisable to see what Hindi does. (I'll check on that.)

Now, this subsection is devoted to ergativity split by tense, not just past tense. The thing is, I've never heard of a split-ergative language that splits it (based on tense) any other way. This could partly be because of the theory I mentioned above. That theory aside, though, this split could work the opposite way: Ergative-absolutive in the present; nominative-accusative in the past. Or maybe even the future. It could be an aspectual split: perfective vs. imperfective. It's perfectly possible. This is just the most common. Georgian does something that really isn't best described as a split system based on tense. This is because what constitutes "tense" in Georgian is incredibly complex. Each verb can be conjugated in 12 or 13 different ways, and these ways are divided into three series: present, aorist and perfect. If I remember right (I'll check my notes and get it straight later), it's the perfect series that displays an ergative-absolutive pattern, whereas the present and aorist series display a nominative-accusative pattern. Anyway, in the case of Georgian, I'd argue that the split isn't based on tense, but on morphological category. The Georgian system is a fascinating system for many reasons. You might go here for more information, or look up Stephen R. Anderson's paper on case in Georgian (though don't take it too seriously).

•4.2 Pronominally-Based Split-Ergativity:

Another common way to have a split system is to have one kind of system that's used with overt nominals, and to have a different system used with pronouns. A prime conlang example of this kind of system is the masterful David Bell's ámman îar (click here to go directly to the part that explains the ergativity of ámmar îar). A lot of ergative languages do this, but often it's mixed with an animacy (or, as Payne calls it, "agency-worthiness") system, which I'll describe later.

The basic concept behind a system where the split is based on whether you have a pronominal argument or an overt NP isn't that hard to imagine. For this example, let's say that Ergato displays an ergative-absolutive pattern for overt nominals, and a nominative-accusative pattern for pronouns. Here are our example sentences:

Kelina sapu. "The woman is sleeping."
Kelina talu. "The woman is dancing."
Kelinam palino lamu. "The woman is petting the panda."
Kelinam palino fisu. "The woman sees the panda."
Kelinam hopokos kitapo kanu. "The woman's giving the book to the man."

I changed the word order to a (in my mind) more natural word order for an ergative-absolutive language. So now there's a dominant SOV word order, but the case marking on the subject changes, so that you get an -m when the subject is an A. Other than the word order, though, the sentences in (23) are identical to those in (7). [Note: I'm going to go ahead and continue using -m as the default ergative marker when A's and P's are marked separately.] Now let's look at our secondary test sentences:

Li palino lamu. "She's petting the panda."
Kitapom palino lamu. "The book's petting the panda."
Kelinam palino lamuri. "The woman petted the panda."

Check out sentence (24a). The only way you know which is the subject and which the object is the word order. But that's not the whole story. So far we've sentences with two overt NP's and one with a subject pronoun and object NP. Now let's look at an intransitive sentence with a subject pronoun, and two transitive sentences, one with a subject NP and an object pronoun, and the other with two pronouns:

Li sapu. "She's sleeping."
Palinom kor lamu. "The panda's petting me."
Li kor lamu. "She's petting me."

In (25), you can see the fully fleshed out version of a pronominally split-ergative language. A and S pronouns are marked just like S and P NP's, and P pronouns have a special accusative marker.

So now we come to valency-reduction. I have no information at hand that addresses what I want to know (e.g. what happens with split-ergative systems and passivization/antipassivization). The only examples that Payne lists of antipassivization in his otherwise fantastic book Describing Morphosyntax are from languages that are entirely ergative-absolutive. Thus, I'll list what a language might do, or could conceivably do:

Li (kelinak) lamuto. "She's being petted (by the woman)."
Kelina (lik) lamuto. "The woman's petting (her)."

What I've shown in (26) is, essentially, a subject controlled valency-reduction system. In other words, depending on what the subject of the sentence is, that determines whether the result is interpretted as a passive (in the case of a pronominal subject) or as an antipassive (in the case of an overt NP subject). It's also possible that you might have two different kinds of systems. So, maybe you have a normal antipassive system for NP's, and then a different kind of antipassive system for pronouns. Either way could work. (Note: David Bell's pronominally split-ergative language ámman îar appears to have taken a semantic approach to valence functions, as opposed to morphological. In other words, you can make any transitive sentence into a passive sentence or an antipassive sentence regardless of case marking. Go here for a thorough account.)

The example I showed above featured an ergative-absolutive system for overt NP's, and a nominative-accusative system for pronouns, but it could easily go the other way. Additionally, you could have different systems for different pronouns, but I'll discuss that in more depth when we get to the section on animacy.

