Causative

For the grammatical case, see Causal case.

In linguistics, a causative (abbreviated CAUS) is a form that indicates that a subject causes someone or something else to do or be something, or causes a change in state of a non-volitional event.

All languages have ways to express causation, but differ in the means. Some languages have morphological devices (such as inflection) that change verbs into their causative forms, or adjectives into verbs of becoming. Other languages employ periphrasis, with idiomatic expressions or auxiliary verbs. All languages also have lexical causative forms (such as English rise → raise).

Typology

Shibatani (2001) lists three criteria for entities and relations that must be encoded in linguistic expressions of causation:

1. An agent causing or forcing another participant to perform an action, or to be in a certain condition
2. The relation between [the] two events [=the causing event, and the caused performing/being event] is such that the speaker believes that the occurrence of one event, the ‟caused event,” has been realized at t2, which is after t1, the time of the ‟causing event”
3. The relation between causing event and caused event is such that the speaker believes the occurrence of the caused event depends wholly on the occurrence of the causing event—the dependency of the two events here must be to the extent that it allows the speaker a counterfactual inference that the caused event would not have taken place at a particular time if the causing event had not taken place, provided that all else had remained the same. (1976a: 1-2)

This set of definitional prerequisites allows for a broad set of types of relationships based, at least, on the lexical verb, the semantics of the causer, the semantics of the causee and the semantics of the construction explicitly encoding the causal relationship. Many analysts (Comrie (1981), Song (1996), Dixon (2000) and others) have worked to tease apart what factors (semantic or otherwise) account for the distribution of causative constructions, as well as to document what patterns actually occur cross-linguistically.

Comrie

Comrie (1981: 158-177) focuses on the typology of the syntax and semantics of causative constructions proper. Crucially, Comrie (and others to be discussed here) distinguish between the linguistic encoding of causal relations and other, extra-linguistic concerns, such as the nature of causation itself, and questions of how humans perceive of causal relations. While certainly not irrelevant, these extra-linguistic questions will, for now, be left aside. Comrie usefully characterizes causative events in terms of two (or more) microevents perceived of composing a macroevent, and encoded in a single expression (of varying size and form). Formally, he categorizes causatives into 3 types, depending on the contiguity of the material encoding the causing event and that encoding the caused event. These are: 1) lexical causatives, in which the two events are expressed in a single lexical item, as in the well-discussed case of English kill; 2) morphological causatives, in which the causing event and the caused event are encoded in a single verbal complex via causative morphology, and, prototypically, morphological marking showing the status of affected arguments. Finally, Comrie discusses analytic causatives, in which the causing event and the caused event are encoded in separate clauses.

Comrie’s work is also noteworthy for having brought the notion of syntactic hierarchy to bear on the typology of causative constructions. A hierarchy of grammatical relations had already been formulated to help explain possibilities for relative clause formation (first presented as Keenan and Comrie’s (1972) NP accessibility hierarchy; see Croft 1990: 147), and Comrie argued that a similar hierarchy was in play, at least in some constructions, in the marking of the original A argument when a base transitive clause is causativized. The hierarchy is as follows:

• subject > direct object > indirect object > oblique > genitive

Comrie’s argument was, in short, that some causativized-transitive constructions mark the new A as belonging to the leftmost available slot in the above hierarchy. Dixon (2000) fleshes out a version this analysis in more detail.

Song

Presenting a typology of causatives and causation based on a database of a staggering 600 languages, Song (1996) is very critical of typological work that depends on statistical inference, citing data from the Niger-Congo family that contradicts some earlier claims that “languages within genera are generally fairly similar typologically” (Dryer 1989: 267)(19) Song therefore culls data from every language for which adequate documentation was available to him, and categorizes the various causative constructions gleaned therefrom into 3 classes: COMPACT, AND and PURP.

Song employs the following terminology:

• [Scause] – the clause which denotes a causing event
• [Seffect] – the clause which denotes the caused event
• [Vcause] – verbal elements of [Scause]
• [Veffect]- verbal elements of [Veffect] (20)

The major differences between Song’s analysis and Comrie (1981) and Dixon (2000), is that Song lumps the range of lexical and morphological causatives together under the label COMPACT (20), in which [Vcause] can be “less than a free morpheme” (e.g., bound morpheme [prefix, suffix, infix, circumfix, reduplication], zero-derivation, suppletion); or “a free morpheme” (28), in which [Vcause] and [Veffect] form a single grammatical unit. I note that most of the examples given look like serial verb constructions, and no in-depth analysis is undertaken for some of the constructions in which [Vcause] and [Veffect] are less formally contiguous. Song notes this non-contiguity, but does not undertake to explain why it might be important.

