1.  Introduction

2. Alsina's Cross-linguistic Parameterization

In Type I, the patient of the causative morpheme is fused with the logical subject of the base verb. Semantically, this means that the causer brings about a change by acting on an individual who is in control of that change event.  In Type II, the patient of the causative morpheme is fused with the patient (or affected argument) of the base verb. Semantically, this means that the causer acts on the base patient by causing an event that affects that individual. If this affected entity is different from the base logical subject, the base logical subject is regarded as an intermediary of causation. Alsina  claims that Type I occurs in causatives in Chamorro and Japanese, while Type II occurs in Marathi and Malayalam, and that both types occur in Chichewa and in Catalan.
In this account, both Type I and Type II have a composite argument structure which is uniformly mapped onto a simplex lexical form of a functionally monoclausal predicate. In Type I, the patient of the causative is mapped onto OBJ, with the base patient (if present) mapped onto OBJ_Th. In Type II, the patient of the causative is mapped onto the OBJ, and the logical subject of the base verb is demoted to an optional oblique (Grimshaw's (1990) argument adjunct). The patterns in the mapping of arguments onto grammatical functions are also given in (1). The regularities of such mapping are accounted for by the theory of Lexical Mapping (Bresnan & Kanerva 1989, Bresnan & Moshi 1990, Alsina 1992) or Functional Mapping (Alsina 1996).
    Additional, independent parameters have been recognized in order to account for full range of variation to be found in different languages. For one thing, the above association patterns are independent of how causatives are expressed in the surface form. The combination of a base verb and a causative verb might be realized as two independent words (as in English and French; the case of periphrastic causatives) or as two morphemes within a single word (as in Japanese, Marathi, and Chichewa; the case of morphological causatives). In Alsina's theory, this difference is captured by the Incomplete Predicate Parameter (Alsina 1997; see also Zaenen & Dalrymple 1996).
    In addition, Alsina (1992) argues that the independently motivated parameter of object symmetry constrains possible passivization patterns of causatives. In Alsina's theory, the patient of the causative (which normally is realized as OBJ) can become a passive subject in both Types I and II. Whether the base patient in Type I can be a passive subject depends on whether the language does or does not allow a secondary object to be a passive subject. If the language is an object-symmetrical language (Bresnan & Moshi 1990), in which both direct and indirect object of ditransitive verbs (i.e., two patient or patient-like arguments) can be [-r] and become a passive subject, the base patient can indeed be a passive subject in Type I causatives. If it is an object-asymmetrical language, in which only an indirect object can be [-r] and become a passive subject, the base patient cannot be a passive subject. By comparing Chichewa and Kichaga, Alsina (1992) argues that this independently motivated parameter of object symmetry predicts whether a base patient in Type I causatives can be a passive subject.
There are, however, some causative phenomena that are not accounted for in this theory. First, the mappings of arguments onto functions given in (1) do not seem to be the only possibilities that are found with Type I and Type II argument fusion. Butt, Dalrymple & Frank (1997) point out that in one type of French causative which Alsina identifies as Type I, the causee is in fact mapped onto OBJ_Th, and the base patient onto OBJ. Alsina & Joshi (1991) themselves recognize another mapping pattern for Type II causatives. In the causativization of 'ingestive' verbs (Masica 1976), the base agent is mapped onto OBJ and the base patient onto OBJ_q, in a way superficially similar to Type I (see sec. 5.3 below).
Second, some have also argued that causatives differ along another dimension of variation, namely functional and/or argument structure complexity (Frank 1996, Matsumoto 1996, Zaenen & Dalrymple 1996). Building on the work of Zubizarreta (1987), Rosen (1989), and others, Frank (1996) argues that French and Italian causatives differ in the complexity of a-structure, namely, French causatives have a complex, biclausal a-structure, while Italian causatives have a simplex or monoclausal a-structure.(Fn1) In my own work I have also argued that such variation in complexity exists within a single language, namely Japanese; morphological causatives in this language are generally biclausal at a-structure, but are in some cases biclausal at f-structure as well (Matsumoto 1996).
In this paper I will discuss this additional parameter of structural complexity in causative predicates by taking a closer look at Japanese morphological causatives.
 

3. Japanese Causatives and the Type II Sase Causatives

3.1. Types of Morphological Causatives in Japanese

(2)   Jon wa       Biru ni   soba o                     tabe-sase-ta.
       John Top    Bill Dat  soba.noodles Acc   eat-Caus-Past
       "John made/let Bill eat soba noodles.'

It has long been noted that sentences like this exhibit certain biclausal properties. In order to account for this, various proposals have been made in which biclausality is expressed in some level of representation. Early transformationalists, for example, posited biclausal Deep Structure (and monoclausal Surface Structure) (see Kuno 1973, 1983, Shibatani 1976, etc.; cf. Miyagawa 1980). In LFG, Ishikawa (1985) argued that Japanese morphological causatives create biclausal f-structure, in which the base verb heads the XCOMP of the causative (see Manning, Sag & Iida 1996 for a proposal in HPSG).
In my own work, I have argued that morphological causatives differ in their biclausal properties according to their subtypes (Matsumoto 1996), as mentioned briefly above. As observed by Shibatani (1976, 1978, 1990), there are potentially four different readings available for Japanese morphological causatives, as summarized in (3). These readings are in part correlated with the dative vs. accusative case marking of the causee when the base verb is intransitive, though this distinction is not operative when the base verb is transitive.

