PhiN 49/2009: 84

Mehdi Vaez Dalili (Isfahan/Iran)

Agreement (AGR) and the Pro-drop/Non-pro-drop Variation: A Meta-analysis of GB and MP accounts

Agreement (AGR) and the Pro-drop/Non-pro-drop Variation: A Meta-analysis of GB and MP accounts
This paper aims to provide a meta-analysis of the role of Agreement (AGR) in the binary variation between pro-drop (e.g. Persian, Spanish, Italian) and non-pro-drop languages (e.g. English, French, Swedish). To this end, seven accounts based on the theory of Government and Binding (GB), and three major accounts formed according to the Minimalist Programme (MP), are investigated. The GB accounts mainly emphasise the relationship between rich agreement morphology and the pro-drop parameter. Also the MP accounts generally focus on checking strong nominal features of subjects, either through moving them out of their base-generated position in Spec-VP to Spec-IP or through forcing them to remain in-situ, respectively characterising non-pro-drop and pro-drop languages. The crux of the present study is to exactly reveal how rich/weak AGR morphology (in GB terms) or weak/strong nominal features (in MP terms) result in the pro-drop/non-pro-drop distinction and determine other subsidiary properties of each language type as a corollary of this distinction.

1 The pro-drop parameter: An introduction

Within Chomsky's Principles and Parameters (P & P) theory (Chomsky 1981a, 1981b, 1986a, 1986b), the study of parameters has received special attention, revealing the nature of UG-based differences between languages. One of the oft-cited parameters which has interested many researchers is the pro-drop parameter. The pro-drop (or null subject) parameter "determines whether the subject of a clause can be suppressed" (Chomsky 1988: 64). In fact, it has two 'settings' or 'values' which allow for a binary crosslinguistic variation. While some languages allow subjects to be phonetically covert others require them to be overtly observable in the sentence. The former are known as pro-drop languages with [+ pro-drop] setting (e.g. Persian, Spanish, Italian, etc) and the latter are called non-pro-drop languages with [–pro-drop] setting (e.g. English, French, Swedish, etc.). For example, unlike English, Persian allows null subjects:

Like all the parameters in the Principles and Parameters (P & P) framework which are supposed to include a cluster of apparently disparate syntactic properties (Chomsky 1981a), the pro-drop parameter is also claimed to consist of four clustered properties: (i) null subjects, (ii) subject-verb inversion, (iii) expletives (or pleonastics), and (iv) that-trace effect (or that-trace sequence) (Chomsky 1981a; Jaeggli 1982; Rizzi 1982). Comparing the Persian clause in (1a) above and its English equivalent (1c), one finds out that the pronominal subject, due to its recoverability from the verb ending -and, is optional in the finite Persian clause; in fact, it can be either present or absent (i.e. null pro). The phenomenon is the same in embedded clauses of Persian, as shown in (2a) below:

The second property of the pro-drop parameter deals with expletives it and there, also known as 'dummies' or 'pleonastics'. Expletives it and there do not have any semantic content. For example, in a non-pro-drop language like English, one might see:

The third property of the pro-drop parameter is called 'subject-verb inversion'. It is the property that allows pro-drop languages like Spanish to have verb-subject order in ordinary declarative clauses. The Italian and Spanish examples in (5) and (6) below show the subject-verb inversion:

The fourth property of the pro-drop parameter is that-trace effect (also known as that-trace sequence or that-trace filter). The restriction on the extraction of a subject preceded by that is called that-trace effect. Ouhalla (1999:315) points out that: "That-trace effect deals with the possibility of extracting an embedded subject across an overt complementiser, for example in question formation". In general, "the sequence of an overt complementiser followed by a trace is ungrammatical" (Haegeman 1991:362) whether the complementiser is that or for. In non-pro-drop languages, that-trace sequence is considered ungrammatical, whereas in pro-drop languages it is allowed. Considering the following Persian sentence and its English equivalent shows the distinction:

The Persian sentence (7a) is completely grammatical, for that-trace sequence is allowed in Persian. However, the ungrammaticality of the English sentence (7c) is due to the presence of the complementiser that. If the complementiser was null, the sentence would be grammatical:

(8) Who_i do you say [_CP e [_IP t_i went]]?

It is argued that each language type, whether pro-drop or non-pro-drop has these prosperities set in a specific way so that the fixed settings form a cluster. Pro-drop languages allow null subjects, subject-verb inversion, that-trace sequences, but do not have expletives. On the other hand, non-pro-drop languages do not allow null subjects, subject-verb inversion, and that-trace sequence, but allow expletives.

2 Syntactic accounts of the pro-drop parameter: GB and MP accounts

In the following, ten major accounts of the pro-drop parameter will be investigated: seven Government and Binding (GB) accounts and three Minimalist Programme (MP) accounts.

2.1 Government-Binding (GB) accounts of the pro-drop parameter

This section deals with seven GB-based explanations of the pro-drop parameter, including Taraldsen's generalisation (1978), Chomsky's Empty Category Principle (ECP) (1981a), Chomsky's Extended Projection Principle (EPP) (1986b), Rizzi's Licensing and Identification (1986), Huang's Generalised Control Theory (1984,1989), Hyams' AG/PRO Analysis (1986,1987), and Jaeggli and Safir's Morphological Uniformity Principle (MUP) (1989).

2.1.1 Taraldsen's generalisation (1978)

Research over the years has uncovered an impressive repertoire of effects attributed to Taraldsen's generalisation (Taraldsen 1978). Taraldsen's generalisation can be put as:

(9) Pro is licensed iff agreement is sufficiently rich to recover its features.

Unravelling the above sentence, Taraldsen suggests that null subjects can occur in languages with rich inflectional morphology. Rich inflectional morphology has often been associated with pro-drop languages. Therefore, it is claimed that pro-drop languages allow null subjects because they can recover the content of pro from the Agreement (AGR) features of Inflection (INFL) on the verb. However, non-pro-drop languages seem to compensate for their comparative lack of inflection on the verb by having visible subjects.

Languages vary with respect to whether the INFL features of T(ense) and AGR(eement) are overtly realised on the verb. (AGR)eement features (also known as the φ- features of INFL) including person, number, gender, and case (Chomsky 1982) are linked to the subject. In richly inflected languages such as Italian, Spanish, and Persian, the φ- features are realised on the verb in the form of overt inflection. Considering the Persian sentence below, this can be seen in a richly inflected, pro-drop language:

As a null subject language, Persian has rich Agreement (AGR) features on verbal Inflections (INFL), implying the 'number' and 'person' of the subject. So, the '–and' ending in the verb 'goft-and' in (10a) can be co-indexed with the pro in the Spec-IP position to recover its content, but in English, as a non-pro-drop language, there are no Agreement (AGR) features on the verbal Inflection (INFL); hence no pro allowed in the subject position. So, it is how the presence of an overt subject is obligatory in Spec-IP in the sentence (10c).

Despite all the excellent considerations inspired by Taraldsen's initiative a number of critiques raised as to the correlation between rich agreement and allowing null subjects. Three major challenges to his idea are:

1. The notion of 'rich agreement' is vague; in fact, it is not precisely defined (Chomsky 1982; Shlonsky 1989; Rohrbacher 1994).

