This essay was inspired by a talk at the Oxford Literary Festival from English lecturer Matthew Reynolds about his book Translation: A Very Short Introduction . It inspired me to think about line drawing as translation. As Manet said, "There are no lines in nature, only areas of colour, one against another."
Or to quote a more recent artist:
So the lines in drawing are an invention. They give the artist a chance to collaborate with the original scene in new and surprising ways, inventing or adapting a language into which to translate them. In this essay, I'll look at a few examples of such languages. I'll also argue that it is not wrong to omit or exaggerate information; in fact, the constraints of the language mean doing so is essential.
Before any linguists shout at me, by the way, I'm using the word "language" loosely, as computer scientists often do. "Notation" or "representation" might be better. I'm not implying that drawing is a language in the sense that English or Chinese are.
One aspect of drawing as translation is choosing the lines that show object boundaries.
Sometimes this must be arbitrary, as with the steam below:
But usually, surely it's obvious where boundary lines should be drawn. Or is it? Look at the boy's lips and chin, and notice how parts have been omitted:
Omitting parts of such lines is a common convention of the time. Here is an entire face in the same style: Doust's was a popular "how to draw" book. In it, he advises that the best way to draw a mouth is often with no edge to the lower lip, because the light catches it, and with emphasis to the lines in shadow. He and the illustrator of Sunshine Stories have omitted the top of the chin, because light catches that too.
The picture below is a clipart drawing of a bar of soap. The internal line which shows edges is broken, perhaps to make the soap look shiny or translucent, or perhaps to indicate that the edges are rounded.
I was discussing this with one artist, and she said that she prefers broken lines even for unrounded edges such as those on a cube. It makes the drawing look more satisfying, somehow, or more sophisticated. Perhaps (my guess, not hers), the brain doesn't like being spoken down to, as if it were a baby to whom every detail has to be spelled out.
Here's a different way in which too much line can be unwelcome. People sometimes speak disparagingly of drawings or patterns that are too busy. It's easy to create these when translating from the original scene. For example, below is a photo of houses from my bedroom window. Imagine if I were to draw it by drawing a line around every brick and every slate. It would look far too busy, far too cluttered.
So instead, one often draws just a few patches of bricks, as in this clipart:
But drawing just part of the texture isn't only for cartoons. Here is an example from a book on learning to draw, showing it's a device used by serious artists:
So a successful translation may delete information. I like to think of this as restoring balance. The language of line drawing is poorer than that of reality. This means that when translating from one to the other, some things can't be said directly, they can only be approximated. That upsets the balance of the drawing. So we then have to restore that balance.
With the bricks, the language of drawing is poorer because it can't express subtle differences in tone. (I'm talking about ink drawing, not pencil.) So drawing lines between the bricks makes those lines stand out much more than they do in reality. To restore balance, we must make them stand out less.
But we can't, because our language isn't rich enough. So we do the next best thing. We remove some of the lines altogether, leaving just enough to tell us that there are bricks.
In mathematics, this would be called a generalised inverse. It's an operation that undoes another operation, not exactly, but as closely as the language permits.
Here's this as a diagram. The vertical arrow represents making the original line drawing, and the horizontal arrow detextures it, removing bricks. They combine to give the diagonal arrow. I'll say more about this in "Drawing as Translation II".
By the way, the diagram uses an image from M. S. Jackson's post "The Exterior - Brick, Stone, and Siding" (3 September 2012) in his blog Our New Home. To make the diagram, I simulated a line drawing by taking this image, converting to greyscale, thresholding, and enhancing lines. I then erased some stones from the right-hand part of the building.
I call the brick-deleting transformation "detexturing". But I also like to call it "appointing a representative". This is a more positive viewpoint. One selects a patch or two of bricks to represent all bricks, then deletes the rest.
It's interesting to think of other examples. Here's some advice on picking representative patches of fur when drawing cute animals, from Christopher Hart:
These cheek ruffles — with a large neutral space between them where they meet under the chin — are the most natural looking. The straight line of neutral space serves as good "pacing" between the ruffles. It's a place where the eye can "rest".
Going back to translation from scene to drawing, here's another example of how the medium affects representation. The fact that line drawings are usually done on white paper makes it hard to draw blond hair. So illustrators often just indicate a few locks at the hairline, as in J. W. Taylor's economically-drawn boy:
This shows again how the medium — the language — dictates balance. Dark hair is much easier to draw on white paper than blond hair. So in a drawing showing people with both hair colours, if there are not many other areas of black, the dark-haired people may stand out too much.
This matters, because black is sometimes used to attract attention to a character. The cartoonist David Langdon is said to have introduced the convention that that in a cartoon, the person with the open mouth is the one speaking. In his cartoon below, the black of the open mouth attracts attention. Other areas of black, drawn just because the artist can, would interfere with that. Langdon was too skilled to make such a mistake, as we can see from his blacks. Notice how the tie balances the bow tie; the moustache balances the open mouth; the shoes match the shadow under the seat. But other artists might not be so careful.
I want to relate the above ideas to the concept of "blooming" in language translation. I came across this in the Language Log post "Blooming, embellishment, and bombs" by Victor Mair (17 August 2015), which refers to a comment by Judith Strauser (3 August 2015) to an earlier post also by Victor Mair (2 August 2015), "French vs. English". Judith says:
French is notoriously longer than English, yes, though I shall not venture into possible reasons (Jeff W makes a good case), but as others mentioned above, it's also pretty clear in this case that the English was translated into French, which does lead to significant expansion, that translation professionals here call "blooming".