The AND causative, for Song, is any construction with a separate [Scause] and [Seffect] i.e., in which “two clauses [are] involved” (35). This, in theory could include larger, multi-clausal expressions of causal relations which many analysts probably would not label a ‘causative construction’, e.g.: ‘It rained yesterday, so they stayed home,’ but the boundaries of the AND causative category are not discussed.

One of Song’s major contributions to the literature is fleshing out an analysis of his PURP causative. These are constructions which encode intended causation on the part of the causer, but which do not encode any outcome: i.e., the speaker encodes [Vcause] and causer intentionality, but remains agnostic as to whether [Veffect] was felicitously effected.

Dixon

Dixon (2000), in his authoritative typology of causatives, discusses the syntax and semantics of all types of causative constructions, in much more detail than can be recounted here. One research question he begins to tackle is the following: Many languages, as he and many others have documented and attempted to categorize, have at least two causative constructions. Leaving aside for now the issue of lexical causatives (except where zero-derivation has been demonstrated to be a productive morphological process), these are often broadly divided into ‘more compact’ and ‘less compact’, with labels, differing by analyst, indicative of relative length of the forms in question (e.g., Comrie’s straightforward ‘morphological’/’syntactic’, or Song’s (1996) ‘COMPACT’/’AND’). Earlier works had attempted to summarize the semantic differences under the vague (though preliminarily useful) rubric of the “Iconicity Principle” (see Huang and Su (2005) for a succinct discussion), which basically posits a correlation between the degree of formal compactness of the linguistic material encoding the causative macroevent and the perceived directness of the relationship between causing event ([Vcause]) and caused event ([Veffect]): i.e., shorter forms, on the whole, were posited to encode more direct causation than longer forms, as in the classic English I killed him. [direct causation] vs. I caused him to die.[less direct causation] examples.

The Iconicity Principle is a good first step, but does not really explain any fine-grained semantic distinctions that may be in play. The first attempt to take the analysis further, to my knowledge, was Comrie’s (1981:164-7) discussion of directness and control, which began looking at the semantics of the causer and causee as possible semantic factors influencing the distribution of different causative constructions. Dixon (2000), however, goes several steps further, identifying and fleshing out a system of 9 scalar parameters along which causative constructions tend to vary, supported by data from his typological survey. These are as follows:

• Relating to the lexical verb involved in the construction:

1. stative vs. active

2. intransitive vs. transitive vs. ditransitive

• Relating to the causee:

3. having vs. lacking control [over the caused microevent]

4. acting willingly vs. unwillingly

5. partially affected vs. completely affected

• Relating to the causer:

6. acting directly vs. indirectly

7. acting accidentally vs. intentionally

8. acting naturally vs. with effort [i.e., initiation of the caused microevent takes less or more effort]

9. involved vs. not involved in the activity [=the caused microevent]

This set of parameters is useful in and of itself, but, importantly, Dixon applies it to all causative constructions in a sample of over 25 languages, and notes that more formally compact causative constructions prototypically tend toward the following values on the above 9 criteria:

Causer acts naturally [8], intentionally [7] and directly [6]; the causee either lacks control [3] or has control but is willing [3 & 4], and is only partially affected [5]. May apply only to active intransitive verbs [or may include some small subclass or transitives] [1 & 2], or to state verbs alone [1]. (Summarized from Dixon (2001:77).)

Dixon thus provides a data-driven account of the prototypical ‘more compact’ and ‘less compact’ causatives (though noting that these prototypes are somewhat artificial, chiefly in that many causative constructions in the languages surveyed involve only some subset of the criteria above), and calls for broader typological surveys to test his model—a call which I have taken into account when selecting the languages for my preliminary survey. One ongoing research goal in this dissertation tests these parameters against this new set of languages: further below, I present a survey of causative constructions from my database, along with notes on which of Dixon’s criteria seem to be at play.

