In 1979-1980 I wrote a paper that was eventually published in the SIL-UND Workpapers in the summer of 1980. That preliminary publication was, oddly enough, for many years the most often-cited thing I had written. I soon came to consider that my thinking on the topic had been only half (well, maybe three-quarters) baked when I wrote the paper, and by May of 1981 I had achieved what was (and is still) for me a more satisfying understanding. That more adequate view was reflected in this paper, which was eventually published in 1985, but has, perversely enough, never attracted the number of citations that its predecessor did. I might express some of this differently today, but I think I still agree with all that I said (which is not entirely true of the 1980 paper). This is the 1985 version of the paper, with only very minor editing changes and the addition of a couple of new footnotes.
G. K. Chesterton speaks of a certain type of mentality endemic to the human race—epidemic, I might add, in the academic community—the mentality of the philosophers. He characterizes philosophers as those who want to diagram the world and figure out how it works. “They have tried,” he says, “to put on paper a possible plan of the world, almost as if the world were not yet made” (1925: 265). He describes how they become increasingly uncomfortable and upset when presented with a diagram which looks less like a mathematical construct and more like a picture, with the distortions inherent in perspective and the movement and arbitrariness of life. He makes the following comment, which I find rather perceptive:
The temptation of the philosophers is simplicity rather than subtlety. They are always attracted by insane simplifications. (1925:135)
I think this describes a lot that has been done in linguistics: insane simplifications by linguists of this philosopher mentality, drawing up possible plans for languages almost as if the languages didn’t already exist, and getting jumpy at anything that isn’t totally formalized and absolutely rigorous. I am convinced (as was Chesterton) that this philosopher’s way of looking at things is legitimate, but partial; it is not wrong, but it is unbalanced, and we who are naturally inclined to that outlook need to learn to balance it. I suffer from the philosopher mentality—that in large part is why I am a linguist—but I want to demonstrate my allegiance to the rest of my humanity by submitting my linguistics, with its tendency to insane simplifications, to the sanity of what I know through common sense, as a speaker rather than as a linguist. The most glorious type of science is the type that can explain without explaining away, that can not only find the simplest diagrams, but can see that they are not the whole story; one that ultimately argues not from simplicity but from inclusiveness and wholeness, avoiding an either/or view and seeking a both-and view, preferring one analysis over another not because it is simpler, but because it shows us what was right with the other analysis as well as what was wrong, illuminates a wider range of facts, and fits in with what common sense, intuition, and dumb experience tell us is so. Much of the best linguistics will consist in demonstrating that Grandma was right after all.
Possessor Ascension and Dative Possessor sentences
Two insane simplifications which are very common in our linguistic heritage and are closely related to each other are the simplifying of syntax by making certain syntactic structures totally derivative of others, as if they didn’t exist in their own right, and the simplifying of semantics by assuming that functional equivalence is the same thing as semantic identity. Both are involved in an analysis that has been around for years in the tradition of Transformational Grammar, in which certain datives in Romance languages are assumed to be underlying possessives, or vice versa. Perhaps the first such analysis was in a paper by Ron Langacker (he numbers it among the sins of his youth) on French possessives which came out in Language in 1968: others have followed. The particular one I ran afoul of was a Relational Grammar analysis involving a structure called Possessor Ascension (PA), in which an “initial” (i.e., pretty much, semantic) possessor is “ascended” (i.e. “raised”) to bear a relation such as the indirect object in a higher structure such as a clause. PA has been claimed to occur in (at least) Albanian, Chamorro, French, Georgian, Stoney, Southern Tiwa, and Tzotzil (Aissen 1979; Crain 1979; Frantz 1979; Harris 1976; Allen, Frantz, Gardiner and Perlmutter 1990), all languages about which I know too little to say much, and David Perlmutter suggested one day in class that it occurs in Spanish sentences like (1-3) below.
Now I grew up speaking Spanish, and my non-philosopher, common-sense side immediately reacted against this. Unfortunately, I know of no place where this particular analysis has been claimed in print for Spanish, so that I am open to the charge of beating a dead horse, or perhaps one that was never born. But I must reply that indeed the horse I am after is very much alive. It is the assumption that paraphrase establishes semantic identity unless proven otherwise, and that syntactic differences between constructions are normally to be accounted for by transformations. PA or something like it is, in the old Transformational Grammar tradition, the obvious way to account for sentences like (1-3).
Perlmutter gave two reasons for adopting such an analysis for Spanish.
