|Home||Feature Inventory||Related Issues||Search||Contact|
Transitivity is a complex, clause-level phenomenon, fundamental to the structure of major clause types. It is best defined as a type of grammatical relationship encoding the distinctness of participants in a situation described by the clause (Næss 2003). It applies at the same time to a certain syntactic configuration in a given language (thus, we can talk of 'syntactic transitivity'), and to a cluster of semantic properties ('semantic transitivity') typically found to correlate with this syntactic configuration. Major clause types are understood here to consist of a predicate and a variable number of predicate arguments; this excludes minor clause types involving two noun phrases, either with or without a copula.
Syntactic transitivity refers to the number and type of core arguments which appear in the clause and which are determined by the predicate's head. Most frequently, the predicate is headed by a verb, though in some languages an intransitive predicate may be headed by a noun or a pronoun. Two clause types may be distinguished (cf. Dixon 1994:113-27, Dixon & Aikhenvald 2000:2): an intransitive clause - with an intransitive predicate and a single core argument which is in an intransitive subject function (S); and a transitive clause - with a transitive predicate and two core arguments which are in a transitive subject function (A) and a transitive object function (O). In some languages a further argument has special, non-peripheral status. Dixon & Aikhenvald (2000:3) label it as E, standing for 'extension to core'. This argument can appear in an extended transitive (or, ditransitive) clause type, or in an extended intransitive clause type. Therefore, considering the number and type of core arguments, we find the following clause types (Dixon & Aikhenvald 2000:3):
As for the participant roles expressed by the arguments, languages vary as to which roles may be represented by the particular grammatical relations. Typically, the A function expresses an agent-like participant, the O function - a patient-like participant which could be patient or theme, or a recipient/beneficiary in extended transitive clauses. The E function can express a recipient/beneficiary or a theme in extended intransitive clauses, and typically a theme in extended transitive clauses. Examples of extended intransitives are intransitive clauses with dative beneficiaries in Polish; or constructions typically used for seeing, hearing, liking and wanting (with E being the thing that is seen or an object that is liked or wanted) in languages such as Tongan (Polynesian), Trumaí (a language isolate spoken in Brazil), Tibetan, Newari (Sino-Tibetan, spoken in Nepal), and Motuna (East Papuan) (Dixon 1994:122-124, Dixon & Aikhenvald 2000:3).
That languages make a fundamental distinction between 'intransitive' and 'transitive' constructions (Dixon 1979) is a basic assumption of linguistic theory, and it is coupled with an assumption that this distinction corresponds to certain specific differences between the kinds of situation that these two types of construction are used to describe (Næss 2003:1). The basic difference is taken to be that between 'one-participant' and 'two-participant' situations. However, the correspondence between this semantic distinction and the formal distinction is far from straightforward. Languages may have several constructions with different formal characteristics which all describe situations of one type (e.g. the same 'two-participant' situation in English can be expressed as either She hit the man or The man was hit) and, conversely, situations characterised by different semantic transitivity may be encoded in the same way with regard to syntactic transitivity (as in the English He slept vs He ate).
In some languages, for example Latin or Dyirbal (Australian), it is possible to classify verbs into 'intransitive' and 'transitive' depending on which type of clause they may occur in. In many languages, however, there is no clear-cut division into verb classes and the same verb can appear in either an intransitive or a transitive clause. An example of such a verb in English is run, traditionally classified as intransitive (I ran away), but equally capable of appearing in a transitive clause (I ran a race) (Kemmer 2003:277); see also §2 below.
Dixon & Aikhenvald (2000:4-5) list the following typical transitivity classes of verbs, found in English and many other languages:
Furthermore, there can be additional divisions. For example, in Tariana (an Arawakan language spoken in Brazil), intransitive verbs divide into two subtypes (Aikhenvald 2000; Dixon & Aikhenvald 2000:5):
Where every intransitive verb is either of type SA or of type SO, verb classes make a so-called split-S system. Otherwise, where some verbs can take either the SA or the SO marking depending on whether or not the referent of the S argument is in control of the activity, languages can be said to have a fluid-S system (Dixon 1994:70-83, Dixon & Aikhenvald 2000:5).
