CONCEPTUALIZATION OF EVENTS, SEMANTIC RELATIONS BETWEEN CONSTRUCTIONS AND TOPOLOGY: A CATASTROPHE THEORETIC STUDY OF GET AND BE

The existence of syncretisms in the marked coding patterns for variations in the domains of transitivity and voice presumably indicate the interrelation between constructions involving different event views and voice alternations and voicerelated phenomena, such as inchoative and stative resultative morphosyntax. This paper will examine deviations from the prototypical event view, variations in transitivity and their relation to voice distinctions in constructions with be and get. It will be argued that the relations between these constructions and the position they occupy in semantic space are best understood in terms of the interaction between the conceptual status of participants as ‘controller’ or ‘affected’ (Klaiman 1988) and the conceptual property ‘relative elaboration of events’ (Kemmer 1993; 1994), correlating with different event views (Croft 1990; 1994).


INTRODUCTION
From a cognitive perspective, we are concerned with the relation between 'events' in the real world, our cognition of events or 'cognitive constructions', their manifestation in a series of semantico-syntactic forms, and the mental space of the self (Langacker 1991a;Bernárdez 1995).As Bernárdez (1995: 271) notes, there is a succesive mapping (g) simple events are endpoint-oriented: possible verbs consist of the last segment (stative), the second and last segments (inchoative), or the whole three segments (causative).(Croft 1994: 91-92) It is hypothesized that verbs or verbal forms prototypically correspond to one of the three types of event or event views, and that subjects and objects correspond to the starting point and endpoint of the causal chain segment representing the verbal form.Verbs typically select different segments of the tri-partite structure on the basis of the type of event view which is profiled.The causative event view profiles the whole segment (causebecome-state) and corresponds to transitive verbs, the inchoative view profiles only the last two segments (become-state) and corresponds to intransitive verbs, and the stative view, profiling only the last segment (state), corresponds to stative verbs or adjectives.
(1) a. Causative: The rock (x) broke the window (y).(Croft 1990: 53-54) Any event may potentially be conceptualized according to the different event views, yielding both prototypical and non-prototypical associations between event view and event class.In this way, with a dynamic verb of creation, for example, deviations from the causative view, typically associated with this type of event, will result in marked constructions: (2) a.The contractors built the cabin in three months.
b.The cabin got built in three months.c.In three months, the cabin was built.(Croft 1990: 57) Similarly, we find the same marked verbal forms in deviations from the basic stative and inchoative conceptualizations: (3) a. John is sick (thanks to the food here).
b. John got sick (from the food).c.The food made John sick.(Croft 1990: 56) (4) a.He soon recovered from his illness.b.The treatment made him recover very quickly.c.He is now completely recovered.
As can be observed, derived inchoative and stative verb forms are marked by elements also found in marked voice or voice-like alternations.Croft (1994: 102) argues that the three event views correspond to "the active, the middle and the stative/resultative basic voice types", the different sets of terms representing "manifestations of the same ICM of events in different grammatical constructions".(Croft 1994: 102;113) In the causative/active event view the entity chosen as starting point, and thus subject, is construed as controller or immediate cause (not subject to an external cause) of the event, and the endpoint entity is the 'principal locus of the action's effects' (Klaiman 1988).In the inchoative/middle event view, the starting point of the event is some entity which is construed as the locus of the event's effect and also as controller, or at least, where external control is defocused.In English we find certain constructions, 'anticausatives' and 'middles' or 'patient-subject' construction 2 , with unmarked verbal forms which lend themselves to an alternative inchoative/middle construal.In some cases, however, English requires a get-construction.

