3.1.1 Response clauses

One adjunct that has been largely ignored in the literature is illustrated in (10). Where mentioned, these adjuncts have been called reason clauses (Betancort, Carreiras & Acuña-Fariña 2006), rationale clauses (Hornstein 1999), justification clauses (Stromdahl 2018), or simply gerundival complements to for (A. Williams & Green 2017). For clarity, and to distinguish them from RatCs, which have also been called reason clauses (A. Williams 2015), I will call the adjuncts in (10) “response clauses.” The reason is that these adjuncts entail that the act of the main clause was meant as a response to what is described by the adjunct (A. Williams & Green 2017; see also Fillmore 1971; Fabricius-Hansen & Sæbø 2011).

Some responses clauses may be understood as conceptual “arguments” of what I will call blame-class verbs, such as blame, praise, and thank, as in (11). The meaning of the verb blame, for example, requires that there is either an explicit or implicit offense associated with the blame (Fillmore 1971). This is not the case with the response clauses in (10). But although their conceptual relationship to the main clause predicate may differ, I assume that the response clauses in either case have the same syntactic status.

There is good evidence that response clauses allow logophoric NOC. In the sentences in (12), for example, controller choice is not limited to c-commanding arguments, as PRO is coreferential with an embedded possessor. If control of response clauses were strictly predicative OC, this should be impossible, since there is no way to form the necessary predication dependency between PRO and a non-c-commanding argument.

In addition, response clauses allow implicit control, which has been argued to be NOC (Manzini 1986; Kawasaki 1993; Landau 2017). In (13a), PRO may refer to the implicit recipient of the medal. Example (13b) shows the same thing, but uses a pronoun to remove the possibility that the response clause adjoins within the object DP.⁷

These examples illustrate that response clauses sometimes exhibit NOC. Importantly, there is also evidence that OC is sometimes possible. The ellipsis site in (14) includes a response clause. PRO in the elided clause requires “sloppy” interpretation (Ross 1967; 1969) under ellipsis, which is suggestive of OC (Pires 2007; Boeckx, Hornstein & Nunes 2010a).

OC is is also seen in response clauses that allow inanimate controllers. It has been argued that NOC requires a [+human] interpretation for PRO (Clark 1990; Landau 2013), which, under the TTC, follows from the logophoric nature of NOC (Landau 2017). And yet PRO in each of the response clauses in (15) allows an inanimate controller, which can only be OC.

In sum, control in response clauses must sometimes be OC, as evidenced by the fact that inanimate controllers are possible, and sloppy identity is at least sometimes required under ellipsis. But there is also evidence that response clauses allow NOC: implicit and non-c-commanding controllers are sometimes possible. This is consistent with the generalization that adjuncts allow both OC and NOC, and the explanation of the duality of temporal and rationale clauses given above based on Landau (2015; 2017) can easily be extended to response clauses. Because they are not selected by the matrix verb, response clauses are free to adjoin with either a predicative or a logophoric structure.

3.1.2 Object-gap purpose clauses

Another adjunct that allows both OC and NOC is the object-gap purpose clause (OPC). Purpose clauses contain one obligatory gap that is coreferential with the theme argument of the host clause (Jones 1991; Kawasaki 1993; Whelpton 1995). In OPCs, this gap is in a non-subject position.⁹ The subject may contain an additional empty category (PRO), which is often coreferential with the indirect object of the main clause, if one is present (16a), or with the matrix subject otherwise (16b). Lexical subjects are also possible (16c).¹⁰

Several authors have claimed that the dependency between PRO in an OPC and its antecedent is OC (e.g. Nishigauchi 1984; Kawasaki 1993; Landau 2013; Hallman 2015). Evidence that OC is possible is that only sloppy identity is available under ellipsis in examples like (17a), which can only be used to say that Chandler got his frying pan with the intent of cooking with it himself. Also, PRO can only have a bound interpretation when controlled by an only-DP in (17b), which entails that no one else bought a frying pan with the intent that they themselves would cook with it; it is still true in a situation where both Monica and Chandler bought frying pans with the intent that Monica would cook with them. Here one is used as the antecedent in order to remove the possibility of a relative clause analysis (see footnote 10).

Inanimate controllers are also possible (18), which again indicates OC.

But there is also strong evidence that NOC is sometimes possible. For example, PRO in an OPC can have a non-c-commanding or syntactically absent controller, as seen in (19), modified from Stromdahl (2018).

OPCs therefore allow both OC and NOC, similar to temporal and rationale adjuncts.