  __                     __                   __
 |                      | explicit permissive   |
 |permissive causatives |          causatives   |
 |                      | implicit permissive   |  biclausal f- and
 |                      |          causatives   |  a-structure
 |                       --                     |
 |                       __                     |
 |                      | persuasive causatives |
 | inducing causatives  |                     --
 |                      | coercive causatives   } biclausal a-structure
 |                       --                       and monoclausal
  --                                              f-structure

(4) a.  explicit permissive causatives

 'permit <Agent, Recipient, Event>'                                    a-structure
                    |            |            |     |
                    |            |            |  'drink <Agent, Patient>'
  _ _ _ _ _ _ |_ _ _ _ |_ _ _ _ |_ _ _ _  _ | _ _ _ _| _ _ _  _ _ _ _ _ _ _
                    |            |            |                |            |
'permit <SUBJ,   OBJ,   XCOMP>      |            |
                                |                |             |            |
                                |            'drink <SUBJ, OBJ>                f-structure
                                |______________|
                                            control
 

      b.  implicit permissive causatives

'permit <Experiencer, Event>'                                        a-structure
                    |                 |    |
                    |                 | 'drink <Agent, Patient>'
  _ _ _ _ _ _ | _ _  _ _ _ |_ _ _ _ _ _| _ _ _| _ _ _  _ _ _ _ _ _ _
                    |                 |                 |         |
'permit <SUBJ,   XCOMP> OBJ  |         |
                                      |          |      |         |
                                      |    raising  |         |
                                      |          |      |         |
                                  'drink <SUBJ,   OBJ>                f-structure
 

      c.   persuasive inducing causatives
 
  'cause <Agent, Recipient, Event>'                                    a-structure
                    |           |             |      |
                    |           |             |  'drink <Agent, Patient>'
  _ _ _ _ _ _ |_ _ _ _| _ _ _ _ |_ _ _ _ _ _| _ _ _ _| _ _ _  _ _ _ _ _ _ _
                    |           |             |                 |            |
'cause <SUBJ,   OBJ,   XCOMP>        |            |
                                |              |                |            |
                                |           'drink  <SUBJ, OBJ>                f-structure
                                |_______________|
                                            control

      d.   coercive inducing causatives

                                                 fusion
                                            _______________
                                            |                             |
              'cause <Agent, Patient, Subevent>'   |                      a-structure
                                |           |              |               |
                                |           |          'drink <Agent, Patient>'
  _ _ _ _ _ _ _ _ _ _ |_ _ _ _| _ _ _ _ _ _ _ _ _ _ _ _ _| _ _ _  _ _ _ _ _ _ _
                                |           |                                       |
  'cause.to.drink <SUBJ,  OBJ,                            OBJpt>     f-structure
 

In the case of explicit and implicit permissive causatives and persuasive conducing causatives, the subject of the embedded clause is functionally controlled by the object of the upper clause (explicit permissive causatives and persuasive inducing causatives involve control (equi) and implicit permissive causatives involve raising). In the case of coercive causatives, the logical subject of the embedded clause is linked to the patient of the upper clause by argument 'fusion' in a-structure.
 

3.2. The Type II Sase Causatives

(5)  a.   Hahaoya wa     akachan ni    kutsushita o      hak-ase-ta.
             mother Top      baby Dat       socks Acc        put.on-Caus-Past
             "The mother put the socks on the baby's feet.'

     b.     Hahaoya wa     akachan ni    miruku o     nom-ase-ta.
             mother Top      baby Dat        milk Acc     drink-Caus-Past
             "The mother fed the baby with milk (in a bottle).'

The base verb hak(-u) "put on' in (5a) refers to the action of putting a clothing item on one's own lower body. The causative form hak-ase(-ru) in its Type II causative reading in (5a) represents the causer's action on the clothing item to the effect that someone else can have it on his or her lower body. (This means that hak(-u) means 'put something on your own lower body' while hak-ase(-ru) means 'put something on someone else's lower body'; note that English put on is misleadingly ambiguous in this regard.) In (5b), nom-ase(-ru), with the base verb nom(-u)  'drink', is used to mean 'feed' or 'give a drink to'.(Fn2)  (These sentences potentially have regular permissive and inducing causation readings as well. However, given that a baby is not independently capable of the caused or permitted actions described, such readings are pragmatically unnatural. The only natural reading one can get is that considered here.).
The semantics of these sentences is that of Alsina's Type II. In (5a), for example, the mother acts on the socks to achieve the intended effect, so that the caused event can be accomplished without the action of the base agent. Note, however, that unlike what Alsina claims with regard to Type II causatives, the base subject in this causative is not an oblique. I will come back to this point later.
This type of -sase  causative is only possible with a restricted set of transitive base verbs, which are listed in (6).

