This dissertation explores the developmental pattern observed in Chinese EFL learners acquiring the three binding properties of English reflexives，c-commanding constraint，subject orientation and locality condition.

It is generally acknowledged that the c-commanding constraint is universal cross-linguistically. In both English and Chinese，reflexives are bound only by the DP which occupies the c-commanding position in the sentence configuration. Therefore，in English sentences as “［Annie_j's sister］_i told the doctor something about herself_i/*j” and their counterpart Chinese sentences，herself is coreferential with Annie’s sister，a c-commanding DP，but not coreferential with Annie，a non-c-commanding DP.

However，it is widely known that reflexive pronouns in English and Chinese differ with respect to subject orientation and locality condition. On the one hand，English reflexives can take local DPs as their antecedents and long-distance DPs are not legitimate antecedents whereas Chinese monomorphemic reflexive ‘ziji’ can be bound either to a local DP or to a long-distance（LD）DP. On the other hand，in English，DPs functioning as subject or object can be legitimate antecedents；in contrast，in Chinese，reflexives can take only subject as their antecedent and sentences with coreferential dependencies between reflexives and object are ruled out.

Given the differences in referential characteristics between English and Chinese，an interesting question arises in the case of Chinese learners' acquiring English reflexives：is it possible for them to acquire the coreference properties of English anaphors？There is positive evidence in the acquisition of the corefe-rence between reflexives and the nonsubject，but in the acquisition of locality requirement，the logical problem holds in adult second language acquisition（White 1989）. Furthermore，if binding properties are acquirable，how do they develop with the gain of English proficiency by Chinese EFL learners？

A truth-value judgment task is employed in this dissertation in that this methodology offers a relatively direct and efficient means of evaluating competence rather than preference. The test format is a story consisting of 2 to 5 sentences followed by a comment sentence. Participants are required to read the 40 stories and then indicate whether the subsequent comment sentence matches or mismatches the situation provided in the story by pressing either ‘y’ or ‘n’ key on the keyboard. The comment sentences following the stories fall into five sentence types to investigate the Chinese speakers' acquisition of three binding properties，c-command，orientation and locality，respectively.

To guarantee that participants are proficient enough to accomplish the experiment task，206 freshmen were selected from 6 natural classes of a top-ten key university in China and a non-key university. Three proficiency levels were set up according to Quick Placement Test administered to all the participants. The top 40 participants are considered as the high scoring group，the 40 in the middle as the mid group，and the 40 at the bottom as the low group. There are big differences among the three groups' mean proficiency scores，with means scores of 49.8，41.7，and 33.9，respectively. According to a one-way ANOVA，the differences among them are statistically significant（F（2，117）=495.6，p=.000），with post hoc Games Howell procedures showing significant differences between the high group and the other learner groups（p=.000），as well as between the low and mid groups（p=.000）.

The results of Mixed RM ANOVAs indicate that the largest three-way interaction between types of sentences，distance of antecedents，and proficiency is not statistical，using the Huynh-Feldt correct（F _{6.59，385.488}=0.556，p=0.781，partial eta-squared=0.009，power=0.234）. This means that participants from the different groups performed similarly on different types of sentences，and they also performed similarly in spite of the distance of antecedents. However，the two-way interaction between the sentence type and proficiency and between the sentence type and distance is statistically different. The difficulty order of the three distinct binding properties is that c-command is easier than locality and orientation for the low group whereas for the mid group and the high group c-command is easier than locality，which in turn is easier than orientation. When locality is instantiated in finite embedded clauses，nonfinite embedded clauses and monoclauses with possessed representational noun phrases（PRNP），the difficulty order is that the resolution of reflexives in PRNPs is more demanding than that in nonfinite clauses，which in turn is more demanding than that in finite clauses for the low group. For the mid group and the high group，the difficulty order is that the PRNP condition is more challenging than the biclausal condition. Of the five sentence types，Types 1 and 3 are the easiest while Type 5 is the most difficult for the low group；Types 1，3 and 4 are easier than Types 2 and 5 for the mid group and the high group. With discourse information taken into consideration，the difficulty order for the low group is 1A/1B/2B/3A/3B and 4A＜2A/4B/5A and 5B；for the mid group 1A/1B/2B/3A/3B and 4A＜4B/5A and 5B＜2A；for the high group 1A/1B/2B/3A/3B/4A and 4B＜5A and 5B＜2A.

There are four developmental patterns of each binding property and each condition of locality exhibited by the three proficiency groups. First，there is significant improvement both from the low-level group to the mid-level group and from the mid-level group to the high-level group. Second，there is no improvement either from the low group to the mid group，or from the mid-level group to the high-level group，or from the low group to the high group. Third，there is no improvement from the low group to the mid group，or from the mid group to the high group；however，there is improvement from the low group to the high group. Fourth，there is only significant improvement either from the low-level to the mid-level group or from the mid-level to the high-level group，but not both.

Though transfer can account for better acquisition of c-command than orientation and locality in the low group，it cannot explain the differences in the same binding property of locality under the three different conditions. The explanation by the extension of the standard binding theory is inadequate and unsatisfactory as well. Two proposals are suggested in the dissertation to account for the diffe-rences between binding properties and the three conditions of locality. It is argued that head category and completeness of phases play an important role in the interpretation of English reflexives，which give rise to the differences in the locality property. This is the first attempt that the phase theory is applied to SLA research.Alternatively，the involvement of distinct interface categories in identifying antecedents results in differences in the three conditions of locality. Additionally，the Interface Hypothesis makes an adequate account for the differences between the distinct binding properties，namely，c-command，orientation and locality. The results indicating that orientation is as demanding as locality implicate that interpretations of reflexives involving interfaces are all challenging no matter whether the interfaces are internal interfaces or external interfaces. This is the new extension of the Interface Hypothesis in that previous studies exclusively include advanced or near-native learners of a second language.

Key words：Reflexives，Principle A of the binding theory，Universal Grammar，the phase theory，the Interface Hypothesis