I try to rejuvenate Cleanth Brooks's old thesis about the 'heresy of paraphrase.' This I do by analysing a couple of well-known poems and by performing thought experiments of the "possible world" kind. They show that paradigmatic examples of poems are not paraphrasable. A prosaic text can be improved with the aid of a paraphrase, but a typical poem cannot. The deeper explanation for the non-rephrasability of poetry is that our understanding of it is basically tacit. In this way I hope to give Brooks's original thesis a more solid foundation.


poetry, paraphrase, tacit knowledge

1. Introducing the problem

In this article, I want to discuss the question whether or not poems can be paraphrased. The idea of the non-paraphrasability of poetry was one of the central tenets of the New Criticism. The fact that this school of criticism does not exist anymore is perhaps one of the explanations for the fact that the thesis of heresy has not been discussed much in recent years. In actual fact there has never been much systematic discussion about the thesis. But as I hope to show, it is well worth probing into and decidedly needs rejuvenation. My aim is to show that there is more than a grain of truth in the thesis that paradigmatic examples of poetry cannot be rephrased in any satisfactory manner. Further, I will try to explain the non-paraphrasability of poems by the means of my contention that our knowledge of them is typically tacit.

The darling of the New Critics, T.S. Eliot, was once asked by a lady what he meant by the line 'lady, three white leopards sat under a juniper tree.' He replied, "I meant 'lady, three white leopards sat under a juniper tree.' " The implication is of course that the line is not paraphrasable. One of Eliot's greatest admirers, the New Critic Cleanth Brooks, coined the phrase 'the heresy of paraphrase.' Another New Critic (and poet), Archibald MacLeish, said, 'A poem should not mean but be.'[1] Obviously, poems cannot be retold if they only are, but do not mean. However, they can be meaningful and at the same time impossible to rephrase satisfactorily.

It seems intuitively plausible that a poem is typically an ambiguous text. Its rhythm, style, sound, images, emotional flavour and intellectual aspects, the denotations and connotations of its words, and even its content and graphic aspects, are inseparable. The whole is bigger than the sum of its parts; a poem is a holistic phenomenon. If this is the case, then form and content cannot be separated in any clear-cut manner, so abstracting the content (the message) from the poem is a risky and none too rewarding business. Further, rhythm often plays an important role for the meaning of a poem and rhythm can hardly be mirrored by propositions. Add to this the importance of metaphors in a host of poems and the difficulties (even impossibilities) of paraphrasing metaphors, and the thesis of heresy seems quite plausible.[2] It does not get any less plausible if we consider the following: we can often improve upon everyday utterances and academic texts by clarifying them. Actually, we can clarify texts by rephrasing them. In contrast to this, it does not make any sense to say that we improve upon a poem by clarifying it, either by paraphrasing it or in other ways. It is actually an open question as to whether we can make changes to a poem without having created a new one.

I think it is high time to take a look at Cleanth Brooks's way of defending the theory of heresy. He did not deny that poems could be paraphrased up to a point, or that such a paraphrase can be useful in some contexts. What he did deny is that a paraphrase can replace a poem or capture its essence. Moreover, attempts at a complete paraphrase of poems tend to need metaphors in order to capture their meaning, but these metaphors in their turn stand in need of a paraphrase. Certainly, many poems contain propositions that are easily rephrased, but we must not mistake them for the inner core of these poems. Actually, such propositions are justified in the context of the poem as a whole, not in connection with a general paraphrase. Thus, a proposition like John Keats's 'Beauty is Truth, Truth is Beauty' gets its precise meaning and significance from its relation to the total context of the poem.[3]

2. Defining the concepts

Brooks was on the right track, and I want to use the remainder of this article to give his thesis a more solid foundation. I will begin by putting forth a stipulative definition of the concept of a non-paraphrasable text on the basis of his analysis: 'A text is non-paraphrasable if and only if a paraphrase neither can replace the text nor capture its essential meaning.' Admittedly, the notion of essential meaning is not exactly clear, I must rely on your intuitive understanding of what the essential meaning of a text is. [4] Suffice it to give the following example: In the lyrics of the Beatles' song 'She loves you,' the core (essential) meaning seems to be that A is telling B that B wrongly thinks that his girlfriend does not love him any more, but that is not right, etc.

