EDITED BY

MARSHALL BLONSKY

The Johns Hopkins University Press

Baltimore, Maryland
 UMBERTO ECO

How Culture Conditions the Colours

,:ve See

I
Colour is not an easy matter. James Gibson, in The Senses Considered as
Perceptual Systems, says that "the meaning of the term colour is one of the worst
muddles in the history of science".1 If one uses the term "colour" to mean the
pigmentation of substances in the emironment, one has not said anything about
our chromatic perception. Johannes Inen, in his Kunst der Farbe, distinguishes
between pigments as chromatic reality and our perceptual response as
chromatic effect.2 The chromatic effect, it seems, depends on many factors: the
nature of surfaces, light, contrast between objects, previous knowledge, and
so on.
I do not have any competence about pigments and I have very confused ideas
about the laws governing chromatic effect; moreover I am neither a painter, nor
an art critic. My personal relationship with the coloured world is a private affair
as much as my sexual activity, and I am not supposed to entertain my readers
with my personal reactivity towards the polychromous theatre of the world.
Thus, as far as colours are concerned, I take the privilege of considering myself
a blind man. I shall be writing about colours from a merely theoretical point of
view, namely, from the point of view of a general semiotic approach.
Since I have assumed myself to be blind or at least a Daltonist, I shall
mistrust my visual experience. I shall start from a verbal text, chapter 26, Book
II, of Aulus Gellius' Noaes Aaicae, a Latin encyclopaedia of the second century
A.O.
To deal with colours by making recourse to a tex1 of this period is rather
1
James Gibson, The Senses Considered as Perceptual SJ,stems (London: Allen & Cnwin,
1968).
2
Johannes ltten, Kunst der Farbe (Ravensburg: Otto Mair, 1961).
 l'MBERTO ECO

