Warning: Undefined property: wpdb::$dmtable in /home/bohemica/public_html/techczech.net/wp-includes/class-wpdb.php on line 783
3 fundamental problems of translating metaphor (or anything else) – Metaphor Hacker
Extended writing Language Metaphor Reading Lists Writing

3 fundamental problems of translating metaphor (or anything else)

How hard is it to translate metaphor?

Metaphor seems like it should be very difficult to translate. But I’d like to argue that what is difficult about translating it is not the metaphor part but rather how it is used. This makes it no different from any other aspect of language. But because it is a rather salient use of language, we can use it to illustrate these general problems of translation.

I have summarised these into three broad classes of problems that a translator has to deal with day in and day out.

  1. Idiomaticity and conventionality
  2. Knowledge and underdetermination
  3. Coherence and cohesion

What is and is not a metaphor?

But first, let’s establish what I mean by metaphor. Drawing on the Conceptual Metaphor Theory, I use metaphor to stand for any non-literal or figurative use of language where two domains of knowledge are projected onto one another to create a new meaning containing parts of both domains. For example, the term computer virus is a result of the medical domain being projected into the domain of computers to create a new concept.

This covers a lot of traditional tropes including simile, analogy, allegory, or even hyperbole or irony. On this definition, they are just different surface representations of metaphor. (Sometimes, metonymy can be also included here).

So I may say, ‘John’s a beast’ (metaphor) or ‘John’s like a beast’ (simile) and there’s a difference in meaning because of how they are expressed. But the conceptual work that goes into understanding what goes on is the same. We confront what we know or don’t know about John with what we know about what is meant by ‘beast’ and form a different impression of John.

The same goes for hyperbole: If I say, ‘Startled, Irene jumped 10 feet high.’ I’m using this domain of height to express the intensity of her feelings. Or irony/sarcasm: “Fred’s a real Einstein”, I’m using the domain of ‘clever scientist’ to project onto Fred (who supposedly isn’t), to create a contrast.

Of course, in translation, the surface representations do matter. For example, as we’ll see below, sometimes the projection of two domains can be translated directly when expressed as a simile and has to be worked around when expressed as a conventional metaphor. But we can best think of these issues in the framework of the three general problems of translation rather than as a special type of problem due to their metaphoricity.

Problem 1. Idiomaticity and conventionality

The fundamental problem for all translation is that actual language use is a not a matter of just combining words according to some rules of grammar. There is a whole other set of conventions about when and how to use certain words and rules. Sometimes, this is a matter of propriety. So you have to know, not to call your teacher ‘dude’ as a matter of social convention. But most often the convention means that certain words used together, the whole has a different meaning than it would if we just combined the meanings of the individual words.

These are called idioms. Often, colourful expressions like ‘kick the bucket’ or ‘the whole nine yards’ are given as examples. But these are easy. The problem is that language is idiomatic all the way down. The words ‘put’ and ‘up’ have certain meanings but there’s no way you can figure out the meaning of ‘I will not put up with you’ or ‘I will put you up with Jane’ without knowing the convention behind them or at least getting enough context.

Most of any language depends on this kind of convention. And most of the metaphors people deal with in translation are also to some extent conventional. So, if I say ‘don’t eat like a pig’ in English, I mean don’t eat so much. In Czech, it means ‘don’t eat messily’. We have the domain of humans eating and pigs eating. That comes with certain imagery of what it looks like and what happens during, before and after. But when we use a metaphor, we are choosing what parts of the domains to project onto one another. And with metaphors that have been conventionalised, different languages and cultures choose different things.

This means that very novel and flashy metaphors and similes are often the easiest to translate. I still remember reading “He looked about as inconspicuous as a tarantula on a slice of angel food cake.” in the Czech translation of a Raymond Chandler novel. And it was very easy for me to google the original now. Because the sort of conceptual work Chandler is doing here is completely original and yet entirely understandable. This might be more difficult if translating into a language that does not have concepts for tarantula or angel food cake but it would be trivial to come up with a combination that does an equivalent job.

