How Language Works

6 Sentences (Part 2: 6.3, 6.4, 6.5, 6.6)

Constituents

We have seen that the situations in the world that people tend to notice and talk about are construed as states or events involving a small number of participants. We can expect the sentences that designate states or events to have a word that specifies the kind of state or event, usually a verb, and a phrase for each of the participants that the speaker chooses to refer to. Each of these makes up a part, or constituent, of the whole sentence. Before we can look at how different languages put sentence constituents together to designate states and events, you will need to know more about what makes up a constituent. In this section we'll be focusing on the constituents that refer to the participants in the state or event. And for these constituents we'll be focusing on their internal structure, what words can make them up.

Let's return to where we left our Grammies. Once they have the insight that it is to their advantage to be able to talk about events and states, the next step might be to come up with words for the different categories of states and events, verbs, that is. In the simplest case, then, we can imagine describing events and states with verbs alone, for example, recovered for a recovery event or gave for a giving event. In some modern languages, such as English, the grammatical conventions don't permit sentences like this consisting only of a verb, but in languages such as Japanese, they are possible. So the following are grammatical Japanese sentences.

1. naotta 'recovered'
2. ageta 'gave'

The first sentence means that somebody recovered; the second means that somebody gave something to somebody else.

But even in Japanese, these one-word sentences can only go so far. Unless it is clear from the context, the hearer may have trouble figuring out who the patient (the person recovering) is for sentence 1 and who the agent and the recipient and what the patient is for sentence 2. Such sentences are massively ambiguous and only interpretable when the context makes it very clear who the participants in the events are. Something else is needed.

Noun phrases

Faced with this sort of ambiguity, the Grammies hit upon the same idea that they had used earlier on to make it easier to refer to things, the principle of compositionality. Remember from the last chapter how this works. Words combine into phrases, and the meaning of the whole phrase depends on the meanings of the individual words. Here's a simple English example.

3. Fred recovered.

We know from Chapter 2 that a proper noun like Fred can refer directly to an individual, making this the easiest way to make the Hearer aware of who is being talked about. By compositionality, we get the meaning of sentence 3 by figuring out the conventional meanings of each of the words, Fred and recovered, which are available in our lexicon, and then applying the appropriate grammatical rule to combine the meanings. We'll talk about the grammatical rule in the next section. For now, the important point is that the meaning of sentence 3 can be derived from the meanings of Fred and recovered and what we know as speakers of English about the grammar of English sentences.

But we also know from Chapter 2 that it is impractical to have a name for every thing we might want to refer to, and in any case, no speaker can know all of the names there are. So if the speaker of sentence 1 hadn't known Fred's name, the sentence would have come out differently, perhaps like this one.

4. The teacher recovered.

Even though the phrase the teacher in sentence 4 and the word Fred in sentence 3 consist of completely different words, they are doing the same work in the two sentences, referring to a particular individual. But there are many other ways the Speaker could refer to Fred, depending on the utterance context, what the Speaker knows about Fred, and what the Speaker believes the Hearer knows about Fred. Here are a few possibilities.

5. He recovered.
6. That teacher recovered.
7. My teacher recovered.
8. That tall teacher recovered.
9. The teacher behind me recovered.
10. The teacher who got fired recovered.

In each sentence the part before recovered refers to the person who recovers (the patient). These phrases do not all mean the same thing, but in one sense they all do the same thing. And with respect to just form and not meaning, we can see that they all share the property of being possible phrases before a verb like recovered in a sentence. All such phrases are called noun phrases (abbreviated "NP"). An NP need not actually contain a noun; for example, the NP in sentence 5, he, does not. But the prototypical NP does contain a noun. This noun is called the head of the NP; this is a term we already saw in the last chapter.

English noun phrases

What else can an English NP contain? One possibility is an adjective, for example, in the tall teacher. In English this always comes before the noun. In fact an NP can have more than one adjective: the tall sick teacher. We can think of an NP as having a set of positions or slots in which words can appear. There is a position for the head noun, and a position for one or more optional adjective modifiers. We can denote these possibilities as follows, where adjective and noun are abbreviated as "Adj" and "N".

Adj* N

The "*" after "Adj" means that some number of adjectives, including zero, are possible.

But note that tall teacher is not really a grammatical NP; it requires a word such as the or a before the adjective. If we looked at a lot of English noun phrases, we'd discover that these words always precede the adjectives in a noun phrase if there are any adjectives. We'd also discover that there are a number of other words that can appear in this same position, that is, at the beginning of an NP and before any adjectives. These words include that, this, some, my, and your. We'd also discover that English doesn't permit more than one of these words in a given NP. We can't say things like the my boss (though such combinations are possible in some other languages). This position in the NP is called the determiner position, and we can also use this word to refer to the words that can fill that position. So now we can denote the structure of NPs that have a common noun such as teacher as follows, using "Det" to mean determiner.

Det Adj* N

In plain English, this expression says that an NP can consist of a determiner followed by some number of adjective modifiers (including none) followed by a head noun.

