You are on page 1of 22

Reprinted from Tre uppsatser om semantisk frndring hos relationella lexem. Nordlund 24.

Smskrifter frn Institutionen fr nordiska sprk i Lund. 25-46.

Lena Ekberg Transformations on image schemas and crosslinguistic polysemy*


1. Introduction
One of the central questions for the study of lexical semantics is whether, and to which degree, the paths of sense developments are predictable, given a certain lexical source. As regards developments of lexemes referring to spatial relations and physical actions, image schemas play a crucial role. In this paper I will argue that so-called transformations of image schemas may generate new, cognitively motivated meaning variants. The image-schematic structure is taken to be the most abstract basis of the lexical meaning connected to a specific linguistic form. Due to its flexible and simple nature an image schema may, and is supposed to, function as a basis for the meaning of whole categories of words. The Path-schema, e.g., is common not only to verbs of locomotion (such as go and come) but also to prepositions and adverbs expressing directionality (cf. to, from; up, down). Furthermore, cognitively founded operations on image schemas are supposed to be reflected in lexical semantics (cf. Lakoff 1987: Case Study 2). The present paper takes the image schema Spatial association as a point of departure for investigating the potential polysemy of lexemes based on a common schema. The schema Spatial association is assumed to structure the meaning of (primarily) spatial prepositions such as with, at, or transitive verbs expressing locomotion such as follow. It will be argued that lexical meaning extensions reflecting transformations of image schematic structure are cognitively motivated, and thus expected to arise cross-linguistically. The paper is written as part of the project The Dictionary of the World1, of which the
*

I am greatly indebted to Carita Paradis, Christer Platzack, Beatrice Warren, and my project colleagues Kerstin Norn and Jerker Jrborg for their valuable comments on earlier versions of this paper. The members of the project are, in addition to the author, Kerstin Norn (project leader) and Jerker Jrborg, both University of Gothenburg. The project is financed by The Bank of Sweden Tercentenary Foundation.

overarching aim is to find general principles effecting semantic change. In this project methods from both the traditional lexicography and Cognitive Semantics theory (Lakoff 1987, Dewell 1994, etc.) are combined. By means of a database of some 60.000 words (lemmas) we are able to test a number of assumptions of semantic regularities in lexical polysemy. The working hypothesis is that a restricted number of general principles operating on abstract semantic structures are responsible for an indefinite number of lexical meaning variants of a given form. The descriptive aim is thus twofold: to account for the input the abstract semantic structure functioning as a source for derivation and to account for the principles operating on the input. The assumption is that semantic structures at a certain level of abstraction, as well as the principles of meaning change, are universal devices for generating new lexical meaning variants. Such a search for universal features has not had priority within Cognitive Semantics (as pointed out by Hawkins 1993). We find it, however, important to emphasize not only language-specific semantics but universal principles that may account for (and explain) lexical meaning variants that are represented cross-linguistically. Here I will deal with one of the assumed principles the image schema transformation investigating cross-linguistic polysemy, based on transformations of various schemas that are part of the superordinate category of Spatial association. As a background I will describe some preliminary properties of the semantic input (the source concept) and the principles operating on these, focussing on the notion of image schema transformation.

2. Semantic input and principles of meaning change


2.1. Description of the semantic input What are the relevant properties of a linguistic expression sensitive to change? What is the relevant level of abstraction on which the processes of change operate? These are the main questions when trying to predict the avenues of sense development. The following is an attempt to narrow down the semantic core of the source concept in the process of creating new meanings. I assume that the semantic input is so abstract that it involves only properties defining classes of predicates (in the sense of Langacker 1987: 97), e.g. predicates denoting spatial inclusion (in, within, into etc.). I will restrict myself to relational 26

predicates (verb, prepositions, adverbs, and adjectives), since the image schema transformations accounted for in the literature relate to this group of semantic predicates. Following Langacker (1987: 183ff), a predication always has a certain scope (a base), described in terms of an experiential domain, or domain matrix (Croft 1993) within which a substructure is selected for designation (the profile). The meaning of finger tip can only be understood in relation to finger (functioning as the characterizing domain of finger tip), which in turn can only be understood in relation to hand, etc. Relating a linguistic expression to a domain thus delimits the possible interpretations of it. The profiled content (the designatum) is furthermore always construed in a specific way.2 As for relational predicates the construal amounts to in the first place the selection of the salient entities in the profile, Trajector (TR) and Landmark (LM), the former referring to the entity being assessed, the latter to the reference entitity (Langacker 1987: 231f). The choice of preposition (in vs. on) in (1) below amounts to the construal of the interconnections between TR and LM, i.e. either the TR is conceptualized as included in the LM (in ) or as located on the surface of the LM (on ) . (Henceforth TR is marked in bold whereas LM is underlined.) (1) The girl is sitting in / on the grass.