3.1.3 Absolutives

A final adjunct that allows both OC and NOC is the absolutive clause, which adjoins as a bare participial phrase, as in (20).

Pires (2007) argues that absolutive adjuncts are OC based on several tests. For example, PRO’s controller in (21a) must c-command it, and (21b) only allows a sloppy interpretation of the elided PRO.

Inanimate controllers are also possible (22), giving further evidence for OC.

However, non-c-commanding controllers are possible given the right context, which indicates NOC. In (23) and (24), PRO can refer to the speaker. This is especially true when the adjunct is left-adjoined, but it is also possible, if slightly more marginal, when it is right-adjoined.

Absolutive clauses therefore also conform to the generalization that adjunct control allows both OC and NOC. Interestingly, though, Landau (2015: 97) mentions in a footnote that bare participles do not project the functional structure needed to support NOC. Although this was only mentioned in passing without additional support, if it is true that absolutives are too small to project a logophoric structure, then the NOC exhibited in (23)–(24) would be surprising. I will return to the issue of adjunct size and its interaction with OC/NOC in §4.2.

3.1.4 Summary of OC/NOC adjuncts

The evidence that response clauses, OPCs, and absolutives allow both OC and NOC is summarized in Table 1, along with a summary of evidence on RatCs and temporal adjuncts found in Landau (2017) and Green (2018). OC is seen when PRO in these adjuncts has an inanimate controller, and when sloppy readings under ellipsis or bound readings with control by an “only”-DP are required. NOC is found when these adjuncts allow PRO to co-refer with a non-c-commanding antecedent, or to refer to an entity not present in the sentence at all, in what I broadly term “implicit control”.

3.3.1 Goal clauses

One adjunct that only allows OC is the goal clause (Huettner 1989; Landau 2013), illustrated in (27).

Although similar on the surface to RatCs such as (28), goal clauses differ in meaning. According to Huettner (1989), the matrix clauses in (27) constitute doing, or attempting to do, what is in the adjunct. For example, under the goal clause interpretation of (27a), Max works hard at staying out of jail, even if he does not literally work hard. With the RatC in (28a), on the other hand, Max literally works hard at some unspecified job because otherwise he will end up in jail. This is further illustrated in the contrast between (27b) and (28b). If the infinitive is a goal clause, then Max puts great effort into avoiding work; but when the infinitive is a RatC, the sentence can only have the strange interpretation that Max literally works hard, because otherwise he would not be able to avoid work.

Goal clauses adjoined to an intransitive matrix clause as in (27) only allow subject control of PRO. Neither (27a) nor (27b), for example, can be used to mean that Max worked hard at someone else staying out of jail or avoiding work. This suggests that OC is involved. Further evidence for OC in goal clauses is that they allow inanimate controllers, as illustrated in (27d) and (29), and that sloppy identity is required under ellipsis (30).

In contrast to RatCs and temporal adjuncts, goal clauses appear not to allow NOC. As the examples in (31) make plain, even when subject control is ruled out, and when the preceding context establishes an available logophoric antecedent, conditions that facilitate NOC in RatCs and temporal adjuncts (Landau 2017; Green in press), OC by a c-commanding argument of the next clause up is the only possible reading.

Implicit control by a passive agent, which Landau (2017) argues is NOC, is also impossible. The goal clauses in (32) are limited to subject control, in contrast to the RatC in (33), which allows implicit control. Because the in order in RatCs is optional, the string in (32b) is ambiguous between containing a RatC or a goal clause. Crucially, though, if PRO is understood to refer to the implicit agent, only a RatC interpretation is possible. The sentence can only be interpreted as the answer to the question of why the plant was cultivated in general (as opposed to being left uncultivated). This is in contrast to the goal clause reading of (32a), in which the plant is specifically cultivated toward the goal of being nutritious, in answer to a more specific question about which goal the plant was cultivated for (among other possible cultivation outcomes). The difference, though subtle, is important.

If temporal and rationale adjuncts allow both OC and NOC simply because their structure is not selected by the matrix predicate, and if goal clauses are similarly unselected, then the strict OC behavior of goal clauses is unexpected.

3.3.2 Subject-gap purpose clauses

The next adjunct restricted to OC is the subject-gap purpose clause (SPC). SPCs such as the one in (34) are similar to OPCs, but without the post-verbal gap.¹³

Jones (1991) categorizes SPCs as OC. And although he does not give a detailed analysis of them, Landau (2015) also states in passing that SPCs involve non-logophoric PRO, which can only be OC. Evidence that this is correct is that SPCs allow inanimate controllers (35), and that PRO requires a bound reading when controlled by an “only”-DP. Example (36) states that only that bookshelf was purchased with the intended purpose of its holding books; it is still true in a situation where the speaker bought other shelves, as long as they were not also meant to hold books.

