THE THEORETICAL PRINCIPLE OF
HOW FAR ABSTRACTION SHOULD REACH AND WHERE IT 
SHOULD STOP

[Rev. Aug. 2, 1996]

© 1996 by Orchid Land Publications

     This is the subject of chap. 2 of Variation in the data: can linguistics ever become a science? (to which, incidentally, the answer is yes, assuming the abandonment of synchronic-idiolectal [minilectal] premises). Some linguists (notably the generative semanticists of yore) have been willing to carry linguistic analysis to a high level of abstraction; but virtually all wish to stop at the level of the idiolect. The DEVELOPMENTAL CRITERION (see Variation in the data, pp. 29-31) stipulates that only true universals (such as mini-length differences in phonetics and the like) should be excluded from analysis; the others, including lectal variation, should be annotated at some level of one's analysis. The principle implies that a polylectal analysis is the only kind that befits a polylectal language. It actually says that all information (other than well-established universals) needed for explaining and predicting new developments--the two jobs of a scientific theory--should be included in the analysis.

This is not as hard as it seems to those who never tried it, since

--Each successively more-monitored style (isolect)--on back to (something as close as the system permits to getting back to) the underlying representation itself is a subset of each adjacent more-monitored style. This is the basis of the pedagogical principle of PEELING THE ONION, which see.

--Implicational models handle this very easily. (Statistical approaches to variation are pseudo-minilectal, not polylectal, as the book just cited makes clear.) The books cited here and on the webpage to which this is attached offer examples of such models; the models prove adequate for dealing with the most complicated problems in variation across the English language and ancient Greek.

--Of the thousands of styles (isolects) that one masters in adolescence (after which grammar-learning diminishes as socialization and polystylism come more into focus) as well as the yet more thousands of isolects that one understands in one's language, most overlap. To show what this means, consider that the informal style of an elderly male is often the same (for some rule) as the next-more-formal (i.e. more-monitored) style of an elderly female of the same class or the next-less-formal (i.e. less-monitored) style of a significantly younger male of the same class, education, regional provenance, etc. Each successive class barrier, regional barrier, age barrier, sex barrier, tempo/style barrier, etc., removes or adds one overlap. See details in Bailey, Variation and linguistic theory (1973), p. 107.

     It is important to note that variation theory does not speak of what pragmatics deals with--the matter of one's competence for choosing the right style for the right situation! One must not confuse "performance" with any of the thousands of (what for the minilectalist are) grammars in one's head: For the minilectalist, each style is a discrete "competence" and a discrete grammar, analysed without regard to its fellow styles and hence often in conflict with one's analysis of them. (Moreover, minilectalists do not tell us which grammar to begin with [they usually use a fairly monitored style], though it should be the most unmonitored style--the superset that includes the others. Again, this is the basis of the principle of PEELING THE ONION.) Of course, minilectal premises don't get one very far toward a scientific linguistics; and the same is of course true of using for analysis monitored styles in the prevalent manner.

     Abstraction has got to begin at some level above raw data, since no two utterances are ever exactly alike. The phonetic symbol [b] is an abstraction over all of the real b's ever uttered. No satisfactory reason has ever been given, so far as the writer is aware, of why abstraction should end at the style-free, variationless idiolect. It therefore needs, according to the principle under scrutiny, to embrace an entire language system. Since this has been done at the phonetological and phonomorphological levels of English, it can be done; the variations of the English sound system are sufficiently complex to warrant the conclusion that if they can be handled polylectally, a similar undertaking with respect to the much more nearly uniform English morphology and syntax is a fortiori a doable undertaking. To carry out the theoretical jobs of explaining and predicting English, e.g., one needs to consider the fact that a native-speaker easily shifts among thousands of styles (different subsystems of rules)--many tempo-dependent and many dependent on the degree of self-monitoring--in one's own speech-production--and the even greater number of such isolects that one understands and deals with "competently." Just as a black-and-white television set cannot receive color, though a color television can receive black-and-white broadcasts, so polylectal models can handle minilectal situations, whereas the converse is unthinkable.

     What warrant is there for treating the styles (each a separate idiolectal "competence" for those who accept this latter notion) of Josephine Speakerine's English the way one would view the difference between Chinese and Arabic? If each minilect is analysed without regard to the others in the same overall system in the speaker-listener's head--and what's there includes all of the kinds of English that one deals "competently" with--the result will be inconsistent. This is shown not only by quarrels over the use of historical information where alternative "synchronic" possibilities present themselves as well as evident reanalysis, back-formations, retrograde developments, etc.--say, in dealing with the English vowel-shift. Inconsistencies are also found in analyses of closely related varieties that result from the minilectal premise, and indeed when the inconsistent analyses are performed by the same person or persons subscribing to the same analytical approach. (Some linguists even quarrel over whether the vowel-shift is phonology or phonomorphology, which it obviously is!). The inconsistencies conflict with the fact that a single speaker commands all of one's myriad styles and shifts easily among them. (See also the mini-essay on pre-emptive neutralization.)

     Page 31 of Variation in the data advises speaking of production errors for coughing, running out of breath--whatever. Even slips of the tongue come tangentially into competence (if one accepts that minilectal concept), since they reveal various unmarkings. (To avoid various kinds of naivete or misconceptions concerning markedness and its relation to change, see Bailey Essays on time-based linguistic analysis, esp. chap. 10.)

     Many paradoxes of minilectal analysis tend to disappear in polylectal analysis. One that often does so is the analysis of neutralization--a grave difficulty for phonemics--the inadequacies of whose models (whether structuralist or generativist) for even minilectal analysis have been known long since on other grounds. (Too bad they are still employed by the glottometrist approach to variation--which compounds the other failings by paradoxically using monolectal models for what purport to be polylectal analyses.)

     See further the HOLISTIC PRINCIPLE on p. 67 of Variation in the data. It forbids analysing one component is such a way as to make hash of some other component of the grammar. (This is not the same as what some advocate--viz. "componentless" grammars.) The principle is also related to the principle of testing theories and models on the most complex data--i.e. not being simplistically satisfied with adequacy for minidata from one (usually monitored) minilect. The last point brings us to the OBSERVER'S PARADOX, dealt with in the books mentioned previously.

---

Search this site    powered by FreeFind

Hits on this website since 11-22-98