We are by no means finished characterizing the structure of English NPs, but it is useful to summarize what we have so far. English NPs have three positions in them that can be filled by words. Just as we can speak of the roles in an event or state, we can speak of the roles in a phrase. But since with a phrase we are concerned with linguistic form rather than meaning, these are syntactic roles, rather than semantic roles. As with semantic roles, we can distinguish the role from the category of things that fills the role. For example, the source and the goal in an instance of move are both filled by things belonging to the category place. For phrases, we can talk about either the categories of the words that make up the phrase or the roles that the words play in the phrase. For NPs, the roles we have seen so far are the following, where "PreMod" means "pre-modifier", a modifier that comes before the head. Note that for the determiner, we use the same word for the role and the category.

Det PreMod* Head

If we look at more NPs, we will see that this structure doesn't fit them. In sentence 9 above, the NP takes the form the teacher behind me. Here the phrase behind me functions as a modifier within the NP. Just as an adjective can modify a noun by narrowing down the category of things to those with a particular attribute, a phrase such as behind me can narrow down the category of things to those members of the category that are in a particular location.

In sentence 10 there is another example of a modifier following the head noun. In this sentence the NP is the teacher who got fired. Within the NP the phrase who got fired behaves like a modifier; it narrows down the category of things designated to members who got fired (that is, who are the patients of a firing event).

The NPs in 9 and 10, then, can both be denoted as follows, where "PostMod" means post-modifier, a modifier that follows the head noun.

Det Head PostMod

There are two important points to note about the syntactic roles in NPs. First, unlike for semantic roles, there is a sequential order to these roles. The expression above does not only mean that an NP can have a determiner, a head noun, and a post-modifier; it means that they appear in that order. Second, the roles can be filled either by single words — the determiner in the teacher who got fired — or by whole phrases — the post-modifier in the teacher who got fired. That is, the constituents of phrases can themselves be words or phrases. In fact, an NP can have another NP inside it; for example, the NP the teacher in front of me contains the NP me. Since an NP can contain an NP, we should not be surprised to see NPs that contain NPs that contain NPs, for example, the teacher next to the door that leads to the exit. In fact there is no obvious limit to this process.

We can also make more complex NPs in English by adding additional post-modifiers, as in the teacher that got fired that I told you about. It is also possible to combine pre-modifiers and post-modifiers in the same NP, as in the tall teacher that I told you about. If we want to capture all of these possibilities in one schema, we can write the following.

Det PreMod* Head PostMod*

This says that an NP can consist of a determiner followed by some number of pre-modifiers (including none) followed by the head noun followed by some number of post-modifiers (including none). We can also diagram this as we did with schemas for events and states. The dotted lines indicate constituents that can be absent, and the arrows indicate the order of the constituents.

Let's summarize what we've learned about English NPs. They serve as constituents in sentences, where (in one of their functions) they refer to the participants of the state or event that the sentence designates. That is, they have a particular function to perform within sentences. They can consist of as few words as one, and there is no clear limit on their maximum length. They can be described as consisting of constituents, parts that are either words or phrases in their own right and that fill particular syntactic roles in the structure of the NPs. For compositionality to work, there should be a grammatical rule for each of these roles that specifies how the meaning of the whole phrase depends on that role. In the last chapter, we saw informally what some of these rules would look like.

Note how NP form and meaning resemble each other. On the meaning end, we have the thing referred to with a number of properties, all of them localized in a single bounded region of space. On the form end, we have a group of words occurring together within a sentence, that is, not separated by words that belong to other constituents. This is a weak example of iconicity because it involves an aspect of form (coherence) that applies to meaning as well. The tendency for the words that make up a constituent to occur together is quite strong among the languages of the world, though there are some exceptions where other principles can override this tendency. The tendency for words that behave as units on the form end to behave as units on the meaning end of language is what I will call constituency.

Before we leave English NPs until the next section, it is worth saying that we are far from characterizing adequately the structure of English NPs. The schema above implies that all NPs with a noun head have a determiner, but there are perfectly grammatical English NPs such as fresh cheese and rocks that do not. It is also possible for a pre-modifier to be a phrase itself, as in the extremely tall, slighly bald teacher. Finally, the determiner can also be a phrase, as in my sister's teacher. But all of this goes beyond the scope of this book.

Lingala noun phrases

All languages seem to resemble each other in having phrases for referring to the participants of events. Most languages have a category of common nouns that can play the role of head in these phrases; that is, these languages have NPs like English. Many, but not all, languages also have a category of adjectives that can modify nouns. And probably all languages have at least some determiners, though many languages have no words corresponding to the English words the and a.

Let's look at NPs in Lingala as an example of a language that differs considerably from English. Here are the Lingala sentences that correspond to two of the English sentences above.

From sentence 11, we see that Lingala is like English in allowing proper nouns to refer to things. From sentence 12, we see that Lingala is like English in allowing NPs with adjectives and determiners. It turns out that Lingala has very few adjectives, using alternative expressions in many places where English uses adjectives, but this won't concern us here. It also turns out that Lingala has no determiners corresponding to English the or a, but, as in the example, it does have words corresponding to English determiners like this and that.