2. Even given clearly rich agreement, pro-drop might not occur. As Bouchard (1984) points out, languages with rich morphology, such as French, are not necessarily pro-drop. It is also the case with Icelandic (Sigurðsson 1993). So, rich agreement is not a sufficient condition for a language to be pro-drop.

3. The most troubling and damning problem, as Deal (2005a) puts, is that even given clearly poor agreement, pro-drop might still occur; in other words, rich agreement is not a necessary condition for a language to be pro-drop, as in Chinese, Japanese, Korean, and Malayalam (Huang 1984, 1989). Mohanan (1983) and Huang (1984) claim that the correlation between rich agreement and pro-drop is at best a tendency and not an absolute principle.

Consequently, on the one hand, the potentiality of Taraldsen's generalisation to bring together all sorts of data from many different flavours of languages is still very attractive to the continuing research project of pro-drop and on the other hand, the generalisation is incomplete with respect to the sum of null subject languages, some of which seem to require other mechanisms or additional stipulations.

2.1.2 Chomsky's Empty Category Principle (ECP) (1981a)

Comparing the sentences below:

one can understand that there is a distinction between the extraction of objects, as in (11), and of subjects, as in (12). This distinction is called 'subject-object asymmetry' with respect to extraction. The phenomenon in (12b) is known as that-trace effect which is considered as ungrammatical in non-pro-drop languages. It refers to the extraction of a subject preceded by a complementiser. In general the sequence of a 'Complementiser (COMP) + trace (t)' is considered ungrammatical. But the question arises as to what the difference between (11b) and (12b) is. Consider the respective partial schematic representation of each of these sentences in Figures 1 and 2 below:

Figure 1: Object extraction

Figure 2: That-trace effect in subject extraction

What is the difference between the two so that the former is considered as a grammatical sentence, whereas the latter is not? The difference derives from the more fundamental concept of 'proper government'. Government is a version of c-command, with two types of restrictions:

Firstly, while any element can c-command another, the ability to govern is restricted to a limited number of 'governors', basically, the set of lexical heads (i.e. N, V, A, and P), though there are a number of exceptions. Secondly, government is restricted compared to c-command in that it is limited from the bottom as well as the top. (Cook & Newson 1996:240)

Along the same line, Cook and Newson (1996) discuss and define the 'proper government' as:

α governs β if and only if: (1) α is a governor (e.g. N, V, P, A, etc.) (2) α and β mutually c-command each other. (Cook & Newson 1996: 240)

Based on the above definition, the lower trace in Figure 1 is governed by the lexical head invite, while the lower trace in Figure 2 is governed by the functional head that. So, according to 'proper government', the former is grammatical, but the latter is not, due to the violation of 'proper government'. In fact, in Figure 2, the complementiser that acts as a 'barrier' which blocks the proper government of the higher trace. Drawing upon the notion of 'barrierhood' of that in that-trace sequences, an additional form of 'government' known as 'antecedent government' is introduced:

α antecedent-governs β iff: i) α and β are co-indexed ii) α c-commands β iii) α is not separated form β by a barrier. (Ouhalla 1999: 267)

Referring back to the differences between Figures 1 and 2, it can be suggested that in Figure 2 the lower trace in the subject position must be 'antecedent-governed' by the higher trace, but as it is observable antecedent government is blocked by that barrier. On the other hand, in Figure 1, the lower trace in the object position is 'head-governed' by the lexical category invite. So, the difference in object and subject extraction can respectively be attributed to the distinction between 'antecedent government' or 'head government' (Cook & Newson 1996: 262).

Thus, employing the new concept of antecedent government, the notion of 'proper government' can be paraphrased as:

α properly governs β iff: i) α governs β and α is a lexical category OR ii) α antecedent-governs β. (Ouhalla 1999: 267)

The bottom line of the discussion on the notion of proper government leads to a principle known as the Empty Category Principle (ECP). Empty Category Principle (ECP) can be defined as (13) below:

(13) An empty category must be properly governed.

Providing a relatively plausible account for the pro-drop phenomenon, Chomsky (1981a) maintains that in pro-drop languages, the empty category pro is properly governed, but it is not the case with non-pro-drop languages. But how?

Based on Taraldsen's generalisation, it is argued that the Inflectional (INFL) category plays a crucial role in pro-drop languages, while it does not in non-pro-drop languages. Drawing upon the critical nature of INFL in explaining the pro-drop phenomenon, the ECP account claims that INFL is the proper governor in pro-drop languages, but not in non-pro-drop languages (Cook & Newson 1996: 61). In fact, having pronominal features (i.e. lexical features) in pro-drop languages, INFL can properly govern the empty category pro (Rizzi 1982). However, in non-pro-drop languages, the INFL position is filled by functional categories (e.g. auxiliaries like will, should, etc.) or otherwise nothing. So, INFL in such languages does not have lexical properties and as is understood from Empty Category Principle (ECP), not having lexical properties equates with not having proper government. A comparison between the Persian sentence (14a) and the English sentence (14c) in the following illustrates the point:

In the Persian sentence the head of IP is the inflectional morpheme '–and' which carries precise lexical properties, as Rizzi (1982) argues, and properly governs the empty category pro; hence it is co-indexed with pro. However, the head of IP in the English sentence is are. It is clear that are is an auxiliary and in turn considered as a functional category which does not have lexical properties; hence it cannot properly govern the empty category pro and such is the case that English, as a non-pro-drop language, does not allow null subjects.

To sum up, what Chomsky (1981a) puts forward under the Empty Category Principle (ECP) can be considered as one of the tenable accounts of the pro-drop parameter within the Government Binding (GB) framework, which again employs the Inflectional (INFL) properties (or more specifically AGR features) of languages to explain whether they allow null subjects or not.

2.1.3 Chomsky's Extended Projection Principle (EPP) (1986b)

Extended Projection Principle (EPP) was initially formulated as a well-formedness condition that requires clauses to have a formal subject. Early proposals concerning such a requirement were formed by Chomsky (1981a) and Rothstein (1983).

With respect to EPP every clause must feature a subject-related element in the canonical subject position, Spec-IP. This element is either an argument with semantic content or an expletive (e.g. it or there) with no semantic content. In other words, even when a verb does not select a semantically active subject, the subject position must still be projected as in the case of clauses having expletive it or there in the Spec-IP position.

Along with the introduction of EPP, some explanatory accounts were developed. In later versions of GB, it is claimed that subjects are base-generated in the Spec-VP. In fact, the subject of a clause moves from Spec-VP position to the initial position of the sentence (Spec-IP or Spec-TP). This movement is claimed to be induced by the Extended Projection Principle (EPP) that requires the subject (Spec-IP) position to be filled. The phenomenon is represented in Figure 3 below:

Figure 3: Movement of subject from Spec-VP to Spec-IP to satisfy EPP

According to the Extended Projection Principle, the I' projection which constitutes the predicate of the sentence, as it can be seen in the above figure, must have a specifier (i.e. Spec-IP). Put in other words, the standard EPP in generative grammar is any grammatical arrangement that serves to guarantee that a predicate is 'anchored', due to the presence of a subject (Chomsky 1981a, 2001).

