Home     Twiglog



Secondary Primaries


Struan Gray



A version of this article in French is available here: Primaires Secondaires.


For a long time now I have been a fan of Henri Gaud's adventures in trichromie (trichromie.free.fr/trichromie).  Partly as a cool revival of an interesting historical technique, but mostly because it seems to open some new drawers in the photographic toolbox.


Briefly, the technique combines seperate monochrome negatives shot through different coloured filters to build up a full-colour image.  Usually, the additive primaries (red, blue, green) are used, or their substractive counterparts (cyan, magenta, yellow).  The primaries chosen are important because anything that moves between the taking of the seperate negatives aquires highly saturated fringes or ghosts whose colour is the complement of the original filter colour.  Also, slight mismatches in the contrast response of the B+W film, or in matching exposure or development, lead to colour casts and crossovers which also tend to take on the hue of the filters or their complements.



Gaud Ferry Fringing

Henri Gaud's fringed ferry passengers
From the Trichromie Blog


To some, these fringes and colour casts are just errors to be regretted, eliminated or sneered at.  But a good modernist easily recognises them as the signature of the technique - a strength to be played to rather than a mistake to be rectified.  The colour fringing around moving objects in particular is an interesting way to introduce time and motion into a still frame without causing blur.  The colour casts can indicate a mood that varies from heightened awareness to a drowsey half seeing, and much in-between.  A digital step in the reproduction chain offers the chance to tailor these expressive colours with precision and subtlety.


Henri Gaud and others have played with these tools in various expressive ways, and most recently Henri has been shuffling the colour planes of his images, swapping the red negative for the green one, or rotating red-green-blue to green-blue-red.  The results are interesting, and tell us much about our own perception of the world, and our visual system's lack of any absolute reference or standard.



Gaud Ferry Permutations

Henri Gaud's colour permutations
From the Trichromie Blog


All of which is exciting and fun, so my rational side gets regularly peeved when my emotional response to individual trichrome photographs is less direct or enthusiastic than would be expected from my blanket approval of the technique as a whole.  It's not just the usual disappointment with technique-led photography - that fascinating tools can make dull pictures - nor I think is it a residual gut instinct that fringes and colour casts are a sign of incompetence.  I can't say I have lost sleep over the issue, but it has niggled away at me when I should be doing other things, and I have at last come up with a reason: I simply don't like the standard colour primaries.


Red, green, blue, cyan, magenta, yellow.  From my first schoolboy painted colour circle onwards these have been presented to me as the primary ur-colours compared to which all others are derivative.  A moment's childlike contemplation of a rainbow should suffice to instill a deep suspicion of the word 'primary', but education takes over and Young, Helmholz, Maxwell and the rest are marshalled as arguments against intuition.


Young, Helmholz and Maxwell were all undeniably great scientists, and I admire the subtlety and breadth of their thinking unreservedly; but their reputation has perhaps done their science a disservice, and helped to harden a view in which the primaries of the human visual system become absolute goods, rather than necessary tools.  A picture illustrates the problem nicely:



Maxwell's Triangle

Maxwell's Triangle
From the excellent efg site


This is the sort of thing I was encouraged to draw at school.  Red, blue and green are at the corners of an equilateral triangle, and other colours are straightforward additions of the corner primaries in proportions corresponding to the geometric distance to the relevant corner.  As a diagram is a deeply satisfying confirmation of the RGBCMY primacy.  White sits in the centre, and the vagaries of perception make the subtractive primaries show up nicely as bright spurs leading to the half-way point on each edge of the triangle.  As expected, cyan is directly opposite red, magenta opposite green and yellow opposite blue.


Henri's shufflings of the colour deck correspond to simple symmetry transformations of this triangle.  Swapping red and blue flips the triangle about a vertical axis: all colours in the image get exchanged with their counterparts on the opposite side.  Rotating the colour planes rotates the triangle 120° so that the corners align, and all the internal colours are mapped onto the next sector by a rotation through the same angle.  Swapping red-green-blue to cyan-magenta-yellow can similarly be represented as a geometrical transformation - a flip about a horizontal axis, or a rotation by 60° - but now the triangle model falls a little short because you have to do something about the way the points of the triangle get mapped into the middles of the sides.  However, the idea is simple enough, and and if you list the possible permutations you quickly realise that there are at most twelve unique versions of any given image.


So why the disappointment? Well, in short, many of my favourite colours appear to be missing.  Where, for example, is the indigo of alpenglow? The sheen of turquoise? The warm brown of burnt umber or polished leather? The blue-green of sunlit water over a sandy seabed? The triangle does an excellent job of displaying the bright, sharp colours of the corporate world, of advertising, and of children's crayon sets, but it obscures the secondaries, the subtle hues of nuanced painting, of pre-industrial, natural dyes and of much of ordinary life.



Josef Albers - Homage to the Square: Soft Spoken

Josef Albers - Homage to the Square: Soft Spoken
The Metropolitan Museum of Art, New York


This might seem like a contradiction, because these straightforward combinations of the three visual primaries should be able to produce every perceptible colour.  Vision is a unique sense in that a combination of colours is not percieved as a combination, but as a single stimulus.  Hearing has its chords and tonal signatures, taste and smell allow the conoisseur to tease out the components of complex mixtures, and even coarse touch can sense an incomplete mixture of hot and cold as two seperate temperatures.  However, colour is a many-in-one-out process, and so combining red, green and blue in suitable amounts should not leave any colour unaccounted for.


In fact, the secondaries are indeed there in the triangle, but they are squeezed into such small areas that they are difficult to see.  Worse, when human nature regards a highly symmetric figure like the triangle it tends to concentrate on the parts which define that symmetry such as the corners and the centres of the edges: the secondaries lie in off-kilter pockets where our attention does not naturally wander, and they are easily overlooked.