As for the syntax of causatives, Dixon notes that for causativized intransitives, the overwhelming tendency is for the old S argument to be marked as O, with the introduced argument functioning as the new A. Somewhat more interestingly, Dixon posits six typological classes based on how languages (constructions?) treat arguments after a new argument has been introduced via the causativization of a base transitive verb.

Talmy

Talmy (2003 v.2: 67-101) contains an in-depth investigation of different types of causal relations. Talmy refers to these as “lexicalization patterns,” a term which remains unclear to me, given that few of the examples given in his discussion are lexical items, and most interpretations of “different types of causation incorporated in the verb root” are in fact wholly dependent on other morphosyntactic material in the clause. Let us first examine his list of possible (semantic) causative types (69-70), with examples:

• autonomous events (non-causative) The vase broke.
• resulting-event causation The vase broke from a ball’s rolling into it.
• causing-event causation A ball’s rolling into it broke the vase.
• instrument causation A ball broke the vase.
• author causation (unintended) I broke the vase in rolling a ball into it.
• agent causation (intended) I broke the vase by rolling a ball into it.
• undergoer situation (non-causative) My arm broke (on me) when I fell.
• self-agentive causation I walked to the store.
• caused agency (inductive causation) I sent him to the store.

One question remaining to be explored is how this set of divisions usefully differs from other analysts’ typologies of the semantics of encoding causal relations. Some overlap in the types of semantic information in play is immediately apparent, however: in cases of instrument causation (‘the hammer broke the cup’), we would certainly expect the ‘causer’ to be acting directly [Dixon’s criterion 6] and to be involved in the activity [criterion 9]; likewise, we would expect instances of caused agency to include more information on causee control on willingness [criteria 3 & 4].

Susceptibility to causativization

Shibatani (2001) classifies verbs into four categories, according to how susceptible they are to morphological causativization:

1. Inactive intransitives [=unaccusatives], e.g., 'faint'
2. Middle/ingestive verbs, e.g., 'eat'
3. Active intransitives [=unergatives], e.g., 'work'
4. Transitive verbs, e.g., 'carry'

In what can be seen as a re-characterization of Dixon’s (2000) criteria (2; transitive/intransitive/ditrasitive), (6; ‘directness of causer action’) and (8; ‘causer effort’), Shibatani argues that this hierarchy of susceptibility to causativization can be characterized in terms of the semantic role of the causer, which he characterizes, in turn, in terms of the difficulty the causer experiences in bringing about the caused event. He argues [in terms reminiscent of Talmy’s (2003) force dynamics] that a patientive causee [S, who prototypically is reassigned to O] poses less resistance than an agentive causee; in the former case, the only counterforce to overcome is the causee’s inertia, either in continuing to be at rest or in continuing to undergo a change. In the latter case (some kind of caused agency), however, causation requires the participation of another entity: in force-dynamic terms, the potential resistance is greater [and, causer control over the realization of [Veffect] is lower], hence the hierarchy:

• inactive intransitives > active intransitives > transitives.

Examples

Indo-European

Sanskrit

In Sanskrit, there is a causative form of the verb (n.ijanta), which is used when the subject of a clause forces or makes the object perform an action. The causative suffix -ay is attached to the verbal root (this may cause vowel sandhi to take place).

• bhū "to be, exist" → bhāv-ay; e. g. bhāvayati "he causes to be"
• khad "to eat" → khād-ay; e. g. khādayāmi "I cause to eat" = "I feed"

Proto-Germanic

In Proto-Germanic, the parent language of the Germanic dialects including English, causative verbs are formed by adding a suffix -j/ij- to the past-tense ablaut of a strong verb, with Verner's Law voicing applied. (All of these characteristics derive from the way that causative verbs are formed in Proto-Indo-European, with an accented -éy- suffix added to the o-grade of a non-derived verb.) Examples:

• *rīsanaN (I) "to rise" → *raizijanaN "to raise", i.e. "to cause to rise"
• *frawerþanaN (III) "to perish" → *frawardijanaN "to destroy", i.e. "to cause to perish"
• *nesanaN (V) "to survive" → *nazjanaN "to save", i.e. "to cause to survive"
• *ligjanaN (V) "to lie down" → *lagjanaN "to lay", i.e. "to cause to lie down"
• *grētanaN (VII) "to weep" → *grōtijanaN "to cause to weep"

In English, to sit / to seat : to settle, and in German, sitzen/setzen : stehen/stellen form pairs of causatives/resultatives.