Dealing with the application of the Extended Projection Principle to the pro-drop parameter, it is observed that many of the world's languages clearly do not have subjects, with both referential pronouns and expletives failing to surface overtly in subject position. In order to deal with the apparent problem of the pro-drop languages, it is therefore proposed that the EPP need not be overtly satisfied, but that it is nevertheless covertly satisfied in pro-drop languages by means of a silent pronominal element dubbed pro. In fact, in pro-drop languages it is pro which fills the Spec-IP position and satisfies the EPP and in non-pro-drop languages, the EPP is satisfied through the occupation of Spec-IP position by an argument or else by an expletive.

More recently, however, it has been emphasised that the EPP does not require an overt or covert element to occupy Spec-IP, and that the introduction into the IP-domain of appropriate inflectional element associated with a verb which undergoes raising to the head I position of IP can be sufficient (Alexiadou & Anagnostopoulou 1998). On this analysis, pro-drop languages such as Persian, Spanish, or Italian which appropriately feature rich verbal morphology can therefore satisfy the EPP without the need for any lexically realised subject. Interestingly enough, it is remarkable that the rich agreement phenomenon proposed by Taraldsen (1978) surfaces again in the EPP account.

Generalising the requirement of having a subject position in Spec-IP position to other types of functional categories, the Extended Projection Principle later developed into 'Generalised Extended Projection Principle' (Chomsky 1998). According to 'Generalised Extended Projection Principle', it is required that not only I but also other functional heads which have uninterpretable selectional features for their specifier positions be filled by proper elements. In other words, it is claimed that CPs, VPs, and DPs should also have some obligatory Spec position.

In a nutshell, the importance of such a universal principle is to the extent that it has continued to be reformulated and investigated by Chomsky (1986a, 1995, 2001) in the Minimalist Programme. However, recently, a number of works have appeared which argue that the EPP can be, and should be, eliminated (Boeckx 2000; Boškovič 2001; Castillo et al. 1999; Epstein & Seely 1999; Grohmann et al. 2000; Martin 1999).

2.1.4 Rizzi's Licensing and Identification (1986)

Capitalising upon Taraldsen's generalisation (1978) on the correlation between rich agreement and allowing pro-drop, one can infer that subjects need not be overt in pro-drop languages because the verbal inflections (specifically AGR features) uniquely define the number and person of pro. In fact, there is still a subject in Spec-IP position, just as if it were an overt pronoun. To capture this idea, Rizzi (1982, 1986, 1990) addresses the
pro-drop question by distinguishing 'licensing' conditions (i.e. what allows the appearance of a null subject) and 'identification' (or recovery) processes (i.e. how the referent of a null subject is semantically recovered).

Rizzi (1982) makes the significant claim that pro-drop languages differ from
non-pro-drop languages in that the former have verbal inflections characterised by a [+pronoun] feature with clitic-like properties. The subject pronoun cliticises onto the verb, permitting recovery of person and number features. This insight is pursued in subsequent works (Jaeggli 1982; Bouchard 1984; Rizzi 1986; Roberge 1990) where the structure is assumed to be such that subject clitics (shown as cl) can be proper licensers of pro as shown in Figure 4 in the following:

Figure 4: Licensing of pro by the clitic-like properties of AGR in pro-drop languages

In Figure 4, the subject clitic properly governs the DP in the Spec-IP position. If no lexical material is present in the DP position, pro fills the position, receiving its licensing by the clitic under AGR. This provides the explanation for the licensing of pro in pro-drop languages which have rich agreement.

Concerning the 'identification' of pro, Jaeggli and Safir (1982) propose two main ways by which null subjects can be identified (recovered). The first type of identification is identification by Agreement (AGR) whereby AGR affixes with the relevant φ-features identify pro in pro-drop languages with agreement morphology.

The second type of identification, known as 'topic-chaining', concerns pro-drop languages which lack agreement morphology (e.g. Chinese, Japanese, Korean, Malayalam, etc.). As for the 'identification' of the null subjects in such languages, Huang (1984) draws a distinction between 'discourse oriented' and 'sentence oriented' languages. While 'discourse oriented' languages such as Chinese, Japanese, Korean, and Malayalam lack inflectional morphology, 'sentence oriented' languages like Persian, Spanish and Italian have rich inflection to identify agreement and tense features of pro. Hyams (1992: 260) explains the difference in identification as below:

Thus, in 'discourse oriented' languages 'topic chains' are employed in order to "delete the topic of a sentence under identity with a preceding sentence" (Huang 1984:549), whereas 'sentence oriented' languages lack the rule of 'topic-chaining' and instead apply the first type of identification – identification of pro by AGR features.

Analysing licensing and identification with reference to pro-drop typology, Deal (2005b) presents a relatively complete analysis of the ways by which licensing and identification (recovery) are characterised in full pro-drop and partial pro-drop languages (Table 1 below). She divides licensing of pro into two categories and distinguishes between: i) licensing 1st/2nd person and ii) licensing 3rd person. As for identification, she distinguishes between identification through agreement and identification through discourse. The discoursal approach to recovery is in turn divided into: i) recovery through topicality and ii) recovery through speech acts.

Deal combines all these data and considers the behaviour of each language in terms of being full pro-drop, partial pro-drop or not allowing pro-drop at all. Italian and Spanish license pro-drop for all persons and recover the content of pro by agreement. Chinese and Old Icelandic also license pro-drop for all persons, but being topic-oriented languages, they recover the content of pro through topicality or 'topic chains' (Huang 1984). As non-pro-drop languages English and Swedish do not license pro for any of the persons, lacking the recovery forms mentioned above. Nez Perce, licenses pro for the 1st and 2nd persons, but not for the 3rd person. It also recovers the content of pro for the 1st and 2nd persons through speech acts. Deal (2005b) claims that if a language recovers pro via speech-act presence, pro-drop should be constrained to 1st/2nd persons. Like Nez Perce, Hebrew and Finnish license pro for the 1st and 2nd persons, but they recover the content of pro through Agreement. Deal's (2005b) categorisation is tabulated in Table 1 below:

Table 1: Licensing and identification in full pro-drop and partial pro-drop languages

To sum up critically, Rizzi's licensing and identification, still based on rich agreement features (Taraldsen 1978), partially works for languages such as Spanish and Italian, which have rich agreement, but runs into difficulties with other languages such as Chinese, Korean, Japanese, or Malayalam that lack agreement features but still license pro.
Huang (1989) addresses this problem through a 'generalised control theory' taking both types of languages into account.

2.1.5 Huang's Generalised Control Theory (1984, 1989)

Nearly all of the accounts presented up to this point have been in one way or the other based on the correlation between rich agreement morphology and the pro-drop parameter. However, these accounts fall short of explaining the pro-drop phenomenon in languages which lack agreement but yet identify null subjects. To address the shortcomings of previous accounts, Huang (1989) proposes an inclusive 'generalised control theory' to explain for the lack of subjects in Chinese-type languages, too.

Huang (1984, 1989) argues that the entities of PRO and pro should be conflated into a single empty category type (Pro), both of which are subject to the Generalised Control Rule (GCR). Drawing upon the notion of 'control' presented in Manzini (1983) and Nishigauchi (1984), Huang defines the Generalised Control Rule (GCR) in the following way:

Generalised Control Rule: An empty pronominal is controlled in its control domain (if it has one). (Huang 1989:193)

He defines 'control domain' as:

α is the control domain for β iff it is the minimal category that satisfies both (a) and (b): (a) α is the lowest S or NP that contains (i) β, or (ii) the minimal maximal category containing β (MMC(β)). (b) α contains a SUBJECT accessible to β. (Huang 1989:193).