A different example of a large system of verb classes with respect to transitivity is found in Motuna (East Papuan) (Onishi 2000). It includes the following verb classes: (a1), (a2), (c1) where the patient is irrelevant or unimportant in the intransitive, (c2) describing a spontaneous process or event in the intransitive, and a class that could be labelled (c3) including a further variety of ambitransitive where the intransitive is reflexive, i.e. S=A=O.
Whether a situation is encoded as transitive or intransitive depends, first of all, on the semantic characteristics of the participants - though, as was mentioned above in §1, the correspondence between semantic transitivity and syntactic transitivity is not straightforward.
A transitive clause typically encodes events which involve two participants. However, transitive semantics depends on the distinctness of the participants, which in turn depends on the individuation and distinguishability of the participants. Thus, formal transitivity is more readily available the more distinct from each other the participants are perceived as being, both in terms of their physical properties and in terms of the roles they play in the event. This principle of maximal semantic distinction of arguments (Næss 2003) provides the basis for deriving a cluster of other semantic factors known to be of relevance to formal transitivity, such as: (a) the volitionality of the agent participant (b) performing a concrete, dynamic action (c) which has a perceptible effect on a specific patient (see e.g. Lakoff 1977, Hopper & Thompson 1980, Kittilä 2002, Lazard 2003).
Thus, a transitive clause will typically involve an agent characterised as volitional, instigating (causing) and nonaffected, and a patient characterised as nonvolitional, noninstigating and affected. Any deviation from this configuration may lead to a less transitive construction. If the participants are not clearly distinct, the agent is affected, or the patient is effected/nonreferential, formally intransitive constructions are often used. Lowered semantic transitivity may be expressible in a basic (non-derived) intransitive clause if the language has an appropriate lexical item that is capable of expressing it. For example, the class of verbs listed above as (c1), which includes S=A ambitransitives, have the capability of making basic intransitive clauses despite expressing transitive semantics (albeit weakened). A formally intransitive clause is then used to highlight that the event described deviates somehow from what is considered to be the transitive prototype (for detailed discussion of this see Næss 2003:Chapter3). Other formal means of expressing lowered semantic transitivity are lexically or syntactically derived constructions such as anticausatives or middles (cf. class (c2) above), reflexives, reciprocals, clauses with oblique objects, or clauses with incorporated objects (see also the entry on 'Diathesis & voice' in this Inventory).
One of the most salient formal aspects of transitive clauses is the formal marking of the arguments, in particular their case-marking (in those languages that have case). Dixon & Aikhenvald (2000:3) summarise the issue of marking as follows. In some languages there is distinct marking for A, O, E and peripheral arguments. In others E and peripheral arguments are treated in the same way. In a further group no distinction is made between O and E. In a few languages all of O, E and peripheral arguments are marked in the same way. Thus, using 'w', 'x', 'y' and 'z' for marking schemes (where 'z' may indicate a variety of markings for various types of peripheral arguments):
Specifically, in Jarawara any nominal phrase that is not in S, A, or O function is marked by the all-purpose preposition jaa (Dixon 2000). In Kinyarwanda O and E follow the verb, and can occur in either order (Kimenyi 1980). In Creek there are two case markers, -t on a subject and -n on a non-subject nominal phrase (Martin 2000).
In many languages, case-marking is an important aspect of the structural pattern of transitive clauses. The two main transitive case-marking alignment systems are the accusative alignment (also referred to as nominative-accusative) and the ergative alignment (also referred to as ergative-absolutive, or ergative-nominative; see the entry on 'Case' for references to publications which discuss this distinction; see also Kulikov, Malchukov & de Swart 2006 for a collection of recent papers on case, valency and transitivity in a variety of languages). The central function of the accusative case is to mark the direct object of a transitive clause, while that of the ergative case is to mark the subject argument of a transitive clause - but what constitues a 'transitive clause' follows from semantic transitivity.
