Mechanisms and pathways of semantic change
The descriptive approach to semantic change described in the previous section is far from ideal. The categories are vague and purely taxonomic, and offer no explanatory insight into the conditions under which meaning change happens. They are also highly informal, and lack clear criteria for their application. This is particularly true for ameliorization and pejorization: whether a meaning change is in a positive or negative direction will often depend on little more than the subjective judgement of the investigator. For example, the change of English knight from the meaning ‘boy, servant’ to the meaning referring to the aristocrat is often described as ameliorization. But this is only the case if the latter meaning is evaluatively superior to the former, a judgement that not everyone would share. Another problem with the traditional categories is that they are also either too powerful or not powerful enough: in the case of generalization/specialization and ameliorization/pejorization, there are many meaning changes which do not seem to fi t, and in the case of metonymy/ metaphor, the categories seem to be able to explain any change, since we can always find some connection of similarity or contiguity between two meanings to justify their treatment as one or the other category. The traditional analysis of semantic change seems worlds away from the kinds of precise explanations that were possible in the study of sound change. Work in the second half of the twentieth century, still continuing, has tried to remedy some of these defects.
One of the features of this more recent work is the attempt to go beyond the mere description of changes, and search for causal explanations. For some scholars, the categories of metonymy and metaphor themselves are cognitively real and hence explanatory: as discussed in Chapter 7, cognitive linguists take metonymy and metaphor as basic cognitive operations which are at work throughout language. From this point of view, it is not surprising that metaphor and metonymy are prominent in meaning change, since they are also the principles behind much synchronic semantics. For someone committed to understanding language in terms of metaphor and metonymy, the contrast between explaining semantic change and just describing or classifying it collapses: the shift from germ ‘seed’ to germ ‘microbe’ is explained, not just classified, simply in virtue of being identified as a metaphor.
For others, however, this type of explanation is not satisfactory. Much modern work on semantic change stresses the role of the conventionalization of implicature as a source of semantic change. Consider the pejorization of English accident, discussed above. The original sense of ‘chance event’ would often have been used in discourse circumstances where the unfavourable nature of the event was strongly implied, as in (3):
The description ‘chance event’ would seem perfectly appropriate to the sense of accident present here. The context, however, strongly implies that the chance event is unfortunate or regrettable, since it ‘discomposes the wisest councils’. According to the conventionalization of implicature theory of semantic change, accident would have become increasingly associated with contexts like (3), and the implication that the event was unfavourable or regrettable would have been progressively strengthened. On encountering accident, speakers would increasingly associate it with con texts like (3), and assume that the event was an unfavourable one. With time, this process of strengthening would change the status of the reading ‘unfortunate/unfavourable chance event’ from an implication to part of the word’s literal meaning. The pejorization of accident is thus the result of the conventionalization of an implicature. This explanation does not deny that there is a relation of ‘contiguity’ between the notions ‘chance event’ and ‘unfavourable chance event’, and is not incompatible with an explanation based on metonymy. But it goes further, by showing the actual discourse mechanisms which allow the contiguity to become relevant.
Another instance where a conventionalization of implicature explanation is persuasive is the transfer of spatial to temporal meanings, as in the case of English go or French venir ‘come’, mentioned in the previous section. Bybee, Pagliucca and Perkins (1991) argue that description of this change as a metaphor is misleading: the transfer of ‘go’ and ‘come’ meanings to temporal uses does not spring from any general analogy or resemblance that speakers exploit between the domains of time and space. They point out that the semantic change only affects the verbs in specific grammatical contexts – not what one would expect for a metaphorical change grounded in a large-scale analogy between the temporal and spatial domains. Instead, the meaning change should be seen as the conventionalization of implicatures generated in particular uses of the spatial verbs. For go, the uses that give rise to the future meaning are those which specifically refer to an agent on a path moving toward a goal. In other words, go on its own is not enough to trigger an implicature of futurity; instead, it must be in an imperfective or progressive construction (see 9.2.2), with an allative (goal-directed) component. The future meaning will then arise as a conversational implicature of the verb under those discourse conditions. As Bybee et al. explain
The temporal meaning that comes to dominate the semantics of the construction is already present as an inference from the spatial meaning. When one moves along a path towards a goal in space, one also moves in time. The major change that takes place is the loss of the spatial meaning. Here . . . the function of expressing intention comes into play. When a speaker announces that s/he is going somewhere to do something, s/he is also announcing the intention to do that thing. Thus intention is part of the meaning from the beginning, and the only change necessary is the generalization to contexts in which an intention is expressed, but the subject is not moving spatially to fulfill that intention. (Bybee, Perkins and Pagliucca 1994: 268).
In other words, go only takes on a future reference in contexts like (4a), which could be used by someone leaving one room to go into another in order to watch television. This context is both imperfective/progressive, describing an action currently unfolding, and allative, in that it involves a specific goal. In contrast, contexts like (4b) (neither imperfective nor allative), (4c) (imperfective but not allative), and (4d) (allative but not imperfective) do not give rise to an implication of futurity:
For Bybee et al., any explanation of the semantics of go based on metaphor cannot explain these constraints. Only by attending to pragmatic, dis course-based factors can we understand the circumstances in which the change from spatial to temporal meaning takes place.
This explanation, then, puts pragmatic considerations at the heart of the understanding of semantic change: to understand why meaning changes, we should not be thinking in terms of broad cognitive operations like metaphor and metonymy, but should look instead at how inferences generated in discourse become part of lexicalized word meaning. The conventionalization of implicature theory of semantic change allows us to add a stage to our earlier generalization about the role of polysemy in semantic change. In light of the role of implicature, we can now describe the process of semantic change of a form between meanings A and B as in Figure 11.1:
The pragmatic origin of semantic change in conventionalized implicatures is now widely accepted. But what about the actual meanings themselves? Figure 11.1 tells us how a change from A > A (+ B) happens. But is it possible to generalize about the meanings A and B themselves? Can any meaning change into any other via a conventionalization of implicature, or are there regularities of semantic change which make some meanings more likely than others to appear as the As and Bs in the figure? Traugott has proposed several general tendencies of semantic change (1987, 1989; Traugott and Dasher 2002), of which we will discuss two.
The first is that ‘[m]eanings based in the external described situation’ shift to ‘meanings based in the internal (evaluative/perceptual/cognitive) described situation’ (1989: 34). We might call this semantic internalization. This category takes in many changes from the concrete to the abstract, particularly the common change whereby words for physical properties are extended to also denote mental ones. Many languages show evidence of this type of change. The change in the Old English verb felan from the meaning ‘touch’ to the meaning ‘feel’ is a case in point: the earlier meaning makes no reference to the internal psychological domain, while the later one does. The history of Greek (Indo-European; Greece) gives many other examples:
QUESTION Can you think of any analogies for these changes in your own native language?
The second tendency is also the more important. This is the tendency of subjectification, which Traugott sees as the ‘dominant tendency’ in semantic change (Traugott and Dasher 2002: 96). This is the tendency for meanings to ‘become increasingly based in the speaker’s subjective belief state/attitude toward the proposition’ (Traugott 1989: 96). Ameliorization and pejorization are prime examples of subjectification: the shift of boor from meaning ‘farmer’ to ‘crude person’ involves the speaker’s subjective attitude being imported into the meaning of the noun, displacing the previously non-evaluative sense ‘farmer’. The ground of the meaning thus shifts from the realm of public observable facts to the subjective opinion and assessment of the speaker. Another common example of this tendency is the development of epistemic modality. Epistemic modality is manifested by may and must in (6):
The speaker uses epistemic modality to indicate that they suspect (6b), or have concluded (6a), that Alfred is guilty: they assert that Alfred’s guilt is possible or likely, rather than an accepted fact. Epistemic modal meanings are thus firmly based in the speaker’s subjective belief state towards the proposition. But an examination of the history of epistemic modal verbs shows that they have not always expressed epistemic meanings. Must, for example, goes back to Old English motan, which meant ‘be able/ obliged to’, not an epistemic meaning, since it concerns the subject’s ability or obligation to do something, not the speaker’s opinion about the likelihood of their doing it. In Middle English, this quite often occurred with the adverb nedes ‘necessarily’; the following sentence, from a mid-fifteenth century text (the end of the Middle English period), illustrates this:
The contribution of nede ‘necessarily’ in this sentence makes an epistemic reading possible as an implicature: not only is a possessor of God’s grace obliged to be good (the original sense of motan ‘must’), but we have good reason to conclude that they are, in fact, good – an epistemic meaning. It would have been this sort of context that encouraged the development of the epistemic meaning of the modal through the conventionalization of the implicature. In sentences like (8) only the epistemic reading is available:
There is no question here of any obligation to be wretched; the speaker/ writer simply concludes that, as a matter of fact, the person who is ‘wicked longest’ will also be the most wretched.
We will end our discussion of semantic change by considering investigation into some more specific pathways of meaning change, specifically those involving verbs of perception and cognition. These types of verb seem particularly likely to feature in semantic change. In an influential study, Viberg (1984) found that the following strong cross-linguistic hierarchy (in a slightly more complicated version) governed the polysemies of perception verbs.
The visual modality of perception, in other words, is always the source, but never the target of processes of polysemy involving other perceptual meanings. Since, as we have seen, semantic change always proceeds via an intermediate polysemous stage, Viberg’s hierarchy can also be interpreted diachronically and taken to express constraints on the possible direction of meaning development. The hierarchy expresses the relative strengths of perception verbs cross-linguistically. A verb whose prototypical meaning belongs to a sense modality further to the left in the hierarchy can get a polysemous meaning referring to some (or all) of the modalities to the right. Verbs principally referring to sight, for instance, can have polysemies of ‘hear’, ‘touch’, ‘smell’ and ‘taste’, but verbs whose prototypical sense is in these other domains don’t take on polysemies referring to vision. Hearing verbs can develop polysemous meanings in the domains of touch, taste and smell, but verbs prototypically from these latter three domains never develop the additional meanings ‘hear’ (or ‘see’).
