OSU08 Morphosemantics Notes 4

Emmon Bach, SOAS, UMass(Amherst)
OSU Linguistics
e m a i l: firstinitialsurname (at) linguist (dot) umass (dot) edu
© Emmon Bach 2008. All rights reserved.
Office hrs: M W 1-3 and by appointment, Oxley 201a
Notes for classes will be posted at / linked to "http://www.people.umass.edu/ebach/courses/osu08-pl.htm"

(9 April) A: Derivational processes II. B: Language Lessons: Western Abenaki.

Part A: Theory

  1. Incorporation 1: Background

    2. History: When European linguists encountered languages of the Americas they were struck by constructions and compounds of the form N+V or V+N in which the nominal element was semantically in some sense an argument (usually direct object) of the verb. Since Western European languages like German or English do not show formations of this sort, at least in any obvious way, this feature was felt to be very exotic. The phenomenon was dubbed "incorporation" (G. Einverleibung gratia Angelika Kratzer). It was erroneously believed that Incorporation was a universal feature of American languages and that it was confined to American languages. [I associate the early discussion with von Humboldt, but haven't checked this out carefully.]

    A.L. Kroeber and Sapir had an exchange about this phenomenon, trying to clarify just what it was, and offering a lot of data about it. They succeeded in correcting the error just mentioned. The exchange is a locus classicus paper and should be consulted by anybody who wants to study incorporation (Kroeber 1909?, 1911; Sapir 1911). Kroeber (1911) concludes thus: (

    The conclusions of the foregoing discussion can be summarized as follows:

    1. "Pronominal incorporation" and "noun incorporation" are different and not connected.
    2. "Pronominal incorporation" is a grammatical or inflectional process.
    3. "Noun incorporation" is, at least sometimes and perhaps always, a compositional [EB i.e. compounding] or etymological process, which differs from the familiar processes of noun composition only in resulting in words of a different part of speech.
    4. All languages belong to one of four classes according as they form compound nouns, compound verbs, both, or neither.
    5. There is no evidence of the existence of any kind of "incorporation" that so far as its process or method is concerned is different from processes occurring in European languages, and it is more reasonable to assume that there can be no such difference than that there must be.

    English analogues do not exist: the closest (for your intuitions) are derivations of noun + verb + -er or -ing into basically nominal compounds: meat-eater, tree-hugger, cigar-smoking (... not permitted here. Note not possible in English:

    1. *They don't meat-eat ostrich.
    2. *I only tree-hug red cedar.
    3. *It is illegal to cigar-smoke Havanas.

    These examples, however, might translate readily into Mohawk or other languages, as examples of one type of Noun Incorporation.

    Following this early discussion, I will exclude from consideration as Incorporation: verbs with Subject or Object marking or cliticization.

    Excursus: Nature abhors a vacuum: the problem of exact synonymy.

    Discussions of such matters as incorporation often turn on the question whether two constructions are synonymous or not. It has often been observed that synonyms tend to diverge over time. Here just a nota bene is in order. Speakers often attribute differences of meaning ad hoc. It sometimes will take quite a bit of work and argument to tell whether non-nonce meanings are "really there."

    Excursus: Etymology

    Etymology studies the origins of words and yields etymologies,theories about how a particular lexeme came to mean what it means. It is often hard to tell whether you are doing etymology or something more synchronic, such as morphological analysis.

    It has often been noted that an item like baby-sit, which can be used as a transitive verb, is a backformation from baby-sitter:

    1. Who is going to baby-sit Trixie?

    Even so, the semantic relation here is not that of an object to transitive verb. Note that such a formation can be the model for direct coinings as in (just coined, EB):

    1. Susan is going to land-sit my Northfield property.

    There was a debate in the 80's about incorporation (Mithun 1984, 1986; Sadock 1980, 1986, 1991; Allen et al. 1984), mainly about the syntactic versus morphological status of incorporation. Mithun is the best source for a survey of types of incorporation. A second point of contention was whether incorporation was ever obligatory (Allen et al., 1984 is a source for data on a language -- Southern Tiwa -- where it was claimed that the only way to get an indefinite object into a sentence was by incorporation.)

