Source
Automatically imported from: http://commons.somewhere.com:80/rre/1995/way.html
Content
This web service brought to you by Somewhere.Com, LLC.
"way"
``` [This is an abstract for a talk at UCSD. I found the abstract interesting and thought you might too. The author's e-mail address is misrael@ucsd.edu.]
Date: Mon, 30 Jan 95 08:21:34 PST From: hare@crl.ucsd.edu (Mary Hare) Subject: Abstract for Tuesday's talk
Here is the abstract for Michael Israel's presentation at the PDPNLP meeting tomorrow.
Time: 4:00 pm Tuesday 1/31 Place: CSB 280
---
This Tuesday, I will be presenting work on the historical development of the English "way"-construction. The modern construction, illustrated in the examples in (1), has been described by Jackendoff as a productive "constructional idiom" which pairs a specialized syntactic form with an idiomatic meaning. The construction is unusual because, among other things, its argument structure seems to be projected not from the lexical verb but rather from the direct object "way". The specific goal of my talk, then, will be to explain how a construction of this sort managed to find its way into the grammar of English.
1. a. Rasselas dug his way out of the Happy Valley. b. The wounded soldiers limped their way across the field. c. Convulsed with laughter, she giggled her way up the stairs.
More generally, I will be concerned with two major theoretical issues. First, what sort of mechanisms or principles do we need to account for language change of this sort? And second, what can such historical phenomena tell us about the cognitive representation of linguistic forms?
I will start by briefly reviewing two traditional notions from historical linguistics, reanalsysis and analogy. Traditionally, reanalysis is understood as an abrupt, structural change in which speakers reinterpret an ambiguous string by assigning it a novel structural description. Analogy, on the other hand, is understood, as a gradual process whereby an already established productive pattern is slowly extended to an increasingly broad class of forms. On the standard account, reanalysis is where change happens, and the role of analogy is simply to help spread the change through the language.
A careful look at the history of the "way"-construction suggests that things are a bit more complex. I will show that the construction started out, sometime in the 14th century, with more or less unexceptional uses of "way" with verbs of motion (2a) and later with verbs of path creation (2b).
2. a. He sets a slender calke, And so he rides his way. (1587). b. As a ship, that through the Ocean wyde....doth make her way. (1594).
In these uses, way designates a literal path traversed or created by the subject. Until the 16th century, such forms were rare, occuring with only a limited range of verbs. The range expanded gradually, as new verbs tended to enter the construction only by close analogy with well-established exemplars: thus, early establishment of "cut" in the construction led to extensions with related verbs like "furrow", "hew", "mow", and "mine". The historical record reveals a number of distinct analogical chains of this sort, in which frequent use with a basic level verb led to extensions throughout the verb's semantic field. This slow analogical process of evolution constitutes strong evidence for the claim that specific learned instances and not just general rules form an integral part of the grammar (Bybee 1988, Langacker 1988): while high-level constructional schemata may have an important organizational function, actual usage is driven more by the availability of specific licensing exemplars. The construction itself thus appears as an emergent property of the instances it licenses: frequent exemplars form the basis for small analogical extensions, and a mushrooming range of exemplars leads to the emergence of increasingly abstract constructional schemata.
Morevoer, the facts suggest that very low-level analogies may be more important in driving historical change than the more abstract process of reanalysis. The implication is that speakers must somehow keep track of both what forms occur most commonly in the construction and of how similar these various forms are to one another. One sort of system that does just this is a neural net. If we think of established usages of the construction at any given time as clustering in some portion of a multi-dimensional representational space, we can view the construction's development as reflecting the cumulative effect of many minor expansions through that space.
Although, I personally have nothing insightful to say about PDP, Whitney Tabor's paper on the development of the degree modifiers "sorta" and "kinda" makes many of the same theoretical points I will be making about language change, and offers empirical (instead of merely speculative) support for the claim that neural nets provide a good processing model for this sort of phenomenon. ```
This web service brought to you by Somewhere.Com, LLC.