Event Views
(5) a.The door closed.
b.This book reads easily.c.The document got lost.
As regards derived voice, Croft (1994: 111) notes that in the event passive with be, "the evolution of middle/stative forms towards the passive involves the reintroduction of the agent (in the agentive passive) and the re-transitivization of the event denoted by the verb".Under this analysis, agented passives would represent the reinterpretation of "a stative or inchoative event view as a transitive one".Stative passive forms and resultative constructions are often indistinguishable from adjectival forms coding stative event views.Though stative passive forms are typically agentless, we nevertheless find instances with an obligatory agent phrase (cf.Marín 1997), as in (6c) 3 .( 6

CONCEPTUAL STATUS AND THEMATIC ROLES: CONTROLLER & AFFECTED ENTITY
Accounts of voice alternation have centred mainly on the semantic roles borne by the arguments involved.The characteristic semantic role of active subjects was assumed to be that of agent (Keenan 1976), and grammatical voice was regarded as a system for promoting to subject status an argument bearing a non-agentive role, typically a patient.As agentivity is not a feature of all active subjects, an alternative approach is that of identifying the thematic roles ascribed to subject and object for different verb classes or situation types (Dik 1989), or positing the superordinate categories of Actor and Undergoer 4 macroroles, which subsume the basic roles of subjects and objects respectively (Foley & Van Valin 1984).In a similar vein, Dowty (1991) characterizes the two proto-roles 5 , Proto-Agent and Proto-Patient, in terms of a set of semantic properties defining the prototypical agent and patient roles.Klaiman (1988: 28), however, holds that the organization of (basic 6 ) voice systems crucially involves the conceptual status of sentential arguments as Controller or Affected Entity, and argues for the need to distinguish these from thematic relations and macro thematic concepts such as Actor and Undergoer.
By affected entity is meant the participant perceived as affected or most affected in consequence of the sententially denoted action (=situation).... controller status can be ascribed to the argument whose participation is seen as determining the course and/or outcome of a sententially denoted action.
She points out that while the conceptual status of affected entity ordinarily converges with the undergoer role, in some voice systems it may be "perceived as affected in virtue of performing, not undergoing, certain actions".Conversely, controller conceptual status is typically associated with the actor role, though "it may accrue to undergoers in some systems" (Klaiman 1988: 28).
4. Foley & Van Valin (1984: 29) describe the Actor as the "argument of a predicate which expresses the participant which performs, effects, instigates or controls the situation denoted by the predicate", whereas the Undergoer "expresses the participant which does not perform, initiate or control any situation but rather is affected by it in some way".
5. The features that characterize these role types, according to Dowty (1991: 572) Klaiman (1988: 71) defines basic voice as "a system for marking identity or nonidentity of the sentential subject with the conceptual status of affected entity.This distinction is encoded on sentential verbs through nonderivational morphological alternations in verbal bases, i.e., diathesis".Derived voice, on the other hand, refers to processes, like passive and antipassive, whose verbal forms take derivational markers.
In this way, Klaiman (1988: 36) views the system of Indo-European voice as "fundamentally a device for marking a sentential verb according to whether its Subject is or is not an affected entity, i.e., whether it is or is not the principal locus of the effects of the sententially denoted action".The middle voice in IE embraced a series of functions where the principal effects of the action accrue to the actor itself (Klaiman 1988: 31-36;Smyth 1956): • Plain/full Middle: conveys the dual status of the subject, as performer of the action and as affected entity.• Reflexive: subject effects the action which affects self.
• Reciprocal: referents of a plural subject do action to one another.• 'Nucleonic' or Indirect Reflexive: subject acts for him/herself or with reference to self; object or affected entity is in, moves into, or moves from subject's sphere.• Deponent: action involves the bodily or physical disposition, emotive or mental disposition of the subject.• 'Passive' Middle: subject is purely an affected entity, while playing no role in effecting the action .• 'Catalytic' Passive (Barber 1975): subject catalyzes the action performed by a separate actor, and, at the same time, it is perceived as the affected entity.Barber (1975) notes that, in English voice, relations are grouped according to an active/passive verbal system, in which the active subsumes all those cases where the subject effects the action or is somehow in control (active, reflexive, reciprocal, full middle) in contrast with the passive where the subject does not effect the action.In English we also find a 'catalytic' passive or get-passive, where the subject corresponds to an affected entity and simultaneously brings the action onto itself, which accounts for the ease with which a reflexive pronoun can be incorporated to the get-passive construction.Barber (1975: 23) argues that the rise of the get-passive construction represents an extension of voice relationships and "provides a logical route by which an active/passive system could regroup to become an active/middle system"7 .Langacker (1991b: 229) characterizes the difference between the two voices, active and passive, in terms of asymmetrical relations and of the order in which conceived entities are characteristically accessed in cognitive processing.In the active, where the relational figure is the most energetic participant, the coalignment of two asymmetries is achieved: "the ranking of participants in terms of their subjective prominence; and directionality in the objectively construed flow of energy".In the passive, on the other hand, "the most salient participant lies downstream in the energy flow.The resulting conflict in alignment is what makes the passive a marked construction; the profiled process receives an unnatural construal, being accessed through a focused participant representing the terminus (rather than the origin) relative to its inherent directionality".

TRANSITIVITY, VOICE & THE ELABORATION OF EVENTS
Transitivity, according to Hopper & Thompson (1980: 253), is best characterized as a complex scalar notion derived from the presence or absence of a series of parameters 8 or components, which "allow clauses to be characterized as MORE or LESS transitive".These components basically refer to the effectiveness and intensity with which the action is carried over or transferred from one participant to another, typically from an agent to a patient.Rice (1987: 78) claims (contra Hopper & Thompson 1980) that the factors involved in transitivity do not exist as independent, objective features, derived from the grammatical properties of the clause and argues for "the need to ground explicitly the transitive event prototype in a schematized event", so that although "many of these factors may be missing, yet the overall event may still be considered transitive".In this 8. Hopper & Thompson (1980: 252) identify the following parameters of transitivity: Participants, Kinesis, Aspect, Punctuality, Volitionality, Affirmation, Mode, Agency, Affectedness & Individuation.9. Rice (1987: 145) proposes a series of components for the two poles of the transitivity continuum: proximity/distance direction location force-dynamic configurational external reaction internal reaction speed or force plus movement pure motion change of location serial position interaction between co-animates action within a setting goal-oriented source-oriented directed approach outward extension independence of participants contingence of participants asymmetrical participants symmetrical participants maximal differentiation of participants minimal differentiation perfective action imperfective situation communicative effect non-communicative effect non-spatial cognitive domain spatial cognitive domain respect, Rice (1987: 156) identifies a series of transitivity components 9 , which refer to the various facets of the prototype transitive event.
The prototypical transitive event is one in which two asymmetrically-related entities are involved in some unilateral activity.The activity requires forceful movement or some energized transfer instigated by one entity resulting in either contact with or some observable effect in the other.
The fact that various events deviating in terms of transmission of energy are assimilated to the transitive marking pattern suggests that "languages typically make reference to a schema that is in some respects more general than the prototype concept" (Kemmer 1994: 191), characterized in terms of a set of parameters of the clause or components of the event.The 'two-participant event schema' for the transitive situation type consists of two participants, Initiator and Endpoint of the event, and an asymmetrical relation between them construed as being directed from Initiator to Endpoint.

A ---------> B Initiator Asymmetrical Relation Endpoint
In the reflexive situation type, the Initiator acts on itself as Endpoint, but the type of event involved is one in which participants are normally distinct entities.In the case of middle situation types 10 , on the other hand, the Initiator and Endpoint refer the same entity with no conceptually distinguished aspects.Finally, in the intransitive situation type, there is only one participant with no conceptual distinction between the Initiating and Endpoint facets.Kemmer (1994: 209) thus proposes the following 'Semantic Transitivity Continuum', in terms of the relative distance from the two active prototype situation types (transitiveintransitive), as a function of the semantic parameter, degree of distinguishability of 10.Kemmer (1993)  participants, or "the degree to which a single physico-mental entity is conceptually distinguished into separate participants" (Kemmer 1994: 206).