Relational predicates fall into four subgroups depending on the conceptual relation between TR and LM (Langacker 1987: 225). These are identification (IDENT), separation (SEP), association (ASSOC) and inclusion (INCL). Transitive verbs typically instantiate a separation relation, i.e. TR is separate from LM, as in Lisa hit Carl. Reflexive verbs however express identification of TR and LM (He shaved himself). Further the locative preposition in denotes inclusion, whereas on typically denotes association. The verb spread is an example of a predicate that may denote different conceptual relations in its different uses, cf. the sentences below. (2) a. Han spred gdseln i ett jmnt lager. (SEP) he spread manure-the in an even layer
2

Compare also the reasoning about the meaning of through in example (7) below.

27

b. Den nya lran spred sig snabbt. the new theory spread itself fast c. Lampan spred ett varmt sken. lamp-the spread a warm light

(IDENT) (ASSOC)

Note that in (2c) TR and LM is dependent on each other but not identical. Connected to the abstract basis of lexical meaning is, finally, the image schema. In the sense of Johnson (1987: xiv) an image schema is a recurring, dynamic pattern [] that gives coherence and structure to our experience. Image schemas are said to be embodied as they emerge through our continuous encounters with the physical environment, and are constantly operating in our perception, bodily movement through space, and physical manipulation of objects (Johnson 1987: 23). As an illustration, the concept of a path, i.e. the Path-schema, underlies the meaning of a range of lexical items, such as go, follow; up, down; to, from. The suggestion I make here is that the above-mentioned properties domain, construal, conceptual relation, and image schema systematically may undergo change, i.e. are sensitive to general principles of meaning extensions. 2.2. Principles of meaning change General meaning changes seem to fall into two major categories: those related to the domain of the lexical expression and those related to the image schematic structure underlying the more specific part of the lexical meaning. Related to the notion of domain are metaphorical and metonymical mappings. Metaphors involve mapping of the image schematic structure of one (source) domain matrix onto another, the target domain matrix. (In the example He is in the middle of life the spatial preposition gets its interpretation in the abstract domain of a human lifetime.) Metonymies, on the other hand, displace the focus from one domain onto another within the same domain matrix (cf. Strindberg is not always easy to read where the literary product is focused instead of the author). Also generalization and specialization of meaning refer to the notion of domain, the former yields enlarging, the latter narrowing of the domain. Finally, the use of predicates of motion to denote location by way of mental scanning of a path usually referred to as subjective, or abstract, motion (e.g. Langacker 1991) involves change of domains, from a spatial to a mental domain. That is, in a 28

context like The road goes from Malm to Lund the motion referred to by go entirely takes places in the mind of the language user. As regards the image schematic structure this can be either specified or transformed (see also Lakoff 1987: Case Study 2). My suggestion is that specifications of the LM (alternatively the TR) do not result in polysemy. Thus contrary to Lakoff (1987: 422), I do not regard the two uses of over in (3) as having separate mental representations. The reason is that spelling out the LM (here hill vs. wall) generally does not seem to influence the schematic meaning of the relational predicate (i.e. over denotes a curved arc-trajectory in both cases). (3) a. Sam walked over the hill. b. Sam climbed over the wall. In contrast, schema transformations, e.g. focussing on some part of the image schematic structure, are more likely to result in polysemy; cf. the sense of endpoint focus in Sam lives over the bridge. In the following section we will take a closer look at what characterizes image schema transformations.

3. Image schema transformations


An important factor in the polysemy of relational predicates is the ability of the image schema to undergo transformations, i.e. cognitively founded operations changing the structure of the schema in a nonarbitrary way. A well-known example of a transformation operating on the Path-schema is the end-point focus, which relates the two senses of over in Sam walks over the bridge and Sam lives over the bridge (Lakoff 1987: Case Study 2). The latter sense of over is consequently motivated by the transformational link between the two image schematic structures underlying the two senses. Elaborating, and improving, Lakoffs (1987) analysis of over Dewell (1994) gives a thorough account of transformations operating on the central schema of over , applicable also to schemas underlying other lexemes. Through these transformations the various senses of over are related to each other in a cognitively motivated way. The list below comprises a selection of general transformations (besides the end-point 29

