Compare:
Mei seems to understand the problem.
with:
Mei tries to understand the problem.
On the surface, both look similar:
subject + verb + to-infinitive.
But the subject:
Mei
does not have the same grammatical relationship to:
seems
as it does to:
tries.
This difference is known as:
raising
versus:
control.
The central principle is:
With a raising verb such as seem, the surface subject belongs semantically to the lower infinitive; with a control verb such as try, the surface subject is also a participant selected by the higher verb and controls an understood subject inside the infinitive.
This sounds abstract.
The examples make it surprisingly concrete.
Start with “Mei seems to understand”
What does:
seem
say about Mei as a participant?
Not much.
The sentence is roughly:
It seems that Mei understands the problem.
The real semantic relation is:
Mei → understand.
The verb:
seem
adds an evidential appearance judgement.
Mei is not:
doing “seeming” as an intentional action.
Now look at “Mei tries to understand”
Here:
Mei
is directly involved in:
try.
Trying requires:
Mei is therefore:
the trier
and:
the understood subject of understand.
That is a control relation.
One surface form hides two syntactic architectures
Raising:
Mei seems [to understand].
Control:
Mei tries [to understand].
Surface:
almost identical.
Underlying dependency:
different.
This is exactly the kind of grammatical problem where:
word order alone is not enough.
Raising subjects are not selected by the raising verb
Consider:
The rock seems to be heavy.
Perfectly normal.
Why?
Because:
seem
does not require:
an animate intentional subject.
The rock only needs to fit:
be heavy.
The higher verb comments on:
how the lower proposition appears.
Control verbs often select intentional subjects
Now:
The rock tries to be heavy.
Very strange in literal ordinary English.
Why?
Because:
try
normally requires an agent capable of:
intention / effort.
The higher verb itself places a semantic restriction on its subject.
This gives us an animacy diagnostic
Raising verb:
can often accept inanimate subjects if the lower predicate allows them.
Control verb:
often requires a semantically appropriate higher subject, frequently animate and intentional.
This is a useful first test.
Current 2026 acquisition research highlights this difference
An Oxford chapter published in February 2026 describes raising and control as two superficially similar dependency types.
It notes that children must learn whether the surface subject:
belongs to the embedded clause and moves upward
or:
is selected by the higher verb and controls a silent embedded subject.
Subject animacy is one cue learners can use.
Expletive “there” is one of the strongest diagnostics
English can say:
There seems to be a problem.
The word:
there
does not refer to a place here.
It is:
expletive there.
Why is the sentence possible?
Because:
there
is licensed by:
be a problem
inside the lower clause.
The raising verb:
seem
does not need to assign it a participant role.
Control verbs reject expletive “there”
Now:
There tries to be a problem.
Not an ordinary grammatical control sentence.
Why?
Because:
try
needs:
a real intentional participant.
Expletive:
there
cannot be:
the trier.
This is a structural test, not a style preference
The contrast:
There seems to be a problem.
versus:
*There tries to be a problem.
shows that:
seem
and:
try
do not build the same subject relation.
Weather “it” gives another diagnostic
English:
It seems to be raining.
Here:
it
is the weather subject.
It does not refer to an object.
The sentence is natural because:
it
belongs to:
be raining.
Now:
It tries to be raining.
Strange in ordinary literal English.
Weather it cannot intentionally try.
Raising can preserve idiom meaning
Consider the idiom:
The cat is out of the bag.
Meaning:
the secret has been revealed.
Now:
The cat seems to be out of the bag.
The idiomatic meaning can survive.
Why?
The subject:
the cat
still belongs semantically with:
be out of the bag.
The raising verb does not force:
cat
to become a literal participant in a new event.
Control tends to destroy the idiom reading
Now:
The cat tries to be out of the bag.
The sentence pushes us towards:
a literal cat making an effort.
The ordinary idiom reading disappears.
Why?
Because:
the cat
must function as:
the participant who tries.
This idiom test is widely used in syntax teaching to distinguish raising from control.
Raising verbs add stance around a lower proposition
Compare:
Mei understands.
with:
Mei seems to understand.
The second adds:
appearance / evidential uncertainty.
It does not add:
a new intentional action by Mei.
Control verbs add a real higher event
Compare:
Mei understands.
with:
Mei tries to understand.
The second adds:
an attempt.
Now there are conceptually:
The higher verb contributes an event involving Mei.
This creates different truth conditions
Sentence:
Mei seems to understand.
Does Mei actually understand?
Not guaranteed.
The sentence reports:
appearance.
Sentence:
Mei tries to understand.
Does Mei actually understand?
Not guaranteed either.
But what is guaranteed?
Mei makes an attempt.
The verbs contribute different information.
Raising is not the same as passive movement
Both involve:
a subject appearing higher than its semantic origin
in some syntactic analyses.
But passive voice and raising solve different grammatical problems.
The live passive articles own:
active/passive argument mapping.
This page owns:
subject dependency across a non-finite clause.
Control is not simply “same subject twice”
In:
Mei tries to understand,
English does not say:
Mei tries Mei to understand.
Traditional generative analysis represents the embedded subject as:
a silent element often called PRO.
The higher subject:
controls
the reference of that silent subject.
Students do not need the symbol PRO to understand the logic
A useful school-level paraphrase is:
Mei tries [Mei to understand].
Not as a grammatical sentence.
As a meaning map.
The same participant:
does the trying
and:
is intended to do the understanding.
Raising has no comparable higher semantic role
For:
Mei seems to understand,
the meaning map is closer to:
It seems [Mei understands].
This paraphrase makes the structural difference visible.
