downplaying the children. Now the important restriction about intransitives is that Agent defocusing only obtains with agentive subjects as attested by the ungram- maticality of examples in Welsh (14) and Spanish (16), both of which involve a non- active subject: (14)*Tyfwud gan y plant yn sydyn. grown with det children pred suddenly ‘There was growing bythechildren suddenly.’ (Perlmutter and Postal 1984:145) (15) Se bail  oo en grande toda la noche. impers/mid dance-pst.3sg in great all.f det.f night ‘(They) danced a great deal all night.’ (16)*Se creci  oo mucho. impers/mid grow-pst.3sg much ‘(They) grew a lot.’ Yet the question is whether all of these constructions involving an intransitive stem are actually passive. Let us therefore consider sentence (12) from German and sentences (11) and (15) from Latin and Spanish in somewhat more detail: (12) exemplifies the presentative construction and (11) and (15) the impersonal active construction (Cennamo 1993; Maldonado 1999). For one, while these constructions show overlap with the passive in that their Agent is made schematic, no passive meaning is put forward. Indeed, crucial to passive constructions (and passive meaning) is the fact that the change-of-state undergone by a thematic subject is made most prominent, and this is not the case in the examples under discussion. What is put in profile is an action being performed, not its terminal point. Note that impersonal constructions are in clear contrast with medio-passive and spontaneous constructions whose subject can be thematic: (17) Se abrieron las puertas. mid open.pst.3pl det.f.pl doors ‘The doors opened.’ Second, the fact that the passive marker extends to impersonal constructions is in itself not sufficient evidence to view them as passive. The impersonal se pronoun, not the passive, can be translated by the English impersonal they or by the French impersonal pronoun on as in On a danc  ee beaucoup ‘They danced a lot’. Moreover, se impersonal constructions designate generic situations where people Figure 32.5a. Get 840 ricardo maldonado in general tend to do different actions. Again, the focus is on the action, not on the result or on the end point of the event. Notice from (18) that the Agent is sche- matically encoded by the clitic se, and lugares ‘places’ cannot be interpreted as the subject since there is no number agreement with the verb: (18) En lugar-es como ese se crec-e/bail-a bien. in place-pl like dem mid grow-prs.3sg/dance prs.3sg well ‘In places like that one grows/dances well.’ Finally in current Mexican Spanish, it is also possible to have impersonal past constructions referring to specific actions developed by unidentified people, as in the dialogue in (19): (19) Qu  ee tal estuvo la fiesta? ‘How was the party?’ Genial, se bebi  oo y se bail  oo en grande. ‘Great, people drunk and danced a great deal.’ No passive reading is attainable here, and no passive interpretation can be obtained in all dialects of Spanish when the clitic se designates an arbitrary interpretation of the Agent equivalent to ‘one’ or ‘we’, as in (20): (20) No contamos con los recursos necesarios, as  ıı que se hizo lo que se pudo. ‘We did not have the funds needed, so we (one) did what we (one) could.’ These data suggest that the ‘‘Agent-defocused’’ strategy may be a source for passive formation in languages that do not have the option of profiling the Theme, while for languages having the two options, the ‘‘Agent-defocused’’ strategy leads to impersonal constructions that need not be interpreted as passive. 3.2. Oblique Agent So far, we have seen that the various strategies to profile the Theme are crucial to the passive construction. Yet the way the Agent is encoded must also be taken into consideration. The way the oblique Agent is linguistically encoded provides crucial informa- tion about the basic strategies to construct a passive. Figures 32.3 and 32.4 represent two ways in which the prominence of the Agent in the active clause can be Figure 32.5b. Get' grammatical voice in cognitive grammar 841 downplayed in the passive construction. In the ‘‘Theme-in-focus’’ construal, there is always an Agent as driving force—with different degrees of prominence. When the Agent is not expressed, it remains in the base; when it is overtly expressed as an oblique phrase, the Agent’s