Linguistics An Essential Introduction (Version 1.5)

We could argue that they are, in fact, possible but that, for various reasons, they simply don’t make sense. The sentence in (3a), for example, omits important information. The predicate corresponding to the verb hand is HAND(x, y, z) — it has three arguments specifying a person who acts on some entity in such a way that another entity receives it —, and the sentence Noah handed only realizes one of these arguments — it does not mention what Noah handed to whom. Conversely, the sentence in (3b) contains too much information — the predicate STUMBLE(x) only has one argument (stumbling is something a person does by themselves), so the sentence Aylin stumbled Noah has a superfluous argument that does not correspond to anything in a stumbling event.

However, such arguments quickly run into problems. Consider the following sentences:

The predicate corresponding to the verb drink has two arguments: DRINK(x, y). Thus, the sentence in (4a) omits important information just like the one in (3a) — it does not mention what Aylin drank. Nevertheless, it is a possible, perfectly normal sentence of English. When we hear (4a), we know that Aylin must have drunk something, but we accept the omission and assume that the omitted information is not important. In the same way, when we hear (3a), we know that Noah must have handed something to someone, so why do we not accept this omission, even in a situation where the omitted information is not important? The sentence in (4b) contains too much information — walking is something that a person does by themselves, it does not involve a second person — WALK(x). Yet, we accept (4b) and interpret it to mean that Noah accompanied Aylin home and that both of them walked. So why do we not accept (3b) and interpret it to mean that Aylin and Noah stumbled together, or that Aylin caused Noah to stumble?

Verbs simply differ with respect to the number of constituents that they can or must be used with. The meaning of the verb — the predicate with its arguments — does not allow us to predict this number.

Verbs also differ with respect to the type of constituent that they can or must be used with. For example, spill must be used with a noun phrase referring to a substance that is spilled (for example, the water), and it can be used with an additional prepositional phrase containing a noun phrase that refers to the location that this substance ends up in (for example, on the floor). This is how it is used in (1b). The location cannot be expressed using a noun phrase instead, as in (3c).

The requirements of a verb concerning the number and type of constituents that it can or must be used with is called its syntactic valence (or sometimes its subcategorization). There are some types of valence that are so frequent that they have their own labels — verbs that do not require or allow any constituents (other than the subject), like stumble, are called intransitive, verbs that require one noun phrase (in addition to the subject), like spill, are called transitive, and verbs that require two noun phrases are called ditransitive. However, there are additional kinds of verbs — for example, verbs that require or allow a specific prepositional phrase, like look for, or verbs that require a noun phrase and a prepositional phrase, like put (Zoe put her phone in her pocket).

It is therefore more useful to represent the valence of a verb by listing the number and type of phrase that it requires or allows. Before we try this, there are two issues to be discussed.

First, we have to decide whether the subject should be included in such a list. Some linguists say this is unnecessary, at least in English, because all English sentences have a subject, so that nothing would be gained by including it. However, while all sentences require subjects, not all verbs do. For example, the predicates corresponding to weather verbs (like rain, snow, thunder, etc.) do not have any arguments: RAIN(), SNOW(), THUNDER(). Or, thinking about it in a different way, they have arguments that are expressed by the verb itself: FALL(RAIN), FALL(SNOW), SOUND(THUNDER). Either way, they do not provide anything that could stand in subject position, so users of English have to insert a meaningless it — a so-called dummy subject in that position:

To distinguish such cases from cases where the verb provides an argument for the subject position, we have to include the subject in the valence of verbs (we will discuss another reason later).

Second, we have to distinguish between phrases that must occur with a particular verb and phrases that can occur with a particular verb — as we saw in the case of spill, it has to have a subject and one additional noun phrase referring to the material that is spilled, and it can have a prepositional phrase referring to the final location of that material. Phrases that have to be used with a particular verb are called its obligatory complements. Phrases that can be used with a particular verb are called its optional complements. These differ from phrases that can be used with any verb — these are called adjuncts. The following sentences illustrates the difference between optional complements and adjuncts:

On Monday and in Paris are adjuncts, as we could add them to any verb. You can see this by adding them to every single example sentence in this book — you will always get a possible (grammatical) sentence! Here are some examples:

In contrast, for her sunglasses and over his notebook are optional complements, as we can not add them to any verb. Again, try to add them to some example sentences from this book — this will not be possible most of the time. Again, here are some examples:

Optional complements must be included in valence descriptions, as they are specific to particular verbs, while adjuncts are, obviously, not included. Let us represent the valence of verbs as follows:

The underscore indicates the position of the verb in an active sentence, so that we can distinguish the subject, which is not part of the verb phrase, from the other complements, which are part of the verb phrase. Parentheses indicate optional complements (for example for drink). Where a PP has to contain a specific preposition, we represent it in the form of a bracketing structure, as introduced in Section 8.3 (for example, for look).

Specifying the number and syntactic type of a verb’s complements allows us to explain why the sentences in (3) above are ungrammatical: while their verb phrases follow the phrase-structure rule for VPs in English, they do so in a way that ignore the valence of the specific verbs in these verb phrases.

Now look at the following sentences:

These sentences follow the phrase structure rule for English VPs and they have the right number and type of complements, but there is still something wrong with them. They are not exactly ungrammatical, but they are not possible sentences of English either (this is indicated by the question marks preceding them).

The problem is that some of the complements have the wrong meaning. The verb put does not just require a PP, it requires a PP that describes a location; however, the PP in (9a) describes a point in time. The verb hand does not just require three NPs, it requires one of these NPs to describe someone who carries out the action, one of them a person who receives something, and one of them an entity that changes location. In example (9b) the second NP refers to a location instead of a recipient. And the verb watch requires a living being capable of visual perception and a visible entity. Example (9b) has both, but they are in the wrong place.

As users of a language, we not only know the number and syntactic type of complements that a verb requires or allows, but we also know the general semantic type that these complements must have. These general semantic types are called semantic roles and the specific combination of semantic roles that a verb requires or allows is called its semantic valence.

It is difficult to determine exactly how many semantic roles there are in a given language and exactly how they should be characterized, since this depends largely on how specific or general we want to make their description. At the most specific level, every verb has a set of very specific semantic roles that may be unique to it or that it may share with only a handful of other verbs. For example, the verb sleep has one semantic role — ‘sleeper’ — that it shares with no other verb, and the verb buy has three semantic roles — ‘buyer’, ‘seller’ and ‘goods’ —, that it shares with a few other verbs (mainly, the verb sell). Such specific semantic roles would do the job of ruling out sentences like those in (9), but they do not allow us to make general statements, for example, about classes of verbs with a similar semantic valence (which we will need to do in a moment), or in order to describe the interaction of semantic roles with grammatical relations like subject and object (which we need to do in the next section).

You will encounter slightly different lists of more general semantic roles, but the following are fairly wide-spread and have proven themselves to be useful:

Using these semantic roles, we can now describe the valence of our verbs more precisely:

Note that such semantic valence patterns are related to predicates, but they are not identical. Semantic valence is tied to specific uses of a verb with particular complements and adjuncts — it specifies the semantic roles of these constituents. In contrast, predicates are purely semantic — they specify the arguments that are present in the event described by the verb.

You can now see a second reason for including the subject in the valence description: subjects, too, have different semantic roles depending on the verb, and by including them, we can specify this role. Such lists of complement types with semantic roles are often called valence patterns, when we want to talk about them independently of a specific verb. For example, we might want to say that see and like have the same valence pattern, so they form a class.

As the following examples show, a verb can have more than one valence pattern:

There are four valence patterns here:

This combination of valence patterns is unique to search — each of the valence patterns also occurs with other verbs, but no other verb has all four patterns. Some examples are shown in Table 8.5.2

Generally, we therefore have to figure out the valence patterns for every verb individually. Linguists, who like making general observations about language, are often not very interested in this kind of work, leaving it to lexicographers.

But linguists are very interested in cases like the following:

Each of the verbs in (11) has two valence patterns, the one in (12a) and the one in (12b):

What makes this case different from the one in (10) is that we can make several generalizations: first, there is more than one verb that has these two valence patterns; second all verbs in (11) are verbs describing a change of state; and third, there is a systematic meaning relation between the two valence patterns — the one in (12b) always means ‘Patient changes’, the one in (12a) always means ‘Agent causes Patient to change’. The basic, intransitive use of the verb dry corresponds to its predicate, which has a single argument: DRY(x). The transitive use contains a second predicate, CAUSE(y, z), that combines with the predicate of dry such that the latter takes the place of the second argument: CAUSE(y, DRY(x)).

Such cases are known as valence alternations (or argument-structure alternations, where argument structure is just a different term for valence). They are sometimes represented as lexical rules, which are similar to word-formation rules for zero derivation, except that they do not change the word class of a word, but the valence. The alternation in (11) could be represented as follows (a very simplified version of the representations you will encounter in the research literature):

CC-BY-NC-SA 4.0. Written by Anatol Stefanowitsch