But because of their conventionality, idioms pose a core problem for the translator, namely picking an equivalent level of conventionality or idiomaticity. I still remember reading in a Western adventure book in my youth a lumberjack calling somebody what to me sounded like a novel curse ‘You god forsaken son of a female dog’ (zatracený čubčí synu). That made it sound like the character was trying to evoke rich imagery through linguistic innovation but I’m pretty sure the original simply had ‘You damn son of a bitch’. The translator simply kept the imagery of the original but did not take into account that it would sound novel to a Czech reader whereas it was entirely conventional in English.

Conventionality can also fly under the radar, blurring the lines between grammar and usage. For instance, ‘a bird singing in the tree’ will be translated as ‘on the tree’ (na stromě) into Czech. English conceptualises the tree as a container whereas Czech as a surface. In Czech, ‘in the tree’ would mean inside the trunk. Both languages have the same conceptual distinction between ‘in’ and ‘on’ but by convention they apply it differently to birds and trees. Equally, an English speaker introducing themselves on the phone will say ‘This is Hana’ but a Czech speaker will say ‘Here is Hana’. Both languages focus on the difference in perspective and distance but one will point at the person and the other at the location. These two examples are not ‘figurative’ in the rhetorical sense but they are conceptual. However, the way they conceptualise the world is a matter of convention, not pure grammar and lexicon.

Problem 2. Knowledge and underdetermination

Conventionality is the underlying matrix of all language in use. But it leads to an even more difficult problem which is underdetermination. Different languages leave different things unexpressed and assume you can figure it out from context. Others require that the speaker always be explicit about that same thing.

For example, languages such as English, German or Spanish have definite articles which specify whether we are talking about something specific or general. But most other languages don’t. There’s a big difference between ‘a horse walked into the barn’ and ‘the horse walked into the barn’. Languages without a definite article can express this difference when it really matters but often leave it implicit. But English always has to express it even when it does not matter in a particular context. Which means a language like Czech or Russian is underdetermined when it comes to definiteness.

So the speakers (and translators) have to rely on knowledge of context, culture or some other area of expertise to make sense of what goes on. And as we saw with the ‘pig’ example, metaphors are underdetermined by their very nature. When we call somebody a ‘pig’ we are not saying they have a little curly tail. We are picking some other similarity. So as a speaker of the language, I have to know not just what a pig looks and behaves like but also the convention of the culture about what aspect of pigness we compare humans to.

But I also have to know enough about the person being called a pig to understand what is meant. So a man could be called a pig because they are annoying (often with sexual undertones) or because he is very fat. Sometimes a bit of both. The expression is underdetermined as to the exact meaning.

When I try to translate that expression, I may not have a similar expression that covers both eventualities. In Czech, calling a man ‘a pig’ without any specification would specify some of the annoyingness (with sexual undertones) but is also underdetermined when it comes to messiness. However, it does not imply fatness. So, if I wanted to specify fatness, I’d have to use a simile ‘he’s fat like a pig’ which overdetermines the original expression. Meaning, I have to pick from multiple underdetermined meanings in the original and pick one.

This requires quite rich knowledge, but it can never be done perfectly. Asking the original speaker what they meant often does not help. Because their language is underdetermined, they may not have wanted to commit to one meaning or another. They may not even have been aware that such a commitment was possible. Their language did not require them to make a choice, so they didn’t make one.

That’s why translation can sometimes reveal faulty reasoning and can sometimes go terribly wrong. And that’s why translators so often agonise over how much of themselves to insert into the translation. Sometimes they have to say more than the author intended and sometimes they cannot say as much.

Problem 3. Cohesion and coherence

So far, we were only looking at metaphors in isolation. But they (as any expression) are always a part of a larger text and context. And the text has to hang together (cohesion) and make sense in context (coherence). This is why it is often more difficult to translate shorter texts than entire volumes. When somebody asks me to translate something, I always want to know the whole sentence and ideally the whole paragraph.

