faculteit der letteren rijksuniversiteit groningen lot winterschool 2005 historical syntax jack...
TRANSCRIPT
Faculteit der Letteren Rijksuniversiteit GroningenLOT WINTERSCHOOL 2005
HISTORICAL SYNTAX
Jack Hoeksema LOT, BCN
University of Groningen
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Focus of the course:
- corpus-based studies
- usage factors vs grammar
- emphasis on modern Dutch (1600-2004)
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday January 10, 2005
Corpus data represent usage, not grammar
Usage data indirectly reflect social and grammatical factors
Social factors have to do with the population of language users and the function of the texts
Grammatical factors have to do with the individual competence of the users
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday January 10, 2005
Strictly social aspects of corpus data
upper class language overrepresented in historical corpora (literacy)
more texts from men than from women
more texts from later than from earlier periods
more from written registers than from spoken registers
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday January 10, 2005
Partly social aspects of corpus dataSuppose texts from some period have 70% variant X
and 30% variant Y• 70% of the population uses only X and 30% only
Y (categorical usage at the level of individuals, variation at the level of the speech community)
• all speakers use X 70% of the time, and Y 30% of the time (variation at the individual level reflecting variation at the group level)
• some more complication relation between individual and group usage
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday January 10, 2005
Variation is relevant at the level of individual behavior and not just at the group level (cf. Guy, Gregory R. 1980. ‘Variation in the group and the individual: The case of final stop deletion.’ In William Labov, ed., Locating language in time and space. New York: Academic Press: 1-36. Contra D. Bickerton)
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Essential variation: variation which remains variation even after all factors influencing it have been controlled for
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Nonessential variation:
example 1
Population: 50% use of A, 50% use of B. All women use A, all men use B. Categorical for each gender.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Nonessential variation:
example 2
Population: 34% use of A, 68% of B. Before vowels, 100% use of A, before consonants 100% use of B.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Essential variation:
Population: 34% use of A, 68% of B. Before vowels, 91% use of A, before consonants 94% use of B. Strong effect of phonological environment, but not a categorical distribution.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
So one goal of variation studies is to determine the controlling factors, and to find out whether these completely determine the distribution or not.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Controlling factors can be internal or external
Internal: vowel vs consonant
main clause vs dependent clause, verb versus noun
External: sex, age, social status, peer group, time
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Statistical distributions are learned by children
Cf. Labov 1989 on “g-dropping”,
the ing-in alternation in English
‘The child as linguistic historian.’ Language Variation and Change 1:85-97.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
I’m working – I’m workin
in < Old English –inde
ing < OE -ing
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
In speech, in is most common in progressives and present participles, less in adjectives, even less in gerunds and least of all in nouns such as ceiling, morning
This distribution is found in the USA, the UK, Canada and Australia
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
The distribution is a clear reflection of the historical origin. The fact that it has survived several centuries, means that is is learned from usage by successive generations
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
01020304050607080
progressive complement presentparticiple
noun
parents child (7)
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Stylistic effect
in is informal, ing is formal
Hence systematically more ing in formal contexts, and more in in informal contexts such as vivid narrative
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
this stylistic factor is found both in the speech of adults and that of children, suggesting that children not only learn statistical distributions, but also associate them with stylistic levels
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Speed of change
Many grammatical changes appear to be glacial from the usage data. E.g. the loss of case marking in Dutch appears to have taken several centuries.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
The same is true, for instance, of the loss of clitic negation in Dutch: 14th century Hollandic Dutch already allows for dropping ne/en clitics, but 18th century Dutch still shows some (by then rare) examples of clitic negation. So the time course spans 4 centuries.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
This could mean a long period of variation, in which the disappearing variant slowly decreases in frequency – or: an abrupt change in the system slowly propagating through the population. Is change gradual, or catastrophic?
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Generative grammar tends to favor catastrophic change, since it usually tries to ignore, or abstract away from, variation.
E.g. the work of David Lightfoot.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Historical linguists of other persuasions, e.g. functionalists, usage-based grammarians, students of grammaticalization, etc. tend to favor gradual change.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Is change gradual if individuals show variation in their usage, within a large time-frame?
E.g. period 1: A uses X 90% of the time, and Y 10% of the time
Period 2: B uses X 50% of the time, and Y 50%.
Period 3: C uses X 10% of the time and Y 90% of the time.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Not necessarily...
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Yet another option: grammar competition at the level of individual speakers – diglossia. Cf. Anthony S. Kroch, 1994,
“Morphosyntactic Variation.” In Beals et al., (eds.), Proceedings of the Thirtieth Annual Meeting of the Chicago Linguistic Society,
vol. 2, pp. 180-201.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Kroch: during transitional periods, languages may show syntactic variation of a type that stable systems do not allow. This is due to the coexistence of two grammars, usually as a result of language contact.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Change in this model is both abrupt and gradual. Abrupt is the switch between two systems, gradual is the process of competition, by which one system ultimately replaces another
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Example: OV vs VO in Middle English. VO more common in the North: Viking influence.