Note how the order of the constituents in Lingala differs from English, however. We would have to look at a lot of NPs to make sure, of course, but the order of the constituents in 12 holds throughout the language. It also turns out that, unlike in English, there are never modifiers before the head noun. So the structure of Lingala NPs, at least those with common nouns as heads, appears to be as follows. The determiner is dashed to indicate that this constituent is often absent in NPs.

So one important way in which languages can differ is in the conventional order of the constituents. Each language will tend to have a preferred order, but the order may vary from language to language.

Armed with some idea of what NPs look like, we are ready to return to events and states. In the next section, we'll be concerned with how verbs combine with NPs to designate events and states.

6.4 Subjects

Let's return to our Grammies and their attempts to talk to each other about events and states. We saw that verbs go part of the way. They allow them to distinguish different kinds of states and events from one another, to say that an event belonging to a particular category such as eat has taken place or that a state belonging to a particular category such as be_hot is true. But, as we've also seen, verbs by themselves can't make it clear who the participants in the event or state are. For this the Grammies need to be able to produce and understand sentences consisting of verbs and NPs.

Given a sentence, what should Hearers be able to figure out about it?

1. They should be able to tell what category of event or state is being described; the verb (and in some cases a predicate noun or adjective) should make this clear.
2. They should be able to tell who the relevant participants in the event or state are, at least those that the Speaker chooses to refer to. This should be derivable from the different NPs in the sentence, though I haven't said much about how they are interpreted by Hearers.
3. They should be able to tell what the semantic roles of those participants are in the event or state.

It is this third property of sentences that we'll focus on in this section and the next two sections. We'll look at some examples of how English solves this problem and how other languages may differ from English. We'll return to this issue again in Chapter 8.

Talking about be states

In this section we'll look only at events or states that have only one core participant, that is, happen events and the simplest be states. In English the most common pattern for be states with only one participant uses the verb be (which takes the different forms am, is, are, was, were, etc.) followed by a predicate NP or adjective. A few other verbs such as seem, taste, smell, look, sound, and feel are also possible. These verbs can be followed by adjectives or by like followed by an NP. Here are some examples.

1. Clark is mad.
2. The woman is a teacher.
3. Your brother looks terrible.
4. That stuff tastes like soap.

Note how be (is, are, am, was, etc.) functions. It has very little content; it just marks this as a be state, leaving the nature of the state to be specified by the adjective or NP that follows. (It also has the possibility of marking the time of the state; this is the difference between is and was.) Spanish, Hindi, and Amharic are like English in this regard; each has a verb that functions like English be (in fact Spanish has two such verbs). Mandarin Chinese also has such a verb, but it is used only in sentences with predicate nouns like 2, not in sentences like 3 that have predicate adjectives in English.

American Sign Language, Tzeltal, and Inuktitut have nothing like English be in either kind of sentence. Here is the Tzeltal sentence corresponding to sentence 2.

Notice that in the English sentences the single participant, the theme, is referred to by an NP that appears right before the verb: Clark, the woman, your brother, that stuff. This position, or syntactic role, called the subject, is basic to English sentences. (To help keep syntactic roles distinct from semantic roles, I'll write them in lowercase.) The subject of an English sentence can usually be identified as the NP that directly precedes the verb. For most personal pronouns, the subject in English also has a special form that it does not take elsewhere in the sentence: I (rather than me), he (rather than him), she (rather than her), we (rather than us), they (rather than them). So we say I am alive, not me am alive (though in some dialects, her and me are alive is grammatical).

In summary, English subjects are distinguished from other NPs in two ways:

• by their position in the sentence
• for personal pronouns only, by their form

Other languages also have a syntactic role that we can call the subject. As in English, it is the role that refers to the themes of be sentences like those above. But as the most basic syntactic role in the sentence, the subject is also associated with one of the core participants in the other semantic schemas that were discussed in the section on semantic roles. As we'll see later, languages tend to agree with one another in what semantic role the subject refers to for these other kinds of sentences, but the agreement will not be at all perfect.

We've seen that for English the subject is marked by its position within the sentence and (for a small number of cases) by its form. These two possibilities apply to other languages as well. In Spanish, as in English, the subject tends to appear before the verb. But in Spanish this is just a strong tendency; the subject can also appear after the verb, a position it can never be in in English. In Tzeltal, as you can see from sentence 5, the subject normally appears after the "verb" (there is really no verb in the sentence but the prediicate noun jp'ijubteswanej 'teacher' acts something like a verb). As in Spanish, though, this is just a tendency; the word order in Tzeltal is relatively free.

Spanish, like English, has a set of special forms for some of the personal pronouns when they are used as subjects; for example, yo 'I' is used only as a subject. Note that we can now add to our discussion of the dimensions along which personal pronouns vary that we began in the section on meaning differences between languages. Remember that we isolated a small number of dimensions, including person, number, and gender, that distinguish personal pronouns in many languages. Just as I and you are distinguished by person, and he and she are distinguished by gender, we need a further dimension to characterize the difference between I and me. This dimension is called case. The case of an NP, including a personal pronoun (which is just a very simple kind of NP), specifies its syntactic role in the sentence.