Persian

In Persian, causative form of the verb is made by adding ân(i)dan to the present stem:

• xordan (to eat) → xor (present stem) → xorândan (to cause/make to eat)
• xandidan (to laugh) → xand (present stem) → xandândan (to cause/make to laugh)

Latin

Latin has inherited a few Indo-European causatives.

   * *iacio "to lay down" vs. *iaceo "to throw"
   * *sedeo "to sit" → *sīdō, "to settle", from an older meaning "to put down"
   

Urdu

Urdu uses the infix "-(l)ā-" and -(l)vā-" to render verbs causative.

• karnā "to do" → karānā "to have done" → "karvānā" → "to have s.o. make s.o. do."
• paṛhnā "to read" → parhānā "to make s.o. read" → "paṛhvānā" "to cause s.o. to make s.o. read."
• hilnā "to move" → hilānā "to have s.th. moved" → hilvānā "to have s.o. make s.th. move."
• pīnā "to drink" → pilānā "to have s.o. drink" → pilvānā "to have s.o. make s.o. drink" e.g.: "Usne naukarānī se bachchõ-ko pānī pilvāyā" - "She had the maid make the kids drink water."

Semitic

In most Semitic languages there is a causative form of the verb. It is postulated that in Proto-Semitic the causative verbal stem was formed by the š- prefix, and this has become ʾa-, hi- or ī- in different languages.

• Syriac: kəθav "he wrote" → ʾaxtev "he composed"
• Arabic: ʿalima "he knew" → ʾaʿlama "he informed"
• Hebrew: tsaxak "he laughed" → hitsxik "he made sb. laugh"

Arabic also has a causative form (Form II) created by gemination of the central consonant of the triliteral root, as follows:

• ʿalima "he knew" → ʿallama "he taught"

The ʾa- form (Form IV), while used in Modern Standard Arabic, is no longer productive in many of the colloquial varieties of Arabic, which uniformly prefer Form II.

Japanese

In Japanese there is a similar causative/obligative inflection:

• 食べる taberu "to eat" → tabesaseru "to make to eat, to feed"
• 読む yomu "to read" → yomaseru "to make to read"

Khmer

Khmer has six prefixes and one infix to derive the causative form of verbs, although they vary in frequency and productiveness. The consonantal prefix p- is one of these affixes.

• coap "joined" → pcoap "to join"
• cum "around" → pcum "to gather"

Finnish

Causative forms are also found in the Uralic languages of Europe, such as Finnish.

• syödä "to eat" → syöttää "to feed"
• täysi "full" → täyttää "to fill"
• haihtua "to evaporate" → haihduttaa "to vaporize"

Notice that the causative suffix is often used irregularly and/or because of historical reasons, e.g. Finnish:

• olla "to be" → olettaa "to assume", not "to make exist"
• kirja- ancient "patterns (of embroidery or text)", but modern "book" → kirjoittaa "to write" ("transform into patterns of text"), not "to transform into books"

Austronesian

Maori

In the Māori language of New Zealand, the whaka- prefix can be added to a verb, for example:

• ako "to learn" becomes whakaako "to teach" (to cause to learn)

Philippine languages

In Philippine languages such as Tagalog and Ilokano, the pa- prefix is added to verbal forms and to adjectives to form causatives.

• dakkel "big (adj)" → padakkelen "to enlarge" (Ilokano)
• kain "eat" → pakainin "to make eat, to feed" (Tagalog)

Guaraní

In Guaraní, an Amerindian language, the mbo- prefix is added to oral verbs, and mo- to nasal verbs:

• puka "to laugh" → mbopuka "to make (someone) laugh"
• guata "to walk" → mboguata "to guide"
• pu'ã "to go up" → mopu'ã "to elevate"

Uto-Aztecan

Classical Nahuatl has a well-developed morphological system of expressing causation by means of the suffix -tia:

• cua "he eats" → quicualtia "he feeds him" (the causative makes the intransitive verb eat into the transitive verb feed, requiring a pronominal prefix, in this case qui- "him")
• calli "it is a house" → caltia "he built a house" (i.e. "he caused it to become a house")

Athabaskan

Rice (2000) makes the following points about morphological causatives in Athabaskan:

# In all Athabaskan languages surveyed [including Hupa, for which an ample data set is presented], the causativizing morphology can causativize at least some intransitive verbs with patientive subjects.