focus already mentioned), accounted for in Lakoff (1987), Dewell (1994) and Ekberg (1995, 1997, 2001). Multiplex / mass: A multiplex collection of objects can be construed as a continuous mass, alternatively as a one-dimensional entity. Cf. He poured the peas / the juice out on the table. The multiplex-mass transformation is based on visual perception, since a collection of objects is perceived as a mass at a specific distance. Multiplex TR / multiplex paths: A single TR moving in various directions can be construed as tracing a multiplex of paths covering the LM. This schema is thus linked to a schema with a multiplex TR covering the LM. Cf. The guards were posted all over the hill / I walked all over the hill. Segment profiling: A path expression is used for referring to a profiled segment of a path (Dewell 1994); cf. He walked around the corner (profiling a semicircular segment), The sun came up o v er the mountains (profiling an upward trajectory), Sam fell over the cliff (profiling a downward trajectory). This transformation is an instance of a whole-for-part-relationship, and thus based on the general ability of visual and mental focussing. Reflexive trajector: The relationship that holds between two separate entities, TR and LM, may be transformed into a relationship holding between different parts of the same entity, i.e. the TR. Cf. He walked around the block / He turned around. Mental rotating: The mental rotating of an oriented path or axis has a direct counterpart in the physical manipulation of objects. An instance of this general operation is the transformation of a vertical axis into a horizontal one, lexically manifested in the use of vertical expressions to refer to movements in the horizontal plane, cf. He walked up and down the corridor (Ekberg 1997). In Lakoff (1987), Dewell (1994, 1997), Ekberg (1995) and others, meaning variants of the same lexeme related by means of image schema transformations are (implicitly or explicitly) regarded as separate senses, i.e. as instantiations of 30

polysemy. This is presumably true for some of the meaning variants, such as the end-point focus which turns a dynamic meaning into a static one incorporating the locomotion as background knowledge. For other meaning variants it is less obvious that we are dealing with psycho-linguistically separate senses. Below I will mention some examples of meaning differences assumed to be the result of image schema transformations which, in my opinion, are so fine-grained that it is questionable whether they give rise to polysemy. The alternative is that they are simply contextual modifications, i.e. instances of monosemy.3 The first example is taken from Dewell (1994: 357), who emphasizes the distinction between the two uses of over in (4), arguing that the sense (my italics) in (4a) should not be confused with closely related instances i.e. those exemplified in (4b). (The transformation responsible for the specific meaning, according to Dewell, is given in brackets.) (4) a. Sam is over the bridge now. b. Sam lives over the bridge. (subjective path) (end-point focus)

The preposition in (4a) as well as (4b) denotes end-point location as a result of the speakers (interpreters) mental scanning of the path. The difference between the two uses is that in (4a) the TR actually has traversed the path, whereas in (4b) there is no indication of a preceding locomotion of the TR. The question is whether this difference is salient enough for the two uses being separate senses in the mind of the speaker/interpreter. My suggestion is instead that both uses are instances of one and the same meaning variant, related to the central meaning of over via end-point focus. Another case where the status of the linguistic outcome of the transformation can be questioned is when a multiplex TR is construed as a mass. It is not obvious that there are different senses connected to the determiner vs. the verb in (5) and (6) depending on how the noun is construed, i.e. as a multiplex or a mass entity. It can just as well be that the alternation between multiplex/mass is

See Sandra (1998) for an insightful and very enlightened discussion of what linguists can, and cannot, tell about mental representations, among them representations of lexical senses.

31

so well integrated in the mind of the language user that it does not give rise to polysemy, perhaps not even vagueness. (5) (6) all people (MX) / all gold (MS) to spread leaflets (MX) / to spread manure (MS)

The third example is taken from another article of Dewell, (1997), where he introduces the notion of construal transformations, accounting for subjective shifts of viewpoint of a given scene. Consider the sentence (7) Lucy ran through the house.