One last thing I want to mention (something that doesn't deserve its own section) is person marking on verbs. Person marking on verbs can work exactly the same way as separate pronouns. My language Sathir is a language that works this way (the language is ergative, but pronominal subjects are marked on verbs, whether they're A's or S's). If we wanted to use Ergato as an example, we could pretend that the pronouns were pronominal suffixes (for one type), and suffixes and prefixes (for a different type). Here's an example where subjects are marked on verbs if they're not overtly specified. The case marking system is ergative-absolutive. This yields:

Kelina sapu. "The woman's sleeping."
Kelinar palino lamu. "The woman's petting the panda."
Sapuko. "I'm sleeping."
Palino lamuko. "I'm petting the panda."

In the above example, the NP's show normal ergative-absolutive case marking (S and P get default marking; A special), but subjects are marked the same way regardless of their status. That's one way it could work. Now imagine a language where NP's are marked in a nominative-accusative way, and verbs inflect for both subject and object. Here's what that could look like:

Kelina sapu. "The woman's sleeping."
Kelina palinor lamu. "The woman's petting the panda."
Sapuko. "I'm sleeping."
Palinor kolamu. "I'm petting the panda."
Kolamupe. "I'm petting you."

The sentences in (28) are essentially a variant on the word order model. The point is that, in transitive sentences, subjects are inflected with a prefix and objects are inflected with a suffix. In intransitive sentences, subjects are marked with a suffix, just like objects in transitive sentences. At the same time, overt NP's are marked in a traditional nominative-accusative way. This same effect could be achieved (and often is) by having different forms of pronominal inflection for the different roles. Here, though, I wanted to keep it simple.

I think that about does it for pronouns. We'll revisit pronouns when we discuss animacy.

•4.3 Semantically-Based Split-Ergativity:

This type of split is extremely common in all the world's languages, though usually in small doses. Essentially, this type of split is a split that causes similar arguments with different semantic roles to be marked differently. The example of this I already discussed is English's sensitivity to verbs of experience in the present tense. But that's not the whole story. Not by a long shot.

Let's start off with something simple. This is what English's pattern might look like in a case-marking language:

Kelina sapu. "The woman is sleeping."
Kelina talu. "The woman is dancing."
Kelina lamu palinor. "The woman is petting the panda."
Kelinas fisu palino. "The woman sees the panda."
Kelina kanu hopokos kitapor. "The woman's giving the man a book."

Above, the word order doesn't change, but notice that the case marking on the subject of (29d) is dative case marking, just like the case marking on the indirect object of (29e). This is a common occurrence in the world's languages, where an experiencer subject gets marked as a recipient of some kind. Additionally, the object of (29d) is marked with the nominative, or default case. Now, the above system, like English, makes sure to line up the subject. A different language, though, might make sure to line up the case, instead, yielding the following:

Kelina sapu. "The woman is sleeping."
Kelina talu. "The woman is dancing."
Kelina lamu palinor. "The woman is petting the panda."
palino fisu kelinas. "The woman sees the panda."
Kelina kanu hopokos kitapor. "The woman's giving the man a book."

The reason for the above would be that, grammatically (or morphologically), palino in sentence (30d) is the subject, and, therefore, should line up with the other subjects. It really depends on how the language defines the notion of subject.

Now how about this. We've seen three different case markers employed in one system: Default, -r and -m. Thus far, though, we haven't seen them all in the same tense. Can it happen? You bet it can. This is what it would look like:

Kelina sapu. "The woman is sleeping."
Kelina talu. "The woman is dancing."
Kelinam lamu palino. "The woman is petting the panda."
Kelina fisu palinor. "The woman sees the panda."
Kelinam kanu hopokos kitapo. "The woman's giving the man a book."

In this admittedly bizarre system, S's are marked the same way as P's (default marking), and A's are marked with -m. Then, possibly for semantic reasons, E's are marked the same as S's and P's, and ST's (stimuli) are marked with a third case, -r. That's really a bizarre system. Here's a more normal one that a large number of natural languages have:

Kelina sapu. "The woman is sleeping."
Kelinam talu. "The woman is dancing."
Kelinam lamu palino. "The woman is petting the panda."
Kelina fisu palinor. "The woman sees the panda."
Kelinam kanu hopokos kitapo. "The woman's giving the man a book."

Here's a system where's there's a distinction drawn between SA's (agent-like subjects) and SP's (patient-like subjects). In (32a) and (32d), the subjects of those verbs are more like patients than agents, so they get default marking, as do normal P arguments. The subjects of (32b), (32c) and (32e), though, are more agent-like (after all, one hopefully doesn't dance by accident). Thus, they're marked with -m. Finally, ST's are marked with -r. (Note: For what it's worth, I think this marking may be optional. Stimuli could very well be marked with the default case—or even with -m, possibly.)

Since we brought up SA's and SP's, I'd like to mention a little fact that can pop up in many different systems. Let's say volitionality is important to a given language. Thus, SA's are marked with an ergative marker (say, -m), and SP's are marked with an absolutive marker (default marking). This could be a hard-and-fast rule, or the language can use the volitionality generalization to its advantage. Consider this possibility:

Kelina sapu. "The woman is sleeping."
Kelinam sapu. "The woman is sleeping on purpose."
Kelinam talu. "The woman is dancing."
Kelina talu. "The woman is dancing on accident."