The root cause of the problem is not really a lack of attention, it is the desire to cram colour perception into a tidy geometric pidgeonhole.  There is no reason whatsoever why our perception of colour should conform to the simple algebraic additions represented by the triangle, and if we step back from our learning for a moment and just look at the world it is clear that it does not.  Aspects of the triangle such as the blue-yellow opposition and the need for three stimuli reflect deep physiological and psychological truths, but it is a mistake to fit those truths into too-tidy a conceptual framework.  Doing so not only blinds us to the richness of colour perception, but it constrains our thinking in counterproductive ways.


It is perhaps worth noting that the persistent contemporary obsession with saturated colours only makes the problem worse.  Boosting saturation pushes colours away from the white centre, first to the edges of the triangle and then along those edges towards the corners.  It doesn't matter if the boost is done digitally or chemically within the layers of a piece of colour film: the consequence is to drive colours towards the primaries and to deprive them of their individuality.  The primaries take on the role of a self-fulfilling prophecy in that they become the only colours the depicted world can contain.


Things become more clear when colours are plotted in a way that makes empirical perceptual differences correspond to a uniform spatial measure.  If you make a plot where colours that seem different by the same amount are the same distance apart - the famous CIE standard observer plots - the blue-red secondaries aquire a bit more space.  Perhaps more interestingly, the bright spurs of the subtractive primaries are still roughly opposite their complementary colours, they no longer bisect the sides of the approximate triangle and so the idea that they are 'half-way' between the primaries, or that there exist simple geometric relationships between them is immediately and obviously made suspect.



CIE 1976 - colours in perceptual space

Colours in perceptual space - the CIE 1976 map
From the efg site


Perhaps the most intriguing consequence of presenting colours in this way is how the red-blue-green and cyan-magenta-yellow triplets no longer appear to be exact complements of each other, and how they do not evenly divide up the colour space as they did in the simplistic Maxwell triangle.  Instead, a third triplet, which I call indigo-orange-turquoise is needed to divide up the colour space into roughly equal sectors around the central white point.  This third set, which I perversely call the "secondary primaries", seems to be necessary to achieve an even coverage of the possibilities allowed by human perception.


To me, it is deeply satisfying that this triplet has as its primaries exactly those secondary colours which I felt were missing from the original Maxwell triangle.  However, that really just shows that even I am not completely immune to the lure of a suductively elegant numerology.  The key point is that no simple geometric figure exactly represents the relationships between colours in perceptual space, and that insisting on a particular scheme - such as the many proposed "correct" colour wheels in the colour theory literature - is at best an ad-hoc working solution.


So what does this have to do with photography? Well, I feel strongly that my own future includes tricolour photography and tricolour printing techniques like colour carbon or gum.  So on a personal note is has been good to muddle through to an understanding of what does and does not fit with my own aesthetic temperament.  When, for example, I do get up off my plump theorist's backside and actually make some tricolour photographs, I shall do so using an indigo-orange-turquoise filter set, and thus my motion fringes and colour casts will take on the colours that I find intrinsically satisfying.



Struan Gray's usurption of Henri Gaud's Colour Variations

Colour variations using all three colour triplets


Second, I have a strong gut feeling that a visual balance is achieved when a set of colour permutations includes the indigo-turqoise-orange triplet.  When only rotations through red-green-blue and cyan-magenta-yellow are present, I get the sense that something is missing.  Filling the gap in the perceptual colour map by including the third triplet makes things complete again.  Above, for example, are the permutations of one of Henri Gaud's photos made by simply rotating the hue values of the image through multiples of 40° (this is easily done in Photoshop's hue adjustment dialog).  This generates nine unique images, six of which are very like those made by shuffling colour planes in the red-green-blue or cyan-magenta-yellow colour spaces (this makes sense: the cyan-magenta-yellow triplet is roughly 40° rotated from the red-green-blue one in perceptual space).  However, three of the images are new, and correspond to the indigo-turquoise-orange triplet.  For me at least, they seem to fill in a necessary gap, and don't create any sense of duplication.  The point is reinforced if the permutations are plotted so that the individual images occupy positions around a circle corresponding to the hue rotation applied: the variation of hue around the circle then becomes obvious.



Struan Gray's usurption of Henri Gaud's Colour Variations on a Wheel

Colour variations using all three colour triplets
Shown on a colour wheel


Finally, thinking through this issue has clarified one aspect of my instinctive dislike of much conventional landscape and nature photography.  The methods and conventions of the genre unavoidably push all colours towards a few bright but uninteresting primaries.  It is Lego photography, and just as with Lego buildings, the available forms are severely restricted by the tools used.  Looking back at my own work in recent years, much of it can be seen as an attempt to rediscover a richer, more subtle palette.  I am not going to claim that secondary colours are necessarily better in a global sense, or that Les Trichomistes or anyone else should do things my way, but it feels good to have reached a rational explanation for a subconscious yearning.




Recommended Reading


While I was writing this I discovered Bruce MacEvoy's Handprint site (www.handprint.com) which discusses in great detail both the usefulness and the fundamentally misleading nature of colour theory and its over-elegant models of human perception.  Bruce uses a different nomenclature to mine, and his pages are a confusing rabbit warren of internested links, but if you resist the urge to click on every link and just follow the pages in series it makes for a wonderful, enlightening read.


For artists: http://www.handprint.com/HP/WCL/wcolor.html


For geeks: http://www.handprint.com/LS/CVS/color.html


Of direct relevance: http://www.handprint.com/HP/WCL/color13.html





Text copyright © Struan Gray 2008.  Images copyright their respective creators.  All rights reserved.