The principles of UG allow parameters of variation between languages.
Differences in a single parameter can lead to a number of surface syntactic consequences. It is believed that pro-drop languages display a cluster of syntactic and morphological properties related to a single parameter; that is, missing subjects, subject-verb inversion and that-trace effects. To further account for the cross-linguistic distribution of null subjects, Chomsky (1982), following Perlmutter (1971), Taraldsen (1978), Jaeggli (1982), and Rizzi (1982), proposed the Null Subject Parameter, which specifies the degree to which languages allow the omission of ponominals in the subject position along with a cluster of syntactic properties. The distribution of null subjects, determined by a process called recoverability (Taraldsen,
1978) or identification (Jaeggli, 1982), connects null subjects with rich verbal agreement paradigm. This generalization captures the difference between [+null subject] languages, such as Italian and Spanish, and [-null subject] languages, such as English and French. According to this theory, Italian contains a rich system of agreement, which can recover the features of person, number, and gender. On the other hand, English doesn’t have a rich agreement system so the AGR is too meager to identify the content of null subject. See the example (10), Italian and Spanish allow subjects to drop from finite clauses, but English and French do not.
(10) a. [e] parla. (Italian) b. [e] habla. (Spanish) c. *[e] parle. (French) d. *[e] speaks. (English)
However, Rizzi (1986) found that there are contexts where pro appears when identification through rich agreement is not possible. For example, the Italian sentence (11b), in which the object is empty raises the problem of identification. Rizzi argues that Italian, but not English, allows pro to occur in object position after certain types of control verbs, which is interpreted as arbitrary in reference.
(11) a. This leads *(people) [PRO to conclude what follows].
b. Questo conduce (la gente) a [PRO concludere quanto segue].
(Rizzi, 1986: 503) In this respect, Rizzi (1986) proposed ‘pro module’, assuming that pro must be both licensed and identified. It is different from the standard assumptions on the theory of pro in which formal licensing and recovery of the content are unified, both being performed through government by ‘strong AGR.’
(12) Rizzi’s ‘pro module’
a. Formal Licensing schema pro is Cased-marked by X0.
b. Recovery convention
Let X be the licensing head of an occurrence of pro: then pro has the grammatical specification of the features on X coindexed with it.
Licensing of pro involves an appropriate licensing head (Infl), which governs the pro in subject position through specifier-head agreement. For instance, in Italian-type languages, Infl is a licensing head, but not in English-type languages. Therefore, if a head is specified positively for the licensing parameter, pro is formally licensed whenever it is assigned Case by the head. One language may choose Infl as a licenser for pro in the subject position and V for pro in the object position (see (13) for the formulation); another language may prohibit licensing altogether.
(Rizzi, 1986:521)
Thus, the class of licensing heads may vary cross-linguistically, from permitting every head to be a possible licenser to allowing no licensing heads. The principle of identification is responsible for recovering referential value of pro. The licensing head is able to identify the content of pro so the features associated with the agreement are able to recover the content of pro. To be specific, licensing by Infl and identification by the agreement morphology on the verb are necessary for pro to occur. However, one point should be noticed that licensing involves verbal features, which applies to all types of null subjects, while identification involves nominal features, which is a requirement imposed on referential null subjects only.
Following Rizzi’s formulation of pro-drop parameter, Hyams (1986) proposed a version of the pro-drop parameter called AG/PRO parameter. Based on the
(13) proi Infl V proj
| |
Agri θj
Government and Binding Principle presented by Chomksy (1981), she claims that in pro-drop languages AG is equal to PRO and cannot be governed; hence AUX must be
lexically empty, for otherwise PRO would be governed (see 14a); in non-pro-drop languages AG is not equal to PRO; therefore, lexical material, such as modals and auxiliaries, can appear in AUX (see 14b).
(14) INFL
a. X：AG＝PRO；Y：(AUX)
X Y b. X：AG≠PRO；Y：AUX
Both Rizzi’s and Hyams’ analyses under the null subjects parameter are still based on the relationship between null subjects and rich agreements. However, the claim that pro-drop languages count on overt AGR to recover the null subjects was challenged by Huang (1984, 1989) since languages without agreement features, like Chinese, can also allow null subjects in finite clauses. This shows that rich agreement is not necessary for a language to permit pro-drop. In this respect, Huang (1989) addresses this problem to suggest that other than AGR identification, antecedent identification be also a way to meet the identification requirement in languages without AGR. He posits a generalized control rule, outlined in (15) with the notion of control domain in (16) defined in Manzini (1983) and Nishigauchi (1984), to explain the behaviors of pro in all languages:
(15) Generalized Control Rule (GCR)
An empty pronominal is controlled in its control domain (if it has one).
(16) 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.
b. α contains a SUBJECT accessible to β.