By 'paraphrase of a poem' I mean 'a prosaic (non-poetic) rewording of a poem.' As for the concept of 'paraphrase,' it would be useful to take a look at a couple of dictionary definitions, the first one being 'rewording of a text, giving the meaning another form.'[5] Another dictionary defines the verb 'to paraphrase' as 'to explain or translate with latitude.'[6] The latter definition obviously indicates that there is a certain connection between the concept of a paraphrase on the one hand, and those of explanation (in the sense of explication) and translation on the other. This seems plausible because when we translate a text we try to say the same thing in other words, which is tantamount to an attempt to paraphrase it. (We might even say, conversely, that paraphrase is a kind of translation.) And when we explicate a text we try to make it easier to understand by using a different set of words to say the same thing but in a clearer manner. Sometimes an explication is closer to a summary, which in its turn tries to give the essential meaning of a text or an utterance.

Based on the discussion above, I define 'paraphrase' as 'rewording of a text, giving the meaning another form, in some cases clarifying the text and capturing its essential meaning.' I will stipulate that (a) a paraphrase must be able to replace the original text and/or (b) capture its essential meaning. Paraphrases of the first kind I call 'reworders,' the second kind 'probers' because they try to probe into the depth of the text as a part of a depth interpretation. Such an interpretation tries to unveil hidden essences of meaning in texts and can in the case of poetry usually be expressed in a few pointed sentences. In order to qualify as a prober, P must be a sort of a translation of the text with the aid of 'the translation manual' of a certain depth interpretation. If the interpretation were, for instance, of the deconstructivist kind, the manual would include some deconstructive rules (or anti-rules).

As for the word 'meaning,' I use it in this article in the sense (!) of 'sense plus connotations (and suchlike[7]) of words, utterances and texts', not 'reference.' In short, I am talking of the meaning we grasp in understanding and ignore completely the question of whether or not this kind of meaning can be ultimately reduced to the truth condition of sentences. The type of meaning I am discussing can both be utterer's meaning and textual meaning.[8]

In contrast to the question of the truth conditions, I will briefly discuss possible answers to the question whether it holds for all utterances, texts and suchlike, that they are somehow not fully paraphrasable. Let us look at some arguments against the thesis that no utterance can ever be paraphrased: In the first place, it seems intuitively strange to maintain that utterances like 'John and Mary went home' as uttered in workaday circumstances are not really paraphrasable or that they are as difficult to paraphrase as poems by T.S. Eliot. Secondly, there are cases where the ability to rephrase utterances is a necessary condition for understanding them.[9] If someone is asked whether he has understood a presidential candidate's speech and he responds by reiterating it verbatim, we congratulate him on his good memory. But we still lack evidence for his understanding. That evidence we can only get if he can rephrase it. Thirdly, I think that the possibility of paraphrasability is built into the very concept of a linguistic expression of an empirical theory.

Can we test the theory if we cannot paraphrase it, for instance ,if we think that its original formulation is not clear enough for testing? So some paraphrasability seems possible, which of course does not prove that there are fully paraphrasable utterances. Consider again the case of John and Mary. In more male chauvinist times than ours, paraphrasing 'John and Mary went home' as 'Mary and John went home' would perhaps not quite have captured the original. The reason is that there could have been a convention saying that a man's name must be mentioned first because males are more important than females. If that were the case, 'Mary and John went home' would have had important connotations different from the original.

3. Analysing an example

Now, it is high time that we illustrate the thesis of heresy with an example. Consider the following lines from Ezra Pound's 'Canto I'[10], which incidentally consists to a large extent of a translation of Homer's Ulysses:

And then went down to the ship,
Set keel to breakers, forth on the godly sea, and
We set up mast and sail on that swart ship,
Bore sheep aboard her and our bodies also
Heavy with weeping, and winds from sternward
Bore us out onward with bellying canvas,
Circe's this craft, the trim-coifed goddess.
Then sat we amidships, wind jamming the tiller,
Thus with stretched sail, we went over sea till day's end.
Sun to his slumber, shadows o' er all the ocean,
Came we then to the bounds of deepest water,
To the Kimmerian lands, and peopled cities
Covered with close-webbed mist, unpierced ever
With the glitter of sun-rays
Nor with stars stretched, nor looking back from heaven
Swartest night stretched over wretched men there.
The ocean flowing backward, came we then to the place
Aforesaid by Circe.[11]

Rephrasing these lines, and for that matter the whole poem, while not easy, does not seem impossible. Let us try our hand at it: 'We went down to the ships, plunged them into the sea and sailed on the ocean, which is of godly provenance. We carried our black sheep (here the expression is to be understood in a literal fashion) aboard at the same time as we boarded the vessel, while in tears. The wind was blowing our way, so the ship sailed rather quickly. This was due to the magical powers of Circe, who is a goddess.' And so on and so forth.