(6) transitive base verbs                      causativized verbs
hak(-u) 'put ... on one's own lower body'   hak-ase(-ru) 'put ... on someone else's lower body'
tabe(-ru) 'eat'                                                tabe-sasd(-ru) 'make ... eat, feed'
nom(-u) 'drink'                                             nom-ase(-ru)  'make ... drink, feed'
shir(-u)  'come to know'                               shir-ase(-ru) 'let ... know, inform'
kik(-u) 'hear'                                                 kik-ase(-ru)  'let ... hear, tell'
mot(-u) 'come to have'                                  mot-ase(-ru) 'make ... have, put in the hand of'

These base verbs all represent 'self-directed' processe---an action conducted with the agent himself as the recipient or beneficiary of a moved entity or influence. The Type II sase causatives of these verbs represent the causation of such self-directed action by the causer acting on the base patient. This means that in Type II sase causatives, the base subject is not so much the agent of the caused action as the recipient of the causer's action. Thus the semantic structure of Type II sase causatives can be represented as something like (7a) or perhaps (7b), structures similar to those which have been proposed for ditransitive verbs in English.

(7) a.
        Cause <Actor, Acted-upon, Result>
                                          |              |
                                          |            Go <Figure, Path>
                                          |____________|          |
                                                                            To <Ground>

     b.
         Cause <Actor, Acted-upon, Result>
                                          |              |
                                          |           Have <Posesser, Possessed>
                                          |_______________________|
 

4.  Evidence for the Purely Monoclausal Nature of Type II Sase Causatives

(8)  Type II sase causatives

        'feed <Agent, Recipient, Patient>'                    a-structure
    _ _ _ _ _ _  |_ _ _ _ _ |_ _ _ _ _| _ _ _ _ _ _ _ __ _
                       |               |              |
         'feed <SUBJ,   OBJ,    OBJpt>'                      f-structure
 
Note that in this a-structure the base verb does not have any argument structure of its own and its 'subject' does not function as a logical subject of any predicate, let alone a grammatical subject.
 In what follows, I will present evidence for this analysis in comparison to implicit permissive causatives and coercive causatives. (The other two subtypes will be ignored here for simplicity of presentation.) Relevant evidence comes from honorific marking, passivization, pronominal binding, control, and adjunct interpretation.(Fn3)
 

4.1.  Subject Honorification

(9) a. PERMISSIVE CAUSATIVE
        Karera wa   ooji ni         sono kutsushita o  o-haki ni         nar-asete  oki-mashi-ta.
        they Top      prince Dat  the    socks Acc     H-put.on Cop become    put-Pol-Past
        'They let the prince (continue to) put the socks on his feet.'

  b. COERCIVE CAUSATIVE
        ?Karera wa   muriyari   ooji ni       sono  kutsushita o
         they Top      forcibly   prince Dat  the     socks Acc
         o-haki ni          nar-ase-mashi-ta.
         H-put.on Cop  become-Caus-Pol-Past
        'They forcibly made the Prince put the socks on his feet.'

  c. TYPE II SASE CAUSATIVE
         *Karera wa     mada  sankagetsu no       ooji ni        sono   kutsushita o
          they Top        yet      three.month Cop  prince Dat   the      socks Acc
          o-haki ni          nar-ase-mashi-ta.
          H-put.on Cop become-Caus-Pol-Past
          'They put the socks on the three-month-old Prince's feet.'  (intended)
 

4.2. Passivization

(10)   ?[PRO  Fukei o         kanshin s-aseru tame]       totemo             muzukashii
                       parents Acc  admire-Caus      purpose   particularly     difficult
            ji ga             (kodomo-tachi ni)  kak-ase-rare-ta.
            letter Nom   child-Pl Dat            write-Caus-Pass-Past
           'In order to impress parents, particularly difficult (Chinese) characters were made [i.e., by the teacher] to be written (by the children).'

Significantly, Type II sase causatives even more clearly allow the base object (as well as base subject) to be the passive subject, just as regular ditransitive verbs do in Japanese. Examples are given  in (11). This again argues for monoclausality.

(11) a.   Sono  rinyuu-shoku wa  mada   dono    akachan ni mo    tabe-sase-rarete i-nai.
              the     baby.food Top      yet      any      baby Dat too       eat-Caus-Pass   Asp-Neg
              'The baby food has not yet been given to feed any child.'

       b.   Sono  akachan wa  mada   dono    rinyuu-shoku o     tabe-sase-rare-ta.
             the      baby Top     yet       any      baby.food Acc      eat-Caus-Pass-Past
             'The baby has not been fed with any baby food.'
 
This means that the arguments of both the causative morpheme and the base verb map onto grammatical functions of a single predicate in f-structure.
 

4.3. Pronominal Binding

(12) a.  Taro_i wa  Jiroo_j ni  sonomama     kare_i,*j / karejishin_*i,j o    bengo s-asete       oi-ta
             Taro Top Jiro Dat   as.it.is           he/himself Acc                  defend do-Caus    put-Past
             'Taro appears to let Jiro continue to defend him(self).'

       b.  Taro_i wa  Jiroo_j ni   muriyari  kare_??i,*j / karejishin_?i,j o   bengo sase-ta.
            Taro Top  Jiro Dat   forcibly   he/himself Acc                    defend do-Caus put-Past
            'Taro appears to let Jiro continue to defend him(self).'

       c.  Anpanman_i wa   akachan ni   kare_*i/karejishin_i o   tabe-sase-ta.
            Anpanman Top  baby DAT    he/himself Acc          eat-Caus-Past
            'Anpanman gave his own body to (feed) the baby.'(Fn5)