According to Dewell (1997: 25), the prepositional phrase, through the house, is polysemous due to various possibilities to construe the LM (the house). Adopting an external vantage point, through only denotes Lucys entering and emerging from the house, whereas her passing through the inside of the house is implied. The second option is to shift from an external to an internal viewpoint (to be able to follow Lucys way inside the house) and then back to an external viewpoint, to watch her emerging out of the house. Thirdly, we may adopt an exclusively internal viewpoint on the house, in which case the outer boundaries are irrelevant for the interpretation of through. Assuming that through is vague as to the location of the endpoints of the path, the internal-viewpoint transformation imposes different, more detailed interpretations of the prepositional phrase, which then is regarded polysemous. As Dewell (p. 25) puts it, [t]he polysemy is not exactly located in either the preposition or the LM, but it arises from their syntactic combination under a particular pragmatic construal of the LM. However, this extension of the notion of polysemy is not justified without support from psycho-linguistic tests. As Sandra (1998) strongly emphasizes, intuition is insufficient when it comes to delimit the boundaries between polysemous senses and variants of a monosemous sense. (See also Tuggy (1993) for a discussion of the boundaries between ambiguity (homonomy), polysemy and vagueness.) With this reservation in mind, we turn back to the role of image schema transformations in the generation of new senses. The question to be answered is: 32

Given a specific image schema, is it possible to account for so-called motivated polysemy? I will argue that it is. If transformations of image schemas are analogs of spatial operations, such as manipulations of physical objects (see Johnson 1987, and cited research), it should be possible to predict which transformations a specific schema may undergo. In the remainder of the paper I will elaborate this argument by illustrating how transformations of the image schemas of Spatial association (Fig. 1) may generate polysemy that is crosslinguistically realized. The image schema Spatial association instantiates the basic conceptual relation of association (ASSOC) (Langacker 1987: 225). It can be regarded as a superordinate schema, embracing a variety of relational predicates encoded as prepositions, adverbs, and verbs, in turn based on various subordinate schemas. Thus the schema Spatial association does not underlie any specific lexical concept but rather generates the image schemas structuring lexical concepts.4 Lexemes denoting spatial association in its central sense are e.g. at, with, together, between, near, follow; thus on an abstract level these lexemes are all related via the schema in Fig. 1 (adopted from Langacker 1987: 230). (A and B are equal to TR and LM, whereas C is a construed, primarily spatial, region including A and B; cf. He stood near the house where C is defined as a region within the visual field including TR and LM.)

B C

Fig. 1. Spatial association.

Given the spatial proximity of two entities, there is a range of possible ways of orientation of the two. Adding a front and a back to the entities they will canonically be oriented face-to-face, as illustrated in Fig. 2 (Clark 1973,
4

The schema Spatial association shows striking similarities with what Hawkins (1993) names a profilable structure, a primitive cognitive structure which exists innately in the human mind before the process begins of acquiring substantive information in particular cognitive domains (Hawkins 1993: 339).

33

Traugott 1986). A second possibility is that they are looking in the same direction, with the consequence that either B will be in front of A (Fig. 3), or A will be in front of B. Lexical instantiations of Fig. 2 and 3, respectively, are against and follow. Finally, A and B might be oriented in an explicitly parallel fashion (Fig. 4), thus instantiating a parallel relation between A and B, exemplified by with. Together, the figures 24 illustrate various specifications of the schema Spatial association.
A B

Fig. 2. Face-to-face orientation. A B

Fig. 3. Sequential orientation. A B A B Fig. 4. Parallel orientation.

In the next section I will account for the developmental paths of each of the specific schemas above in terms of lexical polysemy.

4. Transformations on schematic specifications of Spatial association


4.1. The face-to-face schema The potential polysemy of a face-to-face-lexeme can be exemplified by the development of Old English (OE) wi and its cognates Old Swedish (OSw.) vi, Old Icelandic vir and Old High German widar, henceforth jointly referred to as the with -set. In OE the central meaning of wi was roughly that of Modern English against (Dekeyser 1990). The original meaning of OE wi (and its cognates) was directional, toward (Onions 1966), which in concrete contexts

34

became oppositional (Traugott 1985: 518).5 (The examples in (8) and (9) are from Dekeyser 1990: 39.) (8) a. wi Italia opposite Italy b. Hie gefuhtun wi Walum They fought against the Celts of Britain/theWelsh However with was also used with the meaning together with, that is, the inverse meaning of against, instantiating a parallel orientation. (9) Ferde wum man wi hine Some man travelled with him

Thus OE w i shows a meaning development from face-to-face to parallel orientation. The polysemy of the OSw. equivalent vi suggests that this change in meaning is not unmotivated. Besides the meaning to, toward (10a) and against (10b), vi (viar, wiidh) also had a comitative meaning together with (10c). (See further Ekberg 2002, from which the examples in (10) are taken.) (10) a. fil niar vi ior fell down to earth b. viar min gu against my God c. ath the wille wiidh swerige bliffua that they wanted with Sweden to-stay that they wanted to stay with Sweden In Modern Swedish the meaning variants in (10) have all disappeared from the standard language. Instead, another variant of meaning of OSw. vi, namely at, has become the central meaning of Modern Swedish vid:

The meaning of opposition is presumably derived by pragmatic strengthening of the spatial meaning toward (see Traugott 1989: 51).

