Excerpt
Table of contents
1 Introduction
2 Three approaches to reflexive binding
2.1 The Multivalued Governing Category Parameter Approach
2.1.1 A Study: Makiko Hirakawa 1990
2.2 The Movement in Logical Form Approach
2.2.1 A study: Christie and Lantolf 1998
2.3 The Relativized SUBJECT Approach
2.3.1 A Study: Bennett 1994
3 Discussion
4 Conclusion
5 References
1 Introduction
Within the discourse of generative grammar it has been proposed that Universal Grammar (UG) can be seen as an innate structure that governs the course of learning a language. The domain of UG is often also referred to as the “principles and parameters framework”. While principles can be seen as a set of rules (e.g. the linearity of language) that are applicable for all natural languages, certain parameters make up and have to be set according to different languages. As there is not much disagreement on the role of UG for the acquisition of a first language (L1), many researchers have discussed the question if and how UG constrains second language acquisition (L2A).
A large amount of research has been conducted within the field of the acquisition of reflexives in L2. Earlier approaches explained the acquisition of reflexives in L2A as a resetting of the Governing Category Parameter (GCP). Other approaches accounted the cross-linguistic variation in interpreting domain and orientation of anaphors to different categories of anaphors, namely X0 or head reflexives and XP or phrasal reflexives. It was then hypothesized that interpretation of these reflexives in L2A can either be the consequence of movement in Logical Form or a result of parameter settings regarding the agreement (AGR) parameter resulting in a relativized SUBJECT.
The aim of this paper is to give a short overview of the above mentioned approaches on the domain of reflexivity acquisition by presenting one study for each approach. The results of the different studies will be discussed with emphasis to the question to what extent they show evidence for the non-/availability of UG in L2A. Based on this, a suggestion is stated, that the relativized SUBJECT approach is most sufficient to explain UG-availability in terms of the Full Transfer/Full Access Hypothesis.
2 Three approaches to reflexive binding
2.1 The Multivalued Governing Category Parameter Approach
The Governing Category Parameter (GCP) as proposed by Wexler and Manzini (1987) was of particular interest in the works of linguists focussing on the presence or absence of UG during L2A dating from the late 1980s to the mid 1990s. It defines the domains in which reflexive pronouns must be bound or can be free by means of a mulit-valued parameter. They state that:
α is a governing category of β if α is the minimal category and
has a subject, or
has an INFL, or
has TNS, or
has an indicative TNS, or
has a root TNS (Wexler and Manzini, 1987)
White (1989) simplifies this model by taking only three values, rather speaking of nearest clause, nearest finite clause and main clause as the relevant GCP values for anaphors e.g. in English, Russian and Korean. These values are in subset/superset relationship to each other. In Korean, as the most inclusive (or marked) language, the anaphor can take its antecedent “in the main clause or in any clause embedded in the sentence” (158), while English in this case is the most restrictive (or unmarked) language forming the subset as is shown in Figure 1.
illustration not visible in this excerpt
Figure 1: Different settings of the GCP (White 1989)
In the Subset Principle, it is assumed that language learners start the acquisition with an unmarked setting (subset) first and move to the more marked grammars by positive evidence only.
2.1.1 Study: Makiko Hirakawa 1990
Hirakawa (1990) reports on her study she conducted to examine if learners with an L1 which is marked (Japanese) can successfully reset the GCP value to a target language that is unmarked (English). She hypothesizes four different outcomes. The grammar of her L2 learners could fulfill the Subset Principle by starting right away with the correct unmarked value for English, or they could transfer the Japanese setting to English, resulting in incorrect responses. Other possibilities would be that the learners end up at a GCP setting that is somewhere in between the L1 and L2 or that no UG principle or parameter is involved resulting in a unnatural grammar.
Methodology
The subjects of her study consisted of six groups, representing four test groups of L2 learners of English of different age and exposition to the target language and two control groups containing native speakers of Japanese and English. The mentioned exposition to English was through classroom teaching only, partly of native speakers of English. Students who had abroad experience in an English speaking country were rejected. Hirakawa tests the students’ behaviour on the interpretation of English reflexives with regard to GCP[1] by using five different types of sentences:
illustration not visible in this excerpt
Hirakawa (1990)
The subjects were asked to chose and indicate who the reflexives himself/herself referred to. Test items containing two Noun Phrases (NP) had five response options[2], while sentences with three NPs had nine choices[3]. The test was not timed, but students were asked to chose instantly and not go back to previous test items.
Results
Although correct answers (NP2 in Type A and C sentences; NP3 in Type B and D) are dominating, there is a good number of subjects choosing the wrong antecedent or suggesting ambiguity when there is no, particularly in three-clause sentences[4]. Hirakawa interprets this as evidence against a working Subset Principle. Furthermore she constitutes that her subjects wrongly transferred the widest GCP setting from their L1 to the L2.
With regard to two-clause structures, subjects did best on tensed sentences (Type A), choosing the correct local antecedent more often than on infinite clause sentences (Type C) where the number of non-local responses was higher. This suggests that there is an effect of [± tensed] on the interpretation. As there is a complexity effect with more mistakes happening in three-clause sentences than in two-clause sentences for finite constructions, no such effect can be seen for infinite constructions or when comparing complex embedded tensed sentences (Type B) with rather simple two-clause infinite constructions (Type C). Here the wrong non-local antecedent or ambiguity was taken more often.
All the results do not show any significant difference concerning age or grade effects, which is rather surprising, since it was assumed that older students would show a higher proficiency. Hirakawa explains this with a generally low level, because subjects received only classroom education and had no experiences in an English speaking country. An important observation however is, that 10 subjects did no mistakes at all and 8 subjects committed only one mistake throughout the test. Given these result, she suggests that these learners have been able to reset the GCP value to arrive at a correct L2 grammar.
Hirakawa discusses the possibility that this resetting could be in fact due to the application of a learning strategy like the Minimal Distance Principal of C. Chomsky (1969) rather than an effect of UG. But then students should have obeyed this principle on all test items, choosing always the closest NP to the reflexive. Since they did not on Type E sentence, which was of course correct, she assumes that resetting of the parameter is possible and UG-constrained.
[...]
[1] She also took the Proper Antecedent Parameter PAP into account, but for the limited space in this paper I only refer to the GCP.
[2] NP1, NP2, either NP1 or NP2, someone else or don’t know
[3] NP1, NP2, NP3, either NP1 or NP2, either NP2 or NP3, either NP3 or NP1, either NP 1 or NP2 or NP3, someone else or don’t know
[4] These results replicate the findings of earlier studies conducted on GCP by Thomas (1989) and Finer and Broselow (1986)