challenging. We are facing linguistic terms for colours, but we do not know
what chromatic effects these words refer to. \re know much about Roman
sculpture and architecture, but \'ery linle about Roman painting. The colours
we see today in Pompeii are not the colours the Pompeians saw; ewn if the
pigments are the same, the chromatic responses are not. In the nineteenth
century, Gladstone suggested that Greeks were unable to distinguish blue from
yelYow. Goetz and many others assumed that Latin speakers did not distinguish
blue from green. I have found also somewhere that Egyptians used blue in their
paintings but had no linguistic term to designate it, and that Assyrians, in order
to name the colour blue, could do no bener than transform the noun "uknu",
naming lapis lazuli, into an adjectiYe.
All of this is highly speculative, but we need not test every case. Let me
concentrate on the following passage from Aulus Gellius. The reader is ad,ised
to hold his temper, since the passage is highly confusing.
Gellius is reporting a conversation he had with Fronto, a poet and
grammarian, and Favorinus, a philosopher. Favorinus remarked that eyes are
able to isolate more colours than words can name. Red (rufus) and green
(viridis), he said, have only two names but many species. He was, without
knowing it, introducing the contemporary scientific distinction between
identification (understood as categorization) and discrimination, of which I
shall speak later.
Favorinus continues: rufus is a name, but what a difference between the red
of blood, the red of purple, the red of saffron, and the red of gold! They are all
differences of red but, in order to define them, Latin can only make recourse to
adjectives derived from the names of objects, thus calling flammeus the red of
fire, sanguineus the red of blood, croceus the red of saffron, aureus the red of gold.
Greek has more names, Favorinus says, but Fronto replies that Latin, too, has
many colour terms and that, in order to designate russus and ruber (red), one can
also use fulvus, flavus, rubidus, poeniceus, rutilus, luteus, spadix.
Now if one looks at the whole history of Latin literature, one notices that
fulvus is associated by Virgil and other authors with the lion's mane, with sand,
wolves, gold, eagles, but also with jasper. Flavae, in Virgil, are the hair of the
blond Dido, as well as olive leaves; and the Tiber river, because of the
yellow-grey mud polluting its waters, was commonly called flavus. The other
terms all refer to various gradations of red, from pale rose to dark red: notice,
for instance, that luteus, which Fronto defines as "diluted red", is referred by
Pliny to the egg-yolk and by Catullus to poppies.
In order to add more precision, Fronto says thatfuh·us is a mixture of red and
green, while flavus is a mixture of green, red and white. Fronto then quotes
another example from Virgil (Georgica, III, 82) where a horse (commonly
interpreted by philologists as a dapple-grey horse) is glaucus. Now glaucus in
Latin tradition stands for greenish, light-green, blue-green and grey-blue;
Virgil uses this adjective also for willow trees and for ulva or sea lettuce, as well
as for waters. Fronto says that Virgil could also have used for his same purpose
(his grey horse) caerulus. Now this term is usually associated with the sea, skies,
 HOW CUL TURF. CONDITIO!'\S THE COLOURS WE SEE 159
the eyes of Minerva, watermelons and cucumbers (Propertius), while JuYenal
employs it to describe some sort of rye bread.
And things get no better with viridis (from which comes the Italian rerde,
green), since in the whole of Latin tradition, one can find viridis associated with
grass, skies, parrots, sea, trees.
I have suggested that Latin did not clearly distinguish blue from green, but
f orinus gives us the impression that Latin users did not even distinguish
a\'
blue-green from red, since he quotes Ennius (Anna/es, XIV, 372-3) who
describes the sea at the same time as caeruleus and flavus as marble. Favorinus
agrees with this, since - he says - Fronto had previously described jlarus as a
mi,xture of green and white. But one should remember that, as a matter of fact,
fronto had said thatjlavus was green, white and red, and a few lines before that,
had classifiedj/avus among various gradations of red!
Let me exclude any explanation in terms of colour blindness. Too easy.
Gellius and his friends were erudites; they were not' describing their O\m
perceptions, they were elaborating upon literary texts coming from different
centuries. Can �e say that they were considering cases of poetic invention -
where, by a provocative use of language, fresh and uncommon impressions are
,ividly depicted? If that were the case, we would expect from them more
excitation, more marvel, more appreciation for these stylistic tours de force. On
the contrary, they propose all these cases as examples of the most correct and
precise use of language.
Thus the puzzle we are faced with is neither a psychological nor an aesthetic
one: it is a cultural one, and as such it is filtered through a linguistic system. We
are dealing with verbal language in so far as it conveys notions about ,isual
experiences, and we must, then, understand how verbal language makes the
non-verbal experience recognizable, speakable and effable.
To solve Aulus Gellius' puzzle, we must pass through the semiotic structure
of language. As a matter of fact, colour blindness itself represents a social
puzzle, difficult both to solve and to detect, because of linguistic reasons. Let
me quote this important passage from Arthur Linksz, which is later commented
upon by Marshall Sahlins:

To suppose color terms merely name differences suggested by the visible

spectrum, their function being to articulate realities necessarily and
already known as such is something like the idea that genealogical
relations comprise a defacto grid of "kinship types," inevitably taken in
this significance by all societies, which differ merely in the way they
classify (cope with) such universal facts of "relationship." The point,
however, in color as in kinship, is that the terms stand in meaningful
relations with other terms, and it is by the relations between terms within
the global system that the character of objective reference is sedimented.
Moreover, the concrete attributes thus singled out by the semantic
differentiation of terms then function also as signifiers of social relations,
not simply as the signifieds of the terms. In the event, it is not even
160 L'\iBERTO ECO

necessary that those who participate in a given natural order have the
same substantive experience of the object, so long as they are capable of
making some kind of sensory distinction at the semiotically pertinent
boundaries. Hence the cultural facility of color blinds, functioning on
differences in brightness - in a world that everyone else sees as
;· differentiated by hue. Red-and-green color-blind people talk of reds and
greens and all shades of it [sic] using the same words most of us assign to
objects of a certain color. They think and talk and act in terms of "object
color" and "color constancy" as do the rest of us. They call leaves green,
roses red.Variations in saturation and brilliance of their yellow gives [sic]
them an amazing variety of impressions. While we learn to rely on
differences of hue, their minds get trained in evaluating brilliance ....
Most of the red-and-green blind do not know of their defect and think we
see.things in the same shades they do. They have no reason for sensing
any conflict. If there is an argument, they fipd us fussy, not themselves
defective. They heard us call the leaves green and whatever shade leaves
have for them they call green.People of average intelligence never stop to
analyze their sensations.They are much too busy looking for what these
sensations mean.3

Commenting on this passage in his beautiful essay on "Colors and Cultures",

Sahlins not only insists on the thesis that colour is a cultural matter, but
remarks that every test of colour discrimination is rooted in a sort of referential
fallacy.4 Psychologists frequently assume that classifications of colours and
utterance of colour names are linked to the representation of an actuaf
experience; they assume that colour terms in the first instance denote the
immanent properties of a sensation. Therefore, many tests are contaminated by
this confusion between meaning and reference. When om: l!tters a colour tenn
one is not directly pointing to a state of the world (process of reference), but, on
the contrary, one is connecting or correlating that term with a cultural unit or
concept. The utterance of the term is determined, obviously, by a given
sensation, but the transformation of the sensory stimuli into a percept is in some
way determined by the semiotic relationship b'!tween the linguistic e�'J)ression
and the meaning or content culturally correlatecl to it.
Our problem, to quote Sahlins again, is "how then to reconcile these two
undeniable yet opposed understandings: color distinctions are naturally based,
albeit that natural distinctions are culturally constituted? The dilemma can only
be solved by reading from the cultural meaning of color to the empirical tests of
discrimination, rather than the other way around."

3
Arthur Linksz, P�J•siologJ' ofthe Ere C\ew York: Grune & Stratton, 1952) vol. 2.
4
Marshall Sahlins, "Colors and Cultures", Semiotica, vol. 15, no. 1 (1975) pp. 1, 22.
 HOW CULTURE COJ\'DITI01'S THE COLOURS WE SEE 161

II

I shall begin with verbal language for practical reasons, for it represents the
most p9werful and therefore the most familiar instrument people use for
defining the surrounding world and for communicating to each other about it. It
is not, however, impossible to imagine another sign system in which colours and
other elements of the world were indicated not by words but, say, by fingers (the
thumb means red, the forefinger blue, etc.). Since we have more sounds than
fmgers at our disposal, we are verbal animals. But things might have gone
differently: in the course of evolution we could have elaborated, instead of a
very flexible phonatory apparatus, a particular skill in emitting, by chemical
means, thousands of odours. .
Even if this were the case, our analysis of semiotic
systems would not basically change.
.
If you look into a traditional handbook on communication, you see that such
a process is represented as shown in fi gure 1.
Sender- - - - _., Message- - - - -..Addressee

I ., Code • f
Figure 1

In order for a process of communication to be successful, a common code is the

most elementary requirement. But the notion of code is still rather vague,
abstract.
If dealing with the Morse code, the problem is rather simple: a code is a list
of equivalences by virtue of which a given array of dots and dashes is made to
correspond, element to element, to a given series of alphabetic letters. A natural
language is a bit more complicated. We begin with a paradigm of phonemes
which do not correspond to anything at all, then these phonemes are made to
constitute a repertoire of meaningful units, lexemes. Lexemes are, in tum,
made to correspond, roughly speaking, to certain cultural entities: let me call
them, for the moment, meanings or concepts. Of course, the situation is not this
simple - there are syntactical and co-textual rules, not to mention phenomena
such as homonymy, polysemy and contextual meanings.
In order to make my discourse manageable for the present purpose, let me be
outrageously simple and assume that in order to make communication possible,
one needs a signification system. This principle holds for any sign system, from
natural language to naval flag signals. A semiotic approach attempts to define
the general conditions, for every system of signs, that allow processes of
communication on the basis of a given system of signification. It is highly
improbable to establish a communication process without an underlying
signification system, whereas, theoreticallv speaking, it is not impossible to
invent a signification system without using it in order to communicate - such a
procedure, though, would seem a waste of time and energy.
 UMBERTO ECO

A wry schematic signification system can he represented as follows (figure 2):

Substance
Content Form
-----------------1 j Sign function
Form
Expression Substance
Figure 2

A system of general types of physical entities generates expressions: the general

types are the form of the expression; their concrete and individual productions
and manifestations are substances. 1'atural language, for instance, is based
upon thirty or forty phonemes, organized by a phonological system establishing
abstract types. We recognize these abstract types regardless of the various ways
in which they are physically produced. Types, studied by phonology or
phonemics, refl(esent the emic aspect of language; concrete occurrences,
studied by phonetics, represent the etic aspect.\\'hat is emic involves linguistic
categories, whereas etic involves concrete sounds.
Phonemes are articulated to compose morphemes or - to be less technical -
lexical entities or words. Type-phonemes and type-lexical expressions consti­
tute a system of expression form, emically considered.The form of expression
is used to convey contents, in the sense in which the sign is traditionally defined
as a/iquid quid stat pro a/iquo or, as C. S.Peirce said, so.'llething which stands to
somebody for something else in some respect or capacity. Units of the
expression form are correlated to content units by a sign function.
What is content? Not the external world. Expressions do not signify things or
states of the world.At most, they are used to communicate with somebody about
states of the world. If I say that ravens are black and unicorns white, I am
undoubtedly uttering a statement about a state of the world. (In the first
instance, I am speaking of the world of our experience, in the second I am
speaking about a possible world of which unicorns are inhabitants - the fact that
. they are white is part of the state of affairs of that world.) However, a term like
"raven" or "unicorn" does not necessarily refer to a "thing": it refers instead to
a cultural unit, to an aspect of our organization of the world.
The content of a signification system depends on our cultural organization of
the world into categories. By "world", I do not necessarily mean physical world:
Euclid's world is not a physical one, but a possible universe organized into points,
Jines, planes, angles, and so forth. It is a self-sufficient universe in which there
are neither ravens nor unicorns, but only cultural units such as the concept of
similitude and none such as the concept of love or justice. I can communicate
about the Euclidean universe, making true or false assertions (I can, for
example, assert truly that the sum of the internal angles of a triangle is equal to
180° and falsely assert that two parallel Jines can meet in a given point of that
universe), but the units "triangle" and "line" are, in themselves, neither true
nor false. They are simply the pertinent or relevant elements of the Euclidean
 HOW CULTURE C01'DITI01'S THE COLOURS WE SEE