4.4. Control

(13) a.  PERMISSIVE CAUSATIVE
             Jon_i wa     sono ko_j ni    [PRO_?i,j terebi o           mi-nagara ]     kutsushita o
             John Top  the    child Dat              television Acc watch-while   socks Acc
             hak-asete        oi-ta
             put.on-Caus   put-Past
             'John let the child put on the socks, watching TV.'

         b.  COERCIVE CAUSATIVE
             Jon_i wa     muriyari  sono ko_j ni    [PRO_i,j    terebi o           mi-nagara]
             John Top  forcibly   the     child Dat              television Acc watch-while
             sono    kutsushita o     hak-ase-ta.
             the       socks Acc       put.on-Caus-Past
             'John forcibly made the child put on his socks, watching TV.'

         c. TYPE II SASE CAUSATIVE
             Jon_i wa     sono  netakiri no              roojin_j ni
             John Top  the     confined.to.bed      old.man Dat
             [PRO_i,*j  terebi o                mi-nagara]     sono kutsushita o     hak-ase-ta.
                              television Acc      watch-while the     socks Acc       put.on-Caus-Past
             'Watching TV, John put the socks on the feet of the old man confined to bed.'
 

4.5.  Adjunct Interpretation

(14)   Jon wa     Marii ni   {shibushibu/ooyorokobi de}   kisu s-are-ta.
          John Top Mary by   reluctantly/happily                  kiss do-Pass-Past
         'John was {reluctantly/happily} kissed by Mary.' (either John or Mary is reluctant or happy)

Now, in permissive causatives these adverbials can be interpreted with respect to the base subject, as in (15a), in which either John or the child is happy. The same is true of coercive causatives in (15b). However, this is not the case with Type II sase causatives in (15c), in which only John can be happy.

(15) a. PERMISSIVE CAUSATIVE
         Jon wa        sono ko ni            sono  kutsushita o   ooyorokobi de  hak-asete      oi-ta.
         John Top    the      child Dat    the      socks Acc      happily            put.on-Caus put-Past
         'John let the child put on his socks(,) happily.'

     b.  COERCIVE CAUSATIVE
         Jon wa        muriyari  sono ko ni        sono  kutsushita o  ooyorokobi de hak-ase-ta.
         John Top    forcibly   the    child Dat  the     socks Acc      happily            put.on-Caus-Past
         'John forcibly made the child put on his socks(,) happily.'

      c. TYPE II SASE CAUSATIVE
         Jon wa       sono  netakiri no          roojin ni           sono  kutsushita o
         John Top   the     confined.to.bed  old.man Dat     the     socks Acc
         ooyorokobi de  hak-ase-ta.
         happily             put.on-Caus-Past
        'John happily put the socks on the feet of the old man confined to bed.'

The aggregate of evidence presented above argues strongly that Type II sase causatives are monoclausal in a- and f-structure.(Fn6)
 

5. Similarity to Ditransitive Lexical Causatives

(16)  kise(-ru) 'put ... on (someone else's body), dress' (cf. ki(-ru) 'put on one's body'), abise(-ru) 'pour (over ....)' (cf. abi(-ru) 'be covered with (bathed in)'), kabuse(-ru) 'cover with, put on someone else's head' (cf. kabur(-u) 'become covered with, put on one's own head'), mise(-ru) 'show' (cf. mi(-ru) 'see'), oshie(-ru) 'teach' (cf. osowar(-u) 'learn'), sazuke(-ru) 'endow' (cf. sazukar(-u) 'receive'), azuke(-ru) 'entrust' (cf. azukar(-u) 'be entrusted'), tama(-u) 'give' (cf. tamawar(-u) 'receive'), kas(-u) 'lend' (cf. kari(-ru) 'borrow') , and ii-tsuke(-ru) 'order' (cf. ii-tsukar(-u) 'be ordered')

These are all based on transitive verbs representing self-directed actions. In addition, in all of these ditransitive lexical causatives, the causer acts on the patient of the related transitive verb to bring about the result, just as in Type II sase causatives. Note that the pair ki(-ru) 'put ... on one's own upper body' and kise(-ru) 'put ... on someone else's upper body' exactly parallels the pair hak(-u) and hak-ase(-ru) above.
All ditransitive lexical causatives in Japanese are of this type: there are no ditransitive lexical causatives in which the causer acts on the agentive actor (i.e., Type I association). Further, all transitive lexical causatives can also be regarded as Type II in that only an affected entity not in control of the resulting event can be the object of such a causative (see Alsina & Joshi 1991). Thus, all lexical causatives ((mono)transitive or ditransitive) in Japanese can be regarded as Type II.
 

6. Typology of Causative Predicates

6.1. Variation in Structural Complexity of Causatives

6.2. Structural Complexity and Association Types

(17) a. Type I
                            __________
                            |                    |
    'cause  <x,     y,    event>'   |
                     [affected]   |         |
                                    '...   < x .....>
                                     [most salient,
                                       in control]

    b. Type II
                                 _________________
                                 |                                 |
        'cause  <x,      y,        event>'            |
                           [affected]     |                   |
                                            '....   < (...)  x .....>
                                                           [affected,
                                                            not in control]

Japanese presents several different cases in which the various types of association are realized in causatives of different complexity. Permissive causatives in Japanese can be characterized as Type I in the sense that the most salient function in the embedded clause is linked to a particular function of the causative morpheme. Given that such causatives have biclausal a- and f-structure, they can be placed in slot (a) in (18). Coercive causatives also involve Type I association, but with the association now taking place in a-structure. Given that these are biclausal in a-structure but monoclausal in f-structure, such causatives can be classified as case (b) in (18). Type II sase causatives and lexical causatives, by contrast, are instances of Type II with the association taking place in semantic structure. This is case (f). There are no examples of cases (c), (d), and (e) in Japanese.(Fn7)

