zompist bboard

(If this is too tl;dr for anyone, there's a brief summary for you at the bottom of this post)

I think the time has come for a new introductory polysynthesis thread for novice conlangers/amateur linguists. The original thread, now in the L&L Museum, served its purpose well at the time, but things have changed significantly since most of that discussion took place. At the time, few of us even had any conception of "polysynthesis", and Jeff Burke was essentially the only ZBBer with any deep knowledge of any polysynthetic languages. And he could only speak to those languages he knew, although some people generalized his statements much more widely than they were intended. Furthermore, the purpose of that thread was not initially to provide a thorough introduction to polysynthesis, and so in some areas it's not particularly helpful for that. Finally, it's both very long (the information is scattered over 20+ pages) with a lot of extraneous stuff, and contains some statements that are either misleading or just plain wrong. For all these reasons, a new introduction, aimed at the novice conlangers of 2012, is in order.

I've tried to write this introduction to help new conlangers figure out the basics so they can design a polysynthetic language on their own, but I've also tried to make it interesting and informative for people who are simply interested in linguistics, and at times it may be pretty dense. On the other hand, partly for reasons of length and simplicity, I've avoided discussing a lot of the more recent, more theoretical work on noun incorporation, non-configurationality, etc. Readers should be aware that the descriptions here are designed primarily to give people a broad overview and help them move forward with conlanging or amateur linguistics; they are not meant to withstand a rigorous investigation of all the details, or to represent the consensus view of all linguists. Those who are interested should continue looking into such matters on their own.

Of course, I'd welcome any additional comments, examples, corrections, questions, etc. from anyone!


Table of Contents
1 - What Is Polysynthesis?
2 - A Quick Summary
3 - Polypersonal Agreement
4 - Noun Incorporation
5 - Other Affixes
6 - Free Word Order
7 - Head-Marking
8 - How Does Polysynthesis Arise?
9 - Where To Go From Here
10 - Sources


1 - What Is Polysynthesis?

Lets start with the basics. What does it mean to say that a language is "polysynthetic"? It's actually a rather vague term, and has been used in different ways by different people. Since synthesis refers to the general ratio of morphemes per word in a given language, ultimately, all "polysynthetic" means is that a language has a large number of morphemes per word. In practice, this generally means the language: (a) has verbs that show polypersonal agreement; (b) permits noun incorporation within verbs; (c) makes extensive use of affixes to mark many other categories (e.g., adverbial notions) expressed with separate words in less synthetic languages like English; (d) has relatively free word order; and (e) is predominantly head-marking. Not all polysynthetic languages have all of these characteristics (some Eskimo-Aleut languages, for example, lack noun incorporation under most definitions, but they're highly polysynthetic), but most polysynthetic languages probably have most of them.

The upshot of all this is that you can get entire sentences expressed in a single long verb in polysynthetic languages. In fact, this is a common definition for polysynthesis. For example, one of Evans and Sasse's definitions is, "Polysynthetic languages represent, in a single verbal word, what in English takes an entire multi-word clause." I'll give a handful of examples, to help you get the feel for how such languages work, before moving on to a more detailed discussion of each of the above traits in turn. Examples:

• Caddo: di’ch’áhyahdán’áh
  "they pulled out something round"
  (DEFOCUSING.AGT-PROTHETIC-round.object-away-pull-PERF).
  
  
• Bininj Gun-wok: abanyawoyhwarrgahmarneganjginjeng
  "I cooked the wrong meat for them again"
  (1sg.SUBJ/3pl.OBJ-again-wrong-BENEFACTIVE-meat-cook-PAST.PERF)
  
  
• Greenlandic: aliikusersuillammassuaanerartassagaluarpaal-li...
  "However, they will say that he is a great entertainer, but..."
  (entertainment-provide-SEMITRANS-one.good.at-COPULAR-say.that-REP-FUT-sure.but-3pl.SUBJ/3sg.OBJ=but)
  
  
• Koasati: ostohimilááchihalpíísalahoolimáámimpayon...
  "They say that they all might be able to go and bring it to him, but on the contrary..."
  (go.and-INSTR-DISTRIB-3DAT-arrive-PL-ABILITY-IRREALIS-DEDUC-HEARSAY-CONSEQ-DIFFERENT.SUBJ:FOCUS)

Polysynthetic languages are not evenly-distributed geographically. They're extremely common in North America, and fairly common in Central and South America, Siberia, the Caucasus, northern Australia, and New Guinea. Elsewhere, they are considerably rarer.


2 - A Quick Summary

If you're a real newb to linguistics stuff, or if you just don't want to wade through all the description below, in this section I'll summarize the very basic basics of polysynthetic languages so that you can get started on trying to create your own. But this necessarily means a lot will be oversimplified, so be forewarned (If you don't know what this simplified summary means, you probably need to learn more basics first before tackling polysynthesis):

For the most basic beginner's purposes, a polysynthetic language is a language where verb prefixes and suffixes (together called "affixes") are used to mark a lot of the information that in more familiar languages like English is conveyed with separate words. This usually includes affixes which identify the subject (and in transitive verbs, the object); any tense, aspect, or mood distinctions the language makes; noun roots "incorporated" into or compounded with the verb to qualify it or identify participants like its former object or (intransitive) subject; and a host of other notions like directions, adverbial notions, and so on. Polysynthetic languages revolve around their verbs--most of the work of identifying who is doing what to whom, in what manner, using what instruments, etc. etc., is all accomplished on the verb, and often highly inflected verbs, with no other nouns or other words, can stand as a full, complete sentence.

3 - Polypersonal Agreement

"Polypersonal agreement" means that a verb marks multiple core arguments. To oversimplify for a moment, this basically means that both the subject and object of a clause are marked on the verb. Different languages will distinguish different characteristics that are marked--person, number, gender, and so on--but usually the 'case' (or the grammatical role) of each participant is made clear in some way, via verbal prefixes and/or suffixes.

