Synopsis, Harper & Traugott, Chap. 3

  1. This chapter is concerned with the mechanisms of language change, in particular the mechanisms leading to grammaticalization.


    1. What motivates Grammaticalization
    2. what mechanisms lead to it
    3. what are the probable paths of progression
    4. what are the end results?

    Since particular language changes do not have to go to completion, we are not talking about absolute or obligatory determining factors, rather about factors that enable or facilitate Grammaticalization, either singly or in combination.

    Those mechanisms seem to involve:

    Later we discuss possible motivations of these mechanisms:

  2. Background assumptions about change.

    1. What do we mean by change? What is changing? Language does not exist without speakers, so it does not change without the complicity or participation of its speakers. Hoenigswald says "change is replacement" but perhaps best way to characterize it is as rule change.

    2. Different models of rule change exist, e.g. generative, where

      • rules change at a high level, or
      • as a whole set, not just minor adjustments.
      • Major changes occur in the discontinuity of transmission (generational), especially. during child language acquisition.
      • Assume a homogeneous speech community (hah!)
      • Speakers have a universal capacity for language and language processing;
      • Bring this to the output they receive from previous generations (b> Grammar1 , and
      • Infer an internalized grammar which differs from the one of the earlier speaker, giving Grammar2 (see Fig. 3.1, pg. 34) with an Output2 that differs from Output1

      This model okay as long as we don't assume

      • the rigidity or relatively fixed structure or
      • uniformity of input.

      This model (e.g. Anderson 1973) assumes universal laws, Universal Grammar consisting of unchanging principles etc. with parameters that allow for different "settings" so that changes from one generation to another result from diff. learners selecting different possibilities.

    3. For our purposes, we assume a universal component that does not completely determine the nature of linguistic structure
      • but instead has broader properties,

      • and can be modified by outside stimuli and
      • by functional purposes to which language is put.
      (Note the word functional which has been figuring in discussions of Newmeyer lately.) We wish to avoid the fixity of descriptions or conceptions of UG that act like rule change is complete and abrupt; in fact we see coexistence of both usages/forms etc. in most cases, with relic forms/usages persisting on for long periods, in the grammar of one and the same person. I.e, instead of A --> B, we are more likely to get:

          A >{     } > B
      where A and B coexist for a while before yielding to B.

      Need to distinguish between

      • change and
      • spread of the change, (called generalization )

      i.e. the rule may enter the language. abruptly, but its spread to other situations (other verbs of motion, other verbs of cognition, etc.) is gradual .

      Notice diff. between. spread across linguistic contexts (environments) whatever, and spread across social groups or genres (styles) which is more sociolinguistic.

      Notice also that change in the grammar of one speaker (Andersen) does not mean other speakers have changed, so when can we say that the language has changed? When has the grammar of the speech community changed? Often linguists document the beginning of a change, in the grammar of one speaker, but when can we say it is finished? It is too difficult to document all the tiny changes.

      H&T give ex. of perhaps earliest attested use of OE willan 'want' in a context that can only mean 'later time', i.e. in its modern usage as would/will in reported speech. (This is c. 880, Orosius, p. 37) Later this change is seen more often in ME; if such a 'change' does not show up again, we call it a nonce form or it is attributed to scribal error.

    4. Induction, deduction, abduction

      Difference between. computer languages. and human languages.: In computer languages., one form has one meaning, every 'utterance' is unambiguous. (If it's not, we have 'bugs' and systems crash.)

      Human language has small set of units and constructions that must serve many purposes and functions. Language. is a social thing, one function is to maintain social relationships, and communicate in ways that keep people involved and interested. (I.e. we don't talk like computers; if we do, we lose our audience, are thought of as 'geeky' or 'nerdy' or 'socially inept'.

      Indirectness (e.g.) is necessary for politeness phenomena. We can't (unless we are some kind of tyrant or socially-inept person) just say

      Open that window!

      when we want the window open; we tend to say something. like

      Could I trouble you to open the window a bit? I'm sort of uncomfortable.