 (18)
          a-str   |   biclausal  |  biclausal   |  monoclausal
          f-str   |   biclausal  |  monoclausal |  monoclausal
$(B!!!!!!(B    ________|______________|______________|______________
          Type I  | (a)  OK      | (b)  OK      | (c)  *
          Type II | (d)  *       | (e)  *       | (f)  OK

                                OK:   attested        *: unattested
 
Case (e) in (18) can be sought in other languages. In fact, in Alsina's analysis, where only types (b) and (e) exist, Type II causatives in Chichewa, Marathi, and Catalan are all claimed to be of case (e). A close examination of the data, however, suggests that Marathi causatives are in fact case (f). Consider the Marathi causative sentence in (19) (Alsina & Joshi 1991).

(19)  sumaa-ni  raam-kad5un/*-laa       shaam-laa     maarawle.
        Suma-Erg Ram-by/-Acc             Sham-Acc     beat-Caus
        'Suma made Ram beat Sham. (Suma had Sham beaten by Ram.)'

When a transitive verb is causativized, as in (19), the base subject is expressed as an oblique, with the base object appearing as the accusative object of the causativized verb (Alsina & Joshi 1991). Alsina & Joshi claim that this sentence involves Type II association. (Presumably this means that the goal of causation is to get Sham beaten, with the base subject as an intermediary; see also Saksena 1980, 1982).
Sentence (19) does not seem to have a composite a-structure. T. Mohanan (1988) observes that the base subject in Marathi causatives does not have the status of a logical subject. Consider, for example (20) (Mohanan 1988).

(20)  sumaa_i-ni     raam_j-kad5un  shaamk-laa [aaplyaa_i,*j,*k     ghar-aat]   maarawle.
        Suma-Erg     Ram-by         Sham-Acc  self's                    home-Loc beat-Caus
        'Suma_i made Ram_j beat Shamk in his _i,*j,*k house.'

The Marathi reflexive aapan5  must be bound by its logical subject (Joshi 1989). In sentence (20) the base subject cannot be the antecedent of aaplyaa (genitive form of aapan5 ). This means that Marathi causatives are monoclausal in a-structure (Mohanan 1988), like Japanese Type II sase causatives. The same can be said of Hindi causatives (Mohanan 1994: 38) and perhaps also of Malayalam and Chichewa Type II causatives, in which no logical subject properties of the base agent have been pointed out in the literature.
The missing case (e) in (18) above does seem to be found in Catalan. Alsina (1996) points out that there are two types of case marking in Catalan periphrastic causatives (as in other Romance languages), one dative and the other oblique, as exemplified in (21).

(21) a.   He       fet      netejar  els  lavabos al      general
              I have  made  clean    the  toilet   to the  general
               'I made the general clean the toilet.'

       b.    He      fet      netejar  els  lavabos  pel      general
             I have  made  clean    the  toilet      by the general
             'I had the toilet cleaned by the general.'

(21a) involves Type I association, and (21b), Type II. (In these cases causative verb and base verb are realized as two different morphological words at c-structure.)
Alsina argues that both of these causative types are functionally monoclausal but biclausal at a-structure. They are functionally monoclausal, he argues, given the patterns observed in quantifier floating. He further points out that the base subject in both (21a) and (21b) can be the controller of an adverbial clause subject, thereby showing that it is a logical subject; thus both of the bold-faced NPs can be a possible controller of the italicized verbs in the 'without' clause in (22a) and (22b).

(22) a.   En Pere      els-hi     fara              saludar   el   professor     sense        cridar
                    Pere        them      will make     greet      the  professor     without     shouting
             'Pere will make them greet the professor without shouting.'
 
       b.  En  Pere   els        fara            criticar      (pels  seus  collegues)
                    Pere     them    will make   criticize      by      his       colleagues
            sense       dir       res           a la       direcci
            without   saying   anything to the     management
            'Pere will have them criticized by his colleagues without telling the management anything.'

The conclusion is that Catalan causatives of the type (21b) do appear to represent case (e) in (18).
 (23), an expansion of (18), shows all the attested cases covered in this short review of causatives in various languages. (Fn8)

(23)
          a-str   |   biclausal  |  biclausal   |  monoclausal
          f-str   |   biclausal  |  monoclausal |  monoclausal
$(B!!!!!!(B    ________|______________|______________|______________
          Type I  | (a)  OK      | (b)  OK      | (c)  *
          Type II | (d)  *       | (e)  OK      | (f)  OK
 
                                OK:   attested        *: unattested

     (a)  Japanese permissive causative, English periphrastic causative
     (b) Japanese coercive causative, Catalan causative with dative base subject
     (e) Catalan causative with oblique base subject
     (f) Marathi and Hindi causative, Japanese Type II sase causative

