THOUGHTS ON HOW TO ANALYSE
PRE-CATALOGUED AND
PRE-CLASSIFIED 
LANGUAGE DATA

© 2000 by Orchid Land Publications

C.-J. N. Bailey

[updated 2-6-01]

     To find ORDER in one's data requires some model that gives order and arrangement to one's pre-catalogued and pre-classified data--the point at which theoretical (explanatory and predictive) analysis begins.   Analysis done in the framework set forth in Essays on time-based linguistic analysis (Oxford University Press, 1996) requires the use of markedness theory (one based on naturalness, not on normality--a statistical concept) and of course reversals of values in marked contexts.  One first catalogues the data and then organizes it am appropriate classificatory manner. For example, one could (i) set up a table with the sound or tonal changes for example, in which each column represented a defined  context in terms of a given language system.   For tones, this would of course require universal terms like MF (mid-falling), LR (low rising), HL (high level)--and the more complicated ones like Mandarin ML-L-R.  in, say, column k--a system with four tones of a given type, you may get MF-->L--but in another column q--a system with three tones or four tones different from those of Chinese you may find L-->MR.  One would need the whole picture so that anyone could see in an instant what is what. This could best be an Appendix, but readers should be told at the outset that it is there, so that any reader could constantly refer to during all that s/he reads.  This might be called the reader's ROADMAP or some such. 

CLICK HERE FOR SIMPLE PRESENTATION
OF VARIATION THEORY 

     If the question is about what is borrowed and what represents a natural change, one should not mix oranges and apples.  But if one arrives at implicational patterns, as one does, why should they be distinguished?   Not just anything is borrowable.  It can add to implicational patterns, but can borrowing upset them?  Anyhow, what difference does it make to the pattern whether it a part of it came in by borrowing or by natural development?  But we must distinguish productive from non-productive nonce elements in a pattern or system.
      There are two things to remember:  
(i) Statistics show what is normal; markedness is what characterizes what is natural; the latter is what matters.   
(ii) if the analysis is to go beyond a list or inventory, you need to do the two things, given what  theory "is"--prediction and explanation.  One may come before the other, but there are cases of unexplained predictables and even unpredictable explanations.  One needs to propose theoretical hypotheses (e.g. a mid tone will change to x in a y-type system under z conditions), clearly in a box, and propose it as a tentative theoretical universal.  One then goes looking for  possible counterevidence as well as supporting evidence.  
     Since assimilations are reversals of expectations, those reversed changes should not be mixed in with the others; that would be mixing apples and oranges.  Changes that would violate segmental clusters or suprasegmental sequences would presumably not be borrowable as such.   One needs to check.  As for assimilatory reversals consider these example:  
(i)  Voiceless or heavy stops are the unmarked ones.  But then, how is one to explain more d's than t's between vowels or only z (not s) directly before a nasal?   In English, all of this is affected by internal word boundaries and most especially by syllabic boundaries.  Other languages ignore all of that.
(ii)  If s is natural syllable-initially and sh is natural syllable-finally (don't go by orthography; final sigma in Greek was very likely "sh"--as in South German).  Given that sh can become s naturally by losing a distinction not present in the rest of a language system, whereas s becomes sh syllable-initially only in a palatalizing environment (say, directly preceding "ch" or "y"), we can speak of languages that have only s or only sh, but they have very different explanations (and predictability).  I would not expect to find a language whose only syllable-initial sibilant (or even fricative) was sh.  If ever found, it would require a lot of explanation.  Once you catalogue or inventory the changes, you can proceed to analyse.  
     You will then be able to deduce which of opposed changes is the basic natural change and which is the reversed one.  It won't tell you which is borrowed; that information would have to come from the real world outside of the grammar.   But one could make some good guesses--hypotheses to be investigated.  A change that violates the language pattern would, I predict, usually be unproductive--a nonce thing that stood out as an exception.  Those who don't say preventive like inventive--they say preventative--won't be able to go far with such malformations, say by coming up with inventative.  Preventative is not borrowed but is a "higher-level" analogy that overrides the natural patterns.  What is not known is under what circumstances a borrowing can violate a natural pattern.  Middle English blended French agentive -eur with an Anglo-Saxon form (both from original -or, with long-o) to come up with agentive-instrumental -er as in painter and helper.
      Finding tempo or style differences is an excellent way to ascertain markedness, since a change found when a speaker moves from more formal to more colloquial will be an unmarking (even if a reversed-unmarking).  I don't see how you can proceed without markedness theory and implicational patterning--the model needed to discover markedness.  The early generative linguists showed (and produced many analyses of)  instances in which used to cannot be pronounced as useta, have/has to cannot be pronounced as hafta/hasta, and want to cannot be assimilated to wanna; be gonna is not used the way literally be going to is used.  These are interesting in the way the ordering of syntactic and phonetological materials interact.  But a daughter language of English could borrow unanalysed gonna or yoos as a simple function work in its own right.  Hawai'ian Creole has wen + VERB for anterior forms and go + VERB for posteriority.  Ste ("stay" calquing Portuguese sta) + VERB (more recently VERB-ing) creates a progressive form; stay V(-ing) means "we're singing." 