Everyone here will already be familiar with a language that mark verbs for at least one core participant: even English does so marginally (e.g., the suffix -s in drink-s marks that the subject of the verb is third person singular). Romance languages, which will be familiar to most people here, provides a somewhat better example, with verbs marking the person and number of their subject by means of suffixes (which also mark tense, aspect, and mood (TAM)), as in Spanish habl-o, habl-as, habl-a, habl-amos, habl-ais, habl-an ("speak" with the present indicative suffixes marking the subject as, respectively, 1sg, 2sg, 3sg, 1pl, 2pl, 3pl).

Polysynthetic languages, then, simply extend this type of marking further, and verbs mark both the subject and object (again, this is oversimplifying, we'll get to complications later). For example, take the Baniwa of Içana verb rikapani "he sees him/it". This breaks down as ri-kapa-ni, "he(subj)-see-him(obj)" (or, more accurately: 3sg.NONFEM.SUBJ-see-3sg.NONFEM.OBJ). As you can see, the verb marks the gender (non-feminine), number (singular), person (third), and grammatical role (subject or object) of each argument of the verb.

Not all polysynthetic languages have separate affixes for subjects and objects, however. In many cases the morphemes are partly or entirely fused together, such that a single affix marks the person, number, etc. of both subject and object. A good example is the Iroquoian languages. For instance, the Seneca verbal prefix she- marks both that the subject is second person and that the object is plural, as in ëshéo:wi’ (FUT-2SUBJ/pl.OBJ-tell-PUNCTUAL), "you will tell them." For another example, consider the Cariban language Dekwana, where w- indicates a first person subject and third person object, while y- indicates a third person subject and first person object. Thus, contrast wedanta, "I meet him/her", with yedanta, "s/he meets me."

Finally, note that in many cases polysynthetic languages can mark more than just two participants on the verb. Some other roles that can be indicated include benefactives, indirect objects, and so on. Take, for example, Abkhaz: dbə̀ltart˚’ (d-bə̀-l-ta-rt˚’ = 3sg.MASC.OBJ-2sg.INDIRECT.OBJ-3sg.FEM.SUBJ-give-PURP), "she gives him to you."


A Big Caveat

As I've emphasized a few times, this is a very simplistic overview, and the description above glosses over some important stuff. Here's the main qualification: I've been talking about "subjects" and "objects," but not all languages divide up verb arguments that way. Very basically, languages like English treat the subject of intransitive sentences the same as the subject of transitive sentences, and different from the object of transitive sentences. Languages like English are called "accusative." But many languages treat the subject of intransitive sentences the same as the object of transitive sentences, and differently from the subject of transitive sentences; languages of this type are called "ergative." Some languages are ergative in some places (say, only with third person arguments, or only the perfective aspect) and accusative in others (this is called "split ergativity"). Some languages mark the subject of an intransitive sentence differently depending on whether it is viewed as controlling or instigating the action (in verbs like "jump") or not (in verbs like "sneeze" or "fall down" or "be tall"); languages of this type are called "active", among other names. And so on and so on. So although I've only mentioned "subjects" and "objects" in the above discussion, treating all the languages mentioned as though they were accusative like English, the truth is far more complicated, and you'll need to think about what alignment system you want your conlang to have (there's a quick overview of all this in the Morphosyntactic Alignment thread, though unfortunately Sander's diagrams seem to have disappeared with the passing of years...).

4 - Noun Incorporation

Noun incorporation (NI) is ultimately a form of compounding: it combines a noun root/stem with a verb (a N+V or V+N compound) to create a new verb with a more specific meaning. English has some (marginal) examples, such as "berry-pick" or "mountain-climb": here, a noun is combined with a verb to form a new verb with a narrower scope than the original verb (referring now to specific kinds of picking or climbing). However, the English examples aren't very good analogues to the much more productive systems of languages that make heavy use of NI (for example, in English verbs with incorporated nouns can usually only appear with further derivation, either as actor or activity deverbals, such as "he's a mountain-climber" and "she went berry-picking").


4.1 - Shared Characteristics

Note, before we get into the various types of NI, that all forms share several qualities. For one, NI is normally a valance-adjusting phenomenon. In other words, if a verb incorporates what was previously its direct object, the new verb will have its valency reduced by one (transitive > intransitive, ditransitive > transitive). We can use the English example of "mountain-climb" again to illustrate this: in "I climb mountains", "climb" is a transitive verb, with "mountains" filling the direct object role. In "I mountain-climb", however, the verb ("mountain-climb", with incorporated object) is now intransitive, with "mountain" no longer a core argument of the verb, but rather a PART of the verb, specifying the type of action conveyed by the verb. This isn't as clear in English as in languages which explicitly mark transitive and intransitive clauses separately (for example, by differing the verb conjugation or using an ergative case-marking system), but it works the same way in those languages (in fact, there do seem to be languages where incorporation has no effect on the verb's valence, but this is unusual).^{[NOTE 1]}

Secondly, incorporated nouns are almost always uninflected, unspecified roots. In most languages, nouns roots are never incorporated along with case markers or plurality markers or noun class markers or articles or demonstratives.^{[NOTE 2]} I'll give a brief example to illustrate this point, from Bininj Gun-wok. Note how the noun root, when incorporated into the verb in the second example, no longer has a prefix marking noun class (NB: the notational system the original source uses in the glosses is very hard to understand, and I may have mistranscribed some of it--nonetheless, the important part for our purposes is the loss of the noun classifier prefix when the nominal is incorporated, which I'm confident I got correctly):

• baginjeng gun-ganj
  3sg.SUBJ/3sg.OBJ-cook-PAST.PERF class:IV-meat
  "S/he cooked (the) meat"
  
  
• baganjginjeng
  3sg.SUBJ/3sg.OBJ-meat-cook-PAST.PERF
  "S/he meat-cooked" (="s/he cooked (the) meat")

Very often, as well, the incorporated form of a noun may differ from the form it assumes as a free nominal. This may be a regular process (in Algonquian languages, for instance, the incorporated forms of nouns beginning in m- usually lose the m-), an irregular process, or even a suppletive process with some nouns, where the incorporated form has no obvious phonological connection to the free-standing noun.