I would like to argue that the distribution of attested cases in this table is not accidental, and that the gaps are systematic ones. That is, Type I association does not take place in purely monoclausal causatives (c), and Type II association does not take place in purely biclausal causatives (d).
 Why should this be? First, Type II association is not permitted in biclausal f-structure because such association would posit the control of an object, something unsanctioned in grammar.
On the other hand, Type I association is not allowed in purely monoclausal causatives because of a constraint on semantically possible verbs. As a number of linguists have argued, a lexical or purely monoclausal causative cannot be formed that would represent the external causation of a process which is intrinsically internally caused (e.g., by the will of the actor; see Haspelmath 1993, Levin & Rappaport Hovav 1994, 1995, Matsumoto 1996). It has been observed, for example, that there are no lexical causative counterparts of intransitive verbs such as weep, shout, rejoice, etc., which denote processes caused by factors internal to the subject of these intransitive verbs. Such constraints can be subsumed by the Determinative Causation Condition (Matsumoto 1996) stated in (24), which essentially says that in lexical causatives the causer must be the only individual in control of the resulting event.

(24) In order for an event of causation and its result to be expressed in a lexical meaning, the causing event must be the only crucial cause of the result.

Type I association in a purely monoclausal causative would necessarily result in the violation of constraint (24): the fact that the causer would be acting on the entity in control of the resulting event means that this entity's control would partially determine the result of causation. By contrast, Type  II sase causatives and ditransitive lexical causatives in Japanese do not violate this constraint because of the semantics of Type II association. In tabe-sase(-ru) 'make eat' or 'feed', for example, the causer acts on the food so that the base subject or eater is only the recipient of the action of feeding, not an internal causer of the caused process of eating.
The present typology of causatives can be compared with Baker's (1988) typology of causative predicates. Baker claims that there are basically two types of causatives, reflecting the two ways the base verb can be incorporated: a V-to-C movement type and a VP-to-Comp movement type. He argues that the former type of incorporation occurs in causatives in which the causee is realized as an object and the base patient as a secondary object (as in Alsina's Type I). In causatives formed through VP-to-Comp movement, on the other hand, the causee is realized as an oblique and the base patient as an object (as in Alsina's Type II). What is noteworthy in Baker's account is that his two major types are correlated with some properties related to the issue of mono- vs. biclausality. In the VP-to-Comp type the base patient can be a passive subject, and the base logical subject (causee) does not exhibit subject properties, pointing toward monoclausality. In the V-to-C type, by contrast, the base patient cannot be a passive subject and the causee exhibits subject properties in terms of binding. In Baker's account, however, the difference between grammatical and logical subject properties are not carefully distinguished. Moreover, there are some empirical problems with this approach. For example, grammatical or logical subject properties of a causee are lacking in certain causatives that exhibit the case marking pattern of the V-to-C movement type; in fact, the Type II sase causative above is one such example (see also note 10).

6.3.   Type II Linking and Verbs of Self-directed Action

(25)   sumaa-ni      raam-laa  paan5i   paad(I!(Bzle.
         Suma-Erg     Ram-Acc water   drink-Caus
          'Suma made Ram drink water.'

(Other Indian languages exhibit similar but somewhat different patterns. In Hindi and Malayalam, for example, ingestive verbs allow their subject to be an object as well as an oblique when causativized (Saksena 1980, 1982, Mohanan 1988).(Fn9))
There are, however, some differences between the causatives of self-directed verbs in Japanese and Marathi. First, both involve Type II association but the causative patient is linked to different arguments of the base verb. Alsina & Joshi (1991) argue that cases like (25) too involve Type II association, with the causative patient fused with the base agent, unlike Japanese in which it is linked to base patient. They note that the subject of the ingestive verb (e.g., the drinker) is affected (i.e., nourished) by the event (see also Saksena 1980, 1982), and thus the base subject qualifies for Type II association.
Second, passivization can promote the base patient as well as base subject into a subject in Japanese, but in Marathi it can only promote the base subject.(Fn10) This is partly due to the object symmetry parameter: Japanese is an object-symmetrical language and allows either of the two internal arguments to be [-r], while Marathi is object asymmetrical and allows only one of them to be [-r], in this case the base subject.
One might ask here why it should be the base agent rather than the base patient that is treated as [-r] in Marathi causativized ingestive verbs. A crucial factor seems to be affectedness. As mentioned above, in Marathi causativization, it appears, a special attention is paid to the affected nature of the agent of ingestive processes. This affected nature of the nourished person rather than the intake leads to the [-r] assignment of the base agent in causatives in this language (see Zaenen 1993 for the relationship between [-r] assignment and 'Proto-Patient' properties; see Dowty 1981). One might note here that Marathi and Japanese in this respect represent two different ways of conceptualizing actions like feeding. In contrast to Marathi, Japanese seems to treat feeding or dressing as caused motion; food is a moved entity and the eater is the recipient of the moved entity, and so Type II sase causatives behave like ditransitive verbs. In either case, the agentivity of the eater is suppressed, allowing a purely lexical causative expression.
These two possibilities parallel the alternations manifested by spray/load verbs. In essence, Japanese causativized self-directed verbs are like theme-object spray/load verbs, which in Japanese function as ditransitive verbs, while their counterparts in Marathi are like goal-object spray/load verbs like English feed. Such variation in different languages reflects different preferences in mapping arguments onto grammatical functions, preferences which are presumably independent of causative parameters. Such different preferences might be formalized along the lines of Butt, Dalrymple and Frank (1997); this, however, is beyond the scope of the present paper.