(a) PA is universally available: therefore it is simpler to use. Generalized, the argument is that transformational analyses work in similar cases elsewhere; making one work here would be better because it would be simpler in that it would not complicate the metatheory by positing new devices. I want to claim that this argument from simplicity is one of Chesterton’s insane simplifications, and should be abandoned: sanity must take priority over simplicity.
(b) PA (and other transformational analyses) account for both the identity of meaning between sentences like (1-3) and their English glosses, and their surface differences. The semantic identity is accounted for by giving identical initial (“deep”) structures to each pair, and the surface differences are accounted for by positing the syntactic change in one case but not the other. I want to claim that saying these sentences mean the same thing is another insane oversimplification, and that the way of sanity, while less simple, will show both its truth and its falsity.
Of these two considerations, I think the second is basic: if it can be shown that the sentences and their glosses do not mean the same thing in the required sense, there will be little reason to keep the analysis. So our discussion will center around that question: do sentences like (1-3) and their glosses mean the same thing, or not?
Three kinds of “meaning”
Whether or not they mean the same thing depends on what we mean by meaning. If a word like “mean” is allowed to mean only one thing, then the answer to our question is likely to be a simple “yes” or “no”. But I would claim that most words have more than one legitimate meaning, and that the word “mean” is such a word. I would like to distinguish three kinds of “meaning”. (There are others, of course.)
The first is Functional Meaning. Two expressions “mean” the same thing if they are functionally equivalent in general—if their truth conditions coincide in most cases or if they are used as translations for each other. The distinctions made by this criterion are obviously matters of degree and relative to situation and purpose.
Ball = Sphere
(But not in football.)
Synonyms and paraphrases generally “mean the same” in this sense.
The second kind of meaning is Truth-value Meaning. This is a particular exaggeration or perfection of Functional meaning, by which two expressions can be said to “mean” the same thing if and only if their truth values coincide in every (real world) case.
Some synonyms and paraphrases do, some don’t “mean the same” in this sense. As long as the same thing, event, or situation is uniquely designated by two expressions, the Truth-value criterion will show them as “meaning” the same thing. Meaning is reduced, in other words, to reference.
The third kind of meaning is Imagic Meaning. Its name sounds like it is either magic or imaginary, and some (not my readers, I trust) may think it is both, but I am going to claim that it is the primary kind, the “real“ kind of meaning. (It is essentially the same kind of meaning posited by Ronald Langacker (1979, 1980, 1982a, 1982b) in his Space Grammar model: the term “Imagic meaning” is due to Langacker. Imagic meaning makes even more fine-grained distinctions than the Truth-value criterion. It says that all of the synonymous expressions given above differ in meaning, and that in general synonyms and paraphrases do not mean the same thing, even when they designate the same entity. As a limiting case, complete semantic identity may perhaps occur, but if it does it is extremely rare. The nature of this type of meaning will be discussed more later.
The Dative Possessor sentences and their glosses differ in Truth-value meaning
(1-3) and their glosses clearly “mean the same thing” in terms of Functional meaning. Most situations, including the most common ones, which you could felicitously describe by saying Le ensuciaron el coche you could also felicitously describe by saying They got his car dirty, and in these prototypical cases you would be reporting those incidents in the normal way. Thus it is quite appropriate to use the English and Spanish sentences to translate each other.
However, they do not mean the same thing in terms of strict Truth-value meaning, because there are situations in which the Spanish sentences are true or appropriate while the English sentences are not, and other situations in which the converse obtains.
(1) can be true where the referent of the dative is driving another’s car. Its English gloss cannot be used in those situations: one must say something like They got the car dirty on him. Similarly, (2) can work where it is another’s money that gets stolen from the referent of the dative, and even (3) can be made to work. Actually, I shouldn’t say “even (3)”, since only knowledge of the Imagic meaning makes it surprising that (3) should behave similarly. Here I should simply note that (3) can be used in a situation where the hand is not the hand of the referent of the dative, but where he is responsible for its safety and integrity. I should also say that speakers will resist such a construal. Note as well that in all of these situations a Spanish sentence with an appropriate overt possessive is also possible (sentences in (4)).
Similarly, (1) cannot be true in certain situations where its gloss can. Perhaps the clearest case is when the referent of the dative is recently dead; also there is a strong tendency towards inappropriateness or untruth when he is far away or uninterested in his car. In these situations the sentence Ensuciaron su coche “They got his car dirty”, with an overt possessive and no dative, is still quite appropriate. Similarly, in (2) if the referent of the dative is recently dead, the Spanish sentence is incorrect (unless, however, the money was on the corpse). The English sentence is again correct and a Spanish sentence with a possessive is also correct. (3) is OK in the case of death, but not in the case of dismemberment. If a person’s arm has been cut off and people cut his hand (which is no longer attached to him but is still his), (3) is incorrect, though its gloss is still acceptable and the corresponding Spanish sentence with a possessive is also fine. (This kind of pattern is general with body parts: e.g., one tells the dentist Míreme la muela ‘Look at my tooth’, with a dative ‘me’, but after the dentist has extracted the tooth one says, when showing it to a friend, Mire mi muela, with a possessive ‘my’.)
So (1-3) and their English glosses do not mean the same thing, by the Truth-value criterion. I might add that both the English and the Spanish sentences test out by Lakoff’s (1970) test as vague rather than ambiguous on the relevant parameters: A mí me ensuciaron el coche y a Juan también can be appropriate if they got my car dirty on me and someone else’s on John, or any other combination of ownership; and They got my car dirty and John’s too is OK even if I was driving my car and John was off in Las Vegas and will never hear about what happened to his car.
The nature of Imagic Meaning: examples
Imagic meaning also distinguishes between the two groups of sentences, and its distinctions can be seen as holding across the board, even in the normal cases where the Functional and Truth-value criteria make no differentiation. But before going into how Imagic meaning distinguishes between the two, let me make it clearer what I mean by Imagic meaning and what are the kinds of distinctions it makes. Briefly, Imagic meaning takes into account not only reference but perspective, not only what entity is designated but what cognitive route was taken to arrive at that designation, not just what Thing or Event is viewed, but how it is viewed, what aspects of it are singled out and portrayed most saliently. The same entity may be profiled, to use Langacker’s terms, but it is profiled or foregrounded against a different base. The same scene may be pictured, but it is pictured under different images, from different viewpoints. I believe this outlook fits the phenomenon of meaning as it itself exists. It rings bells with the non-philosopher part of me. To be sure, it introduces tremendous complexities and un-simplifies things drastically, but by doing so it explicates what I know independently to be true. It helps me see the truth in the Functional and Truth-value accounts of meaning, as well as their limitations, and it helps account for intuitions and nuances of felicity that I find. This way lies sanity. I will go through some examples in the hope that they will ring bells with some of you as well.
You can call him Ronnie, Reagan, the man Hinckley tried to kill, the Commander-in-Chief of the Armed Forces, That idiot in the White House, or any of hundreds of other things. In each case the same person will be referred to, and so the Truth-value “meaning” (or the Slander-value meaning) will be the same in every case. But it certainly seems fair to me (and more enlightening as well) to say that in each case you are saying something different about the man. In Ronnie you are viewing him with perhaps unwarranted familiarity and affection, in Hinckley’s target you are viewing him as a participant in a certain historical event, and so forth. In each case you are coming at him from a different angle, by a different cognitive route, profiling him against a different base, and the meanings of the different expressions are different as a result. Similarly, the old Tetelcingo Aztec word for train, tepos-tla-wilāna-ni (iron-unspecified-pull-er) and the new one, trene, do not mean the same thing, although they designate the same thing. ‘Iron pulling machine’ does not mean the same thing as ‘train’. In Seri, spoken in Sonora, Mexico, the conventional expression designating an apple is, literally, ‘thing that the circumcised Chinaman ate’ (datum from Cathy Moser Marlett). Nobody can tell me that means the same thing as apple!
As far as I know no one would propose transformations to relate any of these expressions to the others. However, they do have identical meanings in the same sense that the pairs of expressions related by transformations do: they designate the same class of entities, and so their Truth-values are the same.
Or take the case of all vs. each. All of the sailors are drunk means the same as Each of the sailors is drunk by the Truth-value criterion. Both focus on the same sort of situation, but they view it differently. All views the sailors as a collective mass, and each views them as a collection of individuals. This can help explain both the fact that we “feel” that the two expressions mean something slightly different, and also the fact that we use a singular verb in the one case and a plural verb in the other. (George Lakoff points out [personal communication] that a Truth-value distinction can show up between all and each in sentences such as The sky is so clear that you can see all/*each of the stars in the Milky Way [* “each” would not normally be used in this context].)
Or take the case of the bottle being half full vs. being half empty. Both expressions designate the same amount of liquid, which accounts for their synonymy, but they do so by reference to mutually inverted scales, one a scale of fullness reading from empty towards full, and one a scale of emptiness reading from full towards empty. Notice that here is a strong argument for Imagic meaning: given these two scales it is easy to see that there will be possible an essentially infinite range of synonymous expressions, such as 12.8% full = 87.2% empty. I know of no way that that fact could be predicted from the mere fact that half full and half empty always have the same truth values. Also there are some infelicities and intuitions to be accounted for. David Perlmutter pointed out (personal communication) that it would be odd (though hardly untrue) to tell a waiter pouring you a glass of water to “stop when it’s half empty”. Why does that sound funny? I think that the reason is that when the glass is being filled you naturally scan in the direction of the filling: saying “stop when it’s half empty” forces you to scan in both directions at once, for no reason. Or consider the old proverb about the optimist being the man who says it’s half full, while the pessimist says it’s half empty. Given that in our culture the fullness of containers is generally desirable and encouraging whereas emptiness is discouraging, we can see why the definitions fit. If the expressions really meant exactly the same thing there would be no reason why saying one rather than the other could give us any clue to the character of the speaker. Or consider the fact that one could say, optimistically, of a cod-liver-oil bottle, It’s half empty.
In The Lord of the Rings (Tolkien 1965: 271), Barliman Butterbur, on being asked by Gandalf how many people had been “killed dead,” replied that there were “three and two.” This did not mean the same thing as if he had said simply “five”, though the Truth-value criterion would be hard put to it to show the difference. There were two prominent groups of people at Bree, the big folk and the little folk, and the innkeeper’s reply alluded to that fact. 3 plus 2 and 4 plus 1 and 5 all designate the same number, but they arrive at that number by different cognitive paths, and the results of taking those different paths are to give prominence or not to give prominence to subgroups of various sizes. Notice again that the mere fact of Truth-value correspondence between say 3 plus 2 and 5 cannot account for the fact of there being an essentially infinite number of other “synonymous” expressions, such as 7 minus 2 or 60 divided by 12, or 1.25 plus 1.75 + 2, etc., but the Imagic notion of the number scale and the relations of the locations and distances on it, does predict that fact. Note, too, the strong infelicity or oddness combined with absolute truth in exchanges such as “What’s the square root of 25?” “It’s 3 plus 2.” Knowing the Imagic meaning you can predict the Truth-value and felicity facts, but not vice versa. There is an Imagic distinction behind every Truth-value distinction but not vice versa: the Imagic distinctions hold even when the Truth-value distinctions don’t. Truth-value distinctions are just that tip of the iceberg of Imagic meaning which Truth-value judgments make visible.
Imagic meaning makes subtler connections than Functional meaning can as well as subtler distinctions that the Truth-value criterion can. Take the matter of catching jokes. Most people get “They’re called wisdom teeth because they smart” the first time through. This involves a clear perception of semantic similarity between wise and smart, even though they are synonyms by neither the Functional nor the Truth-value criterion. Or, as the most obvious example, take the kinds of connections involved in metaphors. Calling the neck of a bottle a bottle-neck involved (the first time) clear violation of Truth-value and Functional expectations, in response to a certain Imagic similarity: the similarity of shape between a bottle and an animal or human body. Calling a narrow place in the road a bottle-neck involved the same sort of thing, violating Truth-value and Functional expectations in response to Imagic shape and flow-pattern similarities. These metaphors catch on because other speakers can see the Imagic similarities involved and thus see the appropriateness of the usage. Not all Imagic similarities result in such Functional similarities of usage, but many do. Behind every Functional similarity is at least one Imagic similarity, but not vice versa. Functional similarities are just another tip of the iceberg of Imagic meaning.
Let us take a cross-linguistic case. The Spanish sentences for “he sat down” and “he lay down” are given in (5).
The Spanish sentences are overtly reflexive, and the English sentences have the particle down in them, but such “surface” differences do not hinder their semantics from being the same, do they? By the Functional criterion they clearly “mean” the same thing. They are the normal way in each language to talk about a case where one sits or lies down. By the Truth-value criterion, however, there are some intriguing differences.
Both Spanish constructions are untrue or inappropriate when the person is physically forced (while resisting) to sit down or lie down. (6a-b) are inappropriate in such a context; you must say (6c-d).
The glosses to (6a-b), in contrast to the Spanish sentences, are, at least in my dialect, quite acceptable for such a situation. The Truth-value criterion tells us there is this distinction, but it does not explain it.
Or take a case where a person lying on a couch assumes a seated posture on the couch. Se sentó (5a) is appropriate, but He sat down is not (one must, of course, say He sat up). As far as I know, there is no parallel distinction between acostarse and lie down (5.b). Again, the Truth-value criterion merely tells us there is a distinction: it does not tell us why, nor why it should hold in one case but not in the other.
I would claim that the sentences differ Imagically, however, even in the cases where the Truth-value criterion does not distinguish between them. In the prototypical act of sitting or lying there are two elements which are crucially involved in this distinction. One is a reflexivity: you do something to yourself, changing the posture of your body. The other is a motion away from the vertical posture. The English and Spanish meaning patterns give prominence each to a different one of these.
The Spanish pattern gives prominence to the reflexivity—that (naturally enough) is the significance of the Spanish pattern’s having a reflexive. The English pattern (at least for my dialect) does not give a great degree of prominence to that concept. That explains why the Spanish sentences in (6a-b) cannot and the English sentences can be used to code a situation in which the person is forced to sit or lie: in that situation he is emphatically not doing anything to himself to achieve the final posture. This contradicts the Spanish pattern, but says nothing to the English pattern.
The English pattern, on the other hand, gives prominence to the departure from the vertical: that of course is why the word down is there. The Spanish pattern does not specify such motion. That is why they differ in the case where a person assumes a sitting posture after lying on a couch: he is approximating the vertical rather than departing from it, sitting up rather than sitting down. Thus the situation contradicts a specification of the English construction but none in the Spanish construction. It also lets us see why there is no parallel distinction between acostarse and lie down: the Imagic meaning of lie involves assuming the posture furthest from the vertical. There is no normally conceivable posture from which to lie would be to approximate the vertical: you can’t lie up in the required sense, and that is why there is no Truth-value distinction here. (Note that you also cannot stand down in the appropriate sense.)
Notice that here we have a difference not only between one isolated pair, but between one system and another. There are a number of other English posture verbs which have a particle up or down on them (e.g., jump up, kneel down) and the corresponding Spanish constructions are all reflexive. We habitually take different views on these kinds of scenes, English speakers emphasizing relationship with the vertical and Spanish speakers emphasizing the reflexivity of the action. Even though the sentences are used, and rightly, as translations for each other, they do not mean the same thing.
The Dative Possessor sentences and their glosses differ Imagically
So we finally come back to the sentences with Dative Possessors in them. How are we to account for them Imagically? I would do it like this: in the prototypical cases something happens to a possession of some person and that person suffers or is somehow affected as a result of that happening. As with the sitting and lying case, the Spanish construction emphasizes one feature of such situations, and the English construction emphasizes another feature.
The English construction emphasizes possession. This is, unsurprisingly, coded by a possessive morpheme. This, again naturally, explains why the Truth-value distinctions come up in which the English construction cannot be used but the Spanish can, where the owner of the property is someone different from the person specified. The Spanish construction can be used because it is not talking about possession, whereas the English one cannot because it is talking about possession and its specifications therefore conflict.
The Spanish construction emphasizes the person’s being affected. This person is represented by a dative pronoun. Affectedness, by the way, is a semantic thread running through most if not all Spanish datives. In particular there is a group of datives which have been termed “Ethical Datives” which clearly bear this meaning: I would claim that all “possessor datives” are members of this group.
This notion explains why sentences (1) and (2) are inappropriate in the case of a man who has recently died: he is no longer affected by what happens to his possessions, though they are still his. Similarly, absent and uncaring owners are less likely to be affected and thus situations in which they are designated are less likely to be construed through the image provided by this construction.
Remember that (3) was acceptable in the case of recent death but not in the case of dismemberment. That is explained by the fact that a corpse is still affected by what is done directly to it (in this case, cutting its hand), but once a body part has been severed, what happens to it no longer affects its owner in any direct sense. For the same reason (2) is OK when the dative designates a recently dead man as long as the money was taken off the corpse—again the action can be viewed as affecting the corpse.
Many of you no doubt know that “dative possessor” constructions are supposed to be obligatory in Spanish and French with “inalienable possessions,” particularly including body parts and clothing that is worn. The explanation for this fact, at least in part, lies in the fact that body parts and worn clothing can hardly have anything happen to them without its affecting the person. And note that when those “inalienable possessions” are removed from the person (dismemberment, clothing not worn) he can suddenly be viewed are not affected by what happens to them.
Remember as well that people resist construing (3) with another possessor for the hand. That is explainable by the fact that it is so un-prototypical as to be almost inconceivable that “A” should be the person most saliently affected by what happens to “B”’s hand.
Perhaps this is a good time to return for a moment to the argument that these sentences with “dative possessors” should be accounted for transformationally because similar sentences in other languages are accounted for in the same way. I agree that similarity of structure should go along with similarity of treatment, but it is a matter of degree to what extent those sentences are similar to these. If they are no more similar than the English ones are, they may be rather different in crucial ways. However, assuming them to be the same, I will agree that they should be accounted for in the same way. What this suggests to me, however, is not that this analysis is wrong, but that the transformational analyses in the other languages are probably flawed as well, and that there may well be other factors such as Imagic specification of affectedness rather than possession involved. In any case, putting the Spanish construction in a paradigm of constructions in other languages should not be done at the cost of jerking it out of two relevant paradigms of constructions in Spanish itself, namely the one involving “ethical datives” (which were mentioned above) and one involving the simple omission of a possessor in other places where we in English would expect one. There are a number of ways in which these “dative possessors” clearly fit into those paradigms; some of the evidence is given in Tuggy (1980). The “sane” course, as opposed to the “simple” course, is to give the construction its place in all of the relevant paradigms rather than just in the one that first comes to mind or that makes your analysis work or that makes for a simpler model.
How are Dative Possessor sentences and their glosses alike in terms of Imagic meaning?
We have seen pretty well, I think, how these two kinds of constructions are not the same in meaning, but rather represent two quite different points of view on the situations they encode, in the one case specifying a person as an important participant because he is the possessor and in the other case specifying him as important because he is the person affected by the action involved. But in what senses are these constructions the same? What was right about the analysis that made them identical in meaning?
The most obvious answer is that they are the same in Functional “meaning”. They are in many cases used as Images through which to construe the same scenes. I think this is an important answer. But it depends on actual Function. Suppose English and Spanish had never come in contact, and there had been no occasion for translation between them. It still would be the case that the two kinds of constructions would be semantically similar. There is a sense in which potential Functional similarity is primary over the actual, and that potential similarity of usage depends on Imagic meaning. Two expressions “mean” the same thing to the extent that their Imagic meanings potentially could be used to “fit” the same scenes even if they were not actually so used. This similarity is in a sense extra-linguistic: it is perceived by the same faculty that enables us to perceive that two different paintings, with different perspectives, colorings, etc., are paintings of the same landscape; that different views of the face or the back of the head may “fit” the same person; that two different tastes are both honey; that C chords in different octaves or inversions all “fit” the concept of a C chord; or that both C7 and D7 chords may, in proper contexts, “fit” the idea of a dominant seventh. Similarly, we can judge when and where and to what extent the image of half empty “fits” with that of half full, or the Images of sentarse and sit down “fit” the same conceptual scenes. In this sense (1) and its English gloss are the same even apart from the fact that they are actually used as equivalents: their Imagic meanings are such that they “fit” a wide range of common conceptual scenes, even though they view them from different perspectives, with different Images.
Another relevant issue has to do with the multiple meanings typical of well-entrenched syntactic constructions and other grammatical units (Cf. Langacker 1982a:40-43, 1982b:66, Tuggy 1981:29-32). Meanings arise from (and may be viewed as consisting of) constant associations, and I am sure that there are fairly salient sub-versions of of the most inclusive version of the “dative possessor” construction in which possession is actually specified. Particularly with body-parts, I expect that there are strongly entrenched sub-schemas with with specification of possession as part of their Image, their semantic structure. To say it another way, Spanish speakers are certainly aware that the person affected in these constructions is very often the possessor, and that awareness is part of the meaning of the construction. Thus in another important sense it is right to talk of these cases as involving dative possessors. It should still be noted, however, that the extent to which such a specification or expectation of possession exists is a matter of degree: it is not carved in stone, it is not even stabilized by a morpheme coding it, so we can expect it to fluctuate quite a bit from person to person and from case to case. Also it is important not to arbitrarily cut the continuum from cases with no possession but only affectedness to cases with both affectedness and possession. Yes, there are different sub-patterns, but the differences are not likely to be as salient as is the overall pattern.
Finally, I think a good case can be made for there even being something right about the notion of a transformation linking the two constructions. Noticing the systematic correspondence between two constructions is a cognitive event, and it may achieve quite salient status in a person’s cognitive system. It may become well entrenched and conventionalized, thus becoming, under the conception of semantics I am espousing (Langacker 1982a:23 (d’), 1982b:86) semantic by definition. There may even be a directionality to the correspondence: one construction may be felt to be more basic than the other. Consider, e.g., simultaneous translators at the U.N. Part of their training is to drill such correspondences into their minds until they become automatic units: you feed in an English possessive construction and out comes a Spanish dative construction, almost without thinking. Such transformations can and do exist in people’s grammatical systems even within the same language. I expect that Passive and Dative Movement transformations in English correspond to something real in speakers’ minds, and quite possibly many Spanish speakers have an analogue to a Possessor Ascension transformation, with a dative possessor construction derived from a possessive one, or perhaps more probably they would consider the dative structure more basic and might derive possessive structures from it. Sometimes one syntactic structure does derive from another. But saying this is far from admitting the sort of insane oversimplification that would claim that, therefore, all passives derive from actives and that passives are nothing but transformed actives, or that all sentences like (1-3) derive from sentences with underlying possessors instead of datives, and are essentially nothing but changed possessive constructions. Rather than subtracting from the meaning of a construction, reducing it to mere correspondence with another construction, the transformational dimension, rightly viewed, is an enrichment of it, adding that correspondence to its other specifications. Transformational accounts are not necessarily wrong, but they are not the whole story, and they are misguided when they purport to be.
Summary: discursus on translation
In sum, then, it is a gross oversimplification to state that sentences like (1-3) and their English glosses mean the same thing. It is certainly simpler to assume that they mean the same thing, as evidenced by the paraphrase relationship between them, and to posit a rule of Possessor Ascension or Possessor Raising to account for the ‘surface’ differences between them. It would be simpler yet, and closer to the truth, to assume (as I did in Tuggy 1980) that they do not mean the same thing, arguing in the either/or mode, showing that these are datives and not possessors, involving omission rather than raising of the possessor, and assuming that therefore to posit a transformational relationship between them is totally wrong. To introduce all the complexities of Imagic meaning, including the complexities of distinguishing between the possessor and the person affected, between the schematic overall pattern and the instantiating sub-patterns, between the pattern as it exists in isolation and the pattern construed almost metalinguistically, in comparison with other patterns, is indeed less simple, but I believe it is more sane. It allows us to account for the differences where they exist and the similarities where they exist, for gross differences in meaning and for subtle nuances and degrees of felicity.
Another reason I have for wholeheartedly embracing such an account of meaning is that I am a translator, and it fits so truly what I have experienced in that role. People talk about translation as if it were essentially possible, as if you could take a quantity of meaning from one language, divest it of its form, which is quite incidental to it, and clothe it unchanged in the forms of a new language. This view can be maintained only by defining “meaning” to be “whatever is left unchanged in translation”. You cannot say exactly the same thing in two languages. You will always wind up changing some Image, even if only slightly. Let us say you want to translate into English the Spanish sentence (1), Le ensuciaron el coche. You might try They got his car dirty. You will have changed (among other things, such as specifying a male possessor rather than permitting vagueness as to gender) the Image of affectedness into the Image of possession. Suppose you try They got the car dirty on him. You have gotten a sort of Image of affectedness, but at the cost of specifying probability of non-possession. You try They got his car dirty on him—you are specifying possession again. Besides, the last two sentences sound somewhat odd, whereas the Spanish sentence sounds perfectly normal, and normality is a part of Imagic meaning. There is no way out. You cannot say the same thing in two languages. Of course, you can decide what you think are contextually the most important things that you can say easily in the target language, and say them, but that is a rather different matter from transferring the meaning intact. Translation, especially when you care about the meaning you are trying to translate, is a very humbling experience, involving constant juggling, incurring a great deal of awkwardness to say something you feel is important, while elsewhere you leave out Images or introduce them with no exegetical basis, saying something slightly misleading because it is so much easier and you hope it won’t matter in the end. I am a Bible translator, and perhaps we more than others discover that “the way of the translator is hard.” Or as Grant Showerman put it, “Translation is sin.” (Quoted in Brower 1959:271).
This, of course, fits in beautifully with a long tradition which holds that while translations are worthy and useful things, they inevitably warp the original, and yet by their very striving to be faithful to the original they come out as second-rate specimens of the target language as well. As the Italian phrase complains, comparing them to women, the faithful ones are not beautiful, and the beautiful ones are not faithful. An old Rabbinical saying has it that reading a translation is “like kissing through a veil”; Cervantes compared it to “... gazing at a Flemish tapestry with the wrong side out,” and Wilhelm von Humboldt summed it up by saying: “All translation seems to me to be an attempt to accomplish an impossible task” (the last two quotes taken from Brower 1959: 271, 275.) Yes, and sometimes you succeed in the attempt. Translation can, and it cannot be done. It is an insane oversimplification to say that meaning can be transferred from one language to another; it is an equally insane oversimplification to say that it can not. What is needed is a less simple but more sane view that includes both views, showing both why and to what extent each of them is true. I think that substituting the concept of Imagic meaning for Functional or Truth-value meaning goes a long way toward satisfying the requirements.
Aissen, Judith. 1979. “Possessor Ascension in Tzotzil.” Papers in Mayan Linguistics 88-99. Columbia, MO: Lucas Brothers.
Allen, Barbara, Donald Frantz, Donna Gardiner and David Perlmutter. 1990. “Verb agreement, possessor ascension, and multistratal representation in Southern Tiwa.” In Studies in Relational Grammar 3, ed. by Paul M. Postal and Brian D. Joseph, 321-383. Chicago: University of Chicago Press.
Brower, Reuben A. ed. 1959. On Translation. Harvard Studies in Comparative Literature No. 23. Cambridge, MA: Harvard University Press
Chesterton, G. K. 1925. The Everlasting Man. Image books edition, 1955. New York: Doubleday.
Crain, Catherine. 1979. “Advancement and Ascension to Direct Object in Chamorro.” Linguistic Notes from La Jolla 6. La Jolla, CA: UCSD Linguistics Department.
Donald G. 1979. Grammatical Relations in Universal Grammar.
Supplement to Vol. 23 of the Workpapers of the Summer Institute of Linguistics
at the University of North Dakota. Grand Forks, ND: Summer
Institute of Linguistics.
Harris, Alice. 1976. Grammatical Relations in Modern Georgian. PhD. dissertation, Harvard University.
Lakoff, George. 1970. “A Note on Vagueness and Ambiguity.” Linguistics 3.357-359.
Langacker, Ronald W. 1968. “Observations on French Possessives.” Language 44.52-75.
————. 1979. “Grammar as Image.” Linguistic Notes from La Jolla 6.88-126. La Jolla, CA: UCSD Linguistics Department.
————. 1980. “Remarks on English Aspect.” In Paul Hopper, ed., Tense and Aspect. Amsterdam: Benjamins.
————. 1982a. “Space Grammar, Analyzability, and the English Passive.” Language 58.22-80.
————. 1982b. Chapter I of Foundations of Cognitive Grammar. MS, since published (1987), Stanford CA: Stanford University Press.
————. 1987. Foundations of Cognitive Grammar. Volume I: Theoretical Prerequisites. Stanford CA: Stanford University Press.
————. 1991. Foundations of Cognitive Grammar. Volume II: Descriptive Application. Stanford CA: Stanford University Press.
Tolkien, J. R. R. 1965. The Return of the King. Vol. III of The Lord of the Rings. Boston: Houghton Mifflin.
Tuggy, David. 1980.
“¡Ethical Dative and Possessor Omission Sí, Possessor
Ascension No!” Workpapers of the Summer Institute of Linguistics
at the University of North Dakota 24.97-141. Grand Forks, ND: Summer
Institute of Linguistics.
1985. (The present paper.) “Imagic Meaning and Dative Possessors in
Spanish.” Selected Technical Articles Related to Translation
13.1-14. Dallas: Summer Institute of Linguistics.
[*] This is a slightly revised version of a paper presented in May 1981 at the Conference on Imagery and Language held at U. C. Berkeley. The author wishes to thank David Perlmutter for provoking this paper and discussing some of the ideas behind it, and Ronald Langacker for providing many of the concepts most crucial to the analysis. Thanks are also due to the Summer Institute of Linguistics for permission to reissue, in this electronic format, their 1985 publication of the paper.
The following abbreviations are used in the body of this paper: acc = accusative, dat = dative, PA = Possessor Ascension, refl = reflexive, SIL = Summer Institute of Linguistics, UCSD = University of California at San Diego
 This is not absolutely true, of course: English “possession” is an elastic enough concept to allow the possessive to be used in some situations where the “possessor” is not the owner: e.g. in a race the car Andretti or Foyt drives is “his” car, even though he may not own it.
 The “on him” (meaning something like “so as to affect him”) construction used in this sentence and in (4) below apparently does not occur in some English speakers’ dialects—a reviewer of this paper reacted strongly against it. In my linguistic system it is acceptable for colloquial speech provided that it is pronounced without accenting the pronoun, otherwise the meaning must be that “he” was dirty, and they got the car dirty by brushing against him. It feels a little more natural in intransitive sentences like “My dog (up and) died on me last week” or “It looks like the weather’s fixing to turn bad on us,” rather than in transitive sentences like these.
 A better contrast is I have a present for all of you vs. I have a present for each of you. In the first case there is one present which the group of addressees will have to share, whereas in the second case there are as many presents as addressees.