The Adjunction Prohibition and extraction from non-factive CPs
There is an extraction asymmetry from the clausal complements of factive vs. non-factive predicates.
Factive CPs are weak islands for adjunct extraction (1), while argument extraction is not affected (2).
In this paper, I argue that the adjunct extraction asymmetry in (1) is caused by an extra CP level
associated with non-factives. I propose that an extra CP projection, call it cP, comes between matrix non-
factives and the lower CP, as in (3). This extra projection allows for adjunct extraction in the following
way. McCloskey (2005), following the ideas of Chomsky (1986), formulates the `Adjunction Prohibition'
in (4). Following the Adjunction Prohibition, adjunction to a CP that is directly selected by a matrix verb
(or adjective) is impossible, while adjunction to a CP selected by functional cP is possible. According to
the proposal in (3), CPs are directly selected by a factive matrix predicate, while non-factive predicates
select for cP, which in turn selects CP. Only in the case where there is a cP buffer between VP and CP
(3a) can we get adjunction. I argue that this is the escape hatch for adjuncts out of CP (5). From this
viewpoint, arguments and adjuncts move through different syntactic positions, giving an explanation for
asymmetries in movement possibilities. As arguments don't travel through the same syntactic position as
adjuncts in this account (arguments go through SpecCP, adjuncts through adjunction to CP), argument
movement from out of a factive or non-factive CP is not barred. This approach provides an explanation
for the observed argument/adjunct extraction asymmetry in (1) vs. (2).
Support for this approach comes from a number of sources; McCloskey (2005) shows that Subject
Aux Inversion is possible in Irish English polar questions and WH questions (6), as opposed to factive
predicates (7). Henry (1995) presents data from Belfast English where T-to-C appears to take place in the
complement of a non-factive verb, triggered by WH movement (8). This data is reminiscent of restrictions
on Embedded Verb Second (EV2) in Mainland Scandinavian, where EV2 (which is optional) is only
allowed under a non-factive predicate (9). The embedded verb has raised over negation, showing EV2
movement. The presence of an overt complementizer in (9a) followed by V2 movement has been widely
taken in the literature as evidence for a recursive CP structure, not unlike (3a). The structure in (3a) gives
a fairly straightforward way to account for the EV2 data - the complementizer in cP, V2 movement in the
lower CP. McCloskey (2005) argues that apparent cases of T-to-C movement like those in (6-9) are only
possible when CP is not directly selected. Following Rizzi and Roberts (1989), McCloskey argues that
head movement gives rise to selectional violations. Translated into a derivational system, head movement
can interfere (or interact) with selectional requirements that target heads. For McCloskey, the Adjunction
Prohibition arises from the same problem adjunction creates a complex syntactic object, interfering with
selectional requirements. (10) shows the Swedish factive island parallel to (1), with non-factive, non-
lexically-selected think allowing adjunct extraction, and factive, lexically-selected regret disallowing it.
McCloskey's Adjunction Prohibition and selectional requirement analysis, applied to the structures I
propose in (3), gives a unified explanation for the data above.
Having shown syntactic motivation for postulating the existence of cP, one would expect that there
might be a semantic effect as well. In the parallel structure of vP/VP, the semantic contribution of vP has
been analyzed as agent theta role assignment. I propose that the semantic contribution of cP is that its
presence or absence correlates with whether or not the embedded CP is evaluated as true or not. Progovac
(1994) proposes an operator (Op) in the CP field that licenses `unfixed truth-values'. In embedded
contexts, Progovac argues that Op must be selected by the matrix predicate, as in (11). In factive CP
constructions, where truth-values are fixed, I claim that there is no cP level (3b), while in non-factive CP
constructions cP is present (3a). Given (3), and Progovac's claim that Op must be selected by the matrix
predicate, I revise (11) and propose the structure in (12). The operator, in a way, serves as an `agent' of
truth responsibility. Unless there is a cP with its operator present, the embedded CP gets a default
(speaker-evaluated) factive interpretation. Op is responsible for licensing any lack of truth-value. Under
this analysis, the availability of T-to-C movement in (6) vs. (7) is not surprising. There is no truth-value
associated with the polar and WH question cases where we found inversion (6), so cP is present, buffering
head movement from interfering with selection. We can now categorize these three types of embedded
clauses into two groups: (A) those that are assigned a truth-value (factive CPs) (B) those that are not (non-
factive and question CPs). The present analysis correctly predicts the freer movement associated with (B).
(1) (a) How do you think that you behaved t ?
(b) *How do you regret that you behaved t ?
(2) (a) Who do you think that Bill likes t ?
(b) Who do you regret that Bill likes t ?
(3) (a) [VP think [cP [CP ...]]]
(b) [VP regret [CP ...]]
(4) Adjunction to a phrase which is s-selected by a lexical (open class) head is ungrammatical.
(5) (a) Howi do you [VP think [cP [CP ti that you behaved ti ]]]
(b) *Howi do you [VP regret [CP *ti that you behaved ti ]]]
(6) (a) I wondered was he illiterate. [Irish English]
(b) I wonder what is he like at all.
(McCloskey, 2005:2)
(7) (a) *I found out how did they get into the building. [Irish English]
(b) *The police discovered who had they beaten up.
(McCloskey, 2005:3)
(8) (a) They wouldn't say which candidate they thought should we hire. [Belfast English]
(b) I'm not sure which one I think should we buy.
(McCloskey, 2005:40; taken from Henry, 1995)
(9) (a) Rickard sa att han var inte hemma [Swedish]
Rickard said that he was not home
`Rickard said that he was not home.'
(b) *Rickard ångrade att han var inte hemma
Rickard regretted that he was not home
`Rickard regretted that he was not home.'
(10) (a) Hur tycker du att du uppträdde t ? [Swedish]
How think you that you behaved
(b) *Hur ångrar du att du uppträdde t ?
How regret you that you behaved
(11) I doubt [CP [C' that Op [IP anyone has come.]]]
(Progovac, 1994:67)
(12) I doubt [cP [c' Op [CP [C' that [IP anyone has come.]]]
References:
Chomsky, Noam. 1986. Barriers. Cambridge: MIT Press.
Henry, Alison. 1995. Belfast English and Standard English: Dialect variation and parameter setting.
Oxford and New York: Oxford University Press.
McCloskey, James. 2005. Questions and questioning in a local English. To appear in Cross-Linguistic
Research in Syntax and Semantics: Negation, Tense and Clausal Architecture. Raffaella
Zanuttini, Héctor Campos, Elena Herburger and Paul H. Portner, eds., Georgetown University
Press.
Progovac, Ljiljana. 1994. Negative and positive polarity: a binding approach. Cambridge University
Press.
Rizzi, Luigi, and Ian Roberts. 1989. Complex inversion in French. Probus 1:1-30.
Ross, John Robert. 1967. Constraints on variables in syntax. Doctoral Dissertation. MIT, Cambridge,
Massachusetts.