Degree of Distinguishability of Participants
She claims that the semantic property of degree of distinguishability of participants subsumes the notion of 'subject-affectedness' and allows for the distinction between reflexive and middle in terms of "progressively lower distinguishability".Also, since "the Initiator (controller or conceived source of action) and Endpoint (affected participant) are not separate, but necessarily the same entity", the Initiator is at the same time affected (Kemmer 1994: 209).
In the case of reciprocal events, the relevant property is "distinguishability of the component subevents in a reciprocal action or state" (Kemmer 1994: 210).In ordinary reciprocals the component subevents are viewed as distinct, whereas in naturally reciprocal events, the component events are not separated out, but rather the event is viewed as a 'single unitary event'.Languages without middle systems, like English, use 'heavy marking' in ordinary reciprocals and in the reflexive situation, and unmarked intransitives ('zero marking') for naturally reciprocal events and for the middle situation types, or else make use of certain constructions akin to the passive or expressing changeof-state.
(7) a.They hate each other.
b.They embraced.c.He saw himself in the mirror.d.She dressed (herself)/got dressed.Kemmer (1994:211) views 'distinguishability of participants' and 'distinguishability of events' as two aspects of the more general conceptual dimension 'relative elaboration of events', which "can be thought of as the degree to which different schematic aspects of a situation are separated out and viewed as distinct by the speaker".
The semantic transitivity continuum provides an explanation of the marking patterns in domain of voice.In passive events, for example, the Initiator or Agent participant is defocused.Similarly, in spontaneous processes, the single participant coded is construed as the Initiator and also as the Endpoint, since it undergoes some change of state as well.These situation types thus resemble one-participant events, and as such, their linguistic coding will approximate that of prototypical one-participant events.

SYNCRETISM IN CODING: THE CASE OF 'GET' AND 'BE'
We have observed that deviations from the prototypical event view give rise to marked constructions, involving causativizing, inchoative and stative resultative morphosyntax.We have also observed the existence of a cline in transitivity along which reflexive, passive and middle situation types are located.We will now discuss the constructions with get and be in English, and their distribution in semantic space.

