Linguists classify natlangs according to how they indicate which noun arguments to a verb play what parts in the verb. Your conlang also has to indicate, somehow, which arguments play what parts in the verb — unless you're creating an exotically structured conlang where somehow you don't have a verb with arguments. We'll discuss language structures that play games with lexical/semantic classes like noun and verb in a later section; but meanwhile, if you're going to try to shake things up that much, you may find it helpful to understand, going in to that shake-up, how verb-and-arguments languages can work.
Knowing how a language aligns the arguments to a verb may say surprisingly little about why it works that way. We'll say a bit more about that in this section, but mostly we'll leave it for later sections, because it turns out that the reasons for alignment can be deeply entangled with a wide variety of other deep topics in language structure. It all tends to connect with how sentences fit into larger discourses. So in later sections we'll never be far away from alignment.
Keep in mind that when an area of linguistics is controversial or uncertain, that's an opportunity for conlangers to experiment and see what can work and how. People have looked at the overall arrangement of linguistic terminology in the area of "morphosyntactic alignment" (if you really think about the word "morphosyntactic", it seems to mean the same thing as "grammatical" only more academic-sounding), and concluded that linguists have little or no clue what's going on in this area. This is roughly what linguist Scott Delancey told a 2005 workshop on ergativity:
when we limit a collection to certain kinds of specimens, there is a question whether a workshop on "ergativity" is analogous to an effort to collect, say, birds with talons — an important taxonomic criterion —, birds that swim — which is taxonomically only marginally relevant, but a very significant functional pattern —, or, say, birds that are blue, which will turn out to be pretty much a useless criterion for any biological purpose.
— Scott Delancey, "The Blue Bird of Ergativity", 2005.
For classifying many languages, it's enough to identify three possible roles that a noun argument can play, relative to the verb:
- the sole noun argument to an intransitive verb, called the subject (S).
- The cat sleeps. The jogger runs.
- the noun that acts in a transitive verb, called the agent (A).
- The mouse eats the cheese.
- the noun that is acted on in a transitive verb, called the patient (P).
- The mouse eats the cheese.
Notice that subject is different, in that it really doesn't say anything about how the noun participates in the verb; in
- The mouse eats.
- The cheese is eaten.
mouse is the subject of eats, and cheese is the subject of is eaten (yes, that's passive voice; grammatical voice is another complication, which we'll discuss in a later section). You might suppose that, here, mouse is an agent and cheese is a patient, even though each is the sole argument to an intransitive verb. We're talking, though, about the overall grammatical organization of sentences, and a lot of these fine semantic distinctions of noun usage just don't carry their impact that far: We can define some of the most common language types without going past S A and P, and when those three structural roles aren't enough, other language types tend to be described not in terms of a larger set of structural roles, but in terms of differences from the "simpler" types that still only needed S A and P.
These three are sometimes called "core cases", to distinguish them from the wide variety of other cases — dative, instrumental, locative, etc.
Ways of marking arguments
The different alignment types are distinguished by the logical arrangements they use to keep track of the different noun arguments, but before we get into that it's worth a moment to look at the mechanics of the indicators they use. (We discussed a lot of this at the Intermediate level, of course, and in more detail; but now we want an overview from the perspective of alignment systems.)
The "generic" marking strategy is to inflect each noun to indicate its case, like Latin or Greek does. The word "case" may be used to refer specifically to a noun role in the sentence that is marked by inflecting the noun; someone using the word "case" that way would say that English doesn't have noun cases, by which is meant that it doesn't inflect its nouns for their role in the sentence. Other times you'll find the word "case" used more broadly for noun role indicated by whatever means, and then the same property of English would be described by saying noun case in English isn't marked by inflection. We'll use the word "case" here in the broader (some would say, looser) sense.
Another way to indicate case, then, is by word order; that's the way English does it (although some English prounouns are also inflected for case). Since this is syntax, where noun inflection would be morphology, the difference between an alignment system using inflection and one using position may be made by calling the latter "syntactic"; thus, the alignment used by Latin and Greek is "accusativity", while that of English is "syntactic accusativity".
