This document discusses how metaphors can help explain patterns of polysemy, or multiple meanings, among semantic roles, such as cases and prepositions that encode semantic roles. It argues that certain common metaphors, like the "Companion Metaphor" can account for frequent polysemies between roles like Comitative and Instrument. Taking space as a basic cognitive domain, metaphors provide links between domains that help explain the positioning of semantic roles on semantic maps. While the focus is on metaphors, the document also notes that other processes like analogy and metonymy can also account for some polysemies.
This document discusses how metaphors can help explain patterns of polysemy, or multiple meanings, among semantic roles, such as cases and prepositions that encode semantic roles. It argues that certain common metaphors, like the "Companion Metaphor" can account for frequent polysemies between roles like Comitative and Instrument. Taking space as a basic cognitive domain, metaphors provide links between domains that help explain the positioning of semantic roles on semantic maps. While the focus is on metaphors, the document also notes that other processes like analogy and metonymy can also account for some polysemies.
This document discusses how metaphors can help explain patterns of polysemy, or multiple meanings, among semantic roles, such as cases and prepositions that encode semantic roles. It argues that certain common metaphors, like the "Companion Metaphor" can account for frequent polysemies between roles like Comitative and Instrument. Taking space as a basic cognitive domain, metaphors provide links between domains that help explain the positioning of semantic roles on semantic maps. While the focus is on metaphors, the document also notes that other processes like analogy and metonymy can also account for some polysemies.
This document discusses how metaphors can help explain patterns of polysemy, or multiple meanings, among semantic roles, such as cases and prepositions that encode semantic roles. It argues that certain common metaphors, like the "Companion Metaphor" can account for frequent polysemies between roles like Comitative and Instrument. Taking space as a basic cognitive domain, metaphors provide links between domains that help explain the positioning of semantic roles on semantic maps. While the focus is on metaphors, the document also notes that other processes like analogy and metonymy can also account for some polysemies.
Plotting diachronic semantic maps: the role of metaphors
Silvia Luraghi - University of Pavia
Abstract
The paper discusses patterns of polysemy among semantic roles in cross-linguistic perspective, and the possible way in which such polysemies come into being through meaning extension of morphemes (cases and adposition) that encode semantic roles. Semantic extension is described based on commonly assumed metaphors, which are thought to account for frequent polysemies, such as the Companion Metaphor in the case of the Comitative-Instrument polysemy. Taking space as a basic domain, metaphors are shown to provide links among different cognitive domains, which account for the position of semantic roles relative to one another in the ensuing semantic map. Such map incorporates diachronic information, assuming that metaphorical extension reflects diachronic developments. Although the specific focus of the paper concerns the contribution of metaphors to semantic extension, it is also shown that other processes are involved, such as analogy and various kinds of metonymy, which can account for some specific patterns of polysemy among semantic roles.
Key words: Metaphor, semantic maps, polysemy, meaning extension
2
1. Introduction 1
This paper describes paths of semantic extension undergone by morphemes (cases and adpositions) commonly used for coding semantic roles, based on Lakoff and Johnsons (1980) theory of metaphor. My specific aim is to show how some metaphors, commonly held to explain certain patterns of polysemy, can provide directions for plotting a semantic map that incorporates diachronic information. In doing so, I also offer an explanation regarding cross-linguistically frequent and infrequent polysemies, shedding light on the structure of the conceptual space which serves as the basis for language specific semantic maps. By highlighting metaphorical patterns that provide connections among different domains, I suggest that different sources for the same semantic role indicate different conceptualizations that might account for unexpected polysemies. I also argue that absence of polysemy may be relevant, especially when it frequently holds among semantic roles which are often related to neighboring concepts. In addition, tracing the metaphor responsible for a specific polysemy sheds light on the causes of merger among semantic roles that should in principle be kept distinct, as they are differently located in the causal chain of events (Croft 1991). While this paper specifically concentrates on the role of metaphors in semantic extension and in the development of polysemy, it must be pointed out that not all polysemies are brought about by metaphors. Different types of metonymy (referential and predicational) also have a role, as I show especially with reference to the polysemy of beneficiary and recipient and of instrument and agent. In addition, I argue that semantic extensions may be based on analogical processes not necessarily resulting in metaphors, caused by analogical extension of partial schematicity (see sec. 2.3.1). By
1 I thank Kathryn Allan, Spike Gildea, Bernd Heine and Heiko Narrog for helpful comments on the content and style of this paper. In addition, I am grateful to Tuomas Huumo, Fabrizio Pennacchietti and Manana Topadze, who provided some of the examples. 3
singling out semantic extensions based on different processes one can avoid the risk of overusing the analysis, and set up unwarranted metaphors. As is generally accepted, I assume the domain of spatial relations to be the basic conceptual domain. 2 I argue that spatial relations provide a source for the encoding of semantic roles typical of human beings, but that they are not the only source. Human relations can also provide the source for the domain for metaphorical extension to other human relations. Crucially, patterns of polysemy seem to be conditioned by the source domain: polysemies that do not arise, or are extremely infrequent, when space functions as the source domain, seem to be more easily brought about when the source domain is at the level of human relations. The understanding of the course taken by diachronic developments can also shed light on the frequently neglected issue of possible semantic extensions which result in polysemy vs. those which result in changes, whereby the original meaning is not preserved. Examples for both tendencies are provided by comitative coding. When comitative markers extend to instrument based on the Companion Metaphor, they most often also continue to encode comitative (examples are available from several European languages, cf. Stolz et al. 2006). On the other hand, reported instances of extension from locative to comitative (see Luraghi 2001b and below, sec. 2.2.3) apparently result in the loss of the original meaning by the marker that undergoes the change. The latter type of development might have consequences on what we conceive as frequent vs. infrequent syncretism, as well as on our understanding of (uni)directionality: since diachronic data are often unavailable and can hardly be reconstructed for a great number of languages, what looks like an infrequent semantic extension may in reality only be an infrequent type of synchronic polysemy (see further Narrog 2010a). Similarly, seeming unidirectionality of possible semantic
2 As noted in Croft (1991: 192) locative and directional markers are probably the ultimate historical source of most case marking. Localistic theories have a long history, reaching back to late antiquity; see Hjelmslev (1935) and Anderson (1971) for a comprehensive discussion. 4
extension can be an effect of some meanings being lost while others are not. Such changes also raise problems for a theory of gradualness in semantic change, as I argue in sec. 4. The paper is organized as follows. In section 2 I discuss conceptual domains to which different semantic roles belong. Following Heine et al. (1991), I distinguish between the basic domain of space, the domain of human relations and the domain of non-human and non-spatial relations. After discussing basic spatial relations (2.1), I show how space is mapped onto the domain of human relations (sec. 2.2). I then proceed to non-human relations (sec. 2.3). I argue that it is often the domain of space rather than the putatively intermediate domain of human relations which provides the source for the conceptualization of non-human relations. In sec. 3 I discuss expected and unexpected patterns of polysemy among human relations. Sec. 4 contains a general discussion regarding the role of metaphors in semantic extension, polysemy, and lack thereof. Metaphors described in the paper as responsible for semantic extension and polysemy among semantic roles are listed in Appendix (i) for convenience. 3
2. Encoding semantic roles: source and target domains
Space is generally assumed to offer the source for conceptualization of other, more abstract domains (see Ptz 1996: xi; slightly different views are discussed in Zlatev 2007: 319). According to Langacker (1987), space is a basic domain, as it emerges directly from experience (Lakoff & Johnson 1980: ch. 12; see further Croft 2003). More abstract relations, in particular those pertaining to the domain of causation, are held to be conceptualized as metaphorically derived from spatial relations. Space, then, provides the source domain for the target domain of causation, and this
3 All examples from secondary literature have been checked with informants; I have revised and unified the glosses, unless differently specified. 5
explains polysemy of specific coding devices. According to Croft (1991: 193): [t]he most important relation between causative and locative-directional roles is that of DOMAIN SHIFT or METAPHOR. Heine et al. (1991) claim that semantic extension proceeds from the conceptual domain of space to other domains, following the Chain of Increasing Grammaticalization (reproduced in Fig. 1 for convenience):
FIGURE 1. CHAIN OF INCREASING GRAMMATICALIZATION (from Heine et al. 1991: 159)
This scale presents a number of problems, extensively discussed by Narrog (this volume). Here, I focus on the more general implication contained in the scale as to the direction of semantic extension, that is, that spatial roles constitute the basic source domain for all other relations, and that, among the latter, human relations, i.e. roles typical of human participants, precede inanimate ones as the first stage of semantic extension. This is represented in (1), from Heine et al. (1991: 160):
(1) spatial relation > human relation > inanimate relation
In the next sub-sections, I survey the groups of relations in Fig. 1 in the light of the scale in (1).
Spatial relations are usually described as involving an asymmetrical relation between a figure and a ground, or a trajector and a landmark. Different specific relations between the two can potentially be infinite; in practice, only a small number of spatial relations, including location, direction, source and path, are unanimously regarded as basic, as I discuss below.
2.1.1 Basic spatial relations and the position of path In the light of cross-linguistic coding tendencies, direction (allative), 4 source/origin (ablative), and (static) location (locative) seem to be more basic than path among spatial relations. As argued in Stolz (1992: 30), there is a tendency for case marking relative to spatial relations to exhibit Dreigliedrigkeit, i.e. a tripartite structure featuring dedicated coding devices for location, direction and source. Indeed, path can often be coded through cases/adpositions that usually indicate location, as in English:
(2) Mary walks in the field. / The child is running in the street.
Tyler and Evans (2003: 257) argue that typical landmarks of the English path preposition through are characterized by the spatio-geometric property of having four sides ... and hence possessing an interior and an exterior. Thus, typical landmarks of through are conceived as containers, in very much the same way as typical landmarks of in (see e.g. Vandeloise 1994). It is outside the scope of this paper to provide a detailed cross-linguistic survey of possible coding of path as location. However, it can be observed that path and location share the same conceptual space between source and direction. This is quite straightforward for path, as shown in
4 I prefer the label direction for the allative relation, rater than the label goal due to the polysemy of the latter term, used to indicate patient in certain approaches (e.g. in S.C. Diks Functional Grammar, see Dik 1997), a usage that goes back at least to Bloomfield (1933). 7
Fig. 2, in which I give a schematic representation of a motion event. As shown in Fig. 2, the trajectory that leads from the source to the goal (direction) is located in the span that connects the two endpoints; in dynamic terms, the trajector moves through the landmark. However, the same situation can also be conceived in static terms, not focusing on the trajectors movement from the source to the goal, but rather focusing on the space occupied by the trajectory. Under this perspective, path is conceived in terms of location, that is, the portion of space in which a trajectory is located.
source ----> path ----> direction Tr trajectory location Lm
FIGURE 2. A PROTOTYPICAL MOTION EVENT
Note further that path does not include the starting point and the end point of the trajectory. Path is basically atelic, a trajector moving along a path is performing an ongoing activity, and is not implied in an event which includes a change of state/position (there is no change of position inasmuch as the end result of the motion activity is not in the scope of this semantic role). On the other hand, a trajector reaching the goal of a directional motion or one setting out from a source undergo telic events that imply change of position. The fact of being atelic is another feature that path shares with location, which is stative and thus obviously atelic. The conceptual space of spatial relations can thus be represented as in Fig. 3.
8
source location direction
path
FIGURE 3. THE CONCEPTUAL SPACE OF SPATIAL RELATIONS
2.1.2 A merger without polysemy: location and source Above, I remarked that many case systems display a tripartite nature. This, however, is only part of the story: as indicated by several studies carried out on large language samples, patterns of polysemy among the three basic spatial relations are not random, as languages which code the three relations with only two coding devices point toward easy merging of direction with location in contrast to the extreme rarity of source merging with either location or direction. In other words, source tends to be maximally distinct from other spatial relations. Pantcheva (2010) compares her and others results, as shown in Table 1:
TABLE 1: PATTERN OF SYNCRETISM FOR THE LEXICALIZATION OF LOCATION, GOAL, AND SOURCE (adapted from Pantcheva 2010: 1046; L=location, G=goal [direction], S=source)
The figures in Table 1 point toward a special status of source with respect to the other roles, and to a comparatively frequent merger of location and direction. The loss of a tripartite structure in 9
favor of a bipartite one, in which location and direction are expressed in the same way, is also historically attested: for example, loss of the case system in Late Latin eliminated the distinction between in+ABL (location) and in+ACC (direction). 5 The Romance languages evolved in two directions: some, as e.g. French and Italian, no longer have distinct coding for the two relations and exhibit the pattern L=GS; others, as for example Spanish, restored the distinction by specializing different prepositions for the coding of either relation (en = location, a = direction; see further Luraghi 2011). Pantcheva further notes in passing that she left out of her account the so-called ablative- locative transfer, the historical process where a (originally) source marker starts being used as a locative marker, and remarks that in the framework outlined in her paper this is quite unexpected. Indeed, this type of extension, to which little attention has been paid, seems to be frequent cross- linguistically. In the only extended study of this type of development, Mackenzie (1978) surveys evidence from several branches of the Indo-European language phylum, Israeli Hebrew, and two Austronesian languages, Fijian and Sonsorol-Tobi. Among numerous examples, consider the adverbs in Table 2:
5 Indeed, the three-way distinction was limited to a small number of prepositions that took two cases in Latin; with the majority of prepositions, location was not distinct from direction (i.e. most prepositions could indicate both location and direction without case variation of the governed NP). In addition, a lexically restricted usage of cases without prepositions displayed both a tripartite structure (with a locative, an ablative, and an allative, this last function being fulfilled by the accusative) and a bipartite structure sporting the rare source/ location polysemy. Tri- and bipartite structure depended on inflectional classes of specific nouns; see Leumann, Hofmann & Szantyr (1965) and Luraghi (2009, 2010a) for details. 10
TABLE 2. FRENCH PLACE ADVERBS
The French adverbs above can occur, without further adjunction of primary prepositions, in location or direction expressions. If they occur in source expressions, though, the preposition de from must be added:
(3) Il est/va dehors vs. Il vient de dehors he is/goes outside he comes from outside He is/goes outside. / He comes from outside.
Etymologically, these adverbs already contain the preposition de and consequently their original meaning contained the notion of source. After acquiring the locative meaning, they need to be reinforced again by de in order to express source. Note that, given the extensive polysemy between location and direction in French, the former source adverbs may well have acquired direction meaning at the same time as location meaning. What seems clear from the evidence adduced by Mackenzie, as well as from other instances of the semantic extension described above (cf. for example Bennett 1989, Nikitina & Spano forthc., Luraghi 2009 and 2010a), is that once a marker acquires the locative meaning, it loses the original ablative meaning. Thus, while the extension from source to location is attested, possibly even more frequently than commonly believed, polysemy tends to be avoided. As we will see further on, this 11
situation also holds between other semantic roles. 6
2.1.3 Space and time Before moving on to other types of relation, some remarks about the coding of time are in order. If one takes Heine et al.s scale in Fig. 1 literally, one would be led to think that human relations, or other non-spatial relations serve as a source domain for time. However, this does not seem to be the case as there is extensive evidence for direct mapping of the domain of space onto the domain of temporal relations, see e.g. the many examples mentioned for various temporal concepts in Heine and Kuteva (2002). As pointed out by Lyons (1977: 718): The spatialization of time is so obvious and so pervasive a phenomenon in the grammatical and lexical structure of so many of the worlds languages that it has been frequently noted, even by scholars who would not think of themselves as subscribing to the hypothesis of localism. Indeed, the metaphor TIME IS SPACE appears to be deeply entrenched in human cognition, as shown among other things by the extent to which temporal relations are coded through primarily spatial markers (Haspelmath 1997). 7
2.2 Human relations
Among human relations, Heine and his associates keep agent, comitative and beneficiary separate from possessor and recipient/addressee (for which they use the label dative), which they include in
6 It goes beyond the scope of the present paper to provide a detailed investigation of the reasons why certain polysemies hold while others do not at the level of spatial relations; however, I will give some tentative suggestions in the discussion in sec. 4. Neither is it possible here to establish the direction of semantic extension in the case of the location/direction polysemy. 7 In spite of frequent extension from space to time and almost inexistent extension in the opposite direction, some scholars contend that the cognitive domain of time is as basic as that of space, see e.g. Evans (2004). 12
the domain of human activities. In part, there are diachronic considerations supporting this distinction: for example, recipient seems to originate from beneficiary rather than the other way around. I discuss this issue later on in this paper; for the time being, however, I prefer not to create any partition in the group of human relations. In this section, I survey the most frequent patterns of extension from spatial to human relations, and explore metaphorical mechanisms that provide mapping from the source domain of space to the target domain of human relations.
2.2.1 Two maximally distinct metaphors Let us start from the observation that polysemy frequently holds between source and agent on the one hand, and direction and recipient/beneficiary on the other. Evidence for such polysemies is readily available from numerous genetically unrelated languages: polysemy of source and agent is common in several Indo-European languages, including Romance, Greek, part of the Germanic languages; in the Semitic languages, passive agent is typically encoded through prepositions cognate to Arabic mn from, in Turkish one finds the postposition tarafindan from the side (Kornfilt 1997; for more evidence see Heine & Kuteva 2002: 29-30). In many cases, the diachronic development is also attested, as in the case of the Romance languages. Such semantic extension is based on the metaphor AGENTS ARE ORIGINS: agents, as initiators of events, are conceived as the point in space where events originate (see Luraghi 2000). In the case of polysemy of recipient/beneficiary and direction, the evidence is overwhelming (see Blansitt 1998, Rice & Kabata 2007); the underlying metaphor is RECIPIENTS ARE DESTINATIONS. Hence one can conclude that two separate metaphors account for different semantic roles typical of human participants. As the source domain provides for maximal distinction (source and direction are at the two opposite edges of the conceptual domain of space as represented in Fig. 3 above, and polysemy between the two is virtually non-existent, as shown in Tab. 1), it is expected for languages to avoid merging of these two roles. As we will see in section 3.3, however, this is not always the case. 13
2.2.2 Possession: direction or location? A possessive relation prototypically implies some kind of control of the possessor over the possessee (see Taylor 1989a and Heine 1997a and the discussion therein; control may not apply in case of non- prototypical possession, e.g. inalienable or abstract, see Heine 1997a: 3). Below is a partial list of features of prototypical possession, adapted from Taylor (1989b: 202):
(a) the possessor is a specific human being ...; (b) the possessed is a specific concrete thing ...; (c) the relation is an exclusive one, i.e for each thing possessed there is only one possessor; (d) the possessor has the right to make use of the possessed; ... (e) ... (f) the possessor is responsible for the possessed; ... (g) ... possessor and possessed need to be in close spatial proximity; (h) the relation of possession is a long-term one...
It has been observed by Heine (1997a, b) that the Locative Schema is the most frequent way of expressing possession cross-linguistically. By this schema, possessors are metaphorically conceived as locations, and the locative relation is conceived as implying some sort of control over an entity (a possessor controls a possessee as a consequence of features such as [d] and [f] above). This complex relation can be explained by means of two underlying metaphors: POSSESSORS ARE PLACES and PHYSICAL VICINITY IS CONTROL. 14
Remarkably, Heine also distinguishes a Locative Schema from a Goal Schema. 8 Among instantiations of the latter, he includes languages in which possessor is coded through the dative, and writes: As a source for predicative possession, this schema typically consists of a verb of existence or of location, where the possessor is encoded as a dative/benefactive or goal case expression and the possessee typically is a subject. Since dative/benefactive markers are frequently derived from allative/directional markers, the latter functions may also be part of the case marking figuring in the Goal Schema (1997b: 95). Based on the examples provided in the book, however, there is no evidence for a real Goal Schema separate from the Locative Schema. Consider the following examples, given by Heine (1997b: 95-96; glosses by the author) as instantiations of the Goal Schema:
(4) ur velo c'hlas am eus a bike blue to : me is I have a blue bike. Breton (Celtic, Indo-European) (5) waska tiya- puwan rope exist- for. me I have a rope. Bolivian Quechua (Quechuan, Andean) (6) Le livre est moi the book is to me The book belongs to me. French (Romance, Indo-European)
Remarkably, in all the above occurrences we find verbs of rest (the verb be or exist), so there is no real evidence for the marker to express direction: this could only be the case if motion verbs were
8 Further schemas described by Heine (1997a) are the Source Schema, on which see below sec. 2.2.3, the Companion Schema, which I discuss in sec. 3.5, and the Topic Schema. 15
used. Indeed, frequent polysemy of location and direction markers accounts for the double nature of many cases/adpositions that can accordingly indicate both location and direction. So evidence for a Goal Schema should be sought in languages which present a rigidly tripartite system of local relations. However, as we will see below, one such language, Finnish, only provides evidence for a connection of possessor with location, rather than with direction.
2.2.3 Comitative and the domain of spatial relations Comitative markers display a number of frequent polysemies that I discuss further on in this paper, but they indicate spatial relations only infrequently. This is not to say that space cannot function as a source domain for comitative: indeed, in several cases in which the origin of comitative markers can be detected, location seems to provide the source for the comitative. Unfortunately, evidence for this historical development is limited, and mostly provided by Indo-European languages. Examples are various comitative prepositions in the Germanic languages, such as English with, from O.E. wi against, opposite to and German mit, from Proto-Germanic *medi-, cognate of English middle, Greek me, from Ancient Greek met, original meaning among (see Luraghi 2001b and 2005c on this development), Catalan amb from Latin apud (near)by. In some cases, the spatial origin of comitatives cannot be traced even over several millennia of attested history: Italian and Spanish con, for example, are the outcome of Latin cum with, a preposition which has always indicated comitative and preserves no traces of spatial meaning even in the earliest sources (Leumann, Hofmann & Szantyr 1965: 260). Only comparative evidence, which connects Latin cum with Oscan ku (near)by, allows one to reconstruct the original spatial meaning of the preposition. Apparently, this is another instance in which a newly acquired meaning pushes out the original meaning, as in the case of the extension from source to location. I discuss some possible motivation for this in sec. 4. It must of course be mentioned that sources for comitatives are not limited to spatial relations. As shown in Heine and Kuteva (2002: 91), comitatives often derive from 16
lexical items that do not have spatial reference, as in the case of the Estonian ending -ga of the comitative case, from Balto-Finnic *kansa people, or of various comitative markers that derive from serialized verbs with meaning such as follow, take, etc. (see also Stolz et al 2006). In addition, comitatives may arise from adverbs meaning together, which, in turn, often do not originate out of spatial concepts. Such adverbs often reinforce comitative markers, and a possible development is for them to take over and substitute the latter, in a sort of cycle described in Stolz et al. (2006: 363). Evidence for their frequent non-spatial origin is extensive; suffice it to mention English together based on the root of the verb gather, German zusammen, based on the Germanic source for the word same, which eventually goes back to the PIE numeral for one (*sem, cf. Ancient Greek hn one(neuter)), French ensemble from Latin insimul at the same time. The fact that sources for comitatives are available outside the spatial domain may be part of the answer to the question why the polysemy of comitative and spatial notions is limited. However, the relevance of non-spatial sources for comitatives must not be overstated. Markers of other semantic roles may also have non-spatial origin: recipient markers for example may arise from the verb give. But still, when direction markers extend to recipient, they also retain their original spatial meaning, as in the case of English to (direction and recipient). This peculiarity regarding comitative might owe to the complexity of local relations involving human landmarks. It has been noted by Creissel and Mounole (2011) that human beings are not good landmarks of spatial relations, as they are highly mobile entities. As a consequence, location with respect to a human landmark often really indicates location in his/her habitual space, as with French chez at (only with human landmarks):
(7) Je suis chez mon frre (mais il n est pas l). I am at my brother but he not is not there I am at my brothers (place), but hes not there. 17
Compare an inanimate landmark:
(8) Je suis la gare (*mais elle [=la gare] n est pas l). I am at the station but she [the station] not is not there I am at the station (*but its not there).
On the other hand, comitative generally indicates that the accompanee is present:
(9) Je suis avec mon frre (*mais il n est pas la). I am with my brother (*but he not is not there) I am with brother (*but hes not there).
Thus one can set up the following stages for the extension from locative to comitative:
(10) (generic locative ---> specialized locative for human landmarks --->) comitative
where the parentheses indicate that the former two meanings stop being active as soon as the markers are extended to comitative. Note that there is no need to set up any special metaphor here: a specialized locative marker for human landmarks may come to be understood as comitative simply through pragmatic inference (if person is located by another person, then they are also together).
2.2.4 A spatial metaphor or an extension from other human roles? A third semantic role typical of human beings that is often conceptualized in terms of location is the experiencer. As well known, this semantic role can be coded in a variety of ways. In particular, it is 18
often coded as agent (as in John loves Mary) or patient (as in The thunder scared him); perhaps the most frequent association of the experiencer role with a certain coding device cross-linguistically is the dative (English is quite untypical in not following this tendency). Dative experiencers are reported from Indo-European, Afro-Asiatic, Tibeto-Burman, and Niger-Congo languages, among others. Below is an example from Hungarian (Uralic) of dative encoding experiencer:
(11) Ez tetszik Pternek. this appeal-3SG Peter-DAT This appeals to Peter, Peter likes this.
Indeed, experiencers have something in common with agents, patients, and recipients, as summarized in Table 3:
EXPERIENCER AGENT RECIPIENT PATIENT HUMANNESS + + + - CONTROL +/- + + - CHANGE OF STATE +/- - - + VOLITIONALITY +/- + - -
TABLE 3. FEATURES OF SEMANTIC ROLES (adapted from Luraghi 2010b)
Experiencers are typically human: in this respect, they are similar to agents, as well as to recipients and beneficiaries. Contrary to agents, experiencers do not act voluntarily: in this respect, experiencers are similar to patients. However, there are crucial differences between prototypical experiencers and prototypical patients: in the first place, as already remarked, experiencers are human, because, by definition, they must be sentient beings, while prototypical patients are not. In addition, prototypical patients necessarily undergo a change of state, while experiencers may or may 19
not do so. Situations in which experiencers occur are prototypically states (see Van Valin, LaPolla 1997: 85; Croft 2001: 55-156), however, experiential predicates can also be inchoative, thus implying a change of state by which the experiencer enters the situation (as in e.g. Mary fell ill). For what concerns us here, the fact that experiencers are often static participants is most interesting, because it makes them good candidates for a metaphor that conceives of human beings as locations/containers, such as EXPERIENCERS ARE PLACES (OR CONTAINERS) FOR FEELINGS/SENSATIONS (based on another metaphor: FEELINGS/SENSATIONS ARE THINGS). Hence coding through the dative case might be explained by referring to its link with the locative in the source domain. On the other hand, coding through the dative may owe to similarities between the roles of experiencer and recipient: as discussed above, experiencers share relevant features of recipients, as they are human; in inchoative situations they may be conceived as recipients of sensations or emotions. This similarity can be accounted for in terms of a metaphor according to which EXPERIENCERS ARE RECIPIENTS OF FEELINGS/SENSATIONS (again based on: FEELINGS/SENSATIONS ARE THINGS). Moreover, even though neither experiencers nor recipients are real controllers in the sense agents are, the fact that they are necessarily human implies a certain degree of control, at least inasmuch as they are consciously involved in situations: I have indicated this feature of experiencers with +/- in the control column. 9
9 There is an extensive literature concerning the notion of control as related to experiencers. In principle, experiencers should not be controllers as they simply experience a situation. However, the construal of experience typically allows for varying degrees of control. Depending on the type of experiential predicate, the experiencer can be conceived as more or less actively involved in the situation, hence more or less capable of exerting control. Generally speaking, control is absent in case of bodily sensations, whereas perception and volition are most often conceived as controlled activities. Emotions can be construed in a variety of different ways even within the same language, as shown by possible usage with imperatives in many languages. See Verhoeven (2008: 41-51) for discussion and further reference, and Luraghi & Sausa (2013). As for recipients, the transfer event in which they are involved implies that the entity transferred is in their possession, hence that they can control it at the end of the transfer. 20
Evidence for both possible interpretations (i.e. experiencers as locations or as recipients) comes from several Finnic languages. In Estonian, for example, experiencers can be encoded through the adessive, a spatial case which can also encode location; the same, though to a more limited extent, is true of Finnish (examples (12)-(14) are from Erelt & Metslang 2006 and Huumo 1996: 229):
(12) Mul on hbi / piinlik I- ADESS be:3SG shame:NOM / embarrassing:NOM I am ashamed / I feel embarrassed. (13) Mul on vaja ttada I- ADESS be:3SG need work-INF I need to do some work. (14) Hnell on hyv maku s/he-ADESS be:3SG good taste S/he has good taste.
Note however that encoding of the experiencer with the adessive can be due to the extension of the Possessor Schema as the adessive encodes possessor in these languages (see below): such an extension is cross-linguistically quite frequent (cf. French Jai faim, German Ich habe Hunger, both meaning Im hungry, lit. I have hunger; similar usages are reposted from various genetically unrelated languages, such as Kannada cf. Amritavalli 2004, Tibeto-Burman languages cf. Bickel 2004, Thai cf. Matisoff 1986, Niger-Congo languages cf. Heine 1997a). Thus, it seems more likely that, rather than depend on a putative metaphor by which EXPERIENCERS ARE PLACES, this type of coding is explained through two other metaphors according to which EXPERIENCERS ARE POSSESSORS OF SENSATIONS and SENSATIONS ARE THINGS POSSESSED. 21
Coding of an experiencer through the allative is also frequent both in Estonian and in Finnish, as shown in (15) and (16):
(15) Anne-le meeldi-b dss Anne-ALL like:3SG jazz. NOM Anne likes Jazz. (Estonian, from Erelt and Metslang 2006: 255) (16) Minulla on flunssa I-ALL be:3SG flu.NOM I got a flu. (T. Huumo, p.c.)
Again, one could think of a metaphor that connects the domain of space directly with the domain of experience, 10 but, in light of the common features of experiencers and recipients discussed above (and of the widespread tendency for dative markers to encode recipients cross-linguistically), it looks plausible that experiencers are re-interpreted as recipients. This does not necessarily imply that the metaphors indicated above for such polysemy are always activated. Rather, the semantic extension is based on a gestalt effect, in the sense of Lakoff (1977), whereby common features of both roles, i.e. animacy and sentience without agency, are magnified. Thus, even though recipient markers ultimately go back to direction markers, there is no need to necessarily connect the notion of direction with that of experiencer. I return to the coding of the experiencer role below, sec. 3.7.
2.2.5 Competing metaphors: the case of possessor Location is not the only spatial relation to provide a source for possessor coding. As well known,
10 For example, one might argue that the allative is used if the verb indicates directionality towards the experiencer (T. Huumo, p.c.); however, the verb be is pretty frequent in such constructions.
22
possessors are frequently coded through the genitive case or adpositions with similar meanings, and the typical source for genitives is constituted by ablative markers. This amounts to saying that source also provides a frequent source domain for possessor: this well-known and frequently investigated development is based on the metaphor POSSESSORS ARE ORIGINS (Nikiforidou 1991: 174). The extension from source to possessor is documented in numerous languages: virtually all the Indo-European languages that have lost case marking at least partially attest to this development. When a marker of source extends to possessor and becomes a genitive marker, it often retains its original spatial meaning, but it may also lose it: evidence for both developments is easily available from German (von means both from in spatial sense and of) and English (of is limited to the expression of genitive relations and retains trace of the original separative meaning only when governed by certain verbs or adjectives, as for example in independent of). More examples are discussed by Heine and Kuteva (2002), who caution (p. 35) that most of these examples relate to Indo-European languages; more research is required on the genetic and areal distribution of this process.
source location direction SOURCE DOMAIN: SPACE
path
POSSESSORS POSSESSORS ARE ORIGINS ARE LOCATIONS
TARGET DOMAIN: POSSESSION
FIGURE 4. THE MAPPING OF SPATIAL RELATIONS ONTO POSSESSOR
Thus, in the case of possessor two competing metaphors offer the channel for the mapping of 23
the source domain onto the target domain, as shown in Fig. 4. That different metaphors can provide the source for a certain concept is not surprising: as is well known, for example, agents can originate from various spatial concepts as well as from instruments. Remarkably, the two types of spatial concepts involved here, source and location, are encoded through markers that generally do not allow polysemy between each other. Thus, the two types of metaphor account for possessors that are always distinct from the point of view of encoding.
2.2.6 Recipient and beneficiary Thus far, I have mentioned recipient and beneficiary as a pair. Indeed, polysemy of recipient and beneficiary is frequent cross-linguistically (cf. Kittil & Ziga 2010: 18 among others). Typical examples involve the dative case of various Indo-European languages as well as of languages of other families, cases partly covering the same functions as a dative, as e.g. the Finnish allative, or adpositions that also cover functions typical of the dative. Examples (17) and (18) are from Finnish (allative case, from Kittil & Ziga 2010), while (19) and (20) are from Biblical Hebrew (preposition l- to, also used in direction expressions):
(17) Vanhempi antoi lapselle lahjan parent:NOM give-3SG.PST child-ALL present-ACC The parent gave the child a present. (18) Vanhempi leipoi kakun lapselle. parent:NOM bake-3SG.PST cake-ACC child-ALL The parent baked the child a cake. (19) wa- ttittn gam l- h and she.gave also to-man.her And she gave [the fruit] to her man too. (Gen. 3.6) 24
(20) k YHWH lohem h hannilm l-em because Lord god-your he he.fights for-you Because the Lord himself is fighting for you. (Deut. 3.22)
Whenever the diachronic development of such markers can be traced back in time, it provides evidence for extension from beneficiary to recipient, as indicated by Heine et al. (1991: 159). Evidence for this development is not abundant, perhaps on account of the fact that several languages with a dative case display the polysemy as far back as one can trance their history. In Luraghi (2010c) the following development for the Ancient Greek preposition eis/es is described:
a) Homeric Greek: direction and purpose:
(21) prta mn es Plon elth first PTC to Pylos:ACC go:IMPT.AOR.2SG First go to Pylos. (direction; Hom. Od.1.284). (22) h me ml eis tn koimsate nl hpni PTC 1SG.ACC much to damage:ACC put.to.sleep:AOR.2PL harmful:DAT sleep:DAT You put me to sleep in order to damage me with harmful sleep. (purpose; Hom. Od.12.372)
b) Classical Greek: direction, purpose and beneficiary:
(23) ho Krosos t pn es autn epepoikee ART.NOM Croesus:NOM ART.ACC all:ACC to 3SG.ACC make:PLPF.3SG Croesus had done all that he could for him. (beneficiary; Hdt. 1.85.1) 25
In post-Classical Greek, eis slowly extended to recipient, and in Modern Greek, its outcome s is the common means to indicate recipient (and other functions of the dative case, which disappeared in Byzantine Greek):
(24) Edosa to vivlio ston Antone I.gave the book to+the Anthony I gave the book to Anthony. (recipient).
The Greek example also attests to polysemy of direction and purpose (examples (21) and (22)) as preceding extension to beneficiary and recipient, thus showing a direct mapping of the spatial domain on the domain of abstract relations. This development, which is based on the metaphor PURPOSES ARE DESTINATIONS, explored in sec. 2.3.3 below, is apparently cross- linguistically frequent; it contradicts the predictions in (1), by which extension to human roles should always precede extension to inanimate and abstract roles.
2.2.7 Beneficiaries that are not recipients In spite of the frequent polysemy outlined above, there is extensive evidence for other sources for beneficiary markers, which do not display it. Remarkably, these are markers whose original local function is not direction. In Finnish, beneficiary can be coded through the postposition vuoksi, which also codes cause, as shown in (25) and (26):
(25) Henkilo opettel-i suome-a yksilo-n vuoksi. person.NOM learn-3SG.PST Finnish-PART individual-GEN for A person learnt Finnish for an individual. (beneficiary; from Kittil 2010); 26
(26) Jaatelo sul-i sahkokatko-n vuoksi. ice.cream.NOM melt-3SG.PST power.failure-GEN for The ice cream melted because of the power failure. (cause; T. Huumo p.c.).
The local meaning of vuoksi is through, which may explain its extension to cause (on the meaning and etymology of this postposition see Itkonen, Joki & Pelotla 1978: 1813-1814; 1818). Without going into the details of this development, what is remarkable is that, as already highlighted above, this is not a marker of direction. Similarly, the English preposition for and German fr, which can code beneficiary, cause and purpose, derive from Proto-Germanic *fura, before, cognate with Latin pro. Interestingly, the latter preposition, too, developed a beneficiary (and purpose) meaning out of the spatial meaning before (Luraghi 2005a, 2010a). In the intermediate stage, attested in Early Latin, it indicated an exchange, as in (27):
(27) an tibi malam rem vis pro male dictis PTC you:DAT bad:ACC thing:ACC want:PRS.2SG for bad(ADV) say:PTCP.ABL.PL dari? give:INF.P Do you wish a punishment to be given you for your insult? (Pl. Men. 496).
Sporadically, the preposition could indicate beneficiary, especially of the behalf type, already in Early Latin, as in (28):
(28) ego ibo pro te, si tibi non libet 1SG.NOM go:FUT.1SG for 2SG.ABL if 2SG.DAT NEG like:PRS.3SG Ill go for you, if you dont feel like. (Pl. Mos. 1130). 27
Later, the meaning extended to all types of beneficiary, as well as to purpose, a meaning which was already implicit in the notion of exchange. Still later, the meaning also extended to cause, as shown in (29), from the New Testament:
(29) quia et Christus semel pro peccatis mortuus est iustus pro because even Christ:NOM once for sin:ABL.PL die:PF.3SG righteous:NOM for iniustis unrighteous:ABL.PL Because Christ also suffered for sins once, the righteous for the unrighteous. (1 Pet. 3.18).
Note that the second occurrence of pro in (29) could still have a behalf interpretation, thus pointing to the original meaning of substitution. A partly similar development has apparently taken place in the case of the Ancient Greek preposition hupr over, which, besides having spatial meaning, developed into a beneficiary marker. In much the same way as with Latin pro, this extension started with the behalf type of beneficiary. Similar to Germanic for and Finnish vuoksi, Latin pro and Greek hupr did not code direction. In addition, none of them underwent a semantic extension to also include recipient. Another case of a beneficiary marker that does not code recipient and whose spatial origin is not connected with an allative sense is the Turkish postposition iin, from i inside, which I discuss below, sec. 2.3.3. 11
11 Note that in studies regarding sources for beneficiaries, usually such types of semantic extensions are ignored, or just mentioned in passing, while there is a general tendency to only stress the relatedness of beneficiary and direction. 28
2.2.8 Competing metaphors: beneficiary From the evidence in sec. 2.2.7, one must conclude that the relation between recipient and beneficiary, though well documented and cross-linguistically very frequent, is not such that the two roles cannot be coded independently from one another. Apparently, we are dealing with different metaphors based on the source domain of spatial relations. The first one, which accounts for polysemy of recipient and beneficiary, involves direction; the second, which is not active for recipient but only for beneficiary, involves a more complex locative or perlative relation. Only in the former case does polysemy of beneficiary and recipient seem to arise naturally. This may indicate that, even if beneficiary precedes recipient in semantic extension, the spatial meaning, i.e. direction, is also active when a specific marker extends to recipient. This conclusion finds further support in the fact that markers of recipient/beneficiary for which extension from an original spatial meaning is historically documented also retain the direction meaning (as for example French , English to, the Finnish allative case and many other). The metaphor that maps direction onto beneficiary depicts the latter as the endpoint of a prospective trajectory: it could be stated as A BENEFICIARY IS A DESTINATION; however, it is usually stated in terms of similarity between destinations and recipients RECIPIENTS ARE DESTINATIONS (see e.g. Rice & Kabata 2007). Indeed, the transfer is based on a feature that the beneficiary has in common with the recipient, and that indeed is even more relevant for the latter than for the former. Beneficiaries typically benefit from a certain state of affairs, and often they are the prospective recipients of an entity, even though they might not actually get hold of it in the end. Consider example (30):
In this vein, Schmidtke-Bode (2010: 127-128) remarks that there is cross-linguistic evidence for beneficiary (and purpose) markers that do not indicate direction, but then does not pursue this matter further in his discussion. 29
(30) Mary baked a cake for Paul, but Paula ate it up before he could even see it.
On the other hand, a recipient can be conceived as a special case of beneficiary, who profits by an even by actually getting hold of an entity as its result (Kittil 2005, Luraghi 2010c), as shown in (31):
(31) Mary gave a cake to Paul (???but he did not receive it).
Indeed, in spite of the fact that in the prototypical beneficiary situation there may be no actual transfer (Goldberg 1995: 37), in the default interpretation it is assumed that the intention of the beneficient is to actually transfer the benefactum to the beneficiary (see Luraghi 2010c). In this sense, the polysemy of recipient and beneficiary can be described as a predicational metonymy in the sense of Panther and Thornburg (2007: 246), whereby an intended action is taken as an actual action. See further the polysemy of possessor and recipient/beneficiary, sec. 3.2. Other metaphors available for beneficiary are not based on directional motion. Just a glance at the adpositions mentioned above indicates that there can be more types of conceptualization at work; one, which accounts for the extension of Latin pro in front of and Greek hupr over to behalf beneficiary and then to all types of beneficiary involves a covering relation. Both prepositions can indicate that a trajector is placed in such a position with respect to a landmark to cover it from the perspective of a possible observer, and thus replaces the landmark, as shown schematically in Fig. 5. The relation of replacement is accounted for by the metaphor which equates existence with being in the observers visual field, and can be stated as follows: EXISTENCE IS VISIBILITY (Lakoff et al. 1991). Coverage and replacement induce the idea of exchange and hence of surrogation, which leads to a behalf beneficiary (for further reference see Luraghi 2005a, 2010c, 2013a).
30
observer
Tr Lm FIGURE 5. COVERING RELATION
2.2.9 A Balto-Finnic perspective on the Indo-European dative The Indo-European dative, a case which is often described as typical of human roles, provides little evidence for its ultimate spatial origin. In the earliest sources of the ancient Indo-European languages, spatial usage of the dative is restricted: a few occurrences that might indicate an allative meaning are available from Latin, but it must be noted that in several languages the dative tends to merge with the locative, as in Greek and Anatolian (see Luraghi 2001b and the references therein). This might very well be the origin of the dative case: it was observed by Kuryowicz (1964: 190-193) that the endings of the dative and the locative singular look very much like morphophonologically conditioned variants of the same morpheme. Evidence for the relatedness of datives and locatives, as well as for the close association between dative and animacy in other language families has been provided in Rodriguez Aristar (1996), based on data from several Australian and Amerindian languages. A look at the Finnish cases that correspond to various functions of the Indo-European dative is enlightening. In the Indo-European languages, the dative typically encodes recipient, beneficiary, possessor, maleficiary, among other roles. The recipient construction is usually extended to third argument of verbs that indicate separation, especially in the case that they have human referents and indicate maleficiary, as shown in (32) and (33): 12
12 Such usage of the dative, which seems to be most typical of the Indo-European languages of Europe except English (Haspelmath 1999), is usually described as external possessor, and often occurs with inalienably possessed entities, 31
(32) Mi sparita la borsetta me:DAT is disappeared the purse My purse disappeared. lit.: The purse disappeared on me (maleficiary). (33) Ten pn nm vzal hodn penz DEM man us:DAT took much money That man took a lot of money from us (maleficiary). (Czech, from Janda 1993: 58)
In Finnish, these roles are coded by three different but related cases, that is, the adessive (possessor), the allative (recipient) and the ablative (which codes source, especially with human referents). Examples are (34)-(36) (from Huumo 1996: 74):
(34) Liisalla on kirja Lisa-ADESS be:3SG book:NOM Lisa (possessor) has a book. (35) Liisalle tuli kirje Lisa-ALL come:3SG.PST letter:NOM Lisa (recipient) received a letter. (36) Liisalta katosi kukkaro Lisa-ABL disappear:3SG.PST purse:NOM Lisa (maleficiary) lost her purse.
even though this is not necessarily the case: consider e.g. the following Italian example: Mi hanno rubato la macchina di Giovanni Someone stole Johns car from me. (lit.: They stole Johns car on me.). The dative indicates maleficiary, and the possessive relation is clearly temporary. See further Havers (1911). 32
Remarkably, these three cases are considered local cases more on account of their names than of their usage: as argued in Huumo (1996), they are mostly used for human relations, as shown in the examples, while local relations are usually expressed by the inessive, illative and elative. The Finnish data point toward a difference in the conceptualization of possessor and recipient/beneficiary/maleficiary, based on different parts of the domain of space as the source domain. One might wonder whether polysemy of these two roles is connected with polysemy of location and direction across languages that present it. This is certainly the case in the Indo- European languages. 13
2.2.10 Path as a source for human relations: intermediary In sec. 2.2.1, I have pointed to a lesser basicness of path among spatial semantic roles. This is not to say that path cannot function as an independent concept in the source domain of space for metaphors leading to other target domains, including that of human relations. In particular, a frequent extension of path markers is the role intermediary, as in the case of English through in (37):
13 It is outside the scope of the present paper to provide a detailed description that can support this claim; however, some remarks can illustrate it. Even if a distinction between direction and location markers often exists, a certain degree of polysemy of direction and locative is also normally present. In particular, Indo-European languages with a separate locative case, such as Sanskrit, display constructions in which direction is coded by the locative (see Delbrck 1867); in addition, the typical situation in which spatial relations are coded through adpositions with case variation, as in the case of German two-way preposition, most often features a number of prepositions that can code location and direction with different cases, while source is usually kept distinct by the use of different adpositions. In languages where such alternation occurs (such as Latin or many Slavic languages, for example) one also finds adpositions that do not allow for case variation and, if they are polysemic, their meaning includes location and direction. As for the dative case, as already noted above it seems to present more affinities with the locative than with the allative (which is most likely the original meaning of the Indo-European accusative). 33
(37) They have to speak through an interpreter to be able to communicate effectively.
An intermediary can be viewed as a human instrument. Instrument markers, however, do not easily encode human instruments: a quick cross-linguistic survey indicates that such markers usually encode different relations when they occur with human nouns, which prevent them from being interpreted as encoding instrument. Let us consider, for example, languages in which instrument is encoded through the Companion Metaphor, such as English and many other European languages. In these languages, the instrument marker encodes comitative with animate nouns, as English with: thus, if a human NP co-occurs with with the default interpretation is comitative. Indeed, in (38):
(38) I achieved X with Y[=human noun] / I achieved X with Z [=inanimate concrete noun]
the interpretation of the role taken by the NPs with Y and with Z as comitative and instrument respectively depends crucially on the lexical meaning of the NPs involved. In languages that do not conform to the Companion Metaphor, and consequently do not encode instrument through the comitative marker, on the other hand, there is a frequent tendency for instrument markers to extend to agent (see sec. 2.3.2 below). Thus again, there appears to exist a default interpretation that makes it impossible to use the instrumental case (or analogous morphemes) with such non-prototypical instruments as human beings. The same seems to happen in the case of frequent instrument/location polysemies: the instrument marker usually has another favored interpretation when occurring with a human referent. The Finnish adessive, for example, indicates instrument (or manner) with inanimate nouns and possessor with animates (Huumo 1996: 76-77); the West Semitic instrumental/locative preposition b- occurs with human referents mostly in the plural in the meaning among, as in example (39) from Biblical Hebrew (see further Pennacchietti 1974: 175): 34
(39) hay-yf b-annm the-beautiful in-women Beautiful among women. Ct 1.8.
In sum, prototypical instrument markers appear to be exploited for different purposes when they occur with human nouns, and have default interpretations that rule out human instrument. This comes as no surprise: the fact that a human being is used as an instrument is comparatively rare and unexpected; human beings take roles such as comitative, agent or possessor much more frequently. In addition, an intermediary is not simply an instrument: even if the intermediary acts under someone elses instigation, s/he holds control on the event. In states of affairs that feature intermediaries, one can conceive of agency as being split between the primary agent, who initiates the state of affairs voluntarily and the intermediary who holds the responsibility for the actual carrying out of the action (see Luraghi 1995). Hence the intermediary functions as the channel through which the agent achieves his or her purposes. The notion of path accounts for this conceptualization based on the metaphors AN INTERMEDIARY IS A CHANNEL and AGENCY IS A THING TRANSFERRED. The literature on intermediary is not especially rich, and there are no specific studies that investigate sources and polysemies that involve this semantic role. At least in the Indo-European languages, there seems to exist a frequent polysemy of path, intermediary and means, as in English through. Indeed, means is defined as the role taken by entities used by an agent to achieve his/her goals, with a lesser degree of manipulation than instrument. 14 Typical of means expressions are
14 There are no specific studies devoted to this semantic role, which can also be conceived as a non-prototypical type of instrument. Available definitions are not especially enlightening. Radden (1989:442443), for example, gives the following definition of means: The means role denotes physical or abstract entities which, without being instruments themselves, are essentially involved in bringing about the effects of an agents action. He then concedes that the 35
abstract nouns, which cannot be manipulated, and consequently can be controlled by agents to a lesser degree than prototypical instruments. This feature makes means similar to intermediary: as a human being, an intermediary cannot simply be used and controlled completely by an agent. There are no diachronic studies regarding the relation between intermediary and means. In a study of the Ancient Greek preposition di with the genitive, Luraghi (1989) found that extension from path to intermediary preceded extension to instrument and means, but more evidence is needed in this respect. Remarkably, in Ancient Greek the distinction between instrument and means is not reflected in coding, so the extension from intermediary involves both roles. In any case, the use of markers specifically connected with the semantic role means to prototypical instrument seems to be common (see below, sec. 2.3.1). In several languages, path markers also extend to agent. Possibly, this further extension needs an intermediate stage at which path morphemes encode intermediary. An attested example which might provide evidence is French par, which encodes path and passive agent. In Latin, the preposition per could encode path and intermediary, but it could be understood as indicating an agent only under specific semantic constraints, even with passive verbs (Luraghi 2010a).
2.2.11 Interim summary In sections 2.2-2.2.10 I have surveyed human relations, and shown how they are conceptualized in terms of metaphors based on spatial relations. Patterns of polysemy of semantic roles markers (cases
distinction between the notions of means and instrument is not always easy to draw, and discusses occurrences in which the same NP, money, can be conceived as an instrument or as a means in the same contexts. A more elaborated definition of means can be found in Croft (1991), who, however, focuses on means-clauses (e.g. By doing X somebody achieved Y). Croft further indicates that the occurrence of a means implies the existence of an agent who acts intentionally: the means clause must begin with a VOL[itional] arcthat is, it must be a volitional action. (1991: 178). 36
and adpositions) recurrently attested cross-linguistically provide evidence for these metaphors. In particular, I have shown that the basic spatial relations of source and direction frequently provide the source domain for agent and recipient/beneficiary respectively. Location may provide the source for possessor, but this semantic role can also be encoded as a source: two competing metaphors are at play here. Similarly, beneficiary can merge with recipient, and be conceptualized and encoded as a direction, but in several languages there are alternative means of encoding this role, which point on the spatial relations of location or path as its conceptual source. Experiencers are involved in static situations, and can be encoded through locative markers; in cases of inchoative situations, they can be encoded through direction markers. The extension of markers of spatial relations to this semantic role does not necessarily point toward a direct mapping of spatial on the domain of experience, but it can also be that experiencers are metaphorically conceived as possessors (which in turn can be conceived as locations), or as recipients (which are conceived as directions). Location can also provide the source domain for comitative, but this type of polysemy is synchronically rare: apparently, location markers that develop into comitative markers lose their original meaning when they acquire the new one. Finally, the spatial relation of path often provides the source for intermediary.
2.3 Non-human relations
According to the scale in Fig. 1, extension to non-human relations should follow extension to human relations. I have already discussed the connection between space and time; more discussion on this hypothesis can be found in Narrog (this volume). In what follows, I survey frequent polysemies and metaphors that account for extension to non-human relations. As we will see, in some cases the domain of space is mapped directly onto the domain of non-human relations, while in other the direction of semantic extension contradicts the scale. 37
2.3.1 Source domains for instrument As has already been noted in this paper, the Companion metaphor, according to which AN INSTRUMENT IS A COMPANION, accounts for the extension of comitative markers to instrument. Languages which rely on the Companion Metaphor can be said to be well-behaved with respect to the scale of increasing grammaticalization in Fig. 1, since they display the semantic extension location > comitative > instrument, modeled according to (1). As this metaphor has been exhaustively described in Lakoff and Johnson (1980) I will not dwell longer on it. Another frequent source for instrument markers is provides by locatives. In such cases, even if often human roles are also associated with the same morphemes, there seems to be a direct connection between space and instrumentality. There are several reasons why locative relations can be understood as instrumental. For example, certain types of instruments are also places in or on which human beings can be located: the obvious example for this is constituted by means of locomotion (see Lehmann, Shin 2005 with several examples from different language and language families). Some instruments are shaped as containers, and other as means of support: this fact accounts for the extension of different locative markers to instrument, based on various transfers and extensions. As both containers and means of support are at least in part more or less prototypical instruments (for examples, a cup can be conceived as an instrument for drinking, a bottle as an instrument for carrying liquids, etc.), it is perhaps not especially useful to multiply underlying metaphors. 15 Rather, there seems to be a sort of analogical extension based on a generalization, which can be stated as in (40):
(40) some containers/supports are instruments all instruments are encoded as
15 The polysemy of locative and instrumental morphemes is also frequent in word formation, see Lujn and Ruiz (this volume). 38
instantiating a containment/support relation.
This generalization consists in the extraction of a schema (see Langacker 1987), whereby certain types of instrument have certain spatial dimensions (container, support). The schema is then analogically extended to all types of instrument (see Tuggy 2007: 100-101 on schematicity and analogy in word formation). Examples are provided by Balto-Finnic (Grnthal 2003, Huumo 1996), Semitic, several Australian languages (which however only provide examples of polysemy, but no historical evidence for semantic extension and its direction); some of them are surveyed in Luraghi (2001a). Interestingly, the portion of the conceptual domain of space that provides a source for instrument is the same, both in the case that the mapping is mediated by the Companion Metaphor and in the case that it is direct: it is the space of location, i.e. static relations. Indeed, instrumentality is part of what Lehmann and Shin (2005) call the domain of concomitance, a notion originally proposed by Coseriu (1970), who describes the meaning of a possible German construction mit C as indicating und C is dabei, or unter Dabeisein von C. In other words, a concomitant is a participant which is crucially indicated by its presence during a state of affairs, that is, by a static relation. Luraghi (2001a) suggests that both location and instrument should be grouped together as concomitant roles, that is, roles that are not oriented either from the starting point of the event or toward its endpoint. As already mentioned in sec. 2.2.10, path markers can extend to instrument, possibly as a further step in an extension that goes from path to intermediary to means and then also reaches prototypical instrument. However, it is by no means clear that the extention of path markers to means and instrument is necessarily mediated by a human role, i.e. intermediary. As this type of extension has never really been investigated on a large language sample, I will assume for the time being that, similar to the Companion Metaphor, this type of extension is also well-behaved, and 39
procedes fron path to intermediary to means and instrument, based on the metaphor AN INSTRUMENT/MEANS IS AN INTERMEDIARY. Note however that this assumption is based in its turn on an assumption, that is, that human roles must precede non-human ones, thus partly resulting in circularity. 16 Remarkably, path is located close to location in the domain of space, between source and direction. To sum up, in this section I surveyed three types of semantic extension involved for the conceptualization of instrument, two based on metaphors (the Companion Metaphor and the Channel Metaphor), and one on analogy. By the latter type of extension, markers of spatial relations are used for certain instruments based on their shape, and then extended through analogy to all types of instrument independent thereof. This means that the extension proceeds directly from the domain of space to instrumentality. In the case of the two metaphors, extension is mediated by human relations, comitative and possibly intermediary. In all such extensions, the source domain is located in an intermediate position between source and cause, and points to an analogous position of instrument in the target domain of causation, between antecedent and subsequent roles in the terminology of Croft (1991). Crofts Causal Chain is represented in Fig. 6.
Antecedent Subsequent cause result SUBJECT means OBJECT ---------------> -------->manner ------------------> ---------------------->
16 Alternatively, if it turned out that extension of path markers to instrument/means is not always preceded by extension to intermediary, one should assume that the domain of space can provide the immediate source for the non- human roles of instrument and means. One should set up another metaphor, by which an instrument (or means) is conceived as a channel that conveys agency (i.e. volitionality, control, and energy transfer). An alternative version of the Channel Metaphor could accounts for extension from space without passing through the stage of intermediary: AN INSTRUMENT/MEANS IS A CHANNEL (and, as in the case of the intermediary, AGENCY IS A THING TRANSFERRED). 40
According to Croft, who divides causal semantic roles into two groups, instrument is an antecedent role. In Luraghi (2001a), it is suggested that semantic roles should better be divided into three groups based on the type of involvement in causation, as comitative, instrument, manner and means are better understood as concomitant rather than antecedent. Spatial sources for instrument confirm it consistent tendency to arise from the area of non-directional relations (location) or at least of relations that do not indicate the starting or the end point of motion (path). As we will see in sec. 2.3.3, polysemy can also involve instrument and cause, that is an antecedent role: indeed, polysemies that cross-cut groups of semantic roles are not infrequent, as discusses in the next sections.
2.3.2 Instrument as a source domain A frequent extension of instrument is matter or material: the material of which something is made is understood as the instrument used to make it. Tentatively, I propose the metaphor MATERIAL IS AN INSTRUMENT FOR CREATING OBJECTS to account for this extension. 17 This semantic role of material also has another frequent source in the space domain, that is, source, following the metaphor OBJECTS COME OUT OF SUBSTANCE, explored in Lakoff and Johnson (1980: 73). Both patterns occur for example in Turkish, as shown in (41) and (42) (from Kornfilt 1997):
17 Remarkably, in this case metonymy, rather than metaphor, could be at play as material and instrument are contiguous concepts in the conceptual domain of creation. I am not going to pursue this explanation further here, but see below regarding extension of instrumental markers to agent. 41
(41) tugladan bir ev brick-ABL a house A house of bricks. (42) bu ev tuglayla yapumtr this house brick-INSTR make-PST-PTCP-COP This house is made with bricks.
An often discussed extension of instrument is agent. This pattern of semantic extension, whose direction is clearly attested at last in the Indo-European languages, contradicts Heines Scale of Increasing Grammaticalization in Fig. 1, as it implies that a non-human role serves as the basis for a human one. Polysemy of instrument and agent is also frequent in several Australian languages, both ergative and nominative/accusative (see Blake 1977, Dixon 2002). In such languages, the instrumental case is most often also the locative (see Luraghi 2001b for discussion). The original meaning of the Indo-European instrumental, instead, is usually held to be comitative (see Delbrck 1867): to some extent and virtually limited to Vedic Sanskrit, the extension to the agent role also brought about the infrequent polysemy of agent and comitative, which I survey in sec. 3.4. It is not clear that extension of instrumental markers to the encoding of the agent can be explained in terms of metaphors. In Luraghi (2001b) it is suggested that this type if extension should rather be accounted for in terms of metonymy as agent and instrument are contiguous concepts and an instrument can be seen as a part of an agent. In addition, it is frequently the case that inanimate entities substitute for animate ones in cases of metonymy, precisely when a part substitutes the whole. The occurrence of a metonymic extension here would also explain why this is virtually the only frequent semantic extension that contradicts Heinescale in (1) by moving contrariwise. Finally, polysemy of cause and instrument is also frequent, but directionality is all but easy to 42
gauge: inasmuch as instrumental cases also code cause, one can assume extension starting from instrument, but this matter has never been the topic of any in-depth study, at least to my knowledge. 18
2.3.3 Cause and purpose Following a frequent conceptualization, CAUSES ARE ORIGINS OF EVENTS (Nikiforidou 1991), and PURPOSES ARE DESTINATIONS (cf. e.g. Lakoff, Turner 1989). Instantiations of these metaphors mentioned in the literature operate by mapping the domain of abstract relations directly onto the domain of space, without necessarily implying an intermediate stage at which human relations provide a link between the source and the target domain (see also sec. 2.2.6). The two metaphors map cause and purpose onto the most distant edges of the conceptual domain of space, in very much the same way as remarked for agent and recipient/beneficiary. And indeed, polysemy of cause and agent on one side, and of purpose and recipient/beneficiary on the other side is frequent, as often noted in the relevant literature. Such mapping of the domain of space onto the domain of causation is in accordance with Crofts Causal Chain (see above, Fig. 6). As mentioned in sec. 2.3.2, polysemies ensuing from semantic extensions that cross-cut the Causal Chain and involve antecedent and subsequent roles are also well documented, even though polysemies within the two (or three) groups are much more numerous. However, while polysemy of agent and recipient/beneficiary is not especially frequent (see below, sec 3.3), polysemy of cause and
18 Instrument markers often also indicate manner; this is usually the case in languages that also feature the comitative/instrument polysemy. According to Heine and Kuteva (2002), comitative markers extend to instrument and then to manner, thus complying with the scale in Heine et al (1991) (see above, Fig. 1). However, a cursory look at the Indo-European languages in which instrument is coded through a bare case and comitative through the same case plus a preposition (such as Russian) also feature both instrument/manner and comitative/manner polysemy. Thus the issue seems to be more intricate, and further research is needed in this area. 43
purpose is extremely frequent, as with English for and many other similar markers. Often, this polysemy also involves beneficiary, but typically not recipient, as shown in (43)-(45):
(43) He ran away for fear. (cause) (44) I bought a present for Mary. (beneficiary) (45) Mary went out for dinner. (purpose)
The same polysemy holds for other non-allative beneficiary markers surveyed in sec. 2.2.7 from Indo-European and Balto-Finnic languages. More example are provided by Turkish iin in (46)-(48) (from Kornfilt 1997) and Georgian tvis in (49)-(51) (M. Topadze p.c.):
(46) Sylemek iin geldim say-INF for come-PST-1SG I came in order to say... (purpose) (47) Bayram olduu iin toplar atld holyday be-PST.3SG for cannon-PL employ-PST-3SG Because of the holyday (cause), cannons were shot. (48) sizin iin bir kitap getirdim 2pl-GEN for one book bring-PST-1SG I took a book for you. (beneficiary) (49) ar scalia pikr-is tvis NEG time think(MASDAR)-GEN for S/he has no time to think. (purpose) (50) ertjeradi gamoq'eneb-is-tvis single usage-GEN-for 44
for a single usage (purpose) (51) es bavshv-is- tvis viq'ide this child-GEN-for I-bought I bought it for the child. (beneficiary)
Different languages also provide evidence for non-unidirectionality of the extension between cause and purpose. Luraghi (2005b) shows that Classical Greek di+accusative, which indicated cause, slowly extended to purpose and later to beneficiary (thus also contradicting Heine et al.s predictions on the relative order of human and abstract relations). Its Modern Greek outcome ji is the standard way to indicate purpose and beneficiary, while still also encoding cause and reason. On the other hand, as mentioned above (sec. 2.2.7) Latin pro first extended to beneficiary and purpose, but corresponding prepositions in the Romance languages, such as French pour, also indicate cause. Croft (1991) discusses the cause/purpose polysemy, and remarks that the notion of reason provides an area for possible contact of these two otherwise separate roles. He writes: [e]xpressions of reason, which is a category of intention, not of causation, can represent events that causally follow the verb segment (a goal or purpose) or precede (a source or motivation). (p. 293); he further adds that reason is nondirectional (ib.). Arguably then, the intermediate part of the domain of space, in which location and path are situated, provides a spatial equivalent to reason, hence also giving rise to expressions of cause and purpose. Similar to reason, it is non-directional: either because it is static (location) or because it profiles a path, but not its starting and endpoints. This seems to happen when the polysemy also involves beneficiary, most likely if the latter role is not marked as recipient: the examples discussed above provide evidence for this claim. 19
In Fig. 7 I give a graphic representation of the mapping of the source domain of spatial
19 However, it is remarkable that extension from purpose to cause of allative morphemes is also attested, see Heine & Kuteva (2002: 246-247) for evidence from African languages. 45
relations onto the domain of causation, including the semantic roles of cause, purpose, and reason.
source location direction SOURCE DOMAIN: SPACE
path
CAUSES THE MIND IS PURPOSES ARE ORIGINS A CONTAINER ARE DESTINATIONS
TARGET DOMAIN: CAUSATION reason cause purpose
FIGURE 7. THE MAPPING OF SPATIAL RELATIONS ONTO CAUSATION
Again, there is evidence for an intermediate area between antecedent and subsequent roles, which has a spatial equivalent in the non-directional relation of location and partly also of path, a role which, as already remarked, does not include the starting and endpoints of motion and is not telic. The encoding of the semantic role reason with locative or perlative markers must be accounted for by a complex metaphor, based on instantiations of the container metaphor (THE MIND IS A CONTAINER), and other more specific metaphors that explain the conceptualization of reason in terms of particular spatial configurations (as in the case of Latin pro before described in sec. 2.2.8).
3. Expected and unexpected polysemies among human relations
46
3.1 Agent and possessor
The roles of agent 20 and possessor may share the same source domain, drawing on the spatial relation of source (or origin). 21 Indeed, genitives of agent are common in the Indo-European languages, in which, however, they are mostly limited to nominal forms of verbs. According to Hettrich (1990: 97), such constructions started out from a possessive genitive modifying a verbal adjective; later, they also extended to finite verbs to varying extents in different languages (p. 93-95). A similar development is found in Finnish. Finnish has no real agented passive, but infinitives can occur with genitive NPs which get an agentive interpretation, as shown in example (52):
(52) kirja on minu-n kirjoitta-ma-ni book.NOM be.PRS.3SG I-GEN write-INF3-1SG.POSS The book has been written by me.
(See further sec. 3.3 on a possible interpretation of some genitive agents in Finnish as instantiating the polysemy of agent and beneficiary.) In such construction, too, the infinitive takes a possessive genitive. Thus agents can be interpreted as possessors of actions. The occurrence of nominal forms of the verb in such
20 Cross-linguistically, agents are most frequently realized as subjects of transitive verbs. In nominative-accusative languages, this means that they take the nominative case. Crucially, many transitive verbs do not have agentive subjects; in addition, subjects of all verbs are encoded in the same way in such languages. This points toward a different status of the nominative case: it may well be that nominative case markers eventually go back to agent markers (and possibly markers of source), but their function as markers of a grammatical relation (i.e. subject) makes their possible semantic motivation virtually inexistent. For this reason, I am not going to survey the origin of nominative case markers here. 21 On the relation between source and origin see Nikiforidou (1991) and Luraghi (2003). 47
constructions favors this interpretation: an event which is referred to by means of a nominal form is also conceived as a static entity, a thing, which can be possessed. Metaphors at play here are AGENTS ARE POSSESSORS OF EVENTS and EVENTS ARE THINGS POSSESSED. Not only passive agents, but also ergatives often derive from genitives. According to Lehmann (2002: 98), this type of polysemy arises from constructions of nominative-accusative aligned languages in which a nominal form of the verb takes a possessive genitive: Whenever a (passive) predicate is nominalized - this may occur not only in subordinate, but also in main clauses, namely whenever there is an (analytic) nominal verb form -, its agent may be in the genitive. When such a construction is reinterpreted as transitive, the genitive develops into an ergative. Again, genitive/ergative polysemy is a frequent phenomenon in ergative languages, e.g. in Lak (Caucasian), Eskimo and Sherpa (Tibeto-Burmese). In fact, genitive agents with nominal forms of the verb are usually thought to have provided the origin for ergative alignment in Old Persian (see Hettrich 1990: 95-97; Pompeo & Benvenuto 2011). 22
Polysemy of agent and possessor may also arise as a by-product of both roles being ultimately connected with source in the domain of space: for example, German von from indicates source, possessor and passive agent. However, in such (frequent) cases, the two developments: (a) source possessor (b) source agent can be shown to have occurred historically independent of one another. Thus, this type of polysemy is explained through the common source domains of the two roles, rather than on their own features, once disconnected from their spatial origin.
22 The Old Persian genitive is the merger of the Proto-Iranic genitive and dative. For this reason, it has been suggested, most recently in Haig (2008), that the agent function was connected with the beneficiary meaning of the genitive/dative. While this might in principle be the case, comparison with Avestan and Indo-Aryan suggests that the genitive was the original case in this construction. See Pompeo & Benvenuto (2011) for a detailed analysis of the texts. 48
3.2 Possessor and recipient/beneficiary
Possessor and recipient/beneficiary is a frequent polysemy, and it typically involves the dative case, hence not the type of beneficiary which does not merge with recipient (see sec. 2.2.6 and 2.2.7). Indeed, as I will discuss below, it is recipient, rather than beneficiary, that provides contact with possessor. The direction of semantic extension leads from recipient/beneficiary to possessor: a recipient is someone who gets hold of an entity which is being transferred; after the event of transfer has taken place, the recipient is, at least temporarily, the possessor of this entity. Remarkably, there is no need to set up a metaphor to account for this semantic extension: a recipient is not a metaphorical possessor, rather, the recipient of an entity in a transaction is assumed by default to become its possessor. For this reason, this type of extension can be accounted for in the same way as the extension of beneficiary markers to the recipient role (sec. 2.2.8). Note however that, as this type of polysemy usually involves the dative case, extension to possessor could result from recipient/beneficiary and possessor being neighboring roles in the source domain of space. Indeed, even though the source domain for recipient/beneficiary is provided by direction, while the source domain for possessor seems to be most often provided rather by location (see the discussion in sec. 2.2.5), it is often the case that location and direction merge already in the source domain, as shown by frequent polysemy of location and direction markers (sec. 2.1.2). On the other hand, one can observe partial extension of the dative to non-prototypical cases of possession, such as abstract possession, in languages that otherwise make use of the Locative Schema for prototypical possession, as for example in Russian, as shown in (53) and (54):
(53) Mne dvacat let me:DAT twenty years 49
I am twenty years old. (54) U menja kniga near me:GEN book I have a book.