To take a simple example (from Næss 2003:7-8), "the fact that the verb 'like' in a great many languages takes a different case-frame from e.g. 'kill' ('like' typically taking either a dative-nominative or a 'double-nominative' case-frame, cf. [Næss 2003:]Chapter 10; see also Shibatani 1982) is explained by appeal to the lesser semantic transitivity of 'like', which does not have a controlling agent, does not denote a dynamic action, and does not have a perceptibly affected patient. 'Kill', by contrast, fulfills all these criteria and so typically occurs with a nominative-accusative or ergative-absolutive case-frame, in languages where such case systems exist. Languages will differ in the extent to which they extend the 'transitive' case-frame to clauses which do not fulfill all the criteria for semantic transitivity, but by definition this case-frame will apply at least to those clauses which do fulfill these criteria."
A 'discriminatory' view of case-marking regards core case-marking as distinguishing between the arguments of a bivalent clause, especially in cases where it is not clear from the context or the inherent properties of the referents of the arguments which syntactic function (and thereby participant role) should be assigned to which argument. On this view, definite and/or animate objects are more likely to receive overt case-marking because they are more likely to be mistaken for subjects. An 'indexing' view, on the other hand, regards core case-markers as reflecting certain semantic properties of the referents of the case-marked arguments, e.g. affectedness (accusative case) or control (ergative case). Næss (2003:7; 2006) argues that the discriminatory and the indexing functions of case-marking can be viewed as two aspects of a single function, that of distinguishing between (semantically defined) participants in a transitive situation. Core case-markers canonically refer to participants in the agent-patient opposition, that is, they discriminate between semantically defined entities. Specifically, the accusative prototypically marks an affected patient in a clause with a volitional agent, while the ergative prototypically marks a volitional agent in a clause with an affected patient. So, the canonical function of case involves both the discriminatory and indexing (semantic) aspects. However, it is commonly extended along the discriminatory dimension (e.g. when a core case used only when there is a need for overt discrimination between two participants, regardless of their semantic properties - hence Finnish imperatives without accusative case-marking), or along the semantic dimension (e.g. when accusative case is found on an affected O even when A is not a semantic agent, or ergative case on a controlling A even when O is not a semantic patient; the extreme case are intransitive split-S systems, as in (a1) and (a2) above, which mark all semantic agents and/or patients regardless of whether there is another participant in the clause at all).
Since all languages seem to have verbs that are inherently (i.e. semantically) transitive and verbs that are inherently intransitive, according to Bybee (1985:30), it should be "rare to find a case where valence could be considered obligatory in the sense that every finite clause contains a morphological indicator of the number and role of the arguments." In her sample of 50 languages there were only three languages that could be regarded as expressing valency obligatorily (Bybee 1985:31-32). The first two languages, Kutenai (a nearly extinct language isolate spoken in Southeastern British Columbia) and Maasai (a Nilo-Saharan language spoken in Kenya) both have object agreement markers, and both have instrumental and benefactive markers on verbs. Bybee presumes that these markers occur obligatorily in clauses that contain an expression of an instrument or a benefactor, and the absence of the markers signals the absence of these arguments. However, Bybee hypothesises that there might still be verbs that do not need these markers, but inherently take instruments or benefactors, in which case the inflectional status of valency would be questionable. Furthermore, in Maasai, if object agreement is not present on a semantically transitive verb, the verb is still interpreted as transitive with a third person object. To achieve an intransitive reading, a suffix must be added to the verb (Tucker & Mpaayei 1955). Bybee's third, and best, example of obligatory expression of valency comes from Nivkh (also called Gilyak, a language isolate spoken on Sakhalin Island and along the Amur River in Russia). There, a stem initial consonant alternation coincides with the transitivity of the verb. All transitive verbs have consonants from one series, while intransitive verbs have consonants from another series (Jakobson 1957).
