Abstract

Some researchers have claimed thatwh-movement in ASL is rightward, contrary to the apparent universality of leftwardwh-movement. In contrast to this claim, we argue thatwh-movement in ASL is to a leftward specifier of CP. We account for the occurrence of rightwardwh-elements by independently motivated syntactic and discourse factors which lead to the appearance ofwh-elements in sentence- or discourse-final positions—not by rightwardwh-movement. Our analysis provides an account for a variety of ASL direct and indirectwh-questions and is in accord with cross-linguistic generalizations.

Introduction.

For many years, it has been observed thatwh-movement overwhelmingly bringswh-phrases to a sentence-initial position, now commonly thought of as spec(ifier of) CP. It is quite possible that no language uses rightwardwh-movement (althoughwh-elements may occur on the right edge of a sentence through a different process). This phenomenon remains unexplained, yet its statistical strength is such as to lead an investigator to expect thatwh-movement will be leftward in the next language studied. Thus, it may be rather surprising that a claim has been made for rightwardwh-movement (to a rightward spec-CP), in American Sign Language, inAarons, Bahan, Kegl, & Neidle 1992,Aarons 1994,Neidle, Kegl, & Bahan 1994,Neidle, Kegl, Bahan, Aarons & McLaughlin 1994.

American Sign Language (ASL) is the visual-gestural language used by the Deaf community in the United States and parts of Canada. Over the last thirty years, numerous studies of its structure have come to the conclusion that it has the characteristics of a natural human language (for overviews, seeKlima & Bellugi 1979,Padden 1988b,Wilbur 1987); it is acquired like spoken languages (seeNewport & Meier 1985,Lillo-Martin 1996a); it even breaks down like spoken languages (seePoizner, Klima, & Bellugi 1987). Some effects of the visual modality have been claimed to influence ASL (e.g. in the role that iconicity plays); yet these effects are in general seen to be surface, linguistically minor consequences of the articulatory apparatus (seeLillo-Martin 1996b). As we will show, ASL has many sentence types that have rightwardwh-elements. Is it possible that contrary to what has been found in spoken languages, spec-CP andwh-movement are to the right in ASL? Or are spec-CP andwh-movement actually to the left in ASL, with the rightward occurringwh-elements the result of some other factors?

In this article, we will argue that the latter is correct. We will provide additional evidence thatwh-movement in ASL is leftward, to a leftward spec-CP (in accord withLillo-Martin 1990 andPetronio 1993). In order to support this assertion, we will need to provide an alternative account for the data that motivated the rightward movement analysis. In so doing, it will become clear that a variety of processes affect the position ofwh-elements in ASL, some of which are strictly grammatical, but others of which are stylistic, subject to individual variation. Our goal in this paper is to examine a broad range of data from a variety of native ASL signers and to present a coherent analysis for these different judgments.

As we develop our analysis, we will expand on previous analyses to account for the wide variety of differentwh-sentences that are found in ASL, providing further support for the argument that spec-CP is on the left in ASL. We will show that the most commonly used forms forwh-questions follow from aspects of discourse strategies used in ASL; in particular, its discourse orientation. This discourse structure is supported by specific syntactic structures that are commonly used for focus and presupposed information; these structures enter into the form of questions as well as other sentence types. Finally, we attribute differences between speakers to stylistic and individual differences in discourse strategies and the subcategorization features of lexical items.

1. Cross-linguistic generalizations.

It has generally been assumed since the earliest cross-linguistic research on the formation ofwh-questions that overtwh-movement of interrogative phrases is leftward (e.g.C. L. Baker 1970,Bresnan 1970,Bach 1971, and others, including more recently,Georgopolous 1991).¹ In this section, we will examine the evidence for this generalization and the current theoretical approaches. In addition, we will differentiate truewh-movement from other constructions withwh-elements, such as those involving scrambling or focus movement.

It is commonly assumed that sentences such as 1a illustrate leftwardwh-movement.²

Although truewh-movement moves awh-element leftward, there are other options for interrogativewh-constructions across languages. In numerous languages, including Japanese (an SOV language),wh-words can remain in situ in the syntax, as illustrated in 2.

In yet other languages,wh-phrases can remain in situ, or optionally move to a leftward clause-initial position. In French, for example, awh-phrase may move or remain in situ (only) in matrix clauses. Examples are shown in 3a and 3b respectively.

These differences in the surface form ofwh-questions can be attributed to parametric differences in the level at whichwh-movement applies.Huang (1982) argued that these differences disappear at the level oflogical form (LF), because at that level allwh-phrases occupy the clause-initial scope-bearing position. This contrast can be seen most clearly in the difference between direct and indirect questions in English and Chinese, as illustrated in 4 and 5.

In English, thewh-element overtly moves to the beginning of the clause over which it takes scope. Thus, in the direct question in 4a, thewh-phrase moves to the beginning of the matrix clause, while in the indirect question in 4b, it moves to the beginning of the embedded clause. In Chinese, on the other hand, although thewh-phrase does not move overtly, the scope of thewh-element is reflected in the different interpretations of the direct and indirect questions. Thus in 5a, thewh-element has scope over the matrix clause, while in 5b it has scope only over the embedded clause, like the English counterparts. Huang argued that this follows straightforwardly from the proposal that thewh-phrase in Chinese does move to the beginning of the clause over which it takes scope, but only at LF. Thus, the parametric variation between languages like English, Japanese or Chinese, and French, is in the level at whichwh-movement applies, not whether it applies at all.³

We have seen that languages may differ in the level at whichwh-movement applies. It is generally assumed, however, that there are no differences in the direction of overt movement: it has repeatedly been argued that when overtwh-movement applies, it universally brings awh-element leftward. For instance, a major thrust ofC. L. Baker 1970,Bresnan 1970,Bach 1971, andLangacker 1974, was to account for the cross-linguistic occurrence of leftwardwh-movement and rule out rightwardwh-movement.C. L. Baker (1970) andBresnan (1970) did this through universal rules. For instance, Baker’s ‘Question Universal’ stipulated that the only possible rule for overt question movement moved thewh-element to the sentence-initial position, while Bresnan’s ‘Complementizer Substitution Universal’ stated, ‘only languages with a clause-initial COMP permit a COMP substitution transformation’ (e.g.wh-movement).Bach (1971) andLangacker (1974) in part prohibited rightwardwh-movement by involvingRoss’s (1967) observation that rightward movement rules are (universally) bounded.

In the twenty-five years since this early work, research on the properties ofwh-movement has expanded, yet still the basic observation seems universal: truewh-movement is leftward. In current theoretical terms, this generalization is captured by postulating that the landing site forwh-movement is the specifier of CP (Chomsky 1986). Furthermore, it must be stated that the specifier of CP is universally on the left, even in languages that have been argued to have other specifier positions to the right (seeGeorgopoulos 1991). The source of this generalization remains a mystery, but its strength is striking.⁴

Movement to spec-CP (at whatever level) is accounted for by thewh-criterion (May 1985,Rizzi 1991), as in 6.

The effects of this criterion can be illustrated with the diagram in 7.

In 7, the head of CP (C⁰) has a [+WH] feature. Thus, to satisfy thewh-criterion, the [+WH] object (the wh-operator) must move to spec-CP in order to be in a spec-head configuration with the [+WH] head. This spec-head agreement of [+WH] features is an example of the more general process of spechead feature agreement.

Thewh-criterion also accounts for the movement of awh-element to spec-CP of an embedded clause in indirect questions (e.g. 4b), illustrated in 8. Verbs such aswonder,ask andknow may select a [+WH] complement. In this case, thewh-operator will move to the specifier of the embedded clause where thewh-criterion is satisfied.

In English, which has overtwh-movement, thewh-criterion must be satisfied at S-structure.⁵ Thus, the example in 9 is ungrammatical, since the [+WH] C⁰ and thewh-operator are not in a spec-head configuration within the embedded CP.

If we make the standard assumption that movement to spec-CP in satisfaction of thewh-criterion is true ‘wh-movement’, this allows us to distinguish ‘wh-movement’ from other constructions involving the movement ofwh-elements. For example, in Japanese, scrambling can move various constituents, includingwh-elements, leftward.Takahashi (1993) presents evidence that some instances of scrambledwh-elements do count aswh-movement, whereas others behave like movement to an argument position, not like truewh-movement. Only the cases of truewh-movement involve movement to spec-CP.

Although, as discussed above,wh-elements can be fronted through a process distinct fromwh-movement, it has also been argued thatwh-elements cannot be topicalized (Bresnan & Mchombo 1987,Epstein 1992). According to Bresnan and Mchombo, this is so because a topicalized element must be presupposed, but an interrogative element, since it does not refer to a specific entity, cannot be presupposed.⁶ Rather, Bresnan and Mchombo claim that awh-element is inherently a focus (cf.Dik 1978, cited by Bresnan and Mchombo). This position, thatwh-elements are inherently foci, is also advocated byHorvath (1986),Rochemont (1978,1986), and others. Alternatively,Epstein (1992) argues thatwh-phrases cannot be topicalized because of the syntactic principle of economy. According to this analysis, syntactic topicalization, plus LF movement to the spec of the [+WH] CP, is a two-step process, which is less economical than a single movement to spec-CP.

Apparent exceptions to the generalization that truewh-movement is leftward are found in languages withwh-elements on the right (excludingwh- in situ). In every case that we are aware of, however, there is evidence that these rightwardwh-elements are not the result of truewh-movement to a rightward spec-CP. Furthermore, such languages also showwh-elements in other positions.

For example,Bergvall (1987) shows thatwh-elements in Kikuyu (an SVO Bantu language) can occur sentence-initially, in situ, or sentence-finally. Her examples are given in 10.

Bergvall shows that some rightwardwh-elements can be attributed to sluicing, in the sense ofRoss 1969: the constituents that would have followed thewh-elements are left unexpressed. Another way forwh-elements to appear on the right is through movement, but Bergvall shows that this rightward movement is subject to locality conditions more severe than the usual restrictions on (leftward)wh-movement. For example, she shows that awh-object cannot occur to the right of a manner adverbial, as illustrated in 11a, and awh-adjunct cannot occur to the right of a time adverbial, as illustrated in 11b.

The precise conditions under whichwh-elements can appear rightward will not be explicated here, but Bergvall argues convincingly that ‘there is no righthand analog to the lefthand COMP [our spec-CP] position (a uniform landing site for left-moved question words)’ (p. 47). In fact,Clements et al. (1983) claim that the leftwardwh-elements in Kikuyu obey the island constraints that typically constrain truewh-movement. This is exactly what would be expected if the leftwardwh-elements have undergone truewh-movement to a leftward spec-CP in Kikuyu.

Tangale, an SVO Chadic language, is another language that has rightward-occurringwh-elements.Kidda (1985) notes differences in the possible positions ofwh-subjects andwh-objects, as illustrated in 12 and 13.

In 12a, thewh-subject occurs after the direct object, but before the PP, while in 12b, it occurs sentence finally, after the direct object and PP. In 13a, thewh-object occurs between the verb and the PP, but as shown in 13b, thewh-object cannot occur sentence finally, following the PP.

Kenstowicz (1985), using phonological effects of sandhi rules, argues that Tangale has a postverbal focus position, in which focused NPs andwh-elements can appear. Under his analysis, thewh-elements in 12a and 13a occupy this focus position. Kenstowicz does not discuss the appearance of thewh-subject in final position, nor the ungrammaticality of thewh-object appearing in the sentence-final position.

Tuller (1992), in agreement withKenstowicz 1985, argues that there is a focus position following the verb and direct object in Tangale. Furthermore, she does discuss the sentence-finalwh-elements found in sentences such as 12 b and 13 b. She argues that Tangale has a second focus position, which she claims is a rightward spec-CP, on analogy with the use of the leftward spec-CP for focus in certain other languages. But she cites differences between the characteristics of the sentence-final focus position and the usual characteristics of a leftward spec-CP. Although somewh-phrases and focused non-wh-constituents can appear in the final focus position in Tangale, direct objects and locative goals of motion verbs cannot occur in the final focus position, whether focused non-wh orwh-constituents, as illustrated in the ungrammatical 13b. In comparison, Tuller considers Kanakura, a related Chadic language, with similar properties to Tangale, including two focus positions. Tuller argues that in Kanakura, in contrast to Tangale, spec-CP is on the left, and she shows that any constituent can be focused by movement to this sentence-initial focus position. Thus, rightward movement in Tangale is restricted in a way that leftward movement in Kanakura (and in general) is not.⁷

The restrictions on movement to the sentence-final focus position in Tangale lead us to suspect that this position is not spec-CP. One alternative, in accord with recent work on focus byBrody (1990) and others, is that this position might be the specifier of a separate focus phrase, not the specifier of CP. An analysis along these lines would maintain the hypothesis that crucial characteristics ofwh-movement and the specifier of CP position are upheld cross-linguistically.

In sum, there has been significant evidence that the syntactic process ofwh-movement only bringswh-elements to a leftward, clause-initial position. The scope-bearing position, spec-CP, is a position in which thewh-element’s [+WH] feature agrees with the [+WH] C⁰. Languages may vary in whether or not they applywh-movement overtly, but when it applies overtly, it is leftward. Ifwh-elements are found in a rightward, clause-final position, this is evidently for reasons other thanwh-movement, such as a more general focusing operation, sluicing, or because thewh-element remains in situ. True rightwardwh-movement appears to be unattested.

Recent works (Aarons, Bahan, Kegl, & Neidle 1992;Aarons 1994;Neidle, Kegl, & Bahan 1994; andNeidle, Kegl, Bahan, Aarons, & McLaughlin 1994) challenge this generalization. (Henceforth, when discussing claims that are common to all these works, we will use the acronym ABKN.) ABKN claim that in American Sign Language,wh-movement is to a rightward spec-CP. In the remainder of this paper, we will present data from ASL which dispute this claim, and show that the data which prompted ABKN to make the claim thatwh-movement in ASL is rightward can be accounted for by independently motivated syntactic processes and properties of ASL. Thus, we will argue that the generalization that spec-CP is universally leftward is supported, rather than undermined by the data from ASL, and furthermore that a leftward spec-CP is necessary in order to account for the full range of ASL data.

2. Two analyses ofwh-sentences in ASL.

We begin this section by presenting some common examples of undisputed grammaticalwh-sentences in ASL. We will then show that these examples can be accounted for by either of two hypotheses: the hypothesis that we are advocating, that spec-CP andwh-movement is leftward universally (cf.Lillo-Martin 1990,Petronio 1993); or the alternative hypothesis that spec-CP andwh-movement is rightward in ASL (ABKN). We will argue that even in accounting for these undisputed examples, the leftward movement hypothesis is theoretically superior.

We begin with some simple ASLwh-questions and move into details of the two analyses. Both analyses agree thatwh-phrases in interrogative ASL sentences can remain in situ or move. The following commonwh-questions in 14–16 can be analyzed as having thewh-word in situ under either analysis. (SeeAppendix A for an explanation of the notation/gloss system.)

In 14, thewh-object appears between the verb and an adverb. Under the now common assumption that ASL is underlyingly SVO (Fischer 1974,Liddell 1980,Padden 1988a), both the rightward and leftward analyses use 14 as evidence thatwh-words can remain in situ (whether the adverb is adjoined to either VP or IP). Consistent with the hypothesis thatwh-phrases can remain in situ, 15 shows awh-subject at the beginning of the sentence, and 16 shows awh-object at the end of the sentence.

The leftward and rightward analyses both assume that a nonmanualwh-question marker is associated with interrogativewh-questions (seeBaker-Shenk 1983). The nonmanual characteristics of thewh-question marker include squinted eyebrows, a head tilt, possible slightly forward body position, and possible raised shoulders (Valli & Lucas 1992:283). In 14–16, thewh-question marker occurs simultaneously with all of the manual signs in the questions, as indicated in the glosses by the horizontal line over all of the signs.⁸

Both the leftward and rightward analyses also argue thatwh-questions in ASL can be produced withwh-movement. These contradictory analyses are based in part on identical surface strings. One is a very commonly found type of ASLwh-question, in which awh-element is found both sentence initially and sentence finally. It will be most straightforward to explicate the differences between the two accounts using this type of question, illustrated in 17–19.

In 17 awh-subject appears at both the beginning and the end of the sentence; in 18 and 19, awh-object appears at both the beginning and the end. Note that in 19, there is nowh-word in the position between the verb and adjunct, i.e. thewh-object is not in situ.

In 20a, b we give the structures for 19 under the leftward and rightward analyses (respectively) to illustrate the main differences between them.

The structure in 20a illustrates the leftward analysis fromPetronio 1993, in which spec-CP is on the left andwh-movement is leftward. As discussed in§1, the movement of awh-phrase to spec-CP checks its [+WH] feature by agreement with the [+WH] feature in C⁰. The finalwh-word in 20a occupies the head of CP. This type of sentence-final double also occurs with other focused/emphasized constituents, such as modals, verbs, and quantifiers. These elements are licensed by a [+F(ocus)] feature, which is checked through spechead agreement with a [+F] Operator in spec-CP at LF. More extensive argumentation for the final double, and discussion of its licensing, are presented in§3.1.

We summarize the main characteristics of the leftward hypothesis as discussed so far in 21.

The structure in 20 b illustrates the rightward analysis first presented inAarons, Bahan, Kegl, & Neidle 1992. Spec-CP is on the right andwh-movement is rightward, motivated by the need for the [+WH] C⁰ to agree with a [+WH] spec-CP. According to these authors, the sentence-initialwh-word in 20b is a base-generated topic, similar to other base-generated topics commonly found in ASL.¹⁰ The basic characteristics of ABKN’s original analysis, as well as their later work, are listed in 22.¹¹

Both analyses can account for the data in 14–19, however, the rightward movement analysis makes ASL typologically unusual in at least three ways. First, it makes ASL an exception to the cross-linguistic generalization thatwh-movement is leftward. Second,Bresnan & Mchombo 1987 andEpstein 1992 have noted restrictions onwh-phrases occurring as topics (see alsoBach 1971). Under the rightward analysis, having awh-topic makes ASL an exception to such restrictions. Third, topics in ASL typically occur with a topic marker and are followed by a slight prosodic break (Fischer 1974,Liddell 1980). The initialwh-word in 19 does not have a topic marker, nor is it followed by a prosodic break. Thus, under the rightward analysis,wh-topics differ from other ASL topics.

Sentences such as 23 illustrate another important difference between the leftward and rightward analyses.

In 23, thewh-element appears at the beginning of the embedded clause selected by KNOW. The leftward analysis has a straightforward account for this sentence; KNOW takes a [+WH] complement, and the embeddedwh-word moves leftward to the embedded [+WH] spec-CP. Under the rightward analysis,Neidle, Kegl, & Bahan (1994) claim that in 23 KNOW does not subcategorize for a [+WH] complement. Instead, although thewh-phrase does not have a topic marker, they argue that the leftwardwh-object in 23 is due to embedded topicalization of thewh-phrase. (We will discuss indirect questions in greater detail in§3.3).

The analysis developed below is based on data previously collected from native signers for work presented inLillo-Martin 1990 and inPetronio 1993. These data are supplemented with interview data obtained from eighteen native ASL users. The interviews included watching a videotape of a native ASL model signing differentwh-sentences. The consultants judged the grammaticality of the utterances and discussed reasons for their judgments. They also specified situations in which the sentences might be used and possible corrections for sentences judged ungrammatical. Some of the data came from commercially made videotapes.

From the interviews, it became clear that subtle factors often influenced consultants’ judgments of the sentences presented. Sentences with similar word order, for instance, could be judged differently on the basis of nonmanual markers, positions and movement of the head, pausing and rhythmic patterns, the context of the utterance, and even the direct of eye gaze. These factors have each been noted as important to ASL utterances, but we found that their interaction had an even greater impact than we had anticipated. As one example, we found that the degree to which the eyelids were opened during a brow-raise nonmanual marker could affect the interpretation and judgment of the sentence. Such details are typically missing in the rough glossing systems usually used to transcribe ASL. We believe that this missing information may account for some of the conflicting judgments found in different works on ASL syntax, although true judgment differences undoubtedly exist as well. Our discussion of the data that follow will often mention the importance of context and the nonmanual configurations accompanying the signs, both of which are crucial factors that affected judgments and led to our analyses of differentwh-sentence types.

3. Final doubles in ASL

4. Null arguments and covert wh-words in ASL.

In this section, we will show how the interaction of null arguments and the double construction can result in another type of ASL sentence that has awh-element on the right. ASL has a class of verbs that can be inflected for agreement with subject and/or object (Fischer & Gough 1978,Padden 1988a). Within this class of verbs, an argument with which a verb agrees can be null, like the null subject arguments found in richly inflected null subject languages such as Italian. ASL also displays the characteristics of adiscourse-oriented language (cf.Huang 1984). Even when the verb does not agree with a certain argument (because of properties of the verbal class, irrelevant here) the argument may be null, as long as it is identifiable from the context. Due to these two properties, ASL allows null arguments (subjects and/or objects) in many circumstances. (SeeLillo-Martin 1986 for more discussion of null arguments in ASL.)

Lillo-Martin andFischer (1992) have argued that in the appropriate context, ASL allows nullwh-words. The followingwh-questions are commonly found in ASL.

According to Lillo-Martin and Fischer, thesewh-questions involve awh-phrase in which thewh-element is null. As expected, thiswh-phrase may move to the leftward spec-CP, and it takes scope in the usual way. If a covertwh-element is allowed in short, simple questions such as 56 and 57, why should they not appear in other questions? Lillo-Martin and Fischer argue that they do, although their acceptability is tied to particular contexts. More precisely, what the covertwh-question is asking must be recoverable from the context for it to be used. In this way the covertwh-element is similar to other null arguments. Examples 58–60 show more complex sentences with a covertwh-argument. (We will use ‘e’ to represent the nullwh-element.)

Thus, in addition to having non-wh null subjects and objects, ASL also permits nullwh-elements, under the appropriate context.Aarons (1994) also notes the existence of covertwh-questions in ASL.

Now, let us consider sentences that have a whq marker over the whole sentence and a singlewh-word, which is awh-subject that occurs at the end of the sentence, as in 61.

This type of sentence is always reported as grammatical by ABKN and is a major piece of evidence for their analysis that spec-CP is on the right andwh-movement is rightward. However, we find that when sentences such as 61 are presented in isolation, judgments vary—some signers accept them, but others do not. Given the existence of a nullwh-element in ASL, restricted to appropriate contexts, we can account for sentences like 61 under the leftward movement analysis, while also accounting for the variation in judgments. Our analysis is that 61 has the structure of awh-double (as discussed insection 3.2), and also has a covertwh-subject, as represented by the e in 62.

This analysis is consistent with other ASL sentences, since other elements that can normally appear in a focus position can also show up, in appropriate contexts, with the twin null. An example with a focused modal, CAN’T, is given in 63.

Two other types of evidence favor our analysis over a rightward movement analysis. First, our analysis can account for the variation in judgments for sentences such as 61 when presented in isolation. A signer who will reject a sentence such as 61 when it is presented in isolation will often accept it when it is in the appropriate context, as in 64.

This requirement is in accord with properties that we find for other null arguments. In particular, null arguments, including nullwh-elements, must appear in a context from which the null element is recoverable. Second, under a rightward movement analysis, we would expect that complexwh-subjects such as WHO MOTHER (whose mother), WHO CAR (whose car) or WHICH GIRL could also appear at the end of the sentence. As shown in 65–67, however, complexwh-subjects are not permitted to occur at the end of the sentence.

While the rightward analysis incorrectly predicts that 65–67 should be grammatical, the leftward analysis with the interaction of covertwh-elements and the final double predicts them to be ungrammatical: a complex phrase cannot occur as a double, since the double occupies a head position.

So far, we have seen examples with a nullwh-subject and a sentence-finalwh-double. We can use the same analysis to account for sentences like 68. Under our analysis, this example has a nullwh-object and a sentence-finalwh-double.

As we have seen, examples like 68 are analyzed differently under the rightward movement hypothesis, which takes the sentence-finalwh-element to be in spec-CP. In support of their claim that thewh-element in 68 is in spec-CP,Aarons et al. (1992) observe that an NP cannot normally follow a sentence adverbial, as shown in the ungrammatical 69.

Therefore, since an NP does not normally follow an adverbial, they conclude that thewh-phrase in sentences such as 68 is the result of rightwardwh-movement. However, as we have shown, the interaction of a nullwh-argument and a finalwh-double can account for the same fact.

The rightward and leftward analyses differ crucially with respect to examples such as 70.

We have found that examples with a nullwh-object and an overt phrasal double (such as 70) are ungrammatical. This is completely consistent with our observation insection 3 that a head, but not a phrase, can occur in the sentence-final focus position.

Aarons (1994:146) discusses examples like 70, and claims they are grammatical. She uses this judgment to argue that thewh-double analysis cannot be correct, since it cannot account for a fullwh-phrase following an adverb, but we have consistently found that consultants reject this sentence without a pause before WHICH COMPUTER.

In other recent work, Neidle, Kegl, Bahan,Aarons and MacLaughlin (1994) have argued for rightwardwh-movement and against the leftward hypothesis using the example given in 71, which has a slightly different fullwh-phrase following the adverbial.

We also find that 71 is grammatical, unlike 70. This need not present a counter-argument to the leftward analysis, however. Consider the contrast between the (a) and (b) examples in 72 and 73.¹⁸

Examples 72a and 73a show that a ‘light’ NP may not follow a sentence adverbial, as also seen in 70. Examples 72 b and 73 b, however, show that when the NP becomes more complex, or ‘heavy’, it can follow the adverb. We take these examples as evidence that ASL has a construction similar to the heavy NP shift construction found in many languages. If so, then 71 is not a counterexample to the leftward analysis, as it can be explained using heavy NP shift rather than rightwardwh-movement to a rightward spec-CP.

Our analysis accounts for the various judgments of sentences like 61, by claiming that they involve a covertwh-element, which must be identified in the discourse context. This sentence type is accounted for using structures independently known to exist in ASL. In addition, our analysis accounts for the ungrammaticality of sentences with a full phrasalwh-subject on the right, as in 70 above.

5. Indirect questions and the nonmanual marker.

In this section, we will discuss the structure of indirect questions in ASL. It will first be necessary to explicate some of the differences between the rightward and leftward analyses with respect to the nonmanualwh-question marker. We will then turn to the different predictions regarding indirect questions made by the two proposals.

6. Discourse strategies and multisentence discourses.

In this section, we will examine several examples of strings of signs which end with the nonmanualwh-question marker cooccurring only with a finalwh-element. We will argue that a predominant discourse strategy used in ASL accounts for these structures: this strategy forms topic-comment structures by placing old or presupposed information before new information at both the sentence and discourse level.²⁷

Early work on ASL claimed that the notions of subject and object did not play any role in determining the overt word orders (Friedman 1976,Edge & Herrmann 1977,Ingram 1978). Most early researchers claimed instead that notions such as topic-comment, rheme-theme, and old information-new information determined the word order. Since that time research has shown that the notions of subject and object are relevant for the word order and grammatical structure at the sentential level (e.g.Liddell 1980,Padden 1988a). Taking a broader view we can see the strong effects of discourse strategies that result in old information placed before new information, and topic-comment structures commonly used within sentences and discourses.