It's also possible to put information about the nouns on the verb. This may be redundant information that's also shown on the noun, as with English verbs that "agree" with the number of the nominative. (Curiously, the specific form of the nominative/verb inflection for number is usually a suffix -s that moves between the two, attaching to the verb when the noun is singular or to the noun when it's plural — a feature that, one suspects, might be much admired if it had been invented by a conlanger instead of occurring in a natlang; but, back to alignment.) However, the verb can be inflected for properties of one or another of its arguments that aren't shown on the nouns themselves. How you then determine which noun to apply which verb-marked information to, may then be determined by some other means — such as position, or some other marking that does agree between the verb and noun — or may simply be ambiguous; it's not uncommon for a language to simply leave some things unspecified, and speakers either figure it out from context, or occasionally misunderstand it, without thinking the ambiguity odd or unnatural.
There are also languages that inflect the verb with so much information about the nouns, that the nouns don't even have to appear as separate words; at the extreme, this turns into noun incorporation, and then into polysynthesis where — in theory — there are no separate noun words (we'll discuss polysynthetic languages in a later section). Things like polysynthesis are not, of course, switches that get turned on or off; there's plenty of room for disagreements about whether a given language is "truly" polysynthetic, or the like. The way the language actually works is what matters, and in nature is almost never pure, so if you're building an artlang (so you're not actually trying to produce an unnaturally uniform language), you may want things to work in an artfully impure way.
Simple role-marking alignments
English has two noun cases for these three structural roles: nominative for intransitive subject and transitive agent, and accusative for transitive patient. Since English marks noun case mainly by word order — setting aside possessives, and excepting some of its pronouns — it's hardly surprising that the three structural roles S A P between them are served by only two cases; you can only mark two cases when the marking is simply before-the-verb versus after-the-verb. More interestingly, though, many languages that inflect nouns for case, such as Latin, also use a single nominative case for both subject and agent, and a separate accusative case for patient. Languages that use two case markings this way for the three structural roles are called nominative-accusative languages; accusative languages, or nominative languages, for short. Most Indo-European languages, as well as many others, are accusative.
Some languages split up these three structural roles into two marked cases in the other "obvious" way: by marking intransitive subject the same way as transitive patient. Generally, the subject-patient case is called absolutive, and the agent-only case is ergative. These languages are called ergative-absolutive; ergative languages, or (occasionally) absolutive languages, for short. Ergative languages aren't so common; a notable example is Basque.
The distinction between accusative languages and ergative languages might sound superficial, saying — as the above quote from Scott Delancey suggests — nothing deep about the character of the language. However, even if it is superficial, it can be fairly far-reaching in the ways it interacts with other features of the language. For example, the passive voice we used above, "The cheese is eaten", is by natural affinity a nominative-accusative construct: It converts a transitive verb to intransitive by dropping the nominative argument — the agent — and changing the accusative argument to nominative. Ergative languages can have instead an antipassive voice, where transitive verb converts to intransitive by dropping the absolutive argument — the patient — and changing the ergative argument to absolutive. The reasons you would want to use these voices are different, and really only come into play when you consider large texts, of many sentences. We'll explore voicing in a later section; but it should be clear that how you arrange and use voicing will be all tangled up with noun alignment. Alignment also gets tangled up in the formation of relative clauses, yet another topic for a later section. So this alignment stuff may be superficial in itself, but it matters for how the language scales up to larger texts.
Ergativity was a big trend in conlanging a few years ago, possibly sparked by David Bell's artlang ámman îar. Conlangers have also playfully noted, though, that having divided S A P into two cases by combining S and A, and by combining S and P, one might also try to combine A and P. With a transitive verb, you then wouldn't be able to tell which of the two nouns was the agent and which was the patient. This seemed such an absurd way to do things, they nicknamed it monster raving loony alignment. Except that they then discovered there's at least one example where a natural language actually does that, in a classic instance of anadew. Speakers would presumably cope with the agent/patient confusion by "common sense", considering the context; if it really became a problem for them, one might expect the language to change somehow to resolve the ambiguity — which would be, again, an alignment strategy.