Also, note that there's a restriction in all languages on what nouns can be incorporated. As far as I know, no language regularly allows the incorporation of a transitive subject into the verb.^{[NOTE 3]} Verbs in various languages can incorporate their patient argument (intransitive subject or transitive direct object) or obliques like locations and instrumentals associated with the verb, but agents are never incorporated (actually, I don't know if intransitive semantic agents can be incorporated...does anyone know?). Evidently there are also no known languages in which benefactives or indirect objects (receivers) can be incorporated. All this is probably a side effect of the use of NI to background incorporated participants--see below.

As one final note, be aware that in some languages more than one noun can be incorporated into a single verb. For example, the following Cuiba verb has three incorporated nouns: namaxɨpérɨnadobóbame, "you take off the hair of the skin of the arm" (na-maxɨ-pérɨ-na-dobóba-me = REFL-arm-skin-hair-take.off-2sg).


4.2 - Types and Functions of NI

Most polysynthetic languages make use of NI, though the exact systems and their level of productiveness will vary from language to language (and, of course, not all languages with NI are polysynthetic). Marianne Mithun's work (1984), which I'm drawing on very heavily in this section, distinguishes several types/uses of NI. The first type is the most basic: noun-verb compounds deriving new verbs, with the incorporated noun root serving to narrow the scope of the original verb root. Compare, for instance, the two Yucatec Maya examples below, the first with a separate nominal, and the second with that noun root incorporated into the verb, to narrow the scope of the activity described:

• tinch’akah che’
  COMPL-I-chop-it-PERF tree
  "I chopped a tree"
  
  
• ch’akche’nahen
  chop-tree-ANTIPASS-PERF-I
  "I wood-chopped" (= "I chopped wood")

More advanced types of NI, however, can be used for more complex syntactic and discourse effects. The second type of NI (Mithun's type two) is a way of promoting former obliques to subject or direct object status, by incorporating the former subject or DO of the verb. This strategy is used in many languages, for example, when the underlying subject or DO does not play a major role in the discourse, and the speaker wishes to focus more on the effect the action has on some person or participant who is not an underlying core argument of the verb. For example, English often has body parts as the subjects or DOs of verbs of feeling (as in, "my head hurts" or "she hit my knee"). The second type of NI allows the underlying subject or DO to be incorporated into the verb, which means the affected participant can now be cast as the subject or DO (in the English examples given, this would be the equivalent of, "I head-hurt" and "she knee-hit me", where I'm now the focus of both verbs, rather than my head or knee). All languages which use NI for this purpose also use it in type one situations as described above, to narrow the scope of the verb. A real-life example of type two NI is the following, again from Yucatec Maya. Note how the object changes from the tree in the first sentence to the cornfield (which was initially a locative oblique) in the second:

• kinch’akk che’ ichil inkool
  INCOMPL-I-chop-it-IMPERF tree in my-cornfield
  "I chop the tree in my cornfield"
  
  
• kinch’akche’tik inkool
  INCOMPL-I-chop-tree-TRANS-IMPERF my-cornfield
  "I clear my cornfield"

Here's another real-life example, this time from Blackfoot:

• no’kakíni áisttsiwa
  my-back DUR-pain-it
  "My back hurts"
  
  
• nitáisttso’kakíni
  I-DUR-pain-back
  "I have a backache"

The third use of NI (Mithun's type three) takes place on the level of the discourse, rather than an individual clause or sentence. This is the use of NI to background established nouns: in languages with type three NI, a noun is generally introduced into the discourse with the separate nominal, but in further mentions of it, now that it's established material, it is often incorporated into the verb, unless the noun is highly focused. All languages which use type three NI also use types one and two. As an example of type three, consider the Koryak text below. When the whale is first mentioned, it is referred to with a separate noun; after that, it is incorporated into the verb as established information (I've simplified the orthography a bit to make the example easier to read):



The final type of NI (Mithun's type four) is the incorporation of semantically-broad noun roots in order to classify or specify the verb action; the verb can then in addition still take overt nominals marking subject and object. This is, essentially, the use of NI as a classificatory system. In fact, in many cases, a participant is first introduced with both an overt nominal and classificatory incorporation on the verb, and then for future reference to the participant, all that is needed is to make use of the incorporated classifying nominal. A long but useful example is the following excerpt from a Mohawk story:



Note that the bullheads are first introduced with a separate nominal rabahbót, but also with a classifying incorporated nominal in the verb, -itsy-, "fish". The story continues, telling of how the speaker's uncle found and caught the fish, and then:



In all these subsequent sentences, the bullheads are referred to only through the incorporated noun root -itsy-, functioning as something of a classifier. Mithun reports that languages with productive use of type four NI will have all of types one, two, and three as well.

Mithun's paper was written a number of years ago, and I understand there have been a number of refinements or alternative theoretical approaches proposed by various researchers since then (unfortunately, I don't have access to a lot of them), but the basic rubric is probably close enough to the truth that it can be extremely useful to conlangers, especially novice conlangers who are still trying to sort through all the possibilities that are out there (even given that I left a lot of nuances and detail out).



------------------ NOTES ----------------------

(1) An example of a language in which NI has no effect on the verb's valence is Yanomami. For example, compare the following two examples. In the first there is no incorporation, while in the second the direct object is incorporated into the verb; but in both cases, the subject (if expressed) is treated as ergative--that is, as the subject of a transitive verb:

• (kamijənɨ) sipara japuhii = "I want an/the axe"
  (I-ERG) axe I-want-DYNAMIC
  
  
• (kamijənɨ) jasiparapuhii = "I want it (the axe)"
  (I-ERG) I-axe-want-DYNAMIC

(2) This generalization mostly holds. There are languages, though, where, for instance, adjectives and other modifiers of the (now-incorporated) noun remain outside the verb. See the discussion of type four NI for some examples from Mohawk.

