A student analyzes 'John appears to be happy' and says: 'John is the subject of both appears and be happy, so this must be a control construction like tried to leave.' What is wrong with this analysis?
AThe student is correct — appear and try are both control verbs that assign thematic roles to their matrix subjects
B'Appear' is a raising predicate that assigns no thematic role to its subject position; John originates as the subject of the embedded clause and raises to matrix subject only to satisfy the EPP — there is only one argument, not two
C'Appear' has no subject at all — it is an impersonal verb like 'rain'
DThe error is that 'happy' is an adjective, not a verb, so no movement is possible
The surface similarity of control and raising — both have a matrix NP followed by an infinitival complement — conceals a deep structural difference. In 'appear to be happy,' appear takes no external argument; it is semantically vacuous with respect to its subject. John is not the agent or experiencer of appearing — he is simply the subject of the embedded predicate (be happy) that has moved to the matrix subject position to satisfy the EPP. In 'try to leave,' John IS the agent of trying — try selects a thematic subject. The student's error is treating surface subject position as evidence of an argument relationship, when in raising the matrix subject position is just a structural host.
Question 2 Multiple Choice
Which test most reliably distinguishes a raising predicate from a control predicate?
ARaising predicates require the infinitival marker 'to'; control predicates use bare infinitives
BOnly control predicates can be passivized
CExpletive substitution: replacing the matrix subject with 'it' or 'there' is grammatical with raising predicates (which assign no thematic role to the subject) but ungrammatical with control predicates (which require a referential agent)
DRaising predicates occur only in subordinate clauses; control predicates occur only in matrix clauses
The expletive test works because raising predicates take no external argument — their subject position is a structural requirement (EPP), not a thematic slot. 'It seems that John left' is grammatical; 'There seem to be problems' is grammatical. But 'It tried that John left' and 'There tried to be problems' are ungrammatical because try requires a referential agent in its subject position — a non-referential expletive cannot satisfy a thematic requirement. This test is reliable across constructions and languages because it probes whether the matrix predicate genuinely selects a semantic argument in subject position.
Question 3 True / False
In 'There seem to be many problems,' the expletive 'there' can appear as matrix subject because 'seem' is a raising predicate that assigns no thematic role to its subject position.
TTrue
FFalse
Answer: True
This is a direct application of the raising predicate's defining property: it takes no external argument and assigns no thematic role to its subject. When the EPP (the requirement that TP have a specifier) must be satisfied but there is no NP to raise from the embedded clause in a convenient position, an expletive can fill the slot. 'There seem to be many problems' is grammatical for exactly this reason. The real argument (many problems) remains in the embedded clause. If you tried the same with a control verb — 'There tried to be many problems' — you get ungrammaticality because try demands an agentive referential subject.
Question 4 True / False
In the control construction 'John tried to leave,' John bears mainly one thematic role — agent — because 'try' and 'leave' together form a single complex predicate with a single argument structure.
TTrue
FFalse
Answer: False
Control constructions involve two separate predications, each with its own argument structure. John is the agent of trying (try selects an external argument with the agent role) AND the agent of leaving (leave selects an external argument with the agent role). The null pronoun PRO in the embedded clause's subject position is what the control relation targets — PRO must be coreferent with the matrix subject John. This is why control is analyzed as involving two predicates and two separate semantic role assignments, not a single complex predicate. The two-predicate analysis is confirmed by the fact that try and leave can each independently select different thematic properties.
Question 5 Short Answer
Explain why idiom chunks can appear as the raised subject in raising constructions but not as the controlled subject in control constructions, and what this reveals about the structural difference between raising and control.
Think about your answer, then reveal below.
Model answer: Idioms like 'the shit hit the fan' or 'the cat is out of the bag' have their idiomatic meaning only when the entire idiomatic string is interpreted together — you cannot extract a piece and reassign its reference while preserving the idiom's meaning. In raising, the NP 'the shit' originates inside the embedded clause where the idiom is intact (the shit hit the fan) and merely moves to matrix subject position for structural reasons — the idiom is interpreted at its base position and the meaning is preserved. In control, the matrix predicate assigns a thematic role to its subject: 'the shit tried to hit the fan' requires 'the shit' to be an agent of trying — which an idiom chunk cannot be, since it has no independent referent. The grammaticality of idiom chunks as raised subjects but not controlled subjects is diagnostic evidence that raising subjects receive no thematic role from the matrix predicate.
The idiom chunk test is powerful because it probes thematic role assignment without requiring speakers to have theoretical knowledge of syntax. Idiom chunks are semantically inert as isolated NPs — they only have meaning as part of the idiom. If a predicate assigns a thematic role to its subject, the idiom chunk cannot satisfy that role, and the construction fails. If the predicate merely hosts a structural subject (raising), the idiom chunk passes through and the idiom is interpreted where it was formed. This asymmetry provides direct evidence for the raising/control distinction without appeal to abstract theoretical machinery.