Another straightforward way to handle the three structural roles S A P is to mark the three of them all differently. Languages that do this are called tripartite. Typically, the case of the subject is called absolutive, agent ergative, and patient accusative; so tripartite languages are sometimes called ergative-accusative. Tripartite alignment is quite rare in human natlangs; examples include Nez Perce of North America, Wangkumara of Australia, and Semelai of Malaysia.
Some languages distiguish agent and patient by different cases — like both accusative and ergative languages do — but then may use either case for the sole argument of an intransitive verb. These are called active-stative languages; also called split intransitive languages; or active languages, for short.
Usually (to the extent these sorts of languages are usual at all), the case used with an intransitive verb depends on the verb — each intransitive verb requires a particular argument case. These may be called split-S languages. More rarely, a language may permit choice of intransitive argument case based on semantic considerations, so that the same verb can take either case; these may be called fluid-S languages.
Scattered examples of active-stative languages occur in the Americas (South, Central, and North), Austronesia, Tibet, the Caucasus, and Siberia.
Different parts of a single language may use different alignment systems; such a language is said to be split. Split-S languages are — arguably — an example of this, where the split is between a set of verbs that follow a nominative-accusative pattern and another set that follow an ergative-absolutive pattern. (You'll notice, though, that fluid-S is harder to fit into this notion of "split", while it also seems reasonable to group split-S and fluid-S together into a separate class of active-stative languages.) Pronouns — or at least some pronouns — are often handled differently; even in English, some pronouns (I/me, he/him, etc.) are exceptions to the general lack of core case inflections.
Languages are commonly split along verb tense or aspect. When that happens, the ergative verb form usually isn't present, future, or irrealis. This may have to do with language evolution, and how ergative patterns arise versus how verb forms arise. It's been suggested, for example, that the ergative past tense in Hindi arose because the past tense evolved from a passive voice, which moves the patient into subject role. (Future may sometimes evolve from some form of irrealis, too.) Split alignment can also occur along lines of animacy, such as the first and second person pronouns — presumably the highest point in an animacy hierarchy — following a nominative-accusative pattern while the rest of the pronouns follow ergative-absolutive.
The term split ergative applies to a language with parts ergative and parts not — though not, typically, a split-S language. You may occasionally encounter the term split nominative, but not often, perhaps because nominative is much the most common alignment system while ergative is unusual.
Alignment can also be done using some internal property of the nouns. This occurs in some North American languages, which may have a robust animacy hierarchy for nouns; in Navajo, the two arguments of a transitive verb are positioned with the most animate first, where the most animate usually is the agent and an inflection on the verb indicates when the less animate noun is the agent instead. (Of course, you could interpret the same Navajo alignment scheme as using word order, with an inversion inflection for the verb, and a convention that the more animate noun has to come first. Whether that counts as "caseless" is an argument about terminology, as the language works the same way however it's described.)
A linguistically challenging alignment strategy — native speakers don't have a problem using it, but linguists struggle with how it works — is shared by the Austronesian language family distributed across the Pacific islands, maritime Southeast Asia, and Madagascar. One case is reserved for the noun most emphasized, called by various linguists the focus, topic, or trigger; with various other cases applying to the remaining nouns, and the positions and cases of the noun arguments can be rearranged by means of an array of different voices that can promote most any noun to be the focus.
Much controversy surrounds the nature of the emphasis placed on that one argument; the terms subject, focus, and topic are used by linguists in various ways, on which they may or may not broadly agree, to describe different ways that a noun can relate either to the syntax of the particular clause in which it occurs, or to the larger discourse within which the clause occurs. The term trigger has been proposed by some as a neutral term for the particular sort of emphasis involved, a proposal to which some others have objected.
