Causatives and Causation: A Universal-Typological Perspective. By Jae Jung Song. London & New York: Longman, 1996.


1.      Introduction

Causatives and Causation is an ambitious work, which in the space of 187 pages proposes to create a new scheme for classifying all possible causative constructions, identify a process of language shift which could be used to trace the emergence of causative affixes even in languages with no historical documentation, and ultimately to discover how the human mind understands the concept of causation through the evidence provided by linguistic typology.

In this paper I will attempt a review of major points of the book, including the typology of causatives developed by Song, and the process of historical language change which leads to the development of causative affixes from purposive constructions. In addition, I will briefly review the criticisms made of Song’s work. Finally, I will examine Guaraní causative constructions from the viewpoint of the framework outlined by Song, and the criticisms made of it. This will include a description of purposive constructions which may have a causative meaning in Guaraní, and an attempt to place each type within the classification scheme established by Song.


2. Types of causative constructions

Song claims to have consulted a database of 408 languages in developing the classification scheme used in this book. Song divides the possible causative constructions he has observed in his sample into three general types, which he calls the COMPACT, the AND, and the PURP types. A specific language may employ a variety of constructions, and Song argues later in the book that such variation within a language may be a sign of historical shift in the language. I will deal first with the biclausal AND and PURP types, and then with the monoclausal COMPACT type Song claims is a result of historical change in the previous types.


2.1 The AND type

Song describes the AND as a biclausal causative construction, with one clause containing [Vcause], the other, [Veffect]. In the AND type of causative construction, [Vcause] must always precede [Veffect]. AND is iconic in this respect, because it reflects temporal sequence of events, where a cause always precedes its effect. In the prototypical AND construction, the boundary between clauses is marked with a clause coordinator, represented in the formula as “AND.”


            (1)                    S1(S2(…[Vcause]…)S2+AND+S2 (…[Veffect]…)S2)S1


2.1.1 Overt AND

A number of languages, among them Vata, closely resemble the prototypical AND type. English also allows causative constructions which resemble this model.





(2)                                She whistled     and   the dog came running.

                                    (S2 [Vcause])  AND  (S2        [Veffect])


More common, according to Song, are languages which realize the AND element as an affix. Such affixes may occur on [Vcause], [Veffect], or both. Yet other languages will employ a switch reference system to combine the clauses.


2.1.2 Covert AND

Not all AND constructions use a clause boundary marker. Song calls languages which have zero marking of AND, but retain the other features of the category, “covert AND” constructions. An important feature of the AND type is the set order of [Scause] and [Seffect], and in covert and constructions, it is this iconic order which takes over the job of AND.

Although in the prototypical AND type the sentence is biclausal, Song identifies many languages in which this clausal independence is reduced. In Kobon and Yapese, for example, the verb expressing [Vcause] is fully inflected for tense and aspect, but the [Veffect] may lose some tense or aspect marking. The causee noun phrase, in the Akyem dialect of Twi, moves from being marked as subject, as it would be in a fully independent clause, to being marked as object.

Clausal independence may erode further than it does in the examples given above. Serial verb causative constructions, in which verbs are placed in close proximity and share the same tense, aspect, mood and polarity, are examples of an AND type of causative in which clause boundaries are very indistinct.

Song characterizes AND causatives as always being implicative, and this feature may be helpful in distinguishing it from the other biclausal causative type, the PURP type, described next.


2.2 The PURP type

Song asserts that there is another, previously unrecognized group of causative constructions he calls the “PURP” type. The PURP causative consists of two clauses, one denoting eventx carried out for the purpose of realizing eventy. The clauses may be joined by an element that indicates goal or purpose.


            (3)                    S1 (S2 (…[Veffect]…) S2+PURP+[Vcause]…) S1


S1 (…[Vcause]…S2 (…[Veffect]…) S2+PURP) S1.


Unlike the AND type, the [Vcause] and [Veffect] of the PURP type may occur in either order, probably due to the unidirectional nature of the purposive marking involved. The PURP marking may occur either on the dependent nominal or the governing verb. Verbal marking of irrealis on the [Veffect] may also be employed to indicate the nonfactual nature of the [Seffect]. Irrealis verb markings like future tense, subjunctive mood, or incompletive aspect, then, can be responsible for the goal or purpose meaning, which according to Song is extended to include causation.

PURP causatives are suspected by Song to be implicative in many, but not all, cases, due to the semantic neutralization of the PURP. With neutralization, the goal or purpose meaning is weakened or lost. However, in Chapter 5, he describes the range of possible implicativity of the PURP type as ranging from 0-100%.

Song claims in Chapter 3 that many, if not all PURP type causatives are the result of purposive construction being employed to express causation. Semantic bleaching of the purposive term used (often a directional or applicative affix metaphorically extended to include the meanings of goal or purpose) may lead to the reduction and grammaticalization of the PURP construction into a COMPACT type of causative. This hypothesis of historical change will be outlined in section 3 of this paper.



2.3 The COMPACT type

The COMPACT type includes lexical and morphological causative constructions in which the cause and effect elements are contiguous. Causatives of this type are monoclausal. Song uses this formula to describe the idealized model of causative constructions he groups under the COMPACT heading:


(4)                    S1(…[Vcause]+[Veffect]…)S1


COMPACT causatives may appear in languages in a number of ways. A common manifestation of the AND causative is a causative affix on the verb. In Song’s model, the affix would represent [Vcause] of the formula, and the verb to which it is attached would represent the [Veffect]. The order of affix and verb, and therefore of cause and effect, would vary according to the affixation permitted by a specific language. Song provides examples of languages that employ various affixation strategies in causative constructions.

In chapter 3, Song argues that COMPACT causatives are the result of reduction and grammaticalization of the AND or PURP types, although he speculates that most COMPACT types develop from PURP types, due to factors which will be discussed in section 3 of this paper. Since it is at least possible that a COMPACT causative could be the remainder of an AND causative, the implicativity pattern of COMPACT constructions could resemble either that of AND causatives (100% implicative), or PURP causatives (0-100% implicative).


2.3.1 Causative prefixes and suffixes

Prefixation and suffixation are the most common affixation strategies found cross-linguistically. Bilaan and Abkhaz are cited as examples of languages that use a causative prefix. Basque has a causative suffix, putting the [Vcause] element after the [Veffect] element represented by the verb.

Causative affixation may be observed in Guaraní in the verbal prefix mbo- which has the lexical meaning of “to do” or ‘to make.” (This prefix may originate from the independent verb (j)apo, “to do/make” and possibly also the noun po, “hand.”) Gregores-Suárez and Suárez describe mbo as a factitive prefix, which ignores the obviously causative role it plays in sentences like:




(5)                    Ña Marina       o-   mbopupu   la      y.

Mrs. Marina 3SG makes-boil ART water

Mrs. Marina boils the water.


Guaraní employs a number of other causative and quasi-causative strategies, which will be outlined in section 6.


2.3.2 Causative circumfixes and infixes

Less common affixation patterns used to express causation include circumfixation and infixation. Circumfixation may be represented schematically with the following formula:


            (6)                    [V cause1 [Veffect] Vcause2]


The causative circumfix surrounds the verb.  Georgian is cited as a language that uses a causative circumfix. Another uncommon affixation strategy, that of infixation, can be considered almost the reverse of circumfixation. In the case of infixation, the causative morpheme appears within the [Veffect]:


            (7)                    [Veffect1 [Vcause] Veffect1]


Although infixation for any purpose is rare cross-linguistically, Song mentions Nancowry as a language that employs it to express causation.


2.3.3 Causative free morphemes

The previous AND types resemble one another in that the morpheme which represents [Vcause] is bound to the verb which represents [Veffect]. Song asserts that [Vcause] can also occur as free morpheme, and cites the Romance languages of French, Italian, and Spanish as examples. These languages use a series of verbs to express cause and effect. The verb which represents [Vcause], in the Romance languages as elsewhere, is often that with the meaning of “to do” or “to make.”


            (8)                    Haré                            traer             mi             azada    a        Luís.

Make1SG/FUT            bring            my            hoe       DAT  Luís

I will make Luís bring my hoe.


Because [Vcause] is a free morpheme, it is possible, though sometimes unusual, for elements such as the negative to occur between [Vcause] and [Veffect].


2.3.4 Zero-Derivation causative constructions

It is also possible for a language to have no causative morphology, and express causation with the same structures also used for non-causative purposes. Such a strategy is known as zero derivation, and it can be found in Greek and Modern Chinese. Because the cause and effect elements in such a construction cannot be distinguished or separated, Song describes causatives of this type as demonstrating “extreme compactness” between [Vcause] and [Veffect]. Lexical causatives are also examples of extreme compactness, as in the English causative verb “kill,” which fuses the meanings of “to cause” and “to die.”

Finally, Song notes that there are a few languages which do not seem to have any COMPACT causative construction, among them Djaru and Luo.


