Part 1 / The Grammar of Non-determinism

1. Introduction

Chomsky’s (1995, 2000, 2001, 2008) theory of movement is mainly based on determinism in the sense that a derivation is determined by feature-checking under the last resort condition along with the notion of crash. An alternative theory of movement would be based on non- determinism in the sense that a derivation is not determined by feature-checking under the last resort condition along with the notion of crash but accounted for in terms of various interpretive systems along with the notion of Merge (Chomsky 2008).

In fact, various attempts for non-determinism have been proposed even within the deterministic frameworks. For example, even if Chomsky (2008) basically maintains the deterministic basis of feature-checking for the theory of movement especially for A-movements, he tries to adopt a non-deterministic approach for some A’-movements without much success within his deterministic framework. This monograph shows that systematic non- determinism based on interpretive systems along with the proper notion of Merge is better motivated than an entirely or partially deterministic theory of movement.

Another major claim of this monograph is that the notion of edge feature (EF) posited in Chomsky’s (2008) theory of determinism should be eliminated for the theory of non-determinism. It is proposed in this monograph that the operation of Merge, unmarked or marked, should be the derivational null hypothesis in the sense that it is always free and optional (or costless) not only for the unmarked Merge but also for the marked Merge, given the usual assumption that all the derived structures undergo interpretation at the interface for the degree or kinds of grammaticality.

The postulation of EF for every lexical item (LI) is the basic mechanism of Chomsky’s theory of determinism. Hence, every Merge should be licensed or “determined” by the EF of an LI, which makes his theory inherently deterministic. Since I propose the notion of derivational null hypothesis in place of the notion of EF for my non- determinism, the notion of EF is unnecessary, given the notion of derivational null hypothesis. Implications of non-determinism beyond movement or a possible theory of global non-determinism are discussed.
This monograph proposes that Chomsky’s (1995, 2000, 2001, 2008) theory of movement, which is mainly based on determinism (1), should be replaced by a theory of movement based on non-determinism (2) as suggested in Yang (2011b):

(1) A derivation is determined by feature-checking under the last resort condition along with the notion of crash. (Chomsky 1995, 2000, 2001, 2008 [80%])

(2) A derivation is not determined by feature-checking under the last resort condition along with the notion of crash but accounted for in terms of the architectural conditions of Merge, i.e., Merge types (EM/IM) and interpretive systems of Merge. (Chomsky 2008 [20%], Yang 2011b [100%])

I will show that notions like the last resort condition and crash should be better eliminated, supporting non-determinism (2), given the interpretive systems of non-determinism. Note that the last resort condition is not necessary for optional movements like Scrambling in languages like Japanese, Korean, etc., while their interpretive effects can be ascertained by the interpretive systems as required by non-determinism.

It will be shown below that even for obligatory movements the last resort condition is not necessary, given the Merge types and interpretive systems of non- determinism (2). The notion of crash may also be better covered by the interpretive systems, which are to deal with degrees or types of (un)grammaticality or (un)acceptability within non-determinism. Hence, the postulation of the notion of crash is not only unnecessary within non-determinism but also against the spirit of non-determinism.

The core notion of non-determinism (2) consists of Merge and the interpretive system for it. Hence, strictly speaking, non-determinism (2) is not a theory of movement but a theory of Merge. And the theory of movement should follow from the theory of Merge according to non- determinism (2). In other words, according to non-determinism (2), the theory of movement should be based on the theory of Merge, so that the Merge theory of movement according to non-determinism (2) is better motivated than the traditional theory of movement per se according to determinism (1).

In fact, the notion of Merge has been deterministically introduced. For example, Chomsky (2008) proposes (3), which provides the deterministic characterization of Merge:

(3) Every lexical item (LI) has an edge feature (EF) characterizing the propensity of LIs for being Merged with other elements (LIs).

According to (3), Merge is supposed to operate freely but along with the deterministic notion of EF. In other words, every operation of Merge should be licensed or “determined” by an EF according to (3). However, there is another possibility of characterizing the notion of Merge non- deterministically without positing the notion of EF, as in (4):

(4) Every lexical item (LI) Merges with other elements (LIs), given the usual assumption that every derived structure undergoes interpretation at the interface.