A number of languages have verbal affixes to encode the transitivity of the verbs. For example, in Fijian (Dixon 1988:45, 200-14) and most other Oceanic languages (Lynch 1998:139ff), most verbs are ambitransitive. They take a suffix when used in a transitive clause but lack the suffix when used intransitively. The same situation has been observed in Halkomelem Salish (a Central Coast Salishan language spoken in British Columbia; Wiltschko 2000). Furthermore, in some languages derivational affixes have been argued to serve as markers of transitivity (Dixon & Aikhenvald 2000:5). For example, Amharic (Semitic, spoken in Ethopia) has a number of derivational prefixes, including the intransitiviser tə- and the causativiser as-. Some verbs may only occur with one of these prefixes (with one or the other prefix, but the verb cannot occur prefixless). In these circumstances it can be argued that these prefixes serve as markers of transitivity, similar to Fijian (Amberber 2000). Comrie (2000) describes similar valency-encoding suffixes in Tsez (East Caucasian). In some languages a causative affix has become lexicalised, so that it now has a semi-idiomatic meaning, and may function as a marker of transitivity. This is argued to be the case in Athapaskan languages (Rice 2000).
An example of a language where the causative marker does not function as a transitiviser is Pileni (Polynesian) (Næss 2002:68-69). The transitive marker in Pileni is strictly obligatory with all transitive verbs. Causative transitive verbs are formed from intransitives by means of the causative prefix hoka-/hua-/ha-. Such forms also require a transitive suffix. Causativised verbs with just the causative prefix and no transitive suffix do occur, but they have stative interpretation instead of causative transitive (e.g. tupu 'grow' ∼ huatupu 'be piled up; lie in a pile'). Thus, a causative prefix does not in itself make a verb transitive. Just as any other verbs, causativised verbs are only made transitive through the addition of a transitive suffix. Since Pileni has no case-marking and allows omission of any core arguments which are retrievable from the context, the transitive suffix is the only obligatory marker of transitivity in a clause. The suffix is used when the context clearly implies a definite object, even if this object is not represented by an overt nominal phrase in the clause.
Finally, most languages have some morphological or syntactic devices to indicate increased or reduced transitivity: stativity, anticausativity, reflexivity, reciprocity, passivity, causativity, applicativity, etc. These may include derivational affixes, clitics, or syntactic devices achieving the change of valency (see the entry on 'Diathesis & voice').
In her discussion of lexical, derivational, and inflectional morphology, Bybee (1985:86) observes that derivational processes often create meaning combinations (e.g. unpretty) that are already represented lexically (e.g. ugly), and in such cases it is usual for the derived form to be rejected (Clark & Clark 1979). Valency-changing processes are good examples of this phenomenon, because they are very frequently represented morphologically while, at the same time, differences in valency are also frequently encoded lexically. So, even though transitivising morphology occurs in a large percentage of languages, there are probably no languages in which all lexical verbs are intransitive and all transitives are formed by a morphological process. This is because in the real world certain events are inherently semantically transitive, and not necessarily divisible into an intransitive event plus a transitiviser. Thus, all languages seem to have verbs that are inherently transitive and verbs that are inherently intransitive.
Like many other features, transitivity is determined by the semantics. However, despite the fact that every major clause type encodes a certain number of core participants in its syntactic structure, syntax does not need to know about the transitivity value selected for the clause once its choice has been made at the level of lexical semantics. The lexical items (the verb and the nominal phrases) which are selected enter into the appropriate configuration of the basic clause structure without having to carry a 'transitivity' feature with them. The case-marking encodes the relationship between the predicate and its arguments after the syntactic transitivity frame for the expression of the situation has already been determined by the semantics, taking account of the properties of the selected lexical items. There is no need to posit a morphosyntactic feature of transitivity ([+/- transitive]) because it is not required by the syntax either for the purpose of agreement or government in any of the languages that we know of.
In those languages where transitivity is obligatorily marked, we consider it to be a morphosemantic feature, rather like the feature of tense. In most of Oceanic, where it is claimed that all or most verbs are ambitransitive rather than just intransitive (because many of them are obviously semantically transitive), all verbs are unspecified for semantic transitivity, and the value of transitivity needs to be selected for the verb before it enters into a syntactic configuration. This is similar to the choice of the value of inherent gender for a multi-gendered noun such as the English baby. In those cases where verbs have a fixed value for transitivity, they are like nouns with a fixed value for gender. Inherent gender (as found on nouns) is morphosemantic, but contextual gender (as found on targets of agreement) is morphosyntactic. We have not found any contextual instances of transitivity that would justify regarding this feature as morphosyntactic (see also the discussion of two 'Problem cases' in §6 below).
