section banner


Situation schemas and semantic roles

In this section, we'll look at categories of states and events. We'll see that these categories are more complicated than categories of things because they are in a sense about things. Without things, there can be no notion of state or event. Part of characterizing a particular state or event category is characterizing what sorts of things it can be about, how many things there are (two or three in the case of a relation), and what role the things play in the whole situation.

As with categories of things, most languages have a category of words that designate categories of events, and, less often, categories of states. These words are called verbs. Examples are kiss, write, like, smell, and go. In this section we'll be guided in our consideration of the different categories of states and events by some of the English verbs used to designate them, but our focus will still be on the semantics, that is, on the kinds of things that all languages need to be able to designate. In the succeeding sections, we'll be looking at how events and states are actually designated in sentences in English and other languages.

One word of warning. While there is agreement on the basic ideas in this section, there is no general agreement on the details. The particular categories described here are those that I happen to think do the best job of characterizing the states and events that languages designate.

The do_to schema

What do the meanings of the following verbs have in common: kiss, push, punish, carry, bathe? (Think about the participants in the events and how they relate to the action and to each other.)

Let's begin by considering the event that we might designate with the following sentence.

  1. Lois kissed Clark.

This event has two participants associated with it, Clark and Lois, and they are associated with each other in particular way through a relation we can call kiss. This particular kissing event shares a lot with other instances of kissing. We can abstract over all of these instances of kissing by thinking of a kissing schema, a kind of template for kissing events.

Each kind of event or state carries with it a set of expected participants who play particular roles in it.

We have already seen an example of a schema in the context of the discussion of utterances. Remember that each utterance has a number of roles, form, Speaker, Hearer, location, and time. In the kissing schema, there are two main roles, one for each participant in the event. We can call these roles the kisser and the kissee of the event. I will refer to roles within the schemas for states and events as semantic roles. Semantic roles are defined by their meaning, not by their linguistic form. Later in this chapter, we will see that there is a set of sentence roles, including subject, for example, that are defined by their form. It is important to keep the two kinds of roles distinct.

As with utterances (which after all are a kind of event), I will diagram situation schemas using boxes to show the roles. The figure below shows the kiss schema.


In the schema for the KISS relation, the two roles are not filled by any particular people; we just consider them in terms of the parts they play in the relation. KISS is an asymmetric relation; that is, the parts played by the people filling the two roles are quite different. The KISSER is the participant who consciously initiates the action, who makes contact with the KISSEE. The KISSEE is the destination of the action. While the KISSEE is affected by the action, this participant does not play a conscious part in the event.

In a particular instance of kissing, a particular kissing event, the two roles are filled by particular people. So sentence 1 above describes the event that we can diagram as follows.

lois kissed clark

Now consider another event, the one described by the following sentence.

  1. Lois pushed Clark.

Again there is a relation, push, with two roles, which we can call pusher and pushee. And again the relation is asymmetric: the pusher initiates the action and makes contact with the pushee, and the pushee is affected by the action but plays no conscious part in it. The details of how the two push roles fit into their relation differ from how the two kiss roles fit into their relation, of course. A pusher exerts more force than a kisser, and a pushee normally moves more as a result of the action than a kissee does.

But clearly kiss and push share some properties. Both have two core roles, including one that consciously instigates the action and another that is passively affected by the action. So we can be even more abstract, thinking of a schema for the larger category of actions of this type. Let's call this schema do_to, and let's call the two roles in do_to the agent and the patient. The agent is the conscious initiator of the action, and the patient is the participant that is passively affected by the action. We can diagram the do_to schema as follows.

do to

Other examples of relations that belong to the do_to schema include those that we would express in English using the verbs hit, touch, injure, cure, carry, and bathe, as well as more abstract relations expressed by verbs such as as appoint, insult, and abandon. Note that the agent must be animate (that is, human or animal), or at least construed as animate; otherwise, it can't consciously initiate the action. But the patient can be either animate or inanimate, though some relations, such as appoint require it to be animate or even human.