Most likely, the dative in (53) can be considered close to an experiencer, given the fact that this construction is also used when one indicates bodily sensations, and that experiencer datives are cross-linguistically frequent in languages in which the dative also expresses possession. 23
3.3 Agent and recipient/beneficiary
Polysemy of agent and recipient/beneficiary should in principle be infrequent: the two semantic roles have their source domain in distinct areas in the domain of space, as agent is connected with source and recipient/beneficiary with direction, two roles that tend not to merge in the source domain (see sec. 2.1.2). However, this polysemy exists in languages of different families. As in the case of polysemy of possessor and recipient/beneficiary, the polysemy of agent and beneficiary also seems to involve exclusively the type of beneficiary that also indicates recipient, that is, which is typically coded by the dative. Many Indo-European languages attest to a dative of agent, most often with verbal adjectives, as in example (55) from Latin:
(55) adeundus mihi illic est homo go:GER.NOM 1SG.DAT there be:3SG man:NOM I have to go to that man there. (lit.: That man there must be approached by me.)
23 Note that the construction in (53) is older than the one in (54) as it has cognates in other Indo-European languages. 50
Pl. Rud. 1298.
Hettrich (1990: 64-77) discusses exhaustively numerous examples from various Indo-European languages, and convincingly argues that this pattern must be inherited from Proto-Indo-European, and that is must be viewed as an extension of the recipient/beneficiary function of the Indo-European dative. Indeed, the construction typically occurs in passages such as the one quoted above, where the verbal form expresses some type of agent-oriented modality, most often obligation or necessity (cf. Bybee, Perkins and Pagliuca 1994: 177-179 for the terminology). The action must or can be performed by a human participant, which is conceived as being the target of the obligation or necessity. As recipients are targets of transfer events, they can be understood as being targets of such abstract transfers: this explains the extension of recipient/beneficiary to agent. An underlying metaphor can be set up as follows: AN OBLIGATION IS AN ENTITY TRANSFERRED. 24
Similar to the Indo-European dative, the Finnish genitive also encodes agent with the same types of agent-oriented modality (such constructions are called necessitive in Finnish grammatical descirptions), as shown in (56):
(56) Minu-n tyty-y nyt lhte-. 1SG-GEN must-3SG now leave-INF I must leave now./ I have to leave now. (from Sands & Campbell 2001: 270).
24 In Ancient Greek, too, the dative of agent occurs with verbal adjectives that indicate obligation (see Schwyzer 1950: 150, 409-410). In addition, it is also frequent with perfect verb forms, see Luraghi (2003: 65 with examples). The Greek perfect has stative meaning and, similar to nominal forms of the verb, it refers to an event as to a static entity. This might explain the extension of the dative from verbal adjectives, which are nominal forms, to other verb forms, such as the perfect, which at least partly share features of nominal forms (i.e. staticity). See further Luraghi (2013b). 51
It could be said that this usage is based on the extension of a possessor marker to the encoding of the agent, as in the case of (52) quoted above which is, however, declarative. It can also be that the genitive of agent in Finnish can be accounted for in two different ways, depending on sentence modality. 25
Extension of the dative marker to passive agent occurs in Japanese. According to current views, passive voice in Japanese has developed out of an original middle (Narrog 2010b, Toyota 2011). Passive agents already occurred in Old Japanese, but limited to poetry (Bentley 2001); apparently passive agent phrases could only contain human nouns (Vovin 2005), while they also extended to inanimates at later stages. In origin, middle forms indicated uncontrolled events, thus agent phrases must be a late (though pre-literary) addition. It is possible that a dative could be added to a middle form to indicate the human being affected by the event, and that it was later reinterpreted as an agent. This would also explain why only human agents occur at an early stage. A possible evolution would then be as follows: a beneficiary (dative of interest) is added to a verb form that indicates an uncontrolled event the notion of benefaction or interest is reinterpreted as agency when the verb form acquires passive meaning. In fact, an agent is normally the human being who primarily profits from an event. Also in this case, as for the extension of beneficiary to recipient and of recipient/beneficiary to possessor, there seems to be a default interpretation at play (the human being in whose primary interest an event is brought about is the agent), rather than a metaphor. 26
25 Sands and Campbell (2001: 275) remark that the Finnish genitive often covers the function typical of the Indo- European beneficiary dative, as in: Minu-n on hyv ol-la koto-na (1SG-GEN be-3SG good be-INF home-ESS) Its good for me to be home. In addition, the adessive case is said to indicate agent in Finnish, as in: Is teetti tuoli-n puusep- ll (father make-caus.past.3sg chair-acc carpenter-ade) Father had the chair made by a carpenter (from Huumo 1996: 80), but this is clearly a causee, rather than a passive agent. 26 This conclusion is admittedly speculative; remarkably, however, several occurrences often taken as instances of dative of agent in ancient Indo-European languages can also be explained in a similar way, as shown in Luraghi 52
3.4 Agent and comitative
This type of polysemy is also very infrequent, as discussed in Stolz (2001a). One of the few examples that I am aware of is constituted by the Vedic Sanskrit instrumental, which indicates instrument, passive agent, and to some extent also comitative, even though for this function adpositional phrases were also used already at an early stage (see Luraghi 2001a with examples; Delbrck 1867 and Wenzel 1879 for a thorough discussion of various usages of this case). Consider examples (57)-(59):
(57) hat vrtrm ... ndrena shasa yuj smite:IMPT.2PL V.:ACC I.:INSTR mighty companion:INSTR Smite Vrtra, with the strong Indra as a companion. (Rv. i 23 9 ); (58) abh jahi rakssah prvatena PREV smite:IMPT.2SG R.:ACC.PL stone:INSTR Hit the Raksasa with the stone club. (Rv. vii 104 19 ); (59) hat ndrena panayah sayadhve smite:PTCP.PF.NOM.PL Indra:INSTR P.:NOM.PL lie-down:PRS.MID.2PL Smitten by Indra you, Panis, will sink into death. (Rv. x 108 4 ).
The fact that this type of polysemy is usually avoided can be connected to the nature of prototypical comitative, which refers to a human entity that accomplishes an action together with another agent, as in (60):
(2013b). As Spike Gildea (p.c.) points out, recipient/agent polysemy is also found in the Cariban language family, in which it might go back to Proto-Cariban. 53
(60) Mary prepared dinner with Janet.
The normal interpretation of (60) is that both Mary and Janet are actively involved in the event of preparing dinner, that is, that they are both agents. Given the possible co-occurrence of a comitative and an agent, polysemy is usually avoided. Indeed, the Classical Sanskrit patter, in which the instrumental case indicates the roles of instrument and passive agent, while the comitative is encoded through an adpositional phrase with the instrumental, is typical of the Indo-European languages which retain the instrumental case and display the instrument-agent polysemy. For example, in Russian and most other Slavic languages the instrumental case encodes the instrument and the passive agent, while the comitative is encoded by the preposition s (or its cognates) and an NP in the instrumental.
3.5 Comitative and possessor
The roles of comitative and possessor are usually kept distinct. This lack of polysemy seems connected with another tendency of comitative markers in possessive constructions, described in Stolz (2001b), by which they tend to be re-interpreted as marking the possessum, rather that the possessor. Stolz remarks that such extension of comitatives is based on a conceptualization by which being with something equals to possessing something. He gives several examples that involve predicative possession. Example (61) and (62) are from Swahili and Portuguese (both from Stolz 2001b):
(61) Hamisi a-na kitabu Hamisi 3SG-with book 54
Hamsa has a book. (62) e sentouse porque estava com medo and sit:PRET.3SG+REFL because be:IMPF.3SG with fear And he sat down because he was afraid.
Given the features of prototypical possession reviewed in sec. 2.2.2, this type of extension follows quite naturally as a consequence of the assumption that possessor and possessee need to be in close spatial proximity: if possession implies spatial proximity, then it can also be the case that spatial proximity indicates possession. Thus, the same metaphors that account for the Locative Schema for possession (see sec. 2.2.2) also account for extension of comitatives to possesse markers.
3.6 Comitative recipient/beneficiary
In sec. 3.4 and 3.5, we have seen that polysemy involving the comitative and other human roles is infrequent. This is the case also for polysemy involving the comitative and the recipient/beneficiary, which, according to Stolz et al. (2006) is virtually inexistent. If we consider the regions of space that serve as the source domain for the latter roles and for the comitative, lack of polysemy may look quite surprising: after all, comitatives arise from locatives, and recipients arise from allatives, two spatial semantic roles that are very frequently encoded by means of the same, polysemous marker. However, the relation between comitative and its source domain is completely different with respect to the relation of recipient with its source domain. As I have remarked in sec. 2.2.1, markers of spatial relations (typically direction) which extend to recipient usually also preserve their spatial meaning. But this is not the case for markers of location which extend to comitative, and which generally lose any spatial meaning they may have had before. Even though the comitative relation, 55
being with somebody implies physical proximity, comitative markers seem to be disconnected from the domain of space.
3.7 Experiencer
As remarked in sec. 2.2.4, the experiencer role is encoded in a variety of ways, its trademark being that it is always coded as something else. As repeatedly noted in the literature, cross-linguistic evidence seems to point toward the non-existence of a dedicated case for experiencer: language after language provide evidence for case systems that include cases for primarily coding agent, patient, recipient, possessor, various spatial relations, instrument and other inanimate relations, but case systems typically do not include a case whose primary function is to code experiencer. Virtually the only exception is constituted by the so-called affective case of some Daghestanian languages, whose primary function is indeed that of encoding experiencers with a sub-set of experiencer verbs (some perception, modal and mental verbs). This case exists in Andic languages and in one Lezgic language, Tsakhur (see Comrie, van den Berg 2006, Daniel and Ganenkov 2008). In origin, the affective was a spatial case, as shown by some vestigial usages with toponyms (see Ganenkov 2006); in Tsakhur it is also used for the addressee of speech. The latter is a semantic role typically encoded in the same was as recipient cross-linguistically, through an extension of allative markers (see Daniel this volume), based on the Conduit metaphor (Reddy 1979), a complex metaphor that accounts for communication: IDEAS/MEANINGS ARE OBJECTS; WORDS/SENTENCES ARE CONTAINERS; COMMUNICATION IS SENDING. The occurrence of such a case with the primary function of encoding the experiencer role points toward a conceptualization of experiencers as locations or endpoints of 56
direction. 27
In the discussion in sec. 2.2.4, I have argued that such spatial metaphors could indeed be mediated by other uses of the markers involved. Thus, adessive coding, such as in Finnish, can be mediated by the possessor schema, rather than depend on direct mapping from the domain of space. This conclusion is supported by the extension of the Possessor Schema to experiencer situations in many languages. More in general, the large extent to which experiencers can be encoded through the dative case as recipients/beneficiaries in numerous genetically unrelated languages are very likely to owe to a gestalt effect, by which common features of the two roles (humanness, limited control, being sentient) are focused upon and account for experiencers being understood as similar to recipients/beneficiaries. In sum, the fact that dedicated case marking for experiencers is cross- linguistically so infrequent mirrors a more basic cognitive gap regarding this role, that is, that there are no basic metaphors that map spatial relations onto experiencers. Note that languages abound of spatial metaphors connected with the domain of experience: often, emotions/sensations are conceived as containers, or as things contained in body organs, for example. In addition, the Daghestanian affective case shows that mapping of the spatial domain onto the domain of experience is possible. Apparently, however, no path leading from space to experience is so widespread and arguably entrenched in cognition as metaphors that relate other roles to space, such as AGENTS ARE ORIGINS, POSSESSORS ARE PLACES or RECIPIENTS ARE DESTINATIONS. Possibly, this depends on the fact that metaphors that could account for experiencer (Experiencers are Places/Containers, Experiencers are Destinations) more frequently map space onto other semantic roles (possessor, beneficiary, recipient). A reason for this may well be that the domain of experience is very complex and multifaceted, as it includes very different types of situations (bodily sensations,
27 One can also detect some similarity between the experiencer role and that of addressee, based on the fact that an addressee of speech is normally assumed to hear what s/he is being told, that is, an addressee is normally also a perceiver. 57
emotions, perception, mental processes, intetionality), and that experiencer predicates may be stative or inchoative (see sec. 2.2.4): for this reason, the encoding of the experiencer roles could hardly be based on a single metaphor.
3.8 Interim summary 2
In sections 3.1 through 3.7 I have surveyed some (often unexpected) patterns of polysemy among semantic roles that are typical of human beings. I started with possible polysemies involving possessor: this semantic roles frequently merges with agent on the one hand (sec. 3.1), or with recipient/beneficiary on the other hand (sec. 3.2). I have argued that the existence of two competing metaphors that account for the mapping of space onto the domain of possession, that is POSSESSORS ARE ORIGINS and POSSESSORS ARE PLACES (see sec 2.2.5 and Fig. 4), accounts for possible polysemy with the semantic role of agent, often encoded as a source, and the semantic roles of recipient/beneficiary, often encoded as a direction (see sec. 2.2.1). Based on this latter property of the encoding of agent and recipient/beneficiary, polysemy between these two roles is unexpected. However, as I have shown in sec. 3.3, such polysemy occurs in various languages, in which it seems to be connected with modal or aspectual features of certain predicates, or with (possibly earlier) features of the voice system. The semantic role of comitative is apparently very unlikely to develop any type of polysemy with other human roles. Polysemy of agent and comitative is attested in some ancient Indo-European languages, but seems to be infrequent elsewhere (sec. 3.4), while polysemy of comitative and possessor seems to be excluded by the tendency for (inanimate) comitatives to merge with possessee (sec. 3.5). Finally, Polysemy of comitative and recipient/beneficiary is virtually inexistent, as has already been pointed out in previous studies, in spite of the fact that these two roles are encoded with markers whose original spatial meanings come from neighboring domains. However, as I had already pointed out in sec. 2.2.3, comitative markers do not retain their original 58
spatial meaning, so connection with recipient/beneficiary or, more in general, other roles cannot be provided by the source domain. The experiencer role shows frequent polysemies with several other roles, owing to the multifaceted nature of the domain of experience (sec. 3.7).
4. Discussion
In the above sections, I investigated how space can serve as the source domain for semantic roles typically taken by human beings, and how these can extend through metaphor to other semantic roles, typical of inanimate entities such as instrument or involving human activities such as purpose. The basic spatial relations of source, location and direction each offer a preferred source domain for one or more human relations. In particular, source offers the preferred source domain for agent, while direction offers the preferred source domain for recipient. Given the infrequence of a merger involving source and direction, metaphors mapping agent and recipient onto source and direction remain maximally distinct. And indeed, polysemy of agent and recipient is attested, but it is not manifested in means of encoding (cases, adpositions) connected with both spatial roles. For example, in sec. 3.3, I have shown that agents can be encoded as recipients with verbal forms expressing obligation or necessity. In such cases, it is not the domain of space to provide the source for metaphorical extension. Rather, extension operates within the domain of human relations. Location offers the source domain for several human roles, among which possessor and comitative. Remarkably, the relation of these two roles to their source domain is different, in that locative markers that extend to possessors usually retain their spatial meaning, while this does not happen in the case of locative markers extending to comitative. Beside location, possessor has another frequent source in the domain of space, that is, source/origin. This connects it with agent, thus providing a basis for polysemy among human roles. 59
Notably, source and location are two spatial roles which are typically kept distinct: as remarked in sec. 2.1.2, in spite of relatively frequent extension of source markers to location, the original meaning seems to necessarily be lost when the second develops, in very much the same way as in the case of location and comitative. Thus, languages may rely on both metaphors for indicating possessor, with the two ensuing constructions specialized for different functions (as in the case of external, i.e. dative, possessors in several languages of Europe, see Haspelmath 1999). Similar to possessor, beneficiary also has different sources in the domain of spatial relations, as it may originate from direction markers or from markers of location or path. In the first case, beneficiary typically exhibits polysemy with recipient, while in the second it does not, but it tends to entertain polysemic relations with purpose and/or cause. Remarkably, while purpose is also a frequent meaning of recipient markers, cause is not: cause tends to be related with source in the domain of space, rather than with direction. Although unexpected, the polysemy of cause and purpose is frequent. More research is certainly needed on the mapping of space onto these two semantic roles; from the examples analyzed in sec. 2.3.3, polysemy of cause and purpose seems to arise from markers that do not encode source or direction in the domain of space, but rather location or path, such as English for and Finnish vuoksi. Such markers typically also encode beneficiary, but not recipient (see sec. 2.2.6). Location and path are both located in the same region between source and direction in the domain of space. As argued in sec. 2.1.1, they are atelic, as they do not indicate a change of state, contrary to source and direction, which imply that an entity is starting or ending motion. They provide for an intermediate, non-directional area, which is the source domain for the semantic role reason. The latter role in its turn provides the area of overlap for source and cause. Experiencer seems to have no direct source in the source domain of space: various markers of spatial relations that can code this role in different languages seem to be motivated by its connection to other human roles, in particular recipient and possessor. In any case, cases and adpositions that might point to a relation between the experiencer role and the spatial domain 60
typically encode location or direction, but not source. This is not surprising: experiencers are usually conceived as being affected by experiential situations, rather than initiators thereof. The polysemies reviewed above all show that source and the other spatial semantic roles tend not to merge, and that, accordingly, human roles directly connected with them remain separated. Unexpected polysemies such as the one involving agent and recipient do not have their direct origin in the domain of space, but are based on extensions in the domain of human relations. The predictions in (1) regarding the direction of semantic extension, from space to human relations, to inanimate ones are generally borne out, even though the case that space is mapped directly onto an inanimate relation, with no intermediate human role, is not infrequent. The most obvious example is the mapping of space onto time; in addition, abstract roles such as cause and purpose also often involve direct mapping from the source domain of space. The semantic role of instrument is most interesting in this respect as the instrumental meaning is often acquired by location markers, possibly by a generalization based on the specific shape of certain instruments, which makes them equally available for instrument and for location expressions. Instrument is also involved in the only clear case of semantic extension that contradicts the scale in (1) as instrumental cases or adpositions often extend their meaning to the agent role. Tentatively, I explain this development as based on a referential metonymy rather than a metaphor. The fact that source tends not to display polysemies that involve the other spatial roles does not mean that morphemes encoding source cannot undergo semantic extensions in the domain of space. As I have shown in 2.1.2, extension from source to location and direction is well attested; contrary to many other semantic extensions reviewed in this paper, though, it does not result in polysemy as the original meaning disappears when the new one(s) develop. The data attesting to this development do not point to a gradual change: to the contrary, the fact that polysemy of source and other spatial roles is so infrequent, while the corresponding semantic extension seems to be rather frequent, rather indicates that the overlapping of the source and target meanings is avoided. The 61
same happens for semantic extension of location markers to the comitative role. In this case as well, polysemy is virtually inexistent, and the original locative meaning of comitative markers can often only be reconstructed. In spite of the pervasive role of metaphor, other mechanisms also operate in semantic extension and are responsible for semantic role polysemy. Beside the referential metonymy mentioned above concerning the extension of instrument to agent, I also described predicational metonymies in the case of beneficiary and recipient and of recipient and possessor polysemy. In addition, I argued that extension of various spatial relations, involving containment or support, to the coding of instrument owes to analogical extension. More precisely, I argued that some instruments are indeed shaped as containers or means of support. This offers a conceptualization limited to a set of instruments, which is then analogically extended to all possible instruments. In Fig. 8 below I give a tentative representation of connections among the semantic roles surveyed in this paper, on the background of the source domain of space. The map should better be viewed as being tri-dimensional, thus allowing for a representation of polysemy among non- contiguous roles, such as cause and purpose. In addition, the role of experiencer should be viewed as indirectly connected with the background source domain through other human roles. 62
LOCATION
SOURCE
POSSESSOR
DIRECTION COMITATIVE EXPERIENCER
CAUSE BENEFICIARY AGENT INSTRUMENT RECIPIENT FORCE PURPOSE
INTERMEDIARY
PATH
REASON
FIGURE 8. A MAP OF SEMANTIC ROLES
In Fig. 8, semantic roles are arranged in a way as to account for patterns of polysemy and semantic extension reviewed in the paper. I added arrows in cases in which directionality is made reasonably clear from attested diachronic developments, as for example from comitative to instrument. Extension from possessor to experiencer is documented in the generalization of the 63
possessor schema for feelings and bodily sensations, while extension from possessor to agent can be assumed in cases in which passive agents can be encoded through the genitive. In addition, I tentatively indicated directionality from instrument to agent and from beneficiary to recipient. Note however that, in the latter case, diachronic developments involving the replacement of the dative case by various prepositions in the Indo-European languages seem to involve extension of a directional preposition to all meanings of the dative (that is, to beneficiary and recipient at the same time). As noted above, the four basic spatial relations on the background provide the source domain for (most) other relations. In this respect, this map can be seen as a flat representation corresponding to those in Figures 4 and 7. The four spatial relations are represented as partly overlapping, as in Fig. 3, to account for polysemy among them. This includes polysemy of source and other spatial relations, in the first place location, which is infrequently attested: as I have argued in sec. 2.1.2, merger of source with location is frequent as a diachronic development, even if it does not usually result in polysemy. Again, diachronic extension not resulting in polysemy is also indicated by extension of locatives to comitatives. Most likely, as I argued in sec. 2.2.3 regarding comitative and location, such mergers without polysemy may rest on special complexity of a specific cognitive domain (in this case, of the domain of spatial relations with human landmarks). In the case of source and other spatial roles, it points to an asymmetry regarding source, which has been noted by other authors. Tentatively, I suggest that this asymmetry might be based on an opposition between what can be conceived as being in the visual field of an observer (location, that is where the observer is, and direction, that is a point that lies on a trajectory starting from the position of the observer and developing in front of them), and what is not in the visual field (a position on the back of the observer). Finally, the fact that some diachronic mergers do not result in polysemy leaves the question open whether certain changes are abrupt, rather than gradual. Unfortunately, extension of locatives to comitatives is not fully documented in its development. Concerning extension of source to 64
location, available evidence indicates that there may (perhaps more often than one would think) some synchronic overlap at least limited to specific lexemes. For example, in Classical Latin the preposition ab from could indicate location with nouns indicating spatial regions (back, side, etc.; see Luraghi 2010a for examples and further reference, and 2011 on developments in Romance). Even if I cannot pursue this issue further here, this partial overlap seems to indicate that extension of source morphemes to location is indeed a gradual process. 28
28 A quantitative study of spatial prepositions in a parallel corpus consisting of the Ancient Greek, Latin, Gothic and Old Church Slavonic Gospels shows that some degree of overlap of source with other spatial relations is not limited to Latin, but also attested in the other languages. This, together with cross-linguistically frequent ablative-locative transfers described in sec. 2.1.2, might indicate that such overlap is more frequent than commonly thought, see Eckhoff, Thomason & de Swart (2013). 65
Appendix - List of metaphors discussed in the text AGENCY IS A THING TRANSFERRED AGENTS ARE ORIGINS AN INSTRUMENT IS A COMPANION AN INTERMEDIARY IS A CHANNEL CAUSES ARE ORIGINS OF EVENTS COMMUNICATION IS SENDING EXISTENCE IS POSITION IN SPACE EXPERIENCERS ARE PLACES (OR CONTAINERS) FOR FEELINGS/SENSATIONS EXPERIENCERS ARE POSSESSORS OF SENSATIONS EXPERIENCERS ARE RECIPIENTS OF FEELINGS/SENSATIONS FEELINGS/SENSATIONS ARE THINGS IDEAS/MEANINGS ARE OBJECTS MATERIAL IS AN INSTRUMENT FOR CREATING OBJECTS OBJECTS COME OUT OF SUBSTANCE PHYSICAL VICINITY IS CONTROL POSSESSORS ARE ORIGINS POSSESSORS ARE PLACES PURPOSES ARE DESTINATIONS RECIPIENTS ARE DESTINATIONS SENSATIONS ARE THINGS POSSESSED THE MIND IS A CONTAINER/ CHANNEL THE VISUAL FIELD IS A CONTAINER TIME IS SPACE WORDS/SENTENCES ARE CONTAINERS 66
References
Amritavalli R. 2004. Experiencer datives in Kannada. In P. Bhaskararao and K.V. Subbarao (eds.), 1-24. Anderson, John M. 1971. The Grammar of Case: Towards a Localistic Theory. Cambridge: Cambridge University Press. Bennett, David C. 1989. Ablative-locative transfers: evidence from Slovene and Serbo-Croat. Oxford Slavonic Papers 22: 133-154. Bentley, John R. 2001. A Descriptive Grammar of Early Old Japanese Prose. Leiden: Brill. Bickel, Balthasar. 2004. The syntax of experiencers in the Himalayas. In P. Bhaskararao and K.V. Subbarao (eds.), 77-112. Blake, Barry J. 1977. Case marking in Australian languages. No. 23 in Linguistic Series. Canberra: Australian Institute of Aboriginal Studies. Blansitt, E. L. Jr. 1988. Datives and allatives. In Studies in syntactic typology, M. Hammond, E. Moravcsik and J. Wirth (eds), 173-191. Cambridge: Cambridge University Press. Bloomfield, Leonard. 1933. Language. New York: Holt. Comrie, Bernard and Helma van den Berg. 2006. Experiencer constructions in Daghestanian languages. In Semantic Role Universals and Argument Linking: Theoretical, Typological, and Psycholinguistic Perspectives, I. Bornkessel, M. Schlesewsky, B. Comrie, A. D. Friederici (eds), 127-154. Berlin: Mouton de Gruyter. Coseriu, Eugenio. 1970. Einfhrung in die Strukturelle Betrachtung des Wortschatzes. Tbingen: Narr. Creissel, Denis, and Cline Mounole. 2011. Animacy and spatial cases: Typological tendencies and the case of Basque. In S. Kittil, K. Vsti and J. Ylikoski (eds.),155-182. 67