35

(11)

Han str vid drren. he is standing at door-the

Taking the image schema to be the most abstract basis of lexical meaning, I suggest that the sense development of OE wi, as well as its OSw. equivalent vi, can adequately be described as an image schematic transformation operating on the schema face-to-face orientation. The schema in Fig. 2 has thus been transformed into the schema in Fig. 4, with the result that the asymmetric relation between A and B is abandoned. Also in the at-sense in (10), illustrated in Fig. 5, the asymmetric relation is discarded (or at least bleached). The atmeaning underspecifies the orientation of A and B, although it expresses an unmarked asymmetry between A and B (the one entity, A, normally being smaller and more mobile in relation to the other, B).

Fig. 5. Spatial at.

Interestingly, as regards the Icelandic equivalent v i , all the senses synchronically present in OE and OSw. are retained in present day Icelandic, i.e. to, toward (12a), against (12b), together with (12c), and at (12d). (12) a. hn brosti vi honum she smiled at him (i.e. in his direction) b. berjast vi einhvern (to) fight with someone c. vi rija mann together with two men d. standa upp vi vegg stand at a wall Finally, German wieder etymologically identical to Sw. vid, Eng. with shows another developmental path, namely from face-to-face to back or return.

36

(13) a. Er fiel und stand sofort wieder auf. He fell and stood immediately up again. b. Stell es wieder an seinen Platz! Put it back on its place! I will come back to this particular development in connection with the next set of words, namely form equivalents of Mod. Eng. again. Besides the with-set there were also other Germanic words expressing face-toface-orientation, e.g. OSw. i gen, OE ongean (ongegn), which have the same etymological origin, and belong to what I refer to as the again-set. Broadly speaking, both words show the same development, from OSw., OE against to Mod. Sw., Mod. Eng. again. Below we will take a closer look at the polysemy of OSw. i gen (see also Ekberg 2002). The central meaning of i gen during the Old Swedish period was toward, against (14a). In adverbial use i gen developed two meaning variants that at first glance seem to be unrelated both to the central meaning and to each other, namely closed and back; cf. the Swedish examples in (14b,c). (14) a. Konungin gik honum silfuer ij geen king-the walked him himself toward The king himself walked toward him b. Drren slog igen. door-the hit IGEN The door closed (with a bang). c. Besvren kom igen efter en vecka. pains-the came back after a week With the schema in Fig. 2 in mind it is, however, not surprising that a meaning of closure may develop (14b). A mental and physical movement face-to-face cannot go on forever since the path at some point is blocked by the other entity. The meaning closed arises when the two entities meeting one another close each others path, cf. Fig. 6 below.
Fig. 6. The meaning closed.

37

The meaning back, illustrated in (14c) and Fig. 7, represents another possible development of the face-to-face orientation. (Or, alternatively, a further development of the closure-meaning.)

Fig. 7. The meaning back.

As one entity closes the path of the other, the movement may only continue by making a turn and taking the same trajectory back again. Considering the polysemy of German wieder in the light of the sense development of Sw. igen again will provide us with an explanation of the meaning variants back, return, exemplified in (13) above. Since German wieder shares the same core meaning as Sw. igen, the meaning variants of wieder are motivated on the same grounds as these variants are motivated in the polysemy of the again-set. In Modern Swedish, the central meaning of igen is the repetitive one, i.e. the meaning is equivalent to the meaning of English again . The last link in the meaning chain relates the meaning back to again by emphasizing the feature of repetition present in the back-sense. More specifically, two different movements a coming and a going are reinterpreted into movements of the same kind. Thus, the schema in Fig. 7 is transformed into the schema in Fig. 8 (which is identical with the sequential schema in Fig. 3 above).

Fig. 8. The meaning again.

