This Appendix discusses two issues in more detail: the category of adjectival and verbal passives, and adjectival passives that are apparently derived from active perfect participles.
A.1 The categorial status of passive participles
Throughout this paper I have taken adjectival and verbal passives to differ in category. However, there is another possibility, which is that all
passive participles in English (and other languages) are actually adjectives. This possibility has been suggested by Freidin (1975
), Emonds (2006
), and Lundquist (2012
), among others. The idea is that verbal passives really only appear following two auxiliary verbs, be
, and these two verbs also permit adjectival complements, including adjectival passives:
According to this hypothesis, then, all
passive participles are actually adjectives. Contexts that disallow apparent verbal passives must do so for reasons other than category selection. For instance, adjectival passives are stative, while verbal passives are eventive; it might be that verbs like seem
require stative AP complements and disallow eventive AP complements. (Of course, some explanation would be necessary for why double object APs must be eventive, why unaccusative APs must be stative, and so on.)
) adds to the distributional argument, arguing that several other verbs also select passive phrases, including the promoted object. These are have
, and hear
The players had [the ballboys sent boxes of chocolates for Christmas].
He wants [the table given a thorough cleaning].
She needs [her assistant brought a new smartphone].
You may see [the prisoners given their mid-day meal].
The players heard [insults shouted at them by irate fans].
Emonds argues that the bracketed phrase in each example is a passive phrase, with an underlying object promoted. (I try to use double object verbs, to show that these are verbal passives, and not, for instance, AP small clauses.) According to Emonds, the distributional generalization is that passive phrases can appear in any context that selects either AP or [NP AP]. The verb get
allows either one:
We got [Harry sick]. (NP AP)
Harry got slipped a mickey. (Passive)
We got [Harry slipped a mickey]. (NP Passive)
All of the verbs in (122) also allow [NP AP], for instance She wants the table clean (for the party)
This generalization is not correct, however. There is one (semi-grammatical31
) verb, make
, which allows adjectival complements ([NP AP]), but does not permit a verbal passive:
Such heavy use made [the table dirty].
* I made [Baghdad approached].
* I made [her assistant brought a new computer].
If it were true that passive phrases are permitted wherever AP or [NP AP] is, they should be permitted with make
The correct generalization is that verbal passives can appear in the same position where an active, eventive VP headed by V-ing
can appear. Be
, as do all of Emonds’s verbs:
Harry is handing people leaflets.
The players had [ballboys pitching them curveballs].
The coach wanted [the players throwing each other curveballs].
She needs [her assistant sending people bouquets of flowers].
From here, you can see [spies slipping their contacts secret documents].
The players heard [irate fans shouting insults at them].
In contrast, make
permits neither V-ing
nor passives (though it does allow a bare V), but it does allow adjectives, as previously shown:
Such heavy use made [the table disgusting]. (Adjective)
I made it disgust you. (Bare V)
* I made it disgusting you. (V-ing)
* I made him running around. (V-ing)
* I made [Baghdad approached]. (Passive)
* I made [her assistant brought a new computer]. (Passive)
The real generalization is therefore the following:
Verbal passives are permitted as the complements of verbs that also permit V-ing.
This generalization also covers do
, which allows neither V-ing
nor adjectives, and also does not allow passives.
I contend that the explanation for (128) is that verbal passives are truly verbal, just like (at least some) phrases headed by V-ing. Verbal passives, just like verb phrases headed by V-ing, are selected by verbs which select VPs. (Some verbs impose an additional restriction, selecting only bare VPs; these do not select V-ing or passive VPs.)
In addition to the distributional argument, Emonds (2006) and Lundquist (2012) point to agreement in languages that have agreeing adjectives as an argument that passive participles are adjectives. For instance, in many Romance languages, passive participles agree with their subject in gender and number but not person, just like adjectives and unlike finite active verbs. However, in at least some of these languages, active perfect participles may also agree on the same pattern (e.g., Kayne 1989), so I take this fact only to indicate that verbs and adjectives may share an agreement pattern.
There are other differences between adjectival and verbal passives besides the distributional one that also point to a category difference. For instance, adjectives and verbs differ in their behavior with how
: adjectives must pied-pipe how
, and how
only questions degree; but verbs may not pied-pipe with how
, and how
questions manner. If how
does not pied-pipe with an adjective, the only possible reading is a request for an explanation:
How was he defeated? (questions manner)
How defeated was he? (questions degree)
How was he passed the secret plans? (questions manner)
* How passed the secret plans was he? (pied-piping impossible with V)
How was he unfazed? (only: requests explanation)
How unfazed was he? (questions degree)
The most straightforward account of this difference depends on category, and again points to a category difference between verbal and adjectival passives.
