Deep Structure Semantic Roles

In recent posts I’ve been examining case systems in a few languages; in this post I’d like to introduce the idea of deep structure semantic roles, with a digression on the passive voice in English. Next time I will continue with more discussion of semantic roles.

Here’s an ordinary English sentence: “The shortstop caught the ball”. Here’s another ordinary English sentence: “The ball was caught by the shortstop”. These two sentences are not the same, but they are obviously related. They have the same “meaning” (whatever “meaning” means), but the topic of the first sentence is the shortstop, and the topic of the second is the ball. The first sentence is probably “about” the shortstop—the shortstop is the “topic”—and the rest of the sentence, the “comment”, tells us something about the shortstop—namely, that he caught the ball. The second sentence, the passive sentence, is probably “about” the ball, and the rest of the sentence tells us something about the ball, namely that it was caught by the shortstop. One of the principal functions of the passive voice in English is to move the direct object out of the comment part of the sentence into the topic part of the sentence. The active sentence has been “transformed” into a passive sentence.

We can say more about how this transformation takes place. We can say that the active sentence has two major constituents: a noun phrase, which in this case is a noun plus an article—“The shortstop”—and a verb phrase—“caught the ball”; and we can say that the verb phrase in turn has two parts, a verb—“caught”—and a noun phrase—“the ball”. Schematically the structure looks something like this: [[NP1] + [VP + NP2]]—the subscripts tell us which noun phrase is which.

In order to change this active sentence into a passive sentence, we have to move the NP1 out of its position and put it in a prepositional phrase—“by the shortstop”—at the end of the sentence. That move leaves the first NP slot open, so we move NP2 into that empty slot. Then we add an auxiliary verb—a form of the verb of the verb “to be”. We add the tense marker PAST, which was on the verb “caught” in the active sentence—that gives us the auxiliary form “was”—and now we use the past participle “caught”. Hocus pocus, mumbo jumbo, we’ve got the passive sentence “The ball was caught by the shortstop”.

In traditional grammar we would say that “the shortstop” is the subject of the active sentence (“The shortstop caught the ball”) and “the ball” is the direct object. But in the passive sentence (“The ball was caught by the shortstop”), “the shortstop” is no longer the subject—it’s the object of the preposition “by”, and the prepositional phrase tells us the agent of the action; the subject of the passive sentence is “the ball”. If we say that a word or phrase is the subject of a sentence, that doesn’t tell us everything we want to know. The subject of a sentence can be the person who catches the ball, or it can be the ball that’s caught.

The two sentences describe the same real-world situation—the same player is catching the same ball—but they are directing our attention differently. The active sentence directs our attention to the shortstop and tells us what he’s doing; the passive sentence directs our attention to the ball and tells us what happened to it. Notice, by the way, that we can leave off the agent phrase—we can just say “The ball was caught” without saying who caught it. Perhaps we don’t know who caught it. In some languages the agent of a passive construction must be deleted; in English deletion of the agent is optional.

English nouns, as we have seen in previous posts, mark only two cases, the genitive and the everything-else case. In the sentence “The shortstop caught the ball”, both nouns are in the everything-else case; likewise, the two nouns in the passive sentence “The ball was caught by the shortstop”. But English pronouns (sometimes) do make a distinction between the nominative and the accusative cases. In the active sentence “I caught him”, the subject pronoun, “I”, is in the nominative case, and the direct object pronoun is in the accusative case. In the passive sentence “He was caught by me”, when NP2 is moved over into the first slot, it becomes nominative, and when NP1 is moved to become the object of the preposition, it becomes accusative.

A sentence is made of more than grammar; there is also meaning—that is, semantics—and it will help our understanding of what’s going on if we look at the semantic functions of the words in these sentences. When a shortstop catches a ball, he does something, and more specifically, he does something to the ball. That’s true in both the active sentence, “The shortstop caught the ball” and the passive sentence, “The ball was caught by the shortstop”. When we describe the active sentence, we say that “the shortstop” tells us the agent of the action. But if the shortstop is the agent in the active sentence, he or she is likewise the agent in the passive sentence. The agent can be expressed as the subject of an active sentence, or it can be expressed as the object of a prepositional phrase in a passive sentence. If the shortstop is the agent, it makes sense to label the ball with the corelative term and call it the patient. The patient can be expressed as either the direct object of the verb in an active sentence or as the subject of the verb in a passive sentence.

Thus we see two different distinctions: the distinction between the subject and the object (and the object of a preposition) and the distinction between the agent and the patient. The subject and the object and the object of a preposition are all grammatical concepts, but the agent and the patient are semantic rather than grammatical. But even though semantics and grammar are different, semantic functions can be expressed in grammar. We can say that the agent can be expressed in grammar as the subject of an active sentence OR as the object of the preposition “by” in a passive sentence, while the patient can be expressed in grammar as the direct object of an active sentence OR as the subject of a passive sentence.