In a finite clause in both English-type and Italian-type languages, which have some kind of agreement morphology, the lowest S is the control domain for a subject pro, as the S contains pro and an accessible subject (AGR). However, the two languages are distinguished as to whether AGR can be a controller or not, whether it is strong or weak. In Italian the AGR is strong enough to determine the content of pro, thus (16b) is grammatical, whereas in English the AGR is too weak to control pro, thus (16a) is ungrammatical.

Chinese-type languages differ from both English-type and Italian-type languages with respect to the control domain for pro. The control domain for the pro in the Chinese example below is not the minimal S, since there is no subject (AGR) accessible to it, on the assumption that AGR is systematically absent even in finite clauses in Chinese:

(17) . . . [s pro lai le] (Chinese) (Huang 1989:193)

In such a case where there is no immediate controller, the reference of pro may involve long-distance antecedents, arbitrary reference, or even pragmatic considerations. Huang (1984) argues that Chinese, as a topic-drop and discourse-oriented language, identifies the subject of a minimal S through 'topic-chaining'. This means that pro in (17) is controlled by the topic in the higher clause(s). Hyams (1992: 260) illustrates the topic-chaining phenomenon as:

(18) Discourse Topic_i [topic_i [S pro_i [INFL]…..]

Interestingly, the situation of pro in the minimal S in (17) is parallel to the occurrence of PRO in non-finite clauses in (19) in the following where the minimal clause is not the control domain for PRO, lacking an accessible subject (AGR):

(19) . . . [s PRO to come] (English)

When the minimal S in (17) and (19) is an embedded clause (i.e. contextualised in an environment), both pro and PRO have a control domain. In such cases pro and PRO are controlled, their reference being determined by the controller NP. So, that is why Huang argues that PRO and pro should be conflated into a single empty category.

On the other hand, when the minimal S is a decontextualised one, there is no control domain for pro and PRO and their references are free (i.e. interpreted as an arbitrary person). In fact, when a structure like (17) constitutes an independent clause, there is no control domain for the subject gap; hence its reference is undetermined. The gap cannot be pro, but some other type of empty category called 'zero topic'.

What is important in Huang's theory is that the distribution and reference of pro can be deduced from the Generalised theory of Control, thus eliminating the need for the theory of pro-drop. In other words, the parametric differences regarding pro-drop are reduced to the different properties of AGR and the presence or absence of AGR is crucial for determining the control domain for pro: if AGR is present, pro must be controlled within the minimal S containing it (Italian and English), while if AGR is absent (as in Chinese), pro may be controlled by a controller from outside of the clause containing it. Moreover, for languages with AGR the strength of AGR determines whether it can be a controller or not: the strong AGR in Italian can be a controller, allowing null subjects, but the weak AGR in English cannot, disallowing null subjects.

Although the theory takes both languages with and without agreement morphology into account, it is not as inclusive and flawless as it intended to be. Haegeman (1991) points out that Generalised Control Rule (GCR) implies that pro is possible either in languages with rich agreement or no agreement at all. However, this generalisation does not hold universally, since German and mainland Scandinavian languages have overt agreement and still cannot be considered as pro-drop languages (Platzack 1987).

2.1.6 Hyams' AG/PRO analysis (1986, 1987)

Following Zagona's (1982) work on the relationship between auxiliaries and the pro-drop phenomenon, Hyams (1986, 1987) presents an analysis of the pro-drop parameter known as 'AG/PRO parameter' (AG=AGR). To explain her AG/PRO parameter, Hyams investigates three properties of PRO (Hyams 1986: 34) and claims that all of them are comparable with those of AG. Based on the similarity or dissimilarity of the properties of AG to PRO (i.e. AG=PRO or AG?PRO), she distinguishes between pro-drop and non-pro-drop languages.

As for the differences between pro-drop and non-pro-drop languages, Hyams asserts that "certain differences in the auxiliary systems of pro-drop and non-pro-drop languages (Italian and English respectively) are closely connected to the null subject phenomenon" (Hyams 1986: 26). As a matter of fact, modals and auxiliaries are realised as INFL in later versions of GB (Chomsky 1986a, 1986b). However, since Hyams' AG/PRO analysis is mainly based on Chomsky's early version of (1981a) GB, modals and auxiliaries are realised as AG (or AGR).

According to her distinction between pro-drop and non-pro-drop languages in terms of their auxiliary system, it is noticeable that pro-drop languages (Figure 5 below) regard modals and auxiliaries as affixes base-generated on the main verbs and subsequently AGR remains empty. It is claimed that in such languages AG (or AGR) has the same properties as PRO (i.e. AG=PRO). So, if PRO is ungoverned , then AGR should be ungoverned too and to satisfy this requirement, AGR allows the Spec-AGRP position to be empty and it is how pro is allowed to occupy the Spec-AGRP in such languages. The relation between AG and PRO in pro-drop languages can be expressed through the syllogism in the following:

(20)

AG is ungoverned too Pro is allowed.

Figure 5: Pro-drop languages (AG = PRO)

On the other hand, non-pro-drop languages (Figure 6 below) distinguish modals and auxiliaries from main verbs, considering them as autonomous elements base-generated in AG (AGR) position and main verbs positioned in the head V position. In such languages AG does not have the same properties as PRO (i.e. AG ≠ PRO) because PRO is an empty category while AG is filled by a modal or auxiliary. So, if AG's properties are opposite to the properties of PRO as an ungoverned category, then AG should be a governed element. Consequently, AG is governed by a lexical item in Spec-AGRP (i.e. subject) position and it is so that non-pro-drop languages should have overt subjects. The relation between AG and PRO in non-pro-drop languages can be expressed through the following syllogism:

(21)

AG should be governed Pro is not allowed

Figure 6: Non-pro-drop languages (AG ≠ PRO)

2.1.7 Jaeggli and Safir's Morphological Uniformity Principle (MUP) (1989)

Although Hyams' AG/PRO analysis dispensed with the relationship between rich agreement morphology and the pro-drop phenomenon, Jaeggli and Safir's Morphological Uniformity Principle (1989) still capitalises on the inflectional properties of languages. Moreover, while trying to account for the presence of pro-drop in both languages with rich inflectional systems and those lacking inflectional systems, they attempted to modify Huang's Generalised Control theory.