SPCs do not allow implicit controllers, which would indicate the availability of NOC. The SPC in (37a) clearly allows control by the direct object of donate. But even though that object can be left implicit (37b), control by an implicit object is impossible; (37c) is ungrammatical as an SPC (although it could have a RatC interpretation, in which case the slave driver would control PRO).

Implicit control could potentially be ruled out for other reasons, since even in adjuncts that allow implicit control in some instances, its availability can be restricted (Landau 2017; Green in press). However, non-c-commanding controllers are also ruled out, in contrast to what was seen for OPCs in (19), providing further evidence that NOC is impossible.

Like goal clauses, SPCs appear to counter the hypothesis that unselected adjuncts should allow both OC and NOC.

3.3.3 Telic clauses

A final adjunct that is restricted to OC is the telic clause, illustrated in (39), which express the outcome of the event of the main clause (Whelpton 2001).

These adjuncts, also known as “outcome” clauses, have received little attention in the literature. Where they have been discussed, they have been classified as requiring OC by the matrix subject (Huettner 1989; Whelpton 1995; Landau 2013). That OC is possible is made evident by (39b), which has an inanimate controller, as well as by (40), which requires sloppy identity under ellipsis.

That NOC is not permitted is demonstrated in (41). In each of these sentences, control by non-c-commanding arguments or implicit entities is impossible, despite the strange meanings subject control results in.

Since they allow inanimate controllers and require sloppy identity under ellipsis, and because they do not allow implicit or non-c-commanding controllers, telic clauses appear to require OC.

5.1.1 OC preferred over NOC

According to Hornstein (1999; 2001; 2003), pronominalization resulting in NOC is more “costly” than OC by movement, and is only possible as a last-resort option when movement is blocked. However, the last-resort status of pronominalization as originally formulated is too strong, as it would rule out any case of NOC in adjuncts in favor of OC by the matrix subject, which could arise through sideward movement. Boeckx & Hornstein (2007) acknowledge a similar problem. If pronominalization were strictly last-resort, then (60a) should be ruled out due to the availability of (60b), which Hornstein & Kiguchi (2003) argue arises through movement.

To allow both sentences, Boeckx & Hornstein (2007) propose that pronominalization is only ruled out when movement could not establish the same interpretation (see also Boeckx, Hornstein & Nunes 2010a). Thus, NOC is allowed even when OC is available, as long as NOC results in an interpretation that OC could not have.¹⁸ However, OC may also be preferred more generally. Boeckx & Hornstein assume that parsers are transparent with respect to grammars, meaning that if a grammar has a preference, then the parser will respect that preference. When a comprehender reaches a gap in the incremental processing of a sentence with a control structure, they will prefer to posit a trace of movement if possible rather than a null pronoun, all things being equal. The pronominalization and interpretation in (61b) is not ruled out by the grammar, but because parsers would prefer to posit a trace as the null subject of the adjunct, resulting in (61a).

There is therefore a difference between a derivation being ungrammatical because the grammatical system cannot produce it, and one that is unacceptable because parsers would never choose that derivation over a preferred alternative. The strong preference for OC in adjunct control is due in part to a grammatical constraint requiring OC if it is possible for a given interpretation, and in part due to a parsing preference for positing traces over null pronouns in incremental processing. This combination results in a general preference for OC over NOC.

Boeckx & Hornstein (2007) and Boeckx, Hornstein & Nunes (2010a) argue that parsers are completely transparent to grammatical constraints, and that if it is possible to posit a trace in a given position, parsers must do so. This would explain why examples such as (62) only allow OC by the matrix subject, even when it results in a strange interpretation. Upon encountering the empty category in the adjunct, it is possible to posit a trace linked to the matrix subject, so parsers must do so.

However, this does not explain why the same is not true in any of the examples in (63). Evidently, parsers are able to override the preference for a trace in at least some circumstances, even when a trace in the relevant position would be possible. This cannot be due to structural properties of the sentence alone, since the sentences in (63) permit either a trace (OC) or a pronominal (NOC) without changing the structure (see Green in press). Instead, structural, discourse, and processing factors all compete in determining the probability of OC or NOC.