Summarizing, the sense development of Eng. again and Swed. igen originates in an asymmetric face-to-face-meaning and ends in a symmetric repetitive meaning. The conclusion that can be drawn from the data presented in this section is that the face-to-face-sense has a potential to develop a together with-sense, an at-sense, and a repetitive meaning, respectively. The former two possibilities are shown by the historical development of the with-set, except for the German equivalent wieder. German wieder instead follows the same track as the 38

again -set which speaks in favor of the idea that the potential sense development operates on the abstract image schema, not the more contextual interpretations of the word(s). 4.2. The parallel schema A common path of development of lexemes based on the parallel schema is from the comitative meaning (together with) to the meaning of instrument and manner. For instance, the semantic equivalents Eng. with, Swed. med, Icel. vi and Spanish con all develop both the meaning of instrument and the meaning of manner. Here the symmetric comitative meaning is turned into an asymmetric one by degrading one of the entities to Instrument or Manner, respectively; cf. the examples in (15): (15) a. He threatened her with the scissors. b. She was walking with difficulty.
INSTRUMENT MANNER

Taking into account the possibilities of generating truly inverse meanings out of the face-to-face-schema we expect the same to be true also for the parallel schema. Just to mention one example, Latin contra! against (< *com-tro-) is formed on the preposition com, cum with (Walde 1965: 251, Onions 1966: 209), thus illustrating the opposite development to Eng. with (against > together with). The schema in Fig. 4 is consequently transformed into the schema in Fig. 2.6 Interestingly enough, the polysemy of Latin contra! shows similarities with both the with-set and the again-set; cf. the selection of meaning variants from Oxford Latin Dictionary: (16) The polysemy of contra!: a. in front of one, in the eyes, face to face b. so as to face the enemy, on the other side, against one c. towards, up to, a person, so as to meet him, face to face
6

As com, cum governs ablative in Latin there might be an alternative link between the meaning with and against. Assuming that from (the meaning of the ablative case) and against encode different perspectives of the same directional sense from being source oriented and against goal oriented the against-sense may arise as a result of transforming the perspective of the direction encoded as from into against.

39

d. against or at the enemy e. in opposition, by way of objection, on the other side f. in return, by way of recompense It is especially worth noting that contra! so to speak unites the polysemy of OE ongean and German wieder, i.e. shows both the against-sense (16b, d, e) and the back-sense (16f). Further, the development of the Greek preposition met demonstrates a possible relation between (a variant of) the parallel schema and the sequential schema. In Ancient Greek met could occur with the dative, the genitive or the accusative. With the dative and the genitive met was interpreted as among, with the accusative met could mean either among or after. While the aftersense is clearly sequential, the among-sense can be regarded as a variant of the parallel schema a weaker variant since an indefinite number of entities are involved. Given this analysis, the two meaning variants of met instantiate a relation between the parallel and the sequential schema. According to Luraghi (2001) the polysemy of met is due to the character of the LM. When the LM was multiplex which was the case with the dative and the genitive the meaning turned out as among, but when the LM was a simplex, met instead meant after. The cognitive explanation is that the TR is conceptualized as included in a multiplex LM and thus among the entities referred to by the LM whereas the TR is seen as not included in a LM referring to a simplex entity.7 As accusative LMs with met could be either multiplex or simplex both meaning variants were compatible with the accusative form. At the level of image schematic structure the relation between the two meanings of met is transformational, the parallel schema (or a variant of it) is transformed into a sequential schema. In Modern Greek there are two forms deriving from Ancient met, the one meaning after, the other meaning with. Thus, not only is after and among connected via a transformational link, but there seems to be a developmental path leading from among to with, i.e. from a week to a strong variant of the parallel schema. A further indication of such a path is that Sw. med with shows
7

There is no obvious reason, however, that meaning in the latter case should turn out as behind, after rather than before.

40

the same development as Greek met , i.e. from among (in the middle, between) to with. (Swedish med is derived from PIE *me-dhi (alternatively * me-t ) formed on *me- in the middle, between, among, on which also Greek met is based (Pokorny 1959: 702; Hellquist 1957: 638)). In conclusion, the parallel symmetric schema may give rise to various asymmetric meanings such as instrument and manner, against (based on the face-to-face-schema) and after (based on the sequential schema). 4.3. The sequential schema Finally, we expect the sequential schema to be transformed into either a face-toface-schema or a parallel schema. Arguments for the former case is found in Traugott (1985, referring to Timmer 1967) who gives a wealth of examples of systematic morphological derivation where one word form expresses direction (i.e. sequentiality) and the other opposition, e.g. Arabic k "! l afa to be the successor and kalafa to be contradictory. The latter case the sequential schema turning into a parallel one is illustrated by the Sw. verb flja follow. Prototypically, flja denotes a sequential meaning, A after B (16a), which in certain contexts may be reinterpreted as a parallel locomotion, A moving together with B (16b) (see also Ekberg this volume). (16) a. Han fljde henne uppfr trappan. he followed her up stairs-the b. Han fljde henne till stationen. he followed her to station-the The original schema in Fig. 3 may thus be transformed into the schema in Fig. 4.