1. For intransitive verbs with agentive patients, the family shows a split: only some languages allow morphological causativization in this case.
2. Koyukon (Northern Athabaskan; Alaska) was found to be the only language in the survey allowing productive morphological causativization of transitive verbs.
3. [P]erhaps the presence of the direct object pronoun in the causative construction has something to do with whether the causee is human or animate, or capable of being regarded as such. When the causee or the verb cannot be or is not perceived as a potential controller, then the pronoun is not found [in the Athabaskan languages surveyed].

The semantic factor of causee control; that is, the degree of control that that causee wields over the effecting of the caused microevent (also discussed as parameter #3 on Dixon’s (2000:62) list), and which Rice (2001) finds to be a major factor in other Athabaskan causatives, helps account for quite a bit of the distribution of the Hupa syntactic causative (below).

Hupa

Golla, in his (1970) descriptive grammar of Hupa (summarized in Sapir and Golla (2001)), describes three classes of morphologically-derived causatives:

# causatives from descriptive neuters with ƚ-classifier (176)
ni-whon’ ‘be good, beautiful’ → O ni-(w)-ƚ-whon’ ‘cause O to be beautiful’

1. causatives from primary extension neuters with ƚ- classifier (76-77, 201)
   na-…‘a’ ‘O hangs’ → na-O-ƚ-‘a’ ‘hang O up’
2. causatives from primary intransitive action themes (76-77, 204)
   ti-ch’id ‘grow tired’ → O-ti-ƚ-ch’id ‘tire O out’

While Golla does not generalize about the semantics of verb themes that are compatible with causative ƚ-, several preliminary generalizations can be made. First of all, in the three cases described by Golla, O [the undergoer] is neither controlling nor agentive; O is largely patientive in all cases. Second, the causer appears to be acting directly on O. Third, none of the examples given (including the examples above, of course) involve the causativization of a base-transitive theme.

Devices

There are various ways of encoding causativity.

Periphrastic constructions

There are no regular causative inflections in English, nor in any of the major European languages, which resort to idiomatic uses of certain verbs like English make or have, French faire or laisser, or German lassen. For example:

• She made me eat the vegetables.
• I had John build the house.
• I had the posters taken down.

Note that this type of structure is more complicated than the inflectional causative form exemplified in Sanskrit, since it has two verbs and three arguments: the first is the subject of the first verb; the second is the object of the first verb but also the subject of the second; and the third is the object of the second verb. These arguments can be exchanged using passive voice (in either verb), but the result can be cumbersome or even ungrammatical.

Other complex constructions include the use of subjunctive forms. Spanish uses these often, since it does not allow some simpler constructions that English permits.

• Él hizo que la siguieran. "He had her followed.", lit. "He had (things done so) that they would follow her."
• Hicimos que el perro comiera pescado. "We made the dog eat fish.", lit. "We did (things so) the dog would eat fish."

In the Romance languages, a number of verbs alternate between intransitive (semantically middle voice) and causative transitive, using a pseudo-reflexive clitic pronoun:

• Ella se despierta a las 7. "She wakes up at 7."
• Ella despierta a los niños. "She wakes up the children."

Lexical

In many cases, a language simply uses a different lexical item to indicate a causative form. For example, the causative of English rise is rear, and the causative of eat is feed. English allows a notable freedom in verb valency, resulting in verbs like break, burn or awake, which may be causative or not (he burns it = he causes it to burn). Causativeness is therefore zero-marked in many English verbs.