(3) One possible example of the incorporation of a transitive subject/agent is in the English expression "doctor-recommended." But there's a number of ways in which this is not (in my opinion) a very good example. For one, English does not have very productive NI, except, to a limited extent, in order to describe common habitual activities (often hobbies or jobs, like "mountain-climb" or "berry-pick"). And even then, most examples of "incorporation" in English require the verb to be used in a deverbal form (e.g., "he's a mountain-climber" or "he's off mountain-climbing" are more acceptable than "he mountain-climbs on the weekends"). "Doctor-recommended" is even more restricted, and can only be used in very specific circumstances. I don't know of any incorporating language in which semantic agents/the subjects of transitive verbs can be incorporated regularly, or even in more than perhaps a couple of very marked constructions. In any case it's extremely rare.

5 - Other Affixes

So, to summarize what we've covered so far: polysynthetic languages are (basically) languages in which verbs are frequently very long, consisting of a number of morphemes, and which often are the equivalent of entire English sentences. In part this is explained by the ability of polysynthetic languages to mark the person/number/gender/or whatever of both the subject and the object on the verb, and by the ability of many polysynthetic languages to incorporate into the verb a noun root or noun roots to serve various semantic and discourse functions. But these factors alone don't fully explain the length of verbs in many polysynthetic languages. Most (all?) polysynthetic languages also use verbal affixes to mark other notions that are expressed with separate words in languages like English. These often include a tremendous range of notions: adverbial ideas like "quickly" or "by mistake"; temporal ideas like "again" or "yesterday"; verbal or complementizing ideas like "try to..." or "go and..." or "fall" or "make"; locational and directional ideas like "coming this way" or "underneath" or "into the woods" or "downstream"; instrumental ideas like "with the foot" or "by applying pressure"; evidentials like "I know by sight" or "hearsay"; and many others. The handful of examples of polysynthetic words I gave in the first post above present a few such examples.

Here I will only try to give a few more, with some discussion, to help demonstrate the range of possible notions which can be expressed. This is not meant to be exhaustive. (Note that polysynthetic languages also generally use verb morphology to accomplish a range of valency-adjusting operations like causation and passivization, as well as much of the tense/aspect/mood information of the clause, but since these are more familiar categories, I don't address them here).


5.1 - Ojibwe

Ojibwe is an Algonquian language of the Great Lakes region (and also the polysynthetic language I know best). For our purposes, I'm only going to focus on a handful of the concepts that can be expressed in Ojibwe via verb morphology (in addition to polypersonal marking and NI, both of which Ojibwe has).

For one thing, the Ojibwe verb suffixes which serve to mark transitivity and the animacy of the absolutive participant (these suffixes are called "finals" by Algonquianists) often carry fairly concrete semantic notions as well, expressing things like environmental descriptions or action (-itigweyaa, "river flows", -aashi, "wind blows", etc.), instrumentals (-zo/-de, "by heat"; -bizh/-bidoo, "by hand / pull"; etc.), orientation (-shin/-sin "fall/lie/sit", etc.), motion or movement (-batoo, "run"; -oode, "crawl"; etc.), and so on. (I wrote a little bit about Ojibwe instrumental/environmental finals a few months ago here). In some cases the final plus the verb root can come quite close to verb serialization, e.g., ozhaashaabikishin, "slip and fall on stone" (literally, slip-stone-fall -- see my longer description of such combinations here).

Like other Algonquian languages, Ojibwe also has several classificatory verb affixes, which specify the shape, consistency, etc. of a participant (normally an absolutive or oblique). These include -aabikw, "rock, metal or metallic object", -eg/-iig, "sheetlike object", -aabiig, "stringlike object", -aakw, "wooden or sticklike object (including people)", and -min(ag), "round object." Some examples are: ginwaakozi, "it (sticklike) is long"; miskwegad, "it (sheetlike) is red"; and zhaashaabikad, "it (rocky surface/object) is slippery".


5.2 - Tariana

Evidentiality is the marking of the source of the speaker's information, and several polysynthetic languages make use of sophisticated morphologically-marked evidential systems (though a number of languages with evidentiality are far from polysynthetic). For example, Tariana distinguishes five evendentials, marked with suffixes that also mark tense: visual (the speaker saw the action described), non-visual (the speaker knows the action took place based on non-visual evidence, as by hearing it), inferred (the speaker infers the action took place by observing obvious evidence, but did not observe the action itself), assumed (the speaker assumes the action took place based on more indirect evidence, such as custom or common sense), and reported (the speaker's knowledge is second-hand). Examples illustrating all five evidentials (in the recent past tense):

• Juse irida dimanikaka
  "José played football (we saw it)"
  
  
• Juse irida dimanikamahka
  "José played football (we heard it)"
  
  
• Juse irida dimanikanihka
  "José played football (the football isn't in its normal place, José is gone along with his football boots, and a crowd is coming back from the football field)"
  
  
• Juse irida dimanikasika
  "José played football (it's Sunday, José isn't here, and he usually plays football on Sundays)"
  
  
• Juse irida dimanikapidaka
  "José played football (we were told)"

5.3 - Central Alaskan Yup'ik

Like other Eskimo-Aleut languages, Central Alakan Yup'ik has hundreds of derivational suffixes (called "postbases"), which can be added to noun or verb roots to create new stems. A great many of these postbases have very concrete and specific meanings, however, and a selection of some of them, their meanings, and an example of their use in derivation, is provided below. Note that in many cases constructions like this are very similar to NI or other compounding; the main difference is that these suffixes do not represent free verbs--that is, they are bound suffixes, and cannot be used without attaching to a host root.