As a morphosemantic feature, transitivity appears to have up to two values (as in Oceanic with all ambitransitive verbs): 'transitive' and 'intransitive'. However, in most languages where transitivity is marked, it has only one value, 'transitive', optionally added to the clause. Since such transitive clauses contrast with clauses in which there is no trace of information regarding transitivity, the non-transitive clauses cannot be argued to bear a value of 'transitivity'. Instead, in such situations transitivity is better analysed as an additional piece of information which can be selected optionally for the clause.
At this point, we are not aware of any larger systems, though at least one more value is logically possible (ditransitive), or perhaps even two (ditransitive, and peripheral argument).
When considering values of transitivity, it is useful to bear in mind that sometimes a distinction is drawn between 'transitivity' and 'valency'. Dixon & Aikhenvald (2000:3) suggest the following definitions. Transitivity expresses both the number and type of core arguments in the clause. There are two main transitivity types - intransitive (with core argument S) and transitive (with A and O) - and plain and extended subtypes of each (depending on whether or not E is also in the core). Valency relates only to the number of core arguments, or, the number of overt argument noun phrases a particular verb form is required to combine with in a grammatical clause. Thus, an intransitive predicate is monovalent, an extended transitive is trivalent, and there are two different kinds of bivalent predicates: transitive and extended intransitive.
Another common way to use the term 'transitive' is to see it as part of a family of related terms:
As was mentioned in §2 above, Pileni (Polynesian) has an obligatory suffix which marks transitivity. In addition to it, Pileni shows another interesting morphosyntactic reflex of transitivity by requiring 3SG agreement -i on tense-aspect markers - but only in transitive clauses. Such agreement does not occur with 3SG subjects of intransitive verbs, or on transitive verbs with subjects other than 3SG (Næss 2002:70).
In Biblical Hebrew (as found in the Hebrew Bible referred to as "the MT", the Masoretic Text), there is a marker (particle) 'et, traditionally called the marker of the definite direct object. It has also been referred to as a transitivity marker, but it is rather a marker of the "efficiency of transferring the effect of the verbal action to the nominal unit" (= a "non-agent marker"?, an "affectedness marker"?). It can also introduce adjuncts and adverbials besides direct and indirect objects, so it could be viewed as a "verbal extension marker", or an applicative marker. However, apparently there are also three examples in the MT (out of more than seven thousand) where 'et is not a verbal extension marker (Anstey 1999).
A question has been posed whether the epenthetic -mi in Ndyuka (the main creole of eastern Suriname) could be analysed as a transitive marker. Huttar (1996) observes that "[i]n the creole languages of Suriname, epenthetic -m(i) is obligatorily inserted between certain verbs and certain objects. Phonological and syntactic features of both verb and object enter into the definition of the environments in which such epenthesis occurs, with different sets of features required for the different languages." This morphophonological phenomenon is both synchronically unusual and diachronically puzzling. "It is unusual in that the conditions governing the insertion of epenthetic material are both syntactic and phonological, and both precede and follow the point of insertion. It is puzzling in that, although the languages involved are creole languages for which the historical sources of many of their features are clear, there is no clear source for this insertion - either of the form of what is inserted or of the conditions governing the insertion."
In Mandarin, 'take' (lexicalised from V2 position) is analysed as a transitive construction marker and as a grammaticalised syntactic object marker (Lord 1993). Arguing against an analysis like the one for Mandarin, Ozanne-Rivierre (2004) claims that in New Caledonian languages 'take' has evolved into an applicative transitivising morpheme, and instead of using a reflex of the Proto-Oceanic applicative transitivising suffix, New Caledonian languages use a semantically transparent transitive verb 'take, carry' for associative case-marking.
No instances of agreement in transitivity have yet been found. Transitivity would be regarded a morphosyntactic feature in a language if it participated in agreement. It has been suggested that this may be the case in two languages: Shipibo-Conibo (Panoan) and Pileni (Polynesian).