Constructions with Get
Get-constructions display certain aspectual and modal facets.Get in causative and causative-locative constructions (Givon & Yang 1994) evokes deontic modality, the expression of causation falling somewhere in the middle between direct ('make') and mediated ('cause') cause (Cameron 1990).There is also a correlation between passive and causative in 'indirect passive constructions' (Shibatani 1985).In the get-permissive, "the target of permission is the gramatical subject of the sentence", and "the source of permission is intangible" (Cameron 1990:158).
(8) a.We got him to sign the papers.
b.He got her into the house.c.He got his radio confiscated by the police.d.How come he always gets to go, and I don't?
The aspectual features of 'mutation', 'change' or 'inception' are most obvious with get as a copula (Cameron 1990).The meaning expressed in this case is indistinguishable from the inchoative view of basic causatives and statives.
b.This room gets extremely hot.
We also find reflexive-causative or 'complex reflexive passives', reciprocal, and some middle situation types ('grooming or body care') coded with get in English (Givon & Yang 1994;Collins 1996).
The category identified by Collins (1996) as 'psychological get-passives', which also includes certain non-psychological verbs, is best subsumed under middle situation types such as the emotion middle or the spontaneous middle.This might also be argued for the category of 'formulaic get-passives', which includes idiomatic expressions such as get used to, get fed up with, get accustomed to.
(11) a. ... people she had met and grown to know and got excited by and even loved ... (A12.GL1.) b.She was getting tired.c.Why don't we get rid of her? (A15.GF1.)Bolinger's (1978) semantic characterization of the passive in terms of the transitivity feature 'patient-affectedness' appears to be crucial for get-passives.As Cameron (1996: 151) points out, "get-passives are more transitive semantically than be-passives.Not only must the patient be affected, it must be materially affected".
(12) a. Kathleen got fired from several jobs.b. ?Odette got loved by all.Downing (1996: 189-190), however, presents corpus-based evidence of the existence of get-passives with mental processes of perception, affection and cognition, as well as with certain relational processes.
(13) a. ... people who don't get loved or taken care of.
b.The stocks get over-owned by Americans.
It has frequently been observed in the literature that the passive with get implies initiative 11 and/or partial responsibility of the subject (Hatcher 1949;Lakoff 1971;Barber 1975), or causal involvement of the subject in the event (Chappel 1980).Hatcher (1949: 444) holds that "the original meaning of get as passive auxiliary is not that of 11.According to Collins (1996: 50-51), the agentive role of the subject-referent is evidenced in its behaviour in certain constructions: • Constructions such as try to/and get, go and get, and manage to get imply an agentive role of the subject-referent and typically favour get over the passive with be.• In verb plus non-finite complement constructions, get is disallowed if the subject-referent has no control over the process.• In imperatives, get is more likely to occur than be.
(i) a.He managed to get/??to be transferred to the combat forces.b. Mary was heard/*got heard to insult her parents.c. "Get/*Be stuffed", answered Witcharde.(Collins 1996: 51) Givon (1990: 621) also points out that the initiator role of the subject is reflected in the following features: • The possibility of embedding of the get-passive under verbs of manipulation.
• In co-occurrence with adverbs indicating the presence of a volitional entity, volitionality is attributed to the subject-referent.• With adverbs indicating non-volitional action, the be-passive is favoured.
submitting to an agent, but of using an agent".This attribution of responsibility to the subject favours the choice of referents with high inherent topicality as subjects (Marín 1992(Marín , 1996)).( 14) a.Well, I think I'll go get examined.
b. Have you gotten vaccinated yet?(Hatcher 1949: 436) Cameron (1996: 156), however, notes that "the source of the 'responsibility' overtones in the get-passive is probably more the resonance from the causative get than anything in the inherent semantics of the get-passive".In this respect, it is interesting to note that in co-occurrence with deontic modals, obligation is attributed to the the subject in the get-passive, thus enhancing subject responsibility reading, whereas in the bepassive obligation is attributed to the agent (Lakoff 1971).
(15) a. Radicals must get arrested to prove their machismo.
b. Radicals must be arrested if we are to keep the Commies from overrunning the U.S. (Lakoff 1971: 156) Another interesting feature of the get-passive pointed out by Haegeman (1985: 55), which is probably related to causative resonance, is the possibility of double passivization.This usage, however, is considered archaic.
(16) The blaze of the insurrection is got damped down.
Get-passives are also associated with 'adversative' of 'beneficial' effects of the action on the subject-referent (Hatcher 1949;Chappel 1980) or the speaker's attitute towards the event and its evaluation as 'good or bad' (Lakoff 1971).
(17) a. John got promoted instead of me, ... b. ... he coulda got mugged and killed ... (A12.G1) The original emphasis on the initiative of the subject has extended to cases of 'generic', theoretical responsibility, especially in events of an adversative nature.The (ii) a.They told him to get/*to be fired.
b. John got treated by a bad doctor deliberately.
(18) a. You'll get yourself killed one of these days.b.He got killed in a car accident.
Attenuation in subject control (from volitional participation to non-volitional responsibility to locus of experience in the case of animate subjects) may further extend to cases where "the experiencer is no longer the subject per se, but rather an individual associated with the subject" and even to cases where "the locus of the experience need not be overtly specified or clearly delimited" (Langacker 1997: 21-22).
(19) a. Edith wrote an editorial which never got printed in the Bugle, (A17.G1.) b.After that the camera got mislaid.(A3.G2.) The role of the agent in the get-passive appears to be subordinated, that is, "the agent does not completely dominate the situation" (Hatcher 1949: 436).This probably accounts for the fact that the expression of an agent phrase is rare, particularly if it refers to a human (volitional) entity.Agents, when expressed, tend to be inanimate and nonidentifiable; human agents with a low degree of individuation are also found.The aspectual feature of 'mutation' and the fact that primary responsibility for the event is not attributed to an external (implicit) agent makes the get-passive construction similar to the anticausative in some cases (Marí 1992).In Spanish, both interpretations are readily available with the se marker.(Zandvoort & van Ek 1975: 57;Marín 1992: 220) In this vein Downing (1996: 194) suggests that "the get-passive makes available a typically one-participant causative construction for use with those verbs which do not admit the ergative".The parallel between the two constructions lies in the fact that both focus on "the Medium as the element crucially involved in the process".