7. Conclusion

Notes

2.   One peculiar fact about these examples is that the dative-marked NP is non-agentive. It is possible to have an inanimate entity like a doll as a dative NP in (5a); thus the base 'agent' is not really an agent. This constitutes a major counterexample to the otherwise solid generalization that dative marking in Japanese causatives is allowed only on an agent (Tonoike 1978, Dubinsky 1994). (Back)

3.   I will not use reflexive jibun as a test for the subjecthood of a causee NP. See Iida (1996) for non-grammatical factors involved in jibun binding. (Back)
 
4.  Some might be inclined to attribute the unacceptability of sentences like (9b) to a pragmatic factor: it is pragmatically unlikely that someone would force a person one respects to do something. This account appears to be based on an incorrect assumption on subject honorification. Subject honorification reflects the sense of respect felt by the speaker of the sentence toward the subject of the verb, and there should be nothing strange about someone forcing a person the speaker respects to do something.  (Back)

5.   Anpanman is a benevolent comic character made of bread, who sacrifices himself for the hungry people.  (Back)

6.   One question to be explored is whether there are Type II sase causatives with intransitive base verbs. In this connection, one might note that some sase causatives with intransitive base verbs behave like regular transitive verbs in certain respects. For example, Manning, Sag & Iida 1996 point out that unmarked word order for the arguments of sak-ase(-ru) (bloom-Cause) has the order location-theme, as in (i).

(i)  sono     ame ga       no ni         hana o          sak-ase-ta
      the       rain Nom   field Loc    flower Acc  bloom-Caus-Past
     'The rain made the flowers bloom in the field.'

As they note, this is unexpected, since the order produced by clausal complementation would have the theme (an argument of the causative morpheme) preceding the locative (an argument of the base verb). This pattern seems to hold for certain nonagentive intransitive base verbs. Phenomena like this must be carefully evaluated to see if these are the case of monoclausal case (see also Aoki 1977).  (Back)
7.   The English make-periphrastic causative can be regarded as a further instatnce of case (a).  (Back)

8.   The typology of causatives can also be conceived in terms of the term/oblique status of the base subject. All Type I causatives have their base subject realized as OBJ (i.e., a term) while Type II causatives allow oblique as well as object base subject. Type II causatives with base subject realized as a term are found in Marathi (causativized ingestive verbs) and Japanese (Type II sase causatives). Both of these cases are purely monoclausal, and as far as I know, all cases of Type II causatives with biclausal a-structure and monoclausal f-structure have the base subject realized as oblique. Type II purely monoclausal causatives can also have an oblique base subject, as is the case with causatives of ingestive verbs in Hindi (T. Mohanan, personal communication; see also Saksena 1980, 1982). Given this, the variation in causative predicates can be represented as in (i).

 (i)
            a-str |   biclausal  |  biclausal   |  monoclausal
            f-str |   biclausal  |  monoclausal |  monoclausal
$(B!!!!!!(B  __________|______________|______________|______________
causee    term    |     Type I   |    Type I    |    Type II
 marking  oblique |      *       |    Type II   |    Type II
 
 (Back)
9.   K. P. Mohanan (personal communication) notes that in Malayalam a few psychological verbs behave in a similar way (e.g., wiswasi-ppi-kk-um 'persuade, make ... believe', formed from  wiswasikk-um 'believe').  (Back)

10.   In this respect, this causative would be treated as a case of V-to-C movement in Baker's (1988) theory. It should be noted, however, that here the base 'subject' does not have subject properties, revealing an insufficiency in Baker's typology.  (Back)
 

References

Alsina, Alex. 1996. The Role of Argument Structure in Grammar. Stanford: CSLI Publications.

Alsina, Alex. 1997. A theory of complex predicates: Evidence from causatives in Bantu and Romance. In Alex Alsina, Joan Bresnan & Peter Sells, eds.., Complex Predicates, 203-246. Stanford: CSLI.

Alsina, Alex, and Smita Joshi. 1991. Parameters in causative constructions. In Papers from the 27th Regional Meeting of the Chicago Linguistic Society, 1-15. Chicago Linguistic Society, University of Chicago.

Baker, Mark. 1988. Incorporation. Chicago: University of Chicago Press.

Bresnan, Joan, and Joni Kanerva. 1989. Locative Inversion in Chichewa: A Study of Factorization in Grammar. Linguistic Inquiry 20:1-50.

Bresnan, Joan, and Lioba Moshi. 1990. Object asymmetries in comparative Bantu syntax. Linguistic Inquiry 21:147-186.

Burzio, Luigi. 1986. Italian Syntax: A Government Binding Approach. Dordrecht: Reidel.

Butt, Miriam, Mary Dalrymple, and Anette Frank. 1997. An architecture for linking theory in LFG. In Miriam Butt and Tracy Holloway King, eds., On-Line Proceedings of the LFG 97 Conference. http://www-csli.stanford.edu/publications/LFG/

Dowty, David. R. 1991. Thematic Proto-roles and Argument Selection. Language 57:547-619.

Dubinsky, Stanley. 1994. Predicate union and the syntax of Japanese causatives. Journal of Linguistics 30:43-79.

Frank, Anette. 1996. A note on complex predicate formation: Evidence from auxiliary selection, reflexivization, and past participle agreement in French and Italian. In Miriam Butt and Tracy Holloway King, eds., On-Line Proceedings of the First LFG Conference. http://www-csli.stanford.edu/publications/LFG/lfg1.html.