     The initial catalogue is becomes a sort of potential ROADMAP after one has classified the data at the lowest level of order--the very lowest order being alphabetical.  The THEORETICAL MODEL one uses then offers ways to bring the higher order of explanation and prediction into the mass of data.  One may be able to construct hypermodels for, say, a typological group.   (Cf. the work of Kartunnen or J. Edmondson's work on SE-Asian tonal systems, exclusive of Korean and Japanese--which have a different pattern.)  Generalities about informal vs. formal speech will emerge.        

     To make what I'm thinking more concrete, here's an example that comes to mind:

     The uvular sound of Arabic was borrowed into Spanish to replace the "j," the fronted variant of //g//--both of whose variants were fricatives.  (Spain was under the Umayyad dynasty till the Moorish Muslims conquered it; but the latter spoke Arabic too.)   It's passing strange that the fronted variant of //g// was made more retracted than the system had previously allowed!  During the centuries of Spanish rule in Holland, Dutch (which had, like low German in general replaced [g] with the fronted fricative "j") borrowed from Spanish the uvular trill for "g."   One will naturally wonder:  Could what the Spaniards call the "Arabic sound" be borrowed into a language to replace a stopped "g"?   I would set up a hypothesis to be examined:              

     A voiced uvular trill (Arabic has several throat sounds; some Arabs have the uvular "r" as well as the fricative "th" and other sounds like southern Spanish) got into French or Portuguese (the Arabs controlled some of the Mediterranean coast of France for awhile; and of course they had controlled all of the Iberian peninsula at one time.  It seems to have been a second, voiced "Arabic sound"--the one that got into French or Portugese to replace apical "r."  Whether French got it from Portuguese or conversely, I do not know; it doesn't exist in colloquial southern French (Langue d'Oc)--which is a problem:  Why should it occur in northern, Frankish (Germanic) French (Île de France)?    I doubt that uvular "r" developed naturally in any of the Western European languages other than, possibly, in some kinds of British Keltic, where lenition was even more thorough-going than in Greek and Spanish.  Could it have developed in Keltic-speaking Brittanyand then spread to French and Portuguese, being helped in the latter by the Arab influence?    
     One might set up a hypothesis:   uvular sounds can replace //g// or //r// but not both in a given language. This would be false, since I'm told that Belgian Dutch (Flemish) has borrowed the uvular "r" from the co-national French.   
    It is worth stressing, as my work on the birth of Middle English has shown, that:  
(i)  The prestige language controls the structure.  One must avoid getting confused by the  way words from the language on the ground calque functor and other words of the prestige system; the pseudolineage of English is based on calques--not on language systems. 
(ii) The language on the ground wins out in the pronunciation.
I can give dozens of examples in Europe and daughter languages of English and Portuguese and French.   

       Interestingly for present purposes, further changes have affected Portuguese in Brazil.  There are four ways to pronounce forte--with or without loss of final vowel (pronounced "i") combined with affrication or not of //t//--which are combined with the two ways of pronouncing "r," apical and uvular.  A preliminary some years ago suggested that that not all possible combinations of the three variants is possible; one of the "r" sounds implies, or is implied by, one of the other factors or its absence, i.e. by given combination of the two factors.  This is something that developmental linguists would find worth going into.
     More interesting is the similarity between Spanish and modern Greek.  Greek was a part of the Byzantine Empire before the Umayyads conquered Iberia. The two languages have a like rhythm, accent (not stress--there is no lengthening--but tone), vocalism, and, notably, their similar consonantism--the fricatives, voiced and not voiced, and the lisped (retracted-tongue-tip) sibilants.  Oddly, Spanish has got only voiceless sibilants.   Why?  Since "ss"  and "x" in Hellenistic Greek represented the "sh" sound, it's no surprise that Portuguese uses "x" for that sound.  But Spanish, like French, uses the letter "j" for examples in which Portuguese has got "x"--including sounds derived from Latin consonantal-i (English "y") and of course the palatalized lateral that became the same.  This "j" in Spanish was once like English "j" when it came from a y-sound (cf. Spanish in Seville, in Argentinia, etc.)  or a palatalized ''g''  (the palatalized "d" has a more complex Romance history).  As in French, it has become "zh" in  Portuguese.  Why did it get devoiced in Spanish as a "sh" sound and then get replaced with the Arabic sound?  Or is that the correct temporal sequence?

     Usually rules become more general--with loss of restricting features.  But sometimes they shrink.  The rule dropping //h// before an unstressed vowel ('im for him, vehicle, vehement, etc.) has shrunk in that it no longer applies to the beginning of an orthotone (non-clitic) word.  What used to be "a history : an historical" is now "a history : a historical."  

[PAGE UNDER CONSTRUXION]