In any language that works this way, evidently the sole argument of an intransitive verb is the trigger; so if the agent of a transitive verb is usually the trigger, one might be tempted to think of the trigger case as nominative and classify the language as accusative, or if the patient is usually the trigger, think of the trigger as absolutive and classify the language as ergative. Sure enough, Tagalog — the Austronesian dominant native language of the Philippines, which most often uses patient trigger — has sometimes been classified as ergative.
In a typical accusative or ergative language, passivization (for accusative) or anti-passivization (for ergative) is a valency-reducing operation, and the dropped argument can be reintroduced as an oblique; but Austronesian languages have an array of voices that don't reduce the valency, while allowing agent, patient, or more remote noun roles such as dative, locative, or instrumental, to be promoted to the same case as the sole argument of an intransitive verb.
Discussion of Austronesian alignment on the Conlang mailing list, starting circa 1995, led to a notion of "trigger language" that was meant to be a simplification of how Austronesian languages work. Case markings indicate the roles of all the noun arguments to the verb, except that one of them, designated the "trigger", has no case marking at all, and its case marking appears instead on the verb. A conlang that works this way is Ayeri, started in 2003 (whose author has since been led into an increasingly in-depth exploration of the labyrinthine study of Austronesian alignment simply to exlain how Ayeri differs from it). Another example is the (purely pictographic) conlang X. However, it was pointed out as early as 2006 (in fact, it was pointed out in discussion of this wikibook) by the author of X that this interesting alignment strategy is not Austronesian, as it makes the noun property of trigger-hood cleanly separate from the roles of the noun arguments, whereas Austronesian languages use voicing to rearrange the roles.
Most verbs are either intransitive, with one argument, or transitive with two arguments; and we've been describing alignment systems in terms of how they treat those two most common kinds of verbs. In particular, which of the two arguments to a transitive verb is treated like the one argument to an intransitive verb.
But there may also be a class of ditransitive verbs, with three arguments; the three argument roles are then agent, theme, and recipient. In English, we'd have
- The librarian gave the student the book.
where the librarian is the agent, the book is the theme, and the student is the recipient. But the question is, of the theme and recipient, which of them is treated like the patient of a (mono)transitive verb? This is the same sort of question as asking which of the arguments to a transitive verb is treated like the one argument to an intransitive verb.
A language that treats the theme like the patient is indirective (though you won't encounter that term very much, as apparently most languages work like that). A language that treats the recipient like the patient is called by any of several names: secundative, primary object, dechticaetiative, and — believe it or not — anti-ergative.
From the above example, apparently English is of the less common, secundative type: the recipient is treated like the patient of a transitive verb — positioned immediately after the verb — while the theme is treated in a new way, different from either transitive argument — positioned second after the verb. This is confirmed by what happens when this sentence is converted to passive voice, which drops the agent and promotes the patient to nominative treatment — which in English means the patient gets positioned before the verb.
- The student was given the book.
This makes things sound very tidy, but language is generally organic rather than tidy. Consider the English sentence
- The librarian gave the book to the student.
and its passive form
- The book was given to the student.
What has happened here? The reduction of "The librarian gave the student the book" to "The librarian gave the book" is analogous to passive voice, except that instead of dropping the first argument (the librarian) and promoting the second argument (the student) to its place, this drops the second argument (the student) and promotes the third argument (the book) to its place. The passive form of this would then be "The book was given." Adding a prepositional phrase for the recipient to both of these produces "The librarian gave the book to the student" and "The book was given to the student."
Which may again seem tidy... but not everyone will agree that English is secundative, claiming instead that it's primarily indirective with some secundative mixed in (keep in mind, this is a question of how to describe the language, not how it works); and, more substantively, some dialects of English would say things like "The book was given the student by the librarian" (which most native English speakers would understand even though they mightn't say it themselves).