Some postbases attach to nominal roots and create new nouns, such as:

• -pig = "genuine" (e.g., atpia, "his real name")
• -rrar = "little bit of" (e.g., cuyarraq, "a little tobacco")
• -qlir = "one located in" (e.g., quleqlikacaarr, "the highest one")

Some postbases attach to verb roots but create new nouns, such as:

• -fig = "place" (e.g., misfik, "airport")
• -saraq = "way, device, method" (e.g., yuraryaraq, "how to dance")

Some postbases attach to verb roots and create new verbs, such as:

• -turar = "to keep ...ing" (e.g., nerurallruuq, "he kept eating")
• -ssiyaag = "too much" (e.g., miksiyaagtuq, "it is too small")
• -suit = "never" (e.g., keggsuituq, "it never bites")
• -lngu = "be tired of ...ing" (e.g., nutelnguuq, "he is tired of shooting [but still doing it]")
• -nqigc = "again" (e.g., atunqigtuq, "he is singing again")
• -ngnaqe = "to try to" (e.g., qanengnaquq, "he is trying to speak")
• -nrit = "not" (e.g., qavanrituq, "he is not sleeping")

Finally, some postbases attach to nominal roots but create new verbs (these are the cases that strongly resemble NI), such as:

• -tur = "to eat, use" (e.g., atsarturtuq, "he eats berries"; qayarturtuq, "he uses a kayak")
• -cur/-sur = "to hunt, seek for" (e.g., neqsurtuq, "he is fishing")
• -li = "to make (for)" (e.g., angyaliuq, "he is making a boat")
• -liqe = "to be afflicted in" (e.g., kegguteliquq, "he has a toothache")
• -ngit = "to have no" (e.g., uingituq, "she has no husband")
• -ngu = "to be" (e.g., qayauruq, "it is a kayak")
• -ngqerr = "to have" (e.g., qayangqerrtuq, "he has a kayak")

5.4 - Nuuchahnulth

Many languages of the northwest coast of North America have hundreds of suffixes, called "lexical suffixes", which are bound (like the Yup'ik derivational suffixes above), but which otherwise carry extremely concrete lexical meanings. Aside from their bound nature, and the fact that they normally have no obvious etymological connection to corresponding free nouns, the lexical suffixes with nominal meanings are indistinguishable from noun roots; likewise, aside from their bound nature, and the fact that they normally have no obvious etymological connection to corresponding free verbs, the lexical suffixes with verbal meanings are indistinguishable from verb roots. The same can be said of lexical suffixes with locational, directional, adverbials, etc. meanings. Below, I provide a sample of just a few of the 400-odd lexical suffixes of Nuuchahnulth, a Wakashan language of British Columbia.

Some of the suffixes are very verbal in meaning (and in fact the resulting stem is a verb):

• -ʔay̓imč = "presage ... weather" (e.g., wiiʕay̓imčʔaała, "it is a sign of bad weather", with root wiiq-, "bad")
• -y̓iiḥa = "die from ..." (e.g., c̓axy̓iiḥa, "[people are often] speared to death", with root c̓axʷ-, "hurl point foremost")
• -ḥsaa = "desiring to eat ..."
• -awił = "expecting ..."
• -cḥi = "married to ..." (e.g., masčimcḥinƛ, "she had married a commoner", with root masčim-, "commoner")
• -iic = "belonging to ..." (e.g., ʔiiḥtuupiicukʷaḥ, "[this tama song of mine] belonged to a whale", with ʔiiḥtuup-, "whale" [literally, "big thing"])
• -(n)aanak = "having ... along with one out at sea"
• -aatuk = "making ... sound"
• -cʔakʷ = "acting like ..."
• -cy̓ak = "dressed in ..., appearing like ..."
• -ʕiƛ = "find, come upon ..."
• -ʔinḥi = "waiting for ..."
• -maap = "paying attention to ..."

Others are more nominal in meaning (and the resulting stem is a noun):

• -(š)tuup = "... species, kind of thing" (as in "whale" above, or as in kʷistuup, "supernatural being", from kʷis-, "different")
• -uł = "place of ..." (e.g., k̓uuquł, "hunting ground", from k̓uuq-, "stalk")
• -aʔaq = "... hide, skin"

One of the most notable usage of lexical suffixes, though, is in conveying extremely precise indications of location, space, orientation, direction, and movement. Many of these are body part terms. A sample of some of these locative/directional-type suffixes is:

• -kʷist = "move away (perfective)" (e.g., hičkʷisan̓ap̓aƛ, "they startled them [the birds] off the beach with a light" = hič-kʷist-san̓ap- = illuminate-move.away.PERF-on.beach.CAUS.PERF)
• -ʕaaʔatu = "move down" (e.g., c̓itkʕaaʔat̓asʔaƛ, "it rolled downhill", with root c̓itk-, "roll")
• -cswii = "through" (e.g., kuḥswiiʔakweʔin, "they say [their beams] have holes through them", with root kuḥ-, "hollow")
• -sy̓uč = "exposed, extending out, in view" (e.g., huuʔaksy̓uč̓ičim, "be up (out of bed) early!", with root huuʔakʷ-, "early")
• -is = "on the beach"
• -ʔaaʔa = "on the rocks, in the fire" (e.g., m̓ałʔaaʔamaʔaała, "it's always cold on the rocks", with root m̓ał-, "cold")
• -n̓aaqi = "up on a height"
• -ʔimł = "at the ear"
• -aat = "move downstream, out of the woods"
• -inʕatu = "up the coast"
• -w̓isa = "come out of one's hands, escape"
• -c̓as = "at the crown of the head"
• -(k)swiʔii = "at the teeth"
• -pii = "on the back"
• -stiił = "at the collar bone"
• -p̓iqa = "on the knee"
• -asuu = "under, in liquid"
• -caqs = "at the side of a vessel"
• -wiiʔis = "at the bow"
• -ałc̓a = "at a vertical surface"
• -(q)ḥsa = "at the edge, bank"
• -misa = "on top"

A final example word illustrates how much information can be packed into a single verb this way: ʔaʔaʔaƛqimłḥtimyiłm̓inḥʔaaqƛeʔicuu, "the bunch of you will each move about the house with two dollars on your feet" (PL-two-X.many.round.objects-on.feet-move.about-in.house-PL=INTENT=INDIC=2pl). Or take the following example sentence: ƛiqwiis c̓axy̓ak̓ ƛaatmaqan̓ołʔi, "the spear, which was (long and) made of yew wood, went clear through him on to the beach" (stick.in.vertical.crosswise.position-extend.there-on.the.beach spear sticklike.object.sticking.up-extending.downward-tree-CAUS).