Jaeggli and Safir (1989) maintain that the main property of a null subject language is what they call 'morphological uniformity'. They define 'morphological uniformity' in the following way:

An inflectional paradigm P in a language L is morphologically uniform iff P has either only underived inflectional forms or only derived inflectional forms. (Jaeggli & Safir 1989)

According to their formulation, [+pro-drop] is licensed only in languages which are morphologically uniform and [–pro-drop] is licensed only in languages which have non-uniform or complex inflectional paradigm. They restate this 'licensing condition' as the 'Morphological Uniformity Principle' (MUP):

Null Subjects are permitted in all and only those languages which have morphologically uniform inflectional paradigms. (Jaeggli & Safir 1989)

Analysing the Morphological Uniformity Principle (MUP), Cook and Newson (1996: 287) examine the different present tense inflectional paradigms in English, Spanish, Chinese, and Irish. Table 2 in the following displays the inflectional paradigms of these languages along with those of Persian and French:

Table 2: Paradigms for the present tense inflections in six different languages

Considering the inflectional paradigms of each language, it is evident that where morphologically complex (speakS) and morphologically simple (speak) forms coexist, null subjects are not allowed. The inflectional paradigm is, in this case, a non-uniform or mixed one. So, according to MUP, English should be a [– pro-drop] language. French is also morphologically non-uniform (complex); hence a [– pro-drop] language. In contrast, in Persian, all of the verbal inflections exhibit uniformity in that each of them has an idiosyncratic inflectional morpheme. So, that is how Persian is considered a [+ pro-drop] language. Owing to its morphological uniformity, Spanish as in the case of Persian is deemed to be a [+ pro-drop] language. In Irish all of the paradigms present the same ending –rann. Therefore, being morphologically uniform it is judged to be a [+ pro-drop] language. Chinese like Irish, in which the forms are never morphologically complex, will also allow null subjects; hence a [+ pro-drop] language. In sum, the first two languages which are morphologically complex are non-pro-drop languages, but the other four languages showing morphological uniformity are pro-drop.

Like Huang's Generalised Control theory, Jaeggli and Safir's MUP accounts equally well for Persian-type and Chinese-type languages (i.e. both for languages with rich inflection and without inflection). However, "the morphological uniformity solution raises as many problems as it solves" (Cook & Newson 1996: 289). For example, it does raise the new problem of why German, which has rich inflection, is a [– pro-drop] language. To solve the problem, Jaeggli and Safir (1989) resort to an explanation which draws a clear line between 'licensing' and 'identification' of pro. Morphological uniformity, as they assert, characterises just the 'licensing condition' presenting the context in which pro can occur. But pro, like any null category, must also be identified (i.e. its reference has to be available). On their analysis, pro can be identified by: i) local AGR (which must include Tense) or ii) a c-commanding nominal or iii) a topic. Analysing the German example, Jaeggli and Safir (1989) argue that, while pro might well be licensed in German, it could not be identified because the verb second phenomenon separates tense and AGR into different nodes. Therefore, violating the first approach to identification of pro, it is impossible to identify pro in German and that is why this language is a non-pro-drop language.

Some L₁ and L₂ acquisition studies have drawn upon the MUP account (Hilles 1991; Lakshmanan 1991; Meisel 1991; Hyams 1992), but none of them precisely proves the plausibility of the Morphological Uniformity Principle (MUP) account.

2.2 Minimalist Programme (MP) accounts of the pro-drop parameter

The advent of the Minimalist Programme (MP) affected many aspects of the generative enterprise. Not excluded from the impact of such linguistic development was the pro-drop parameter. Some generative linguists interested in pro-drop phenomenon tried to investigate it from the MP viewpoint. Below, three MP accounts of the pro-drop parameter are discussed: Speas' Economy of Projection Principle (1994, 1995), Rohrbacher's Full Paradigm Condition (1992, 1994, 1999), and Radford's feature-checking account (1997, 2004).

2.2.1 Speas' Economy of Projection Principle (1994, 1995)

Speas (1994, 1995) considers the pro-drop phenomenon from a Minimalist stance, but her theory still captures the crosslinguistic generalisation that while subjects are omitted in languages with rich agreement morphology (e.g. Persian, Spanish, Italian) and languages with no agreement (e.g. Japanese, Chinese, Korean, Malayalam), they must be overt in languages with weak agreement morphology (e.g. English, Swedish). Speas capitalises upon Chomsky's Economy of Projection Principle to differentiate between pro-drop and non-pro-drop languages. Chomsky's (1991) definition of the 'Economy of Projection Principle' is:

(22) Project XP only when XP has content.

Simply, this principle requires that a maximal projection always contain more than its complement. Therefore, the principle prohibits structures as in Figure 7 below in which both the specifier of XP and the head of XP lack content:

Figure 7: Economy of projection principle violated

The central point is that in order for a maximal projection to have content, at least (minimally) its head or specifier position must be lexically (phonetically) or semantically filled. Of course, both the specifier and the head may also have content at the same time. As a general rule, according to the Economy of Projection Principle, the Spec position of maximal projections such as VP, TP, AGRP, or any others may remain empty provided that the head of the same maximal projection is filled by phonetic or semantic content. So, for instance, if at the VP level the verb provides the head of VP with content, no overt subject in Spec-VP is needed to license the VP maximal projection. The maximal projection TP is also licensed by its head since T has independent semantic content, even when it does not contain phonetic material. Thus, Spec-TP may also remain empty. However, dealing with AGR, Speas contends that what separates pro-drop languages from non-pro-drop languages is where inflectional affix (i.e. AGR) is generated, i.e. either separately from the verb in the head AGR position or directly on the verb. However, this does not mean that if, in non-pro-drop languages, the inflectional affix is generated on the verb, there is no need for AGRP projection. That is why Speas assumes that the AGRP projection is necessary if and only if a language has some sort of agreement, no matter how residual. Therefore, both pro-drop and non-pro-drop languages must project AGRP.

Speas argues that in pro-drop languages which have rich AGR, the inflectional affix is base-generated in the head AGR position of AGRP. She calls this type of agreement 'strong' agreement. So, when the head position of the maximal projection AGRP is filled by lexical content, the Economy of Projection Principle is satisfied and the Spec-AGRP is licensed to be empty (i.e. pro). The schematic representation of the 'strong agreement' phenomenon in pro-drop languages is as follows:

Figure 8: Generation of inflectional affix in AGR in pro-drop languages

However, in non-pro-drop languages which have weak AGR, the inflectional affix is base-generated on the verb. According to Speas, the type of agreement base-generated on the verb is called 'weak' agreement. So, when the head AGR position is empty, the Economy of Projection Principle requires that at least (i.e. minimally) the Spec-AGRP be filled either by moving an NP to it or by inserting a pleonastic in that position. That is how the subject position is filled and pro is not allowed. Figure 9 below displays the schematic representation of the 'weak agreement' phenomenon in non-pro-drop languages:

Figure 9: Generation of inflectional affix on the verb in non-pro-drop languages

Now that the pertinent differences between languages with 'strong' and 'weak' agreement have been clarified, the question remains as to how languages without agreement allow for null subjects. As mentioned previously, it is necessary for languages with strong or weak agreement to project AGRP, no matter how residual their agreement morphology is; but the idea for languages which lack agreement morphology is that AGRP is not projected at all, since agreement is missing altogether. Given that AGRP does not have to be licensed if it is not projected, the subject (Spec-AGRP) is not required to be overt in such languages and Spec-TP is the place for overt (i.e. NP) and null subjects. The schematic representation of the head-final languages that lack agreement morphology is shown in Figure 10 in the following:

Figure 10: No AGRP projection in head-final languages without agreement morphology

In sum, the crosslinguistic differences among the three groups of languages mentioned above is as follows:

a. A language is [+ pro-drop] if AGR affix is base-generated in head AGR position (e.g. Persian, Spanish, Italian)

b. A language is [– pro-drop] if AGR affix is base-generated on the verb (e.g. English, German, Swedish)

c. A language is [+ pro-drop] if it has no AGR at all (e.g. Japanese, Chinese, Korean, Malayalam).