For Shipibo-Conibo (a West Central Panoan language spoken in the area of the Ucayali River in the Peruvian Amazon), Valenzuela (1999) gives examples of place and manner adverbials which vary according to the transitivity of the verb. Specifically, place and manner adverbs as well as nominals in these functions exhibit different forms depending on the transitivity status of the verb or clause. In example (1) below (from Valenzuela 1999:359), the locative adverbials carry the additional marker of transitivity, -xon, which is obligatory when the verb is transitive. In example (2) (from Valenzuela 1999:358), the locative adverbial does not carry the marker, because the verb is intransitive; the use of the marker on either the verb or the adverbial would be ungrammatical:
A similar phenomenon has been observed by Næss (2002:70-72) in Pileni, a Polynesian outlier spoken on several small coral islands in Temotu, easternmost province of the Solomon Islands. Pileni is typical of Oceanic langages in having a transitive marker which is affixed to all transitive verbs - that is, bivalent verbs with a specific referential object. The marker is one of the forms of the Polynesian -(C)ia (-(C)onsonant-ia) suffix. With only very few exceptions, some form of the suffix (-na, -a, -ina, or -(a)kina) is found on all transitive verbs in Pileni. In clauses with manner modification by means of a stative verb following the main verb (which is a common phenomenon in Pileni), the modifying verb 'agrees' with the main verb in transitivity, taking the productive transitive suffix -ina if the main verb is transitive, while showing no suffix if the main verb is intransitive. Sentence (4) below shows an example (from Næss 2002:71) of the transitive marker (TR) appearing on the modifying manner verb vakao 'encircle' which describes the manner of action of the main verb, meaning that the action is performed so as to encircle something; TA indicates a tense-aspect marker:
Næss (2002:71) argues that it is not plausible to characterise the verb vakao as semantically intransitive in sentence (3) and transitive in sentence (4). Rather, it takes the transitive suffix -ina in (4) because the verb it modifies, motia, is itself transitive. Thus, there is agreement in transitivity between the main verb and its modifying manner verb. This type of modification construction is highly grammaticalised in Pileni and the modifying verbs always show the productive form of the transitive suffix (i.e. -ina) and no other, even when the same verb takes a different variant of the suffix when it functions as the main verb (Næss 2002:71). For example, the verb t(h)emu 'be quiet' takes -akina as its transitive suffix in regular transitive contexts, such as under causativisation. However, when it modifies a different verb (e.g. kip-ina transitive 'keep'), it takes the productive suffix -ina (e.g. kip-ina themu-ina 'keep it [the truth] quiet'). According to Næss, this is the case with all verbs occurring in this construction as modifiers of manner.
However, the question remains whether the multiple occurrence of the transitivity marker within the same clause, both in Shipibo-Conibo and in Pileni, can be regarded as agreement. In all languages studied so far where transitivity is morphologically marked, it is a semantic value added onto the clause, an option chosen by the speaker who decides to express the transitive meaning through the clause. In Shipibo-Conibo and Pileni, this additional information has to be expressed on more than one element in the clause - rather like semantic case may have to be expressed on all elements of the noun phrase for which it is selected. Hence we call transitivity a morphosemantic feature.
We have not found any justification to regard transitivity as a morphosyntactic feature. Morphosyntactic features are found on targets of agreement, or on governees due to government. The fact that there are two or more elements in the clause marked for transitivity does not in itself constitute agreement. It means that the feature is marked more than once, but not that the adverbial modifier or the stative verb is the target of transitivity agreement imposed by the main verb. We take as definitional of agreement that it requires different controllers systematically co-varying with the targets. Transitivity in Shipibo-Conibo and Pileni does not fall under this definition because no other value of this feature (e.g. intransitivity) is expressed or marked in a comparable way - there is no evidence that the marking of transitivity is a member of a two-way opposition, rather than being an expression of optional information. We argue that transitivity is better analysed as an extra piece of information that is multiply marked in the clause in Shipibo-Conibo and Pileni. In this way it is similar to the phenomenon referred to as 'negative concord'. When the negative value is chosen by the speaker, it may also be marked in more than one place in the clause in some languages. However, when the clause is not negative, we do not want to say that it is marked for 'positive' feature value; in a 'positive' clause the negative value is not there without a trace, because the speaker simply did not choose to express (i.e. optionally add) this meaning.
How to cite this entry: