Chapter 1, Hopper and Traugott

Handout for LING/SARS 319/519

Get print version for this document here.

  1. The first issue to be dealt with in discussions of Grammaticalization is the problem of Polysemy: (a.k.a. multiple meaning.)

    going to go (go to VB) AUX --> gonna go
    going to London (go to NP) (Motion) *gonna London

    When AUX expresses immediate future, it derives from a motion VB in a certain CONTEXT, and this does not apply to other contexts. The reduction of going to to gonna does not happen in the context go to NP.

  2. Grammaticalization: focuses on how grammatical constructions

    We are concerned with the question of

    There is a tension between unconstrained lexical structure and constrained morphosyntax and morphological structure.

    We are interested in providing a conceptual structure for a principled account of the relative indeterminacy in language and the basic non-discreteness of grammatical categories.

    Grammaticalization is therefore 2 things:

    1. the actual phenomena that the theoretical framework of Grammaticalization seeks to address; and
    2. the processes involved in Grammaticalization .

  3. Since Saussure, there has been a dichotomy between

  4. Grammaticalization can be studied 2 ways:

    Harper & Traugott combine these, but emphasize the historical. (I would try to balance these more: for a diglossic language, you can see things both ways. HS)

  5. Gonna/going to illustrates several facets typical of Grammaticalization .

    1. change in a very local context: purposive directional construction with non-finite complements:

      • I am going to do X (travelling, leaving)

      • *I am going to London (to do X) (directionality with locative adverb.)

    2. Change is possible because there is an inference of futurity with purposives: I am Xing to do Y means Y is in the future. Without overt directional phrase futurity takes over, and becomes salient .

    3. Reanalysis: Shift from purposive to AUX (immed. future.) involves reanalysis of

      • of the 'be going to' phrase

      • of the verb: [I am going [to marry Bill]] --> [I am going to marry Bill].

      • Aspect changes from progressive aspect to immediate future.

    4. Reanalysis is discoverable (manifest) only when VB following be gonna is incompatible with purposive meaning, e.g.

      • I am going to like Bill

      • I am going to go to London

      the contexts of "be going to" have been generalized or analogized to previously unavailable contexts.

    5. Once reanalysis happens be going to can undergo phonological reduction typical of AUX:

      is ---> 's, do not ---> don't, can ---> cn, have --> 've
      so going to --> gonna i.e. there is no phrasal bracket between [go]-[ing]-[to], it's just [gonna];


      and therefore we are at a whole new level of analysis: a phrase consisting of lexical items has been replaced by a new monomorphemic item with different meaning, and perhaps grammatical "meaning" as well.

    6. These stages all coexist in modern English, even though the process began in Middle English.

      The original purposive meaning continues to constrain the use of AUX. be gonna is the future of intention, plan, schedule, and can occur where will can't:

      • If you're gonna be obstinate about it, we'll have to take another look.

      • * If you will be obstinate about it, we'll have to take another look.

      The persistence of the older meaning gives a kind of overlap; they coexist, and the new one reinforces the older meaning.

    7. The Main verb go is rather general; it means any kind of motion away from the speaker, so such a verb can be recruited to to participate in this process. (A more specific kind of verb couldn't be used for this! )

    8. During Grammaticalization some of go's concrete meanings have been lost ("bleached out") such as the notions of motion and directionality but new meanings have been added: they are

      • more abstract

      • speaker-based (i.e. temporal meanings are based in speaker time; the meanings are more subjective, more colloquial; they express a point of view (?), or a motivation.

  6. What is a Grammaticalized form? Let's distinguish between

  • Types (or classifications) of Grammaticalized'd forms:

    Another way to see this is as a cline of grammaticalization:
    shift is not abrupt, but moves slowly, in stages. transitions are gradual (similar in type across languages.), such as:

    There is disagreement as to where, how many points there are on the cline. Some argue that the path should be: context item (lexical item)--> grammatical word --> clitic --> inflectional suffix.

    as in X full of Y --> X-ful --> hopeful.

    Hopper & Traugott are much concerned with the cline of grammaticality, and its conceptualization. Some (Heine) are concerned with how one thing seems to imply another; whether there are channels that are followed (paths), or push-pull chains (?), or inferencing.

    H&T say that at least we must recognize:

    This needs to be worked out (below).

  • One more point: note the dichotomy between the periphrastic and the bounded:
    		    /  \
    	  	   /    \
           	          /	 \	
    		 /        \
           		/          \
    	Periphrastic     Bounded
    	have waited     waited   TENSE
    	of the X        X's	POSSESSION
    	more curious   curioser COMPARISON

    < What starts out as periphrastic becomes "synthetic", i.e. is realized by affixation; the diachronic tendency seems to be toward affixation. Periphrastic constructions (they say) coalesce over time and become morphological. Other examples:

    1. Definite articles becoming affixes in Scandinavian, in Istro-Romanian, e.g. Danish -en and -et ( dreng-en 'boy' and hus-et 'the house' are definite articles that were previously postposed in Old Norse. Also passive as in Scandinavian.