Note that every wrongly Merged structure will be filtered out by the general principles of interpretation at the interface. In fact, (4) conforms with the derivational null hypothesis in the sense that the derivational operation of Merge may be free and costless. Note that null hypothesis indicates a situation where certain properties of the grammatical system “come for free” ? that is, do not require stipulation, given the basis parameters for what the grammar is supposed to do.

I mean the derivational null hypothesis to be a null hypothesis with respect to a derivation: within a derivation, an operation may “come for free” if all the relevant operations are free and costless.

Hence, the notion of EF may be eliminated under the derivational null hypothesis within non-determinism, simplifying the minimalist theory. In fact, the notion of EF is peculiar even for or in Chomsky’s (2008) grammatical system since EF is not a feature for checking or semantic interpretation. Furthermore, it is assumed in Chomsky (2008) that EF inherently is assigned to every LI and has to be automatically deleted as soon as its Merging function is found to be no longer necessary. Obviously, the complicated conditions related to EF will disappear when EF is eliminated.

Further problems with the EF theory are as follows. It is difficult to characterize even the distinction of the so-called unmarked and marked Merge in terms of the EF theory. Note that the unmarked Merge is optional Merge whereas the marked Merge is obligatory Merge. And under the derivational null hypothesis along with (4), I may assume that Merge, unmarked or marked, is obtained simply by (4), the distinction between unmarked and marked Merge being made by the lexical specification to the effect that only the Merging head of the marked Merge is assigned the feature [+marked], which makes the Merging function of the Merging head obligatory.

In other words, only the assignment of the feature [+marked] on the Merging head of the marked Merge will make the distinction of the unmarked and marked Merge under the derivational null hypothesis along with (4). But the EF theory has to carry the additional burden of dealing with the EF in accounting for the distinction of unmarked and marked Merge, as follows. According to Chomsky (2008), it is maintained that when the Merge is unmarked or optional the EF of the Merging head may not be always deleted, or optionally deleted, whereas when the Merge is marked or obligatory the EF of the Merging head should never be deleted. Note that EF is necessary for every Merge.

A similar situation obtains in accounting for other phenomena involving the unmarked and marked distinction as in (5):

(5) The unmarked Merge is the norm while the marked Merge is rather exceptional across languages.

Again we may assume the derivational null hypothesis along with (4) to account for phenomena like (5) in the sense that the unmarked Merge is most common since it is purely due to the derivational null hypothesis, whereas the marked Merge is rather exceptional since it involves a marked Merging head with [+marked] in addition.In fact, the EF theory is no longer crucially involved in further development of the Merge theory as we see in (6):

(6) EM yields generalized argument structure (theta roles, the “cartographic” hierarchies, and similar properties); and IM yields discourse-related properties such as old information and specificity, along with scopal effect. (Chomsky 2008)

Here we see that the theory of Merge should be developed into either the theory of external Merge (EM) with respect to the generalized argument structure or the theory of internal Merge (IM) with respect to the discourse-related properties. Neither of these properties of EM and IM crucially involve the EF theory. For both of the theories of EM and IM, Chomsky (2008) proposes the principle (7):

(7) Every Merge induces a (new) interpretation. (Chomsky 2008)

Indeed, with (7) we may begin to develop the theory of EM and IM as follows. According to (7), every Merge induces an effect on the argument structure for EM or an effect on the discourse-related properties for IM. In fact, I claim that the system of discourse-related properties for IM is the so-called system of interpretive effects, which I claim would replace the minimalist movement theory (1) discussed earlier.
I propose (8a, b) as the exhaustive list of interpretive effects of IM for the theory of Merge/movement under the non-determinism (2):

(8) a. Discourse Effects: (a) topic, (b) focus, (c) givenness, (d) null effect
b. Semantic Effect: (e) scope

Among the five interpretive effects listed in (8a, b), the null effect (d), as induced by Subject-Raising in English, is the only interpretive effect realized only at PF whereas all the other interpretive effects may be realized at LF as well as PF. In other words, the null effect is manifested only phonologically without any semantic effects.