I could use other verbs that would make more sense here, but I'd rather not use too many different made-up words. Instead, I'll make up different contexts. So, for (33b), let's say the woman isn't so much a woman, but a young girl. It's Sunday morning, and she's woken up, but she knows tomorrow is Monday, and she remembers how nice it is to just laze about in bed. But she hears that her mother has awakened... And her mother wants to make her go to church, thereby ruining her lazy morning. As if on cue, in walks her mother to say, "Get up, Hildegarde: It's time for church." Oh, but young Hilde's concocted a fiendish plan. "Perhaps if I pretend I'm asleep," she thinks, "my mother will leave without me, not wanting to be late." And thus, Hildegarde attempts to sleep on purpose, as to fool her mother. That's context number 1 for sentence (33b). [Incidentally, this rarely works. I've heard.]

Now, for (33d). Imagine a dance at a high school gym—let's say, Pacifica High School's gym, located in sunny Garden Grove, CA. Now imagine that there's a woman (or girl) there who doesn't want to dance because she's afraid she won't be that good and doesn't want to embarrass herself. She's by no means unpopular. Several boys (yes, and even a girl or two) have asked her to dance, but she's systematically declined each one, citing the weather, an obscure religion, uncomfortable heels, a full bladder, etc. Unbeknownst to her, though, the ants that live beneath Pacifica High School in the Realm of the Ant have plotted against her. "Foolish human!" squeaks the queen of the ants. "She thinks she can attend a dance and not dance!? We'll see about that. My minions!" The queen's armies snap to attention, "Yes, your highness!" "This night we shall teach that wallflower a lesson. If I'm not mistaken, I spotted a cookie crumb that somehow fell onto that young girl's dress. Your queen desires a late night snack. If you have any love left for your queen at all, you'll bring me that crumb, do you hear!" "Right away, your highness!" And with that, the ants go marching one by one. Hurrah! Hur—"AHHHHH!" screams the young girl, as she spies the benighted trail moving slowly yet persistently up her calf. To get them off, she jumps; she twists; she flails wildly, and...as if by accident, the young girl is dancing! Young and sweet; only seventeen...

So there's your context. Languages that work this way are rather neat, because you can handle something so common, yet so rarely encoded morphologically, simply by changing the case of the subject.

This is by no means the end, though. After all, if there are different names for each of these types of semantic arguments (SA, SP, P, A, E, ST...), couldn't there be a language that marks each one separately? Yes, there certainly can. I'll show you two different examples. In natural languages, this is rare, but attested. The most common of those types attested looks something like this:

Kelina sapu. "The woman is sleeping."
Kelinam talu. "The woman is dancing."
Kelinam lamu palinor. "The woman is petting the panda."
Kelina fisu palinor. "The woman sees the panda."
Kelinam kanu hopokos kitapor. "The woman's giving the man a book."

In the example above, SP's are marked with default case marking, SA's with -m, and objects (regardless of status) are marked with -r. This is a common enough pattern. But we can go further. Though I don't believe it's attested among natlangs, you can imagine a language like the following:

Kelina sapu. "The woman is sleeping."
Kelinak talu. "The woman is dancing."
Kelinam lamu palinor. "The woman is petting the panda."
Kelinap fisu palinol. "The woman sees the panda."
Kelinam kanu hopokos kitapor. "The woman's giving the man a book."

I had to make up some case markers on the fly in this one. Okay. Above, SA's are marked with default marking. SP's are marked with -k. A's are marked with -m (there are two. No language marks the agent of a transitive verb differently from the agent of a ditransitive verb. But one can imagine...). P's are marked with -r. Indirect objects are marked with -s. E's are marked with -p. And, last but not least, ST's are marked with -l. Now that's a very precise language. I'd like to point out that though this type of thing is attested, it's generally meted out differently than either of the two examples above (more on that when we get to animacy).

We're almost done with this section, but there's one bit left. We've talked about SA's and SP's, but consider the following English sentences:

"The woman's petting the panda."
"The book's petting the panda."
"The wind's petting the panda."
"The panda's being petted (by the woman)."

Those four sentences have four different types of subjects—two of which we haven't really talked about before. The first in (36a) is simply an agent. The last in (36d) is a subject that is, in fact, a patient (i.e. the subject of a passive). The second subject in (36b) is something we've talked about, but not directly. Remember the story about the woman with the clean panda? The woman is still the one initiating the petting action, but the book is the instrument used to perform the action. Thus, the subject is an instrument (SI). In (36c), unless the wind is some kind of sentient being, the wind is neither an instrument nor an agent, but simply a force of nature: a non-volitional subject (I'll call it SN). One could imagine a language where all four of these are marked differently, as in these sentences below:

Kelinam lamu palino. "The woman's petting the panda."
Kitapok lamu palino. "The book's petting the panda."
Makipos lamu palino. "The wind's petting the panda."
Palino lamuto (sa kelinak). "The panda's being petted (by the woman)."