3. Historical sources of causative affixes

Song outlines a four-step diachronic model to explain the development of causative morphemes from AND/PURP constructions:

Stage I: A non-causative construction (such as a purposive construction) begins to be used to express causation. The original [Vcause] marker is still present.

Stage II: The strong association of the purposive construction with causation allows the option of omitting the [Vcause] marker.

Stage III: The [Vcause] marker no longer appears in the AND/PURP construction, which has completely taken over the function of causation.

Stage IV: The AND/PURP element becomes reduced to a derivational causative affix, and possibly later, combines with verbs to form lexicalized causative verbs.

He goes on to eliminate AND causative types as a source of causative affixes, since, according to his description of the type, the AND element tends to be semantically empty, and quite often not even present in the construction.

In a chapter devoted to the topic, Song looks at Korean as an example of a language possibly at Stage II of this process. Korean has suppletive lexical causatives, morphemes that derive causative verbs from noncausative ones, and syntactic causatives that employ a complementizer –ke, and the verb meaning “to do.”  Song notices that there is an indistinct division between non-causative sentences that use –ke as a complementizer, and sentences which employ –ke to express causation. He takes this as evidence that Korean is undergoing the above process, and that a new COMPACT causative morpheme may be the eventual end result.


4. Cognitive basis for causative constructions.

            Song argues that the differing causative constructions which occur cross-linguistically can provide insight into the way in which the human mind understands causation. He identifies three components of causation which are highlighted or suppressed to varying degrees in different causative constructions:

i.                     Perception of some desire or wish

ii.                   Deliberate attempt to realize desire or wish

iii.                  Accomplishment of desire or wish.

These three components have an intrinsic temporal progression which is maintained even when one of the components is suppressed in a particular type of construction.  Song describes AND causatives as highlighting ii. and iii., while i. is suppressed. Consider example 2 again.


(2)                    She whistled     and   the dog came running.

                                    (S2 [Vcause])  AND  (S2        [Veffect])

                                                ii                                      iii


[Vcause] reflects the attempt to realize a desire or goal (ii), and [Veffect] reflects the accomplishment of this goal (iii). The desire or goal of the subject is not stated in this sentence; that is, i. is suppressed.

            PURP constructions are described by Song as highlighting i. and ii. and suppressing iii. This may be the dynamic at work in the following Guaraní construction:


            (9)                    A-    ju      Paraguái   pe      a-     mba’apo ha§ua.

1SG come Paraguay ALL  1SG    work       for.

                                                ii                                           i

                                    I come to Paraguay to work.


This construction is non-implicative. What is being expressed in the second clause is the desire to work, not the accomplishment of this goal.

            Song identifies a parallel description in the “chain of causality,” first proposed by DeLancey in his work with Lhasa Tibetan:


            (10)                  Volition -> Event -> Resulting State


5. Reviews of the book

            A major complaint about this book has to do with the sample used. Although the total number of languages Song claims to have included is 408 (Toops questions whether even a fraction of those were really examined closely), some reviewers of the book questioned whether the sample does indeed reflect the entire range of variation possible in human languages. Moore and Polinsky argue that Song sacrifices depth for breadth in electing to use such a large sample. In spite of the sample size, it may not be completely representative of all languages, considering that Ethnologue estimates there to be 6,700 languages in the world, and is highly probable that languages which could disprove Song’s classificatory scheme were missed. Toops notes that Balto-Slavic languages are underrepresented in the database, despite a large body of available research. Whaley notes that the sample of convenience used by Song may fail to control  for genetic and aereal biases.

            The breadth vs.depth controversy also appears in the doubts which some reviewers have about the accuracy of Song’s interpretation of what constitutes a causative construction in many languages. Moore and Polinsky charge that he misrepresents some non-causative sentences by glossing them as causative to prove his point.

            Other reviewers (De Haan),  argue that the AND/PURP/COMPACT categories are so vague as to be unworkable, and that the relationships between them are unclear. Moore and Polinsky question whether the categories Song proposes are “motivated,” or whether they were created around the data Song had available, what they term an “analytic convenience.” That Song goes on to use the new and unproven typology to support other new hypotheses is, to Moore and Polinsky, objectionable.


6. Causative constructions in Guaraní

Although Song included Guaraní in his database, no example from the Guaraní is given in the book. It may be presumed that Guaraní was one of the languages used “only minimally” by Song because its source was limited to only morphological expressions of causation. Indeed, Song’s source, Gregores-Suárez and Suárez, describes the only suffixes-gui and -uka as causatives. Other explanations of Guaraní grammar not consulted by Song do describe syntactic causatives, and in the final section of this paper, I will attempt to provide a more complete account of possible causative constructions in Guaraní.