Let us assume that we show this paraphrase to a stranger, who neither knows Pound's poem nor Homer's epic and ask him what kind of a text this is. Is it a paraphrase of a novel, everyday discourse in a pre-modern civilisation, a fairy-tale, an epic or a modern poem? Would it not require quite a leap of imagination on his part to determine that this is actually from the two last categories?

Contrast this to a paraphrase of a scientific paper or book. In most cases, we would have no problem identifying the nature of the object of the paraphrase. A paraphrase of Einstein's book on the theory of relativity could hardly be mistaken for the paraphrase of a novel or a book about politics. So rephrasing a scientific text is probably more rewarding than paraphrasing a poem, a fact that ought to count in favour of the thesis of heresy. The attempted paraphrase of 'Canto I' also counts in favour of the thesis. The paraphrase seems like an empty shell, far removed from the complex organicity of Pound's poem. It cannot capture the particular rhythm of the poem, which suggests among other things the movements of the ship. Further, Pound's use of Homer creates a set of associations, which are hardly paraphrasable in any fruitful manner. And just the very Verfremdungseffekt (the effect of estrangement) of taking this part of Homer and wedding it to a completely different type of discourse, as Pound does at the end of the poem[12], is also not easily captured by a paraphrase. Pound's poem can therefore be subsumed under my definition of a non-paraphrasable text.

4. Visiting possible worlds

To be sure, I do not doubt that there are cases where poems could be paraphrased in a similar fashion as non-poetic texts. Take a look at the following example, a poem by William Carlos Williams:

This Is Just to Say
I have eaten
the plums
that were in
the icebox
and which
you were probably
for breakfast
Forgive me
they were delicious
so sweet
and so cold[13]

We can try to paraphrase the poem (make a rewording of it) in the following fashion: 'I ate the plums that were in the icebox. In all probability you kept them in order to eat them at breakfast. I am sorry, the plums were tasty, they were very sweet and cold.' This is a reworder of the poem and seems to be a good enough replacement (it would be an even better replacement if we made its rhythm conform to that of the poem). Therefore, this is an adequate reworder, given my stipulation. Admittedly, the task at hand is a bit more difficult if we want to make a prober out of it. Such a prober could be based on a depth interpretation of the poem as conveying the message that we human beings are easily led into temptation, but in most cases that is something we have to live with it. C'est la vie!

The problem is the obvious one that the number of possible depth interpretations is infinite and, mutatis mutandis, the same holds for the number of probers. But who says that there is no such thing as the correct depth interpretation? And who says that such depth interpretations are always useful? Brooks certainly would have contested their usefulness. In our case the result of such a contesting is something Brooks would not have liked; namely that if a poem can only be reworded, not given any fruitful depth interpretation, then it is paraphrasable for all intents and purposes. So if there is such a thing as paraphrasable text, Williams's poem is, unless its soft rhythm plays a decisive role in its meaning.

Now, does this mean that the thesis of heresy is wrong? No, not at all. In the first place, by just looking at a paraphrase of the poem we would not have the slightest clue that it was a paraphrase of a poem, unless told so. So it seems that a paraphrase of this poem has certain features in common with paraphrases of more 'poetical' poems such as Pound's 'Canto 1.' Secondly and much more importantly: If there is such a thing as a paraphrasable poem, Williams's poem cannot be a paradigmatic example of a poem for the simple reason that such poems are parasitic upon non-paraphrasable ones. Poems like the one by Williams get their identity from being challenges to high modernistic poetry or even poetry as such. It is almost as if the American poet were teasing us by implicitly saying, 'Look, I have presented this text in a poetic form, therefore it is a poem, since poetry has no essential nature. Whatever is presented as a poem is a poem.'[14] The identity of such a poem consists among other things in their balancing between being poems and anti-poems; the aesthetic device used is the Verfremdungseffekt, the shock of having something unpoetical presented as poetry.