I'm fairly certain that such a language as that in (37) doesn't exist, but it could. For that reason, I wanted to bring it up. And that, unless I think of something else later on, will finally conclude this section on semantically-based split ergativity.

•4.4 Animacy-Based Split-Ergativity:

It's been alluded to several times in the text above, so here it is: The section on animacy. Animacy really interested me for a long time because I didn't understand it. I don't claim to be a master on the subject now, but I do understand what people say about it. I've also intended Sheli to be a language that's sensitive to the animacy of its subjects and objects.

Anyway, so a quick question: What do people mean when they discuss animacy as it relates to language? Well, some languages encode animacy into their grammar. It can be done in many different ways, some of which aren't related to ergativity, per se. The essential point is this. Let's say you have a verb and two noun phrases. Let's say they're this: "eat", "sandwich", "man". In English, these can be arranged in two different ways, giving you "The man eats the sandwich", or "The sandwich eats the man". But leaving out cartoonish contexts, which one of these sentences is really the more likely to be uttered by a human being? Chances are it's the first one. This is because, speaking of reality as we know it, it's not only possible, but highly probable, that a human will eat a sandwich. It is impossible, though (or, at the very least, highly improbable), for a sandwich to eat a human. For that reason, is it even necessary to say which is the direct object and which is the subject in any way (either with cases or word order)? According to a lot of languages, no. (For a fascinating example, see Payne's discussion of the language Sierra Popoluca in his book Describing Morphosyntax.)

So, how does this relate to ergativity? Well, some languages use animacy to split up case assignment. Thus, some types of arguments will get one type of marking, and the rest will get the other type of marking. So here's a simple example:

Kelina lamu hopokor. "The woman's petting the man."
Hopoko lamu kelinar. "The man's petting the woman."
Kelina lamu palino. "The woman's petting the panda."
Palinom lamu kelinar. "The panda's petting the woman."
Palinom lamu kitapo. "The panda's petting the book."
Kitapom lamu palino. "The book's petting the panda."

In the example above, human beings are marked with a nominative-accusative system, and everything less animate than a human is marked with an ergative-absolutive system. The result is that in a sentence like (38c), the subject and object are marked with the same case. But this isn't a problem. Why? Because the more likely subject is the most animate one, which is the woman. Thus, it doesn't matter that there seems to be fixed word order in the sentences above. All six sentences below in (39) could only mean "The woman's petting the panda":

Kelina lamu palino. "The woman's petting the panda."
Palino lamu kelina. "The woman's petting the panda."
Kelina palino lamu. "The woman's petting the panda."
Palino kelina lamu. "The woman's petting the panda."
Lamu kelina palino. "The woman's petting the panda."
Lamu palino kelina. "The woman's petting the panda."

In fact, a language that uses this system has the advantage of achieving relatively free word order without having heavy-handed case marking as seen in a conlang like Zhyler (cases everywhere in that language! And it doesn't even have free word order!).

That's the basic idea behind an animacy system as it relates to case marking. So, a question: Is this the only way it can be split (i.e. one type of marking for humans, another type for the rest?). Absolutely not. So what are the ways to split it up? Well, there are two answers. The first is: Anyway you can imagine it. If you can dream it up, it's possible. Now, what's common among natural languages? For that there's a different (and rather definite, it seems) answer. According to Payne, there's a grand hierarchy of agent worthiness which I will try my darndest to reproduce here (I think I'm going to need to use a table...):

1 > 2 > 3 > 1 > 2 > 3 > Proper Names >	Humans > Non-Human Animates > Inanimates
Agreement > Pronouns	Definite > Indefinite

So...as I understand it...the table above is... Hmm... Okay, I get it. Odd he did it that way, though... Okay, the reason that 1, 2 and 3 are up there twice, is because the first set of 1, 2 and 3 refer to first, second and third person verbal agreement markers. The second set refers to pronouns. I guess it would've been too difficult to repeat everything after "proper names" twice, though, because those only appear once. Essentially, this is how to read that table. Let's take "proper names". Proper names will always be considered to be of higher animacy than humans, non-human animates and inanimates (regardless of definiteness [I guess in this table, proper names are always assumed to be definite—not necessarily an uncontroversial claim]). However, both pronominal verbal agreement, and personal pronouns will be considered more animate than proper names. For that reason, if you had a proper name and a pronoun as two arguments, the pronoun would be construed as being the subject, and the proper name the object (to indicate otherwise, an inverse marker, or something like it, would be required).

This relates to case marking because of a universal claim that Payne makes. So let's say that in a given language, everything to the left of proper names will be marked one way, and everything that's to the right of the last 3 will be marked a different way. According to Payne, it will always be the case that what's to the left of "proper names" will be marked with a nominative-accusative system, and what's to the right of the last 3 will be marked with an ergative-absolutive system. Why? I can't seem to find a good answer. I'm sure something metaphysical can be guessed at, though.

Now, as it relates to conlangs, does this have to be the case? Certainly not. It could be the opposite. But, given the fact that no natural language has been discovered (thus far) that does things the opposite way (i.e. contrary to Payne's claim), there seems to be some sort of connection between high animacy (or high topic-worthiness) and a nominative-accusative system. So, if a natural language is what you're after, you might want to stick with the universal. If you want to be daring, though, you have my blessing. Not that you need it.

Anyway, I could spend a long time showing you every possible example of where the hierarchy could be split, but instead I'll show you just one interesting example. This is an Ergato version of a language Payne describes called Cashinawa. Cashinawa has a system where first and second person pronouns are marked one way, third person pronouns another way, and full NP's are marked yet another way. Here's what that might look like in Ergato:

Ko sapu. "I'm sleeping."
Ko lamu per. "I'm petting you."

So those are the first and second person pronouns, and they're marked with a nominative-accusative system. Now here are the third person pronouns:

Li sapu. "She's sleeping."
Lim lamu lir. "She's petting her."

Above you have a three-way system, where each argument is marked differently. Again, this is only with third person pronouns. Now here's what the NP's look like:

Kelina sapu. "The woman's sleeping."
Kelinam lamu hopoko. "The woman's petting the man."

And, to round it off, the NP's are marked with an ergative-absolutive system. Now, here's something to notice: To what does the pronoun li refer in the sentences in (42)? I guess the default assumption would be a human, but there's no reason why it couldn't be a female panda, or some other female animal. Despite the semantics of its referent, though, the pronoun will always be higher up in the hierarchy. This is why Payne objected to the terms "agentivity hierarchy" and "animacy hierarchy". It doesn't really depend on the animacy of the referent—or, at least in this system. Rather, it depends on the morphological status of the argument. In that way, a less-animate third person pronoun will be higher up in the topic-worthiness hierarchy than an animate human NP. Now, it doesn't have to work this way for a conlang. You could easily imagine a system like this:

Li sapu. "She (human)'s sleeping."
Li sapu. "She (animal)'s sleeping."
Li lamu lir. "She (human)'s petting her (human)."
Li lamu li. "She (human)'s petting her (animal)."
Lim lamu lir. "She (animal)'s petting her (human)."
Lim lamu li. "She (animal)'s petting her (animal)."

A system like that above would surely help to disambiguate pronouns in certain situations. But, then again, you might have a whole different set of pronouns for different types of NP's. After all, in English we have "he", "she" and "it".

Another thing to remember is that these claims of universality are for the natural languages spoken on this planet we live on. One can easily imagine a language spoken by a race of intelligent (yet still quite cleanly) cats. In this language, perhaps there would be a new category: sentient non-humans. And perhaps NP's referring to sentient non-humans would be higher up in the hierarchy than humans. Additionally, there's always androids and robots, or talking trees. Or one can also imagine a highly-sexist matriarchal society where women are seen as more animate (and more worthy of being the topic of discussion) than men, dividing humans into male humans and female humans (and maybe the same is true of animals and pronouns). Thus, maybe a female flea would be considered more animate than a male human. The possibility for flux in the hierarchy is limited only by the reality you want your language to live in. So in that respect, think of the above as a guide, rather than a set of rules to follow.

•5.0 Mixing Systems:

To quote the great linguist Thomas Wier, "every language shows some features of ergativity and some features of accusativity" (click here for that discussion). Thus, a good system will include some elements from all the sections discussed above. I've already mentioned (dozens of times) how English makes a distinction between experiencer and non-experiencer verbs in the present tense. Another famous example is the -ee suffix, summarized below:

Escape (intransitive verb) + ee = escapee, "one who escapes" (nominalizes intransitive subject)
Nominate (transitive verb) + ee = nominee, "one who is nominated" (nominalizes transitive patient)
Nominate (transitive verb) + or = nominator, "one who nominates" (nominalizes transitive subject)

In the example above, you can see a clear ergative-absolutive pattern. This only applies to one tiny little corner of English grammar, but, then again, the same can be said of experiencer verbs in the present. This is part of what goes into creating a realistic language. Not everything is perfect, and not every pattern jumps out and draws attention to itself. Another simple pattern from a natural language can be seen with French. In French, there's a distinction in (what is now) the simple past tense between verbs that take an SA and verbs that take an SP. Take a look at this example:

J'ai dormi. "I slept." (SA)
Je suis arrivé. "I arrived." (SP)

In the example above, the subject is enacting the sleeping event (to an extent), whereas in the second sentence, the verb is something that happened to the subject. "Appear" is another verb like this. The difference between the two is a difference between two auxiliaries. If you translated it directly into English, the first would be "I have slept", and the second "I am arrived". In fact, we had something kind of like this in some version of English way back when. Ever heard the expression, "I am come", e.g. "I am come to partake of thy lodging and victuals!" I'm pretty sure that's English.

