More on Semantic Roles

In a previous post (Deep Structure Semantic Roles, posted 17 Jan) I showed that the sentence “George weighed the potatoes” has a passive partner, “The potatoes were weighed by George”, but the sentence “George weighed two hundred pounds” doesn’t have a passive partner, “*Two hundred pounds were weighed by George”. I proposed that the phrase “the potatoes” has the semantic role of “patient”; a patient can be, and often is, the object of an active transitive verb, but a patient is also allowed to go into the subject slot in a passive construction—so if the patient “the potatoes” is the object of the active sentence “George weighed the potatoes”, it is also allowed to be the subject of the passive construction, “The potatoes were weighed by George”. I then proposed that “two hundred pounds” in the sentence “George weighed two hundred pounds” does not have the semantic role of patient; instead, it has the role of what we can call “measure”. In this post I will explore some more sentences to see if we can find more semantic roles.

Consider, for instance, the sentences “Tom is cold” and “Chicago is cold”. These seem to have the same form; in traditional grammar we would say that they are copulative sentences, that is, they are formed with a subject, a copulative verb, and a predicate adjective. But notice that we can say “It’s cold in Chicago” but we can’t say “It’s cold in Tom”. What’s the difference? These sentences suggest that the word “Tom” and the word “Chicago” have different semantic roles. “Chicago” tells us where it’s cold, and “Tom” tells us who is cold. In these sentences, “Chicago” has the semantic role “locative”, and “Tom” has the semantic role “experiencer”.

Now consider the sentences “Sue opened the door” and “This key opened the door”. Again the sentences seem to have the same form—Subject+Verb+Direct Object—but there’s a difference which shows up in related sentences: we can say “Sue opened the door with this key”, but we can’t say “*This key opened the door with Sue”. Sue is the agent who opens the door, while the key is the instrument which is used to open the door. In English an instrument is often expressed in a prepositional phrase, such as “with this key”, but in addition an instrument can be the subject of a sentence. An agent can be the subject of a sentence, of course, but it can’t go into an instrumental prepositional phrase. In some situations, however, people can be thought of as instruments—“The general attacked the fortifications with the infantry”.

Many languages (including Proto-IndoEuropean) can express the locative and the instrumental with grammatical case forms that appear in the surface of the sentence. English doesn’t have these cases, but the semantic differences are there—in the deep structure.

Another common grammatical case is the dative, which expresses the recipient of a gift or more generally the beneficiary of an action—“Mary gave Tom a book”, “Mary made Tom a cake for his birthday”. This case occurs in Greek, Latin, German, and so on, but not in English. But English does mark the semantic function in word order with an appropriate verb: the recipient, in these cases Tom, comes right after the verb, and the direct object comes at the end of the sentence—or else the recipient is in a separate prepositional phrase. You can’t say “*Mary gave a book Tom” but you can say “Mary gave Tom a book” or “Mary gave a book to Tom”.

In English the recipient can become the subject of a passive sentence—“Tom was given a book by Mary”—but in some languages, including Latin and German, this construction is not allowed.

So far we have identified seven semantic functions: agent, patient, experiencer, measure, locative, instrumental, and recipient. Are there more? How many semantic functions could there be?

The meanings—and the kinds of meaning—expressed by language are almost without limit. It would be impossible to mark every kind of meaning with its own semantic function. Inevitably, some functions are grouped together, especially if they seem in some way similar. Let’s consider the semantic function “possession” and what it might include. Many languages express possession with the genitive case; traditional grammars of English often use the term possessive case. (English can also mark possession with the preposition “of”, but that detail needn’t concern us here.) So let’s look at some of the uses of the English genitive or possessive. Here’s a list:

Jane’s car; Jane’s opinion; Jane’s arm; Jane’s brother; Jane’s dog; Jane’s honesty; Shakespeare’s sonnets; the car’s headlight; Chicago’s climate; Canada’s parliament; the president’s executive order; the country’s elderly; Canada’s capital; the jar’s contents; the book’s price; the book’s pagination; the book’s plot.

