Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

PLIN0009 Semantic Theory Spring 2020

Lecture Notes 5

1 Notes on Syntax

We have been assume syntactic structure like (1) as the object that the semantic compo-
nent of grammar operates on. Besides the omission of Tense and Aspect, mentioned in
Lecture 3, you might wonder if the structure should have some non-branching nodes and
look like (2).

(1) S (2) S

John VP DP VP

loves Mary John V DP

loves Mary

We can incorporate such non-branching nodes without changing our analysis substan-
tially, by assuming that a non-branching projection does not change the meaning. This is
a natural assumption given that in a non-branching constituent, no new meaning is added.
So we assume the following equality, no matter what A and B are.
4 <M
A
(3) For any model M, 5 = = vBwM
B

Then, the above two trees will clearly mean the same thing. To be more precise, we can
state (3) as a compositional rule, which we call the Non-Branching Node Rule.

(4) Non-Branching Node Rule: For any model M, If A is a non-branching node with
M M
B as its sole daughter, then vAw = vBw .

To simplify the exposition, we will always omit non-branching projections, but it should be
kept in mind that this is harmless, because if syntacticians told us to put the non-branching
nodes back, we could always do so without changing our semantic analysis substantially.
(That said, syntacticians these days are also skeptical about non-branching projections, so
they probably won’t tell us to have them back anyway!)

You might also wonder if our labels are correct. Maybe you want to use IP or TP, instead
of S, for example. But it should be noticed that syntactic labels like these play no semantic
role in our semantic theory. That is, the interpretation only depends on the hierarchical
structure and the semantic types of the constituents, and the syntactic labels are simply
completely ignored. So we could actually omit syntactic labels altogether without chang-
ing anything in our semantic analyses. We will however sometimes put them for the sake
of readability.

1
2 Nouns

What do nouns denote? The irst thing to notice is that just like verbs, nouns can be used to
describe properties that entities have. For example, compare the following two sentences.

(5) a. John smokes.


b. John is a smoker.

These two sentences mean almost identical things. We have an analysis of the irst sen-
tence (5a), according to which ‘smokes’ denotes a function of type xe, ty. More speci ically,
we have assigned the following denotation to it.

(6) For any model M, vsmokeswM = λx P De . 1 iff x smokes in M

By analogy, let us assume that the noun smoker in (5b) denotes the following function of
type xe, ty.

(7) For any model M, vsmokerwM = λx P De . 1 iff x is a smoker in M

This would certainly work, if the sentence (5b) were completely parallel to (5a) and looked
like (8).

(8) S

John smoker

But there are words in (5b) that do not occur in (5a), namely, is and a. So the structure
that we need to account for is actually something like (9).

(9) S

John VP

is DP

a smoker

This forces us to say something about the semantics of these little extra words.

But before getting it should perhaps be noted that in a number of languages (e.g. Russian),
the sentence corresponding to (5b) in fact looks like John smoker, at least on the surface.
In such a language, we might not have to say anything more than (7). Words like is and
a, which do not seem to be there for grammatical purposes and do not seem to have sub-
stantial lexical content, are often called functional items (as opposed to lexical items).

2.1 Semantic Vacuity

Here, let us pursue the idea that is and a are semantically vacuous, which is another way
of saying that they mean nothing. What does it mean to mean nothing? Here, we equate

2
meaning nothing to denoting an identity function.

An identity function is a function that returns the exact same thing as the input. Identity
functions come in many different semantic types, but since the input and the output are
always the same, they all have a semantic type that looks like xτ , τ y for some type τ . The
one we need in the analysis of (5b) is of type xxe, ty , xe, tyy, which you can igure out by
reasoning about semantic types as follows. First, ill in the semantic types we already
know.

(10) S t

John e VP ???

is ??? DP ???

a ??? smoker xe, ty

The sister constituent of the subject DP should be of type xe, ty.

(11) S t

John e VP xe, ty

is ??? DP ???

a ??? smoker xe, ty

Now, since we are assuming that ‘a’ denotes an identify function, it should take the deno-
tation of its sister constituent and return the same thing as is. This means that it will take
a type xe, ty function, and return it as is. So we can ill in two more unknown types:

(12) S t

John e VP xe, ty

is ??? DP xe, ty

a xxe, ty , xe, tyy smoker xe, ty

Now, there’s only one choice left for the semantic type of ‘is’: it should do the same thing
as ‘a’!