There are many, many ways you could create a mixed system. One way might be to have a nominative-accusative system to mark pronouns in the present tense, and an ergative-absolutive system to mark NP's in the present, while all arguments, pronoun and NP alike, are marked with an ergative-absolutive system in the past tense. And then maybe, in all tenses, the cases are flipped for verbs of experience (i.e. nominative marks pronoun stimuli, and accusative marks pronoun experiencers, in the present, and everywhere else, the ergative case marks stimuli, and the absolutive marks experiencers). The theoretical possibilities are endless (though certain possibilities become more difficult to justify linguistically than others).

•6.0 Something Else to Consider: Ditransitives:

One thing that often gets ignored in a discussion of ergativity is the marking of secondary objects in ditransitive clauses. As it turns out, it's by no means simple. Below I'll summarize a description of possible types of indirect object marking laid out explicitly in a paper by Matthew S. Dryer entitled "Clause Types" (warning: that link is to a .pdf).

So far in the nominative-accusative ditransitive examples I've shown, the direct object (P) has always been marked with the accusative case -r, and the indirect object (R) has always been marked with the dative case -s. Does this necessarily have to be the (excuse the pun) case, though? As it turns out, no. Actually, there are three different possibilities. First let's detail the common (to us) pattern. This is a pattern like Latin. This is an example where the direct object of a transitive verb is grouped together with the direct object of a ditransitive verb:

Kelina sapu. "The woman's sleeping."
Kelina lamu palinor. "The woman's petting the panda."
Kelina kanu kitapor palinos. "The woman's giving a book to the panda."

The ordering of the indirect object and direct object in (47c) can vary, but nevertheless, this is a very Latinate kind of pattern. Now let's take a look at a different kind:

Kelina sapu. "The woman's sleeping."
Kelina lamu palinor. "The woman's petting the panda."
Kelina kanu palinor kitapos. "The woman's giving a book to the panda."

In the example above, the cases on the objects of kanu, "to give", flip-flopped (as did the order, just to keep everything in line). A language that does ditransitives like this will usually mark that last argument with an instrumental, as opposed to a dative, case. Nevertheless, it is a different case, as opposed to an oblique, like in the English "I gave the book to her". In that English example, the "to her" part isn't as much a part of the argument structure as the R is in the counterpart sentence "I gave her the book".

For a final example, we can see a pattern that looks a lot like the last English example I gave:

Kelina sapu. "The woman's sleeping."
Kelina lamu palinor. "The woman's petting the panda."
Kelina kanu palinor kitapor. "The woman's giving a book to the panda."

As you can see, now there's only two cases operating in the (c) sentence. How do you know which is the direct object and which the indirect object? Strict word order. So, in the above example, there'd be some kind of rule that states that the first object in a ditransitive clause would be interpreted as the indirect object, and the second the direct object. This is exactly how it works in English, in a phrase like, "You gave me him" (an odd sentence, I know. And why? Because of animacy!), "me" is always interpreted as the indirect object, and never as the direct object. (Note: There are dialects where the opposite is still productive, thus the indirect object in, "Give it me, I say!" is "me", not "it".)

So those are three possibilities for nominative-accusative systems. What about ergative-absolutive systems? Well, there's three possibilities for them, as well, and they match up nicely with the three systems above.

The first ergative-absolutive system is one where the absolutive argument of a transitive clause is marked the same as the direct object of a ditransitive clause. This is what it looks like:

Kelina sapu. "The woman's sleeping."
Kelinar lamu palino. "The woman's petting the panda."
Kelinar kanu kitapo palinos. "The woman's giving a book to the panda."

This should look just like the system in (47), only with -r's flipped around. This would be like ergative Latin, which I call Nital. Pretty straightforward. Next system:

Kelina sapu. "The woman's sleeping."
Kelinar lamu palino. "The woman's petting the panda."
Kelinar kanu palino kitapos. "The woman's giving a book to the panda."

Again, this is like the examples in (48). Perhaps a helpful way to think of the ditransitive verbs in sentences like these is that kanu isn't defined as "to give (something)", but rather "to give to (someone)". The extra case, then, specifies what's being given (again, usually something like an instrumental). Now for the last example:

Kelina sapu. "The woman's sleeping."
Kelinar lamu palino. "The woman's petting the panda."
Kelinar kanu palino kitapo. "The woman's giving a book to the panda."

And, again, the way you tell which object is which in (52c) is strict word order.

That wraps up this discussion of ditransitives. There's more to them, to be sure, but this is all that presently concerns us. Again, it's just something to think of. The status of indirect objects is something I certainly didn't think about in many of my languages, and I believe they're the less realistic for it.

•7.0 Impossibilities:

There are certain patterns deemed to be impossible, which makes them immediately interesting. I'll just mention them here.