Some of these are worth comment. Jane’s car could be hers because she paid for it and owns it; but we could also use this phrase for a car Jane has rented or borrowed; and when we talk about “the president’s limousine” I don’t think we have any sense that the president owns or rents or has borrowed the car. If Jane owns a car, then presumably Frank doesn’t own it (unless they bought it together). But if Jane has an opinion, anyone else can also have that opinion. I don’t think Jane owns her arm, but in these days of organ transplants I suppose the ownership of part of a body might be a question. Jane certainly doesn’t own her brother; she owns her dog, but with legal restrictions on what that ownership means. Shakespeare doesn’t own his sonnets—a grammar book might call this semantic role the genitive of source; these days authors have legal ownership of their works, at least for a while. The car doesn’t own its headlight, nor is it the source of the headlight, nor does it have any legal rights to it. I suppose this genitive just expresses a kind of association of two things—but I don’t think we would talk about the headlight’s car. I leave the other examples for your consideration.

The point here is that the English genitive covers a lot of different semantic relationships. It would be impossible to have a different case form for each of these. Different languages make different decisions about which semantic roles to group under the same case; linguists make decisions about which semantic roles have some degree of generality, which roles should be grouped together. All languages, say, have the semantic role “agent” and “patient”; how many more roles should be distinguished? Linguists disagree. Some make long lists, some make short lists.

I don’t think there’s one list which will please every linguist or suit every purpose. It’s a matter of judgment. In this post and the next one I will use the list proposed by Robert Longacre, in his wonderful book “An Anatomy of Speech Notions”. (He uses the term “case” instead of “semantic role”—if you care about terminology see my terminological note at the end of this post.) Longacre lists ten nuclear cases and six peripheral cases. I will be mostly interested in his nuclear cases. Here they are: Experiencer (E); Patient (P); Agent (A); Range (R); Measure (M); Instrument (I); Locative (L); Source (S); Goal (G); and Path. (Path)

(His peripheral cases are Place; Time; Manner; Accompaniment; Cause; and Purpose; I won’t be concerned with these for now.)

Some of these we have already noted. I will give an example of each along with an occasional comment.
Experiencer: I’m hot, I’m nervous.
Patient: “The shortstop caught the ball.” “The window was broken accidentally.”
Agent: “John broke the window.” In the sentence “Tom listened to the owl”, Tom is both Agent and Experiencer.
Range: “The role assigned to any surface structure nominal that completes or further specifies the predicate; the product of the activity of a predicate.” “Caruso sang a song.”
Measure: “It weighs six pounds.” “This piece of equipment costs $500.00”
Instrument: “John cut the rope with a knife.”
Locative: “The ship sank at sea.” “Harriet is travelling in Europe.”
Source: “The locale which a predication assumes as place of origin; the entity from which physical sensation emanates; the animate entity who is the original owner in a transfer.” “Mary obtained her visa from the Australian Embassy.”
Goal: “The locale which is point of termination for a predication; the entity towards which a predication is directed without any necessary change of state in that entity; the animate entity who is the non-transitory or terminal owner.” “The boat drifted to the right bank.” “The department obtained a visa for Dr. Smith.”
Path: “The locale or locales transversed by a predication; the transitory owner.” “Tom threw the knife across the room.” “Mr. Smith sold Tom a convertible for his wife.” Tom is the path by which the car gets to his wife.

Semantic roles are not isolated features of a noun. They are determined by the whole sentence, and most especially by the relationships of the nouns to the verb. That’s what I will begin to explore in my next post.

A Footnote on Terminology: I have been using the term “semantic roles” for these features of nouns, but most linguists call them “cases”. This general approach to syntax is often called “Case Grammar”; one of the most influential articles in this field was Charles Filmore’s seminal paper, “The Case for Case”, and next week I will be examining what Robert Longacre calls “Case Frames”. So I’m definitely going against the main steam. Perhaps a compromise is possible: cases which are marked in morphology (such as six cases in Latin) could be called “surface cases”, and those which show up in other ways (such as the locative or instrumental in English) could be called “deep cases”. That’s what I will do in future posts.