      Other non-literal things like metaphor are part of language (Johnson & Lakoff 1980) and can't be done away with. People know how to reason from the form of what is said to the intent of what is said, so humans can interpret the two utterances above as the same in some way, and accept the second as 'more polite' than the first, and a more effective way of doing linguistic business.

    5. Deductive vs. Inductive reasoning can account for explicit and direct utterances, but can't account for

      • indirectness
      • expressiveness
      • or change

      For these, a type of reasoning known as abductive is necessary (Pierce 1931) (Cf. also Andersen 1973). These types of reasoning exemplified by three propositions of a syllogism:

      • The Law (e.g. all men are mortal)
      • The Case (e.g. Socrates is a man)
      • The Result (e.g. Socrates is mortal)

      Deductive reasoning applies a law to a case and predicts a result (as in above) Conclusion assumes/asserts nothing not given in premises, so if premises are true, conclusion is also true.

      Inductive reasoning: proceeds from observed cases and results to establish a law.

      Abductive reasoning proceeds from an observed result , invokes a law, and infers that something. must be the case. This is the kind of reasoning that occurs when speakers make assumptions based on outputs of other grammars and make inferences from them.

      Pierce said this was a weak form of reasoning (i.e. not the ideal) but was useful to study because it is the basis of human perception and the only kind of reasoning by which new ideas can get started. Andersen saw this as essential to development of cultural patterns, including language; and H&T see it as the mode of reasoning that leads to reanalysis.

      Consider the development of cargo cults. During WWII (and even before; the earliest cults seem to date to the period of 19th century colonialism), Americans and Australians built airstrips and landing fields in New Guinea, and then cargo planes dropped supplies, munitions etc. on those fields to be used in the war against Japan. After the war, anthropologists discovered that the people of New Guinea were building their own airstrips, hoping that supplies would then magically arrive. They used abductive reasoning:

      If you build it, they will come.
      The original reasoning, which was "We need to drop supplies for the war effort against Japan, therefore we will build airstrips" was turned around: if we build airstrips, cargo will arrive (after all, it happened for the Australians and Americans, didn't it?) Even though cargo droppings did not result, people continued to try to find the formula that would work, i.e. in ritual and primitive religion, rather than in logic.

    6. Langacker, discussing syntactic change, defined reanalysis as:

      "change in the structure of an expression or class of expressions that does not involve any immediate or intrinsic modification of its surface manifestation"

      Since this (1977) the idea has become common that reanalysis means shift from one parametric setting to another.

      • one simple kind: merger or fusion of two or more forms across word/morphological boundaries, such as the development of words like childhood, kingdom, manly by incorporating suffixes ( hood, -dom, ly ) that were once independent words meaning "state, realm, condition, body" etc. Affixes were once lexical items having very general meanings. In many languages., body parts are incorporated to function as emphatic markers. Then, these may become reflexive markers or pronominal forms.

      • Fusion involves changes in the assignment of boundaries, i.e. rebracketing such as Tamil {pookal]+[aam] being reinterpreted as [pooka]+[laam] , and eventually [poolaam]. But rebracketing is not always same as fusion . Ambiguity (opacity) allows structure to be analyzed as before, but new analysis may coexist with it. Otherwise, old forms would disappear.

        Reanalysis is the result of abduction (conclusion from the output, working backward to the Law or principle, restating the principle):

        [[back] of the barn] --> [back (behind) of [the barn]]

        The hearer matches output (result) with possible NP structures ("the case"), after matching with possible nominal structures ("the Laws"), and the output structure is compatible though not identical with previous analysis.,

      • Other languages: French future, derived from Latin habeo cantare " I have (these things) to sing" --> "I will sing these things". This form came to compete with cantabo and finally replace it. French output: chanter-ai the infinitive (as in Latin) plus the 1st singular. of avoir (have). H&T go through this in detail, showing how La. habere used to mean "possessing, belonging" but also a kind of general locative meaning "belong, be in possession of". (Parallel to Dravidian iru meaning "be located" in general, but also, with dative or locative, have as in en-kiTTe X iru-kku "I have X". (Kannada nan-hatra X ide "I have X; I am in possession of X"?)