Furthermore, it is due to the marked Merge only, whereas all the other interpretive effects are not. The marked Merge induces a marked interpretive effect, behaving as an obligatory process, whereas the unmarked Merge induces an unmarked interpretive effect, behaving as an optional process. We may conclude that the null effect is the marked interpretive effect that is realized only at PF, which is rare across languages.

Note that the null effect is characterized within the Merge theory but without involving the notion of EF. In English the expletive undergoes the obligatory Subject-Raising and I claim that it induces the null effect according to the exhaustive list of interpretive effects (8a, b) within non-determinism. Hence, we can account for why an expletive undergoes only the obligatory movement of Subject-Raising in English and I claim that the expletive is subject to the PF interpretive effect only according to my theory of interpretive effects.

Given that the interpretation applies phase by phase (Chomsky 2008), we can account for the interpretive effects of a phase according to (9) under the assumption that a trace does not carry phonological information:

(9) Phonological and semantic interpretations should go hand in hand.

Note that usually within a phase the final link of a successive-cyclic Merge/movement will carry the phonological and semantic interpretations including semantic interpretive effects according to (9). (9) will also account for cases of covert Merge/movement and why QR(Quantifier Raising), a covert movement, should be local. (9) will also account for the theta effect on the initial link under the assumption that an overt element like a resumptive pronoun is possible at the initial link.

I have claimed earlier that the derivational null hypothesis along with (4) accounts for phenomena like (5) in the sense that the unmarked Merge is most common since it is purely due to the derivational null hypothesis, whereas the marked Merge is rather exceptional since it involves a marked Merging head with [+marked]:

(5) The unmarked Merge is the norm while the marked Merge is rather exceptional across languages.

The derivational null hypothesis inherently implies the free and optional operation, as discussed earlier. Hence, the non-determinism essentially based on the derivational null hypothesis can be considered roughly as a “free and optional” system. And Chomsky (2001) makes a claim (10) to the effect that the notion of interpretive effect that is considered as an essential part of the non-deterministic grammar as discussed above is of the inherent property of the optional operation:

(10) Optional operations can apply only if they have an effect on outcome (Chomsky 2001).

(10) implies that optional operations should induce the interpretive effect whereas obligatory operations do not necessarily do so, which conforms with our discussion above in the sense that the marked Merge (obligatory operation) may induce the PF interpretive effect only. I propose that we may rephrase (5) as (11):

(11) The optional operation is the norm while the obligatory operation is rather exceptional across languages.

Chomsky (2008) partially adopts non-determinism for some A’- movements, as opposed to Yang (2011b), which fully adopts non- determinism. Chomsky might be simply assuming that pure non- determinism is too idealistic in the sense that at least some phenomena in natural language are apparently “deterministic” like A-movements in English, as argued in Chomsky (2008). But Chomsky’s (2008) “partial” non-deterministic theory of grammar is losing the significant insights of non-determinism (2).

First of all, his theory never captures the fact that his fragmentary non-deterministic phenomena of A’-movements should not be “exceptional”, but due to the regular systematic contrast of the marked vs. unmarked Merge, as claimed in non-determinism (2). Secondly, his theory never captures the fact that his “exceptional” fragmentary non-deterministic phenomena of A’-movement represent the unmarked Merge, which should be the “norm” across languages, whereas his major deterministic phenomena of A-movements represent the marked Merge, which should be “exceptional” across languages.

Hence, Chomsky’s (2008) “partial” non-deterministic theory of grammar is seriously misleading: it never represents the true picture of the unmarked vs. marked distinction in natural language. Non-determinism (2) is not a theory restricted to movement. It can and should be extended beyond movement. It will be discussed how non-determinism is to be extended beyond movement. This monograph will be the basis for the extension of the non-determinism beyond movement.