In Japanese, there are a large number of verbs that alternate in various semi-regular patterns between intransitive forms and causative transitive forms, for example:

• agaru "to go up, to rise" → ageru "to raise"
• magaru "to turn" → mageru "to bend"
• kowareru "to be broken" → kowasu "to (cause to) break"
• kaeru "to go back" → kaesu "to send back"

For the purpose of syntax, a derivation that turns an adjective or noun into a "verb of becoming" works the same as a causative construction for intransitive verbs. For example, in English the derivational suffixes -(i)fy can be thought of as a causative:

• simple → simplify = "to make simple", "to cause (sthg.) to become simple"
• object → objectify = "to make into an object", "to cause (sthg.) to become an object" (figuratively, that is)

Changes of state

In languages with stative verbs (equivalent to English adjectives), the acquisition of a quality, or changes of state, can be expressed with causatives in the same way as with regular verbs. For example, if there is a stative verb to be large, the causative will simply mean to enlarge, to make grow. The reflexive form of this causative can then be used to mean to enlarge oneself, or even as a middle voice, to grow.

Syntactic causative constructions

A causative form or phrase can be thought of as a valency-increasing voice operation, which adds one argument. If the original verb is intransitive, then the causative construction as a whole is transitive: to fall → to make (sbdy./sth.) fall, to topple (sbdy./sth.), or indeed, to fell, a fossilised form from when causatives were an inflexional part of English grammar. If the original verb is transitive, the causative is ditransitive: to eat (sth.) → to make (sbdy.) eat (sth.), to feed (sth.) to (sbdy.).

Causative voice

The causative voice is a grammatical voice promoting the oblique argument of a transitive verb to an actor argument. When the causative voice is applied to a verb, its valency increases by one. If, after the application of the grammatical voice, there are two actor arguments, one of them is obligatorily demoted to an oblique argument.

Japanese and Mongolian are examples of languages with the causative voice. The following are examples from Japanese:

Tanaka-kun	ga	atsume-ru
Tanaka	nom	collect-pres
Tanaka collects them.
Causative
Tanaka-kun	ni	atsume-sase-yō
Tanaka	dat	collect-caus-cohort
Let's get Tanaka to collect them.

kodomo	ga	hon	o	yom-u
children	nom	book	acc	read-pres
Children read books.
Causative
kodomo	ni	hon	o	yom-aseru
children	dat	book	acc	read-caus-pres
(They) make children read books.

Bibliography

• Comrie, B. 1981. Language Universals and Linguistic Typology: Syntax and Morphology. p. 158-177. Chicago: University of Chicago Press.
• Croft, W. 2003. Typology and Universals, 2nd ed. New York: Cambridge University Press.
• Dixon, R.M.W. and Alexandra Y. Aikhenvald. 2000. “Introduction”. In Changing Valency: Case Studies in Transitivity, Dixon, R.M.W. and Alexandra Y. Aikhenvald, eds: 1-28. New York: Cambridge University Press.
• Dixon, R.M.W. 2000b. “A typology of causatives: form, syntax and meaning”. In Changing Valency: Case Studies in Transitivity, Dixon, R.M.W. and Alexandra Y. Aikhenvald, eds.: 30-83. New York: Cambridge University Press.
• Goertz, G. et al. 2006. “Use of causatives in Navajo: Syntax and morphology.” In Santa Barbara Papers in Linguistics, Volume 18: Proceedings from the Ninth Workshop on American Indigenous Languages.
• Huang, S. and Lily I-Wen Su. 2005. “Iconicity as Evidenced in Saisiyat Linguistic Coding of Causative Events.” Oceanic Linguistics, Vol. 44, No. 2 (Dec., 2005): 341-356.
• Rice, Keren. 2000. “Voice and valency in the Athabaskan family”. In Dixon, R.M.W. and Alexandra Y. Aikhenvald, eds. 2000. Changing Valency: Case Studies in Transitivity. New York: Cambridge University Press.
• Shibatani, M. (ed.) (2001) The grammar of causation and interpersonal manipulation. Amsterdam and Philadelphia: John Benjamins
• Song, J.J. (1996). Causatives and causation: A universal-typological perspective. London and New York: Addison Wesley Longman.
• Song, J.J. (2001) Linguistic Typology: Morphology and Syntax. Harlow and London: Pearson (Longman).
• Talmy, L. 2000. Toward a Cognitive Semantics Volume 1: Concept Structuring Systems. Cambridge: MIT Press.
• Talmy, L. 2000. Toward a Cognitive Semantics Volume 2: Typology and Process in Concept Structuring. Cambridge: MIT Press.

External links

• What is a causative?