LINGUIST List 4.888

Thu 28 Oct 1993

Disc: Infixes

Editor for this issue: <>

Directory

Message 1: 4.873 Re: Origin of Infixes

Date: Wed, 27 Oct 93 07:27:35 HS4.873 Re: Origin of Infixes
From: <stampeuhunix.uhcc.Hawaii.Edu>
Subject: 4.873 Re: Origin of Infixes

Content-Length: 6236

The Russel Ultan paper "Infixes and their origins" (In: Linguistic Workshop, ed. Hansjakob Seiler, Munich: Wilhelm Fink, 1975), cited by Alexis Manaster-Ramer (LINGUIST 4.872), isn't available here, and I'm having a hard time imagining how infixations can be entirely explained as due to metathesis or reanalysis. Maybe someone can post a summary.

The best-known examples of infixation don't involve metathesis or reanalysis at all. In the common pattern of infixation in "Austric" (Austroasiatic and Austronesian) languages, certain vc affixes go before the first vowel of a word, i.e. they are prefixes of V-initial words [vc[V...]] but infixes of C-initial words [C[vc]V...]. But not all vc prefixes are infixed in CV forms. For example, in Sora, an Austroasiatic (Munda) language of Orissa, India, verbs like id `scratch' and pO (O=open-o) `pierce' form Ar-id (A=wedge) `scratching tool' and p[Ar]O `piercing tool'. But beside Ar-id `scratch each other' we have not *p{Ar]O but Ar-pO `pierce each other'. That is, Ar `instrument' is infixed in consonant-initial words, but Ar `reciprocal' is just prefixed.

Even if all vc prefixes were infixed in CV forms it is difficult to see how to relate this to metathesis, which is normally understood as a reversal of phonological segments. But there is no evidence that unprefixed forms of the shape V1C1C2V2... ever became C2V1C1V2... in either Austroasiatic or Austronesian. Even if they had, it would not have been a reversal of segments, but of a syllable and a segment.

In our article "Rhythm and the holistic organization of language structure" (In: Parasession on the Interplay of Phonology, Morphology, and Syntax, ed. John Richardson et al., Chicago: Chicago Linguistic Society, 1983), Patricia Donegan and I briefly mentioned infixation as a characteristic of proto-Austroasiatic (the ancestor of Munda and Mon-Khmer), and suggested that its purpose was to avoid [vc[CV...]] structures which, by forming a closed syllable, would attract accent to the affix from its canonic and iconic position on the lexical root CV.... This explains why cv-prefixes are NOT infixed: they don't close a syllable and attract the accent: [cv[CV...].

[We did not explain why some vc-prefixes are not infixed, but I'll give that a try. Our theory implies that a vc-prefix whose meaning is such that it OUGHT to be accented should not be infixed. In Sora the vc-prefixes that are infixed are those that nominalize verbs: p[Ar]O `piercing tool', p[An]O `a puncture'. In themselves they contribute no meaning. The ones that are not infixed, however, do contribute meaning -- Ar-pO `pierce each other', Ad-pO `not pierce' -- and it is therefore fitting that they should attract the accent, much as the syntactic complement of a verb can attract the accent. In fact, the negative prefix Ad-, when it is prefixed to a vowel-initial verb, is unique in taking the form Adn- (e.g. Adn-id `not scratch'), as if to force a syllable closure and keep the accent, which in Sora goes on the syllable bearing the second mora in the word.]

This account is based on the well known association of accent with lexical as opposed to merely grammatical content. The mechanism is phonological, or more precisely prosodic. It's hard to see why infixations like these, which are quite typical of Austroasiatic and Austronesian, require any appeal either to metathesis or reanalysis.

(The following material may not be suitable for children.)

In his article "Where you can shove infixes", the infamous Quang Phuc Dong, writing under the pen name James D. McCawley (In: Syllables and Segments, ed. Alan Bell & Joan Bybee Hooper, Amsterdam: North-Holland, 1978), noted that -fuckin'- and other infixable epithets optimally go between a light and a heavy beat, as in fa3nta1stic (where 1=primary accent, 2=secondary accent, etc.): fa2n-fu3ckin-ta1stic. Now, this is just the same rhythmic pattern as in Adjective Noun constructions like du2mb yo1kel. Epithets are "infixed" even here: du2mb fu3ckin yo1kel. Like non-lexical elements such as clitics and affixes, epithets are backgrounded by being placed, as Wackernagel put it, in the accentual shadow of lexical elements. There is already a place in a 2 1 beat pattern for a minimally (3) accented element, namely on the 3-rest that comes between them, exactly as in common time music (1 3 2 3 1). If infixing epithets in phrases is natural, for accentual reasons, then it is a natural extension to infix them also in words that have multiple beats. After all, because of the association of accent with lexical meaning, we often treat such words as compound: alcoholic (workaholic), hamburger (veggieburger), helicopter (helipad, jetcopter), and so on for hundreds of examples.

(The following material may not be suitable for grownups.)

I suppose there are other kinds of infixation which can't be explained as a way of keeping meaningless elements away from the accent. Dwight Bolinger knew more accent and meaning than anyone, and if Dwight didn't write about it, then maybe we are barking up the wrong tree trying to explain infixation this way. But there is another supporting argument, based on a peculiar but nearly universal variety of infixation. If meaning is associated with accent, then the best way to obscure our meaning ought to be to remove the accent from meaningful to meaningless elements. This is how kids use infixing in "secret languages", in which the accent is displaced by infixing a meaningless but ACCENTED element before each vowel, sOBo nOBobOBodOBy cOBan OBundOBerstOBand thOBem OBat OBall.

David Stampe