6.1 Classification tests

To classify each construction, I will use four tests, based on the characteristics of each causative type, as established by Song. To separate the COMPACT causatives from the other two types, I will first determine the number of clauses required in each construction. As a first test to distinguish AND types from PURP types, I will determine the canonical order of [Vcause] and [Veffect] in each construction. As noted previously, AND constructions always follow the iconic order [Vcause] + [Veffect], while PURP causatives may follow either order. A second test of AND vs. PURP will be to determine the implicativity of the causation, since according to Song, AND causatives are always implicative and PURP causatives may vary in their implicativity. A final test will be to determine if a purposive, directional, or irrealis marker is present in the construction which could tentatively identify it as PURP.


6.2 Gui

-Gui is recognized by Gregores-Suárez and Suárez, Meliá Lliteras, and CHP as a causative suffix, although it is also recognised as a directional affix with ablative meaning:


(11)           Ha-á       ta             kavaju ári gui.

1SG/fall/FUT horse- on  ABL

I’ll fall off the horse.


            -Gui is also used in comparative constructions like


(12)           Xe       vare’a  ve        ka’i    Botánico gui.

1SG     hungry-more monkey  zoo-   ABL

I’m hungrier than a zoo monkey.


In causative constructions -gui appears on the [Vcause], which usually is ordered last in the biclausal construction:


(13)           Ndokýi                        o-     jere  gui   yvytú.

NEG/3SG/rain/NEG   3SG turn-ABL wind

It didn’t rain because the wind changed.


The cause (wind change) is ordered after the effect (no rain) and marked with the ablative affix. The causation in this example is implicative. Thus, according to the tests set up in 6.1, this would make the –gui causative a member of the PURP group.

-Gui may also appear on a nominal, usually one which can also be used as a verb:


(14)           O-     mano ñembyahýi-               gui.

3SG  die      hunger(be hungry)  ABL

He died of hunger.


There are cases, though, when –gui appears on a nominal. This erosion of clausal boundaries may be evidence of movement toward a COMPACT type of causative.


(15)           Xe    kuerái  ko   clase- gui.

1SG  bored  this class ABL.

This class bores me.


6.3 Rupi

-Rupi is a suffix on the [Vcause] which indicates past causation. Like –gui, it is also a directional suffix, usually glossed as “by” in English and “por dónde” in Spanish. It is only recognized as a directional affix by Gregores-Suárez & Suárez and Meliá Lliteras CHP cites a causative meaning, among several others:


(14)           A-     ha- rupi   o-      pa   la        lio.

1SG  go since  3SG  end  ART  fight

Since I left, the fight ended.


The order of this bicausal construction is [Vcause] + [Veffect].  It is implicative. Although the classification of this construction is uncertain due to the use of a directional affix, lack of evidence for a possible [Veffect]+[Vcause] order would suggest that this is an AND type of causative.


            6.4 Haguére/Hague

            Because of their similarity of use and meaning, I will consider haguére and hague together.  These two suffixes are used with verbs of feeling or sentiment, and mark the irrealis [Vcause] which causes the feeling or emotion. Hague is used in expressions where the [Vcause] is non-factual (NF).


(15)           A-     mbyasy xenupahague xe     sy.

1SG   sad       1SG/hit/NF   1SG   mother

I’m as sad as if my mother had hit me.


Haguére is used with future events.


(16)           A-      vy’a       o-      mendahatahaguére José rehe.

1SG   happy    3SG   marry/ FUT/ NF      José with

I’m glad she’s going to marry José.


Both these constructions are biclausal. They are both implicative. Both bollow the pattern [Veffect]+[Vcause]. Their only function seems to be to mark the causative event  as nonfactual. Therefore, these two constructions belong unambiguously in the PURP group.


            6.5 Ha§ua

            Ha§ua is described by Gregores-Suárez and Suárez as a conjunction, but it seems to function as a  purposive marker with verbs that express a wish or desire. It follows  the verb that expresses the desired effect.


(17)           He’i         xéve         a-sê   ha§ua   hógagui.

3SG/say   me/ALL   1SG PURP  house/ABL

He told me to leave his house.

The ha§ua expression is nonimplicative. The expression is biclausal, and puts [Vcause] before [Veffect]. It may be confidently classified as a PURP construction.


            6.6 Râ

            is usually defined as a “future possessive” suffix rather than a causative, although it can also express the meaning “to become,” as seen in the example below.


