In work of Humboldt (1822), e.g., there was a concern for the evolutionary nature of language, from isolating to agglutinative to synthetic, which was supposedly the highest stage. This was part of the ideas about typology at the time. Later concern for historical linguistics took over and notions about the "evolution" of grammar became swept up in the Neogrammarian tradition, where everything had to be "exceptionless" so no concern any more for the evolution.
Gabelentz also concerned with grammaticalization (1891); he talked about how forms begin vigorously (like burocrats) then fade, grow pale, become bleached out, need new paint; some become mummified. He suggested there were 2 competing tendencies:
Gabelentz also noted that the process is not linear, but cyclical with constant reiteration of the process; the conditions are always present in the language, i.e. it doesn't reach a final, ideal stage. Or, it's not so much cylical (bec. things would come back to the same place) but a spiral.
Meillet then took up the notion, used the term grammaticalization, and recognized it as a central area in the theory of language change (esp. morphological change.) But anchored in a positivistic idea of "being able to know" something for sure:
Meillet said there were two ways that new grammatical forms emerge:
Quotation pg. 22Thus the example of be-going-to ---> gonna where all morpheme boundaries are erased, and phonological reduction occurs. The phrasal collocation may take over from a reduced form that has become "commonplace" and as sort of "lost its zip" (become banal.) Later Meillet also talks about the possibility of grammaticalization being extended to syntax, since syntactic patterns also have meaning and gives example from Latin, where various kinds of word order became SVO, and it is clear only from the order which is the subject and which the object. (dog bites man/man bites dog etc.)
This word-order business shows two of the hallmarks of grammaticalization :
thelô ina --> thelô na --> thena -- tha "I wish that"Semantic development is from 'wish, desire' to 'future.' The phonolog. weakening is there, is the semantic change a "weakening"? Meillet tends to stress 'deficits' i.e. things getting worse (loss, weakening, attrition). Is this part of the classical attitude that lg. change is deterioration? (esp. w. classical lgs.)
But there are limits to arbitrariness, i.e. numerical systems could theoretically consist of a different 'word' for every number in the lg., but in fact they don't: they use a small, finite no. of morphemes and recombine them into an infinite number of combinations. (The morphs for 1-9, 10, 'teens', multiples of 10 (20-90), hundred, thousand, etc. are all that is needed.) Other examples of derivation, reusing similar affixes etc., or case-systems, tense-systems which theoretically could use a different morpheme for every word, but don't.
imagic iconicity: systematic resemblance between item and its referent by some (visual? pictorial? audial?) characteristic. and
|kuruDan avane paattu siriccaan||original meaning:||'having seen him, the blind man laughed'|
|blind-man him hvg.seen laughed-he||reanalyze as:||blind-man him - AT - laughed-he|
|'The blind man looked at him and laughed'||New meaning:||'The blind man laughed at him'|
In other words, the past-participle of the verb paar "see" can be reanalyzed as meaning "direct the attention at" or just 'at' since with a subject like 'blind-man' there can be no interpretation of 'seeing.' This can be seen with many other postpositions in Dravidian which have started out as verbs. (See more examples from Tamil here, esp. postpositions taking the accusative case.