Imagine a possible world P.W.1 where we only have what we in our world call 'paraphrasable poems' (let us call them 'p.p.s' for the sake of convenience; a paradigmatic example would be Williams's poem). Is it certain that this possible world would need the notion of poetry in order to classify the p.p.s in meaningful and useful way? No, because the difference between prose and what we call 'poetry' would be so unclear that the inhabitants of P.W.1 would have no use for the notion of poetry.

The arch-p.p., Williams's poem, might be called 'a parody of 'real' poem' in our world. In P.W.1 it would not make sense to call it 'a parody.' In a similar fashion, it does not make sense to ask whether there would be parodies of "Mona Lisa" in a possible world where the painting does not exist. A painting which in our eyes would look exactly like "Mona Lisa" but with a moustache would not be a parody in the possible world in question.

Let us look at another possible world example: We might imagine a possible world P.W.2 where art has always been like the avant-garde art of the twentieth century. In P.W.2, a person called Marcel Duchamp could display a urinal at an art exhibition, sign it R. Mott, and call it "Fountain," but it would still not be the same artwork (or anti-artwork) as in the actual world. For Duchamp's "Fountain" is whatever it is by virtue of being an implicit attack on prevalent notions of art. Those notions would be quite different in the possible world discussed. In an analogous manner, Williams's text (poem?) would not have any shocking effect in P.W.1, in a glaring contrast to the effect it could have had in our world some decades ago. Even if P.W.1 actually employed the concept of a poem, Williams's text would not count as an anti-poem, since it has not broken with any tradition. Add to this the fact that a host of people in our world would flatly deny that Williams's text is a poem but rather see it as a piece of everyday discourse, undeservedly called 'a poem.' So whatever virtues this poem might possess, being a paradigmatic poem is not one of them. This means that if my earlier description of the typical features of poems is correct and Williams's poem is a typical example of a p.p., then we can conclude by saying that non-paraphrasable poems are paradigmatic and the paraphrasable ones are parasitic on them.[15]

5. Discussing tacit knowledge

I think that the deeper explanation for the non-rephrasability of typical poems is that our understanding of them is basically tacit. If we possess propositional knowledge of a given X, then we can describe X adequately by putting forth a series of true propositions about the nature of it. However, if we cannot rephrase a certain poem, then we cannot describe it adequately with the aid of propositions (or indeed anything else). Therefore, our understanding of it cannot be entirely propositional.

The very idea of tacit knowledge evokes the image of Michael Polanyi, who famously said, 'We know more than we can tell.'[16] Our knowledge of a physiognomy, for instance a face, is tacit. The reason is that we know a physiognomy as a whole, without being able to identify its different parts.[17] I can know the faces of friends and relatives with a somnambulistic certainty without being able to describe them. Faces are Gestalts and so perhaps are poems; we have seen earlier that typical poems must be regarded as being wholes, which are greater than the sum of their parts. To be sure there are important differences between the grasping of a poem and the perception of a face. We cannot grasp a poem unless we are able to identify its different parts. Nevertheless, if the knowledge of faces is tacit by virtue of us perceiving them as being bigger than the sum of their parts, it seems tempting to think that the same holds for poems.

In order to vindicate this claim, I seek inspiration in the thought of certain followers of Wittgenstein. Wittgenstein never used the expression 'tacit knowledge', but there is no doubt that it was on the tip of his tongue. Norwegian philosopher Kjell S. Johannessen points out that Wittgenstein says in §78 of the Philosophical Investigations that we can perfectly well know how a clarinet sounds, but this knowledge cannot be put in words. Johannessen thinks that our knowledge of artworks, of which the knowledge of the sounds of clarinet can be an example, is a brand of tacit knowledge, which he terms 'knowledge by familiarity.'[18] Such a knowledge is not a know-how, it is in some ways a (tacit) know-that (our knowledge of faces must be some kind of a tacit know-that). Obviously, knowing that a given sound is the sound of a clarinet is a different brand of knowledge from that of knowing how to play the instrument.

Besides being tacit, knowledge by familiarity is autotelic, i.e. has its aim in itself, and cannot be used for anything but to be related to by the knower in experience and reflection. Thirdly, knowledge by familiarity can only be acquired by personal acquisition and must be personal knowledge, Johannessen says.[19] According to him, our knowledge of artworks is of this kind. We can be very familiar with a painting or a symphony without being able to describe it in words.