Texts have to be cohesive locally as well as globally. They have to make sense and their different parts have to connect to each other. Otherwise they would just be lists. This connective tissue of text is built up from different types of constructions. For example, there are connectives like ‘thus’, ‘but’, ‘however’ that establish causal or other logical relationships between parts of the text. But the most important is anaphora which is used to point back (and sometimes forward) in the text to establish a relationship between different parts.

This can be done in all sorts of ways. Pronouns as in “Aisha went into battle. She won. Her soldiers revered her.” But often this is just done by repetition (which is often alternates with pronouns). So the next sentence in the previous example may go like this: “And Aisha deserved their respect, because she took good care of her soldiers.”

Cohesion is not the same in all languages or genres. Sometimes, repetition is avoided or even forbidden, sometimes it is encouraged or required, some languages require more specific causal signalling than others. In other words, we see echoes of the problems of conventionality and underdetermination raising their head.

But if that were all, cohesion would not require a special mention. The problem for the translator and particularly the translator dealing with metaphor is that cohesion is not always very straightforwardly textual. It relies not just on what was said but also on what else is known.

This is particularly the case with so called ‘anaphoric islands’. They are a sort of metonymic phenomenon. For example, ‘I speak Russian, but I’ve never been there’. ‘There’ refers to ‘Russia’ which was never mentioned but is metonymically linked to the language of ‘Russian’. This example itself would not cause a problem for the translator. But when the metonymy becomes enmeshed with metaphor, it often is. Because we can now link together not just things from one domain but from two.

For example, take ‘The band exploded onto the scene, and the reverberations are still being felt today.” Here the author is taking the domain of explosions in the initial conventional metaphor and draws on it some more by taking other images from what happens after the explosion. But the translator may not have been able to translate the initial idiom using the same domain in explosion. For example, in Czech you can only use the word for ‘explode’ (vybouchnout) to mean the equivalent of the English ‘bomb on stage’. So, the domain of explosions was never activated and reverberations would make no sense.

This sort of thing is not that difficult to overcome within a single sentence but sometimes the entire text is built around a metaphor that only shares some of the domain across languages. So the translator may have to leave things out or try to make up a completely different metaphor and hope it will not distort the meaning of the original too much. Journalistic and academic texts are often full of these structuring metaphors and it is almost impossible to keep up with them throughout the text.

Side note: Notice how the subtle grammatical difference makes all the difference between metaphor and literal language. ‘Exploded onto the scene’ is very different from ‘exploded on the scene’.

All the other problems of translation

Translation is hard. And it is impossible if we expect perfect translation that goes both ways. If we imagine that a hypothetical perfect translator translates a text from L1 to L2, we would expect another hypothetical perfect translator to be able to take that text in L2 and translate it back to L1 and the we could get the exact same result as the original. That is only possible on the simplest of texts. But if we went back and forth a lot, threw in some other languages into the mix, we would be fairly far from the original.

All the different problems of conventionality, underdetermination and cohesion would compound so that we would see a very different text. But at the same time, it is not that hard. Because we could probably by and large keep the basics together. As we know, the Bible is a translation of a translation and it’s not all that different across languages. But it did take teams of careful translators with deep knowledge of the original and its context years or even decades to complete. Which is in great contrast to most of the translation done by hurried, underpaid and often underinformed translators.

Language is so redundant that bad translations make less difference than one might think. It is remarkable how many mistranslations there are in subtitles or dubbings of popular TV shows and people still love them. I’m sure there are legal documents, psychological texts and more that also contain mistranslations. Sometimes they can be consequential but often they’re not.