5. Conclusion: meaning development in terms of conceptual networks


Semantic change is usually far less systematic and general than changes in phonology, morphology and syntax. This does not mean that a search for regularities also in the area of semantics would be fruitless. Modern research within the field of historical lexical semantics and grammaticalization in fact has 41

provided arguments that meaning change is motivated by cognitive principles independent of specific languages. Above a number of examples have been given of sense developments of lexemes some related, others unrelated sharing the same abstract core of Spatial association. The meaning changes accounted for are all motivated by cognitively founded image schema transformations, and are thus (in the sense of Lakoff 1987) natural changes. In other words, it is no wonder that they turn up in language after language. It is, however, not possible to predict that they will turn up. Which meaning variants are in fact realized ultimately depend on a range of linguistic, and extralinguistic, factors, among which the image schema transformation is only one. Image schema transformations affecting the orientation of (parts of) the image schematic structure can obviously give rise to opposite meaning variants of the same lexical item. Lexemes instantiating asymmetric meanings, like face-toface-orientation, may develop explicitly symmetric meanings, cf. OE w i against, which ends up as Mod. Eng. with . On the other hand, lexemes instantiating a parallel meaning may end up with a face-to-face-meaning, such as the formation of Lat. contr "! against < com, cum with.8 In addition, meaning changes of words of spatial orientation may either strengthen or neutralize the inherent meaning. In the case of OE wi, as well as OE ongean, the spatial meaning toward is strengthened when the meaning is extended to against (opposition). On the other hand, when the Swedish cognate vid develops the meaning at, the asymmetry between TR and LM is neutralized. Finally, also symmetric meanings can be either strengthened or neutralized (weakened). The former seems to yield Greek met, developing from among to with, whereas the latter yields when the OSw. preposition med , with the primary meaning together with, is interpreted in the presence of (mz [med] thwa aff brdrom in the presence of two of the brothers) (Ekberg 2002). The lists below comprise various examples of the developmental paths of the notions of face-to-face, parallel, and sequential orientation, respectively.

Ekberg (2002) following Traugott (e.g. 1986) proposes that a principle of symmetry triggers sense developments eliminating the asymmetric relation, whereas a principle of asymmetry triggers opposite developments, from symmetric meanings to asymmetric ones.

42

(17) Face-to-face-orientation: OE wi OSw. vi Eng. again, Swed. igen; Germ. wieder Eng. confront

against > together with against > at against > back, again to face in hostility (1588), to adjoin on equal borders (1601), to parallel (1641)9

(18) Sequential orientation: Sw. flja OE ongean, OSw. i gen Arabic kalafa to be the successor

sequential > parallel orientation in a direct line with > toward > against k"!lafa to be contradictory (= sequential > face-to-face)

(19) Parallel orientation: Eng. with, Swed. med, Span. con Greek met Eng. contest

Lat. com, cum > contra!

comitative > instrument and manner among >after to assert or confirm with the witness of an oath (1579), to dispute (1603)10 with > against

The data presented suggest that general principles of sense development seem to operate on the underlying abstract semantic structure (the image schema) rather than on the more specific lexical meaning. Thus, given the notion of face-toface-orientation !underlying the meaning of OE wi, Eng. again, Swed. igen; Germ. wieder the meanings together with, back and repetition are likely to develop in preference to other meanings. I suggest that the avenues of semantic change are best described in terms of a conceptual network where the nodes are linked by transformations of the image schematic structure. Such a conceptual network is assumed to cut across both language-specific lexicons and
9

Traugott (1985) Traugott (1985)

10

43

lexical networks within a specific language. For the basic notions of spatial orientation the conceptual network would have the hypothetical structure given below:
FACE-TO-FACE

PARALLEL

NEUTRAL

SEQUENTIAL

INSTRUMENT/ MANNER MANNER

FACE-TO-FACE

SEQUENTIAL

FACE-TO-FACE

PARALLEL

Fig. 9: The conceptual network of Spatial association.

The nodes of the conceptual network illustrated in Fig. 9 are all specifications of the schema Spatial association. I regard this schema as a superordinate cognitive structure, which functions as a means of categorization of perceptual information into semantic categories, in turn structured by the specific instantiations of Spatial association. Along with other superordinate schemas some of which are still to be investigated Spatial association is a potential universal source of lexically manifested spatial relations. The generation of the more specific spatial relations is carried out by means of specification or transformation of the schema. Not only the schema but also the transformations of the schema are universal, in the sense of being language independent and motivated by human cognitive ability. Thus polysemy based on image schema transformations is universal and cross-linguistic as regards the schematic, abstract meaning underlying the rich, lexical meaning.