QUESTION What tests do you think could be used to work out what the prototypical sense of a perception verb is?
Polysemies in numerous languages conform to Viberg’s hierarchy. In Swahili (Niger-Congo; Tanzania), the basic sense of ona is ‘see’, but it can polysemously convey ‘taste’ when followed by the noun ladha ‘taste’ in the phrase ona ladha, literally ‘see taste’. The verb sikia basically means ‘hear’, but can also be used for ‘touch’. In Kurdish (Indo-European; Turkey and Middle East) dîtin ‘see’ also means ‘touch’, and can mean ‘smell’ and ‘taste’ when paired with nouns for these words in the same way as Swahili ona ladha. In Luo (Nilo-Saharan, Kenya) winjo basically means ‘hear’, and has polysemies of ‘touch’, and ‘taste’ and ‘smell’ when combined with the appropriate NPs. As discussed by Evans and Wilkins (2000), Australian Aboriginal languages also conform to Viberg’s hierarchy: to give just two of many possible examples, Mayali bekkan ‘hear’ also means ‘touch,’ and Guugu Yimidhirr nhaamaa ‘see/look’ also means ‘hear’, with an additional polysemy to ‘think’.
This Guugu Yimidhirr polysemy shows an extension out of the domain of perception proper into that of cognition. This looks rather like a pat tern of semantic extension familiar in Indo-European languages, in which verbs of cognition like ‘know’/ ‘think’/ ‘understand’ are often derived from verbs of seeing. This is seen again and again in Indo-European. For example, Proto-Indo-European *weid- ‘see’ not only gives many IE daughter languages their word for ‘see’ (Latin video, Russian videt ‘see’, etc.), but also becomes the verb for ‘know’ in many languages (Ancient Greek oida, Dutch weten, German wissen), and gives other parts of speech related to cognition in other languages (English wise, wit; Irish fios ‘knowledge’).
Sweetser (1990) explained this pattern of extension as an instance of what she termed the mind-as-body-metaphor: the persistent equation of the physical and the inner self. The world of concrete physical experience serves as an analogical model for talking about abstract mental phenomena like knowing and understanding. For example, many Indo European languages recruit their vocabulary of intellection (under standing) from that of physical holding. English grasp is a perfect illustration of this polysemy. Other examples include Latin comprehendere ‘seize’, which gives French comprendre ‘understand’ and Ancient Greek katalambanō ‘seize’, which had the metaphorical sense of ‘grasp, understand’, and is the source of Modern Greek katalabaino ‘under stand’.
For Sweetser, the mind-as-body metaphor is what lies behind the extension of ‘see’ verbs to ‘know/understand’. Vision is a concrete phenomenon which serves as the metaphorical model for knowing, conceived of as inner ‘vision’. Sweetser says that this metaphor is based in the primary status vision has as a source of information about the world: since vision gives us our most certain knowledge of what is outside us, physical seeing is a natural model for inner, mental understanding. Sweetser points out that the use of physical, vision-based vocabulary for abstract domains of intellection is by no means limited to verbs. Thus, we speak in English of ‘a clear presentation’, ‘opaque statements’ and ‘transparent ploys’, in all cases appropriating the vocabulary of visual experience to describe more abstract mental domains. Sweetser’s conclusion (1990: 45) is that the vocabulary of physical perception shows systematic metaphorical connections with the vocabulary of internal self and internal sensations. These connections are . . . highly motivated links between parallel or analogous areas of physical and internal sensation . . . . The internal self is pervasively understood in terms of the bodily external self, and is hence described by means of vocabulary drawn (either synchronically or diachronically) from the physical domain.
A natural question to ask is how widespread the mind-as-body metaphor is as a source of polysemies and semantic changes. In this light, the Guugu Yimidhirr ‘see, hear, think’ polysemy mentioned above suggests that the same vision > intellection shift may be observed in an entirely different language family. However, Evans and Wilkins (2000) found that it is in fact the meaning ‘hear’, not ‘see’, which leads to inner, cognitive meanings in Australian Aboriginal languages. A representative survey of ‘see’ words in Australian languages reveals that the extension to ‘understand’ is only found when the verb also means ‘hear’. ‘Understand’ is typically associated with the verb for ‘hear’, as in this example from Pitjantjatjara (Pama Nyungan, central Australia):
The link between hearing and cognition is a frequent polysemy pattern in Australian languages, not restricted to verbs alone. Evans and Wilkins give the following list of reflexes of the Proto-Pama-Nyungan word for ‘ear’, *pina:
Vanhove (2008) shows that the link between audition and cognition is a widespread polysemy in the languages of the world.