One that I may have mentioned already has to do with split-tense systems. In all the split-tense systems that have been found, the present tense has a nominative-accusative pattern, and the past tense has an ergative-absolutive pattern. Based on this evidence, experts have deemed the opposite impossible. While it may be easier to come up with a historical explanation for the opposite, it's by no means unworkable.

Related to tense, if you read up on this stuff, you'll notice that the only tenses that are mentioned are present and past, or, at the most, past and non-past. The future tense is never discussed. And I'm sure any conlanger can think up more tenses than even past, present and future. As far as I know, there are no universals for what kind of marking you get in the future (well, except maybe that it probably looks like the present). That's something to think about.

Let's say that we are working with just past, present and future (no aspect). That's three tenses. The reason why nominative-accusative and ergative-absolutive works so well with present and past tense is because they line up: Two systems, two tenses. But what do these terms stand for? In a sentence with three basic arguments, S, A and P, nominative-accusative stands for the system that groups S and A together to the exclusion of P. Ergative-absolutive, on the other hand, stands for a system that groups S and P together to the exclusion of A. Do you see what I see? There's a third pattern not mentioned here, and, coincidentally, a third tense that doesn't get to play. So imagine, if you will, the following: Nominative-accusative in the present; ergative absolutive in the past; and in the future (using -sa as an impromptu future marker)...!

Kelinar sapusa. "The woman's gonna sleep."
Kelina lamusa palino. "The woman's gonna pet the panda."

Oh, yeah! This is a system that, paradoxically, groups A and P together to the exclusion of S! This kind of system is unattested in natural languages, and judged impossible. Thus (to my knowledge), it hasn't been officially named. Therefore, I'm going to name it. What ties together the subject of a transitive verb and the patient of a transitive verb...? Well, how about this: In a transitive clause, there are two arguments; in an intransitive, there's one. Thus, the case assigned to both the subject and object of a transitive verb is the duative, and the case assigned to the single argument of an intransitive verb is the unitive. Yeah! That sounds good. Thus, I dub the above pattern a duative-unitive system. I named them this way because the pattern seems to be that the case that's assigned to the subject of a transitive verb is the one that goes first. Hee, hee... Now I wish I had a language that used this pattern. I'll have to work on that...

(Quick Note: On the CONLANG list, this pattern was dubbed the "Monster Raving Loony", or MRL, pattern. The case names were called the "intransitive" and "transitive" cases. I don't like this naming strategy, because both "intransitive" and "transitive" already mean something, and confusion could easily ensue. Go here to see the various related posts.)

(Update: It appears that such a system does exist in the wild. In the language Rushani, the agent and object are marked with an oblique case, while subjects are marked with the absolutive. This occurs only in the past tense. It apparently arose from the reanalysis of a split-ergative system, and the system is now in decline with younger speakers. The literature apparently calls this system either a transitive system or a double-oblique system. Personally, I still prefer duative-unitive. Oh well. For more information on the Rushani system, go here.)

Some other impossibilities have been touched on in the animacy section. Here's an idea. Referring to the hierarchy mentioned in the animacy section above, why not have two splits. And not like the kind I described for the Cashinawa system. This is a system where the section in the middle is marked one way, and the sections on either end are marked another way. So let's say that all pronouns are marked with a nominative-accusative system, as are everything to the right of humans, and then humans and proper names are marked with an ergative-absolutive system. That would be strange, and definitely would violate the universal Payne proposed.

Another impossibility one can imagine is with ditransitives. In all six examples above, the indirect object and direct object could be marked in various ways, but they were always marked differently from the subject. Why not mark the indirect object the same way as the subject? In fact, let's do these three possibilities with a duative-unitive system, just for kicks:

Kelinar sapu. "The woman's sleeping."
Kelina lamu palino. "The woman's petting the panda."
Kelina kanu kitapo palinos. "The woman's giving a book to the panda."

In this pattern, the direct object of both transitive and ditransitive verbs are treated alike. And, as you can see, they're both marked with the duative case. The subjects of the transitive verbs are as well. The subject of the intransitive is marked with the unitive, and the indirect object in (54c) is marked with the dative. Now for the next one:

Kelinar sapu. "The woman's sleeping."
Kelina lamu palino. "The woman's petting the panda."
Kelina kanu palino kitapos. "The woman's giving a book to the panda."

Same thing here as with the "give to (someone)" verbs we've seen before, where the R is assigned the objective case, which is in this case the duative. And here, the -s probably stands for an instrumental case. Last one:

Kelinar sapu. "The woman's sleeping."
Kelina lamu palino. "The woman's petting the panda."
Kelina kanu palino kitapo. "The woman's giving a book to the panda."

And this is about as duative as you get. Here the subject of the intransitive verb in (56a) is marked with the unitive, and everything else is marked with the duative, the status of each object being determined by word order in (56c).