The case for the subject in English and most (but not all) other languages is called the nominative case (abbreviated nom). In English, nominative case is marked explicitly only for personal pronouns. I, he, and we are nominative personal pronouns; me, him, and us are not. The pronouns that are not normally used as subjects, me, him, her, us, and them, are called objective.

Remember from our discussion of the dimensions that distinguish personal pronouns that languages differ in terms of which dimensions are relevant for a given set of words. Just as gender is relevant for English, Spanish, Japanese and Amharic pronouns but not relevant for Mandarin Chinese, Hindi, Lingala, Tzeltal, Inuktitut, or American Sign Language pronouns, case is relevant for English, Hindi, Spanish, and Inuktitut pronouns but not for Lingala, American Sign Language, or Japanese pronouns. Lingala, ASL, and Japanese use the same forms for personal pronoun subjects and non-subjects.

Talking about happen events

Like the be states discussed above, happen events have a single core participant, the patient. Most languages use the subject for this role too.

Many types of happen events have particular English verbs associated with them, for example, fall, boil, and die. Here are some examples.

6. The milk boiled.
7. My goldfish died.

Note that there a clear relationship between some be states and some happen events. For each type of state, there is an event that results in that state. There are two ways this relationship is reflected in English sentences. One possibility is for a particular verb to be used with the happen event and an adjective that is related to that verb to be used in the be state. Here are some examples of this possibility. For each, the sentence describing the event appears first, then the sentence describing the state.

8. Rex died. Rex is dead.
9. The water froze. The water is frozen.
10. Lois woke up. Lois is awake.

In some languages, such as Amharic and Japanese, the relationship between the state and the event is made even more explicit. The event and the state use the same verb; in fact the two may be indistinguishable from one another. Here are two Amharic examples; only the context can make it clear whether the change of state or the resulting state is intended.

Another possibility in English (and many other languages have a similar possibility) is for an adjective or NP to be used in both the state and event sentence, with the verb be for the state and the verb become or get for the event. Here are two examples, again with the event first.

13. Clark got sick. Clark is sick.
14. Jimmy became a journalist. Jimmy is a journalist.

Notice how the relationship between these two sentence patterns in each case simplifies matters for the language learner. Instead of learning a separate word for each type of state and a change into that type of state, the learner is only responsible for one word (an adjective such as sick or a noun such as journalist). In order to produce both kinds of sentences, the learner only has to learn the general rules that relate the two kinds of sentence patterns. Here is an informal way to state those rules.

be adjective (state) ↔ get adjective (change of state)

be a noun (state) ↔ become a noun (change of state)

This is yet another example of productivity, the property of language that permits speakers to make new expressions using a general rule. In this case, the rule is based on a generalization that is made in English and many other languages: for every type of state, there is a corresponding change of state that results in that state.

In this section we've seen some of the ways that languages describe events or states with only one core participant. The focus has been on the syntactic role used for that participant, the subject. There will be lots more to learn about subjects, but that will have to wait until we've looked at how other participants are handled within sentences. That's the topic of the next two sections.

6.5 Direct objects

Composition and sentences

Now what happens when the Grammies need to refer to events or states with two core participants? Let's start with instances of the do_to schema because the semantic roles are very clear-cut and because children seem to learn about these sentences relatively early.

Say an instance of eating has taken place, the agent (the eater) is Clyde, and the patient (the eaten thing) is a mango. Now one Grammie wants to tell another about what happened. They have a verb for eating — let's call it eat — and nouns for the two participants — let's say Clyde and mango. The speaker could just string the words together: eat!, mango!, Clyde!. Assuming these three words are in the hearer's lexicon, the hearer could figure out that an instance of eating is being described and that the participants in that eating are Clyde and a mango.

But the two participants in an eating play very different roles. One is an agent, the other a patient. Understanding what the speaker is saying involves also figuring out "who did what to whom", that is, which participant fills which semantic role. In our eating example, this would not be difficult. The hearer could use knowledge of what eating is to infer that Clyde is the agent (eater) and that the mango is the patient (eaten thing). Eaters must be animate; eaten things are often inanimate and are almost always members of the general category of food, which includes the more specific category mango.

But things would not be as simple for an event such as a hugging. In a hugging, both participants are normally animate. So if a Grammie speaker referred to a hugging by stringing the verb and the two nouns together in a random order — hug!, Clyde!, Lois! — the hearer would not know whether Clark hugged Lois or Lois hugged Clark (unless this was somehow clear from the context). What's needed is a convention for how the semantic roles of a two-participant event correspond to the NPs in a sentence describing the event.

But this issue should be familiar from the last chapter. It's just what we expect from compositionality: language-specific grammatical conventions specify how the meaning of a phrase is derived from the meanings of the words or phrases that make up the phrase. The problem we face with sentences describing two-participant events is exactly the same sort of problem that must be solved for English noun + noun phrases of the type we learned about in this section. Given the unfamiliar phrase bag box, for example, a hearer has to be able to figure out whether this refers to a box that has something to do with bags or to a bag that has something to do with boxes. An English-speaking Hearer knows what to do because of a grammatical convention of English: it is the second noun in these phrases that is the head, that is, the noun that specifies the category for the phrase's meaning. (More precisely, the meaning of the phrase is a subcategory of the meaning of the second noun.)

Note that the solution for noun + noun phrases involves the notion of syntactic roles, particular positions within the phrase that play particular roles in the meaning of the phrase. The syntactic roles for noun + noun phrases (and adjective + noun phrases) are modifier and head.

For sentences the problem is the same. A compositional solution to the problem is one that says that the meaning of a sentence is some combination (specified in a grammatical convention) of the meanings of the verb and the other constituents of the sentence. Since the meaning of a sentence, at least as laid out in the theory of semantic schemas earlier in this chapter, involves semantic roles, we need a convention that relates particular sentence syntactic roles to particular semantic roles. We already have one sentence syntactic role for the NP in a one-participant sentence, the subject. We saw that this role is marked by its position or its form or both its position and its form in different languages. Most, if not all, modern languages use this same role within sentences that refer to multiple participants.

We need one more syntactic role, then, for the other participant in sentences describing do_to events. The syntactic role that does this is the direct object. Like the subject, the direct object is thought to be a universal role, something that all languages have, though the way in which the direct object can be identified differs somewhat from language to language. A sentence with both a subject and a direct object is called a transitive sentence. Transitive sentences contrast with intransitive sentences, which do not have direct objects. The grammatical case for the direct object in most (but not all) languages is called the accusative case (abbreviated acc).

Direct objects in English

Let's start with English. The English sentence that designates a hugging event in which Clyde is the agent (the hugger, that is) and Lois is the patient (the "huggee", that is) is of course the following.

1. Clark hugged Lois.

We know from the position of the NP Clark in this sentence that it is the subject. The other NP in the sentence, Lois, is the direct object. The normal position for the direct object in an English sentence is directly after the verb. (In the next section, we'll see one situation where the direct object is separated from the verb by another constituent.) Thus, like the subject, the direct object in English is usually identifiable by its position. Unlike the subject, the English direct object doesn't have a set of special forms for the personal pronouns; forms such a me and him are also used in other contexts. That is, English doesn't really have accusative pronouns. However, forms such as me and him are clearly not the subject. So consider the next sentence, which could describe the same event as sentence 1 does.

2. He hugged her.

In this sentence we know that the agent is the man not only because of the position of he before the verb but because the form of the pronoun is he and not him. By the same token, her is clearly the direct object, and not the subject, because of its position, but also because it is in the non-subject, objective form her.

To summarize, English has two syntactic roles that Speakers use to refer to the core participants of a two-participant event. These roles, subject and direct object, are identifiable mainly on the basis of their position with respect to the verb. In sentences describing do_to events, that is, events with a clear-cut agent and patient, the subject refers to the agent and the direct object refers to the patient.

The relationship between the syntactic roles of a sentence and the semantic roles of the event or state that the sentence describes is an example of a mapping. The notion of mapping is a concept from mathematics that is very important in cognitive science. For a mapping there are two sets of elements, and each of the elements in one set corresponds to (or "maps onto") an element in the other. In our case each of the syntactic roles in the transitive sentence maps onto a semantic role in the situation. The figure below illustrates this. Because we are concerned now with the relationships between form (syntax) and meaning (semantics), there are now two large boxes. The green arrows denote these relationships. The one connecting the two large boxes denotes the fact that the sentence describes the event. The ones connecting the syntactic roles in the sentence with the semantic roles in the event denote the correspondences in the syntax-semantics mapping. The "<" symbols between the syntactic roles represent their usual order.

Direct objects in Japanese and Spanish

Now let's look at subjects and direct objects in Japanese, which are a bit more complicated. One possible Japanese translation of sentence 1 above would be the following sentence.

Notice first that the order of the sentence constituents is different in Japanese. The prototypical position for the verb in a Japanese sentence is at the end, so both the subject and the direct object precede it. The subject usually precedes the direct object, though this order is not so rigid. Second, notice that the Japanese sentence has two words that don't correspond to anything in the English sentence, wa and o. These words, sometimes called particles, specify the role of the NP they follow in the sentence. The particle o specifies that the previous NP is in the accusative case, that is, that it is the direct object of the sentence (though the detailed function of o is somewhat more complicated than this). A word like this is called a case marker. The particle wa has a different sort of function; it specifies that the NP it follows is the "topic" of the sentence, roughly what the sentence is "about". Notice that in this case wa follows the subject of the sentence. It turns out that Japanese also has a nominative case marker, ga, but when the subject is the "topic", the nominative case marker is replaced by the topic marker wa.

The figure below illustrates the syntax-semantics mapping for do_to sentences in Japanese. The dashed line separates the subject and direct object from the verb, which they normally precede. The "<" symbol between the subject and direct object is fuzzy because this ordering is only a tendency in Japanese.

What is important about all this for our purposes is that Japanese has special words, case markers, to indicate which NP is the subject and which the direct object. That is, even when the order of the subject and direct object deviates from the default order (subject first), a hearer can figure out which NP is which.

What is also important about Japanese is how it is like English. For the do_to schema the syntax-semantics mapping is the same: the subject maps onto the agent, and the direct object maps onto the patient.

Let's look at one more language before we consider other types of states and events. Spanish is similar to English in the default order of the constituents — subject, verb, direct object — but as already mentioned, there is much more freedom in Spanish to deviate from the default. Spanish is also like Japanese in having an accusative case marker, though this is normally limited to direct objects referring to humans. Here is one Spanish translation of sentence 1.

The accusative case marker a appears before the NP Lois, marking it as the direct object of this Spanish sentence. The figure below illustrates the syntax-semantics mapping for Spanish do_to sentences. The accusative case marker is in a fuzzy box because it is only used in certain situations.

Change of state and causing of change of state

For the relationship between states and changes of state, we saw that a language can simplify the task for the learner by making a generalization that recognizes what states and changes of state share. Being crazy and becoming crazy are similar in one way, and the learner only has to learn one new word for each new state and change of state. English makes a similar sort of generalization concerning changes of state and acts of an agent (that is, instances of do_to) that result in the changes of state. Consider the following two sentences concerned with a breaking event.

7. The vase broke.
8. Jimmy broke the vase.

The first sentence is intransitive; it mentions only the patient. The second sentence is transitive; it mentions both the patient and the agent. Notice that the patient appears in different syntactic roles in the intransitive and transitive sentences: the subject of the first sentence, the direct object of the second. This is an example of a generalization that English could, but does not make (though other languages do). That is, English fails to treat the patient in the same way in the two kinds of sentences. On the other hand, English does use the same verb for the two cases, in effect recognizing that the same kind of event is involved.

So what does a speaker of English need to remember about the verb break in the lexicon? One possibility is that there are two separate verbs, one for intransitive sentences like sentence 7 and one for transitive sentences like 8. But this would miss the obvious generalization that is made in English; if these are two completely different verbs, it is very surprising that they are pronounced and written exactly the same. The other possibility is a single verb with a single meaning (whatever it is that breaking involves) and two different patterns for how the syntactic and semantic roles map onto one another.

Here is one way of writing the two syntax-semantics mappings for break. Each of the smaller boxes represents a different mapping, and only one of these appears in a given sentence with break as the verb. The upper box shows the mapping used in sentence 7; the lower box shows the mapping used in sentence 8.

But the English generalization goes beyond this. There is a whole set of verbs with this same property: boil, freeze, cook, fry, bake, steam, soften, thicken, shatter, split, rip, tear. For all of these verbs the same two sets of mappings illustrated in the figure above apply. So English speakers probably learn a general rule that allows them to produce and understand both transitive and intransitive sentences with verbs similar to these even if they have never heard the sentences before, as in the box at the beginning of this subsection.

This English example illustrates a more general point. While we've seen only similarities between languages so far in terms of the syntax-semantics mappings, these are conventions that do differ from one language to another. There will be a number of examples of these differences in the next section. But even within languages, particular verbs may have unpredictable mappings associated with them. For this reason, it is usually assumed that knowledge about how syntax and semantics map onto each other belongs in the lexicon, where it is associated with individual verbs or with clusters of similar verbs (such as break and boil). We'll look at some more syntax-semantics mappings for verbs in English and other languages in the next section.

6.6 Adjuncts

Prepositional and postpositional phrases

As we saw in the section on semantic schemas, an instance of the be or happen schemas can have more than the single core participant, and an instance of the do_to schema can have more than the two core participants. There are additional, peripheral participants that are not central to the particular category of state or event but may be worth noticing or mentioning nonetheless. The semantic roles for these participants include instrument, beneficiary, sufferer, location, time, and manner.

For a given event, a relatively wide range of these roles is possible. For example, a transfer event can have any of the above roles. For this reason, when the Grammies realized they would want to talk about these participants, they decided that they would need more different ways of marking the roles than they had for the subject and direct object. They created a new category of words whose function was to specify the semantic role for these peripheral participants. In English we use such a set of words; we call them prepositions because they appear before (pre-) the NP referring to the participant. Here are some examples.

1. The soup was cooking on the stove.
2. Lois tickled Clark with a feather.
3. Clark cooked some soup for Lois.
4. Lois sang until dawn.

In sentence 1 the preposition on begins the phrase referring to the location of the cooking. In sentence 2 the preposition with begins the phrase referring to the instrument of the tickling. In sentence 3 the preposition for begins the phrase referring to the beneficiary of the cooking. And in sentence 4 the preposition until begins the phrase referring to the time of the singing. Semantically, we can see that the function of a preposition is to specify some relation between a state or event and a thing (the thing that fills the peripheral semantic role). A preposition may also function to specify the relation between two things, as it does for the relation between a cat and a table in the following sentence.

5. The cat is under the table.

Syntactically a preposition combines with an NP to form a larger phrase, called a prepositional phrase. In the last five examples, the prepositional phrases are on the stove, with a feather, for Lois, until dawn, and under the table.

Many other languages, including Spanish, Lingala, and American Sign Language, also have prepositions. But in other languages there are words with a similar function that follow the NPs that they are relating to the rest of the sentence. Japanese and Hindi are such languages. These words are called postpositions, and when we want to group prepositions and postpositions together, we call them adpositions. Here is the Japanese sentence corresponding to sentence 4. Note that the postposition made corresponds to the English preposition until.

Prepositional or postpositional phrases within a sentence are called adjuncts to distinguish them from subjects and direct objects. We'll meet some other kinds of adjuncts later in this section. In the rest of this section, we'll look at how adjuncts are used to refer to particular peripheral roles and how these peripheral roles can also be realized in other ways. In the process we'll discover some interesting ways in which languages can differ from one another. One kind of difference involves a distinction that is made in one language but not in another. In other words, two different sentences in one language correspond to one ambiguous sentence in the other language. Another kind of difference involves alternate ways of conveying the same information. One language may express in adjuncts what another language expresses within the verb itself.

Talking about move events

Remember that move events have a source, a goal, and path, any of which might be worthy of attention and mention. English typically uses the preposition from for the source and the preposition to for the goal. There are a number of other possibilities for the goal, however. Here is an example with under.

7. The ball rolled under the table.

But this sentence is in fact ambiguous, and it's an ambiguity we can express in terms of semantic roles. We could use the same sentence if the rolling took place under the table, that is, if the area under the table was the location rather than the goal of the rolling. (The location interpretation probably seems odd because it is hard to visualize what caused the rolling in this case.)

Often when a language exhibits ambiguity in a sentence, that sentence corresponds to different sentences in some other languages. In this case, for example, Japanese would convey sentence 7 differently, depending on whether the area under the table is the goal. Here are the two Japanese sentences corresponding to sentence 7, the first with the goal interpretation, the second with the location interpretation.

The path in English is often expressed by one of a small set of words that behave somewhat like prepositions, except that they occur freely without a following noun phrase. These include up, down, out, and off, as well as some words that can also function as prepositions such as in and around. I will refer to these words as directional adverbs. Here are some examples.

10. The boll rolled down the stairs.
11. The ball rolled down.
12. The balloon floated up the chimney.
13. Clark strutted around the room.

In languages such as Spanish, on the other hand, it is more common to express the path as a part of the verb. That is, Spanish verbs for move events often include a specification of the direction of the movement. Here are some examples.

The verbs in these sentences include the notions of falling or rising, so we can say that Spanish tends to lexicalize aspects of the path in verbs for move. Of course English has verbs such as descend and rise too, but these are not nearly as common as alternatives like go up and go down and are not normally learned by young children. The point is that the basic verbs used for motion in Spanish tend to express path, whereas the basic verbs for motion in English do not.

Instead English has a tendency to lexicalize manner in its verbs for move events. Examples in the above sentences include roll and float. Note that neither of these verbs specifies the path; the rolling or floating could be in any direction.

So what does Spanish do with manner in sentences corresponding to 10-13 above? It appears that Spanish speakers are less likely to refer to manner in sentences like this than English speakers are. When they do refer to manner, it tends to appear in an adjunct of a kind we have not seen yet, one that is a form of a separate verb expressing manner. Here is a Spanish translation of sentence 12. The adjunct flotando expresses the manner of the movement.

For more on how languages can be grouped on the basis of what aspects of situations are lexicalized, see the work of Leonard Talmy, the linguist who pioneered work of this type.

Talking about transfer events, indirect objects

Recall that transfer and information_transfer events are like move events in having a source and a goal, that is, the receiver. For the subcategory of events where the source of the transfer is also the agent, for example, in giving and telling, English has two common patterns. One uses the preposition to to mark the recipient. Here are two examples.

17. Clark lent his laptop to Lois.
18. Lois told her story to her officemates.

The other possibility is to refer to the recipient in an NP that has no preposition indicating its role and appears right after the verb. This syntactic role is called the indirect object. It represents a third core syntactic role, along with the subject and direct object, in English and many other languages. Notice that when there is an indirect object in an English sentence, it appears before the direct object. Here are sentences 17 and 18 reformulated using indirect objects.

19. Clark lent Lois his laptop.
20. Lois told her officemates her story.

Since sentences 17 and 19 and sentences 18 and 20 have different forms, however, we would expect their meanings to differ in some way. But the difference is very subtle and not well understood. And the choice of one pattern over the other depends on other factors not directly related to the meaning of the sentence, such as the length of the NP referring to the recipient. That is, sentence 19 would be much less acceptable if Lois were replaced by the woman he met on the train to work last Friday. In this case English speakers would prefer the pattern with the NP referring to recipient after the direct object and preceded by to.

Note that English makes no distinction between the form of the indirect and direct objects; that is, the same set of personal pronouns, the objective forms, is used for both. Spanish does make such a distinction, however, though only for the third person, where accusative case is distinguished from dative case, the form used for the indirect object. Thus there are two Spanish words corresponding to him: lo, the accusative form, and le, the dative form.

In English, the indirect object is also used for beneficiaries with some verbs. Here is an example.

20. Lois knit Billy a sweater.

Note that for beneficiaries the alternate form with a prepositional phrase adjunct instead of the indirect object uses the preposition for rather than to.

21. Lois knit a sweater for Billy.

In Chapter 8 we'll see how Lingala provides an elegant way to create new forms of verbs that take indirect objects for a wider variety of roles than is possible in English.

Talking about experience

Modern languages tend to be fairly consistent in the syntax-semantics mappings for the do_to schema, with the agent represented by the subject and the patient represented by the direct object. What happens with other schemas with two or more core participants?

Let's try to get some insight into this by continuing our fictitious history of how language started. The Grammies first started using transitive sentences to refer to do_to events, events that are easy to observe and whose participants' roles are easy to understand. When they first felt the need to refer to human mental experience, it was not so clear how to proceed.

As we saw in the section on situation schemas, these states or events have two core participants, just as do_to does, but the roles they play are quite different. Rather than creating completely new syntactic roles for the experiencer and the experience theme, the Grammies realized that they could make do with the two core syntactic roles they already had, subject and direct object. Since the verb of a sentence could make it clear that the sentence was about an instance of experience rather than an instance of do_to, all that was needed was conventions for how the two semantic roles were to be referred to by the NPs in the two syntactic roles.

But the sensible thing would be to use the subject and direct object in a way that resembled their use in the familiar do_to sentences. That way each syntactic role would tend to have a consistent semantic interpretation. There seem to be two ways to go, however. In one way, the experiencer is like an agent; it is typically animate and intimately involved with the event or state. That is, when you see, believe, or hate something, you seem to have much more to do with what is going on that the thing that you see, believe, or hate does. On the other hand, the experiencer is like a patient in the sense that is affected by the state or event in a relatively passive way. Seeing or believing are processes that happen to you, not processes that you are in control of.

Since both of these associations make some sense, it is not surprising that the Grammies failed to agree on a single way to map the experiencer roles onto their syntactic roles. Some of them decided to use subjects for experiencers (as well as agents), while others decided to use other syntactic roles for experiencers. Still others treated some kinds of experience one way and other kinds another way. Modern languages also exhibit this sort of disagreement. Let's look at a few examples.

For the most part, English went with the option by which experiencer is like agent and theme is like patient. That is, it is the subject that typically refers to the experiencer and the direct object to the theme. Here are two examples.

22. Lois sees Clark.
23. Lois likes Clark.

Spanish agrees with English for sensory experiences like seeing and hearing: the Spanish subject refers to the experiencer, and the Spanish direct object refers to the theme. But look at a natural translation of sentence 23 into Spanish.

In this sentence Lois is referred to both by the NP Lois and the dative (indirect object) pronoun le. Though Lois is the first NP in the sentence, it is not the subject. In sentences like this, Spanish treats the experiencer like it treats a recipient. The subject of the sentence is Clark, referring to the theme.

Japanese differs even more from English. For sensory experiences, as in sentence 22 above, Japanese tends to put the experiencer in an NP marked with the postposition ni, roughly 'to'. A natural Japanese translation for sentence 22 is the following.

The theme, Clark, is referred to by the subject, and the experiencer, Lois, appears in a phrase marked just as Japanese marks the recipient of a transfer. Japanese sentences for liking and hating states are more complicated, so I won't include them here.

Syntax-semantics mapping revisited

In this section we've seen that for some categories of situations, the syntax-semantics mapping is reasonably predictable. That is, for events with clear agents and patients, there is a strong tendency within and across languages to map the subject onto the agent and the direct object onto the patient. For some of the peripheral semantic roles, it is also possible to make generalizations about their syntactic realization that apply to sentences with many different verbs. For example, the English prepositions that are used for source, goal, location, and time can be used in almost any sentence where it is appropriate to mention these semantic roles. And the instrument can almost always be conveyed in English with the preposition with.

However, even for these semantic roles, things are not always so simple. Some verbs even permit them to appear in one of the core syntactic roles. Here are some English examples.

26. The stone broke the window.
27. The 1990s saw the end of the Cold War.
28. Perry splattered the floor with paint.

In sentence 26, the subject, the stone, refers to the instrument apparently used by some unspecified agent. In sentence 27, the subject, the 1990s, refers to the time of the event that is realized as the direct object, the end of the Cold War. In sentence 28, the direct object, the floor, refers to the goal, whereas the object of the preposition with refers not to the instrument of the splattering but rather the patient.

We have also seen that the syntax-semantics mappings are less predictable for some classes of situations. Within languages, experience states and events may be treated differently for different verbs, for example. In general, as noted in the last section, it's safest to assume that each verb is associated in the lexicon with its own syntax-semantics mapping(s).

Let's look at some more examples of these mappings. First, given the realization of the instrument in a sentence like 26, we see that there is a third mapping for verbs like break, in addition to the two discussed at the end of the last section. Here is a more complete set of mappings for break than appeared there. The optional instrument phrase is included in the last mapping to show how it differs from the realization of the instrument in the second mapping.

Here are the two possible mappings for give, one of the verbs that can take indirect objects referring to recipients.

Finally, here are syntax-semantics mappings for English and Spanish verbs of liking.