Semantic roles and syntactic relations

A verb is the centre of a clause. A verb may refer to some activity and there must be a number of participants who have roles in that activity (e.g. Sinbad carried the old man); or a verb may refer to a state, and there must be a participant to experience the state (e.g. My leg aches).

A set of verbs is grouped together as one semantic type partly because they require the same set of participant roles. All GIVING verbs require a Donor, a Gift and a Recipient, as in John gave a bouquet to Mary, Jane lent the Saab to Bill, or The Women’s Institutes supplied the soldiers with socks. All ATTENTION verbs take a Perceiver and an Impression (that which is seen or heard), as in I heard the crash, I witnessed the accident, I recognized the driver’s face. AFFECT verbs are likely to involve an Agent, a Target, and something that is manipulated by the Agent to come into contact with the Target (which I call the Manip). A Manip can always be stated, although it often does not have to be, e.g. John rubbed the glass (with a soft cloth), Mary sliced the tomato (with her new knife), Tom punched Bill (with his left fist).

We are here working at the semantic level, and it should be stressed that each type has a quite distinct set of roles. There is nothing in common between Gift (that which is transferred from one owner to another) and Impression (an object or activity that is seen or heard), or Perceiver (a person who receives visual or auditory sense impressions) and Agent (a person who wields a Manip to come into contact with a Target), and so on.

There are about thirty semantic types associated with the Verb class. Some verbs, such as those in the GIVING and AFFECT types, have three semantic roles. Some, like ATTENTION, have just two. And some have just one (CORPOREAL verbs like breathe, and MOTION verbs like fall). Altogether, it is necessary to recognize forty or fifty semantic roles.

Turning now to syntax, we find that every language has a limited number of syntactic relations. Subject and Object are probably universal relations, which apply to every language. But just as the criteria for the major word classes Noun and Verb differ from language to language, so do the ways in which syntactic relations are marked. In Latin, for instance, the Subject occurs in nominative case (e.g. domin-us ‘master-NOMINATIVE’) and the object in accusative case (e.g. serv-um ‘slave-ACCUSATIVE’). Words can occur in many different orders in a clause in Latin, so that Dominus servum videt, Servum videt dominus, Videt dominus servum, etc. all mean ‘The master sees the slave’. In English, nouns have no case inflection and grammatical relations are shown primarily by word order, Subject before the verb and Object after it.

The roles of each type, at the semantic level, are mapped onto syntactic relations, at the grammatical level. For ATTENTION verbs, for instance, the Perceiver is grammatical Subject and the Impression is marked as Object.

There are quite often several different ways in which semantic roles may be associated with grammatical relations. With the GIVING type either the Gift may be Object, as in Jane lent the Saab to Bill, or the Recipient may be, as in Jane lent Bill the Saab; for both of these clauses the Donor is Subject. It is also possible to have Recipient as Subject, and then a different verb is used: Bill borrowed the Saab from Jane. Borrow is the semantic converse of lend; both verbs belong to the GIVING type and involve the same three semantic roles.

For AFFECT verbs the Agent is usually the Subject and the Target the Object, with the Manip marked by a preposition such as with—John hit the pig with his stick. But we can have the Manip in Object slot (this often carries an implication that the Manip is less strong than the Target, and likely to be more affected by the impact)—John hit his stick against the lamp post. Or, as a third alternative, the Manip can be placed in Subject relation—John’s stick hit Mary (when he was swinging it as she walked by, unnoticed by him); use of this construction type may be intended to imply that John was not responsible for any injury inflicted.

Verbs fall into two broad subclasses—those that require only one role (intransitive verbs) and those which require two or more roles (transitive verbs). There is considerable difference between intransitive subject and transitive subject. We will need to refer a good deal to these relations, so it will be useful to employ abbreviatory letters for them, and for object:

S—intransitive subject

A—transitive subject

O—transitive object

If a verb has only one role, at the semantic level, then it must be mapped onto S relation, at the syntactic level. Some of the roles in S slot can control the activity in which they are involved (e.g. walk, speak) but other S roles exercise no control (e.g. break, die, grow).

If a verb has two or more roles, one will be mapped onto A and another onto O. It is the role which is most relevant for the success of the activity which is put in A relation; compare Bill tried to borrow the Saab from Jane with Jane tried to lend the Saab to Bill. And it is the non-A role which is regarded as most salient for the activity (often, the role which is most affected by the activity) which is put into O relation—compare Mary cut the cake into slices with Mary cut slices off the cake.

There is a subset of transitive verbs which require a further argument (sometimes called E, for extension to the core). The extended transitive (or ditransitive) verbs in English include give, show and also put. (One cannot say just *I put the teapot, it is necessary to specify where it was put, e.g. on the table or here.)

A number of verbs have dual transitivity; that is, they can be used both transitively and intransitively. These ‘ambitransitive’ (or ‘labile’) verbs fall into two types. Some ambitransitives identify O in transitive with S in intransitive use, as in John (A) broke the glass (O) and The glass (S) broke. Others identify A with S, as in Mary (A) has eaten lunch (O) and Mary (S) has eaten.

Besides the major intransitive and transitive clause types, there is a minor but important type, copula clause. This involves two further grammatical relations:

CS—copula subject

CC—copula complement

In English, the CS has similar properties to S and A; however, in some languages CS is like S and unlike A, and in at least one language it is like A and unlike S. What follows the copula verb (generally be or become) is the CC; for example, the CC is enclosed in brackets in My son is [a doctor], You are [generous], The dog is [in the garden]; It is important to note that the CC is a type of noun phrase in grammatical relation with the (copula) predicate, similar to S, A, O and CS. A Copula Complement has sometimes been described as a ‘nominal predicate’; this is not a useful designation, and can be highly confusing.

The core syntactic relations are Subject, Object and Copula Complement. Other, peripheral relations are in English marked by a preposition— these can refer to a place or time setting (in the morning, at the races) or to some additional participant (for Mary, with a hammer).