Constructions with Be
The basic stative event view is found in be-constructions, taking various types of relational expressions as complements, and in existential be (Déchaine 1995;Langacker 1997).
b. Mary is the teacher.c.Peter is in the kitchen.d.God is.Langacker (1991b) observes that three semantic variants of the perfect participial morpheme 12 may be identified in be+PP constructions.The first type [PERF1] is formed on intransitives and derives stative participles.The resulting be+adjectival participle construction corresponds to so-called subjective resultative constructions and resultatives of translational motion events (Nedjalkov & Jaxontov 1988), and is indistinguishable from the stative view of basic inchoatives.The second type [PERF2] also derives stative participles, but involves transitive verbs and additionally effects a figure/ground reversal, in that the landmark of the verbal stem becomes the trajector of the adjectival participle.This type is found in so-called objective resultative constructions, which are indistinguishable from the stative view of basic causatives.
(23) a.One foot was badly swollen, his flanks were red and inflamed by evening.
(A18.R2-3) b.Now that the girls were gone with their young attendants, (A4.R1.) c.Even if America's burned -even if it's destroyed -(A14.R15-16) The variant [PERF3] also effects a figure/ground reversal, but derives a processual participle.This is the type which appears in event passives (agented and agentless), formed both from perfective predicates, and in stative passives, formed from imperfective predicates.
12. Langacker (1991a:203) describes the family resemblance between the three variants in the following terms: PERF1 and PERF2 are alike in restricting the profile to a single state that results from a participant undergoing a change, while PERF2 and PERF3 are alike in elevating to trajector status the participant corresponding to the standard's primary landmark.An additional similarity is that all three variants enhance the salience of a "downstream" element [...].In the case of PERF1 and PERF2, this element is the final state, which is downstream from the others with respect to the flow of time, and surpasses them in salience by virtue of being profiled.With PERF2 and PERF3, the element in question is a 'terminal participant', i.e. one that lies downstream from another with respect to the flow of energy (or some analog thereof), and it is prominent by virtue of being made the relational figure (trajector).
(24) a.I was invited by Harry's doctor, Shafik.(A5.11.) b.Shall champagne be served?(A6.2.) c.The church and the churchyard were hidden by trees.(A1.S12.)Langacker (1997: 20) notes that in the case of the be-passive, the profiled relationship is the one expressed by the participle, and "the participial subject is ipso facto the subject of the entire construction", or passive undergoer; we thus have a situation of full transparency where the auxiliary be simply adds sequential viewing to its complement.As for the get-passive, "the evolution of get involves progressive attenuation in both the nature of the profiled relationship and the degree of control exercised by its subject.The result is that get comes close to being just a passive auxiliary, like be, serving only to provide the sequential viewing required for the head of a finite clause".However, "get-passives are non-transparent 13 in that the subject must in some way be implicated in the experiential relationship, not necessarily as the locus of experience, but maybe just by providing a link to the implicit experiencer" (Langacker 1997: 22).

CONCEPTUALIZATION OF EVENTS: TOWARDS A CATASTROPHE-THEORETIC ANALYSIS
In this paper we will attempt to characterize the relations between the above constructions in topological terms as "spaces characterized by gradual, continuous dynamics interrupted by sudden discontinuities (catastrophes)" (Bernárdez 1995: 273).As Bernárdez (1995: 268)