An example of the interaction of such constraints is given in Boeckx & Hornstein (2007), where it was argued that NOC is available in (60a) despite the availability of OC in (60b) because at the point where the empty category is encountered, if parsers posited a trace, they would not yet be able to assign it a referent; the only positions that something could have moved to from that position come later in the sentence. The preference to quickly assign reference to dependent elements (Nicol & Swinney 1989; Badecker & Straub 2002) is in competition with the preference for a trace, resulting in the possibility of either a trace or a pronominal. Similar competition may be involved in NOC cases of adjunct control.

5.1.2 Difference between complements and OC/NOC adjuncts

The MTC assumes that OC is preferred over NOC due to a grammatical constraint that rules out NOC for any interpretation that could have arisen through movement, and to a parsing preference rules out NOC in general when OC is available. For some adjuncts, this preference can be overridden in the proper context. But the same is not true for complements. Why should complements and adjuncts differ in this respect under the MTC? The only difference in the derivation of complement control and adjunct control is that the former involves upward movement and the latter sideward movement. I propose that adjuncts are more likely to permit NOC than complements because sideward movement is for some reason more “costly” than upward movement. This would lead to the preference scale for the derivation of control represented in (64).¹⁹

The additional cost of sideward movement reduces the parser’s preference for traces over pros, which in turn makes it more likely that NOC will be permitted if contextual or other constraints favor it. But what is the source of the preferences in (64)? Why would sideward movement be more costly than upward movement? As discussed above, the preference for OC over NOC is in large part due to processing. This accounts for why NOC is lowest on the scale. Might the preference for upward versus sideward movement also be the result of processing biases? Perhaps sentences involving sideward movement are simply more difficult to process than sentences with upward movement. One reason why this might be the case begins with the fact that adjuncts are optional. Because they are optional, they may be less predictable. And if they are less predictable, a comprehender will have less of an expectation, at the matrix verb, of needing to form a control dependency, than when the verb selects a control complement. But even if this is true, why should online predictability have anything to do with what interpretations are licit or available offline? There is no logical connection here, so the effect would simply have to be stipulated. Unlike the preference for OC over NOC, this effect cannot be reduced to other known parsing preferences, such as the preference for local versus distant antecedents or the preference for syntactic versus discourse dependencies. Therefore, stating that (64) is due exclusively to processing preferences is tenuous.

It is also not immediately clear what grammatical constraints would result in (64). Both upward and sideward movement involve copy and re-merger of an element of the growing syntactic structure. Neither involves more steps or even different underlying operations that could lead to a greater cost. There is only one real difference between the two that could cause a cost difference: sideward movement involves copying an element from one tree and moving it to the root of another (65a), while upward movement involves copying from and re-merging to the same tree (65b).

(65)	a.
	b.

Perhaps sideward movement incurs a derivational penalty akin to the one that has been posited for moving out of islands (Chomsky 1972; Lasnik 2001). This could, perhaps, simply be the result of the fact that sideward movement requires forming a syntactic dependency between two separate workspaces, so to speak, in the derivation. Or it could have something to do with constraints on proper movement. Although I do not speculate further on the exact nature or source of this penalty, let us simply suppose by hypothesis that there is a penalty for moving an element from within one tree to the root of another, unconnected tree. Suppose, furthermore, that this penalty is weak,²⁰ and does not cause the derivation to crash, but instead results in a lower “degree of grammaticality” or acceptability (Epstein 1990; see also Chomsky 1965; 1986; Lasnik & Saito 1984). Because both upward and sideward movement can lead to convergent derivations, both may result in grammatical strings; those involving sideward movement are simply lower on the scale between completely acceptable by the parser and completely unacceptable.

If this hypothesis is on the right track, the preference for upward OC versus sideward OC is due to a weak grammatical constraint against movement to a non-c-commanding position. Assuming there is some performance penalty associated with this grammatical violation, the lower grammatical status of sideward OC makes its appeal relative to NOC weaker than upward movement’s. This in turn leads to a greater probability of both OC and NOC being accepted in OC/NOC adjuncts than in complements, since adjuncts involve sideward movement, while in complements, movement is upward.

Admittedly, positing that sideward movement is more costly than upward movement is somewhat stipulative. But something like it as part of the scale in (64) would be required in order to capture the differing availability of NOC in complement and adjunct control under the MTC. And interestingly, the idea that sideward movement results in a greater likelihood of NOC has independent support in the examples from Boeckx & Hornstein (2007) in (60), repeated in (66). This structure allows both OC and NOC, and importantly, Boeckx & Hornstein argue that the OC derivation in (66b) involves sideward movement. This is consistent with the idea that OC derived from sideward movement is less strongly preferred to NOC than OC derived from upward movement is, and that this contributes to NOC also being possible.