    2. Use of aux. verbs to form tenses periphrastically, such as Hindi pres. tense, Romance languages use of formerly periphrastic tenses e.g. cantare habemus (Latin. we have to sing) --> Ital. canteremo 'we will sing.' ( cantarabemus --> cantaremos .. ).

    3. Maybe Tamil -aam modal is an example of this kind of development from aakum i.e. pookal + aakum 'going will happen' --> pookalaam 's.o. may go.'

  • Second diachronic tendency that makes periphrasis/bondedness distinction important is renewal i.e. tendency for periphrastic forms to replace morphological ones over time, in a kind of cycle, which they illustrate with examples from I-E, Latin, and French on pg. 9. The peripheral. form becomes reduced and morphologized, then it is replaced by another peripheral. form, which is then morphologized, and which is then it is replaced by another peripheral form, etc. Each new form has some nuance of meaning diff. from the previous, but eventually it takes over, coalesces/reduces, and then is replaced. (Cite Hodge 1970 for many examples of this; see D.N.S. Bhat on evolution of tense forms in S. Dravidian, which show incorporation of forms of iru as past tense markers, which then get reduced and require renewal.)

  • More examples .

    1. Let's (let us) --> lets (let it be the case that; let it happen that) with a meaning of 'adhortative' or 'first person imperative' or even 'condescending encouragement' of 2nd person ("lets eat our liver now, Betty").

      This shows various stages:

      • full verb let has altered its semantic range somehow. Perhaps Grammaticalization in early stages often/always shows shift in meaning , and only in a spec. context, here the imperative. Permission or allowing has become extended (in part of the paradigm) to encouraging or suggesting s.o. to do s.t. Let has become less specific, but also more speaker-centered or centered in speaker's attitude to the situation . (Began in 16th century or earlier.)

      • First-person plural. pronoun us became cliticized ( let's and from word-plus-clitic we get single word lets . This is fine as long as still used with plural subjects, but when used with singular. subjects, we can't do this any more. Final s of lets loses its status as separate morpheme. This illustrates a general shift of
        word --> affix --> phoneme
      • Next stage brings more phonological reduction i.e. [ts] is often reduced in rapid speech to [s] as in le's go or even 'sgo ; come on, let's go --> c'mon, sgo [kmã:sgo]

      • Next we have the routinization or fixation of a meaning or discourse function for lets that was formerly freer; it's on its way to Grammaticalization , singling out something from a whole paradigm of forms and specializing it as (here) the adhortative :

        Let'm go to hell, I don't care.
        This form does not have the meaning Permit him to go... but rather "I suggest he go ..." H&T say this is provisional and relative, not permanent; it may not survive, but for now it is available and helps to build interactive discourse (this is what they mean by these things being more attitudinal, speaker-centered, discourse-centered and interactive.

      H&T show this as part of a tendency in English toward phrasal expression of modalities of the verb, rather than morphological as in Old English (example p. 14), e.g. use of may, be going to, keep Verb-ing.

    2. H&T's next example is of a process in Ewe very similar to Tamil! There is a verb [bè] which means 'say' which is being grammaticalized into a complementizer whenever a verb other than [bè] is used in the sentence: this is obviously similar to use of Tamil en- as a complementizer ( enru ) such as 'He said that he would come' i.e.

      avan naan vareen-NNu sonnaan
      For more discussion of this in Tamil, look here . As they note, the verbs involved are verbs of speaking, cognition and perception; they are similar to a verb meaning 'say' because they can have objects that are propositions. Same in Tamil/Dravidian, where the verbs that can use enru as the complementizer or embedding marker are verbs like nene, nambu, utteesamaa iru etc. ('think, hope, have the intention, etc.') But, the meaning and the morphology of the 'say' verb (in both Ewe and Tamil!) is essentially lost in the Grammaticalization of it as complementizer. In Tamil, of course, it also is phonologically reduced by loss of the initial vowel e . Syntactically, as they illustrate on pg. 16, there is also shift from [x[y]] to [x[y[z]]]. In Tamil, it would be change from :

      [[avan naan vareen]-NNu sonnaan] .

      [[[avan naan vareen]-NNu] sonnaan] .
    3. Agreement Markers. Here H&T give an example of an already grammaticalized form becoming more grammatical. They give example from French where il which is the masc. pronoun has become (in non-standard Fr.) AGR and is bound to the verb, not signaling gender:

      ma femme il est venu
      my wife AGR has come

      (This example not very satisfying to me.) How about the example of Tamil accusative marking becoming the marker of definite-ness? as in:

      naan pustakam paDicceen 'I read a book'
      naan pustakatte paDicceen 'I read the book.'

    Summing up: Grammaticalization raises many questions, many of which are as pertinent for Tamil as they are for EWE or for other languages. I wonder whether we focus too much in descriptions on standard languages and ignore the colloquial, where we find (e.g. in English) interesting things that tend to be ignored; in Tamil we don't have this luxury if we are to adequately describe the language most people speak but don't treat seriously. Perhaps grammar is a way-station, constantly being reorganized, and Grammaticalization is a voyage that makes stops from time to time.