Notes On Lasnik "The Forms Of Sentences"

Mastery of language requires several different kinds of knowledge. One of these is the knowledge of syntax.

Syntactic Structure

"An initial assumption about syntactic structure might be that speech (like written English) is broken into sentences, and the sentences into words." This abstracts over phonology, but is not abstract enough to be useful to us. "A sentence cannot be analyzed as simply a sequence of words, but rather must be regarded as having a certain hierarchical structure as well." This hierarchical structure reflects categories over groups of words.

Constituent structure can be depicted with "upside down tree" graphs, called phrase markers, which can be generated by using reqrite rules called phrase structure rules. Knowledge of a language requires knowledge of such rules. Constituent structure permits phrase structure rules to extend language by using recursion to embed one kind of constituent into another. "By providing finite means for generating an unlimited number of sentences, recursive rules supply a crucial part of the answer to the fundamental question of the creative use of language."

Deep And Surface Structure

We still require further abstraction; phrase structure by itself is not enough. For example, proerties of particular words help determine syntactic well-formedness. For example, transitive verbs require direct objects, while such objects cannot be linked to intransitive verbs. "The large lexical category V is thus divided into smaller lexical subcategories, each with its own special privileges of occurrence. ... These subcategorization requirements are linked to thematic relations (notions like understood subject, understood object)."

Details, now. Why does the sentence "This problem, John solved" seem reasonably grammatical, when verbs like "solved" require an NP? It is because the phrase "This problem" seems to function as if it really were in the VP. This is a deep structure property; Lasnik here introduces the notion that each sentence will have two different kinds of phrase marker representations, one for its surface structure, the other for its deep structure. Surface structure is created by applying one or more transformation rules to the deep structure. Transformation rules convert one phrase structure into another.

A Case Study: Interrogative Inversion

One example of this is interrogative inversion -- when changing a declarative sentence into a question, there is wh- movement and reversal of order of the auxiliary verb. What rule governs this transformation?

First stab at a rule is "Beginning with a declarative deep structure, invert the first and second words to construct an interrogative." But this fails., because the rule does not depend only on word order, but must be sensitive to more abstract categoires.

Second stab at the rule is "Beginning with a declarative deep structure, move the auxiliary verb to the front ot construct an interrogative. This is fine, until you encounter a declarative sentence with more than one auxiliarry verb.

Third stab at the rule is "Beginning with a declarative deep structure, move the first auxiliary verb to the front to construct an interrogative." This is an improvement, but still fails for more complex sentences (i.e., sentences with embedded clauses).

"The right generalization is (a priori) much more complicated, relying on structured hierarchical organization." "Beginnint with a declarative deep structure, move the first auxiliary verb following the subject tot he front to construct an interrogative." "Like the failed structure independent characterizations, this rule demands analysis into words, and categorization of the words. But it additionally requires abstract structural analysis, of the sort discussed earlier in this chapter."

What about converting other kinds of declarative sentences into questions -- specifically, declarative sentences with no auxiliary verbs? A first glance might lead us to believe that quite a different rule is involved. But speakers of English disagree. Why? Basic poing -- in these sentences, the inflectional ending of the verb (which, for eg., sets its tense) moves similar to the auxiliary verb in the earlier examples. When this happens, the inflectional ending can't stand on its own, it has to be added to a verb that serve no other role than structural (i.e., the verb do!).

A complete account of this requires us to assume that an S is headed by Aux, just as NP is headed by N, and Vp is headed by V. "If Aux is indeed the head of S, then even a sentence without an auxiliary verb is expected to have an Aux. Further, if it is the inflectional ending that moves to the front...and if the inversion process actually fronts Aux, then it is reasonable to conldue that..the inflectional ending is the Aux."

So here we get "affix hopping: -- the inflection moves to the front of the sentence, it can't stand alone, so it is attached to the neutral verb do. "In the dialect we are considering, when the main verb of a sentence is be, the pattern diverges from that found with other verbs, such a sleep." In this case, the transformation causes it to be raised to the inflection. "To instantiate the connection between Infl and V, if V is be it raises to Infl (V-raising). Otherwise, Infl lowers to V (affix hopping).

The bottom line of all of this is that the same general rule applies, once the notion of Aux in every sentence is entertained, combined with the notion that inflectional endings serve the role of Aux in sentences that have no explicit auxiliary verb.

Further Evidence: Negative Sentences

This general notion, of Aux being the head of S, can be used to deal with lots of other phenomena. For example, it helps determine where the word "not" appears in negative sentences. The informal rule for this is that "in a negative sentence, not occurs immediately after the first auxiliary verb following the subject." Lasnik fleshes this out formally, but we don't need to know this kind of detail for this course.

VP Deletion

Another example dealt with by this appraoch is VP deletion, where a VP is deleted if it has an antecedent (that is, another VP to which it is identical). So, we say "Harry can swim, and Susan can too" instead of "Harry can swim, and Susan can swim too". The second VP (swim) is deleted.

VP deletion is senstive to the output of V-raising, in that only the material following a raised have or be deletes (see eg.s on page 305). What happens when there is not auxiliary verb? As with previous examples, "we wind up with a form of do in the target sentence."

Basic points of all of this:

  1. Sentences have abstract structure.
  2. Deep and surface structure are both required.
  3. Transformational rules are also required.

"These structures and processes were not accessible to direct inspection, but, as is common in scientific inquiry, were posited based on the role they played in accounts of facts that we were able to collect."


"On the one hand, sentences are far more complicated than they at first appear. ... On the other hand, the overall pattern must have a fundamental simplicity as far as learners and speakers of the language are concerned."

Pearl Street | "An Invitation To Cognitive Science" Home Page | Dawson Home Page |