The rule specifies “the environment in which pro/PRO (he treats both the same) must have a local, unique, non-arbitrary antecedent (Huang, 1989: 194),” eliminating the need for PRO Theorem (Chomsky, 1981) which agued that PRO must be ungoverned according to Binding theory. In this rule, the presence or absence of SUBJECT/AGR determines the control domain for pro. If SUBJECT/AGR is present, pro must be controlled within the minimal S containing it, like Italian and English. However, the two languages are distinguished as to whether AGR can be a controller or not. As mentioned earlier, AGR in Italian is strong enough to determine the content of pro, while in English AGR is too weak to control pro. Huang also proves that languages in which the AGR is absent, such as Chinese, a pro/PRO is allowed if it has no control domain at all (Huang, 1989). Apart from GCR, there is another analysis that tried to account for the occurrence of null subjects in both languages with rich inflectional systems and those without an inflectional system, which will be discussed in the next section.
2.2.2 Morphological Uniformity Principle (MUP)
In order to account for the diversity of inflectional system related to the recoverability of null subjects in different languages, Jaeggli and Safir (1989) argued that the idea of using rich agreement to encompass all null-subject languages is problematic and need to be examined. Therefore, they proposed Morphological Uniformity Principle (MUP) as a universal principle to demonstrate the pro-drop phenomenon. They stated that “null subjects are permitted in all and only languages with morphologically uniform inflectional paradigms (1989: 29).” A morphological paradigm is uniform if all its verb forms are morphologically complex or none of them are. If the paradigm is mixed, that is, some of the forms are morphologically complex while others are bare stem, then it is not uniform. The MUP is stated in (17)
below.
(17) Morphological Uniformity Principle
An inflectional paradigm P for a category K in a language L is morphological uniform iff L has either only underived inflectional K-forms or only derived inflectional K-forms.
Italian-like languages, for example, have all their verbs inflected for number, person, tense and mood; therefore, they allow null subjects due to the feature of [+uniform +inflected]. Similarly, the inflectional systems in Japanese-type languages are also regarded as [+uniform] despite the fact that they are different from Italian-type languages in that Japanese verbs are inflected for tense, mood, aspect and negation. Chinese-type languages can also allow null subjects since they have no inflection affixation at all and uninflected verb stem is the only possible form. As a result, they are considered to have the feature of [+uniform -inflected]. According to MUP, null subjects are said to be licensed in these [+uniform] languages. On the other hand, English-type languages are non null-subject languages since they have morphologically mixed verb paradigms, which receive the feature of [-uniform].
However, some languages have morphologically uniform paradigms but partially allow null subjects (e.g. German), and still some languages have mixed morphological paradigms but do allow pro-drop (e.g. Bengali). Therefore, Jaeggli and Safir (1989), following the θ-criterion (Chomsky, 1981:36), proposed the identification requirement, which states exactly when null subjects will occur in a language, as defined in (18).
(18) Identification Requirement
A thematic null subject must be identified.
According to (17) and (18), failure to satisfy either of the two necessary and
sufficient conditions, namely the licensing condition of morphological uniformity and the identification requirement, will result in the prohibition of null subjects in a language. Jaeggli and Safir (1989) further proposed two types of identification for thematic null subjects. The first type of identification is via AGR, which applies to null subject languages with rich agreement, like Italian and Spanish. The second type of identification obtains in languages such as Chinese and Japanese which uniformly lack person-number agreement2. In these languages, null subjects can be identified not merely by an overt c-commanding nominal but also by a topic or possibly null topic.
Their analysis derives from Huang (1984) who has suggested the difference between
‘discourse-oriented’ languages (e.g. Chinese and Japanese) and ‘sentence-oriented’
languages (e.g. English and French). As mentioned earlier, Huang, following Taso (1977), suggested that the discourse-oriented languages have a rule of Topic NP deletion, “which operates across discourse to delete the topic of a sentence under
identity with a topic in a preceding sentence. The result of such a deleting process is formally a Topic Chain (1984: 549)”. Based on the generalization, Jaeggli and Safir (1989) proposed three mechanisms for the identification of null arguments:
(19) a. by local AGR, including a tense feature (e.g. Italian) [s proi [infl Agi/Tense]…]
b. by a Topic (e.g. Chinese)
Discourse Topici [topici [s ti…]]
c. by a c-commanding NP (e.g. Chinese) Subjecti verb [s proi VP]
2.2.3 The Null Argument Parameters
Following Huang’s (1984) argument, Lillo-Martin (1991) and Wang et al. (1992)
2 AGR features must include person markers as they are only real identifiers (Borer, 1983; Gilligan, 1987; Jaeggli & Safir, 1989; Rizzi, 1982).
postulated two separate parameters to account for cross-linguistic variation with respect to the occurrence and distribution of null arguments. The first parameter is the Null Pronoun Parameter (NPP), which basically follows Rizzi’s (1986) null pronoun parameter with some modification. This parameter concerns null pronominal arguments (i.e. pro). Just as Rizzi’s ‘pro module’, null arguments may require licensing and identification. The first part of NPP is the Null Pronoun Licensing Parameter which requires the language learner to determine independently for each case-assigning node whether it licenses pro (see Rizzi, 1986). The second part of NPP is the Null Pronoun Identification Parameter, which accounts for cross-linguistic variation in the extent to which the reference of pro is identified by morphological marking on the licenser. The two parts of NPP are given in (20) and (21).
