Sentence structure: Syntax.

 

One of the fundamental properties of human language is that it can be used to create an unlimited number of utterances given a limited number of pieces. At the syntactic level, this property follows directly from a mathematical property of language called recursiveness. We’re all familiar with recursiveness (even if not with the term). It is found in language, and computer programs, and even in children’s stories, as in (1).

In (1), the process or rule that creates a relative clause (here, the clauses beginning with that) has applied repeatedly to the noun phrases inside other relative clauses. This repeated application of the same rule to create more and more complex sentences is an example of recursiveness. The children’s story is amusing precisely because we all know that there is no theoretical limit to the application of this rule. Any speaker of English can add to the story by generating another relative clause at the beginning, as in (2).

The only limitations on the number of relative clauses are practical and notlinguistic: the speaker may run out of things to say, or out of breath, or time, or memory. It is because the rules of syntax are allowed to apply recursively that language is non-finite: there is no limit to its expressive power. It is important that this recursiveness applies to create structures which are embedded inside other structures, to create subordinate clauses, such as the that relative clauses in (2). These embedded subordinate clauses involve more complex structure than coordinate clauses, which are illustrated in (3).

Embedding distinguishes subordinate clauses of the sort exemplified by The House that Jack Built from simple coordination. A simplified diagram of the structure of recursive subordinate versus coordinate clauses is given in figure 22.1. The rule that makes a relative clause – put a sentence inside a noun phrase– produces a structure in which one sentence is inside another sentence. This is the key to recursion: by putting one phrase inside another of the same type, there is in principle no limit to the length of a sentence. Human language is not limited to simple conjunction of one phrase after another (like that in (3)); it has both coordination and subordination. If someone found a human language that allowed only coordination, not subordination, this would shake our fundamental assumptions about what human language is. Thus, it should be rather surprising that exactly this was claimed about ASL in the early days of its study. Thompson (1977) attempted to discover the mechanisms for subordination in ASL, and, not finding what he considered to be evidence for it, decided that it was systematically missing. If this is correct, then either the character and structure of human language is not as has been commonly assumed, or signed languages are significantly different from spoken languages, missing recursivity, which is often taken to be a defining property of language.

Later research has made it unnecessary to choose between these two puzzling options. Liddell (1980) argued convincingly that Thompson’s claim was incorrect, and Liddell’s analysis has been substantiated by many researchers since. Thompson had looked for several kinds of indications that ASL has subordination. For example, he looked for overt complementizers – an analog to English that in “I know that Susan will win” – and found none in ASL. He looked for relative clauses (like those in (1–2)) – and instead found sequences of signs as in (4).

He looked for clausal complements to adjectives like “happy” (cf. “He’s happy that she passed.”) and found instead unembedded sequences, as in (5).

In some ways, these and other observations of Thompson’s were correct. However, his conclusion that ASL had no grammatical means of embedding to create subordinate clauses was shown to be wrong. Liddell noticed that relative clauses are indeed grammatically marked in ASL, contrary to Thompson’s claims – not by function words such as that, but by non-manual grammatical markers. Liddell identified the non-manual marker of relative clauses as consisting of raised brows, a backward head tilt, and a tensed upper lip. This non-manual marker co-occurs with the material of the relative clause. In addition, Liddell argued that relative clauses in ASL are not like those of English. In particular, he showed that the noun that is modified by the relative clause occurs inside the clause in ASL, as in spoken languages like Diegueño, rather than outside it as in English. While in English the subordinating word that indicates a relative clause, in ASL the nonmanual markers that extend over the clause perform essentially the same role. A comparison of these aspects of relative clauses in ASL and English can be observed in (6).

In the notation, the line marked “rc” indicates the scope of the relative clause non-manual marker. The lower case subscripts are indices for DOG and CAT. The “a” index on DOG and COME is expressed formationally in the language, indicating that it was the dog and not the cat that came home. Clearly, the reason why Thompson thought ASL has no relative clauses was that he expected them to look more like English. However, once the characteristics of relative clauses in languages other than English are considered, it becomes clear that ASL does have relative clauses, formed by a rule of subordination that allows recursion.

Although the relationship between (7a) and (7b) is very similar to the relationship between (7c) and (7d), only the first pair are both grammatical. Making a question out of the noun phrase following a preposition is (usually) grammatical in (colloquial) English, but making a question out of a noun phrase which is coordinated with another noun phrase is not. In fact, the restriction on questions like (7d), which disallows extraction of material out of a coordinated structure, is virtually universal, and it has been proposed that a general universal constraint – the Coordinate Structure Constraint – prohibits it. Whether or not one accepts the hypothesis that such constraints are innately specified, the fact that they are not easily deducible from the input but appear to be ubiquitous in spoken languages makes them key exemplars of the human language capacity. It is therefore important to determine whether or not they hold for sign languages as well. If so, we may conclude that the constraints are truly universal, and that sign languages have the same properties as any other natural language. In fact, several researchers have argued that this is the case. The Coordinate Structure Constraint demonstrated in (7) provides the clearest example. Padden in 1988 shows that coordinate structures are allowed in ASL, as illustrated in (8).

Furthermore, as expected, ASL, like English, clearly prohibits violations of the Coordinate Structure Constraint, as illustrated in (9). The “t” on the line over FLOWER in (9) indicates a nonmanual marker for topicalization, the process by which the topic, FLOWER, is extracted and moved out of the coordinated structure, to the beginning of the sentence.

In several domains of syntax, the constraints proposed to be universal (including the Coordinate Structure Constraint) can be demonstrated to apply to ASL as well as to spoken languages – and it is expected that other signed languages will also show adherence to these constraints.