2.2.2 Rohrbacher's Full Paradigm Condition (1992, 1994, 1999)

In the Minimalist Programme (Chomsky 1995), it is assumed that the V-to-AGR movement (also known as V-to-I movement or verb-raising) is driven by the strong features of I, which need to be checked by the movement of V (Radford 1997, 2004). It is observed that most pro-drop languages exhibit the V-to-AGR movement. However, contrary to syntactically driven accounts, Rohrbacher (1994, 1999) and Vikner (1995, 1997) hypothesise that the V-to-AGR movement is morphologically driven so that "rich morphological paradigms trigger strong feature values, while impoverished morphology results in weak feature strength" (White 2003:160). Bobaljik (2002) calls this "Rich Agreement Hypothesis".

Like Speas (1994), Rohrbacher (1992, 1994, 1999) maintains that in languages with rich agreement, affixes have independent lexical entries while in languages that lack agreement or have weak agreement, affixes are base-generated on the verb. However, unlike Speas, who is primarily concerned with constructing a theory for null subject, Rohrbacher's work is principally concerned with characterising the morphological trigger for V-to-AGR movement, and only secondarily concerned with null subjects. However, he concludes that the conditions under which AGR is 'strong enough' to trigger V-to-AGR movement are the same ones as those in which AGR is 'strong enough' to license referential null subjects. He suggests that in null subject languages AGR has its own lexical entry, and hence both triggers V-to-AGR movement and can in principle license null subjects if and only if it has a 'Full Paradigm'. He defines 'Full Paradigm Condition' as follows:

INFL is a referential category with lexically listed affixes in exactly those languages where regular subject-verb agreement minimally distinctively marks all referential INFL-features such that a. and b.: a. In at least one number and one tense, the person features [1st] and [2nd] are distinctively marked. b. In at least one person of one tense, the number feature [singular] is distinctively marked. (Rohrbacher 1994:118)

Simply put, unlike Jaeggli and Safir's MUP (1989), the Full Paradigm system does not require every cell in a full paradigm to have an overt agreement affix. Instead, a Full Paradigm minimally distinctively marks the first and second person features in at least one number of one tense. The crux of Rohrbacher's proposal is that in languages that meet the 'Full Paradigm Condition', the agreement affixes of such a paradigm each have individual lexical entries. According to Speas (1994), this property permits these affixes to project their own AGR phrase, thus providing the AGR head with content and this in turn allows the Spec-AGRP to remain empty, leading to the possibility of null subjects.

Nevertheless, as mentioned above, Rohrbacher is not primarily concerned with the licensing of null subjects via overt agreement morphology but his aim is to capture the intuition that rich subject-verb agreement is responsible for V-to-AGR raising. He observes that languages with V-to-AGR movement always have minimal distinctive marking of the first and second person features, which is consistent with his requirement on Full Paradigms licensing null subjects. Consequently, he assumes that if a paradigm is sufficiently rich its referential agreement affixes are inserted under AGR, which gives content to AGRP, (leading to referential pro subjects) and this in turn triggers movement of the verbal stem to AGR in order to join the verb to the stranded affix that is base-generated in AGR.

Belletti (1990) makes such a claim for Italian. In languages where the agreement affixes are base-generated in AGR, verb movement seems to be a straightforward way to join the verb with its agreement affix. As such, it provides an explanation for the crosslinguistic observation that null subject languages with rich verbal agreement exhibit verb-raising to AGR.

Thus by coupling Speas' (1994) Economy of Projections Principle with his own Full Paradigm Condition, Rohrbacher (1999) arrives at an agreement-based parameter which unifies null subjects and V-to-AGR movement. This parameter predicts the following crosslinguistic differences. If a language meets the Full Paradigm Condition, its agreement affixes are lexically listed, resulting in pro-drop and verb-raising. On the other hand, if a language does not meet the Full Paradigm Condition, its agreement affixes are base-generated on the verb and neither null subjects are licensed, nor is V-to-AGR raising triggered.

Although Rohrbacher's Full Paradigm Condition overcomes some of the fatal empirical flaws of the Morphological Uniformity Principle (MUP), there are a few problems with his approach. Three problems of this account are investigated in the following:

Firstly, based on the Full Paradigm analysis, in L₁ acquisition of verb-raising languages, it is expected that the emergence of verbal agreement results in setting the feature strength to [strong] and subsequently setting strong features is followed by the emergence of verb-raising. In fact, the acquisition of verb-raising in the absence of rich agreement morphology is deemed impossible (White 2003). However, Verrips and Weissenborn (1992) show that German and French children acquire verb-raising well before the full agreement paradigm is being used.

Seconly, White (2003) argues that while it seems to be the case that languages with rich agreement have verb-raising, it is not the case that languages with impoverished agreement morphology necessarily lack verb-raising. There are some languages which allow V-to-AGR movement while having weak morphological paradigms, for example, Afrikaans
(du Plessis et al. 1987), a dialect of Swedish spoken in Finland (Platzack & Holmberg 1989), Capeverdian Creole (Baptista 1997) and Middle Danish (Vikner 1997).

Thirdly, Speas (1995) contests that there is a logical equivalence between verb-raising and null subjects. She notes that, although the generalisation that languages that permit null subjects also have V-to-I movement seems to be correct, the reverse is not true. In other words, she objects to Rohrbacher's prediction that languages with V-to-AGR movement necessarily permit null subjects. There do seem to be languages that have V-to-AGR movement yet do not allow null subjects. For example, French is not normally described as allowing null subjects but clearly has V-to-I movement. In addition, Yiddish and German are considered to have rich agreement but allow only pleonastics to be null.

To avoid the above-mentioned pitfalls, Speas (1995) asserts that verb movement should be dissociated from a rich agreement paradigm. More importantly, she contends that verb movement should be dissociated from the licensing of null subjects.

2.2.3 Radford's feature-checking account (1997, 2004)

As is inferred from the previous account, Rohrbacher (1992, 1994, 1999) and Vikner (1995, 1997) consider rich agreement as the licensing condition for the missing subjects in pro-drop languages. Furthermore, they relate V-to-I movement to rich agreement morphology. In fact, they analyse the syntactic phenomenon of V-to-I movement as a morphologically driven one but some have challenged the correlation between verb-raising and inflectional morphology (Speas 1995; Bobaljik 2002; White 2003; Bentzen 2003). Contrary to morphologically driven accounts which are based on rich (i.e. strong) agreement, there are some accounts which introduce syntactic arguments for V-to-I movement in pro-drop languages.

Radford (1997, 2004) considers V-to-I (or V-to-T) movement as a syntactic phenomenon which is based on checking theory. The languages which have rich agreement morphology (e.g. Persian, Spanish) are considered to have strong agreement features and in turn allow null subjects, while languages which have weak agreement morphology (e.g. English) are supposed to have weak agreement features and do not allow null subjects. Radford (1997, 2004) explains the two types of languages through some systematic examples.