Croft, William. 1991. Syntactic Categories and Grammatical Relations. Chicago: Chicago University Press. Croft, William. 2001. Radical Construction Grammar: Syntactic Theory in Typological Perspective. Oxford: Oxford University Press. Croft, William. 2003. The role of domains in the interpretation of metaphors and metonymies. In Metaphor and metonymy in comparison and contrast, R. Dirven, R. Prings (eds), 161- 205. Berlin/New York: Mouton de Gruyter. Daniel, Michael A. and Dmitri S. Ganenkov. 2008. Case marking in Daghestanian. In Handbook of Case, Malchukov, Andrej and Spencer, Andrew (eds). Oxford: Oxford University Press. Daniel, Michael A. this volume. Addressee of speech: a Goal or (merely) a metaphor? Delbrck, Berthold. 1867. Ablativ localis instrumentalis im Altindischen. Berlin: Dmmler. Dik, Simon C. 1997. The Theory of Functional Grammar. 2 vols. 2nd edn. Edited by K. Hengeveld. Berlin: Mouton De Gruyter. Dixon, R.M.W. 2002. Australian languages: their nature and development. Cambridge: Cambridge University Press. Eckhoff, Hanne Martine, Olga A. Thomason and Peter de Swart. 2013. Mapping out the Source domain. Studies in Language 37/2: 302355 Erelt, Mati and Helle Metslang. 2006. Estonian clause patterns from Finno-Ugric to standard average European. Linguistica Uralica 42/4: 254-266. Evans, Vyvyan. 2004. The structure of time. Amsterdam/Philadelphia: Benjamins. Ganenkov, Dmitry S. 2006. Experiencer coding in Nakh-Daghestanian. In Case, Valency and Transitivity, L. Kulikov, A. Malchukov, P. de Swart, 179-202. Amsterdam/Philadelphia: Benjamins. Geeraerts, Dirk, and Hubert Cuyckens (eds). 2007. The Oxford Handbook of Cognitive. Linguistics. Oxford: Oxford University Press. 68
Goldberg, Adele. 1995. Constructions. A Construction Grammar approach to argument structure. Chicago: University of Chicago Press. Grnthal, Riho. 2003. Finnic Adpositions and Cases in Change. Helsinki: Societ Finno-ougrienne. Haspelmath, Martin. 1997. From space to time. Mnchen/ Newcastel: Lincom Europa. Haspelmath, Martin. 1999. External possession in a European areal perspective. In External Possession, D. Payne and I. Barshi (eds), 109-135. Amsterdam/Philadelphia: Benjamins. Havers, Wilhelm. 1911. Untersuchungen zur Kasussyntax der indogermanischen Sprachen. Strassburg: Trbner. Heine Bernd. 1997a. Auxiliaries. Oxford: Oxford University Press. Heine, Bernd. 1997b. Cognitive Foundations of Grammar. Oxford: Oxford University Press. Heine, Bernd and Tania Kuteva. 2002. World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Heine, Bernd, Ulrike Claudi and Friederike Hnnemeyer. 1991. Grammaticalization. A Conceptual Framework. Chicago: University of Chicago Press. Hettrich, Heinrich. 1990. Der Agens in passivischen Stzen altindogermanischer Sprachen. Gttingen: Vandenhoek & Ruprecht. Hjelmslev, Louis. 1935/37. La catgorie des cas. Acta Jutlandica 7.2, 9.1. Huumo, Tuomas. 1996. Domain shifts and the grammaticalization of case: A case study of the Finnish adessive. Folia Linguistica Historica 17: 73-95. Itkonen, Erkki, Aulis J. Joki, Reino Peltola. 1978. Suomen kielen etymologinen sanakirja. Vol. 6. Helsinki: Suomalais-Ugrilainen Seura. Janda, Laura. 1993. A Geography of Case Semantics: The Czech Dative and the Russian instrumental. Berlin/New York: Mouton de Gruyter Kenesei, Istvn, Robert M. Vago and Anna Fenyvesi. 1997. Hungarian. London: Routledge. Kittil, Seppo. 2010. Beneficiary coding in Finnish. In S. Kittil and F. Ziga (eds), 245-270. 69
Kittil, Seppo. 2005. Recipient-prominence vs. beneficiary-prominence. Linguistic Typology 9/2: 269-297. Kittil, Seppo and Fernando Ziga. 2010. Introduction: Benefaction and malefaction form a cross- linguistic perspective. In S. Kittil and F. Zuiga (eds),1-28. Kittil, Seppo, Katja Vsti and Jussi Ylikoski (eds). 2011. Case, Animacy and Semantic Roles. Amsterdam/Philadelphia: Benjamins. Kornfilt, Jaklin. 1997. Turkish. London: Routledge. Kuryowicz, Jerzy. 1964. Inflectional Categories of Indo-European. Heidelberg: Winter. Lakoff, George. 1977. Linguistic Gestalts. Papers from the Thirteenth Regional Meeting of the Chicago Linguistic Society. Chicago: Chicago Linguistic Society, 236-287. Lakoff, George and Mark Johnson. 1980. Metaphors we Live By. Chicago: University of Chicago Press. Lakoff, George and Mark Turner. 1989. More than Cool Reason: A Field Guide to Poetic Metaphor. Chicago: University of Chicago Press. Lakoff, George, Jane Espenson and Alan Schwartz. 1991. The Master Metaphor List. Available on line at: http://araw.mede.uic.edu/~alansz/metaphor/METAPHORLIST.pdf. Langacker, Ronald W. 1987. Foundations of Cognitive Grammar. Vol. 1, Theoretical Prerequisites. Stanford: Stanford University Press. Lehmann, Christian. 2 2002. Thoughts on Grammaticalization. Second, revised edition. Arbeitspapiere des Seminars fr Sprachwissenschaft der Universitt Erfurt 9. Erfurt: Seminar fr Sprachwissenschaft der Universitt. Lehmann, Christian and Yong-Min Shin. 2005. The functional domain of concomitance. A typological study of instrumental and comitative relations. In Typological studies in participation, Ch. Lehmann (ed), 9-104. Berlin: Akademie. Leumann, Manu, J. B. Hofmann and Anton Szantyr. 1965. Lateinische Syntax und Silistik. 70
Mnchen: Beck. Luraghi, Silvia. 1989. Cause and Instrument expressions in Classical Greek. Mnemosyne 43: 294- 308. Luraghi, Silvia. 1995. Prototypicality and agenthood in Indo-European. In Historical Linguistics 1993, H. Andersen (ed), 259-268. Amsterdam/Philadelphia: Benjamins. Luraghi, Silvia. 2000. Spatial metaphors and agenthood in Ancient Greek. In 125 Jahre Indogermanistik in Graz, Ch. Zinko & M. Offisch (eds), 283-298. Graz: Leykam. Luraghi, Silvia. 2001a. Syncretism and the classification of semantic roles. Sprachtypologie und Universalienforschung 54/1: 35-51. Luraghi, Silvia. 2001b. Some remarks on instrument, comitative, and agent in Indo-European. Sprachtypologie und Universalienforschung 54/4: 385-401. Luraghi, Silvia. 2003. On the Meaning of Prepositions and Cases. A Study of the Expression of Semantic Roles in Ancient Greek. Amsterdan/Philadelphia: Benjamins. Luraghi, Silvia. 2005a. Prepositions in cause expressions. In Papers on Grammar 12/2: 609-619. Luraghi, Silvia. 2005b. Paths of Semantic Extension - From cause to beneficiary and Purpose. In Historical Linguistics 2003, M. Fortescue, E. Skafte Jensen, J. E. Mogensen, L. Schsler, 141-157. Amsterdam/Philadelphia: Benjamins. Luraghi, Silvia. 2005c. The history of the Greek preposition met: from polysemy to the creation of homonyms. Glotta 81: 130-159. Luraghi, Silvia. 2009. The evolution of local cases and their grammatical equivalent in Greek and Latin. In The Role of Semantics and Pragmatics in the Development of Case, J. Bardal and S. Celliah (eds), 283-305. Amsterdam/Philadelphia: Benjamins. Luraghi, Silvia. 2010a. Adverbial Phrases. In A New Historical Syntax of Latin, Ph. Baldi and P. Cuzzolin (eds), 19-107. Berlin/ New York: Mouton de Gruyter. Luraghi, Silvia. 2010b. Experiencer predicates in Hittite. In Ex Anatolia Lux, R. I. Kim, E. Rieken, 71
N. Oettinger, M. J. Weiss (eds), 249-264. Beech Stave Press: Ann Arbor. Luraghi, Silvia. 2010c. Where do beneficiaries come from and how do they come about? In Historical Cognitive Linguistics, K. Allan, H. Tissari, M. Winters (eds), 93-131. Berlin/New York, Mouton-De Gruyter. Luraghi, Silvia. 2011. Human landmarks in spatial expressions: from Latin to Romance. In S. Kittil (eds.), 207-234. Luraghi, Silvia. 2013a. The mapping of space onto the domain of benefaction: Beneficiaries that are not Recipients and their sources. Paper read at the 10th International Conference of the Association for Linguistic Typology (Leipzig, 15-18 August 2013). Luraghi, Silvia. 2013b. The dative of agent in Indo-European languages. Paper read at the Workshop historisch-vergleichende Sprachwissenschaft (Salzburg, 22-24 November 2013). Luraghi, Silvia and Eleonora Sausa. 2013. Hate and anger, love and desire: The construal of emotions in Homeric Greek. Paper read at the 21th International Conference on Historical Linguistics (Oslo, 5-9 August 2013). Lyons, John. 1977. Semantics. Volumes I and II. Cambridge: Cambridge University Press. Mackenzie, J. Lachlan. 1978. Ablative-locative transfers and their relevance for the theory of case- grammar. Journal of Linguistics 14: 129-375. Matisoff, James A. 1986. Heans and minds in South-East Asian languages and Engsh: an essay in the comparative lexical semantics of psycho-collocations. Cahiers de linguistique asie- orientale 15: 5-57. Narrog, Heiko. 2010a. Authors reply to The Best of Two Maps by Sander Lestrade. Linguistic Discovery 8/1: 257-258. Narrog, Heiko. 2010b. Voice and non-canonical marking in the expression of event-oriented modality A cross-linguistic study.Linguistic Typology14/1:71-126. Narrog, Heiko. This volume. The chain of increasing grammaticalization of case functions 72
Extensions of semantic roles vs. universals of grammaticalization. Nikiforidou, Kiki. 1991. The meaning of the genitive. Cognitive Linguistics 2/2: 149-205. Nikitina, Tatiana and Marianna Spano. Forthcoming. 'Behind' and 'in front' in Ancient Greek: A case study in orientation asymmetry. In Ort und Bewegung: Linguistischen Arbeiten zu Sprachen der Antike, S. Kutscher & D. Werning (eds). Berlin/New York: Mouton de Gruyter. Noonan, Michael. 2009. Patterns of development, patterns of syncretism of relational morphology in the Bodic languages. In The Role of Semantics and Pragmatics in the Development of Case, J. Bardal and S. Celliah (eds), 261-282. Amsterdam/Philadelphia: Benjamins. Pantcheva, Marina. 2010. The syntactic structure of Locations, Goals, and Sources. Linguistics 48/5: 1043-1081. Panther, Klaus-Uwe and Linda L. Thornburg. 2007. Metonymy. In D. Geeraerts and H. Cuyckens (eds), 236-263. Pennacchietti, Fabrizio. 1974. Appunti per una storia comparata dei sistemi preposizionali semitici. Annali dellIstituto Orientale di Napoli 34/2: 161-183. Peri Bhaskararao and K.V.Subbarao (eds). 2004. Non-nominative Subjects. Vol. 1. Amsterdam/Philadelphia: Benjamins. Pompeo, Flavia and Maria Carmela Benvenuto. 2011. Il genitivo in persiano antico. Un caso esemplare di categoria polisemica. Studi e Saggi Linguistici 49: 75-123. Ptz, Martin. 1996. Introduction: Language and the cognitive construal of space. In M. Ptz and R. Dirven (eds), Berlin/New York: Mouton de Gruyter, xi-xxiii. Radden, Gnter. 1989. Semantic roles. In A Users Grammar of English, R. Dirven and R. A. Geiger (eds), 421-471. Frankfurt am Mein: Peter Lang. Reddy, Michael. 1979. The Conduit Metaphor. In Metaphor and Thought, A. Ortony (ed), Cambridge: Cambridge University Press. 73
Rice, Sally and Kaori Kabata. 2007. Crosslinguistic grammaticalization patterns of the allative. Linguistic Typology 11: 451514. Rodriguez Aristar, Anthony. 1996. The relationship between dative and locative: Kurylowiczs argument from a typological perspective. Diachronica 13: 207-224. Schmidtke-Bode, Karsten. 2010. The role of benefactives and related notions in the typology of purpose clauses. In F. Ziga and S. Kittl 2010 (eds), 121-146. Schwyzer, Eduard. 1950. Griechische Grammatik. Band 2.: Syntax. Mnchen: Beck. Stolz, Thomas. 1992. Lokalkasussysteme. Wilhelmsfeld: Gottfried Egert Verlag. Stolz, Thomas. 2001a. Comitatives vs. instrumentals vs. agents. In Aspects of Typology and Universals, W. Bisang (ed), 153-174. Berlin: Akademie Verlag,. Stolz, Thomas. 2001b. To be with X is to have X: comitatives, instrumentals, locative and predicative possession. Linguistics 39/2: 321-350. Stolz, Thomas, Cornelia Stroh and Aina Urdze. 2006. On comitatives and Related Categories. A Typological Study with Special Focus on the Languages of Europe. Berlin/New York: Mouton de Gruyter. Taylor, John R. 1989a. Possessive genitives in English. Linguistics 27, 4/6: 663-86. Taylor, John R. 1989b. Linguistic categorization: Prototypes in linguistic theory. Oxford: Clarendon Press. Toyota, Junichi. 2011. The Grammatical Voice in Japanese: A Typological Perspective. Cambridge: Cambridge Scholar Publishing. Tuggy, David. 2007. Schematicity. In D. Geeraerts and H. Cuyckens (eds), 82-116. Tyler, Andrea and Vyvyan Evans. 2003. The Semantics of English Prepositions: Spatial Sciences, Embodied Meaning, and Cognition. Cambridge: Cambridge University Press. Van Valin, Robert and Randy LaPolla. 1997. Syntax. Cambridge: Cambridge University Press. Vandeloise, Claude. 1994. Methodology and analysis of the preposition in. Cognitive Linguistics 74
5/2: 157-184. Verhoeven, Elisabeth. 2007. Experiential Constructions in Yucatec Maya. Amsterdam/Philadelphia: Benjamins. Vovin, Alexander. 2005. A Descriptive and Comparative Grammar of Western Old Japanese. Folkestone, Kent: Global Oriental. Wenzel, Heinrich. 1879. Ueber den instrumentalis im Rigveda, Tbingen: Laupp. Yanagida, Yuko and John Whitman. 2009. Alignment and Word Order in Old Japanese. Journal of East Asian Linguistics 18 Zlatev, Jordan. 2007. Spatial semantics. In D. Geeraerts and H. Cuyckens (eds), 318-350. Ziga, Fernando and Seppo Kittil (eds). 2010. Benefactives and Malefactives. Case Studies and Typological Perspectives. Amsterdam/Philadelphia: Benjamins. 75
List of abbreviations
1 first person 2 second person 3 third person ACC accusative ADESS adessive ADV adverb ALL allative AOR aorist ART article COP copula DAT dative FUT future GEN genitive GER gerund IMPT imperative INF infinitive INSTR instrumental Lm landmark MID middle NEG negation NOM nominative P passive PART partitive 76