    Baker's book (1988) incorporates "Incorporation" into the (P & P) model of the time and gives it a new meaning. It is a thorough-going syntacticization of the process. Along with a lot of new ideas, Baker invokes the UTAH principle, which is something like the Relational Grammar principle of Perlmutter and Postal, which posits a uniform semantic source for structures exhibiting the same Grammatical Relations:

    ...within the framework of Relational Grammar it was proposed that there is a uniform mapping from semantic properties of lexical items to the relational structures associated with them in the initial stratum. This Uniform Alignment Hypothesis (Perlmutter and Postal, 1984; Rosen, 1984) was adapted to the X-Bar framework by Baker (1988) and stated in terms of Thematic Roles as the Uniformity of Theta Assignment Hypothesis (UTAH): every lexical item with a given cluster of thematic role assignments will appear in D-Structures of a determined type. (Bach, 1994)
    Baker's work led to a continuation of the morphology - syntax debate, couched in terms of lexicalist vs non-lexicalist ideas. For example, Di Sciullo and Williams, 1987, come down on the nonsyntactic side.

    An idea that comes up in various forms in many of the discussions is the idea of multiple representations, most directly in Sadock's autolexical theory, in the idea of "corepresentation" in Di Sciullo and Williams (and now blossoming out in Williams' latest book Representation Theory). In a sense, the syntactic accounts like Baker's bow to this idea as well, in that the relevant pieces of an incorporation structure are "in different places" (in a derivation) or at once.

  2. Incorporation 2:

    3. Mithun, 1984, considers Noun Incorporation to be a kind of compounding. Compounding is taken to be a morphological process which combines free forms ("words" maybe) into new free forms. She distinguishes four types (the italicized descriptions are Mithun's, the categorial characterizations are mine):

    1. Type I: N + TV > IV Lexical Compounding
      2. pseudo-English examples: meat-eat, berry-pick,...
      nearest analogues: synthetic compounds: meat-eating, letter-writer,...

    2. Type II: N + TV > IV/XP The Manipulation of Case
      3. In this type a noun is combined with a transitive verb into a verb that takes an argument that is marked by some oblique case or functional element like an adposition.

    3. Type III: N + TV > IV [??] The Manipulation of Discourse Structure

    4. Type IV: N + TV > TV Classificatory Noun Incorporation

    Note that the incorporation of a noun in this type does not change the valence of the transitive verb. The effect of the operation is to restrict the class of arguments that the verb takes. Chung and Ladusaw (2004) have written extensively on this kind of incorporation. As M's descriptions show, the classification is in part a functionalist one. One of her main theses is that different ways of expressing the "same" content in languages are put to different uses. Thus the main difference between Types I and III appears to be in what the process is used for: for Type I to yield a lexical item indicating some culturally important and stable activity, for Type III to produce structures in which part of the content is backgrounded.

    The categorial perspective is helpful here, I think. The primary classification of processes or rules or whatever will be guided by the categories of the grammar. Thus unlike X-bar type theories (as in Selkirk, 1982, and other writers) the rules or processes will not yield members of classes like Verb, but rather TV, IV or whatever their analogues are in the word-grammar. (This amounts to requiring that the output of the rules in the X-Bar systems include subcategorization information.)

    On the other hand, the categorial treatment needs to define supercategories like Verb in some way. We do this in the case at hand by observing that verbs end up as categories that take term phrases (TP: DP) to make sentences. If we abbreviate this highest category as IV (as we have done stillschgweigend above), then a verb will be anything that ends up as this category: TV = IV/TP, ditransitives will be (IV/TP)/TP etc. IV (in English type languages) is TP\S. A notation for this kind of category will be IV$, and we can use it also for abbreviations like IV$/TP (for the use of such abbreviations, see various of Mark Steedman's writings).

