OSU08 Morphosemantics Notes 3

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"

Derivational processes: I; Languages: Haisla, Western Abenaki.

Part A: Theoretical

  1. Two Grammars

    2. Picking up from our discussion last time, let us assume that there are two grammars, each with its system of "tactics" (how to put things together), its phonology and its semantics:
    1. phrase grammar (PG)
      2. This is the grammar for the domain of syntax, in the traditional sense. It uses categories that correspond to notions like Noun, Verb, DP, VP, in various frameworks. We will lay out the categorial analogues or correspondents to these classes below.
    2. word grammar (WG)
      3. This is the account of morphological words. It traditionally uses categories like Root, Stem, Suffix, Prefix, and the like. We need to ask how the morphological categories and their classes of items are related to categories and types of the phrase grammar.
    Orthogonal to these two systems is the lexicon and the system of rules, principles, etc. that enable the creation and analysis of lexical items.

  2. Categories of Phrase Grammar (Syntax)

    3. In a categorial grammar, the system of categories is built up just like the type theory, by giving a list of Basic Categories, and a recursive definition of derived categories. For example, one scheme for a simple concatenative grammar is this (compare Montague, 1973: PTQ), giving a definition of the set CAT of syntactic categories:

    1. BCAT = {s, n};
    2. BCAT is included in CAT;
    3. if a and b are members of CAT, then so are (a/b) and (b\a);
    4. nothing else is in CAT.
    (outermost parentheses will be suppressed)

    (compare Notes 2 for a more general formulation of this definition). Here the slashes indicate directionality: the slash "leans toward" the argument, and the notation signifies that a member of the class of expressions indexed by the category -- I will say "member of the category" for short -- a/b (b\a) can be concatenated with a member of the category b to its right (left) to form a member of the category a, in shorthand: REPEAT WARNING: different writers on categorial grammar use different notations. another one is to always put the argument category on the right, so these writers would notate thus: a/b and a\b.

    1. a/b b ==> a FA Forward Application

    2. b b\a ==> a BA Backward Application

      3. In addition, we have "harmonic function composition":

    3. a/b b/c ==> a/c
      4. FC Forward Composition.

    4. c\b b\a ==> c\a BC Backward Composition.

      5. TL Order-preserving type-lifting:

    5. a ==> b/(a\b) TLF Forward Type-Lifting
    6. a ==> (b/a)\b BLF Backward Type-Lifting

      7. To help remember this: see the category for term-phrases below: t/(e\t)
      Think of type-lifting like this: given a thing of a certain category (type) a, then it can be interpreted as a function from (functions that take a's to b's) to b's. Keeping the specification of directionality ensures that you won't get anything new by this route.

      Easy arithmetical example: Square is a function that takes numbers and delivers their square: Square(2) = 4, etc. Fat-2 is a type-lifted 2: so Fat-2(Square) = 4, Fat-2(Cube) = 8, etc.

    In addition to the above setup, most workers in CG assume a scheme or schemes for conjunction, often cross-categorial.

    Note that the schemata showing the composition or analysis of complex expressions in categorial grammars are NOT to be interpreted as phrase structure diagrams. They are instead like the analysis trees of Montague's PTQ. Their closest analogues in the transformational generative tradition are the Transformation Markers of early classical transformational grammar, vintage pre-Aspects (i.e. before 1965).

    Some Phrasal Categories: Here are some important examples of syntactic categories from the literature, with their closest analogues in some versions of X-Bar theory. For the time being I will use lower-case letters for the categories, to avoid confusions. Recall I assume basic categories {n, s} ("name", "sentence") mapped into types e and t, respectively. We ignore worlds and intensions for the time being:

    1. n - expressions for individuals: Pat, Kim, it, the hurricane Esmerelda
      2. (no analogue in X-bar theory)
    2. n\s - tensed intransitive verb-phrases: swims, walks fast, loves Pat, ran to the store, gave Kim a popsicle
      3. abbreviation: ivp - something like VP
    3. s/(n\s) (nominative) term-phrases: everyone, some fish, whoever Kim saw, .. and category-lifted (type-lifted) versions of all expressions of category n
      4. abbreviation dp - something like DP, depending on how we treat case
    4. (n\s)/n - tensed transitive verb phrases: sees, persuaded to leave, loved unrequitedly
      5. abbr: tvp some analyses in the X-Bar tradition and in early TG have this or something like this category.
      Note that complex member of this class require special treatment to account for the placement of the object right after the (morphological) verb.

      Two further categories are needed for (1) common noun phrases like fish, boat that Pat built and so on, and (2) determiners that take such phrases to make tp's (DP's). Montague assigns common noun phrases to a category that maps into predicates (like intransitive verb phrases), but is syntactically distinct. For now let's just adopt an ad hoc label:

    5. cnp - common noun phrases: fish, house that Jack built
      6. corresponds to N, NP, N-bar, NOM, etc. Now we can say
    6. dp/cnp - determiners: the, every, some
      7. abbr: det - Det

    aExcursus: Predicative expressions.