There's a list and discussion here of all the lexical suffixes in a different NW Coast language, the Salishan language Saanich.


5.5 - Case Study: Koasati

Koasati is a Muskogean language of Texas and Louisiana, with one of the most elaborate verbal systems I've seen. The verbal template can be represented as follows:

PREFIXES
slot 9) Indefinite
slot 8) Directional
slot 7) Instrumental
slot 6) Distributive
slot 5) Indirect object
slot 4) Direct object
slot 3) Specific locative
slot 2) General locative
slot 1) Subject/negative markers for verbs of class 1A

ROOT

SUFFIXES
slot 1) Adverb
slot 2) Diminutive/intensive
slot 3) Habitual
slot 4) Intention
slot 5) Ability
slot 6) Realis/irrealis
slot 7) Deduction
slot 8) Modality
slot 9) Dubitative
slot 10) Hearsay
slot 11) Auditory
slot 12) Tense
slot 13) Consequence
slot 14) Discourse functions
slot 15) Enclitics

I'll provide some explanation/examples of some of the prefix and suffix slots below.

• Indefinite prefixes (slot 9). These are aat- (and various allomorphs), "someone" and naas- (and various allomorphs), "something", which marks the object of the verb as being indefinite, as in atchimathátlok "you are afraid of people..."
  
  
• Directional prefixes (slot 8). There are two of these, oht- (and allomorphs), "go and..." (i.e., a translocative) and iit- (and allomorphs), "come and..." (i.e., a cislocative). For example, itthopótlit "he came and passed through..."
  
  
• "Instrumental" prefixes (slot 7). This slot includes the prefixes s(t)- "by means of, with" which is a general instrumental (e.g., choyyí nihá stilihamóhlit, "he rubbed himself with pine resin"), and the prefixes mat- "at a distance" and mas(t)- "full, solid, containing something".
  
  
• Distributive or iterative prefix (slot 6). The distributive prefix is ho-/oh-, and the iterative prefix is hoho-/ohoh-, as in ohachóólit, "they all sewed it and..."
  
  
• Specific locative prefixes (slot 3). There are nine of these:
  
  • itta-, "action on the ground or in fire" (e.g., stittathómmit, "it whipped the ground")
  • oo(w)-, "action in water" (e.g., achoowíllilaho, "I will drown (lit., die in water)")
  • paa-, "action on a raised, artificial, or non-ground surface" (e.g., paachayáhlilifóók, "[one could hear the ice cracking as] I walked across it")
  • on-, "action on or in a vertical plane" (e.g., onólfat, "[the pine tree] sent up a shoot from the side of its trunk")
  • itta-, "in the middle of, in two" (e.g., ittathopótlil, "I pass through the middle of it"; ittakawáthkat, "it snapped in two pieces")
  • ibii-, "action on the human face" (e.g., hokchók chabiihókchot, "the stinkbug squirted foul liquid in my face")
  • ichoo-, "action on the mouth" (e.g., ilichoosáhlillaho, "I will shave my beard (lit., cut my hair around the mouth)")
  • nok-, "action on the human throat" (e.g., konoktithííkalaho, "they will strangle us (lit., press down on our throats)")
  
  
• Adverb suffixes (slot 1). There are a number of these, with meanings such as "very, almost", "really", "in the same way", "all the time", "must, to be obliged to", "to be metaphorically like", etc. A few examples: stootamátlibáhnááhimpat, "they say they might be obliged to trade with them"; ískofíntílk, "we drink too much"; ááthimááli, "[the cypress also] gives fruit in the same manner."
  
  
• Intention suffixes (slot 4). These are -ááhi and -a, as in kááhááhitok, "[he says he forgot] what he would have said"; chokfááthihilkon immánkalááhimp, "he says that he would like me to tell him a traditional narrative."
  
  
• Ability suffixes (slot 5). The most common of these is -halpiisa, "be able to", as in okolchá hóhchalihalpíís, "I can dig a well."
  
  
• Deduction (slot 7). This is -:li, as in ifóóliis, "one might guess that it is a dog."
  
  
• Modality (slot 8). There are several suffixes in this slot, although none are very common, including -áápi, "almost", and -mááthi, "must, would", as in chatámmáápit, "I almost fell"; áhískamááthon, "were you to give me [some cake, I'd eat it]."
  
  
• Dubitative (slot 9). The most common of these is -máámi/-má "perhaps, maybe," which indicates doubt on the speaker's part about the possibility of the action, as in kalifóóniyafoolimáám, "[the oranges] may be from California."
  
  
• Hearsay (slot 10). This is the suffix -mpa, and indicates that the speaker's information is second-hand or was said by another person, as in tálwalimp, "he says that I sing"; wáykat nakáthtohoolimpak, "[Hummingbird] flew off, so it is said."
  
  
• Auditory (slot 11). This is -hawa, and is similar to the Tariana non-visual evidential above, as in aksóhkaha, "it sounds like [the meat] is charring"; fapliyaalihawat, "one could hear the wind occasionally."
  
  
• Consequence (slot 13). There are four of these:
  
  • -:p, "if; when", marking consequential events that are potential or unreal (e.g., akohchóósip kááhaap achóóbon ónk, "now if one says 'akohchóósi', it means an old person")
  • -:k, "if; when", marking consequential events that are a generalized possibility (e.g., ilhóósiik máálon haytanáhkat othááchito, "when people got lost, they turned around and went back the same way")
  • -ska, "because, for, since" (e.g., chokfík chokííboschoolisk immathátlik..., "since Rabbit was quite small, he was afraid of him, and..."; máámoosok ilanawíhkóóthaskan..., "and then, because hunting season was arriving...")
  • -tika, "but" (e.g., alikchí mók hohííchatik sánkon..., "they looked for a doctor also, but were unable to [find one], and...")
  • -y, "but", contrary to expectation (e.g., ...íhsok haaláhkayok..., "...[she pulled herself out], but even though she took it and moved it, [her leg was injured]").

