On ‘On binding’

In 'On binding', Chomsky reformulates the Propositional Island Condition and the Specified Subject Condition as conditions on logical form, called the Nominative Island Condition (NIC) and the Opacity Condition (OC), respectively. He also replaces cyclic nodes by 'bounding' nodes in the formulation of the Subjacency Condition. An investigation of NIC and OC shows that the claim that they are universal is virtually empty and that they cannot be regarded as natural conditions. Furthermore, they account for a quite limited range of data and the mechanisms that have been proposed to supplement them are ad hoc or problematic in other ways. Subjacency is no more satisfactory. It seems unable to account for various relevant facts. A consideration of these issues suggests that some of Chomsky's most basic assumptions should be abandoned and that more concrete frameworks are likely to prove more adequate. On binding' (Chomsky, 1980) represents an important new stage in the development of the ideas that Chomsky has been exploring since 'Conditions on transformations' (Chomsky, 1973). It introduces two important innovations. One, which forms the central theme of the paper, is the reformulation of the Propositional Island Condition (PIC) and the Specified Subject Condition (SSC) as conditions on logical form, called the Nominative Island Condition (NIC) and the Opacity Condition (OC), respectively. The other, which is rather less prominent but of comparable importance, is the replacement of cyclic nodes by 'bounding' nodes in the formulation of the Subjacency Condition. In this paper, we will take a critical look at the theoretical framework which Chomsky outlines in his paper. Chomsky suggests that the On binding' framework constitutes 'a highly unified theory, with a few abstract principles governing a wide range of phenomena' (1980: 10). We will argue that this assessment is Linguistics 19 (1981), 1133-1164. 0024-3949/81/0019-1133 $2.00 © Mouton Publishers 1134 R. D. Borsley and G. M. H rn untenable. We will develop a number of criticisms of NIC and OC, and the Subjacency Condition. These criticisms not only argue against Chomsky's assessment of the On binding' theory but also cast doubt on the general approach to syntax which Chomsky is pursuing. In our view, this approach compares unfavourably with the less abstract approaches developed in Bresnan (1978) and Horn (1979, to appear). We cannot discuss this matter in detail here but we will touch on it in a number of places. Since On binding' was written, numerous papers have appeared which revise and extend Chomsky's proposals. We will make no attempt to survey this literature. We will, however, consider some of the changes which Chomsky himself proposes in his Lectures on Government and Binding (Chomsky, 1981). In this work, which appeared when the present paper was largely complete, Chomsky develops a somewhat different framework, which he calls the government binding theory. Some of the criticisms that we advance against the On binding' theory are not applicable to the new theory. Others, however, remain valid. We will comment on this matter where appropriate. The paper is organised as follows. In section 1, we will consider the basic nature of the conditions. We will argue that the claim that they are universal is virtually unfalsifiable and that they cannot be regarded as natural conditions. In section 2, we will consider the supposed motivation for the conditions. We will show that they leave many facts unaccounted for and that the additional mechanisms that have been proposed to supplement them are ad hoc or problematic in other ways. Finally, in section 3, we will consider the Subjacency Condition. We will show that it too is effectively unfalsifiable and, more importantly, that it cannot provide an illuminating account of the various facts that it is supposed to handle. 1. The nature of NIC and OC In essence, the reformulation of PIC and SSC as conditions on logical form is an attempt to make concrete the view advanced in Chomsky (1977) that they should be seen not as general conditions on rule application but as conditions on anaphora. The new conditions are formulated as follows: (1) The Nominative Island Condition: A nominative anaphor cannot be free in S. (2) The Opacity Condition: If α is in the domain of the subject of β, β minimal, then α cannot be free in β.