As a French translator (of English), I've been taught the blooming of a text as you translate it represents generally about a 6-10% volume expansion when going from English to French for a long text like a novel. It's (apparently) generally considered that on a long text, a blooming factor of over 15% hints at a bad (literarily speaking) translation job. There are other blooming ratios for other language pairs, all serving as reference points, but I forgot them.
Both posts contain discussion of why blooming occurs. Is it just that some languages are more concise? But "Guy" (17th August 2015) comments:
Although I'm sure some portion of "blooming" is due to some languages being overall more concise than others, I imagine some of it is simply an inevitable result of translation — that in general, given two equally "concise" languages, translations in either direction can in general be expected to be longer than originals. I expect this because the translator will often feel obliged to retain nuance that is easily expressed in one language but not the other, but will less frequently condense. I suspect that condensation in translation would be less common both because a) the original author chose to express themself in a way that is tailored to the original language and 2) condensing a complex expression into a simpler one when possible in the target language but not the source language requires a greater degree of insight or skill on the part of the translator, with a mastery of the idiom of the target language.
I haven't seen the notion of blooming applied to drawing, but I think it's a worthwhile intellectual leap. It would be interesting to look for different kinds of blooming. Note Guy's point "the translator ... will less frequently condense". J. W. Taylor condenses when he reduces blond hair to a few strokes at the hairline. The cartoonists who draw just a few patches of brick condense. They have mastered the idioms of their target language.
An essay that I'd recommend to every translator is Douglas Hofstadter's "Analogies and Roles in Human and Machine Thinking", from his book Metamagical Themas. In the PDF, this starts on page 547. It has many insights about what to do when you can't translate a text exactly. How, for example, would you translate "The First Lady of Britain" from American to Thatcher-era English? Hofstadter gives many conceptual experiments — many examples of analogies and "slipped" analogies — to clarify the problems involved.
Hofstadter uses stepping stones as a metaphor for translation, as he explains in his caption to the figure:
FIGURE 24-7. A metaphor for translation. A stream (symbolizing reality) has two sets of stepping-stones (symbolizing the basic ingredients of a language, such as words and stock phrases) in it. The black stones (Burmese, say) are arranged in one way, and the white stones (say, Welsh) in some other way. A pathway linking up a few black stones (a thought expressed in Burmese) is to be imitated by a "similar" pathway joining up white stones (translated into Welsh). One possibility is the speckled pathway, located at nearly the same part of the stream as the original pathway but not terribly similar in shape to it (a fairly literal translation), while a rival candidate (a more literary translation, needless to say) is the pathway located a distance upstream and resembling the original in some more abstract ways, including patterns in some of the "overstones" of the main stones (the similar archipelagos in Burmese and Welsh stones running roughly parallel to the far bank).
In terms of these stepping stones, a translation that's bloomed would be one with more stepping stones than the original. One whose translator has recondensed it would have the same number of stones. I'm not convinced that the metaphor works well, though. Should a recondensed translation have the same shape as the original? Of course, it depends what exactly we decide the stones and paths actually stand for.
When translating from scene to drawing, how do we indicate form? Here's a quote from another instructional book by L. A. Doust:
The next essential fact always to have at the back of your mind is "form", or the fact that the figure which you are drawing has thickness as well as outline.
If you study closely the great masters of figure drawing in outline, you will be amazed to discover that they manage to indicate the "form" of a body without the use of shading. How is this done? The secret of these clever drawings is often in certain lines on the figure or head and not actually on the outline — a fold, a collar, a cuff, a crease. Look at these lines carefully, and you will observe that they are very correctly drawn, sometimes even more than the actual outline. A simple illustration of this point is in Fig. E, Plate 12.
And here's one of his examples:
In the above, Doust gives information about curvature of the hat cross sections, thereby telling us about the tilt and shape of the head. He tries to keep this accurate. However, artists sometimes exaggerate. Here are three caricatures by Joyce Grenfell. Notice the chin of the dark-haired fellow in the second caricature. It's a kind of bent U which intersects slightly with the left-hand side of his neck. I suspect that not all of that U line was actually visible, and that Grenfell has lengthened what was visible. She may also have distorted the left of the U further to the left so that it would overlap the neck, thereby giving another cue to the form.
Very well known as Roald Dahl's illustrator, Quentin Blake exaggerates in a different way. His visual language emphasises facial expression, posture, and gesture, including hand positions.
I suggest that like omission, exaggeration is not deceit. It's saying what has to be said, as closely as possible given the constraints of the language.
When I draw, I'm often translating from specific words or concepts as well as from the scene. Below, I've reproduced three pages from my sketchbooks. I drew each figure at least partly "blind", so the drawings are rough. But that's not the point. As I drew, I was mentally annotating each line with a quality, action, or object. I wrote those annotations down alongside the drawings. They make up part of the source language I was translating from. Some of them will have affected my lines as I drew them; others require more thought, and would probably only be evident in a finished drawing.
Finally, here is another rough sketch. Again, my lines reflect what I wanted to "say". In this case, it was mainly form rather than, for example, the quality of fabric. So there are quite a few cues to shape such as L- and T-junctions, concave curves indicating saddle points, and what I suspect are apparent ridges. For an explanation of all but the last, see Visual Intelligence: How We Create What We See by Donald D. Hoffman. Apparent ridges are a recent addition to computational image interpretation, described in "Apparent Ridges for Line Drawing" by Tilke Judd, Frédo Durand, and Edward Adelson.
These final two sections may seem to be rather different in ttheme from the rest of this essay. But they all try to answer the question "what does an artist mean to say?". Or, as some books on drawing instruct, "make every line count". But count for what?