LINGUIST List 4.366

Wed 12 May 1993

Sum: Theme/Patient

Editor for this issue: <>


  1. , Summary: Theme/Patient

Message 1: Summary: Theme/Patient

Date: 11 May 1993 12:07:01 -0600Summary: Theme/Patient
From: <>
Subject: Summary: Theme/Patient

A while ago I posted the following query to the LINGUIST and to

> Some scholars dealing with thematic roles, (aka theta roles, semantic
> cases, deep cases) use the term "theme", others say "patient". A
> few,
> such as Van Valin, claim that the two are different categories, and
> that it matters in some languages.
> I have never seen any language in which alleged "themes" were treated
> any differently than alleged "patients". Does anyone know of a
> language where it is necessary to distinguish them? Or where anyone
> presents any evidence on the matter?

I received several replies from people informing me that _theme_ is
widely used as the approximate equivalent of _topic_, then contrasting
with _rheme_. I knew this, of course, but it had nothing to do with my

Several respondents tried to explain the difference in terms of
semantics. This was interesting, but of limited use. Since I regard
thematic roles as syntactic categories (the syntactic implementation of
semantics), I insist on syntactic evidence for their existence, and
regard any identification of thematic roles based on semantics alone as

Mark Jones <> wrote:

>I'm no expert, but this is what I have found:
>Both Theme and Patient are acted upon & are most often found
>as Direct Objects.
>The only difference I could find is that Patient is internally
>changed by the action, Theme is not
>Therefore, If I move the ball-- the ball is the Theme
> If I set the ball on fire-- the ball is the Patient

<> wrote:

>I think part of the problem stems from the confusion betyween
>semantic roles and theta-roles. The latter are syntactic nodes
>that correspond to a set of semantic roles, which seem to be a
>larger set. Grimshaw collapses the semantic roles theme and
>patient into the theta-role theme in English. Whether this is
>universal is debatable. There could be languages that do
>differentiate them. Semantical;ly, a patient answers the
>question 'what did you do to X.' X is a patient.
> The ball rolled down the hill.
>Ball is a theme; it is in motion. It does not answer to teh
>question 'What did you do to the ball?' There is no syntactic
>difference between these to semantic roles:
> The ball unraveled.
>, hence they are realized as a single theta role. By
>convention, I guess, it is called a theme.

If there is indeed no syntactic difference, as de Armond claims, then
we don't need two thematic roles. As for semantics: categories such as
agent, patient, and (if distinct semantically) theme are grossly
inadequate for expressing semantics. Let R be the semantic relation
between a theme argument and the predicate _fall_. No one would claim
that the relation between a theme argument and the predicate _exist_ is
also R. There is a commonality, but there is not identity. And the
commonality is achieved by abstraction, i.e. by loss of content. By
positing two categores instead of one, we have added precisely one
binary feature to the syntactic representation of the relationship.
That's nice, I suppose, but since the information is also available
from the semantics of the predicate, it's of little value: even with
theme distinct from patient, semantics gets short shrift.

Several responders referred to Jackendoff's analysis of thematic roles.
Andrew Rosta <> answered my query about languages
where theme and patient must be distinguished as follows:

>English, maybe. In Jackendoff's framework, the object of
>_kick_ wd be patient but not necessarily theme, and the
>subject of _die_ wd be theme but not patient, & the object of
>_kill_ wd be both.

Steven Schaufele ( wrote:

>If i remember correctly, Jackendoff in his Semantic Structures
>distinguishes very carefully between 'theme' and 'patient' as
>'Theme', at least with a verb of motion, refers to that which
>is actually moving.
>'Patient' refers specifically to one of the two essential
>arguments of a semantically transitive verb - the one other
>than the 'agent' (Jackendoff prefers 'actor').
>Thus, in 'I hit the window with a rock', 'window' would be the
>'patient', but 'rock' would be the 'theme'. In 'The car hit
>the tree', 'tree' would be the 'patient', but 'car' would be
>the 'theme'.

A few respondents gave other references. Don Ringe
<> wrote:

>Prof. Connolly-- You might take a look at Stephen Pinker's
>book of 1989,
>*Learnability and Cognition*, esp. ch. 3. He's basically
>trying to figure out how 1st-lg. learners learn which argument
>structures are admissible with which verbs, but to do that he
>has to develop a fairly explicit theory of the concepts
>underlying arg. str. (relying in part on the work of others,
>copiously cited), and he finds that a distinction between
>"theme" and "patient" is well motivated by the types of
>alternations between structures that one encounters. Whether
>any language marks the difference *on the surface* seems
>doubtful, but
>(a) if you need it anywhere in the grammar, then you need it,
>period, and (b) if we're not still behaviorists, there's no
>good reason to be suspicious of and/or uneasy about such
>covert categories. Hope this is some help.

I have not yet consulted Pinker's book, but this is a good lead. BTW,
as a case grammarian, I'm not afraid of covert categories; I just want
to see the evidence.

Randy LaPolla (HSLAPOLLA%TWNAS886.BITNETpucc.Princeton.EDU) wrote:

>If you were not convinced by Van Valin's account of the need
>for two roles, I can give you one clear example: in Chinese
>only verbs of motion, i.e. theme-taking predicates, allow that
>argument (the theme) to appear postverbally (e.g. lai che le
>'lit.: come car ASP'). Patients cannot do this unless part of
>a larger serial verb structure (e.g. da po le beizi 'hit break
>ASP cup' "broke a/the cup"). I don't work on English, but it
>seems to me the "spray/load" verbs in English
>(that is, the alternation possible with these verbs) is only
>possible with theme-taking predicates.

This is fascinating, but it's hard to say quite what it means, since it
sounds as if *stative* verbs do not permit the construction. If that
is true, then it must still be shown that the presence of a theme
(rather than patient or whatever) is a cause or even a
more-than-coincidental sine qua non.

I should also add that the reference to the 'spray/load' verbs in
English is not particularly helpful. While the thematic roles are not
entirely clear, one can perfectly well analyze the material as a
theme/patient and the destination as a locative, so that the syntactic
evidence for a distinction between theme and patient is still

Thanks to all who responded. It has been most helpful to me.
--Leo Connolly
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue