Abstract
It has been observed that “words do not have meanings but rather meanings have words.” But even more, words and phrases, usually thought of as the basic units of meaning, actually should be seen as having only usages within contexts. These contexts themselves are the elements of communication that activate mental and bodily states and processes, and can be properly thought of as the construal of meaning. Sometimes these contexts of usage are shaped by phonological or grammatical patterns (see B. Bergen), sometimes by sociological factors such as reciprocity (see E. Goffman), and of course most commonly by lexical (usually radial) patterns and categories (traditionally presented as “semantic domains”). Cognitive linguists (e.g., Lakoff, Turner) have been hinting at something like this for a long time, but a clear restatement of what constitutes semantics is overdue.
Introduction
It has frequently been observed that “words do not have meanings but rather meanings have words.” But even more, words and phrases, usually thought of as the basic units of meaning, actually might better be seen as having usages within contexts. These contexts themselves are the elements of communication that activate mental images or ideas or mental and bodily states and processes, and can thus be properly thought of as construing meaning. This article will therefore attempt to relocate “meaning” from the traditional term or phrase to contexts of usage.
Sometimes these contexts of usage are determined by phonological patterns, sometimes by grammatical structures, at other times by sociological factors such as reciprocity, and of course most commonly by lexical patterns and categories (traditionally presented as “semantic domains”). As we shall see, these diverse semantic architectures are constituted by the very words and phrases that derive their meaning from these contexts.
Cognitive linguists have been hinting at something like context-based meaning for a long time (e.g., Lakoff 1987; Fauconnier and Turner 2002), but a clear restatement of what constitutes semantics is overdue. This article will restrict its scope and examples mainly to the translation of biblical texts, but to the degree that the ideas are found to be valid there, they should be applicable everywhere.
What is meaning?
According to a number of cognitive linguists, meaning should be thought of as the prompting for a mental representation of some sort in the receptor—an image, a relation, or a state of mind (Fauconnier and Turner 2002, 353ff.). Thus, meaning is not to be found in a term or phrase, but in the activation of expected mental processes in the mind of the receptor in a communication event. This might often be simply prompting for a specific image. George Lakoff (2004) has famously used the example of the phrase “don’t think of an elephant” to show how easy it is to illustrate this. In fact, it is seemingly impossible not to think of some sort of elephant once the word has been uttered. Mark Turner (Fauconnier and Turner 2002) has written extensively on what he calls blending—the prompting of relations between semantic networks which can be both innovative and at the same time immediately obvious (he uses the example of “jail bait” to refer to underage girls considered from the point of view of older males, among others).
These examples so far show prompting for similar mental representations in the mind of the hearer and that of the speaker. There is abundant evidence that human beings come wired for matching the mental representations of others—a cooperative principle (Tomasello 2000, as well as Ramachandran 2011, on mirror neurons), if you will. But what about states of mind? If someone is trying to humiliate another, the goal is to prompt not only for the feelings of arrogance and superiority that are in the mind of the mocker and his friends, but also for feelings of shame and humiliation in the mind of the one being mocked. But as many students of theory of mind or cognitive perspective-taking (e.g., Hynes, et al. 2006) have shown, humans develop at a young age (around four years) the ability to guess (or logically deduce) what others are likely to be thinking or feeling in context, and distinguish it from their own thoughts and feelings. This can be extremely useful as well as hurtful, as it allows for the prompting of fully reciprocal, but non-similar states of mind.
Lexical semantics and its problems
1. Lexical semantics is too lexical
While linguistic aspects (particularly pragmatic ones) other than the lexicon have at times been included in semantic analysis, most semantics is in fact thought of as lexical semantics. An exception to this general practice can be found in the examination of deixis related to the discussion of tense and aspect. An example would be: “As John drove to his friend’s house, he remembered he had left the stove on in his kitchen.” From the point of view of the speaker, there is a moment in the past (drove . . . remembered) being referred to, but also another prior moment (had left the stove on). The moments of deixis in this utterance are not contained by any lexical uses of “this” or “that,” but rather are realized in the temporal (past/past perfect) grammatical relations represented.
As we shall see below, there are not only other semantic structures besides radial lexical structures to be accounted for, but also different semantic networks based on linguistic/semiotic/communicative elements ranging from phonological to grammatical to sociological. Thus, semantics should be seen as being distributed across the whole range of linguistic phenomena, and not restricted only to the lexicon. But, perhaps more importantly, semantic construal should be seen as located in the context of the various linguistic phenomena, and not so much in the phenomena themselves.
2. Lexical semantics is too radial
George Lakoff (1987, 83ff.) maintains that most lexical semantics tends to be too focused on radial categories. Such categories are often thought of as prototypical. Lakoff uses the example of mother, which starts out as prototypically referring to biological mothers (e.g., John’s mother), but quickly can be and is expanded to metonymic (e.g., foster mother) and even metaphorical usages (e.g., the mother lode, the mother of all battles). Thus, the central, prototypical usage gives rise to motivated (if not entirely predictable) extensions, which are envisioned as radiating out from the central prototype.
But the radial model is not the only semantic structure, even when restricted to lexical semantics. One alternative structure, according to Lakoff, is family resemblances. While he does not offer examples, the idea has been echoed by other cognitive linguists, who refer to semantic networks (preferred by many over domains) that do not have a radial structure, but which all relate to a center or hub that has no linguistic representation, as “centroid.” That is, unlike the radial structure, there is no lexical item at the center of the cluster of related lexical items.
An example of this in Spanish might be the words associated with what we would call “wood” in English. While one can be more precise by talking about “firewood,” “lumber” (unfinished for building), “timber” (felled trees in a forest), and so forth, it is all wood. But if one starts with what seems to be the general, central equivalent term in Spanish (madera), one encounters a problem: no one burns madera unless they are an arsonist. Firewood is leña, and arguably is the only kind of wood (or term-usage for wood) one can burn in a fireplace or at a campfire, for example. Yet Spanish speakers still think of madera as the central term for wood. How can it be? Madera may feel like a “central” term to speakers, but their usage patterns betray the fact that leña is not a simple radial extension of madera (as is “firewood” from “wood” in English), but occupies an independent yet related node in a semantic network with a centroid structure. The relationship of leña to madera can be represented as adjacent nodes—they are related, but neither is central in relation to the other. The graphic below attempts to represent this idea of semantic networks of nodes, with a radial structure on the left for the English terms and a centroid structure on the right for the Spanish terms. We shall see more centroid structures later with other examples.
3. Lexical semantics is often too essentialist either by statistics or by etymology/essence
Translators often complain that a word or phrase in the biblical text is not translatable. They will say, “We do not have a word for this in our language.” This way of framing the problem is lexical and essentialist—that is, it views the word or phrase as having some essential properties or belonging to a “natural” category that may not exist in the translator’s language. This locks the translator into thinking about semantics from a perspective that is too narrow. Terms and phrases have contexts of usage based on a number of different patterns, and these usage patterns can be mapped to one degree of specificity or another. It is within these “usage maps” (i.e., contexts of usage based on different structural patterns) that together we negotiate the construal of meaning.
Lexical semantics is essentialist from at least two different perspectives. The first is the traditional location of senses for a term or phrase in context by statistical analysis. If a term or phrase occurs one thousand times in what appears to be a given context, then we can mark that term in these contexts as having a certain meaning. This is not unlike what we are calling usage maps. The other perspective one could call “core” essentialist. The idea in this case would be that there is a core (or original) meaning for a term or phrase with radial extensions, and the core meaning would be found by weighing the senses (finding their radial core or central node) rather than arrived at by counting or by statistical analysis.
But in both cases, talk about senses and meanings of terms is too term/phrase oriented. It is better (for various reasons dealt with throughout this article) to talk about contexts prompting for the construal of meaning, and about terms and phrases (and many other linguistic elements) simply having usage maps corresponding to those contextual construals. Whether by weighing or counting, the idea of terms having senses simply is not as helpful as the idea that it is in the contexts of usage that meaning is construed. Freeing terms and phrases from having basic senses in themselves will also free translators to look for solutions to semantic problems beyond the terms and phrases themselves to the contexts which will have to be reconstructed in the translating (or receptor) language.
The term nefesh is often translated as “soul,” though in the plural it is often simply taken as referring to (a number of) people (e.g., Gen 12.5). By statistical count, something like “life principle” is usually understood, thus the common translation of “soul.” But there are a few cases that point to the etymology of the term as “neck” or “throat,” and derivationally as “appetite” (especially in Ugaritic sources). One can see this in the NRSV rendering of Ps 69.1, for example: “waters have come up to my neck.” Statistical semantic analysis would list these embodied references as a different sense of the term, and move on. But essentialist semantics would take the “throat” meaning as central and derive the “life principle” and “person” meanings from there (one breathes through the throat, thus “life,” or even “person” can easily be derived). But for contextualists, these are not arguments that need to be settled. Rather, there is a usage map that connects all these “meanings” (or usages) in a semantic network without any need to prioritize any of them, either by counting (statistics) or weighing (core or essentialist meaning).
The term hesed presents another similar example. By statistical measures, the term is most often taken to mean “faithfulness,” or “loyalty.” It is slightly more nuanced, as usually the term is understood as applying to someone with superior social standing in a relationship. Thus, a common subject of this hesed is God, and the people are the objects, or beneficiaries. A prototypical case is Ps 136, where NRSV translates hesed as “steadfast love.” But the essentialist semanticists focus largely on one occurrence in Isa 40.6. Here, the NRSV text reads, “All people are grass; their constancy is like the flower of the field.”
The statistical semanticist will argue that in Isa 40.6 we have a metaphor, and while the social aspect of “faithfulness” does not apply to plants, it simply does not matter, since metaphors only borrow what they need for their comparisons. The essentialist will reply that it is simpler to start with the core meaning of constancy, and when the term is used in contexts of personal interaction (including with God), you then get the added social dimension for free (thus “faithfulness”). Contextualists again will claim that it does not matter; there are simply different usage patterns and contextually determined meanings, and not different meanings (or senses) for the term itself. This is not to say that one cannot detect semantic networks of various kinds in contextual usage maps. But these are usually “just so” stories that we tell ourselves after the fact. Again, according to Lakoff, they reveal motivated, not predictable, patterns.
Semantic domains
Mixed domains/networks: “key” < > llave
The word “key” will doubtless bring to mind for a vast majority of English speakers the image of a key to a lock or door. This is due to prototype effects, or the fact that for many words there is in the minds of the speakers of the language a “best” representative of that word (Lakoff 1987, 58ff.). But there are many other kinds of keys—to pianos, computers, maps, codes, music, and even to problems. If one were to try to map an essentialist domain for the word in English, there would likely emerge some idea about access, maybe from small to large. Going back to the prototype, the key to the house is small, and gives access to the larger house, or locked box, or whatever.
But we get into trouble as soon as we try to map this domain over into Spanish. The key to a house or lock in Spanish is indeed llave. But llave also means the handle to a faucet, or even a wrench. So, the “domain” of llave is more like something you handle and turn to open and shut things, or turn them on and off. Worse, keyboard keys (computer, piano) are teclas, or button-like objects you push or manipulate with your fingers. There is a domain in Spanish having to do with access as such, but here the word is clave (related historically to llave, but now independent; Bascom 2000, 8–9). The point here is that the word “domain” does not seem to be appropriate to describe the various semantic relations and structures that are interconnected networks, often leading back to embodied concepts (turning hands, pushing fingers, little–big access). And as always, it is contexts and usage that drive the networks, not the other way around (though once established, networks may well facilitate further development of the semantic structures).
Here is a simpler representation of the cross-linguistic semantic relations charted above:
wrench (llave) faucet (llave) key = llave (key) tecla (key) clave
Mismatched cross-linguistic domains: tson and seh in Biblical Hebrew
Among the various terms for sheep and goats in Biblical Hebrew, two terms stand out for the translator. These are tson and seh. What makes them unusual is that they do not distinguish between sheep and goats. While tson usually refers to a group of animals, and seh to an individual in the tson, they can both refer to either sheep or goats (or both). This is not a particular problem for Biblical Hebrew (or any other language), but presents unique problems for the translator. Unless there are good contextual clues to make a definitive choice, one is simply forced to decide without adequate evidence. Interpreters have usually opted for “sheep.” Yet the lexicon entries maintain the category of “flock of sheep and/or goats” for tson, and “a sheep or a goat” for seh, which, to English ears at least, allows an unaccustomed amount of ambiguity. We are faced with the fact that apparently the words tson and seh in Biblical Hebrew prompted for categories that only vaguely correspond to those in English, for example (such as “flock”).
Other linguistic and communicative spheres where semantic structures can be found
1. Phonology and phonaesthemes
Benjamin Bergen (2004), among others, has drawn attention to an interesting semantic structure which seems to be based on phonology: phonaesthemes. Examples are given below, along with approximate semantic mappings. One can easily see that many English terms that begin with gl- have something to do with light. It is hard to classify those that prompt for a reflection of light and those that are imagining light emanating from a source. This emanation was considered to be from the human eye itself (e.g., “glower”), reflecting a concept going back to ancient times of how the eye interacted in the world with light. Meanwhile, many terms beginning with sn- are nearly onomatopoetic, having to do with noses (or snouts). And many sl- terms have a couple of semantic nodes (at times overlapping), one being “slippery” and the other a negative evaluation. Of course, these phonologically-based semantic structures do not govern all instances of the sound patterns. Gl-, sn-, and sl- function as phonaesthemes at times, while at others they simply are the beginnings of unrelated terms (e.g., gloat, snip, slake).
gl-
glitter, glisten, gleam, glint, glimmer, gloss (reflection of light) glower, glare, glance, glimpse, glow, glory (transmission of light)
sn-
snarl, snout, snicker, sniff, snort, snore (nose)
sl-
slide, slick, sled, slip (slippery) slack, slouch, sludge, slime, slosh, slash, sloppy, slug, sluggard, slattern, slut, slang, sly, slither, slow, sloth, sleepy, sleet, slipshod, slay, slant, slovenly, slough, slum, slump, slobber, slaver, slur, slog (negative)
2. Grammar
If one starts in Spanish, lexical semantics works quite well for this example. For an inside corner, rincón is the word one should use. But for an outside corner, esquina is the term. In both cases, the preposition is the same: en (“in” or “on”). But when we switch to English, something interesting happens. Here the term for both inside and outside corner is simply “corner.” So how do English speakers represent the difference between the two types of corners? By the preposition used with the term “corner.” For an inside corner, “in the corner” is used. But for an outside corner, it is “on the corner.” One can argue that prepositions are also in the lexicon, but of course it is the combination of the preposition and the noun that prompts for the image of either an inside or an outside corner.
My friends and companions stand aloof from my affliction,and my neighbors stand far off. (NRSV) My friends and neighbors will not come near me, because of my sores;even my family keeps away from me. (GNB)
This example again seems like a simple term-level mapping mismatch from Biblical Hebrew to English, and might have been listed above with other examples from Hebrew (like hesed, nefesh). But arguably it is not the variable meaning of the waw that makes the translators into English choose either “and” or “even” as their renderings. It is rather the structure of the semantic parallelism itself. In that semantic context, the waw changes from being a simple conjunction to meaning something more like “and more so” (see Kugel 1981 and Alter 1985). The best rendering in English in each case of semantic parallelism will depend on the logical relation between the first and second lines of the couplet. In this case, the renderings of “friends,” “neighbors,” “companions,” and “kinsmen” depend on how seriously the translators take rendering the waw according to the fundamental rule of semantic parallelism in Biblical Hebrew poetry (“A, and what is more, B”; again, see Kugel and Alter).
3. Pragmatics/relevance—implication/implicature
The difference between grammatical (semantic) implication and pragmatic implicature is often used by relevance theory advocates as an example of how relevance operates. A particularly good example of this can be found in Matt 1.25:
but [he] had no marital relations with her until she had borne a son.
But he had no sexual relations with her before she gave birth to her son.
Y sin haber tenido relaciones conyugales, ella dio a luz a su hijo. And without having had conjugal relations, she gave birth to her son.
The simple changes from “until,” to “before,” to “without having had” shows a range of meaning from grammatical-semantic implication (NRSV) through logical (relevant) but weaker implicature (GNB), to finally no necessary relation at all between the Greek clauses (DHH, which appears to be trying to avoid, either by implication or by implicature, saying anything about Mary’s virginal status after Jesus’ birth). And all are defensible renderings of the Greek ἕως, understood and translated variously in the three different ways outlined above. This might simply be seen as a mismatch of usage mapping from Greek, but interestingly crosses over from grammar (implication) to pragmatics (implicature/no necessary relation) in the various translations.
It should be noted that the location of semantics more or less exclusively within the lexicon is not universal, even among linguists. Relevance theorists tend to see much of semantics taking place at a pragmatic level, for example. Semioticians locate semantics within a communication system, or, as they would put it, within a system of signs. Various cognitive linguists (notably Lakoff 1987) try to account for semantics within the broader contexts, or frames, of communication and human interaction (following Goffman 1974).
4. Sociology—Reciprocity
Perhaps the most unusual or unexpected area in which one can find semantic structures is one that is technically not even under the umbrella of linguistics (even taken broadly). But sociology (or cultural studies) nevertheless does deal with human interaction and communication, and so might well be expected to contain semantic structures within its purview. And indeed, it does have such structures, it would seem.
A common concept in cultural studies is the idea of reciprocity. It can be seen in the matching of forms (or at least register) in the greeting and response of salutations. For example:
“Hey” <> “Hey” or “Hi” “Hi” <> “Hi” or “Hello” “Good morning/afternoon/evening” <> “Good morning/afternoon/evening” or perhaps “Hello”
One can see from these examples that there is some room for flexibility, but people generally stick to the same social register when replying to a salutation. Reciprocity goes far beyond salutations. Goffman has described the negotiation of reciprocity in almost mercantile terms. He uses the concept of the exchange of various kinds of tokens within the contexts of human interaction. Goffman deserves to be quoted at length, even more so than we do below. But the idea is clear from the following quote, in any case:
The acts or events, that is, the sign-vehicles or tokens which carry ceremonial messages, are remarkably various in character. They may be linguistic, as when an individual makes a statement of praise or depreciation regarding self or other, and does so in a particular language and intonation (Garvin and Riesenberg 1952); gestural, as when the physical bearing of an individual conveys insolence or obsequiousness; spatial, as when an individual precedes another through the door, or sits on his right instead of his left; task-embedded, as when an individual accepts a task graciously and performs it in the presence of others with aplomb and dexterity; part of the communication structure, as when an individual speaks more frequently than the others, or receives more attentiveness than they do. (Goffman 1956, 477)
What concerns us here are mostly linguistic tokens, but framed in the context of a communication event, itself framed by the context of human interaction.
A specific case is the usage of bitte in German. Most semantic analyses of bitte would have at least two senses listed. A request best glossed as “please” (or “excuse me”), and a quite different meaning (or sense) of “you’re welcome.” But if we apply Goffman’s theory of social tokens, we can better understand the semantic structure of the usages of the term. It is much less helpful to simply say there are two senses here than it is to talk of the usage in terms of the negotiation of social relations via the exchange of linguistic tokens. To put it succinctly, it appears that in the context of a certain kind of social interaction, at the beginning one often gives a “Bitte!” and a little later gets a “Bitte!” back.
Conclusion
Is the idea of contextual semantics an example of circular reasoning?
One criticism of the thesis of this article is that contextual semantics (with varied architectures) is not required to explain the phenomena, and that traditional semantics can handle all the examples well enough. But the contextual approach brings all the various phenomena together under a simpler explanation, while at the same time allowing for a greater diversity in what is considered semantics than has heretofore been the case.
A more serious objection is that contextual semantics seems circular. That is, contexts are made up of elements (words, phrases, etc.), and at the same time these elements each provide the contexts for the others. Thus, for example, “John forcefully struck the key(s)” immediately prompts for the image of someone playing a piano. One does not normally strike computer keyboard keys, nor does it make sense to talk of striking keys to maps, codes, and so forth. So, “strike” provides the context for interpreting “key(s).” But “key(s)” just as much provides the context for “strike.” This “strike” is not that of the one prompted for when talking of striking a match, or of workers going out on strike, etc. It is a particular action made with piano keys, while the other uses of “strike” prompt for different construals of meaning based on matches or workers or whatever. So, “strike” does for “key” what “key” does for “strike,” without one taking precedence over the other.
But this circularity is not logical; rather it is structural. Contexts are accessed as wholes, with each part contributing to the context for the other parts, and vice versa. Thus, the negotiation of the whole comes together in the interaction of the parts, without one necessarily being more fundamental or critical than the others. The elements determine the contexts, and the contexts define the elements. And whether we see the elements as words or phrases, or as contexts, simply depends on when and where in the process we choose to analyze them.
What is the architecture of meaning negotiation within linguistic and social communities?
The major points of this article have been that:
Meaning is the prompting for an image, a relation, an idea, or even a state of mind, and this prompting is always negotiated.
The construal of meaning is contextual, and is not encapsulated in a term or phrase.
Contexts are not limited to those of lexical items, but can and do include networks formed within virtually all areas of linguistics, from phonology to grammar, pragmatics, and relevance.
Ultimately, the contexts of meaning construal go beyond traditional linguistics, and include semiotics, communication, and human interaction (e.g., sociology) in general.
Most of the time we can ignore this complexity and treat meaning as if it were lexical, radial, statistical (or essentialist), and as having only minor differences or mismatches cross-linguistically (in either term/phrase or domain). But for translators, especially those working with languages in language families that are dissimilar from their source (in this article, often shown between Semitic and Indo-European, but at times even within these families), the traditional approaches to semantics begin to break down. Anyone who has worked with translation will recognize the limitations of the unhelpful idea that one must simply find the right word or phrase in order to get a translation “right.” Freed to use all the communicative resources for the contextual construal of meaning, translators may well find both creative and apt solutions to traditionally vexing problems.