    Examples: IV/TP (TV), (IV/TP)/TP (TTV), (IV/TP)/PP are all instances of IV$. An immediate consequence of this categorial treatment is that there will be no such thing as Subject Incorporation. Why not? Because Noun Incorporations (for the types that change the valence of the verb) yield verbs by satisfying an argument position, the result then being a Verb that has one less argument position to fill. But for intransitives this would yield a Sentence and not a verb. It would be interesting now to study apparent exceptions to this claim. Some putative examples in Sapir, 1911, from Paiute: 266 (CW V: 43), all referring to natural phenomena or states. Conjecture: "sentential verbs." (Kroeber, 1911, mentions the idea that incorporating a subject would just yield a sentence.)

    In general, Mithun and others (Di Sciullo and Williams, 1987) go for a lexical account, meaning non-syntactic, no Movement or whatever. But Sadock has presented compelling evidence that "remnants" of DP's can occur outside of the incorporating word. Although some of the examples in the literature can be argued to be combinations of a incorporating word and an independently generable DP, this won't work for examples like this one:

      (Greenlandic Eskimo, Sadock [which reference!?], 96-97):

    1. kunngip panippassuaaqarpoq
      2. kunngi-p panik-passuaq-qar-poq
      king-ERG daughter-many-have-INDIC/3s
      `There are many king's daughters (i.e. princesses).'
    2. *kunngip takuvunga
      3. kunngi-p taku-vunga
      king-ERG see-INDIC/1s

    Note that Sadock's examples from Greenlandic do not involve combining free forms or stems. In fact they cannot, as Greenlandic like Yup'ik and Haisla (Wakashan in general) does not allow compounding. So Sadock is considering putting together bound affixes and stems the same sort of animal as the classical cases of incorporating a Noun into an independently useable Verb. But note also that Sadock's general theory of Autolexical Syntax makes this a very natural thing to do. An affix like -qar `have' is syntactically independent but morphologically dependent.

    Strong lexicalist theories such as that espoused by Di Sciullo and Williams (1987) predict the examples like the ones just given are impossible, hence they are falsified or require some epicycles.

    The general questions here have to do with the organization of a grammar and the relations between phrase grammar and word grammar. I believe that a lot of confusion has come from the identification of "morphological" with "lexical." A full discussion of the problems here requires discussion of a range of issues involving phrasal affixation, the possible arguments of bound affixes, and issues about clitics. The questions have their semantic side.

  3. Haisla: Some Lexical Suffixes and Processes:

    4. The first of these suffixes is relevant to the incorporation puzzle. Similar bound forms in other languages have been treated by one or another writer as implicated in Incorporation in Baker's sense, the others don't seem to be at all:

    1. -[g]ila making and going
      2. One suffix or two? The notation [x]Y stands for segments that appear or not depending on the phonological form of the host duentila make name (duent)
      gélwʼgila make canoe (gélwʼa)
      gugʷasila make or found a village (gugʷáas)
      laugsila logging, make "logs"

    2. =ilh and =isinside and outside
      3. The marking = (from Boas) signifies that the suffix induces "softening" (voicing) of the final obstruent of the host.

      These two are extremely productive and widely used, verging on grammatical affixes. For the lexical-conceptual meanings we might compare the English lexeme to picnic which contains the idea of having a meal plus some indication of having the meal al fresco or outside. Similarly, the Haisla word hamzilh means "to eat in a restaurant." In the present context, the relevant aspect of the meaning is that they are not in anyway related as arguments to the host.

    3. +lʼit is said...
      4. This suffix is something like a discourse particle or evidential. It is a question whether it should be considered a derivational affix at all rather than a clitic in the grammar proper.

Part B: Language Lessons

  1. Algonquian and (Eastern Algonquian) Western Abenaki
    2. General reference: Bach, Emmon. n.d. Sketch of Western Abenaki Grammar.
    Bibliographical information on Western Abenaki:     Bach [2001] .

    On the importance of philology in the endangered languages industry: Bach 2004b.

    Here's a map of the traditional Western Abenaki lands.

    A bit about the history of Algonquian and Western Abenaki studies,

References: will be found at here at "http://www.people.umass.edu/ebach/courses/mrphrefs.htm". They will be given in short form e.g.: Frege 1892