To make it clearer what do_to involves, let's consider some event types that are not instances of this general category. Events designated by verbs such as love, see, know, and understand do not belong to the do_to schema both because there is no conscious initiator of the process and because there is not necessarily a participant that is affected by the process. That is, they have no agent and no patient. For example, we normally see or know something without a conscious effort on our part, and the thing that we see or know is not affected by our seeing or knowing. The event described in the sentence Clark became a reporter is not an instance of do_to because there is only one participant, and the state described in the sentence Clark resembles Superman is not an instance of do_to because there is no event to be consciously initiated by a participant in the first place.

It is important to note at this point, however, that a category such as do_to is similar to the other kinds of categories we've seen in the book. That is, there will be prototypical instances of the category (for example, the event designated by the sentence Lois hit Clark), and other cases where it is not so clear which category an event belongs to. We will also see cases where an event can be described as a blend of more than one category.

The do schema

Consider the events designated by the following sentences.

  1. Clark worked all day.
  2. Lois played all day.
  3. Jimmy just eats.
  4. Perry was singing.
Agent is a general semantic role that can occur with or without a patient.

In each of these cases, there is a conscious initiator of some process, that is, what we could call an agent, but no second participant that is affected by the process. We obviously cannot call these events instances of do_to, but they resemble do_to because of the presence of an agent. I will call the schema for such events, as well as for do_to events, the do schema. This more general schema is characterized by the presence of an agent, that is, something animate that consciously initiates some process. When there is also a patient, the event belongs to do_to, a subcategory of do.

The experience schema

Each of the following pairs of sentences describes the same situation. How does the construal differ for the two sentences?
  • Lois is looking at the bird. Lois sees the bird.
  • Lois is smelling the rose. Lois smells the rose.
  • Lois is listening to a rustling in the backyard. Lois hears a rustling in the backyard.
Sentences that look similar in English may describe quite different kinds of events or states.

An event or state may involve two participants, as in the do_to schema, but with a very different relation between the two than for do_to. Consider the situations designated by the following sentences.

  1. Clark saw Lois and Jimmy.
  2. Clark heard something strange.
  3. Clark likes Lois and Jimmy.
  4. Clark knows Perry.
  5. Clark understood the word.

In each of these cases, there is an animate participant, Clark, who experiences something that he doesn't directly control and therefore doesn't initiate, so he is not an agent. I'll call this participant the experiencer. The second participant can be almost anything, and, unlike the patient in the DO_TO schema, it is normally not affected by the experience, though it does in some sense lead to the experience. I'll call this participant the theme. We'll see this term later. In general, it is used to refer to a role that is central to the meaning of the state or event but does not participate directly in it. Situations like these belong to what I'll call the experience schema.

One point to note about instances of experience is that, though they are probably best viewed as states, they also have properties of events. To the extent that we see these experiences as unchanging situations, we are viewing them as states. To the extent that we see them as mental processes that take place or we focus on the beginning of the process, we are viewing them as events.

Note how the meanings of sentences involving sensory experience such as 7 and 8 above differ from similar sentences in which the experiencer plays a more active role.

  1. Clark looked at Lois and Jimmy.
  2. Clark listened to the music.

In these cases, Clark is consciously attending, either visually or auditorally, to something. That is, Clark is the agent of the process. However, in these cases the other participant is not affected by the action of the agent, so we cannot call it a patient. Rather the second participant can be seen as a kind of target for the agent. I will again call it the theme to emphasize its similarity with the theme of the experience schema. In both cases, this participant is providing the input to an experience. In a sense, then, sentences 12 and 13 designate events that are a blend of the experience and do schemas.

The happen and move schemas

Each of the following pairs of sentences could describe the same situation. How does the construal differ for the two sentences?
  • The water boiled. Clark boiled the water.
  • The ball rolled. Clark rolled the ball.
  • The shirt got clean. Clark washed the shirt.