I haven't read through this yet, but you are an awesome person for doing this, Whim.

Yes, good idea for a thread and I await seeing what else you have to say on this subject.

_________________

"There was a particular car I soon came to think of as distinctly St. Louis-ish: a gigantic white S.U.V. with a W. bumper sticker on it for George W. Bush."

My conlang Squalipsh is mildly influenced by Wakashan languages, but currently, its form is not polysynthetic. I'm very interested in reading more, so I can learn how polysynthetic languages work, as I might make Squalipsh polysynthetic in some way.

_________________
[bɹ̠ˤʷɪs.təɫ]
Nōn quālibet inīquā cupiditāte illectus hoc agō
Yo te pongo en tu lugar...
Taisc mach Daró

you
are
a
god.

_________________
næn:älʉː

I am very much pleased to see this, and I look forward to seeing it completed. In particular the section on how polysynthesis arises, as I am currently actively working on just that in a conlang and there's a lot in that area that I'd like to be more certain of.

As you know I've also offered to write a section or two to help take some of the load off, and one of those is waiting in your inbox.

Excellent! I find the noun incorporation type section especially useful

Haven't read through all the case studies yet, but I will soon. Having the necessary information gathered together in one place like this makes trying to make a polysynth conlang much less intimidating!

_________________
High Eolic (PDF)

Very interesting stuff.



So how come all these qualities tend to come together? (or maybe I've missed it in the above)

_________________
It was about time I changed this.

Haven't read it yet, but great job. It seems really complete, useful and it must have taken a lot of work to do, so thanks. Definitely going to read this soon.

_________________


Languages I speak fluently
English, עברית

Languages I am studying
العربية, 日本語

Conlangs
Athonian

This is awesome.
I'm looking forward to seeing the next sections.

On noun incorporation:

Is it typically equally productive with all verbs, or is it usually restricted - or more common with - a subset of verbs? (the latter seems to be the case, sort of, with Yup'ik postbases.)

Thanks everyone!


To some extent I think it varies depending on the language in question. If a language uses NI extremely productively for types three and four (i.e., to manipulate discourse prominence), then it probably allows incorporation with just about any semantically-appropriate verb. But a lot of the time, especially for the first two types of NI, there's a more limited use of the process to form new vocabulary for unitary, culturally-important activities (like, "mountain-climbing", as opposed to "climbing" in general)--in such cases, the number of incorporating verbs and nouns will probably be more limited. Likewise, the second type of NI is by far the most common as a means of possessor raising, so it normally occurs with body part terms, and verbs associated with such constructions. And there seems to be a general tendency for NI to be most common with more semantically empty or broad verbs ("take", "have", "get", "eat/consume", and so on).



Depends who you ask. Mark Baker has tried to argue that they can all be traced to a single macro-parameter (this is part of the terminology and assumptions of the Principles and Parameters school of grammatical theory, don't worry about it too much for now) called the "morphological visibility constraint", which very basically says that all verbs need to index or mark all of their arguments on the verb itself, either through person-marking or noun incorporation, and that this then has other effects in other areas of the grammar. There's some problems with this theory, though: for one thing, it's based mostly on data from Mohawk, rather than polysynthetic languages in general; for another, it fails to capture the similarity between languages like Mohawk and languages like Eskimo-Aleut ones; and finally, as a number of researchers have pointed out, it requires postulating null-realization of agreement markers in some situations (that is, the agreement markers are there, but have no phonetic realization), which is really iffy, theoretically-speaking.*

Personally, I'd suggest that these traits don't necessarily cluster together as much as we sometimes think. It's just that when they do cluster together in a language, the result is a language with really long verbs, which we call "polysynthetic" (the exception is free word order, which is to a considerable extent possible because the grammatical relations of the clause are all expressed on the verb, which will be explained more fully when I've finished that section). Polysynthesis is also very much an areal phenomenon, so it may be that, say, most North American languages have come to share these traits by areal influence rather than because the traits all naturally go together.

This is an area, though, where there's a lot I don't know, so if anyone knows more about why these traits might tend to cooccur I'd be interested to know!




*Before anyone asks, this isn't the kind of null agreement marker where like, third person singular realized as zero is part of a larger, clear paradigm, in which case positing a null agreement marker is fairly reasonable. This is where, sometimes the agreement markers Baker's theory predicts just don't show up, so he has to posit null agreement in those cases to salvage the theory.

Yesss awesome this is great


A minor quibble: I wouldn't call any Caucasian languages polysynthetic other than the NW ones, which account for about a tenth of the total.
Speaking of, I'm fairly familiar with the NWC langs, and have a good number of examples, so if you need some data from Abxaz or Kabardian or something feel free to ask =)

Are you familiar with the phenomenon where some languages (Chukchi and Mohawk are ones I've heard of) can incorporate a noun into the verb and leave behind an adjective? As in John caught the big fish → John fish-caught the big. This is pretty much what you describe as the fourth type of NI, but in another flavor, I guess.

I'm looking forward to seeing more!

Ah, okay, thank you. My main forte is with NA languages, I know barely anything about Caucasian languages. (I'll edit that section accordingly.)


Yes, and I was planning on mentioning that in the NI section, but then it got way out of control and I forgot about it. I'll add it in soon. Thanks!

ROCK'N'ROLL!

This is just so ... awesome. Thank you, Whimemsz!

_________________
...brought to you by the Weeping Elf
Tha cvastam émi cvastam santham amal phelsa. -- Friedrich Schiller
ESTAR-3SG:P human-OBJ only human-OBJ true-OBJ REL-LOC play-3SG:A

Thanks for that. I wasn't sure if these traits "naturally" came together to form common structures among some languages, or whether these qualities were indentified in a few situations.