이 저서는 미국 MIT의 Chomsky교수의 소위 최소주의 이론(Minimalist Program, 1995)을 더욱 발전시킨 새로운 언어학 이론을 모색한다. 특히, 원래의 최소주의 이론보다 더욱 최소주의적인 언어학 이론을 제시한다. 다시 말하면, 이 저서에서 새롭게 제시하는 소위 비결정론(Non-determinism)은 Chomsky교수의 최소주의 이론에서 결정론적(deterministic) 요인들을 가능한 한 제거한 비결정론적 이론이다. <br/><br/>따라서, 비결정론(Non-determinism)은 종래 최소주의 이론에서 분명히 결정론적(deterministic) 요소인 주변 자질 (edge feature)의 개념을 완전 제거하고 그 대신 영가설(null hypothesis)의 문을 열어 놓는다. 다시 말하면, 영가설 이론의 도입은 종래 최소주의 이론에서 주변 자질을 완전 제거하는 비결정론적 이론의 출발점이 된다. 따라서, 앞으로 소위 최소주의 이론은 비결정론적 이론으로 발전되어 갈 것이다. 비결정론적 이론의 또 하나의 특성은 파탄(crash)의 개념의 재정립이다. 최소주의 이론을 포함하여 종래의 모든 생성문법론에서 비문법적 문장은 파탄의 개념으로 규명되었었다. <br/><br/>즉, 모든 문장은 문법적이거나 비문법적인 문장으로 구분되고, 문법이란 문법적인 문장만을 구분해내는 결정론적 규칙들의 집합으로 간주되어 왔다. 따라서 모든 비문법적인 문장은 그 도출이 파탄나 있다고 보았다. 그러나 비결정론에서는 모든 도출의 파탄 여부뿐 아니라 그 (비)문법성의 성격까지 해석부(interpretive system)에서 평가된다고 본다. 따라서, 비결정론에서는 통사부(syntax)보다 해석부의 역할이 더 크게 부각된다.<br/><br/>This monograph presents a new non-deterministic linguistic theory developed from the so-called Minimalist Theory (Chomsky 1995) by eliminating from it such deterministic elements like the edge feature. Chomsky’s Minimalist Theory is a rather deterministic theory in the sense that it is based on the deterministic assumption that every linguistic structure is formally or syntactically determined. <br/>It is, however, shown in this monograph that the strict determinism cannot be maintained. Hence, a non-determinism is proposed in place of the determinism in this monograph. For the non-determinism the following proposals are made. First, the edge feature or the core deterministic property of the determinism is eliminated as suggested above. <br/><br/>Second, due to the elimination of the edge feature the notion of the null hypothesis has to be introduced, which leads to the explanatory value of the non-determinism. Third, the deterministic notion of crash is replaced by an interpretive notion of the types of uninterpretability.* This monograph was recommanded and financially supported by the Korean Research Foundation, to which I remain grateful.<br/><br/>In other words, the notion of crash is to be considered as a kind of the interpretive notion rather than a syntactic notion to be syntactically defined.<br/>Specifically, Chomsky’s (1995, 2008) minimalist theory is mainly based on determinism, which may be represented as (1), whereas non- determinism may be represented as (2):<br/><br/>(1) A derivation is determined by feature-checking under the last resort condition along with the notion of crash. <br/><br/>(2) A derivation is not determined by feature-checking under the last resort condition along with the notion of crash but accounted for in terms of the architectural conditions of Merge, i.e., Merge types (EM/IM) and interpretive systems of Merge. <br/><br/>Part 1 The Grammar of Non-determinism presents arguments and evidence for the superiority of non-determinism as opposed to determinism. It is argued that non-determinism prevails not only in syntax but also in non-syntax areas such as semantic interpretations and lexicon. In fact, it is further argued that non-determinism leads to an optimal grammatical theory. <br/><br/>In Part 2 Global Non-determinism, it is argued that non-determinism applies globally across linguistic levels, leading to better linguistic explanations. Specifically, in Part 2-1 Global Non-determinism for Merge it is argued that Merge should apply globally across linguistic levels. In Part 2-2, The Global Nature of Interpretive Effects of Merge it is argued that the proper interpretive effects of Merge may be properly captured across linguistic levels.<br/><br/>In Part 3 Non-determinism and Null Hypothesis, it is argued that non-determinism leads to the issue of null hypothesis in generative grammar. Specifically, in Part 3-1 Derivational Null Hypothesis, it is discussed why and how the notion of derivational null hypothesis should be characterized in generative grammar. In Part 3-2 Null Hypothesis in Generative Grammar, it is further discussed how the notion of null hypothesis has to be recognized in generative grammar. <br/><br/>To conclude, with the notion of non-determinism we can develop an optimal minimalist theory of grammar.