u nh·erse. T�us a signification system allows its possible users to isolate and
name what is relevant to them from a giwn point of Yiew.
Let us consider a classic example gi,en by many semantic handbooks:
Eskimos apparently have four words to designate four types of snow, while
Eu rope�ns have only one word and consider releYant only one specific state of
H20 in opposition to other states like ice and water. Of course a skier can
recognize different qualities of snow, but he always sees, and speaks of, the
same cultural unit, considered from different points of view according to certain
practical needs. Eskimos, on the other hand, see, perceive and think of four
different things in the same way in which I perceiYe, and speak of, two different
things when, about to skate on a lake, I ascertain whether there is water or ice.
This means that a given culture organizes the world according to given
practices, or practical pu�oses, and consequently considers as pertinent
different aspects of the world. Pertinence is a function of our practices.
According to a suggestion made by Luis Prieto,5 if I 'have on a table before
me a large crystal ashtray, a paper cup and a hammer, I can organize these
pieces of furniture of my limited world into a twofold system of pertinences. If
my practical purpose is to collect some liquid, I then isolate a positive class
whose members are the paper cup and the ashtray, and a negative class whose
only member is the hammer. If, on the contrary, my purpose is to throw a
missile at an enemy, then the heavy ashtray and the hammer will belong to the
same class, in opposition to the light and useless paper cup. Practices select
"'pertinences. The practical purpose does not, however, depend on a free
decision on my part: material constraints are in play, since I cannot decide that
the hammer can act as a container and the paper cup as a missile. Thus
practical purposes, decisions about pertinences and material constraints will
interact in leading a culture to segment the continuum of its own experience
into a given form of the conten�. To say that a signification system makes
communication processes possible means that one can usually communicate
only about those cultural units that a given signification system has made
pertinent. It is, then, reasonable to suppose now that one can better perceive
that which a signification system has isolated and outlined as pertinent.
Let us imagine an archaic community which has only two terms to designate
every possible kind of human being: a term equivalent to "man" and a term like
"barbarian" or "alien". The members of the community have two cultural units
at their disposal: for them, the many-coloured universe of featherless
two-legged mammals (among which we might distinguish black and Chinese,
Dane and Dutch, European and American, East and West German) is a black
and white universe split into "us" and "the others". Let us for the moment
disregard the fact that further properties can be associated to these cultural
units, namely that "men" are rational and friendly, while "barbarians" are
stupid, irrational and dangerous.The problem of the organization of content is,
of course, more complicated than this, and from the perspective of
5
Luis Prieto, Pertinence et pratique (Paris: Minuit, 1975).
 UMBERTO ECO
contemporary compositional analysis, what I called cultural units are more
finely subdivided into a network of minor semantic properties. Yet even when
we limit the domain of semiotic problems to be discussed, our fictional
community retains a note of verisimilitude. The ancient Greek subdivision of
humankind into Hellenes and "oi barbaroi" is more or less similar to mv
fittional model, as is the Nazi reorganization of humankind into Aryans and
"inferior races".
Imagine a "sci-fi" situation in which our planet is invaded by monsters from
two different galaxies. The aliens of Galaxy no. 1 are round, greenish and have
three legs and four eyes; the aliens of Galaxy no. 2 are elongated, brown and
have six legs and one eye. Certainly we would be able to distinguish and
deseribe both species, but as far as our defence is concerned, the aliens are all
"non-human". When the men of the terrestrial outpost first encounter the alien
avant-garde, they will probably perceive and signal those they meet as simply
aliens or monsters. Before we forge new terms to define their differences, we
would need scientific interaction, and at some point we would enrich and
reformulate our content form. But without such a collective reshaping of our
content system, our very ability to recognize aliens will be strongly influenced bv
our cultural categories.
In the same way, for the members of our fictional ancient society, it will be
difficult to ascertain the difference between a Viking and a Phoenician, as well
as the difference between their languages: at first they will all be "barbaroi" ,
speaking a non-language. Eventually, at _a more advanced stage of inter-racial
contact, someone will discover that Vikings are more aggressive and
Phoenicians more eager to entertain commercial relationships, thus facilitating
the reformulation of the content, the discovery of new pertinences, and the
invention_ of new expressions to designate these pertinences. A sort of
underlying discriminative ability will lead to a more refined system of
identification and categorization. But in the early stages of contact, categoriz­
ation will overcome discrimination.
At this point, l must introduce a new concept, the opposition between
restricted and elaborated code. In a further stage of inter-racial contacts, our
fictional society could split into two castes. Priests and merchants will be able to
distinguish Vikings from Phoenicians, probably for different purposes (mer­
chants because they are interested in dealing with the Phoenicians, and priests
because they suppose that Vikings can be easily converted). These two castes
will reorganize their content form, and coin new expressions to name these
different cultural units. But the rest of the citizenry, in order to be employed as
warriors, will still share a more restricted code; for them, "men" versus
"barbarians" will remain the only pertinent opposition. Thus at the same
moment in the same society there will be two different levels of social
organization; therefore there will be two different ways of thinking, perceiving
and speaking, based upon two different systems of signification or, better, upon
two different stages of complexity of the same system. As the Italian playwright
Dario Fo once said: the worker knows a hundred words, his master a thousand
 HOW CULTURE CONDITIOKS THE COLOURS WE SEE

- that is why he is the master. To know more words means to conceive of a

more refined organization of the content. \\ben our instinctive tendency to
discriminate produces a more subtle categorization, we acquire a more
powerful world view. In the course of this improvement, one changes one's
codes. ' .
Of course, such a passage from restricted to elaborated code happens not
only infra-culturally but inter-culturally as well, and in space as well as time.
Take the rodent universe. In Latin there is only one name to indicate two
different kinds of animal that the English call, rcspectiYely, "mouse" and "rat"
(figure 3). •

Latin English
:'\1ouse
Mus . . . . . . . . . . ., .
Rat

Figure 3

In Italian, we have two names, "topo" and "rano", but many Italians today
confuse the terms, using "topo" for both animals. This linguistic simplification
deters them from paying attention to the morphological differences between a
"little mouse" and a "big one" - an attitude that can produce a number of
· sanitary and social consequences.
Thus it is possible to say that the Latin term mus (and perhaps topo today)
referred to a sort of homogeneous pertinent portion of the content, while
the English names "mouse" and "rat" refer to two different pertinent units
(figure 4):

Latin: English: E C
�

EE] mouse

rat
XJ

Xz

Figure 4

The organization of content has to do with the empty cases I have filled up with
variables. The important semiotic problem here is how to describe the content
of these empty cases, as we are obliged to analyse them through other
expressions - in their tum having to be analysed by other expressions, and so on
ad infinitum. I return to the problem of colour and to the page of Aulus Gellius I
mentioned earlier; the problem of the categorization of colours involves such
empty cases.
1 66 UMBERTO ECO

III

Perception occupies a puzzling position, somewhere midway between semiotic

categorization and discrimination based upon mere sensory processes. Jean
Petitot, who is working on the material roots of linguistic categorization based
upon the mathematical theory of catastrophes, suggested to me that categoriz­
ation llhd discrimination do not interact in the universe of sounds as thev do in
the universe of colours.6
We can, it seems, identify sounds with remarkable precision, but once we
have perceived the emic difference between, say, pa and ba, we have difficulty in
discriminating between the different etic ways in which pa and ba can be
pronounced. Petitot suggests that this ability to categorize sounds is not
culturally but innately grounded, and postulates a brain mechanism called
"perceptual categorization" which would explain why verbal language is such a
paramount semiotic system. Such an innate ability in sound identification, and
such a difficulty in sound discrimination, are crucial for human language.
It is important that we can identify the thirty or forty phonemes which
constitute the phonological paradigm of a given language, but it would be
embarrassing (linguistically speaking) to be exaggeratedly sensitive to minimal
individual differences between the etic ways of uttering the same phoneme.
That is why, were I speaking, you could understand your native language even
though many of you would be able to guess that I was not nurtured at Oxford.
Your ability to discriminate accents has nothing to do with your etic competence
- at most it has to do . with paralinguistics or tonemics, which are entirely
different. The more you were to focus your attention on my sounds, thinking of
them as phonemes of your native tongue, the more you would be recognizing
them emically, independently of the accent; you would forget the accent and
directly catch phonological categories. Of course there are individuals specially
trained in discriminating tonemes (that is, the subtle nuances in the etic
production of sounds), such as actors or social workers interested in people's
national or regional origins. But theirs is an etic training which has nothing to
do with the emic training connected with the acquisition of a language as an
abstract type.
Our discrimination ability for colours seems to be greater: we can detect the
fact that hues gradually change in the continuum of a rainbow, though we have
no means to categorize the borderlines between different colours.Nevertheless,
when a given subject is exposed to a continuum of sounds ranging from the
syllable ba to the syllable pa, uttered in many etic ways, "k" will be the

