Previous Empirical Studies on the Three Universal Hypotheses

A productive paradigm of research on English RCs has been inspired by the three

universal hypotheses. Its focus is mainly on testing the validity of the three

theoretical claims by investigating whether the accuracy order of ESL/EFL learners’

performance of various RC types actually corresponds to the difficulty hierarchy as

predicted by the three. Generally speaking, studies on the three predictor hypotheses,

particularly the PFH and the PDH, are characterized by mixed results.

Notwithstanding preliminary support from his own study (1974), in which

significant differences in L1 children’s comprehension were observed between

English RCs with non- and parallel function to head NPs, Sheldon’s PFH has been

directly countered by evidence from L2 studies (e.g. Gass & Ard, 1980; Flanigan,

1994). Flanigan (1994) found that his L2 child subjects encountered even greater

difficulties with SS relativization than with OS in both comprehension and

sentence-combining tasks, thereby refuting the parallel function of heads and

relativized nouns as a relevant factor determining the difficulty order of English RCs

and suggesting that other factors such as center-embedding play a more important role

in learning relativization.

Likewise, Kuno’s (1974) PDH has not found consistent empirical support in the

SLA literature, albeit positive findings from, among others, Cook (1973), Ioup and

Kruse (1977), Schumann (1980), Kubota (1993), and Izumi (2003). Ioup (1983), for

example, used a different elicitation measure─a sentence-combining task─but failed

to replicate her earlier findings in 1977, which substantiated the PDH. Wei (1997)

also provides counter-evidence from her study, in which right-embedded RCs (OS/OO)

appeared to pose greater difficulties than did center-embedded ones (SS/SO) for L2

learners in comprehension.

As opposed to conflicting findings regarding the PFH and the PDH, L2 studies

on Keenan and Comrie’s (1977) NPAH have uniformly yielded positive results.

Table 4 on page 52 summarizes representative SLA studies on English RCs in terms

of their support for the three universal hypotheses. As indicated in Table 4, the

NPAH is supported by the majority of SLA research, followed by the PDH and then

the PFH.

As a valid predictor of the difficulty order of English relativization, the NPAH

also serves as an alternative explanation for L2 learners’ errors involving the use of

resumptive pronouns and their behavior of avoidance, both of which have long been

held as resulting from L1 interference. In view of the fact that pronoun copying

(resumption) is found in L2 learners whose L1 does not employ nominal reflexes in

RCs and even in native speakers of English, resumptive pronouns can be viewed as a

universal initial strategy with which to reduce processing difficulty²⁶ inherent in

structural complexity (Pavesi, 1986). The extent to which resumptive pronouns are

used is claimed to concur with the NPAH (Keenan & Comrie, 1977; Gass, 1979,1980;

Flagnigan, 1994). That is to say, the lower is a noun to relativize on the hierarchy,

the harder is it to process the resulting RC on that position, and the more likely is one

to employ an overt resumptive pronoun to make up for the increased processing

difficulty. The psychological validity of the NPAH also manifests itself in

26 Keenan (1988: 37) argues that resumptive pronouns facilitate processing of RCs because they allow the logical structure of an embedded clause to be preserved, thus obviating the need to reconstruct the

accounting for L2 learners’ underproduction of RCs, especially those lower on the

accessibility hierarchy, e.g. prepositional object and genitive relatives (Gass, 1979,

1980; Gass & Ard, 1984; Eckman et al, 1988; Yip, 1991).

Strictly speaking, the disparities in empirical support for the three predictor

hypotheses, as revealed in Table 4, need to be interpreted with some caution, as they

may have been a consequence of methodological differences. For one thing, studies

on universal factors often vary in their research focus. Most of them do not

investigate all the three hypotheses. Instead, often assuming that the support for one

hypothesis would indirectly falsify the other two, they tend to examine only one

particular hypothesis with little consideration for the others. For another, these

studies differ to a great extent in their participants’ L2 proficiency and their elicitation

measures. Some included advanced learners or employed productive tasks (e.g.

sentence-combining, free writing, speech production), whereas others, intermediate

learners or receptive tasks (e.g. comprehension questions, interpretation tasks,

grammaticality judgments). It can be argued that different proficiency levels and

testing instruments may yield different results concerning the validity of a particular

hypothesis.