Just as we can have events with an agent but no patient, we can have events with a patient but no agent. Consider the events designated by these sentences.

  1. The lake froze.
  2. My watch stopped.
  3. This shirt is fading.

In these examples, there is a single participant that undergoes a kind of change of state. Because this participant is directly affected by the process, it is a patient. I will refer to this as the happen schema; something happens to the patient. Note that we can see do_to as a subcategory of happen as well as a subcategory of do because it has a patient as well as an agent role.

When the patient of a happen event is animate, this schema resembles the experience schema, except that there is no second participant.

  1. The cricket died.
  2. I got sick.

An important subcategory of happen involves a change in the location of the patient. I'll call this the move schema. It is illustrated by the following sentences.

  1. The ball rolled off the ledge.
  2. A rock fell in my lap.
  3. The tractor crossed the field.

In addition to the patient, the move schema has three other roles that specify the movement. When something moves, it has a place where the movement begins, the source; a place where the movement ends, the goal; and a way of getting from the one to the other, the path. For a given construal of an instance of move, any of these may be seen as worthy of attention. Sentence 19 above mentions the source, sentence 20 mentions the goal, and sentence 21 mentions the path.

When the patient of a move instance is animate, it often has control over the movement and the event becomes a subcategory of the do schema rather than the happen schema. This possibility is illustrated in the next two sentences.

  1. Clark jumped off the ledge.
  2. Lois crossed the field.
It is possible for the same participant to fill two different roles at the same time.

In such cases the patient is also an agent; for example, in the event described by sentence 22, Clark is both affected by the process and the initiator of the process. But there may also be an agent that is separate from the patient; in such cases we have instances of do_to as well as move. This possibility is illustrated in the next three sentences.

  1. Clark put the book on the desk.
  2. Clark took the book from the table.
  3. Clark pushed Simon off the ledge.

The transfer and information_transfer schemas

Another common schema combines move and do_to, but in this case there is a second animate participant, the recipient, which functions as the goal for the patient. The source is also animate. I'll call this the transfer schema. It is illustrated in the next four sentences.

  1. Lois gave Clark a CD.
  2. Lois gave a CD to Clark.
  3. Lois sold Clark her computer.
  4. Clark borrowed a computer from Lois.

A transfer is actually quite complicated. Before the event there is a state in which the source of the transfer is in control of the patient. For example, for sentence 27 Lois owns the CD before she gives it to Clark. After the event who controls the patient has changed; it is now the recipient of the transfer. For example, for sentence 27 Clark owns the CD after Lois gives it to him. Either the source or the recipient can be the agent of the event. In sentences 27-29, the agent is the source; in sentence 30, the agent is the recipient.

Another type of transfer is also possible, one involving the "moving" of information from the agent to the recipient. In information_transfer, there is really no patient, however, because the information itself is not affected by the transfer. I'll refer to the information as the theme. And the agent doesn't really lose the information in the transfer, as is the case with the transfer of physical objects. Here are some examples of information_transfer.

  1. Lois told Clark her password.
  2. Lois informed Clark of the meeting.
  3. Lois asked Clark the time.

Sentence 33 is a little complicated. Here the information that is transferred, the theme, that is, is not the time itself but rather Lois's desire to know the time.

The be schema

How is the noun doctor used differently in the following two sentences? That is, how is the Hearer expected to use the information provided by the word (the category of people, doctor)?
  • A doctor came in the room.
  • Carla is a doctor.

Finally, we have what is in a sense the simplest schema, the one for prototypical states. I will call this the be schema and will refer to the participant that is in the state as the theme. Instead of verbs for each particular type of state, the different states are defined mainly by the nouns and adjectives we already discussed in Chapters 2 and 5. Here are some examples.

  1. Lois is sick.
  2. Lois is a Hoosier.
  3. This is a keyboard.
  4. The keyboard seems sticky.
  5. The keyboard smells funny.
Nouns are not only used to refer to things. They may also designate a state that a given thing is in.

Though we are not focusing in this section on the forms that languages use for the different schemas, it is important to see how the nouns and adjectives at the ends of these sentences are used differently than they were in the examples we saw in Chapters 2 and 5. Previously we only considered nouns and adjectives as a means of referring, of pointing to particular individuals in the world or the mind of the Hearer. But a Hoosier in sentence 35 is not meant to point to particular instance of the category hoosier. Rather it functions to call the hearer's attention to a state that Lois is in, the state of belonging to the category hoosier.

Similarly, compare sick in sentence 34 with sick in the sentence the sick child is improving. In this second sentence the adjective has an attributive function; it narrows down the range of possibilities for who is being referred to so the Hearer can figure out what child is meant. But in sentence 34 sick is used to call the attention of the Hearer to the state that Lois is in, the state of being sick. Sick in sentence 34 and Hoosier in sentence 35 are said to perform a predicative function rather than an attributive or referential function.

There are two very important subcategories of the be schema that are relations. One of these involves possession or control of one participant over another. I'll call this the have schema. Here are some examples.

  1. Lois has a digital camera.
  2. Clark has a lot of cash.

I'll refer to the two roles in have states as the possessor and the theme.

Another important subcategory of be is also relational; the state concerns the spatial or temporal relation between the two things. We can call this schema be_at. Here are some examples.

  1. Lois is behind Clark.
  2. Lois is next to the car.
  3. Bloomington is south of Indianapolis.
  4. The concert is on Monday.
  5. The concert is after the class.

There are two participants in the states designated by each of these sentences. It is normally possible to reverse the way the relations are expressed. So corrsponding to sentences 41 and 43, for example, we have the following possibilities.

  1. Clark is in front of Lois.
  2. Indianapolis is north of Bloomington.

Such reversals do not change the relation between the two participants at all. Hence there is no good reason for distinguishing the two participants, and I'll refer to both as themes. Of course this does not mean that sentence 41 is completely synonymous with sentence 46. The difference has to do with the perspective that the Speaker is taking on the state, which participant is being singled out to say something about. This sort of factor won't concern us in this chapter though.

Note that when the relation is a temporal one, as in sentences 44 and 45, the two "things" that are related are actually events, in 44, the concert and a particular Monday. But by referring to the events with nouns, the Speaker of the sentence is treating them as abstract things that are "located" at different points in time.

Other semantic roles

How does the semantic role of Clark differ in the following three sentences?
  • Lois sent Clark a message.
  • Lois did Clark a favor.
  • Lois did Clark an injustice.
Participants can vary in terms of how central they are to the particular kind of event or state.

We can identify a number of other semantic roles that can occur together with the basic schemas described above. The roles we have seen so far are those that are central to each of the schemas; I will refer to the things playing these roles as the core participants in the states or events. But there is also the possibility of some relatively peripheral participants, those that are not central to the state or event that is being thought or talked about. For example, all states or events can have a location or time. Here are some examples that make these roles explicit.

  1. Lois kissed Clark behind the barn.
  2. Lois kissed Clark before the class.
  3. Lois is sick around Jimmy.
  4. Lois was sick on Tuesday.

In instances of the do schema, especially the do_to schema, there is the possibility of a participant that is used by the agent to achieve the process. This role is called the instrument. Here are some examples illustrating instruments.

  1. Lois broke the window with a stick.
  2. A stick broke the window.
  3. Lois used outdated software to write her report.

The sticks referred to in sentences 52 and 53 play the same role in the breaking events; both are instruments. The main difference is that the agent is not mentioned in 53. But we can assume that even in this case, there was an agent behind the action; that is, the stick didn't break the window of its own accord. In sentence 54, the software is the instrument for the writing of the report.

In instances of do, and again especially do_to, another possible role is an animate participant that benefits from the action, a beneficiary. Here are some examples in which Clark is the beneficiary.

  1. Lois fixed the window for Clark.
  2. Lois fixed Clark a great meal.
  3. Lois sent Perry a letter for Clark.

Note that in sentence 56 there is both a recipient, Perry, and a beneficiary, Clark.

An event may also harm, or otherwise adversely affect, someone other than the patient. I'll call this participant the sufferer (there is no generally agreed on name). Here are some examples, but, as we'll see later, English does not offer as direct a way as languages such as Japanese and Amharic to refer to the sufferer.

  1. Lois lost her checkbook.
  2. Lois and Clark went and eloped on me.

In sentence 55, the checkbook is the patient, and the loss of the checkbook has an adverse effect on Lois. In sentence 56, Lois and Clark (together) are the agent, and their action has an adverse effect on the speaker.

A final possibility is a semantic role that is not really peripheral but that results from viewing an event as having a more direct and a less direct cause. The agent is the participant mostly directly responsible for the process, but there may be another participant that causes the agent to carry out the process. I'll call this the causer. Here are some examples.

  1. Lois got Clark to write the report.
  2. Lois had the report written.

Notice that in setence 58, no agent is mentioned — we do not know who actually wrote the report — but we know that somebody other than Lois did it, so Lois must be seen as the causer rather than the agent.

Summary of semantic roles

Let's summarize the different semantic roles we have met in this section. Remember that for each role there are clear-cut, prototypical examples and other, less clear cases where a participant may share properties with multiple roles.

The core participant that directly causes the event. Prototypically human, or at least animate, in which case there is conscious intention behind the act, and the question what did agent do? is appropriate.
It is easily confused with an experiencer, which is also animate. The difference is that an experiencer isn't consciously behind the experience; it just happens to the experiencer.
The core participant that is affected by the event. That is, the patient changes in some way. A patient can be animate or inanimate.
For "going" events, the patient can also be the agent (the agent moves itself). Participants in addition to the patient may also be affected by the event; these include recipient, beneficiary, and sufferer.
The core participant that experiences the event or state without having direct control over it. Since experience requires a nervous system, an experiencer must be animate.
Since the experiencer is affected by the state or event, it can be seen as a subcategory of patient. Some kinds of events, such as "looking" and "listening", combine experience with conscious action; I am treating the participants that cause these events as agents, but they overlap with experiencers.
The core participant that possesses or controls a second participant in a state. The other participant is the theme. The possessor is prototypically animate.
A core participant that neither causes the state or event nor is directly affected by it. The notion of theme is a weakness in this account of semantic roles because it is defined almost completely in negative terms; it is really a place to put core participants that don't belong to some other role. Spatial relation and temporal relation states have two themes.
A core participant that causes an agent to cause the event, though the agent may not be explicitly referred to. A causer is normally human.
source, goal, path
Three roles that define the spatial aspects of a move, transfer, or information_transfer event. In a transfer or information_transferevent, the source and goal are the animate participants in the event; the goal is the recipient.
The animate goal of a transfer or information_transfer event. In a "taking", the recipient is also the agent.
The participant that is used by the agent to achieve the event.
Where there is a separate causer and agent, the agent is a sort of instrument for the causer, but the agent can also have its own instrument: Lois had Clark take off the tire with the wrench.
beneficiary, sufferer
Animate, relatively peripheral participants that are affected either beneficially or adversely by the event.
These need to be distinguished from patients, which are core participants but can also be animate and affected either beneficially or adversely, for example, in Carla killed Frank.
location, time
Inanimate, peripheral participants that situate the state or event in space or time.

The set of schemas and semantic roles that I have described is not just an arbitrary way of dividing up the states and events that people think and talk about. The distinctions between the different schemas and roles are reflected in the forms that different languages use to designate different kinds of events and states. This is the topic of the rest of this chapter.

TOC Email author Printer Friendly Next