This chapter presents the major findings and theoretical advances in the area of Control. We describe the different types of control (complement, adjunct, obligatory, nonobligatory) and illustrate their profiles in several languages. It is shown that while certain features of Obligatory Control (OC) are common – nullness of PRO, nonfinite complements – they are not universal, hence should not enter its core definition. Comparing approaches to the choice of controller based on lexical meaning postulates with those based on embedding of speech acts, we conclude that the latter provide deeper insights into the core properties of OC. The fundamental semantic distinction between clauses denoting a property and those denoting a proposition proves to be important: It affects both the possibility of Partial Control in complements and the possibility of Non Obligatory Control in adjuncts. These insights are integrated in the Two-Tiererd Theory of Control, laid out in the final sections.

A series of recent works (Chomsky 2021, Chomsky et al. 2023, Chomsky 2024) offers a novel theory of Obligatory Control (OC), based on the operation Form Copy (FC), which applies in movement chains as well. In this article I offer a critical assessment of the FC theory, focusing on its empirical consequences. I argue that this theory faces significant challenges. First, it offers no satisfactory answer to the fundamental question of what makes control obligatory in many cases. Second, it ignores the mass of evidence that the controllee is a pronoun rather than a lexical NP. Third, it unnecessarily complicates phase theory. Fourth, it overgenerates structures in which the controlled copy is produced by movement (Internal Merge); and fifth, it cannot adequately handle Partial Control. For all these reasons, judgment as to the implications of (or support from) the FC theory of control for the Strong Minimalist Thesis must be suspended.  

Intensive research on Obligatory Control (OC) in the past 2 decades has revealed a rich crosslinguistic terrain of deviations from the classical format. Five types of noncanonical OC are surveyed here: Finite control, controlled overt pronouns, partial control, proxy control and crossed control. Each one is described and illustrated, paying attention to methodological difficulties in establishing its characteristic empirical signature. We then turn to a critical assessment of leading theoretical accounts of these phenomena, pointing to merits and faults, and indicating how they can be integrated with broader concerns of syntactic theory.

Recent studies reveal that the values of finiteness, tense, modality and polarity in a clause elided under sluicing may be distinct from their correlates in the antecedent clause. Focusing on CP ellipsis in Hebrew (an instance of Argument Ellipsis), we first demonstrate that it is distinct from both Null Complement Anaphora and (null) pronominalization, and then show that the values of force (declarative, imperative, interrogative) can be distinct between the antecedent and the missing clause as well. Possible mismatches are bidirectional, ruling out "subset" theories of identity in ellipsis and challenging certain accounts of the semantics of polar questions. Implications for the general theory of ellipsis are discussed and evaluated.

Argument Ellipsis (AE) is a productive process in Hebrew, but some arguments resist ellipsis—precisely those that do not denote individuals. This constraint, a reflection of a general constraint on variables in natural language, is captured if AE sites are descended from a pro element that is derivationally replaced by a constituent recoverable from the antecedent. This must occur after spellout (to escape pronunciation) but prior to LF (to allow overt subextraction). The proposed analysis integrates novel data as well as recent findings from studies of AE in East Asian languages, and offers a new derivational path to ellipsis, which invokes neither PF-deletion nor LF-copying.

Theories of argument ellipsis (AE) based on PF-deletion or LF-copying do not generate predictions as to possible constraints on the semantic type of the elided argument. Yet such constraints obtain, as documented in Landau 2023 (NLLT): Only <e>-type arguments can be targeted by AE. Focusing on quantificational arguments here, we show that when they yield readings expressible by <e>-type denotations, they may elide, but when they denote genuine generalized quantifiers, they may not. Utilizing the restricted range of interpretations made available by choice function binding and E-type pronouns, the analysis derives a number of peculiar scopal properties of indefinite NPs, quantifiers and exceptive phrases under argument ellipsis.

Advocating a verb-stranding-VP-ellipsis analysis for object-gap sentences, Andrew Simpson (“In defense of verb-stranding VP ellipsis”) argues that under negation the “adjunct reading” is missing because it depends on focal stress, which cannot be realized on unpronounced material. No such condition holds, I maintain, and the absence of the adjunct reading reflects a syntactic absence: argument-ellipsis sites contain no VP adjuncts. The adjunct reading emerges in some languages when the antecedent sentence is negative too, an inexplicable contingency for the verb-stranding-VP-ellipsis analysis; in fact, these constructions involve polarity ellipsis (of TP), in which VP adjuncts are included. However, this derivation is not available to all languages, explaining some crosslinguistic differences in adjunct readings under negation. Finally, an optional adjunct reading may emerge in affirmative object-gap sentences due to pragmatic enrichment, a process sensitive to context in ways that go beyond the predictions of the syntactic analysis advocated by Simpson.

Sentences like They tricked him into believing them and They charged him with abandoning them raise interesting issues for selection and control. We show that these two sentences exemplify two distinct classes, subsuming P-gerund constructions that are formed with seven distinct prepositions: implicative vs. nonimplicative constructions. The first class displays a cluster of restrictions, both syntactic and semantic, which are absent from the second class: It resists partial control or embedded lexical subjects, and it bans object drop and movement of the P-gerund phrase. The existence of these two classes, as well as their empirical profiles, follow from Landau’s (2015) theory of control and challenge alternative approaches.

Control in adjuncts involves a complex interaction of syntax, semantics, and pragmatics, which so far has resisted systematic analysis. This book offers the first comprehensive account of adjunct control. Extending the framework developed in my earlier book, A Two-Tiered Theory of Control, the present book analyzes ten different types of adjuncts and shows that they fall into two categories: those displaying strict obligatory control (OC) and those alternating between OC and nonobligatory control (NOC). It explains how and why adjuncts shift between OC and NOC, unifying their syntactic, semantic, and pragmatic properties.

It is shown that the split between the two types of adjuncts reflects a fundamental distinction in the semantic type of the adjunct: property (OC) or proposition (NOC), a distinction independently detectable by the adjunct's tolerance to a lexical subject. After presenting a fully compositional account of controlled adjuncts, we test and confirm the specific configurational predictions for each type of adjunct. We describe the interplay between OC and NOC in terms of general principles of competition—both within the grammar and outside of it, in the pragmatics and in the processing module—shedding new light on classical puzzles in the acquisition of adjunct control by children. Along the way, the book addresses a range of empirical phenomena, including implicit arguments, event control, logophoricity, and topicality.

Gribanova (2013) observes that when the antecedent clause for an object gap in Russian contains a disjunction of ditransitive VPs, the gap is interpreted as if it contained the disjunction, thus favoring a V-stranding VP-ellipsis (VSVPE) analysis over a simple Argument Ellipsis (AE) analysis. We show that VSVPE fails to predict the attested interpretation of the gap when the antecedent VP contains a conjunction. The data are fully explained by combining AE with a pragmatic (scalar) inference triggered in situations of indeterminate ellipsis resolution, leading to multiple “cycles” in the dynamic semantics.