The subject (glossing abbreviations: SUB or SU) is, according to a tradition that can be traced back to Aristotle (and that is associated with phrase structure grammars), one of the two main constituents of a clause, the other constituent being the predicate, whereby the predicate says something about the subject.23 According to a tradition associated with predicate logic and dependency grammars, the subject is the most prominent overt argument of the predicate. By this position all languages with arguments have subjects, though there is no way to define this consistently for all languages.4 Even in languages such as English, there is not always a perfect match between the semantic predicand and the subject, as a predicate may be predicated on an argument in another clause (see raising).
From a functional perspective, a subject is a phrase that conflates nominative case with the topic. Many languages (such as those with ergative or Austronesian alignment) do not do this, and by this definition would not have subjects.
All of these positions see the subject determining person and number agreement on the finite verb, as exemplified by the difference in verb forms between he eats and they eat. The stereotypical subject immediately precedes the finite verb in declarative sentences and represents an agent or a theme. The subject is often a multi-word constituent and should be distinguished from parts of speech, which, roughly, classify words within constituents.
In the example sentences below, the subjects are indicated in boldface.
The subject is a constituent that can be realized in numerous forms. The following table lists the many forms that they may take in English:
There are several criteria for identifying subjects:5
Of these three criteria, the first one (agreement) is the most reliable. The subject agrees with the finite verb in person and number (and sometimes in gender as well). The second and third criterion are merely strong tendencies that can be flouted in certain constructions, e.g.
In the first sentence, all three criteria combine to identify Tom as the subject. In the second sentence, which involves the subject-auxiliary inversion of a yes/no-question, the subject immediately follows the finite verb (instead of immediately preceding it), which means the second criterion is flouted. And in the third sentence expressed in the passive voice, the 1st and the 2nd criterion combine to identify chemistry as the subject, whereas the third criterion suggests that by Tom could be the subject of the verb if it were changed to its active form (i.e. Tom is studying Chemistry) because Tom is an agent.
The fourth criterion is better applicable to other languages, the exception being the subject and object forms of pronouns, I/me, he/him, she/her, they/them.
The fifth criterion is helpful in languages that typically drop pronominal subjects, such as Spanish, Portuguese, Italian, Latin, Greek, Japanese, and Mandarin. Though most of these languages are rich in verb forms for determining the person and number of the subject, Japanese and Mandarin have no such forms at all. In other languages, like English and French, most clauses should have a subject, which should be either a noun (phrase), a pronoun, or a clause. This is also true when the clause has no element to be represented by it. This is why verbs like rain must have a subject such as it, even if nothing is actually being represented by it. In this case, it is an expletive and a dummy pronoun. In imperative clauses, most languages elide the subject, even in languages which typically requires a subject to be present, e.g.
One criterion for identifying a subject in various languages is the possibility of its omission in coordinated sentences such as the following:7 The man hit the woman and [the man] came here.
In a passive construction, the patient becomes the subject by this criterion: The woman was hit by the man and [the woman] came here.
In ergative languages such as the nearly extinct Australian language Dyirbal, in a transitive sentence it is the patient rather than the agent that can be omitted in such sentences: Balan dyugumbil baŋgul yaraŋgu balgan, baninyu 'The man (bayi yara) hit the woman (balan dyugumbil) and [she] came here'
This suggests that in ergative languages of this kind the patient is actually the subject in a transitive sentence.
There are certain constructions that challenge the criteria just introduced for identifying subjects in English. The following subsections briefly illustrate two such cases: 1) existential there-constructions, and 2) inverse copular constructions.
Existential there-constructions allow for varying interpretations about what should count as the subject, e.g.
In sentence 1, the first criterion (agreement) and the second criterion (position occupied) suggest that there is the subject, whereas the third criterion (semantic role) suggests rather that problems is the subject. In sentence 2, in contrast, agreement and semantic role suggest that problems is the subject, whereas position occupied suggests that there is the subject. In such cases then, one can take the first criterion as the most telling; the subject should agree with the finite verb.8
Another difficult case for identifying the subject is the so-called inverse copular constructions, e.g.9
The criteria combine to identify the boys as the subject in sentence 1. But if that is the case, then one might argue that the boys is also the subject in the similar sentence 2, even though two of the criteria (agreement and position occupied) suggest that a chaotic force around here is the subject. When confronted with such data, one has to make a decision that is less than fully arbitrary. If one assumes again that criterion one (agreement) is the most reliable, one can usually identify a subject.
