A Companion to Chomsky. Группа авторов
and (30) are inherently different in meaning. In (29), Kate is understood as the agent of helping Meghan event, but not of the seeming to help Meghan event (in fact, the latter event has no agent whatsoever). In (30), however, Kate is understood as the agent of both the helping Meghan and the promising to help Meghan event.
1 (29) Kate seems to help Meghan.
2 (30) Kate promises to help Meghan.
This difference in meaning suggests that seem does not assign an agentive thematic role to its subject Kate, whereas promise does. Because this is confirmed by various independent syntactic tests (see Landau 2013), Rosenbaum (1967), Postal (1974), and many researchers since them assume that this difference between seem‐ and promise‐type predicates is reflected in syntactic structure.11 The example in (29) is analyzed as derived by a transformation called raising that “moves” Kate from the subject position in the embedded clause (i.e. immediately preceding to) to the subject position in the root clause (i.e. immediately preceding seems) (31). The gap in (31) is therefore a trace of the position that Kate occupied before raising occurs (see endnote 4).
| (31) | Kate seems tKate to help Meghan. | (where tKate is the trace of Kate's raising) |
| (32) | Kate seems PRO to help Meghan. | (where PRO is silent pronoun interpreted as “Kate”) |
By contrast, Rosenbaum treated the gap in (30) as an unpronounced nominal (later analyzed as a silent pronoun called PRO; see Chomsky 1981 for a thorough treatment), which enters into a nonlocal dependency with Kate and is understood as co‐referent with her, yet is not related to Kate through a derivational history (32). In the linguistic jargon, Kate controls the interpretation of PRO.
PRO and standard, pronounced nominals show a complementary syntactic distribution. However, many languages display silent pronouns that can occupy the same positions that pronounced pronouns can, such as the subject position of a finite clause (33). To distinguish them from PROs, these silent pronouns are referred to as “little” pros (Chomsky 1981). Proof that the “missing” subject in (33) is indeed instantiated syntactically by pro (rather than being merely pragmatically inferred) comes from the fact that it engages in a syntactic dependency with the verb for first‐person singular agreement.
| (33) | Δpro | sen‐i | gör‐dü‐m | [Turkish] |
| 2SG‐ACC | see‐PST‐1SG | |||
| “I saw you.” | ||||
Throughout this section, we have used the symbol Δ to represent gaps, which could give the impression that all gaps are atomic, and hence devoid of internal structure. Although some gaps are indeed atomic (e.g. null simplex syntactic heads, see Section 4.3.1), generative research has shown that many are nonatomic, and therefore have internal complexity. An instructive example comes from the domain of ellipsis. Ellipsis is a family‐resemblance term in generative linguistics, whose archetype is post‐auxiliary predicate ellipsis in English. This refers to sentences in which a predicate – often a verb phrase – is missing yet inferred (34). Evidence that the missing predicate is a syntactic phrase with internal structure comes from the discovery that these silent phrases can contain the tail of a nonlocal dependency (35) (see Haïk 1987 for detailed discussion). If the missing predicate were a simple, atomic syntactic element, then no nonlocal dependency would be established in (35) and the sentence would be unacceptable, contrary to observation.
1 (34) David should arrive on time and Fiona should Δ, too.
2 (35) I know who will arrive on time and also who won't arrive on time Δ.(where strikethrough represents the silent, elliptic predicate)
Beyond ellipsis, there is evidence that gaps in certain A′‐dependencies also have internal structure. This evidence comes from the phenomenon of reconstruction, which was mentioned briefly in Section 4.3.2. Reconstruction describes situations in which the top member of an A′‐dependency behaves as though it occupies the tail position of its dependency. Consider the second sentence in (36), in which the reference of her first PhD student is dependent on the reference of every professor (i.e. Professor Brown remembers Olivia, Professor Black remembers Davina, etc.). Recall from (1) that her must be c‐commanded by every professor for this co‐variant interpretation to be possible. On the surface, her first PhD student is not c‐commanded by every professor in (36). However, this phrase is engaged in an A′‐dependency with a gap that is c‐commanded by every professor, and is interpreted as occupying this position.
1 (36)
Although reconstructed interpretations can arise from different sources (Lechner 1998), it is uncontested that many reconstructed interpretations come from the fact that the gap in certain A′‐dependencies has internal structure, and is actually a verbatim yet silent copy of the A′‐dependency's top member (37) (Chomsky 1995 [1993]). Thus, the existence of reconstructed interpretations reveals that many syntactic “gaps” are merely standard yet unpronounced syntactic phrases.
1 (37) … but her first PhD student, every professor will remember her first PhD student.
4.3.4 Section Summary
By viewing generative syntactic research through the lens of nonlocal dependencies, we presented in this section approximately 35 enduring discoveries of generative syntax (depending on how one counts). Primarily, generative research has uncovered that (i) grammatical function is typically reflected in hierarchical structure, (ii) natural language is awash with nonlocal dependencies of various types, and (iii) there exist a variety of linguistic elements that contribute to meaning and participate in syntactic dependencies yet are not pronounced.
4.4 Conclusion
The foundation for contemporary generative syntactic research is a broad class of generalizations about natural language syntax that were (primarily) discovered between the late 1950s and the early 1980s and that (primarily) concern nonlocal syntactic dependencies and the hierarchical structure on which they are instantiated. These generalizations are “enduring” because they are cross‐linguistically robust, they can be organized into typologies that encourage reductivist analysis, and they serve as useful diagnostic tools. The fact that generative syntax has uncovered such a large number of generalizations in such a short span of time (relative to the progress made in the centuries that preceded its inception) is directly attributable to its conceptual and methodological foundations, which were established by Noam Chomsky in his earliest work on language.
Endnotes
1 1 Our intended readership is broad, extending