J ean Petitot, work in progress (personal communication), with references to the work of
6
A. Liberman, N. Studdart-Kennedy, K. Stevens (on perception of the speech code),
Eimas, Massaro and Pisoni (on selective adaptation and features detectors), Eimas and
Mehler (on innate bases of categorial perception).
 HOW CL'LTl.RI CO!\'DITI01'S THE COLOL'RS WE SEE

"catastrophic point", where so-called feature detectors in the human brain

isolate the threshold between two emic categories:

..
ba.............................. pa
Figure 5
Our innate capacity for perceptual categorization enables a subject to perceiYe a
clear opposition between the two emic entities, ba and pa, and disregard etic
discrimination. But if the same subject is intensiYely e:\-posed to the stimulus ba,
the catastrophe point will slip to the left when he is once again intensiYely
c:\-posed to the full range of sounds represented in fi gure 5. This phenomenon
is called "selective adaptation" : the subject will acquire a quite seYere notion of
the emic type ba (and, probably, a snobbish sensitiYit to accents).
y

Th e opposite happens with colours. Let us consider two colours, a and b,

which are mutually adjacent in the spectrum. If the stimulus a is repeated, the
catastrophe point will slip to the right rather than to the left. This means that
the more a subject becomes acquainted with a stimulus, the more eager he will
be to assign similar stimuli to the category to which he has assigned the original
stimulus. Categorial training produces categorization ability for both sounds
and colours; but sound categories become more restricted, while colour
categories become more tolerant, and sensitiYity in discrimination decreases.
Of course a painter can be trained more in etic discrimination than in emic
categorization, but in the ell.-periments aboYe, the reaction of the subject is
determined by the fact that he is not freely concerned with sense data, but is
influenced by the aims of the laboratory ell.-periment. He is encouraged to isolate
categorial entities and reacts with categories already defined by language, eYen
though he speaks only to himself.
These experiences have nothing to do with what I previously said about
sign functions: to perceive phonemes or colours has to do with the emic analysis
of expressions, not with the correlation between e:\-pressions and contents. But I
smell in these ell.'J)eriences the presence of a more complicated semiotic
question.

IV

It has been said that colour discrimination, under laboratory conditions, is

probably the same for all peoples no maner what langu age the)' speak, though
psychologists also suggest that there is not only an ontogenetic but also a
phylogenetic increase in discriminatory competence. The Optical Society of
America classifies a range of between 7. 5 and 1 o million colours which can
theoretically be discriminated.
A trained artist can discriminate and name a great many hues, which the
_ pigment industry supplies and indicates with numbers, to indicate an immense
variety of colours easily discriminated in the industry. But the Farnsworth-
1 68 UMBERTO ECO
Munsell test, which includes 1 oo hues, demonstrates that the average
discrimination rate is highly unsatisfactory.Not only do the majority of subjects
have no linguistic means with which to categorize these 1 oo hues, but
approximately 68 per cent of the population (excluding colour defectives) make
a 'total error score of between 20 and 1 00 on the first test, which involves
rearranging these hues on a continuous gradation scale. Cases of superior
discrimination (only 1 6 per cent) scored from zero to 1 6. The largest collection
of English colour names runs to over 3000 entries (Maerz and Paul),7 but only
eight of these commonly occur (Thorndike and Lorge). 8
Thus average chromatic competence is better represented by the seven
colours of the rainbow, with their corresponding wavelengths in millimicrons
(figure 6):
Average
chromatic competence
Red
[8oer-650
64er-590 Orange