driving force is downplayed with respect to the main change-of-state designated by the passive. The presence of an oblique phrase is not just the result of mechanical demotion, but it is meaningful. In the case of English, the by-phrase lexicalizes the source of energy bringing about actions (21b) as well as effected objects (21a): (21) a. The sculpture is by Zu ´ n ˜ iga b. Bragging by officers will not be allowed. (examples taken from Langacker 1982: 69) Both sculpture and bragging constitute the trajector of the clause; they only differ in that the former is nominal and the former processual. As the by-phrase may be the source of energy bringing about a new object or an action, it constitutes the natural choice for coding a downplayed Agent. Its input in a passive construction in (22) is the same as in the nominal and the processual examples in ( 21a) and ( 21b): (22) The keynote was delivered by Talmy. This characterization allows us to differentiate passive and stative construc- tions. Consider (23): (23) a. His antics amuse me. b. I am amused at his antics. c. I am amused by his antics. As observed by Langacker (after Postal 1971), (23c) is a passive, but (23b) is not. The latter conforms to the stative, adjectival value of perf 2 . The prepositional phrase at simply designates a location always accessible for contact. The experiencer can either receive some impulse from, or have access to, antics for the sensation of amusement to come about. In contrast, in (23c), the by-phrase designates the source of energy effectuating a change-of-state in the Experiencer, as coded by perf 3 . To the extent that the oblique phrase depicts an initiating cause, a perf 3 construal obtains, resulting in a passive construction. Figure 32.5c. Get'' 842 ricardo maldonado The contrast between initiating source and noninitiative location is of course gradual. Inanimate entities can be construed as agentive if conceptualized as lit- erally extending from an animate Agent (e.g., a stone can be viewed as agentive if it is construed as extending from the stone thrower). As such, the polish Agent/ Instrumental contrast from Słon ´ (2000) is thus common in Romance languages, English, and many other languages: (24)a.Okno zostało stłuczone window.nom become.pst:perf:3sg break.pst part:perf:sg przez kamien ´ . by stone.acc ‘The window was broken by a stone.’ b. Okno zostalo stłuczone window.nom become.pst.perf.3sg break-pst part.perf.sg kamieniem. stone.inst ‘The window was broken with a stone.’ Inanimate elements may become ‘‘sources of energy’’ (and thus assume the properties of an Agent), while the opposite transformation—Agents becoming Instruments—is quite uncommon (Słon ´ 2000). Thus, across languages (24a) tends to be acceptable, but *The window was broken with John tends to be rejected, as can be expected from a general tendency to humans as Agents not as Instruments of some other initiating force. Common to all situations rendered by the passive construction is that the second most prominent participant of its active counterpart is coded as the tra- jector of the event in the passive. This situation is typical of accusative-dominant languages, which basically trace an event using an out > in strategy. In languages using the alternative in > out strategy, we would expect antipassives, which render the reverse type of construal, to be quite productive. This phenomenon is ad- dressed in the following section. Figure 32.6a. by nominal grammatical voice in cognitive grammar 843 4. Antipassives Antipassives are best conceived as the mirror image of passive constructions. While in the passive the Theme becomes most prominent, in antipassives the Theme is downplayed giving the Agent maximal prominence. Antipassives are commonly, but not exclusively, found in ergative-dominant systems, where, as I have suggested previously, the event is naturally traced from the core > out, that is, from the change-of-state undergone by the Theme to its event-initiating driving force. Thus, under normal circumstances the Theme is first accessed to trace the event and as such it receives special prominence. While in accusative-dominant systems the passive construction attributes secondary status to the Agent, in ergative-dominant systems the antipassive construction gives less prominence to the Theme, resulting