I will not discuss the finer points of Johannessen's analysis. Instead, I will utilise it for my own purposes. As suggested, my view is that our understanding of a non-rephrasable poem is essentially some kind of tacit know-that. A poem that cannot be rephrased cannot be adequately described because a paraphrase seems to be a kind of description of a poem's main content and possibly also its basic formal features. Further, such a description would consist of a series of propositions. At the same time, Brooks has correctly shown that a paraphrase does not have to be without worth in helping us to understand a poem. In a similar fashion a face or the sound of a clarinet cannot be adequately described, but descriptions can be of aid in identifying them: 'John has a round face', 'the clarinet sounds like the song of certain birds' etc. Both the paraphrases and the descriptions are put forth in propositions and are in Brooks's terminology 'pointers'[20]: they point towards the truth.

Well, I certainly hope that my theories point in that very direction. Actually, my theory about the tacit nature of our understanding of poetry gives my analysis the virtue of testability. There must be some ways of finding out in an empirical fashion whether or not our understanding of non-paraphrasable poems is tacit. If it turns out that it were not the case, that fact would weaken my defence of the non-paraphrasability of typical poems. And if there also were serious logical flaws in my possible world examples, then it would be hard to see how I can uphold my defence. So whatever lack of merit my arguments might have, I am not moving in vicious circles. Nobody can accuse me of just deciding by fiat that poems are typically not paraphrasable.

It is time to sum up this article in a few words: Paradigmatic examples of poems are not paraphrasable. A prosaic text can be improved with the aid of a paraphrase, a typical poem is only destroyed. The deeper explanation for the non-rephrasability of poetry is that our understanding of it is basically tacit. Thus, we have strengthened the foundations of Brooks's original thesis.

Only the anti-heretic can undertake the logical mystery tour of poetic interpretation. The heretics will be left behind in the grey land of paraphrases and empty phrases.[21]


1. Archibald MacLeish: 'Ars Poetica,' in Geoffrey Moore (ed.): American Literature (London Faber & Faber, 1964), p. 1069.return to text

2. This concludes my description of the nature of a typical poem.return to text

3. Cleanth Brooks, The Well-Wrought Urn (London: Dennis Dobson, 1968), pp. 157-176.return to text

4. The same goes for notoriously unclear notions like those of 'poetry' or 'everyday understanding.'return to text

5. The New Bantam English Dictionary (New York: Bantam Books, 1979), p. 655.return to text

6. Concise English Dictionary (London: Tophi Books, 1990), p. 293.return to text

7. 'Suchlike' can be 'the intonation of utterances', 'the rhythm', or 'the graphic form of texts.'return to text

8. In order to avoid misunderstanding, 'textual meaning' is not to be understood as 'sentence meaning.' 'Textual meaning' is simply the sense of the text, with all its connotations, which we try to grasp without worrying about the author's intentions.return to text

9. I use 'rephrase' and 'paraphrase' as synonyms. The same holds for 'paraphrasability' and 'rephrasability.'return to text

10. Ezra Pound: "Canto I", www.americanpoems.com/poets/ezrapound/12625return to text

11. It must be said in all fairness that using a more archaic language in the attempted paraphrase would probably have made it more like the poem itself.return to text

12. In the last lines of a poem 'the Homeric voice' is suddenly silenced and a modern voice speaks instead.return to text

13. William Carlos Williams, "This Is Just to Say,"return to text


14. Anybody knowledgeable about William Carlos Williams knows that he himself would never have said such a thing. But the point is that someone who writes such a poem could have said it without sounding inconsequent. The author of the Idylls of the King could not have.return to text

15. I take my chance at maintaining that this (and everything else I say in this article) holds for poetry in general in our empirical world, not just poetry in our epoch and culture.return to text

16. Michael Polanyi, The Tacit Dimension (Garden City: Doubleday, 1966), p. 4.return to text

17. Ibid., p. 6. The example was inspired by the findings of the Gestalt psychologists.return to text

18. It is important to emphasize that this type of knowledge has nothing to do with the traditional empiricist notion of perceptual knowledge.return to text

19. Kjell S. Johannessen, "Language, Art and Aesthetic Practice," in Johannessen and Tore Nordenstam (eds.), Wittgenstein-Aesthetics and Transcendental Philosophy (Vienna: Hölder-Pichler-Tempsky, 1981), pp. 108-126.return to text

20. Brooks, op.cit, pp. 196-197.return to text

21. This article was originally presented at the XXIst World Congress of Philosophy in Istanbul, Turkey, August 10-17, 2003.return to text

Lillehammer College, Norway