Читать книгу A Companion to Chomsky - Группа авторов - Страница 30

Chomsky (1981) is a fundamental contribution to the study of human language in its effort to develop a new theory of what is universal across languages and what is variable. The main change affected the notion of filters, which came to be replaced by parameters. Parameters were seen as providing the solution to two issues: How can we capture the observed variation across the world's languages, and how do humans know so much given the limited evidence that is available to us (Plato's problem). The idea was that the child only had to set the correct value, which mostly was thought to involve a choice between two options, much like a switchbox as James Higginbotham put it. The head parameter is a simple example of this: You look at whether the verb precedes or follows the object, which gives you the two main word orders across the world's languages: verb–object, or object–verb. As Chomsky pointed out:

If these parameters are embedded in a theory of UG that is sufficiently rich in structure, then the languages that are determined by fixing their values one way or another will appear to be quite diverse (Chomsky 1981, p. 4).

Importantly, on this approach, the space of linguistic variation is part of UG.

An essential part of the framework was clustering, that is, parameters should represent clusters of properties. Chomsky (1981, p. 6) put it as follows: “[I]deally we hope to find that complexes of properties […] are reducible to a single parameter, fixed in one or another way.” A prominent example of this reasoning is provided by the Null Subject Parameter (Rizzi 1982; see also D'Alessandro 2015 for a review), which governs the realization of silent subjects, like in Spanish and many other languages. An example is provided in (11), where the subject does not have to be pronounced. This contrasts markedly with the same sentence in English, where the subject cannot be left out.

1 (11)(Voi) state leggendo un libro.(Italian)you.PL are reading a book“You are reading a book.”(D'Alessandro 2015)

Rizzi (1982) argued that being a null subject language is correlated with several other properties: i) the subject can be moved out of a finite embedded clause that is headed by an overt complementizer, (12), ii) they allow for subject inversion (Kayne 1980), (13), iv) these are languages with so‐called rich agreement on the verb (Taraldsen 1980; Alexiadou and Anangostopoulou 1998), in which both referential and nonreferential (so‐called expletive) null subjects are licensed, unlike in non‐null subject languages, (14).

1 (12)¿Quiéni dijiste que ti salió temprano?(Spanish)who say‐PRET.2.SG THAT leave‐PRET.3.SG early‘Who did you say that left early?”(Perlmutter 1971, p. 103)

2 (13)a.È arrivato Gianni.(Italian)b.*Est arrivé Jean.(French)c.*Has arrived John.(English) (Roberts 2007, p. 28)

3 (14)a.It rains frequently in April.b.*Rains frequently in April.c.*Ello/Lo llueve a menudo en abril.(Spanish)it.STRONG/it.CLITIC rain.3SG frequently in Aprild.Llueve a menudo en abril.‘It rains frequently in April.’(Judy and Rothman 2010, pp. 200–201)

This clustering highlights another virtue of the clustering idea, namely that properties that are hard to observe for the child are linked to properties that are easy to observe. It is easy for a child to determine whether the language in question is a null subject language or not, but it is not easy to observe the relevant data that will tell her that extraction out of a finite embedded clause with an overt complementizer is licit or not, as in (12). Such data are not very frequent, and this may create problems when it comes to extracting the right generalization from them when they are observed. If the clustering claim is true, the model really would have solved Plato's problem in abstraction: Learners do not have to compare whole grammars; they need to learn one parameter after the other. This would be a small number of choices, each of which would not involve many options. However, this only holds in abstraction because the model was never supplemented with an actual theory of how the child would learn the relevant parameter settings from the input.

We said “if true” because a lot of work since has cast a lot of doubt on the particular clustering that Rizzi (1982) argued for (e.g. Haspelmath 2008, Jaeggli and Hyams 1988, Newmeyer 2005, Rothman and Iverson 2007; Rothman 2009a, Sheehan, Chapter 11 of this volume). Baker (2008, p. 352) claims that “[h]istory has not been kind to the Pro‐drop Parameter as originally stated.” Since Chomsky (1981), the mainstream view on parameters has changed quite a bit, as Sheehan (Chapter 11) illustrates and which we will return to briefly below.

The first part of Principles and Parameters (P&P) was immensely productive, in particular from a cross‐linguistic point of view. Hornstein (2013, p. 399) puts this well in the following quote:

It is important to appreciate how fecund and productive this period of research [up to the 1990s] was. Arguably, we learned more new facts about more typologically diverse languages than ever in the history of the study of language. We learned a tremendous amount about how languages and grammars operate, what they have in common and how they differ.

This is an important result that is often underappreciated by Chomsky's critics.

The Principles and Parameters theory came to have two different instantiations or models (cf. Freidin and Vergnaud 2001, Hornstein, Nunes and Grohmann 2005, Lasnik and Uriagereka 2005, Boeckx and Uriagereka 2006, Lasnik and Lohndal 2010, 2010, Freidin 2012). The first one was called Government and Binding, and this was the model that was developed and used throughout the 1980s. In the beginning of the 1990s, Chomsky developed a new model, called the Minimalist Program. We will discuss the essentials of Government and Binding in the next section, and then in section 3.5, explain the transition to the Minimalist Program.