in decreased accessibility to the energy transmission from the event-initiating driv- ing force and, correspondingly, increased difficulty to fully complete the intended event. In other words, the antipassive indicates that the action is not fully carried out since there is a certain degree of difficulty for the Agent to have the intended effect on the object (Dixon 1980; Cooreman 1994). In the following examples, the degree of affectedness of the object decreases in the antipassive construction: (25) Chamorro a. Un-patek i ga'lago. erg.2sg-kick the dog ‘You kicked the dog.’ b. Mamatek hao gi ga'lago. ap-kick 2sg.abs loc dog ‘You kicked at the dog.’ (Cooreman 1988: 578) (26) Samoan a. Sa 'ai e le tiene le i'a. pst eat erg det girl det fish ‘The girl ate (all of ) the fish.’ b. Sa 'ai le tiene i le i'a. pst eat det girl loc det fish ‘The girl ate some of the fish (lit.: at the fish).’ (Mosel 1989; cited in Cooreman 1994: 61) Figure 32.6b. by processual 844 ricardo maldonado The ergative-absolutive behavior of these examples is evidenced by the change of the ergative marking on the Agent in the transitive samples (23a) and (24a) to its absolutive marking in the antipassive construction in (25b) and (26b). Note that in Samoan, as in many other ergative languages, the absolutive is zero-marked. Alter- natively, in (25b) and (26b), fish and dog are oblique. In languages that do not have an antipassive marker per se, locative, dative, and genitive are common antipassive markers. The inability rendered by antipassive constructions to fully carry the intended effect on the object can be viewed as conceptual distance. While in the active the subject makes contact with, and imposes some change in the object, in the anti- passive, the downplayed object, now an oblique, is at a relative distance from the subject’s action and is not available for contact. This contrast between active direct and antipassive can be represented as in figures 32.1 and 32.7. Since in the antipassive construction the downplayed object is not available for contact, energy transmission is not totally projected onto the object (the dotted arrow), and as such no affectedness is predicated (no squiggly arrow). The notion of ‘‘conceptual distance’’ provides a natural way to account for a variety of meanings normally associated with the antipassive construction. Ex- ample (23b) is the prototypical case of an antipassive whose object is not as affected as the direct object would be in an active-direct construction. In this example, the subject may not have been lucky enough to hit the dog when kicking it, or he or she may have not have affected the dog as much as he or she would have expected. Example (26b) represents another common use of the antipassive: the partitive reading. While the active involves eating the whole fish, the antipassive involves eating only a portion. We can see, then, that conceptual distance may operate at two levels: (i) it may diminish the degree of affectedness of the oblique object, or (ii) the event may not be taken to full completion. The partitive reading in Samoan (26b), in fact, contains both: the complex object is not totally affected, and the event is not fully completed. Yet either reading may occur independently: according to Bittner (1987), in Greenlandic Eskimo the antipassive may depict noncompleted events: (27) Greenlandic Eskimo (Bittner 1987) a. Jaaku-p illu taa-nna sana-pa-a. Jacob-erg house this.sg-abs build-trns.ind-3sg.erg/3sg:abs ‘Jacob built/was building this house (may but need not have finished).’ b. Jaaku illu-mik taa-ssuinnga sana-Ø-pu-p. Jacob house-inst this-sg.inst build-ap-int.ind-3sg:abs ‘Jacob was/is building this house (has not finished yet).’ In other languages, it designates low degree of individuation/identification of the oblique object, such that the Patient is not accessible for the Agent to interact with. Thus, the antipassive designates a low degree of affectedness on the Theme. In Chamorro, the antipassive is obligatory when the object is indefinite (Cooreman 1987); as such, the antipassive in ( 28a) contrasts with the ergative construction in (28b), which contains a definite object (examples taken from Cooreman 1994: 54): grammatical voice in cognitive grammar 845 (28)a.Mgnonne' (guihan) i peskadot. ap.catch (fish) det fisherman ‘The fisherman caught a fish.’ b. Ha -kone' i peskadot i guihan. erg.3sg -catch det fisherman det fish ‘The fisherman caught the fish.’ Furthermore, in Greenlandic Eskimo the antipassive may be employed when the speaker has a nonspecific referent in mind or some specific referent that he or she does not want to specify (see Bittner 1987): (29)a.atuartut ilaat ikiur-tariaqar-pa-ra. of.students one.of.them.abs help-must-trns.ind-1sg.erg/3sg.abs ‘I must help one of the students.’ b. atuartut ilaannik ikiu-i-sariaqar-pu-nga. of.students one.of.them.inst help-ap-must-intr.ind-1sg.abs ‘I must help one of the students.’ (any student will do) The extreme case of low individuation/identification occurs where the object is not expressed at all. The antipassive is commonly used when the object is not identified. In the case of Mam (Mayan), for instance, the antipassive construction does not allow for the unknown or unspecified object to be explicitly encoded (example taken from Cooreman 1994: 53; as cited in England 1988: 533): (30)a.ma Ø- -w -aq'na-7n-a. perf abs.3sg -erg.1sg -work-ds-1sg ‘I worked on it.’ (something) b. ma chin aq'naa-n-a. perf abs.1sg work-ap-1sg ‘I worked.’ (no implication of what was worked on) The lack of an overt object is represented in figure 32.8 by the dotted circle, which means that the object remains implicit in the base; the downplayed object construal is represented in figure 32.7'; here, the object is a regular part of the base. We can see that the object may be left implicit for a wide variety of pragmatic reasons, most notably to designate general tendencies and routine actions. It may also be left implicit when the object is easily recoverable from context or when the speaker wants to keep the object unspecified although it may be definite. One may Figure 32.1'. Active 846 ricardo maldonado be tempted to suggest a grammaticalization continuum where the object first is downgraded to oblique and then is deleted; however, Cooreman (1994) points out that no language in her survey uses the antipassive to downplay the oblique object without also allowing optional or obligatory object deletion. It seems reasonable to suggest, then, that the conceptual path of antipassive object demotion goes from implicit to oblique rather than the other way around: (31) antipassive object demotion: object > implicit > oblique In other words, there seem to be two general pragmatic motivations for the object’s absence. Either it is recoverable from the verb’s meaning or the action is generic. In contrast, when the antipassive downgrades the object to oblique, it does so to des- ignate more specific situations where the object is not totally affected or the event is not totally completed. Given these meanings, we may expect important aspectual correlates. While passives are most commonly associated with perfective aspect (perfect, resultative, and stative), 2 antipassives tend to take imperfective aspect (durative, continuative, repetitive, and habitual) (Tsunoda 1988). Imperfective antipassives tend to depict general tendencies for things to happen as well as habitual and repetitive actions. The habitual is represented by the Tsutsujil (Maya) example (from Dayley 1985: 346), and the repetitive is from Chamorro. Both examples have been taken from Cooreman (1994: 57–58): (32) Tsutsujil Ja nuutee7 b'ara ´ ata nk'ayin wi7. the my.mother cheaply 3sg.abs.sell.ap emph ‘My mother sells cheaply (at low prices).’ (33) Chamorro Mang-galuti gue' ni ga'lago. ap-hit abs obl dog ‘He pounded on/repeatedly hit the dog.’ The passive and the antipassive correspond to construals where the default most prominent participant or trajector is downplayed to profile a landmark. There is an alternative strategy where a secondary participant may gain prominence without downgrading the trajector to oblique. Inverse constructions, as introduced in the next section, accomplish that purpose. Figure 32.7. Antipassive grammatical voice in cognitive grammar 847 5. Inverse Inverse voice involves coding a secondary participant as the most prominent el- ement in an event without downgrading the participant naturally chosen as event trajector. The construction involves the presence of a marker indicating that the default representation of an event has been reversed in favor of a secondary par- ticipant. Crucially, grammatical relations—subject and object—remain unchanged. In Olutec, a Mixe-Zoquean language from Mexico, first person is the default pro- minent element, in that it usually is more prominent than third person; as such, when the default event representation is reversed, with the subject as third person and the object as first, the inverse marker is used: (34) Olutec (Zavala 2003) a. direct 1:3 Tan-tze:k-ku ¨ x-u ja7. erg1.scold-3pl-compl 3pro ‘I scolded them.’ b. inverse 3:1 Ta-tze:k-ku ¨ x-u ¨ -w-a7. abs1-scold-3pl-inv-compl-3pro ‘They scolded me.’ Likewise, in the classical Algonquian voice system (Dahlstrom 1986), the third- person arguments Agent and Patient contrast in terms of topicality: the more topical is case-marked as proximate, and the less topical as obviate. In the active direct clause, the Agent is the proximate and the Patient is the obviate. The inverse is used when there is a topicality switch; that is, the Patient becomes the proximate, and the Agent is the obviate, as in (35b) (from Givo ´ n 1994: 20; citing Dahlstrom 1986), where the young man is more prominent than his father—note that the Agent is still the grammatical subject and that the verb remains active: (35)a.direct Aya.hcinniw-ah nisto e.h-npaha.t awa na.pe.sis Blackfoot-obv there kill/ dir-3/obv this boy/ prox ‘The boy [prox] killed the Blackfoot (men) [obv]’ Figure 32.7'. Antipassive 848 ricardo maldonado b. inverse osa.m e sa.khikot ohta.wiy-ah wa o.skini.kiw much love/inv/obv-3 his/father-obv this young.man/prox ‘(For) his father [obv] loved this young man [prox] too’ In general, the obviate can be said to mark a participant conceptualized as distant from the norm (Langacker 1990: 248): a participant may be physically, socially, or temporally distant from the speaker; or he or she may be distant in the sense of being dissimilar—in terms of individuation, animacy, agentivity, and so forth—from the prototypical representation in a given domain. As such, in (35b), where the Patient is more prominent than the Agent, the inverse construction signals dissimilarity from a prototypical situation where the Patient participant would be less prominent than the Agent. In a similar vein, Givo ´ n(1994: 23) has proposed tentatively that the inverse signals ‘‘a norm reversal vis-a ` -vis the expected relative topicality of event partici- pants,’’ and Cook (1997) has extended that definition of the inverse as a system marking norm reversal to various other domains. A crucial case is that of Samoan, in which the inverse suffix -ina (and its allomorph -a) covers a wide array of domains. One relevant domain is certainly animacy. In particular, entities which score high on the animacy scale are prototypically marked by the ergative. In contrast with humans, then, an animal marked for ergative is distant from the expected norm; thus, it takes the inverse marker, as in (36b): (36)a.Na opo e le tama le tiene. past hug erg the boy the girl ‘The boy hugged the girl.’ b. 'Ua etoeto-ina lona lima e le pusi. perf lick-inv his hand erg the cat ‘The cat licked his [a person’s] hand.’ Inverse marking is determined by an animacy scale where humans score higher than animals. This pattern is also found in Mayan languages and is also expected in other languages with an inverse system. Cook shows, following Chung (1978), that the use of the inverse suffix -ina does not signal a change in grammatical relations. In an inverse construction, the participant marked for ergative remains the trajector of the clause. This can be tested in at least two evident situations. In an equi-construction, it is the ergative Figure 32.8. Antipassive implicit object grammatical voice in cognitive grammar 849 . rendered by the passive construction is that the second most prominent participant of its active counterpart is coded as the tra- jector of the event in the passive. This situation is typical of accusative-dominant languages,. affected the dog as much as he or she would have expected. Example (26b) represents another common use of the antipassive: the partitive reading. While the active involves eating the whole fish, the. proximate, and the less topical as obviate. In the active direct clause, the Agent is the proximate and the Patient is the obviate. The inverse is used when there is a topicality switch; that is, the Patient