References
Clark, Herbert H., 1973: Space, Time, Semantics, and the Child. Cognitive Development and the Acquisition of Language, ed. by T. E. Moore. New York: Academic Press, 2763.

44

Croft, William, 1993: The role of domains in the interpretation of metaphors and metonymies. Cognitive Linguistics 44, 335370. Dekeyser, Xavier, 1990: The prepositions with, mid and again(st) in Old and Middle English. Belgian Journal of Linguistics 5, 3548. Dewell, Robert B., 1994: Over again: Image-schema transformations in semantic analysis. Cognitive Linguistics 54, 351380. Dewell, Robert B., 1997: Construal Transformations: Internal and External Viewpoints in Interpreting Containment. Lexical and Syntactical Constructions and the Construction of Meaning, ed. by M. Verspoor et al. Amsterdam: John Benjamins Publishing Company, 1732. Ekberg, Lena, 1995: Image Schemas and Lexical Polysemy: The Case of Swedish runt around. Working Papers in Linguistics 25, 2342. Ekberg, Lena, 1997: The Mental Manipulation of the Vertical Axis: How to go from up to out, or from above to behind. Lexical and Syntactical Constructions and the Construction of Meaning, ed. by M. Verspoor et al. Amsterdam: John Benjamins, 6988. Ekberg, Lena, 2001: Transformations on the Path-schema and a minimal lexicon. Studia Linguistica 55:3, 301323. Ekberg, Lena, 2002: Symmetri eller asymmetri? Kognitiva principer fr betydelseutveckling, med exempel frn fornsvenskan. Arkiv fr nordisk filologi 117, 197219. Ekberg, Lena, this volume: Semantiska frndringsprinciper. Exemplet flja. Gyri, Gbor, 2002: Semantic change and cognition. Cognitive Linguistics 132, 123166. Hawkins, Bruce, 1993: On universality and variability in the semantics of spatial adpositions. The Semantics of Prepositions. From mental Processing to Natural Language Processing, ed. by Cornelia Zelinsky-Wibbelt. Mouton de Gruyter, 327349. Hellquist, Elof, 1957 [1922]: Svensk etymologisk ordbok. 3 uppl. Lund: Gleerups frlag. Johnson, Mark, 1987: The Body in the Mind. The Bodily Basis of Meaning, Imagination and Reasoning. Chicago and London: The University of Chicago Press.

45

Lakoff, George, 1987: Women, fire, and dangerous things: What categories reveal about the mind. Chicago and London: The University of Chicago Press. Langacker, Ronald W., 1987: Foundations of Cognitive Grammar, vol. 1. Theoretical Prerequisites. Stanford, CA: Stanford University Press. Langacker, Ronald W., 1991: Concept, Image, and Symbol. The Cognitive Basis of Grammar. Berlin/New York: Mouton de Gruyter. Luraghi, Silvia, 2001: The Ancient Greek Preposition met: A Cognitive Account. Abstract for the 7th International Cognitive Linguistics Conference, July 2227 2001, University of Santa Barbara. Onions, C. T. (ed.), 1966: The Oxford Dictionary of English Etymology. Oxford: Claredon Press. Oxford Latin Dictionary, 1982, 2000. Ed. by P. G. W. Glau. Oxford at the Clarendon Press. Pokorny, Julius, 1959: Indogermanisches etymologisches Wrterbuch. Vol. I. Bern: Francke Verlag. Sandra, Dominiek, 1998: What linguists can and cant tell you about the human mind: A reply to Croft. Cognitive Linguistics 94, 361378. Timmer, Esther H., 1967: A semantic comparison of the third form in Arabic with the prepositional form in Swahili and the applied form in Kanuri. Unpublished M, A. thesis, Stanford University. Traugott, Elizabeth Closs, 1985: Confrontation and association. Papers from the 6th International Conference on Historical Linguistics, ed. by Jacek Fisiak. John Benjamins. 515526. Traugott, Elizabeth Closs, 1986: On the Origins of and and but Connectives in English. Studies in Language101, 137150. Traugott, Elizabeth Closs, 1989: On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65:1, 3155. Tuggy, David, 1993: Ambiguity, polysemy, and vagueness. Cognitive Linguistics 43, 273290. Walde, A., 1965: Lateinisches Etymologisches Wrterbuch. Heidelberg: Carl Winter Universitetsverlag.

46

You might also like