3
(13) S t

John e VP xe, ty

is xxe, ty , xe, tyy DP xe, ty

a xxe, ty , xe, tyy smoker xe, ty

Now we have given types to ‘is’ and ‘a’, it’s easy to ind out their denotations. We are assum-
ing that their denotations are identity functions and they should be of type xxe, ty , xe, tyy,
so:

(14) For any model M,


a. viswM = [λf P Dxe,ty . f ]
b. vawM = [λf P Dxe,ty . f ]

These functional items denote the same identity function. Their job is to take a function
of type xe, ty and return the same function without change it. You will see why this leads to
‘mean nothing’, when you compute the denotations. First, let’s compute v[DP a smoker]wM .
4 <M
DP
5 = = vawM (vsmokerwM ) (Branching Node Rule)
a smoker
= [λf P Dxe,ty . f ](vsmokerwM ) (Lexicon)
M
= vsmokerw

This computation proves that v[DP a smoker]wM = vsmokeswM . Let us now compute the
denotation of the VP.
<M
4 <M 
4
VP
DP
M 5 = 
6 >
6 >
6 is
6 DP >
> = vis w   (Branching Node Rule)
5 = a smoker
a smoker
4 <M 
DP
 = 
= [λf P Dxe,ty . f ] 5  (Lexicon)
a smoker
4 <M
DP
=5 = (λ-conversion)
a smoker
= vsmokerwM

4
These computations demonstrate the following equality.
4 <M
VP 4 <M
6 > DP
= = vsmokerwM
6 >
6 is DP > =5
6 >
5 = a smoker
a smoker

Therefore, the words ‘a’ and ‘is’, although syntactically and phonologically real, have no
semantic effects. They just pass the meaning of their sister constituent up to the next
level, without changing it.

2.2 Transitive Nouns

Nouns like ‘smoker’ are intransitive nouns, in the sense that they only take one argument.
They are similar in meaning to intransitive verbs like ‘smokes’, which also only take one
argument. Just like there are transitive verbs like ‘studies’, there are transitive nouns too.
For example, ‘student’, as in (15b), takes two arguments, Mary and linguistics.

(15) a. Mary studies linguistics.


b. Mary is a student of linguistics.

According to the analysis we developed in the previous lecture, the transitive verb ‘studies’
can be given the following function of type xe, xe, tyy as its denotation.

(16) For any model M, vstudieswM = [λx P De . [λy P De . 1 iff y studies x in M]]

Here we assume (without argument) that ‘linguistics’ is a proper name denoting some
entity. Of course the denotation of ‘linguistics’ is not a concrete person or object, but a
more abstract kind of entity. What exactly such abstract entities might be is an important
question, but to simplify the discussion here, let’s not dwell on this issue.

In analogy to (16), we can analyze transitive nouns as denoting functions of type xe, xe, tyy,
as in (17).

(17) For any model M, vstudentwM = [λx P De . [λy P De . 1 iff y is a student of x in M]]

Again there are a couple of words in (15b) that do not occur in (15a), namely, ‘is’, ‘a’, and ‘of’.
We have already discussed ‘is’ and ‘a’ above and analyzed them as semantically vacuous. It
seems that ‘of’ is also a kind of word that has no meaning. So let us analyze it as semanti-
cally vacuous too. However, it’s semantic type should be different from the semantic type
of ‘is’ and ‘a’. To see this, consider the following tree diagram.

5
(18) S

Mary VP

is DP

a NP

student PP

of linguistics

By assumption ‘linguistics’ denotes an individual, so vlinguisticswM is of type e. Then, in


order to combine vofwM and vlinguisticswM via the Branching Node Rule, vofwM needs
to be a kind of function that can take vlinguisticswM as its argument. This means its
semantic type will look like xe, y. Recall now that we want to analyze vofwM to be se-
mantically vacuous, which is to say that it denotes an identity function. Since an identity
function simply returns the input, its semantic type will be xe, ey. Therefore, the denota-
tion of ‘of’ is as in (19).

(19) For any model M, vofwM = [λx P De . x ]

Let’s make sure that the semantic types will work out:

(20) S t

Mary e VP xe, ty

is xxe, ty , xe, tyy DP xe, ty

a xxe, ty , xe, tyy NP xe, ty

student xe, xe, tyy PP e

of xe, ey linguistics e

Using the denotation in (19), you can now prove the following equality, which is left for an
exercise here.
4 <M
DP
4 <M
6
6
>
> NP
6a NP
> 6 >
6 > 6 >
> =6 student PP
6 > >
6 6 >
student PP
6 > 5 =
6 >
of linguistics
6 >
5 =
of linguistics

6
4 <M 
PP
 = 
= vstudentwM 5 
of linguistics

= vstudentwM (vlinguisticswM )

2.3 Implicit Arguments

You might be wondering what happens in sentences like (21).

(21) Mary is a student.

This sentence does not explicitly say what Mary is a student of. If student is a transitive
noun, how should we deal with a sentence like this? There are a couple of possibilities
here.

One possibility is to assume that there are intransitive and transitive versions of the noun
student, and what occurs in (21) is the intransitive version, unlike the transitive version
used in (15b). The denotations of these two nouns can be written as follows.