It's certainly and interesting idea, I shall definitely read through your work here.

_________________
It was about time I changed this.

Awesome!


What are some techniques for analyzing word boundaries? How do you tell the difference between a polysynthetic language and a less-synthetic language written without spaces?

I would imagine that phonological processes come into play, since those often operate only within particular bounds such as the word. One would also need to consider whether bound morphemes qualify as proper affixes or merely clitics as well.

_________________

"There was a particular car I soon came to think of as distinctly St. Louis-ish: a gigantic white S.U.V. with a W. bumper sticker on it for George W. Bush."

That can be tricky for languages you do not speak, and to some extent the notion of word boundaries may not even apply fully to all languages. But for the most part is is generally possible to look at some of the behaviors that words display and usually they will line up with each other. The more that they don't, the more arbitrary the idea of word boundaries becomes.

Prosody: a word has at most one primary stress, so if there's more than one primary stress in a string, it has at least two words.. Languages also commonly base their stress location rule on a word's syllable count, but much more rarely on a phrase's syllable count.
Unity in speech production: speakers who are interrupted or who need to take a breath stop at the end of a word, normally not in the middle of one, and if they make a speech error they start over from the beginning of the word. Language games that apply rules to linguistic units, like Pig Latin, generally do not break up strings any deeper than to the word level.
Morphophonemics: Some morphophonemic processes occur much more often within words than across word boundaries. I'm not clear on which, or by how much.
Syntactic indivisibility: if two morphemes can never have any other material appear between them, they are part of the same word. If material can intervene, and there is good reason to believe that material consists of words (as opposed to morphology), then the first two morphemes belong to separate words.

I can only reiterate what others have said; this is a great thread!


Baker definitely says that intransitive agents can't be incorporated, and he also attempts a theoretical explanation of why this generalization holds. We can't accept that explanation as a definitive truth because it's a bit circular (it rests on the assumption that NI is fundamentally syntactic - as opposed to lexical -, which is what he wants to prove in the first place), but his reasoning sounds logical and the data he provides shows that there is at least a very strong tendency cross-linguistically for agentive intransitive subjects not to be incorporable.

6 - Free Word Order

One consequence of the fact that so much information about an event and its participants is encoded onto the verb in polysynthetic languages is that word order can be much more fluid than in languages like English, where word order is needed to distinguish subjects from objects. By no means do all polysynthetic languages have free word order; however, they are more likely to have freer word order than a more analytic language of the English type. In most cases in languages with relatively free word order possibilities, the actual order in a given sentence will be determined primarily by factors such as the discourse prominence of various participants, a desire to emphasize or background a participant or event, etc. In general, newly-introduced characters or topics will be expressed with nouns placed near the beginning of the sentence, while old, established characters/topics will be expressed with nouns toward the end of the sentence, though of course this varies depending on the language, and the precise nuances needed, and so on.

Some languages go even further than this, and have completely free word order--the order of elements is determined solely by the current discourse needs. Such languages are known as "non-configurational." There's a few characteristics that tend to go along with non-configurationality, including extensive omission of noun phrases (NPs), and discontinuous phrases. I'll discuss them briefly; to go very deep here would require getting into a lot of theoretical stuff that novices probably aren't ready for, so if you want to know more, I'd encourage you to keep looking into it on your own.

NP Omission. This is an extension of a property many of you will be familiar with from Standard European languages, termed "pro-drop." In these languages (like Spanish, for instance), where the verbs are inflected to mark their subject, subject pronouns are rarely used in neutral sentences. This is possible because the subject is already marked with a verb affix, so no additional specification of the subject is needed to make the utterance intelligible (warning: I'm glossing over a lot of complications and counter-examples here). So, in (most dialects of) Spanish, the normal way to say "I'm singing" is canto. In pro-drop languages, when they appear, pronouns tend to be used for purposes of emphasis: yo canto generally means something equivalent to English "I'm singing" with emphatic stress. Non-configurational-type polysynthetic languages, then, extend this pattern in two ways. First, since they mark both subject *and* object on the verb, there's no need to express either with overt pronouns. But secondly, they frequently drop entire NPs, not just pronouns. Even in mildly-configurational polysynthetic languages, like Ojibwe, it's very rare for both the subject and object of a verb to be expressed with NPs.

Discontinuous phrases. Also called "discontinuous constituents," this refers to the property whereby the individual elements of a phrase (say, the demonstrative and noun of an NP) do not need to occur next to one another. I'll provide a couple of examples from Ojibwe:

• bezhig eta ngii-nsaa saawe
  "I caught (lit., killed) only one perch"
  (one only I.killed.it perch)
  
  
• ngoding shkinweg gii-gkinoohmaagziwag niizh maa gkinoohmaadiiwgamgong...
  "Once, two young men studied there at the University..."
  (once young.men they.studied two there at.the.school)

One final note about non-configurationality or free word order languages: while many polysynthetic languages have these characteristics, many non-configurational/free word order languages are not polysynthetic. For instance, non-configurationality was first pointed out in Warlpiri, a non-polysynthetic Australian language, as in the following example (I've simplified/changed the gloss to make it easier to follow):

• kurdujarrarlu kapala maliki wajilipinyi witajarrarlu
  "Two small children are chasing the dog"
  (child-two-SUBJ AUX dog.OBJ chase-NONPAST small-two-SUBJ)

And here's another example of a discontinuous phrase in a non-polysynthetic Australian language, Wambaya (again, I've simplified/changed the gloss):

• ngaragananguja ngiya gujinganjangani jiyawu ngabulu
  "(His) mother gave (him) milk with grog in it"
  (grog.OBJ-PROPRIETIVE she-PAST mother.SUBJ give milk.OBJ)

Question: if a nominal participant is incorporated, how do polysynthetic languages tag a relative clause onto it? I mean, if "Mom gave me milk" is expressed something like "Mom milk-me-gave," what about "Mom gave me the milk that she bought at the supermarket"?

And kudos for the thread. Love it!