Three Universal Hypotheses for English Relativization Difficulty

Different hypotheses have been proposed to account for the relative ease and

difficulty of various RC types, regardless of L2 learners’ native background. Among

them, three universal hypotheses in particular have stood up well in the SLA literature.

They are the Parallel Function Hypothesis, the Perceptual Difficulty Hypothesis, and

the Noun Phrase Accessibility Hierarchy Hypothesis, all based on different rationales

and predicting different difficulty orders for English relativization.

The Parallel Function Hypothesis (PFH), posited by Sheldon (1974), is based

on a cognitive-processing interpretation of the relationship between the grammatical

function of the head NP in the matrix clause and that of the relative pronoun in the

embedded. In examining the difficulty order in the comprehension of four RC types

─SS, OO, SO, and OS²⁰, as categorized in terms of the syntactic functions of the

head NP and its co-referential relative pronoun─by children learning English as their

first language, Sheldon found that these children understood RCs better when the

functions of the head NP and the relative pronoun were the same. Hence, he

formulated the PFH, which claims that RCs with a non-parallel function (SO/OS) are

more difficult than those with a parallel function (SS/OO).

A second predictor hypothesis of English RC difficulty is Kuno’s (1974)

Perceptual Difficulty Hypothesis (PDH). Deriving its validity from perceptual

considerations of the constraints on human short-term memory, the PDH essentially

states that RCs involving center-embedding²¹ pose greater difficulties than those

involving left-²² and right-embedding, since in the former, there is interruption in the

processing of the matrix clause. Compared to Sheldon’s PFH, Kuno’s PDH is

concerned only with the position of the RC in the matrix clause─whether it is

20 The first alphabet refers to the grammatical role of the head NP in the main clause, the second, that of the relative pronoun in the embedded clause, as exemplified in the following:

(i) SS: The dish that fell on the floor broke in half.

(ii) OO: The child ate the cookies (that) the neighbors baked.

(iii) SO: The candy (that) Billy gave me tasted good.

(i) Center-embedding: The cheese that the rat that the cat chased ate was rotten.

(ii) Right-embedding: The cat chased the rat that ate the cheese that was rotten.

22 Left-embedding does not exist in English relative constructions.

center-embedded or non-center-embedded─rather than the functions of the head NP

and the relative pronoun. The difficulty order predicted by the PDH for the four RC

types in the PFH is thus as follows: SO/SS sentences are more difficult than OS/OO

sentences.

The last, and arguably the most influential, predictor hypothesis is put forward

by Keenan and Comrie (1977). They examined different RC formations among

more than 50 languages in the world regarding syntactic functions relative pronouns

can serve and based on the typological markedness obtained, proposed the Noun

Phrase Accessibility Hypothesis (NPAH). The NPAH claims that the ease with

which RCs are formed follows a particular hierarchy of ordering in terms of the

grammatical functions of their relative pronouns, as is indicated in (64):

(62) Subject＞²³Direct Object＞Indirect Object＞Object of Preposition²⁴

＞Genitive＞Object of Comparison²⁵

Among these six RC types, subjects are universally easier─or more accessible─to

relativize than direct objects, which are more accessible than indirect objects, which

are in turn more accessible than prepositional objects, and so on. In contrast to the

23 ＞ means “more accessible than.”

24 Keenan and Comrie (1977) use the term “Oblique Object” in place of “Object of Preposition” in English.

25 Object of Comparison is included only for the sake of completeness, since such sentences as (i) and (ii) are marginally rare in English and even native speakers of English are unsure of their grammaticality.

(i) ?The man who Mary is shorter than is my brother.

first two hypotheses, the NPAH focuses primarily on the relative clause itself, with no

attention given to the matrix clause, and predicts another different difficult order for

the four RC types aforementioned: SO/OO sentences are more difficult than SS/OS

sentences (though this difficulty order is not explicitly stated in the NPAH but

deduced by reason only for a direct comparison with the PFH and the PDH).

Figure 1 below summarizes the three universal hypotheses in terms of their

NPAH; ease of acquisition is predicted at the accessible end of the hierarchy.

SS/OS＞OO/SO

Figure 1: Three hypotheses for English relativization difficulty (adapted from Doughty, 1991: 438)

Disparate as these three hypotheses may seem at first glance, their predictions

are all grounded on the relative psychological ease or difficulty of processing of

various RCs. Specifically, they are in line with Slobin’s (1973) universal operating

principles, which claim that “interruption” or “rearrangement” renders sentence

processing difficult. For instance, though originally motivated by typological

markedness, the NPAH is also processing-based, with certain psychological validity.

As Keenan and Comrie (1977) suggest, RCs formed in the lower position of the

hierarchy are more difficult, because they become harder to process, with their

underlying word order being more susceptible to rearrangement in the surface

structure.