Finally, Hornstein (2001) assumes an Uriagereka (1999)-style adjunction in which adjuncts are linearized and therefore become islands before being merged with the main clause (see also Nunes 2001). If this is the case for all adjuncts, then all cases of OC in adjuncts will necessarily involve sideward movement. And because sideward movement is “costly”, NOC through pronominalization should be possible in the right context. Under these assumptions, the MTC makes the incorrect prediction that all adjuncts should allow both OC and NOC. The MTC therefore does just as well as the TTC on adjuncts that allow both OC and NOC, while struggling with strict OC adjuncts. In the following section, I outline a proposal for how the MTC can deal with low-adjoining strict OC adjuncts.

5.3.1 Speaker-oriented adverbials

It was noted above that PRO in SOAs such as those in (77) can only refer to the speaker, which is evidence for their being restricted to logophoric NOC.

The explanation offered for the TTC was that since SOAs adjoin high in the clause, above the position of the matrix subject, as in (78), the adjunct is too high for a sentential argument to predicatively control PRO. Logophoric control is therefore the only option. However, if having a small clause size results in strict OC, as briefly suggested in Landau (2015) and as explored above, then SOAs still present a problem for the TTC, since they can adjoin as bare participials and still exhibit strict NOC.

(78)

Intuitively, it seems that a height-based explanation of the strict NOC status of SOAs should be able to apply to the MTC. With the structure in (78), PRO is not c-commanded by any matrix argument. Movement out of that position to the main clause would then result in an ill-formed chain, since the head of the chain would not c-command the tail (Chomsky 1995). Although the same is true initially for all cases of sideward movement, here the problem is more serious, since even at the end of the derivation, the final position of the moving element does not c-command the trace in the adjunct.

However, Hornstein & Kiguchi (2003) argue that c-command is not required for sideward movement. They argue that examples like (79) involve OC via sideward movement, and yet PRO’s controller does not c-command it, even on the surface. And this is not limited to psych-verbs, as might be assumed based on theories such as that of Belletti & Rizzi (1988), since PRO in (80) is also not c-commanded by its OC antecedent (Norbert Hornstein, p.c.).

If OC does not actually require c-command, then it is not clear why in SOAs the matrix subject should not be able to sideward move from the adjunct into the main clause, as in (81).

(81)

Importantly, however, it is also unclear how the TTC would explain the OC properties of (79) and (80). If predication requires c-command, then predicative OC should be impossible. Logophoric OC would also be insufficient, since it is unclear what would restrict the logophoric context of the non-finite clause to one representing Mary’s perspective. In fact, the sentence could easily be representing the opinion of a separate individual, such as John in (79). Therefore, if Hornstein & Kiguchi’s arguments are right, and (79) and (80) do exhibit OC, then the TTC must also admit that OC does not require c-command, leaving no good explanation of SOAs. If, on the other hand, Hornstein & Kiguchi are incorrect, and these examples do not involve OC, then the same explanation proposed for the TTC will work in the MTC, namely that SOAs adjoin too high for OC to occur.

Under either account, the restriction of SOAs to NOC may also be due to other factors. One possibility is that there is a functional head high in the structure where SOAs adjoin that requires the adjunct to represent the perspective of the speaker (Cinque 1999). Unless the matrix subject is also the speaker, this would require NOC in order to satisfy the demands of that functional head.

5.3.2 Telic clauses

Finally, telic clauses likely involve sideward movement, and yet appear to be restricted to OC. Strict OC in an adjunct is only expected under the MTC if that adjunct attaches lower than the base position of the controller and is not linearized prior to adjunction, thus allowing control via upward movement. But as discussed in §4.1, Whelpton (1995) provides evidence that telic clauses adjoin at the IP level, well outside the merge position of the subject within the vP. Because of this, the only way for OC to occur would be through sideward movement, as in (82).

(82)

Because sideward movement is required, the account presented here predicts that NOC should be possible in telic clauses given the right context. In order to account for the strict OC status of telic clauses, the MTC would either have to say that they attach lower than Whelpton (1995) thought, claim that NOC is ruled out for some other reason, or provide evidence that has eluded me that NOC is possible. Importantly, though, the TTC account in §4 is also unable to account for telic clauses, since there is no way that they could be arguments of the matrix predicate, and they appear to be at least as large as a CP, which should be sufficient to allow a logophoric structure. I leave further investigation of telic clauses to future research.