Considering the 'successive cyclic fashion' of head movement, Radford (1997) puts forth the example below to show that the verbs in the Early Modern English (EME, also known as Elizabethan English or Shakespearean English) move from the head V position of VP to the head I position of IP and in turn to the head C position of CP:

(23) [_CP Know_i [_IP you t_i [_VP not t_i [_DP the cause]? (Radford 1997: 118)

The corresponding schematic representation of the above sentence indicates V-to-I movement through the 'successive cyclic fashion' of movement:

Figure 11: V-to-I movement in Early Modern English (EME) (Radford 1997: 118)

Then the question arises as to the reason underlying the verb-raising phenomenon in EME. Also it is worth further investigation into the origin of the difference between Early Modern English (EME) and Modern Standard English (MSE). It is claimed that the difference between the two originates in Chomsky's 'strength' metaphor. Using the strength metaphor, Radford (1997: 118) argues that the finite verb know in the above example carries strong agreement features and it is obvious that features are checked off through head-Spec relationship. So, to check the strong AGR features of Spec-IP you, the verb moves from V to I. Furthermore, to make an interrogative sentence it moves from head I position of IP to C position in CP.

Thus, the difference between Early Modern English (EME) and Modern Standard English (MSE) is considered to be due to the strength of features; in Early Modern English (EME) the verbs carry strong AGR features, hence movement is triggered. However, in Modern Standard English (MSE) it is not the case because verbs carry weak AGR features. Investigating the determining factor for 'strong' or 'weak' AGR features, Radford puts:

A plausible answer is that this is correlated with the relative richness of the agreement inflections carried by the finite verbs, in that finite verbs have strong agreement-features in languages in which they carry rich agreement inflections, and weak agreement-features in languages in which they carry impoverished agreement inflections. (Radford 1997: 119)

To consider the truth of the correlation between strong AGR features of verbs and rich agreement inflection, he examines the inflectional system of the Early Modern English (EME). In Shakespearian English, as a sample of Early Modern English, there are three present tense inflections: second person singular +st (e.g. Thou seest how diligent I am) and third person singular +th (e.g. She taketh most delight in music) and +s (e.g. It looks ill, it eats drily). However, in Modern Standard English, these inflections have reduced to just one for the third person singular +s.

Therefore, it is seen that rich agreement inflection results in strong agreement features and in turn strong agreement features induce movement and this serves to V-to-I movement. On the other hand, weak agreement inflection is correlated with weak agreement features and weak agreement features apparently cannot trigger V-to-I movement.

3 Conclusion

While most of the GB-based accounts of the pro-drop parameter depend on the relationship between rich agreement morphology and licensing pro-drop, MP-based accounts view parametric variation as a result of functional categories and their features. In fact, MP accounts mainly focus on checking 'nominal' or 'verbal' features of 'subjects' and 'verbs' through moving them out of their base-generated position or forcing them to remain in situ.

Dealing with subject movement, Herschensohn (2000) discusses that in non-pro-drop languages, the Tense (T) of the clause has strong nominal features (F_N), so absorbing the nominal features (F_N) of the subject, it has to move out of its base-generated position in Spec-VP to the Spec-AgrSP to check the nominal features (F_N) of the Tense (T). On the other hand, she maintains, in pro-drop languages the nominal features (F_N) of the Tense (T) are weak, so they do not cause the subject to move out of the Spec-VP and the subject remains in situ, raising covertly at LF (i.e. it does not have any phonetic realisation). This analysis can illustrate the distinction between the two major types of languages with reference to the pro-drop parameter:

Concerning verb movement in pro-drop and non-pro-drop languages, Radford (2004) argues that in pro-drop languages, the finite main verb moves from the head V position of VP into head T position of TP to check the strong verbal features (F_V) of Tense (T) in the head T position of TP. However, in non-pro-drop languages, due to weak verbal features (F_V), the verb does not raise to head T position of TP. In such languages, the verb raises covertly to T at LF. So, it is obvious that MP accounts are mainly based on feature checking analysis of subjects and verbs. Whereas Pro-drop languages have weak F_N and strong F_V, non-pro-drop languages have strong F_N and weak F_V.

In sum, taking both GB and MP accounts of the pro-drop parameter into account, it is concluded that "the difference between pro-drop and non-pro-drop languages is then a choice concerning the nature of AGR" (Cook 1993: 163): while GB accounts scrutinise the correlation between rich/weak agreement morphology and pro-drop/non-pro-drop languages, MP accounts analyse this parameter in terms of weak/strong nominal features (F_N) of Tense (T). Presumably, it is due to the leading role of Agr in the investigation of syntactic structures, pro-drop parameter included, that makes most syntacticians claim that the syntax of agreement constitutes the cornerstone of current syntactic argumentation, whether you believe it or not, whether you are concerned about it or not, and whether you are an "AgrOphile" or "AgrOphobe" (Radford 1997).

References

Alexiadou, Artemis & Anagnostopolou, Elena (1998): "Parametrizing Agr: Word order, V-movement and EPP checking", in: Natural Language and Linguistic Theory, 16, 491–539.

Baptista, Marlyse (1997): The morpho-syntax of nominal and verbal categories in Capeverdian Creole, Unpublished PhD dissertation, Harvard University.

Belletti, Adriana (1990): Generalized verb movement, aspects of verb syntax, Turin.

Bentzen, Kristine (2003): "V-to-I movement in the absence of morphological cues: Evidence from adult and child Northern Norwegian", in: Proceedings of the 19th Scandinavian Conference of Linguistics 31/3, 573–588.

Bobaljik, Johathan D. (2002): "Realizing Germanic Inflection: Why Morphology Does Not Drive Syntax", in: Journal of comparative Germanic Linguistics 6/2, 129–167.

Boeckx, Cedric (2000): EPP eliminated, Ms., University of Connecticut.

Boškovič, Željko (2001): A-movement and the EPP, Ms., University of Connecticut.

Bouchard, Denis (1984): On the Content of Empty Categories, Dordrecht.

Castillo, Juan Carlos / Drury, John / Grohmann, Kleanthes K. (1999): "Merge Over Move and the Extended Projection Principle ", in: Aoshima, Sachiko, Drury, John, Neuvonen, Tuomo (eds.), University of Maryland Working Papers in Linguistics 8, 63–103.

Chomsky, Noam (1981a): Lectures on Government and Binding, Dordrecht.

Chomsky, Noam (1981b): "Principles and parameters in syntactic theory", in: Hornstein, Norbert, Lightfoot, David (eds.), Explanations in Linguistics, London.

Chomsky, Noam (1982). Some Concepts and Consequences of the Theory of Government and Binding. Cambridge, Mass.

Chomsky, Noam (1986a). Knowledge of Language: Its Nature, Origin and Use. New York.

Chomsky, Noam (1986b): Barriers, Cambridge, Mass.

Chomsky, Noam (1988). Language and Problems of Knowledge: The Managua Lectures. Cambridge, Mass.

Chomsky, Noam (1991): "Some notes on economy of derivation and representation", in: Freidin, Robert (ed.), Principles and parameters in comparative grammar, Cambridge, Mass., 417–454.

Chomsky, Noam (1995): The Minimalist Program. Cambridge, Mass.

Chomsky, Noam (1998): "Minimalist Inquiries: The Framework", in: MIT Occasional Working Papers in Linguistics 15.