Grimshaw, Jane. 1990. Argument Structure. Cambridge, Mass.: MIT Press.

Haspelmath, Martin. 1993. More on the typology of inchoative/causative verb alternations. In Bernard Comrie and Maria Polinsky, eds., Causatives and Transitivity, 87-111. Amsterdam/Philadelphia: John Benjamins.

Iida, Masayo. 1996. Context and Binding in Japanese. Stanford: CSLI Publications.

Inoue, Kazuko. 1976. Henkeibunpoo to Nihongo [Transformational Grammar and Japanese]. Tokyo: Taishukan.

Ishikawa, Akira. 1985. Complex Predicates and Lexical Operators in Japanese. Doctoral dissertation, Stanford University.

Jackendoff, Ray S. 1972. Semantic Interpretation in Generative Grammar. Cambridge, Mass.: MIT Press.

Jacobsen, Wesley M. 1992. The Transitive Structure of Events in Japanese. Tokyo: Kurosio Publishers.

Joshi, Smita. 1989. Logical subject in Marathi grammar and the predicate argument structure. In E. Jane Fee and Katherine Hunt, eds., Proceedings of the Eighth West Coast Conference on Formal Linguistics, 207-219. Stanford: The Stanford Linguistics Association.

Kitagawa, Yoshihisa. 1986. Subjects in Japanese and English. Doctoral dissertation, University of Massachusetts, Amherst.

Kuno, Susumu. 1973. The Structure of the Japanese Language. Cambridge, Mass.: MIT Press.

Kuno, Susumu. 1983. Shin Nihon Bunpoo Kenkyuu [New Studies in Japanese Grammar]. Tokyo: Taishukan.

Kuno, Susumu. 1987. Honorific marking in Japanese and the word formation hypothesis of causatives and passives. Studies in Language 11:99-128.

Levin, Beth, and Malka Rappaport Hovav. 1994. A preliminary analysis of causative verbs in English. Lingua 92:35-77.

Levin, Beth, and Malka Rappaport Hovav. 1995. Unaccusativity: At the Syntax-Lexical Semantics Interface. Cambridge, Mass: MIT Press.

Manning, D. Christopher, Ivan Sag, and Masayo Iida. 1996. The lexical integrity of Japanese causatives. In Takao Gunji, ed., Studies in the Universality of Constraint-Based Phrase Structure Grammars, 9-37. Osaka: Osaka University.

Marantz, Alec. 1984. On the Nature of Grammatical Relations. Cambridge, Mass.: MIT Press.

Masica, Colin P. 1976. Defining a Linguistic Area: South Asia. Chicago: University of Chicago Press.

Matsumoto, Yo. 1996. Complex Predicates in Japanese: A Syntactic and Semantic Study of the Notion 'Word'. Stanford: CSLI Publications & Tokyo: Kurosio Publishers.

Miyagawa, Shigeru. 1980. Complex Verbs and the Lexicon (Coyote Papers 1). University of Arizona Linguistics Circle, Tucson.

Mohanan, Karvanuur P. 1983. Move NP or lexical rules? Evidence from Malayalam Causativization. In Lori Levin, Malka Rappaport, and Annie Zaenen, eds., Papers in Lexical-Functional Grammar, 47-111 Bloomington: Indiana University Linguistics Club.

Mohanan, Tara. 1988. Causatives in Malayalam. Ms., Stanford University.

Mohanan, Tara. 1994. Argument Structure in Hindi. Stanford, Calif.: CSLI Publications.

Ono, Tsuyoshi. 1992. The grammaticalization of the Japanese verbs oku and shimau. Cognitive Linguistics 3:367-390.

Rosen, Sara T. 1989. Argument Structure and Complex Predicates. Doctoral dissertation, Brandeis University.

Saksena, Anuradha. 1980. Affected agent. Language 56: 812-826.

Saksena, Anuradha. 1982. Topics in the Analysis of Causatives with an Account of Hindi Paradigms. Berkeley & Los Angeles: University of California Press.

Shibatani, Masayoshi. 1976.  The grammar of causative constructions: A conspectus.  In Masayoshi Shibatani, ed., Syntax and Semantics, Vol. 6, The Grammar of Causative Constructions, 1-40.  New York: Academic Press.

Shibatani, Masayoshi. 1978. Nihongo no Bunseki: Seiseibunpoo no Hoohoo [An Analysis of Japanese: The Method of Generative Grammar]. Tokyo: Taishukan.

Shibatani, Masayoshi. 1990. Languages of Japan. Cambridge: Cambridge University Press.

Tonoike, Shigeo. 1978. On the causative constructions in Japanese. In John Hinds and Irwin Howard, eds., Problems in Japanese Syntax and Semantics, 3-29. Tokyo: Kaitakusha.

Zaenen, Annie. 1993. Unaccusativity in Dutch: Integrating syntax and lexical semantics. In James Pustejovsky, ed., Semantics and the Lexicon, 129-161. Dordrecht: Kluwer.

Zaenen, Annie, and Mary Dalrymple. 1996. Les verbes causatifs 'polymorphiques': les predicats complexes en francais. Langages  30, 122:79-95.

Zubizarreta, Maria L. 1987. Levels of Representation in the Lexicon and in the Syntax. Dordrecht: Foris.