Within sentences ASL productively uses topicalization to front constituents. Topicalization is so common that when a declarative sentence is presented in isolation, many people will reject the underlying SVO order. For example, when the sentences in 91 are presented in isolation many people prefer 91b-d, which have a topicalized constituent.

Topicalization can also apply inwh-sentences, and can result in a sentence in which thewh-question marker occurs only with a rightwardwh-subject, as in 92 and 93. This type has a topic marker over a sentence-initial VP, and awh-subject that occurs at the end of the sentence.

The surface order of these sentences might be interpreted as deriving from rightward movement of awh-subject. However, taking into consideration the nonmanual markers, it is clear that these sentences illustrate VP topicalization.²⁸

In addition to discourse strategies (such as the use of topic-comment structures) affecting individual sentences, we see the effects of these strategies on the organization of information within a discourse. A discourse usually starts with a topic (old information) and then presents the comment (new information). This type of discourse organization appears very frequently in ASL, especially as compared to English.

It is relevant to our analysis ofwh-questions, to consider how placing old, presupposed information first in a discourse may affect the structure of thewh-questions. Often a presupposition will be given and then a question relating to this presupposition will be asked. For example, when asked for an ASL translation of the English sentence, ‘Who does John like?’, many native ASL consultants constructed multisentence discourses such as those in 94–95.

Notice that in 94, all of the presupposed information is presented in the first, declarative sentence, which is followed by the simple question, ‘Who is it?’ In 95, the topicalized constituent WOMAN_aINDEX presupposes the existence of a woman. In the second sentence, WHO_aINDEX, the identity of the woman is questioned.

In 94 and 95, thewh-word is followed by an INDEX, which functions as a predicate. Counterparts to these examples, without the INDEX, are given in 96–97.

Examples 96 and 97 clearly have the meaning ‘Who is it?’ We propose that these examples, like 94 and 95, involve awh-question that is a separate sentence, with the difference that in 96 and 97 the predicate is unexpressed. In ASL, this can result in a singlewh-word question on the surface.

English also allows singlewh-word questions in context.

The final question word in these sentences clearly constitutes its own sentence; the question word is not an argument of the first sentence, but questions something mentioned in the first sentence.

A similar phenomenon of missing constituents was investigated byRoss (1969); he gave it the namesluicing. According to Ross, in these question examples it is possible to sluice off parts of a sentence which are recoverable from a previous sentence in the discourse. We believe the ASL examples can be explained as a type of sluicing.

We take the sentences in 100 and 101 to be further examples of this phenomenon: one presents the presupposed information, the second thewh-question.

In 100–101, there is a clear difference in the nonmanual marker between the first and the second sentence. Only the second has a whq nonmanual marker. The first sentence may or may not have a head nod, but we claim that it is a separate sentence in all of these cases.²⁹

We are now in a position to discuss a slightly different type of example. In the appropriate contexts, sentences such as 102 b and 103 b are acceptable. (102a and 103a are repeated from 100a and 101a).

We now claim that 102b and 103b are two-sentence discourses, similar to 102a and 103a.³⁰ The only difference is that in the (b) examples, a null argument is used in the first sentence. It has independently been argued that null arguments are available in ASL (see§4 above, andLillo-Martin 1986), in the appropriate context. As expected, we find that examples like 102b-103b are accepted only in appropriate contexts. Thus, according to our analysis, 102b-103b consist of multisentence discourses in which the first sentence contains a null argument, and the second is a singlewh-word question.

In our analysis, multisentence discourses may also occur with a separatewh-element as the first sentence, as illustrated in 104.

We argue that each of the examples in 104 consists of three separate sentences. The first, thewh-element itself, is a question directed by the signer to the signer. The second sentence gives the presupposition the signer is pondering. The final sentence, the finalwh-element itself, can be either a direct question addressed to an addressee, or a continuation of the self-talk.³¹

Evidence that the firstwh-element is a separate sentence comes from extraction facts. As illustrated in 105, a topic can cooccur with awh-question.

In contrast, it is not possible for an element of what we have analyzed as the second sentence to be topicalized over the initialwh-element in 104, as shown in 106.

Thus, we find that multisentence discourses can contain utterances with a nonmanualwh-question marker only over a singlewh-element, which we interpret as a single-sign sentence.

ABKN also note examples of strings in which the nonmanualwh-question marker occurs only with a rightwardwh-element, as in 107 (fromAarons et al. 1992:108).³²

According to ABKN’s analysis, examples such as 107 are considered single sentences with thewh-element moved rightward, rather than multisentence discourses. Because the rightward spec-CP has lexical material, according to their analysis the nonmanualwh-question marker need not spread. ABKN’s analysis predicts that examples such as 107 should be grammatical even without context, since they would be simple examples ofwh-elements moved to a rightward spec-CP. As we explained earlier, we find that judgments on these sentences are very dependent on context; without the appropriate context many consultants will usually judge them ungrammatical, while other consultants are able to construct an appropriate context that makes the sentence grammatical. Our analysis predicts such a dependence on context, since the sentences employ null elements that are known to require identification in the context or discourse.

In the next section, we will consider possible sources of judgment differences between individual consultants, particularly in regard towh-questions with leftwardwh-elements.

7. Individual stylistic and idiolectal differences.

The literature reports differences in judgments reported for sentences that have a single, leftward, sentence-initialwh-object. To maintain a leftward analysis, we must account for the varying judgments for this type of sentence. In §7.1 we look at simplewh-questions with a single sentence-initialwh-object and account for the different judgments by attributing them to individual stylistic preferences that are in accord with the discourse-oriented strategies of ASL discussed above. In§7.2 we examinewh-questions involving long-distance extraction out of an embedded clause and argue that different judgments, in part, indicate idiolectal differences in the subcategorization properties of different verbs, that is, if the matrix verb is a bridge verb or not.

8. Summary and conclusion.

One of our goals in this article was to evaluate the claim that spec-CP andwh-movement is to the right in ASL. We have considered this claim both with respect to cross-linguistic observations about the nature ofwh-movement and with respect to a variety of data from ASL itself. We found a general observation made in studies of spoken languages: in languages with overtwh-movement, spec-CP is clearly on the left. We examined potential counterexamples of rightwardwh-elements in the literature on spoken languages, and it appeared that in every case the rightwardwh-elements were due to factors other than rightwardwh-movement to a rightward spec-CP. The strength of this generalization led us to consider whether the cases of rightwardwh-elements in ASL could also be accounted for by factors other than rightwardwh-movement to a rightward spec-CP.

We found independently motivated features and constructions in ASL that interact withwh-elements in such a way as to result in the appearance ofwh-elements on the right. For example, a general focusing construction was shown to apply to non-wh-elements, resulting in a double of the focused element occurring in sentence-final position. By this same construction, a double of an interrogativewh-element (inherently focused) can appear in the rightward, sentence-final position. We noted that ASL independently allows covertwh-elements and showed that a covertwh-element can also occur in the double construction, which then results in a surface form in which only a single overtwh-element (the rightward, sentence-final double) appears. We discussed discourse strategies commonly employed in ASL, which include placing old, presupposed information before new information in both sentences and discourses, and result in multisentence discourses in which the presupposed information is presented in one sentence, followed by a sluicedwh-question, consisting of a singlewh-element, which occurs at the right of the discourse.

Using these independent features, we showed that examples ofwh-elements on the right in ASL do not require the postulation of rightward movement to a rightward spec-CP. We showed examples ofwh-elements on the left in ASL, which the rightward movement hypothesis is unable to account for. We demonstrated that the rightwardwh-movement analysis is problematic both in that it overgenerates ungrammatical examples and undergenerates grammatical examples. The leftward movement hypothesis, together with independently motivated mechanisms, correctly predicts the grammaticality of rightward and leftwardwh-elements in ASL. In addition to this empirical superiority in accounting for the ASL data presented here, the leftward movement hypothesis further supports the cross-linguistic generalization that truewh-movement and spec-CP are to the left. Our conclusion that ASL does not present a counterexample to this cross-linguistic generalization also supports the idea, as would be expected, that true linguistic universals hold across modalities.

Acknowledgments

This research was supported in part by research grant number 5 R01 DC 00183–12 from the National Institute on Deafness and Other Communication Disorders, National Institutes of Health. Earlier versions of portions of this work were presented at the 1995 annual meeting of the Linguistic Society of America in New Orleans, and at colloquia at the University of Rochester, Harvard University, and Cornell University. We would like to thank the respondents at these presentations for their helpful comments and suggestions. We would also like to thank Zeljko Bošković, Heles Contreras, Susan Fischer, Howard Lasnik, Don Metlay, Yutaka Osugi, and two anonymousLanguage referees for comments on an earlier version of this paper. In addition to the many other ASL signers who assisted us with data, we particularly want to thank Jimmy Challis, Jan DeLap, Phil Quibodeaux, Doreen Simons-Marques, Ted Supalla, and Clayton Valli.

Appendix A: Notation/transcription system for ASL

Appendix B: Sources for data in exx. 30–42

BCASL:Humphries et al. 1985.

Green Book:Baker & Cokely 1980.

Vista:Smith et al. 1988.

Footnotes

Contributor Information

Karen Petronio, Eastern Kentucky University.

Diane Lillo-Martin, University of Connecticut.

REFERENCES