Table 4: Summary of previous SLA studies on English relative clauses in terms of their support for the PFH, PDH, or NPAH (adapted from Izumi, 2003: 293-294)

Study Subjects Data elicitation method Hypothesis confirmed

*Studies which specifically examined the effect of instruction on English RC acquisition.

2.2.3 Summary

This section has so far explored two potential sources of difficulties encountered

by ESL/EFL learners in acquiring RCs: L1 interference and universal factors. On

the one hand, some researchers, especially those following the paradigm of the

contrastive analysis (e.g. Schachter, 1974; Schachter et al., 1976; Bley-Vroman &

Houng, 1988, cited in Kamimoto et al., 1992; Zhao,1989, cited in Kamimoto et al.,

1992; Li, 1996; Wei, 1997; Gisborne, 2000; Yin, 2001; Chan, 2004a, b), have pointed

out L1 interference as a significant or even the sole cause of RC error patterns and

underproduction (either conscious avoidance or subconscious underuse) by L2

learners. They argue that L2 learners whose native language employs different

relativization processes or has different RC distributions and functions often

experience great difficulty acquiring English RCs. The validity of these studies on

L1 interference, however, suffers from problems in their experimental design.

Moreover, the significant role of the native language on RC acquisition has been

disaffirmed by counterevidence (e.g. Berkau, 1976; Ioup & Kruse, 1977; Chiang,

1980; Gass, 1980; Sadighi, 1994; Liu, 1998; Chen, 2003; Liu, 2005). As an

alternative to L1 transfer, others (e.g. Cook, 1973; Ioup & Kruse, 1977; Gass,

1979,1980; Schumann, 1980; Gass & Ard, 1984; Yip, 1991; Wolfe-Quintero, 1992;

Kutoba, 1993; Sadighi, 1994; Flanigan, 1995; Izumi, 2003) have approached the same

issue from the point of view of universal factors. Specifically, L2 learners’ difficulty

with English RCs, regardless of their L1 background, is universally attributable to

processing problems posed by structural complexities of RC constructions.

Underlying these processing problems are three major hypotheses─Sheldon’s (1974)

Parallel Function Hypothesis, Kuno’s (1974) Perceptual Difficulty Hypothesis, and

Keenan and Comrie’s (1977) Noun Phrase Accessibility Hierarchy Hypothesis─each

predicting a different difficulty order for acquisition of various RC types. Empirical

support for the three predictor hypotheses varies, with the NPAH enjoying the most

endorsement from SLA researchers, followed by the PDH and then the PFH. These

mixed results may in effect be a direct consequence of differences in research

methodology.

It should be pointed out that the aforementioned literature on English RC

acquisition by L2 learners on the whole has centered on the syntactic aspects of RC

acquisition only (with the exception of Bley-Vroman & Houng, 1988, Zhao, 1989,

and Li, 1996, all of whom studied the pragmatic differences between English and

Chinese in their RC functions): while some studies examine RC structural difficulties

afflicting L2 learners from a cross-linguistic point of view (e.g. right-branching in

English as opposed to left-branching in L1), others investigate RC structural

complexities confronting L2 learners from a universal processing-based perspective

(e.g. center-embedding versus right-embedding). Furthermore, all these studies are

limited in scope to the case of restrictive RCs. The issue of how L2 learners acquire

(or use) non-restrictive RCs is rarely touched upon.

To put the matter simply, the review of the previous literature suggests that

although a great deal of effort has been made to shed light on EFL/ESL learners’ RC

acquisition from different perspectives, what seems to be still lacking is further

research (1) which inquires into L2 learners’ acquisition of NRRCs, the marked type

of RCs in English; and (2) which probes into the functional aspects of RC acquisition,

namely, the issue of how effectively L2 learners are able to employ English RCs for

such pragmatic/discourse functions as identifying, characterizing, presenting,

interpolating, and backgrounding. It is mainly to fill in these two gaps in RC

research that the present study is conducted.