Adjectival and verbal passives also differ in coordination: adjectival passives can be coordinated with underived adjectives, but verbs cannot be (Lundquist 2008
). Verbal passives cannot be coordinated with adjectives, either:
He made Bond angry and unconvinced that we were right.
* He made Bond angry and hand Xenia the secret plans.
* He made Bond angry and given a sedative.
Apparent examples of coordination of adjectives and verbal passives, like (131a) below, are probably coordination of larger categories, for instance some kind of non-finite clausal category. An auxiliary can be included in the verbal part of the coordination, for instance (131b):
With Bond semi-conscious and given a dose of truth serum, …
With Bond semi-conscious and being given a dose of truth serum, …
The second conjunct in (131b) could not possibly be an AP. Where such larger categories are not possible, as in (130c), coordination of verbal passives with underived adjectives is ungrammatical. Again, I take the simplest explanation for this difference to be a category difference.
According to Emonds (2006), there is also a difference between adjectival and verbal passives in Dutch, in their behavior in final verb clusters. Once again, the simplest explanation says that verbal passives are verbs, but adjectival passives are adjectives.
If all of this is correct, there is a real distributional difference between adjectival passives, which have the distribution of APs, and verbal passives, which have the distribution of VPs. It is not correct to view all passive participles as adjectives.
A.2 Apparent active participles
Levin and Rappaport (1986
) hypothesized that adjectival passives derived from unaccusative verbs were derived from active perfect participles instead. These were discussed in the main text. Bresnan (1995
) claims that active perfect participles can generally be turned into adjectives, subject to various constraints, citing the following examples:
a confessed killer (= a killer who has confessed)
a recanted Chomskyan (= a Chomskyan who has recanted)
(un)declared juniors (= juniors who have (not) declared [majors])
an unbuilt architect (= an architect who has not built [buildings])
a well-prepared teacher (= a teacher who has prepared well)
a practiced liar (= a liar who has practiced)
According to Bresnan, these have the meanings indicated in the parentheses, and so must be derived from active perfect participles, as in the paraphrase.
Embick (2004a:note 6) dismisses these as special meanings only available when the stativizing head attaches directly to the root, but his evidence for this, the claim that these meanings disappear with un-prefixation, is not correct (examples below). Moreover, at least one class of these is quite productive, as I will show. For these reasons, I do not think such examples are so easily dismissed, and they must be accounted for.
However, there is much reason to doubt that English generally allows adjectives to be formed from active perfect participles. Alongside the examples above, close synonyms and closely analogous phrases are ill-formed (see also the ill-formedness of unergative adjectival passives in (68) above):
* a (well-)readied teacher (cf. a well-readied meal)
* an exercised athlete (cf. an over-exercised dog)
* an unbuilt inventor, *an unwired electrician, *an unprogrammed computer scientist
I therefore conclude (with most of the literature) that it is not possible to form adjectives out of active perfect participles. (I assume that this is because they are formed by adding a higher aspectual projection, above Voice, and the Adj head cannot combine with this aspectual head; it only combines with Voice.) This leaves Bresnan’s examples above to be accounted for. As a start, I will divide these into different categories, adding additional examples where possible. Bresnan’s first three examples fall into the first class, which consists of verbs of declaration. These quite productively form such adjectives, and some can be prefixed with un-
Participles based on verbs of declaration
a(n) (un)confessed killer
a(n) (un)recanted Chomskyan
a(n) (un)declared candidate
a(n) (un)committed evangelical
Bresnan’s suggested paraphrase for these (‘a killer who has confessed’) is not quite right, however. A confessed killer
actually means ‘someone who has confessed to being a killer’. That is, in some way (namely, as something predicational or propositional), killer
is the internal
argument of the participle, not the external argument. This is made abundantly clear by the following sentence, found on the internet:
Heck, I’m actually a recanted-Catholic who became a born-again Christian, only to recant that one too a few years later.
That is, what is recanted is being a Catholic/born-again Christian
. The same holds for confessed
; what is confessed is being a killer
Additionally, if Bresnan’s suggested paraphrase were correct, we would expect the adjective to be able to appear in predicate position, but this is not correct for most of them:
* She is confessed/recanted/avowed/declared/sworn/admitted.
She is (un)committed/undeclared (in academic sense).