After a stable period of OV (Old English), there is a transitional period of variation, followed by another stable period of VO (Modern English)
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
OV correllates with Particle < Verb order, VO with Verb < Particle order.
If two grammars compete, we expect to see the position of the object to parallel that of the particle
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Alternatively, if we assume an optional rule of object movement and an optional rule of particle movement, there is no a priori reason why utterances should show a correlation between the two orders.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Study of word order variation has found significant correlations between object < verb and particle < verb order
Cf. S. Pintzuk, 1991, Phrase Structures in Competition. Variation and Change in Old English Word Order. Diss. University of Pennsylvania
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
The choice between the grammar competion model and language-internal variation is a highly theoretical one, and can only be settled in the context of well-defined grammatical frameworks.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
S-shaped change:
curves representing change in usage tend to have a so-called S-shape. Change slow at first, then quick, and slow again at the end.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
010
2030
405060
7080
90100
1 2 3 4 5 6 7
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
S-shaped curves can be modelled (by curve-fitting) to various mathematical functions, including the logistic function
ln(p/(1-p)) = k + st
where t is time, p is the probability of the advancing form, k is the intercept and s a constant representing the slope of the curve (see Kroch 1989)
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Constant Rate Hypothesis
(Kroch 1989):
changes spread at the same rate in all environments
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
That is: the intercepts of two curves may differ, but the slopes are the same, assuming the two curves represent two conditions of a single change
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
010
2030
4050
6070
8090
100
1 2 3 4 5 6 7 8
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Primary motivation for the CRH: DO-support in English
Slope parameters for 5 environments:Negative declaratives 3.74Negative questions 3.45Transitive Yes/No questions 3.62Intransitive Yes/No questions 3.77Affirmative WH-object questions 4.01
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
A Dutch example: lexical replacement of ogenblik by moment
3 environments:
op het ogenblik/moment: series 1
elk/ieder ogenblik/moment: series 2
geen ogenblik/moment: series 3
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
0102030405060708090
100
1 2 3 4 5 6 7 8 9 10 11 12
Series1 Series2 Series3
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
The constant rate hypothesis holds very neatly for the two related minimizer uses: geen moment and ieder moment
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Compare
The train may arrive any moment/second/minute now.
#The train may arrive any day now.
##The train may arrive any year now.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Universally quantified minimizing nouns have a special immediate future readingCf. Jack Hoeksema, 1997,
"Ieder moment: Scalaire universele
kwantificatie", Tabu 27-4, 161-170.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
But op het ogenblik/op het moment shows a much slower rate of change.
Possible explanation: op het ogenblik / op het moment is a fixed expression with idiomatic properties
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Op het moment ben ik vrolijk
at the moment am I cheerful
#Op de dag ben ik vrolijk
Op die dag ben ik vrolijk
Op de dag dat ik vertrek ben ik vrolijk
Op deze dag ben ik vrolijk
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Deictic readings:
op het moment ben ik vrolijk
= nu ben ik vrolijk
hij had op het moment geen geld = hij had toen geen geld
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Other expressions
op dat moment /op dit moment
op een gegeven moment
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
0
20
40
60
80
100
1 2 3 4
Series1
Series2
Series3
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
So...
constant rate effect holds for all cases except op het ogenblik
All environment go from < 5% moment to > 95% with a century, but op het ogenblik has only reached about 40%
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Caveat: this is hard to check with Google, or text cd-roms, because the figures are only valid for expressions without modifiers
Cf. Literom hits (2002)op het moment2323op het ogenblik 1797
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
op het moment op het ogenblik
2323 1591
minus: op het moment/ogenblik dat
526 1240
minus: op het moment van
296 1176
minus: op het moment waarop
228 1163
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Conclusion:
op het moment/ogenblik van de ramp
op het moment dat zij aankwam
etc are not idiomatic, but compositional
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Question: is the Constant Rate Hypothesis validated or invalidated?
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Neither
The evidence from the Constant Rate Hypothesis so far is rather slender (mainly the 3 case studies in Kroch 1989, and some later studies by Kroch and some of his students)
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Usefulness of mathematical modelling of usage curves:
• may help estimate missing data points
• helps the analyst to decide if s/he has enough data
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Caveat
The logistic function tails off asymptotically at the beginning and at the end. This is not realistic for changes with a definite beginning and a definite end.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
Note that the logistic function yields a symmetrical S-shape, so in principle, the time course of a change is predictable from only the first, or the second half. This makes it possible to make predictions for changes that are half-way to completion, or to reconstruct part of a change if there is a gap in the historical record.
Faculteit der Letteren Rijksuniversiteit GroningenLecture 1, Monday Jan 10, 2005
It is not known, whether this is always a realistic assumption. If the S-shape represents a change spreading through a population, we might expect its second half to be faster due to the effect of mass media (for recent changes), or slower, due to the effect of schooling and standardization.