 (18 )                Rafaela o-      studia maestrarâ

                                    Rafaela 3SG  study   teacher-become

                                    Rafaela is studying to become a teacher.


This construction is monoclausal, with –râ marking the hoped-for future effect, usually a noun. No verb of [Vcause] is present in the construction, and –râ seems to take over the causal function. This affix seems to be derived from the benefactive suffix –guarâ, and to share its benefactive connotation:


(18)           Ko  haiha   ndéve guarâ

This pencil you    BEN

This pencil is for you


The existence of a causative construction that employs a benefactive affix in place of the [Vcause] may be evidence for Song’s theory of the origins of COMPACT causative constructions. It may be an example of a Stage III type of construction, and as such, could be classified as being at an intermediate stage between a PURP construction and a COMPACT construction.


            6.7 Mbo-

            Mbo- is a factitive verbal prefix that can be added to intransitive verb stems to make them transitive.


(19)           A-     mbojahu    la       mitâme.

1SG  FAC/bathe ART  child DAT

I bathe the child.

CHP and Meliá Lliteras recognise the causitive meaning of mbo- and its allophones. The addition of mbo- shifts the subject of the intransitive verb to the position of object the transitive verb. Compare:


(20)           O-    gue              la        tata

3SG  extinguish  ART               fire (NOM)

The fire went out.


(21)           Re-  mbogue               la        tata.

2SG FAC/extinguish  ART   fire (ACC)

You put out the fire.


The causative meaning of  mbo- in (21) is unmistakable. As previously mentioned, Mbo- may be related to the verb (j)apo, meaning “to do” or “to make,” a fact which is consistent with the way causatives are formed in other languages, through the juxtaposition of the verb meaning “to do/make” with the verb denoting the desired effect.  This prefix can then confidently be classified as a COMPACT causative, as it combines [Vcause] and [Veffect] within the same word.


            6.8 –Uka

            The final causative morpheme I will examine is the verbal suffix –uka. Whereas Mbo- is used only with intransitive verbs and nonverbal elements like adjectives, -uka is affixed to transitive verbs to denote forced or otherwise direct causation.


(22)           Don Marciano o-     ñotyka            hikuái mandyju kokuépe.

Don Marciano 3SG  plant/CAUS   3PL    cotton     field/LOC

Don Marciano made them plant cotton in the field.


If mbo- is factitive, then –uka can be considered double factitive, as it adds a third argument to the clause. It is possible that the affix –uka is related to the verb (j)uka “to kill” in the same way that mbo- may be related to the verb (j)apo, but any such connection must remain tentative until more evidence becomes available. However, -uka, like mbo-, can be classified as a COMPACT causative which joins cause and effect in the same word.


7. Conclusion

            The classification scheme developed by Song is not unworkable. Objections raised by critics about the superficial nature of the sample and the ad hoc classificatory scheme can be overcome by a more concentrated examination of specific languages, and how they justify or contradict the proposed system. The assertions Song makes about the historical source of causative affix and the implications his model has for our understanding of human cognition must likewise be proven, piece by piece,  by specialists in these areas. Such an ambitious set of assertions as those found in this book cannot be accepted as more than speculation without withstanding such tests. This examination  of a single language’s strategies for forming causative constructions provides a bit of positive evidence for Song’s model.



De Haan, Ferdinand. Review: Causatives and Causation: A Universal-Typological Perspective. Southwest Journal of Linguistics, 17, 109-111.


Guarani II and Guarani III. Paraguay, CHP International.


Krivoshein de Canese, Natalia and Feliciano Acosta Alcaraz. 1991. Ñe’eryru: Diccionario Guaraní-Español. Asunción, Paraguay: Colección Ñemitÿ.


Meliá Lliteras, Bartolomé, Alfonso Pérez Peñasco and Luís Farré Maluquer. 1960. El Guaraní A Su Alcance. Asunción, Paraguay: Ediciones Loyola.


Moore, John and Maria Polinsky. 1998. Review Article: Causatives and Causation. Linguistic Typology, 2, 231-251.


Song, Jae Jung. 1996. Causatives and Causation: A Universal-Typological Perspective. London: Longman.


Gregores-Suárez, Emma, and Jorge Alberto Suárez. 1961. A Description of Colloquial Guaraní. Ann Arbor: University Microfilms.


Toops, Gary H. 1998. Review: Causatives and Causation. Language, 74, 670-671.


Whaley, Lindsay J. 1999. Review: Jae Jung Song, Causatives and causation: a universal-typological perspective. Journal of Linguistics, 35, 229-230.