The existence of subject-less clauses can be construed as particularly problematic for theories of sentence structure that build on the binary subject-predicate division. A simple sentence is defined as the combination of a subject and a predicate, but if no subject is present, how can one have a sentence? Subject-less clauses are absent from English for the most part, but they are not unusual in related languages. In German, for instance, impersonal passive clauses can lack a recognizable subject, e.g.
Gestern
yesterday
wurde
was
nur
only
geschlafen.
slept
Gestern wurde nur geschlafen.
yesterday was only slept
'Everybody slept yesterday.'
The word gestern 'yesterday' is generally construed as an adverb, which means it cannot be taken as the subject in this sentence. Certain verbs in German also require a dative or accusative object instead of a nominative subject, e.g.
Mir
me-DAT
graut
is uneasy
davor.
about it
Mir graut davor.
me-DAT {is uneasy} {about it}
'I am uneasy about it.'
Since subjects are typically marked by the nominative case in German (the fourth criterion above), one can argue that this sentence lacks a subject, for the relevant verb argument appears in the dative case, not in the nominative.
Impersonal sentences in Scottish Gaelic can occasionally have a very similar form to the first German example where an actor is omitted. In the following sentence, the word 'chaidh' ("went") is an auxiliary carrying tense and is used in an impersonal or passive constructions. The word 'falbh' ("leaving") is a verbal noun.
Chaidh
went
falbh
leaving
Chaidh falbh
went leaving
'They/Everybody left.'
The subject receives a privileged status in theories of sentence structure. In those approaches that acknowledge the binary division of the clause into a subject and a predicate (as is the case in most phrase structure grammars), the subject is usually a daughter of the root node, whereby its sister is the predicate. The object, in contrast, appears lower in the structure as a dependent of the/a verb, e.g.10
Subjects are indicated using blue, and objects using orange. The special status of the subject is visible insofar as the subject is higher in the tree each time than the object. In theories of syntax that reject the initial division (as is the case in most dependency grammars), the subject is nevertheless also granted a privileged status insofar as it is an immediate dependent of the finite verb. The following trees are those of a dependency grammar:11
The subject is a dependent of the root node, the finite verb, in both trees. The object, in contrast, appears lower in the second tree, where it is a dependent of the non-finite verb. The subject remains a dependent finite verb when subject-auxiliary inversion occurs:
The prominence of the subject is consistently reflected in its position in the tree as an immediate dependent of the root word, the finite verb.
Comrie (1989), pp. 105–106. ↩
See Conner (1968:43ff.) for a discussion of the traditional subject concept. ↩
The division of the clause into a subject and a predicate is a view of sentence structure that is adopted by most English grammars, e.g. Conner (1968:43), Freeborn (1995:121), and Biber et al. (1999:122). ↩
See Tesnière (1969:103-105) for the alternative concept of sentence structure that puts the subject and the object on more equal footing since they can both be dependents of a (finite) verb. ↩
See Biber et al. (1999:123) for a similar list of criteria for identifying subjects. ↩
For basic discussions of subject-verb agreement, see for instance Barry (1998:68f.), Fergusson and Manser (1998:36f.), and Jurafsky and Martin (2000:366f.). ↩
Discussion in Comrie (1989), pp.111ff. ↩
For a discussion of the subject status of existential there, see Biber et al. (1999:944). ↩
For in depth studies of inverse copular constructions, see Moro (1997) and Mikkelsen (2005). ↩
Phrase structure trees similar to the ones produced here can be found in numerous introductory texts to grammar and syntax, e.g. Payne (2011). ↩
Dependency trees similar to the ones produced here can be found in *Ágel et al. (2003/6). ↩