      • Stage Three: with a gerundive, Habeo + Inf. meant some sort of obligation, or future orientation, which was transferred (or 'bled') from the gerundive to the habeo : aedem habuit tuendam 'he had a house to attend to (cf. English 'he had to look after the house! which implies future obligation.)
      • Stage 4:
        haec habeo cantare -->
        1 2 3
        haec cantare habe
        1 3 2

        New word order, and with new meaning OBLIG, or future-oriented. Cf. with English:

        'what can I say?' ('what do I have to say') --> 'what will I say?'

        By abduction: [[cantare] habeo] --> Late Latin [cantare habeo] --> French [chant-er-ai] "I will sing".

      • Another Example: S. Dravidian -ee "emphatic" in locative expressions:

        Literary Tamil
        Gloss Spoken Tamil
        viiTT-il in the house viiTTule
        meel on top meele
        kiiR below kiiRe
        anku there ange
        vaDakku (to the) north vaDakke
        meerku (to the) west meekke
        terku (to the) south tekke
        kiRakku (to the) east keRakke
        veLi outside veLiye

        When the LT and ST forms are compared, the ST forms appear with a final -e(e) which traditional grammarians analyze as the "emphatic particle" and simply state that ST locatives all have to appear with this final emphatic. In fact, however, if emphasis is needed, an additional e(e) has to be added: ange-y-ee "right there" etc. What has happened is that the emphatic e(e) has been incorporated into the locative, and has in some sense become the marker of locative, since it is the common feature of all the ST forms above, whether explicitly marked with le or not, as in "semantically locative" forms (most of the above: postpositions, points of the compass, etc.). Note also how the points of the compass are marked with a now redundant dative marker ( -ku ). The emphatic has now been fused with the locative, and bracketing between it and the locative is gone. The points of the compass are, in the case of 'east' and 'west' built upon and incorporate the locational postpositions meel 'up' and kiiR 'down' (since the geography of Tamilnadu is such that west is 'uphill' and east is 'downhill.') (Cf. perhaps Kannada forms? Cf. Gai etc.)

      • Another example of fusion: Kannada maaDu-v-ud-illa originally a negative verbal noun with future marking:. maaDu-v-udu 'making' from maaDu 'make' -vu 'future' + -du 'thing' + illa 'negative' meaning quot;not going to make or do". Through reduction this has become

        • (stage 1): reduction of the sequence -uvu- to -oo --> maaDood-illa and
        • (stage 2) further deletion/reduction to maaDolla 'not going to make/do' (immediate or habitual negative).

        This contrasts with maaDlilla 'didn't/doesn't make/do'. Reduction and deletion of the intervocalic lax consonants results in a short form without overt future marking or verbal noun marking, but the meaning is the same.

    7. English Modal Auxiliaries: Lexical verbs may, can, must, do become AUX.

      • In Late Middle E: question inversion and negation without do and Transitive clauses with verbs can or may + NP:

        • Know you X? He knows not X. etc.
        • "She could much about travel" (She knew a lot about travel)

      • LME also had past participles of modal verbs, e.g. "had mought to escape" (had been able to escape).

      • LME also allowed sequences of modal verbs:

        • "not to may will have" (to be able to want to have)
        • (Note relics of modal sequences in southern American, e.g. "might could.")

        • Next stage: early Modern English:

          • Above-mentioned possibilities are gone
          • Do-constructions replacing them.
          • may, can, shall, do are a new category: the MODAL
          • Inversion and Negation now have reference to this category.