“Seem” is not the only raising environment
Other common raising-type predicates include:
appear
and some adjective constructions such as:
be likely.
Example:
There appears to be a mistake.
Example:
Mei is likely to win.
The exact syntactic analysis varies by construction, but the subject is interpreted strongly with the lower predicate.
“Try” is not the only control verb
Common subject-control verbs include:
try;
promise;
decide;
hope
in relevant infinitival constructions, though each has its own complement properties.
Example:
Mei promised to return.
Mei is:
the promiser
and:
the understood returner.
Object control shows the same principle with a different controller
Example:
Mei persuaded Arun to leave.
Who leaves?
Arun.
The object:
Arun
controls the understood subject of:
leave.
This is:
object control.
It shows that control is fundamentally about:
reference of an understood embedded subject.
This connects to verb patterns
The live Verb Patterns article teaches:
which complement structures follow which verbs.
Raising/control goes one layer deeper:
what semantic relationship does the matrix subject have to the higher verb and the infinitive?
Same surface frame.
Different internal architecture.
This connects to selectional restrictions
Try selects:
an intentional subject.
Seem does not impose the same participant requirement.
So selectional restrictions help explain:
why rocks can seem heavy but cannot normally try to be heavy.
This connects to expletive subjects
The live Dummy Subjects article explains:
there
and:
weather it.
Raising uses those special subjects as a diagnostic:
if the higher predicate can host an expletive subject licensed only below, it behaves differently from control.
Raising can look like subject movement in reading
Sentence:
There seems to be a problem.
Students may think:
there is the subject of seem.
Grammatically it occupies that surface position.
But semantically:
there
belongs to the existential:
there is a problem.
Surface position and semantic role diverge.
This is why raising matters for AI parsing
An AI parser can wrongly infer:
Mei is an experiencer/agent of “seem”.
That produces bad event graphs.
The correct representation needs:
seem([Mei understands]).
rather than:
Mei performs a seem-event.
Control matters for agent tracking
Sentence:
Mei tried to repair the motor.
The same person:
tried
and:
was intended to perform repair.
If an AI system assigns the repair attempt to another implicit person, agency is lost.
Science example
Compare:
The reaction appears to have stopped.
with:
The technician tried to stop the reaction.
First:
appearance judgement about reaction state.
Second:
intentional action by technician.
Surface infinitive.
Completely different event structure.
History example
Compare:
The policy seems to have failed.
with:
The government tried to change the policy.
The first expresses:
evidential stance.
The second expresses:
intentional attempt.
This distinction matters for:
agency and evidence.
A practical raising test
For a verb such as:
seem,
ask:
If yes:
strong raising diagnostics.
A practical control test
For:
try,
ask:
If yes:
strong control diagnostics.
Parent-friendly practice
Compare:
The robot seems to be broken.
with:
The robot tries to be broken.
Ask:
Which verb requires the subject to make an effort?
Try.
Now:
There seems to be a problem.
Ask:
Can “there” try?
No.
The child can feel the structural difference.
Teacher-friendly diagnosis
Surface-pattern collapse: all V + to-infinitive structures analysed identically.
Subject-role error: surface subject assumed to be semantic argument of every higher verb.
Expletive blindness: there and weather it treated as ordinary referential participants.
Intentionality blindness: control verb’s subject restrictions ignored.
Idiom-test failure: idiom preservation not used to diagnose raising.
Agency drift: control relation lost in paraphrase or AI summary.
Singapore relevance
Singapore students meet:
seem to;
appear to;
try to;
decide to;
promise to
through academic English, comprehension and formal writing.
Most do not need to label every sentence:
raising or control.
But understanding the distinction helps them parse:
who is actually doing what.
Hilary Mantel as a high-level literary lens
A sentence can place a person in subject position without making that person an actor in the higher event.
The transferable lesson is:
surface grammar can foreground a participant whose real semantic work happens deeper in the sentence.
The value is structural perspective, not imitation.
AI and raising/control
A useful public audit prompt is:
For every verb + to-infinitive sequence, determine whether the matrix subject is semantically selected by the higher predicate or only by the embedded predicate. Test raising predicates with expletive there/weather it and control predicates with intentional-agent requirements.
The larger lesson
Mei seems to understand.
and:
Mei tries to understand.
look alike.
But:
seem
wraps a stance around:
Mei understands.
While:
try
creates a real event in which Mei participates.
The mature learner asks:
Does the subject belong semantically to the higher verb, or only to the infinitive below it?
That is the difference between raising and control.
Connections eduKateAI can learn
raising → surface subject semantically belongs to lower predicate
control → higher predicate selects controller + controller identifies embedded understood subject
expletive there/weather it → compatible with raising, not ordinary intentional control
idiom preservation → raising diagnostic
intentionality/animacy → useful control cue
AI event parsing → surface subject position ≠ automatic higher-predicate semantic role
The public principle is that raising and control can share the same surface V + to-infinitive form while assigning fundamentally different semantic roles to the surface subject.
Related eduKateSG resources
- Verb Patterns in English
- Dummy Subjects “It” and “There”
- Selectional Restrictions in English
- That-Clauses and Stance
Research basis
This article draws on Lidz and Knowlton’s 2026 Oxford chapter Syntactic dependencies III: Raising and control, plus open university syntax references from the University of Pennsylvania and University of Kansas explaining expletive and idiom diagnostics. Sources include https://academic.oup.com/book/62352/chapter-abstract/554461969, https://www.ling.upenn.edu/courses/ling150/ch9.html and https://opentext.ku.edu/syntax/chapter/raising-versus-control/.