notes:
A CATASTROPHIC PROCESS is understood as the transition from a structurally stable state to another equally stable state: from a FORM to another.A non-catastrophic process, on the other hand, is the continuity of a stable state through time (and space).A stable state is dynamic, i.e. it is modified with time, so that it is not exactly 'the same' at two different, successive moments of time.
Our conceptualization of events may be understood in terms of two stable states, a causative-transitive state and a stative-intransitive state, with a breach of stability or catastrophic transition from one to another represented by an area of change of state or inchoativeness.
The representation of the features of space involved and the dynamics of qualitative changes may be carried out by means of a topological map, in this case corresponding to the CUSP catastrophe (Figure 1), which is characterized by the presence of two stable states with a transitional area between them (Wildgen 1982;1994).Three parameters may be identified: two governing the horizontal axes ('external' or 'control' parameters), and one governing the vertical axis ('behaviour') (Bernárdez 1995: 269).These cognitive parameters regulate changes of meaning-form as a function of the variations in the values adscribed to them: (i) Parameter 1: First horizontal dimension, 2-1 participant event, representing the degree of transitivity of the event in terms of the 'degree of distinguishability of participants' (Kemmer 1994), according to which we find the following continua: This parameter represents a cline describing gradualness from events where there are two participants in an asymmetrical relation involving transmission of force, to oneparticipant events where there are no initiating or endpoint facets.
(ii) Parameter 2: Second horizontal dimension, Controller-Affected Entity, representing voice distinctions in terms of the conceptual status of the participant coded as subject in the event as controller or affected entity.

G-CAUSATIVE
This parameter represents a cline ranging from greater to lesser participation of the Initiator participant, and correspondingly from less to greater foregrounding of the Endpoint entity as affected entity.
There is an interaction between these parameters such that an increase or decrease in the values of the first two parameters will bring about changes in values of the third vertical dimension.At one extreme, in the causative-transitive space, the degree of distinguishability of participants and of controller status of the subject is highest, whereas the stative-intransitive space is characterized by the lowest values in these parameters.
(iii) Parameter 3: Vertical dimension, Cause-Change-State, which correlates naturally with the previous values, and which refers to the internal structure of events and the type of event view which is profiled, involving prototypical and non-prototypical associations between event class and event view (Croft 1990): This characterization approximates a 'natural' explanation for the gradualness and continuity in the different spaces and the different linguistic forms found in the various subspaces (Bernárdez 1995).As we may observe, there is a clear meaning-form transition from the semantic space involving the regions of reciprocal, reflexive, middle, inchoative and passive get, to the semantic space of the event passive, and to the space of the stative passive, stative and resultative constructions with be.
As regards the extensional structures of these constructions, although the issue is beyond the scope of this paper, it is interesting to note that the various extensional sequences posited in the literature for the be-passive and the get-passive 14 may be accounted for in terms of the topological representation in Fig. 1.

CONCLUSION
In this paper we have observed a relation between a series of constructions with get and be, coding a variety of instances of deviation from the prototype (prototypical event view, transitive prototype, unmarked voice).
We have discussed a series of parameters which are crucially involved in voice distinctions (the status of the subject-referent as controller or affected entity), and in the transitivity continuum (the degree of distinguishability of participants in an event), and identified their correlation with the various event views (causative, inchoative, stative).Finally, we have proposed a characterization of the interaction between these parameters through a catastrophe-theoretic study, which appears to account the gradualness and continuity of the different constructions found in the various semantic spaces as well as for the extensional pathways posited in the literature.
We may conclude that the cognitive dimensions which subsume the various parameters of voice and transitivity are intimately linked to our conceptualization of events, as the existence of syncretisms in the marked coding patterns for these domains seems to indicate.
) a. Kleopatra was swept by a red surge of anger.(A4.11) b.Too much was being imposed on him.(A5.1) c.About 70% of the world's continental shelves are now covered by relict sediments.(B6.S22) d.It was soaked under the arms with dark stains of anxious perspiration ... (A1.R13) e. but by this time her eyes were quite shut.(A3.R9)

(
20) a. */?The woman got run over by John.b. */?The window got broken by John.c.The woman got run over by a drunken driver.d.The woman got run over by a car.

(
21) a.Another of our new cups got broken last night.b.Another of our new cups broke last night.c.Otra de nuestras tazas nuevas se rompió anoche.