          There is a new definition of the constituent AUX, including Tense and Modal and Modal is expanded as can, could, will, would, must, shall, should, may, might, do. Note that the past forms of modals (preterits) themselves take on special meanings, e.g. "conditional" and/or other meanings: should is not just the past of shall but has " obligative" meanings etc.

          H&T refer to this set of modals as 'closed' which it is to some extent; but it's not completely closed, since newer modals are being added: gonna, hafta, sposeta, gotta etc. Now for interrogation and negation, do and other modals are preposed/inverted, other verbs can't be (except for the quasi-modals they mention: need, dare, ought to, be to etc.) which have special problems:

          • Can sometimes invert, sometimes not; depends on the person?
          • Can/can't have do

          So we get examples such as:

          • You needn't go (instead of 'you don't need to go'?)
          • Need I say more? (Do I need to say more?) (pragmatic difference? politeness?)
          • *Need he go?
          • Need it be said?
          • You oughtn't to say that (*You don't ought to say that)
          • Ought you to do that? (*Do you ought to do that?)
          • You hadn't ought to do that; ?you had ought to do that.
          • Dare I suggest that the two parties have no case? (Do I dare suggest...? pragmatic/politeness?)
          • *Dare you go?
          • [arch? dialect?) he dasn't do it; ?he daren't do it

            Meanwhile, these new candidates for modal status ceased to have the freedom of taking NP objects, and have special morphologies ( may/might; can/could ) "preterit presents" (sic?) i.e. a cluster of factors is involved: H&T say this may be crucial and that we will find it often, or maybe always.

        • But: reanalysis and Grammaticalization are independent: Sometimes reanalyis leads to something other than new affixes:

          • With compounding we sometimes get a new lexical item (e.g. 'bo'sun' from 'boatswain', 'hussy' from 'huswif' ('housewife'), 'sheriff' from 'shire-reeve', etc.) These have not been reanalyzed as new gramm. morphemes, but as new lexical items: lexicalization .

          • Or: a grammatical item (e.g. prepositions) spawns a verb, as in English 'up, down, out' as verbs: 'up the ante,' 'he downed it in one gulp', to 'out' somebody (reveal their secret identity), etc.

          • Example of Estonian emphatic particle ep (from a bound clitic).

          • Word order changes: Cf. previous discussion of Meillet's, where he talks about the 'grammaticalization' of expressive (pragmatic) word order, and changes in word order bringing with them certain typological characteristics:

            • Verb-initial languages are prepositions;
            • the adjective follows the noun;
            • possessive, relative. clause follows the noun;
            • Aux precedes the verb;
            • yes-no Q particle is clause-initial.

              Verb final languages. (e.g. Dravidian) usually. have all these reversed, but not as consistent. But when inflections develop in OV languages., they do so via reanalysis of clitics, bound forms (with fusion), phonological. attrition of already bound forms.

              When new periphrastic constructions arise in shift from OV to VO (if this happens), usually. through reanalysis of lexical material . But no constraints depend solely on word order using lexical material: rather, the constraints are:

              • semantic suitability
              • inferences (conversational and logical) from context;
              • potential constructional ambiguities coming out of inferences. H&T give example of Latin habeo developing both into a future and perfect e.g. French chanter-ai 'I will sing' and j'ai chantł 'I have sung.'

    The path to future was via obligative and future-oriented sense while path to perfect via locative-possessive-existential (as in Dravidian iru becoming perfect.) More examples of English word order changes from VO to OV.

  3. Analogy and Rule Generalization

    Which thing is the attractor, and which the attracted: things get more general . Kiparsky: analogy is rule extension , i.e. the rule spreads to more and more general environments, less specific. Optimization. (Some relics may remain, e.g. English plurals: -s plural very general, but some relics: foot/feet etc.)

  4. Reanalysis, analogy, reanalysis: example of French negation.
  5. Conclusion:, last modified Jan. 31, 2000