My favourite childhood book was the The Coot Club by Arthur Ransome. In it, some children are learning to sail and they are spending a lot of time naming the sides of the boat. These are ‘port’ and ‘starboard’ in English and they are notoriously difficult to learn. But in Czech they are simply left side (levobok) and right side (pravobok). Children often confuse left and right, so it makes sense that they would make some errors. But I still remember thinking as a child that these kids were particularly dimwitted. Furthermore, two characters (twins) had nicknames ‘Port’ and ‘Starboard’ which was simply rendered into Czech as ‘Lefty’ and ‘Righty’ and it was never connected with the nautical term. There is nothing the translator could have done to rescue all the internal connections and cultural allusions in the text. And still I liked the books so much that I actually moved to Norfolk where the action took place.

There are many many more sub-categories of problems that translators have to deal with. But most of them could be thought through the lens of one of the three I mentioned. Translating figurative languages is hard but only because translation is hard. It is impossible to convey every little nuance, turn of phrase, hint and allusion without a lot of footnotes. Fiction is harder to translate than non-fiction, short pithy phrases are harder to translate than sprawling texts. Translation is hard. Metaphors are hard. But we still get by.


I wrote this after attending the conference on Metaphors in Translation. I was disappointed that nobody was thinking of metaphor as a complex category.

Summary of main points

  1. Metaphor on its own is not a special problem for translation but it requires lots of background knowledge on the part of the translator.
  2. The key issues for any translation that are just as important for metaphor as anything else are:
    1. Idiomaticity of many expressions that requires finding an equivalent expression not using the same words because of different conventions
    2. Underdetermination that requires the translator to fill in or take out something in the original text (often because of the idiomatic convention difference between the two languages)
    3. Cohesion where a metaphor or some other conventional expression is referred to throughout the text in different ways that are impossible to replicate in both languages.
  3. Because of the issues above, it is impossible to create a perfect translation in such a way that two ‘perfect’ translators, one translating a text from L1 to L2 and the other then translating the text back from L2 to L1, would get back to the exact replica of the original. This is especially true for figurative language of all kind but there are elements of figurativeness in all language.

More readings on metaphor

Even though I tried to explain key concepts as I went along, I necessarily took some shortcuts and may have taken things for granted. Here are some links to other things I wrote about metaphor in various contexts.

How metaphor works

Hacking a metaphor in five steps

First post I ever wrote on this blog that outlines all the key things I think are important to know about metaphor.

How we use metaphors

A quick reproduction of a taxonomy I developed early on in my research on metaphor (published in a 2005 paper). It is a classification of the different ways in which metaphor is used in actual texts. The paper (in retrospect quite clunky) gives more examples.

3 burning issues in the study of metaphor

There are just too many assumptions we make about metaphor. But a lot is still unknown. Here I try to outline the key questions that still need answering.

Metaphor as ordinary language

My main preoccupation on this blog (despite its name) is that metaphor is nothing special and it is all pervasive. Here are few posts that make the case in more detail.

Fruit loops and metaphors: Metaphors are not about explaining the abstract through concrete but about the dynamic process of negotiated sensemaking

Shows on the example of an extended text that metaphors are intermixed with non-figurative language to create rich meaning.

Poetry without metaphor? Sure but can it darn your socks?

What it looks like when the ‘literal’ language is used to evoke poetic imagery.

Metaphor is my co-pilot: How the literal and metaphorical rely on the same type of knowledge

An example of when understanding a literal statement requires the same kind of mental work as understanding a metaphor.

Anthropologists’ metaphorical shenanigans: Or how (not) to research metaphor

Snarky take down of overinterpretation of metaphor in work of anthropology due to over-translation of conventionalised tropes.

What is not a metaphor: Modelling the world through language, thought, science, or action

Quite dense but gives a few more examples of how hard it is disentangle metaphor from other types of language.

Metaphors and other tropes

Binders full of women with mighty pens: What is metonymy

In this post, I explain what metonymy is but then unpick some of that to link it to metaphor rather than put it in opposition to it as is commonly done.

Not ships in the night: Metaphor and simile as process

This post illustrates how metaphor and simile work together to construct an elaborate textual and conceptual structure. It is an analysis of a single extended paragraph.