(22) a. vstudentintr wM = [λy P De . 1 iff y is a student in M]


b. vstudenttr wM = [λx P De . [λy P De . 1 iff y is a student of x in M]]

But this might look a bit too uninspiring. Another possibility is that one of the meanings
is the basic one (perhaps the transitive use, because it’s more complex), and the other use
is derived by some general mechanism. Notice that some verbs show similar transitive-
intransitive alternation.

(23) a. Mary studied at UCL.


b. Mary studied linguistics at UCL.

(24) a. John ate.


b. John ate pizza.

That the transitive-intransitive alternation is observed across categories suggests that there
is indeed some general mechanism behind it. This is a very interesting topic, but it is be-
yond the scope of this module (it would be a good topic for your Long Essay!).

2.4 Summary

To summarize, we analyzed nouns on a par with verbs. Just like there are intransitive
and transitive verbs, there are intransitive and transitive nouns. Intransitive verbs and
intransitive nouns denote functions of type xe, ty.

(25) For any model M,


a. vsmokeswM = λx P De . 1 iff x smokes in M
b. vsmokerwM = λx P De . 1 iff x is a smoker in M

It is easy to come up with similar lexical entries for other intransitive nouns.

7
(26) For any model M,
a. vboywM = λx P De . 1 iff x is a boy in M
b. vlinguistwM = λx P De . 1 iff x is a linguist in M
c. vtablewM = λx P De . 1 iff x is a table in M
d. vcitywM = λx P De . 1 iff x is a city in M

Transitive verbs and transitive nouns denote functions of type xe, xe, tyy.

(27) For any model M,


a. vstudieswM = [λx P De . [λy P De .1 iff y studies x in M]]
b. vstudentwM = [λx P De . [λy P De . 1 iff y is a student of x in M]]

It is routine to generalise this analysis to other transitive nouns (can you come up with
more transitive nouns?).

Sentences containing nouns often contain ‘is’, ‘a’, and ‘of’. We analyzed them as semanti-
cally vacuous, meaning they denote identity functions. Identity functions simply return
their inputs without changing them.

(28) For any model M,


a. viswM = [λf P Dxe,ty . f ]
b. vawM = [λf P Dxe,ty . f ]
c. vofwM = [λx P De . x ]

Here is an example computation. Let us take a speci ic model, say M8 , and assume vGregwM8 =
G and vlinguisticswM8 = L.

4 <M8
S
6 >
6 >
6 Greg VP >
6 >
6 >
6 >
is DP
6 >
6 >
6 >
6 >
a
6 >
6 >
6 >
6 >
6 >
6 student PP >
6 >
5 =
of linguistics

8
4 <M8
VP
6 >
6 >
6 is DP >
6 >
( )
6 >
6 >
=6 a vGregwM8 (BNR)
6 >
>
6 >
6 >
student PP
6 >
6 >
6 >
5 =
of linguistics
4 <M8
VP
6 >
6 >
6 is DP >
6 >
6 >
6 >
=6 a (G)
6 >
>
6 >
6 >
student PP
6 >
6 >
6 >
5 =
of linguistics
4 <M8 
DP
6 
6 
>
6 a 
>
6 
>
M 8 6 
>
=visw 6  (G) (BNR)
>
6 
>
6 student PP 
>
6 
>
5 
>
=
of linguistics
4 <M 8 
DP
6 
6 
>
6 a 
>
6 
>
6 
>
=[λf P Dxe,ty . f ] 6  (G)
>
6 
>
6 student PP 
>
6 
>
5 
>
=
of linguistics
4 <M 8
DP
6 >
6 >
6a >
6 >
=6 (G) (λ-conv.)
6 >
>
student PP
6 >
6 >
6 >
5 =
of linguistics

9
4 <M8 
6 > 
6 > 
 student > 
=vawM8 6 PP >  (G) (BNR)
6 = 
5 
of linguistics
4 <M 8 
6 > 
6 > 
 student > 
=[λf P Dxe,ty . f ] 6 PP >  (G)
6 = 
5 
of linguistics
4 <M 8
6 >
6 >
6 student
=6 PP >
> (G) (λ-conv.)
5 =
of linguistics
4 <M8 
PP
 = 
=vstudentwM8 5  (G) (BNR)
of linguistics
( )
=vstudentwM8 vofwM8 (vlinguisticswM8 ) (G) (BNR)
( )
=vstudentwM8 vofwM8 (L) (G)
=vstudentwM8 ([λx P De . x ](L))(G)
=vstudentwM8 (L)(G) (λ-conv.)
=[λx P De .[λy P De . 1 iff y is a student of x in M8 ]](L)(G)
=[λy P De . 1 iff y is a student of L in M8 ](G) (λ-conv.)
=1 iff G is a student of L in M8 (λ-conv.)

10

You might also like