Now consider the active sentence “John weighed the potatoes”. This has the basic structure [[NP1] + [VP + NP2]], so it has a corresponding passive sentence, “The potatoes were weighed by John”. (This sentence on its own may sound a little artificial, but it might work better as part of a coordinated pair of sentences, “The rice was weighed by Andrew, but the potatoes were weighed by John”.) What about the sentence “John weighed two hundred pounds”? Does this have a corresponding passive: “*Two hundred pounds were weighed by John”? That sentence seems not just odd, but quite unacceptable. (It’s conventional to mark unacceptable sentences with an initial asterisk; a questionable sentence is marked by an initial question mark.) The sentence “John weighed two hundred pounds” seems to have the right structure to make a passive—[[NP1] + [VP + NP2]]—but somehow the passive transformation doesn’t work. When John weighs the potatoes, he picks up the potatoes and puts them on the scale. John is the agent of the action and the potatoes are the patient of the action. We can make either an active or a passive sentence to describe the situation.  But when John weighs two hundred pounds, those two hundred pounds aren’t the patient of the action. The phrase “two hundred pounds” isn’t really the direct object of the active sentence, and therefore it can’t become the subject of a passive sentence.

These sentences use two different verbs “to weigh”, or at least two different ideas about weighing. Weighing potatoes is what John does but weighing two hundred pounds is what John is. Here’s another test: Let’s imagine that someone wants to buy two hundred potatoes (say for a banquet where each person gets a baked potato) and that the potatoes are sold by weight. John counts out two hundred potatoes and then he weighs them, and his co-worker Bill checks to make sure he got the right price. I can say “John weighed two hundred potatoes, and Bill weighed them, too.” If I’m talking about how much John weighed last year before he went on a diet, and I’m comparing him to Bill, I could say “John weighed two hundred pounds, and so did Bill”, but it would not be acceptable to say “*John weighed two hundred pounds and Bill weighed them, too”.

These examples show that the phrase “two hundred pounds” in the sentence “John weighed two hundred pounds” is not the patient of the action, and that’s why it can’t become the subject of a passive sentence. If it’s not the patient, what is it? We can say that the semantic role of this phrase is to tell us a measure. So we could say that in addition to the semantic roles of agent and patient, we need to add the semantic role of measure. In English there is no case to mark this semantic role, but we can detect it by using sample sentences as a test. We can find other examples of this role: The active sentence “This book costs five dollars” can’t be turned into the passive sentence “*Five dollars is costed by this book”.

Now for a slight digression on passive sentences. The complications of passive constructions are sometimes hard to explain. Some passive constructions are clearly acceptable (“The ball was caught by the shortstop”), some are clearly unacceptable (“*Five dollars is costed by this book), and others are questionable—perhaps acceptable but odd.

For instance, the active sentence “Tony likes cookies” makes the questionable passive “?Cookies are liked by Tony”. But with a slight change the sentence becomes acceptable—“Cookies are liked by people all over the world”. If I have a particular cookie in mind, then I can say “The cookie I was saving for Bob was eaten by Tony”, but if I mean cookies in general, the passive sounds odd, at least to me—“?Cookies were eaten by Tony”.

Compare “Sam ran the Boston Marathon”, “?The Boston marathon has been run by Sam” and “The Boston Marathon has been run by thousands of people”. The passive seems to be better the more people are involved.

Some idioms don’t make good passives: “Tom kicked the bucket” (meaning that he died) sounds odd as a passive, “?The bucket was kicked by Tom”; but if there was an actual bucket that Tom kicked, then “The bucket was kicked by Tom” is acceptable, perhaps because there is an actual object which can act as the patient of the verb. Some constructions with reflexive pronouns won’t make an acceptable passive: “Tom weighed himself” won’t give “Himself was weighed by Tom”; “He recognized himself in the photograph” won’t give “Himself was recognized in the photograph by Tom”.

Why are some passives better than others? More than one factor may be involved: First, some semantic roles, such as measure, may not be available for the passive transformation. Second, since one of the primary effects of the passive transformation is to move N2 into the topic position of the sentence, the transformation may work best when it is appropriate for N2 to be the topic. Third, the more effect the situation has on the patient, the better the passive.

In this post I’ve introduced a distinction between grammar and semantics, I’ve looked at a few curious passive constructions, and I’ve suggested that in order to understand these constructions we need some semantic concepts—agent, patient, and measure—in addition to the usual grammatical concepts of subject and object. In my next post I will look at more sentences which will require some additional semantic concepts.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s