    In standard predicate logics, there is a family of predicative expressions: one-place, two-place,... Or if schoenfinkeled (curried) a series of functions of types (e, t), (e (e, t)),... Focussing on one place predicates, these are taken sometimes to correspond to natural language intransitive verbs, and also (non-relational) predicative adjectives and (non-relational) common nouns. Montague (PTQ) followed this tradition for common nouns, giving them a category which mapped into the type (e, t), but syntactically distinguished: t/e, t//e (using his non-directional categorial notation). Adjectives are another story we'll come back to. The setup just given can be taken to be a notational variant of the PTQ idea. It is possible to pursue other ideas about common nouns, as much discussed especially in the philosophical literature (a basic reference: Gupta, 1980).

    Excursus: Crosslinguistic Syntax.

    There has been a long discussion about whether or not the core or all syntactic categories are universal, and in what sense. For example, is there just a universal stock from which all languages draw, or do all languages use the same set? These questions are closely related to questions about the Universal Base Hypothesis, put forward in the 70's of the last century by writers such as G. Lakoff, E. Bach, J.D. McCawley, and others, abandoned by some as not making any empirical claim, more recently revived or assumed (I think) in some form or other. Some of the languages of the Pacific Northwest have played an important part in this discussion, early and late. Claims have been made and debated that the Wakashan languages, such as Nuuchahnulth and Haisla do not have a distinction between common nouns and verbs. This idea is based largely on the fact that any member of the open class of stems can be used predicatively. We'll take this up as a special topic later in the term, and look at some of the recent discussions by writers such as W. Jacobsen, Jelinek, Demirdache and Matthewson.

    For now, look at these examples from Haisla:

    1. Duqʷelanugʷa mayaas.`I see a cat.'
    2. Mayaasnugʷa.`I am a cat.'
    3. Guxʷ ḡuda.`That one (remote) is a house.
    4. Qalhela ḡuda.`That one (remote) is taking a walk.'

    X-Bars and the like.
    In X-bar systems, there is the idea that category labels are shorthand for combinations of feature specifications, such as [+N,-V, Bar-2], where the Bar feature represents a kind of vertical dimension, starting with the value 0 for the lexical items associated with categories like N ([+N, -V]) and so on. Montague's PTQ used a different notion of Basic and Phrasal categories and the first statement of the grammar -- not a rule -- said that for all categories A, the set of Basic members was included in the set of Phrasal members of that category.

  3. Derivation vs Inflection

    4. The distinction between Derivation and Inflection is often stated like this:

    If we take the definition of a language (grammar) as a set of basic elements (the lexicon) and a set of rules cum operations from which we can derive (generate, specify, enumerate) linguistic signs for each of the (syntactic) categories of the language, each sign k-tuple including interpreted phonological elements, a syntactic category, and an interpretation, then any change in the lexicon is a change of the language. Thus, this dimension of linguistic theory deals ultimately with a diachronic process. That at any rate is what we will start with in our deliberations today. (See Dowty 1978, 1979: Chapter 5.)

  4. Synchrony and Diachrony

    5. Throughout the 19th century, linguistics was historical linguistics.

    Compare the great German linguist, Hermann Paul's pronouncement:

    "Es is eingewendet worden dass es noch eine andere wissenschaftliche Betrachtung der Sprache gäbe als die geschichtliche. Ich muss das in Abrede stellen." [It has been objected that there is another scientific way of looking at language besides the historical. I have to disagree with that.]

    Prinzipien der Sprachgeschichte.

    It has been a matter of principle in modern linguistics -- basically through the whole of the last century and well into this one, to give priority -- to the description of a language as a synchronic system, with lipservice usually to Saussure. Note that the move we have just taken, to focus on lexeme formation as a diachronic process requires a rather radical rethinking of a lot of data and questions. For a similar move in phonology see Juliette Blevins' book on Evolutionary Phonology (Juliette Blevins 2004).

    So in effect we are removing lexeme-formation from the grammar. Note that this doesn't mean that operations on words (in some other sense, say morphological) are excluded from the grammar. That's still up for grabs, it certainly is required that we modify morphological words for inflections, on at least one way of doing inflectional morphology, my currently favored way, which we will pursue in our next class.

  5. Are words made out of words?
  6. Are words made out of phrases?
  7. What is a base? a root? etc

  8. Categories of Word Grammar (Morphology)

    1. Vertical Categories:

      2. I mean here categories like Root, Stem, Base, Prefix, Suffix, (morphological) Word, etc. We might view these as analogous to the bar-features of X-Bar theories.

      It would seem that there is no substantive semantic import to these categories or features that might be used to encode these distinctions. In phrasal syntax as well there is no semantic uniformity associated with similar distinctions (contra Jackendoff, 1977):

      Excursus (from Bach, 1994):

      There are two ideas about the relationship between syntax and semantics that have been enunciated by works in the tradition of X-Bar syntax. First, the Crosscategorially Uniform Semantic Interpretation Hypothesis (CUSIH) was proposed by Jackendoff (1977, see Williams, 1981): it says that the interpretation of items at a given bar-level will be fixed and independent of the particular category that the item belongs to. So, for example, all lexical items will have interpretations of the same type, since they are all of the bar-level 0. I believe that this idea is best understood (within the present context) as claiming that the semantic operations associated with a particular construction at a given level will be the same across "horizontal" categories. For example, nominal and verbal counterparts will be interpreted as functions taking their complements as arguments, optional elements at level 1 will be interpreted as restrictive modifiers, and so on. We'll return to this question below in the context of comparing the denotations of verbs and nouns. (In Montague grammar and extensions thereof these two examples are both instances of the single semantic operation of functional application.)

      Second, 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. This hypothesis seems to be a revival of some form of the hypotheses about the relationship of semantics and syntax associated with the program of Generative Semantics. Unfortunately, it is difficult to test these claims without some explicit semantic theory, so far lacking in both frameworks.

      (On these questions, see now Culicover and Jackendoff 2005.)

    2. Horizontal Categories:

      3. These will be whatever are the analogues to Noun, Intransitive Verb, TV, etc. One view of these is that they are exactly the same as or drawn from the same set as the basic categories of phrasal syntax. I will argue against this view in the subsequent discussion.

    3. Lexical (Listeme) Categories

      4. Given the view of the lexicon as a repository of all the arbitrary conjunctions of meaning and form, we might expect that the elements of the lexicon -- the listemes -- could belong to any category of the phrasal grammar (and analogously for the word grammar). This is largely true, but in addition many lexemes go beyond this, being strings or structures excerpted from the phrasal combinations. For example, in Notes 1 we mentioned lexical items like put up with, which does not belong to any syntactic category at all. Here are some more examples:

      1. take ... to task
      2. it beats the shit out of ... `it baffles ...'
      3. keep tabs on ...
      4. get ... 's goat
      5. x hold x's tongue
      6. x hold x's breath

        7. Di Sciullo and Williams (1987) claim that lexemes are confined to categories of the syntax. To do this they assume that items with holes in them like those just listed actually have structures like these:

      7. take DP to task
      8. get DP's goat

        9. (Here and below I update `NP' to `DP.') Here are more examples from Di Sciullo and Williams (1987: 6), intended to show that lexical items can come from all kinds of syntactic units:

      9. AP: all wet
      10. PP: in the dark about DP
      11. S: the cat has got DP's tongue
      12. N': that son of a bitch [sic: presumably `son of a bitch']
      13. DP: The Big Apple
      Of course, many or most lexical items will be members of the large open word classes such as V, N, A, and so on, or their analogues in various frameworks.

    Part B: Language Lessons

    1. Basics on Haisla: notes from a short dictionary

    2. Word Formation in Haisla

      1. General

        1. Vertical categories: Root, Stem, Base, Affix, Clitic

        2. Horizontal categories: pred, pred\pred; ??: pred\n, pred\v,...

          3. Here's where the famous controversy about lexical categories comes in

        3. No compounding: words are built out of combining Roots, Stems, or Bases
        4. (almost) all affixes are suffixes: so categorially all affixes are of categories b\a plus segmental adjustments in the phonology, either automatic or governed

        Examples of one derivational rule: R-338 (

        ItemGlossRootDerived formGloss
        saakgrizzly√sax-sasenʼágo_after grizzly
        kʷenáaqmink √kʷenqkʷakʷenqʼágo after mink
        kʷíkʷenaqfrog √kʷenqkʷakʷenqʼágo after frog

      2. Examples of affixal rules

        1. R-338 `Going after X'
          2. MCAT: root\root
          PhonOP: Suff!(attract-stress(a))(RED-a): where
          RED-a(α)
          shorthand for RED-a:
          C1VC2(C3) ==> C1aC1aC2(C3)
          SEM-OP λP[GO-AFTER(P)]
          Suff! = Harden last segment of host (refer to hardening table, mostly glottalize, but with wrinkles)

          What we need to know minimally for an affixal rule:

          1. Identifier:
          2. Phonological form: here: Consonantism; Vocalism
          3. Morphological form: here: Root:
          4. Semantic (value)

    3. If time: Introduction to Algonquian: Focus on Western Abenaki (Eastern Algonquian):
      4. notes on grammar and a paper by EB.