It is also telling that most of these can be prefixed with self-
, with essentially the same meaning:
a self-committed do-gooder
a self-declared enemy of immorality
?? a self-recanted Chomskyan
a self-avowed Christian warrior
In fact, some participles take on this use just when prefixed with self-
Prefixation with self-
is telling because most other past participles with self-
are clear passive participles, with the internal argument externalized:
self-caused, self-controlled, self-deceived, self-educated, self-engrossed, self-generated, self-governed, self-guided, self-insured, self-motivated, self-obsessed, self-ordained, self-propelled, self-raised, self-satisfied, self-taught
The prefix self-
seems in these cases to attach to an adjectival participle, formed as described above, and specify that the implicit external argument is identical to the internal argument.32
If this is correct, then that is probably what it is doing in self-confessed
and the other examples in (137), and the externalized NP is the internal argument, not the external one as Bresnan took it to be.
I suggest that examples like a
are similar to adjectival passives like alleged
, analyzed above. Confess
takes a nominal small clause, and both the NP subject and the NP predicate move, as described above for alleged
. Most of these verbs can, in fact, take ECM/raising or small clause complements:
He confessed himself a killer.
* He recanted himself (being/to be) a Chomskyan.
He declared himself a biology major.
He committed himself *(to being) an evangelical.
He avowed himself a communist.
He swore himself (to be) an enemy.
He admitted himself a murderer.
The only difference between the confessed class and the alleged class is that with the confessed class, it is usually understood that the external argument is the same as the subject of the small clause, and this can be explicitly marked using self-. (But note that this is not necessarily the interpretation; I have the intuition that with phrases like a confessed/admitted murderer, the person could have been identified as such through someone else’s confession or admission.) I suggest that when the reflexive internal argument is missing from verb phrases headed by these verbs (e.g., He confessed that he was the murderer), the verb has been reflexivized in some way, perhaps in the same way that is accomplished with inherently reflexive verbs like wash and shave.
The analysis I am suggesting for confessed, then, is exactly the same as for alleged, given above. There is a nominal small clause, and both the predicate and subject of that small clause move and abstract over the structure they attach to. The noun (e.g., killer) forms the property argument, while the external argument of the resulting NP is interpreted as the subject of the small clause. The only difference is that confess is usually understood reflexively, so that the implicit external argument is taken to be the same as the subject of the small clause.
If this analysis is correct, it explains why these adjectives cannot be used predicatively (*She is confessed): as was the case with the alleged class, above, the adjective must combine with a noun in order to satisfy its property argument.
This accounts for three of Bresnan’s six examples, which turned out to belong to a productive class. They are actually like other adjectival passives, and are built from passives, not active perfect participles.
As for an unbuilt architect, I think that the analysis suggested by Caroline Heycock in Bresnan’s footnote 8 (page 14) is actually correct. According to this suggestion, ‘architect’ stands in for ‘the architect’s designs’, the same way They never play that composer anymore actually means ‘the works by that composer’. This would make an unbuilt architect exactly parallel to an unpublished author, which means ‘an author whose works have not been published’, and not ‘an author who has not published anything’. An unbuilt architect may have built many things, the same way an unpublished author may have published many things in a previous career as a publisher; what is important is that no one has built the buildings that architect has designed. (In fact, architects typically do not build their own designs—builders do.) In other words, this participle is passive again, and not active: architect (and author in unpublished author) corresponds to the internal argument of the verb. If one could generally form adjectival passives from active perfect participles like this, it should be possible to say *an unbuilt inventor or *an unwired electrician, but these are not possible.
This leaves only prepared and practiced. Prepared may also be reflexive, like confessed. That is, it may be derived from prepare oneself. If this is correct, it is on a par with inherently reflexive verbs like wash and shave, which can be used as reflexive adjectival passives (an unwashed/unshaven man), as discussed briefly above. As for examples like a practiced liar, I suggest that the noun is the internal argument again. That is, a practiced liar is someone who has practice as a liar. A practiced eye is an eye that has been practiced (or made to practice). This would make practiced another instance of an adjectival passive with a missing active input, since one cannot practice an eye or a liar in the relevant sense (see Sect. 7.2 on missing inputs). I suggest the same analysis for experienced in an experienced sailor, which seems quite similar to practiced but does not have an active perfect paraphrase (*a sailor who has experienced).
If all of this is on the right track, adjectives apparently formed from active perfect participles are all actually adjectival passives, and are not active. Active perfect participles cannot generally be turned into adjectives, and there are plausible analyses for all of these cases that treat them as passives.