(20) The Null Pronoun Licensing Parameter
a. AGR when it Case-marks [is/is not] a licensing head for pro b. V when it Case-marks [is/is not] a licensing head for pro c. P when it Case-marks [is/is not] a licensing head for pro
(21) The Null Pronoun Identification Parameter
I. Let X be the licensing head of an occurrence of pro:
Then pro has the grammatical specification of no features
the number features on X coindexed with it.
the person and number features
II. Null referential pronominal arguments require Identification by morphological marking of
no features
the number features on the licensing head.
the person and number features
(Lillo-Martin, 1994: 302)
The second parameter, based on Huang (1984), is the Discourse Oriented Parameter (DOP), which states that languages vary according to whether empty topics is coindexed with a preceding discourse topic. Discourse-oriented languages, such as Chinese, allow null arguments without rich agreement system; in these languages, the reference of the null argument is determined by a discourse topic. Besides, it should be noted that the DOP also permits languages with discourse-oriented properties to have both null subjects and null objects. Though the Generalized Control Rule predicts that null arguments in object position are not allowed in Chinese (also in English and Italian), Huang (1984, 1989, 1991) shows the prediction is workable in Chinese. Chinese does allow a null object; however, such a null subject “should be analyzed, not as a pro, but as a variable bound by a null topic, or more generally, by a null operator,” given that “α is a variable if and only if it is locally A’-bound and in an A-position” (Chomsky, 1981: 330).
(22) [TOP ei], [Zhangsan shuo [Lisi bu renshi ti]]
Zhangsan say Lisi not know
‘Zhangsan said that Lisi does not know *[him/her/them/you…]’
(Huang, 1984: 542) Huang (1984) further states that while a null object is a variable, a null subject can be either a variable or a pro, depending on where the null subject occurs. To be specific, when the null subject is c-commanded by the matrix antecedent, it can be a pro, while in other contexts, null subjects are variables bound by zero-topics.
2.2.4 UG and Second Language Acquisition
Universal Grammar (UG), proposed by Chomsky (1986), suggests a set of rules intended to explain language acquisition in L1. UG consists of principles, which are invariant constraints on all human languages, and of parameters, which permits a
limited degree of variations among languages and the various parameters of UG are triggered according to the input data from the L1. In other words, UG constitutes the child’s initial state, the knowledge that the child is equipped with in advance of input.
However, whether the adult L2 learners, like the L1 learners, have access to the principles of UG has been a controversial issue. Among the issues of the availability of UG in L2A, there are three theoretical options: (i) ‘Full Access to UG’, (ii) ‘Partial Access to UG’ and (iii) ‘No Access to UG’ (summarized by Cook, 1988, 1993;
Epstein et al., 1996; Lakshmanan, 1994; White, 1989), summarized in Table 2.1. The first position is the Full Access to UG, which involves two conditions. On the one hand, UG is available in L2A and works exactly as it does in L1A, rejecting the theory of critical age (Full Access No Transfer). On the other hand, L2 learners initially transfer the setting of the L1, but are still able to tap UG and parameter resetting is possible (Full Access Full Transfer). As for the Partial Access position, parameter resetting is impossible: UG is inaccessible but any aspects of it available in L1 can be used in the L2. The third claim states L2 learners cannot have an access to UG but utilize ‘acquisition strategies’ (Clahsen, 1988); therefore, ‘wild grammars’ (impossible errors) can occur.
Table 2.1 Alternative hypotheses regarding access to UG in L2A
Hypotheses Learning Process
Partial Access L2 input L1-based UG IL grammar
No Access
parameter-setting model of pro-drop with two settings [+pro-drop] or [-pro-drop]
([+pd]/[-pd]) (Adapted from Braidi, 1999: 63-64).
Table 2.2 Parameter model of acquisition of pro-drop by SLA Full Access No Transfer
If L2 learners have a full access to UG, two possibilities will arise: Firstly, we would expect L1 and L2 developmental sequences to be identical; secondly, UG is accessible but learners initially transfer the setting of the L1, and the L1 still can be reset to the L2 parameter. If L2 learners have a partial access to UG, UG is accessible but only via the setting of the L1, and the L1 can only be reset to the L2 parameter by L1-based UG. If L2 learners have no access to UG, parameter resetting can not take place.