Chomsky, Noam (2001): "Derivation by phase", in: Kenstowicz, Michael (ed.), Ken Hale. A Life in Language, Cambridge, Mass., 1–52.

Cook, Vivian J. (1993): Linguistics and Second Language Acquisition, Basingstoke.

Cook, Vivian J. & Newson, Mark (1996): Chomsky's Universal Grammar: An Introduction, Oxford [Second Edition].

Deal, Amy R. (2005a): Pro-drop, topic-drop, and the functional lexicon: A constructional account of null arguments, Unpublished MA thesis, Brandeis University.

Deal, Amy R. (2005b): "Pro-drop and subject (non-)recoverability: The case of Nez Perce, in: Generative Grammar @ Geneva 4, 93–111. [Retrieved August 5, 2007, from http://www.unige.ch/lettres/linge/syntaxe/journal/Volume4/Deal_NzPprodrop200522215303.pdf]

du Plessis, Jean / Solin, Doreen / Travis, Lisa / White, Lydia (1987): "UG or not UG, that is the question: A reply to Clahsen and Muysken", in: Second Language Research 3/1, 56–75.

Epstein, Samuel D. / Seely, T. Daniel (1999): SPEC-ifying the GF "subject"; eliminating A chains and the EPP within a derivational model, Ms., University of Michigan and Eastern Michigan State University.

Grohmann, Kleanthes K. / Drury, John / Castillo, Juan Carlos (2000): "No more EPP", in: Proceedings of the Nineteenth West Coast Conference on Formal Linguistics, Somerville, Mass., 139–152.

Haegeman, Liliane (1991): Introduction to Government and Binding Theory. Oxford.

Hawkins, Roger (2001): Second Language Syntax, Oxford.

Herschensohn, Julia (2000): The Second Time Around: Minimalism and L2 Acquisition, Amsterdam.

Hilles, Sharon (1991): "Access to universal grammar in second language acquisition", in: Eubank, Lynn (ed.), Point counterpoint: Universal Grammar in the second language, Amsterdam, 305–338.

Huang, C.-T. James (1984): "On the distribution and reference of empty pronouns", in: Linguistic Inquiry 15, 531–574.

Huang, C.-T. James (1989): "Pro-Drop in Chinese: A Generalized Control Theory", in: Jaeggli, Oswaldo, Safir, Kenneth (eds.), The Null Subject Parameter, Dordrecht, 185–214.

Hyams, Nina (1986): Language Acquisition and the Theory of Parameters, Dordrecht.

Hyams, Nina (1987): "The theory of parameters and syntactic development", in: Roeper, Thomas, Williams, Edwin (eds.), Parameter Setting, Dordrecht.

Hyams, Nina (1992): "A Reanalysis of Null Subjects in Child Language", in: Weissenborn, Jürgen, Goodluck, Helen, Roeper, Thomas (eds.), Theoretical Issues in Language Acquisition, Hillsdale, New Jersey, 249–269.

Jaeggli, Oswaldo (1982): Topics in Romance syntax, Dordrecht.

Jaeggli, Oswaldo & Safir, Kenneth (1989). The null subject parameter and parametric theory", in: Jaeggli, Oswaldo, Safir, Kenneth (eds.), The null subject parameter, Dordrecht, 1–44.

Lakshmanan, Usha (1991): "Morphological Uniformity and null subjects in child second language acquisition", in: Eubank, Lynn (ed.), Point Counterpoint: Universal Grammar in the Second Language, Amsterdam, 389–410.

Manzini, Maria R. (1983): "On control and control theory", in: Linguistic Inquiry 14, 421–446.

Martin, Roger (1999): "Case, the Extended Projection Principle, and minimalism", in: Epstein, Samuel, Hornstein, Norbert (eds.), Working minimalism (pp. 1-25) .Cambridge, Mass., 1–25.

Meisel, Jürgen (1991): "Principles of Universal Grammar and strategies of language use: Some similarities and differences between first and second language acquisition", in: Eubank, Lynn (ed.), Point counterpoint: Universal Grammar in the second language, Amsterdam, 231–276.

Mohanan, K Karuvannur P. (1983). Functional and Anaphoric Control. Linguistic Inquiry, 14, 641-674.

Nishigauchi, Taisuke (1984): "Control and the thematic domain", in: Language 60, 215–250.

Ouhalla, Jamal (1999): Introducing Transformational grammar: From Principles and Parameters to Minimalism, London.

Platzack, Christer (1987): "The Scandinavian languages and the null-subject parameter", in: Natural Language and Linguistic Theory 5, 377–402.

Platzack, Christer & Holmberg, Anders (1989): "The Role of AGR and Finiteness", in: Working Papers in Scandinavian Syntax 43, 51–76.

Radford, Andrew (1997): Syntax: A Minimalist Approach, Cambridge.

Radford, Andrew (2004): Minimalist Syntax: Exploring the Structure of English, Cambridge.

Rizzi, Luigi (1982): Issues in Italian syntax, Dordrecht.

Rizzi, Luigi (1986): "Null objects in Italian and the theory of pro", in: Linguistic Inquiry 17, 501–557.

Rizzi, Luigi (1990): Relativized Minimality, Cambridge, Mass.

Roberge, Yves (1990): The syntactic recoverability of null arguments, Buffalo, NY.

Rohrbacher, Bernhard (1992): "English AUX^NEG, mainland Scandinavian NEG^AUX and the theory of V to I raising", in: Proceedings of the 22nd Western Conference on Linguistics (WECOL92).

Rohrbacher, Bernhard (1994): The Germanic VO languages and the full paradigm: A theory of V to I raising, PhD dissertation, University of Massachusetts, Amherst.

Rohrbacher, Bernhard (1999): Morphology-driven syntax: A theory of V to I raising and pro-drop, Amsterdam.

Rothstein, Susan (1983): The Syntactic Forms of Predication, PhD. dissertation, MIT.

Shlonsky, Ur (1989): Null and displaced subjects, Bloomington. [Revised version of 1987 MIT dissertation.]

Sigurðsson, Halldór A. (1993): "Argument drop in Old Icelandic", in: Lingua 89, 247–280.

Speas, Margaret (1994): "Null arguments in a theory of economy of projection", in: North East Linguistics Society 24,179–208.

Speas, Margaret (1995): Economy, agreement and the representation of null arguments, Ms., University of Massachusetts, Amherst.

Taraldsen, Knut T. (1978): On the NIC, vacuous application, and the that-trace filter, Bloomington.

Verrips, Maaike & Weissenborn, Jürgen (1992): "Verb movement in early German and French: The independence of finiteness and agreement", in: Meisel, Jürgen (ed.), The acquisition of verb placement: functional categories and V2 phenomena in language acquisition, Dordrecht, 283–331.

Vikner, Sten (1995): Verb movement and expletive subjects in the Germanic languages, Oxford.

Vikner, Sten (1997): "V-to-I movement and inflection for persons in all tenses", in: Haegeman, Liliane (ed.), The new comparative syntax, London, 189–213.

White, Lydia (2003): Second Language Acquisition and Universal Grammar, Cambridge.

Zagona, Karen (1982): Government and Proper Government of verbal Projections, PhD dissertation, University of Washington/Seattle.