Oh, one thing I forgot about: What about a valency reduction system in a duative-unitive system? This would be odd, because in this case (and in this case only), the case that would be reduced would be the unmarked/default case, rather than the marked/special case. (Well, that is if the duative is the unmarked case.) Anyway, the result is that the transitive verb becomes intransitive, and the duative argument becomes a unitive argument. But which duative argument?! You don't know. Therefore, the resulting verb would mean something like, "Y is a participant (either agent or patient) in an X action". Thomas Wier suggested this might be like the Ancient Greek middle voice construction (see his post to CONLANG by clicking here). In any case, here's what it'd look like in Ergato:

Kelina lamu palino. "The woman is petting the panda."
Kelinar lamuto (palinok). "The woman's petting (the panda)/being petted (by the panda)."
Palinor lamuto (kelinak). "The panda's petting (the woman)/being petted (by the woman)."
Kelina hopokos kanu kitapo. "The woman's giving the book to the man."
Kelinar hopokos kanuto (kitapok). "The woman is giving to the man (and what she's giving is a book)/being given to the man (by the book)."
Kitapor hopokos kanuto (kelinak). "The book is giving to the man (and what it's giving is a woman)/being given to the man (by the woman)."

Given a system like the above, one can easily imagine that discourse context and animacy would help you decide which reading is the correct one (for example, if giving is the act, and you're talking about a woman and a book, it's pretty likely that the book's the one being given). Anyway, that's what a duative-unitive system would look like, in toto (I believe). As for the valency-reduction system, if you already have passive and antipassive, then I propose that the name of this system should be an ambipassive, since it can apply to either of the arguments in a transitive clause.

Here's a thought I don't think I've run across before: What if the subjects of intransitive verbs, tranisitive verbs, and ditransitive verbs all had different subject marking? This would be treating the subjects of ditransitive verbs as something inherently different from transitive verbs. This is probably unattested, but nevertheless, a possible pattern.

Those are some ideas to mull over. There's a lot more that's possible than is attested in the world's languages (though they do do a lot more than most universalists would have you believe).

•8.0 Conclusion:

The intention of this section has been to document the basics of ergativity. It's my hope that this is a starting point. If you have more information, or if you think I've made a mistake (or if you spot any typos—I know there are tons!), my hope is that you'll e-mail me, so that I can further improve this section. Though I did write all this, I prefer to think of this as a collaborative effort, since I got my information from many different sources. I hope you've got something from this section on ergativity, and that if you have something to share, you'll let me know, so I can make improvements in the future.

•9.0 References and Thanks:

These are a list of references I used and some shout outs:

Bell, David. ámman îar Reference Grammar.
http://www.graywizard.net/Conlinguistics/amman_iar/amman_iar.htm, 2004.

Dryer, Matthew S. "Clause Types". SUNY Buffalo, 2001. Download .pdf

Henning, Jeffrey. Langmaker.com. http://www.langmaker.com, 2004.

Hillery, P.J. The Georgian Language: An outline grammatical summary.
https://www.richw.org/armazi/georgian/, 2004.

Loos, Eugene E. (gen.ed.). Glossary of linguistics terms.
https://glossary.sil.org/term, 2004.

Payne, Thomas E. Describing Morphosyntax: A Guide for Field Linguists. Cambridge University Press, 1997. See Amazon.com Info

Rye, Justin B. Learn Not to Speak Esperanto: Case.
http://www.xibalba.demon.co.uk/jbr/ranto/r.html, 2004.

Wier, Thomas. Re: ergative + another introduction.
http://listserv.brown.edu/archives/cgi-bin/wa?A2=ind0411c&L=conlang&F=&S=&P=31479, 2004.

Wier, Thomas. Re: Ergativity Reference Done.
http://listserv.brown.edu/archives/cgi-bin/wa?A2=ind0411D&L=conlang&P=R12196, 2004.

I'd like to thank all those who contributed to the recent discussion of ergativity on the CONLANG list (well, recent as of November 28, 2004), as well as all those who've discussed ergativity many, many times on CONLANG over the years. In particular, I'd like to thank Thomas Wier for reminding me of the escapee example, which, despite its fame, always seems to elude me in times of need. I'd also like to thank Roger Mills for reminding me of David Bell's section on ergativity in ámman îar. I'd also like to thank Taliesin for his design advice (as you can probably tell, I'm not too good a judge of what is and is not easy to read on the screen), and Jeff Rollin for pointing out something I missed (there will be more). And, of course, I'd like to thank Christophe Grandsire for providing me with webspace. Vive la France!

Update (09/17/2019): Thanks to an email from Kirill Elagin, I've updated several of the dead links on this page. (I'll also note that I now have my own webhosting.)

Update (07/30/2020): A big thank you to Bruno Behling! He spotted two errors in the text—one quite serious, and I'm surprised neither I nor anyone else caught it up to this point—and also pointed me towards a natlang example of a duative-unitive system. Delighted to be able to update the text!