[ s 8erss o Yellow

[ 54er-490 Green

480-460 Blue
[450-440 Indigo
430-390 Violet
Figure 6
Square brackets indicate the thresholds where, according to modem experi­
ments, there are clear jumps in discrimination.This segmentation does seem to
correspond to our common experience, though it was not the experience of
Latin speakers, if indeed it is true that they did not clearly distinguish between
green and blue. It seems that Russian speakers segment the range of
wavelengths we call "blue" into different portions, goluboj and sinij. Hindus
consider red and orange a unified pertinent unit. And against the 3000 hues
that, according to David Katz,9 the Maori of New Zealand recognize and name
by 3000 different terms, there are, according to Conklin, the Hanun6o of the
Philippines, with a peculiar opposition between a public restricted code and
more or less individual, elaborated ones:
Color distinctions in Hanun6o are made at two levels of contrast. The
first, higher, more general level consists of an all-inclusive coordinate,
7
A.Maerz and R. Paul, A Dictiona,)' of Color (New York: Crowell, 1953).
8
E. L.Thorndike and I. Lorge, The Teacher's Word Book r!f30,000 Words (l'iew York :
Columbia University Press, 1962).
9
David and Rose Katz, Handbuch der Ps)'chologie (Basel: Schwabe, 1960) vol. 2.
 HOW CULTURE CO!':DITI01'S THE C OLOURS WE SEE 169
four-way classification which lies at the core of the color system. The four
categories are mutually exclusiw in contrastive contexts, but may overlap
slightly in absolute (i.e., spectrally) or in other measureable terms. The
second level, including several sublevels, consists of hundreds of specific
color ;· categories, many of which overlap and interdigitate.
Terminologically, there is "unanimous agreement" (Lenneberg, 1 953, p.
469 ) on the designations for the four Level I categories, but considerable
lack of unanimity - with a few e:\.'])lainable exceptions - in the use of terms
of Level 11. 1 0
Let us disregard Level II, which seems a case of many elaborated codes
differing from males to females and even from individual to individual. Let us
consider the various formats of Level II as idiolectal and quasi-professional
codes.
The three-dimensional color solid is divided by this Level I categorization
into four unequal parts; the largest is mabi:ru, the smallest malatu;• [see
figure 7]. While boundaries separating these categories cannot be set in
absolute terms, the focal points (differing slightly in size, themselves)
within the four sections, can be limited more or less to black, white,
orange-red, and leaf-green respectively. In general terms, mabi:ru
includes the range usually covered in English by black, violet, indigo,
blue, dark green, gray, and deep shades of other colors and mixtures;
malagti, white and very light tints of other colors and mixtures; marara,
maroon, red, orange, yellow, and mi:\.tures in which these qualities are
seen to predominate; malatuy, light green and mixtures of green, yellow,
and light brown. All color terms can be reduced to one of these four, but
none of the four is reducible. This does not mean that other color terms
are synonyms, but that they designate color categories of greater
specification within four recognized color realms. 1 1
Hanun6o segmentation follows our basic English paradigm only to a limited
extent, since it involves black, white and grey in different ways. What is
important for our present study is that the pertinentization of the spectrum
depends on symbolic, i.e. cultural principles. Note that these cultural
pertinentizations are produced because of practical purposes, according to the
material needs of the Hanun6o community.
The basis of this Level I classification appears to have certain correlates
beyond what is usually considered the range of chromatic differentiation,
and which are associated with linguistic phenomena in the external
environment.
First, there is the opposition between light and dark, obvious in the
contrast of ranges of meaning of lagti and biru. Second, there is an
10
Harold C. Conklin, "Hanun6o Color Categori es", Southwestern Journal ofAnthro-
pology, vol. II (1955 ) pp. 339-44; see p. 34 1 .
1 1 Ibid.,
pp. 34 1-2.
 UMBERTO ECO

oppos1t1on between dryness or desiccation and wetness or freshness

(succulence) in visible components of the natural environment which are
reflected in the terms rara and latuy respectively. This distinction is of
particular significance in terms of plant life.Almost all living plant types
possess some fresh, succulent and often "greenish" parts.To eat any kind
of raw, uncooked food, particularly fresh fruit or vegetables, is known as
sag-laty-un (/atuy). A shiny, wet, brown-colored section of newly cut
bamboo is malatuy not marara. Dried-out or matured plant material such
as certain kinds of yellowed bamboo or hardened kernels of mature or
parched com are marara. To become desiccated, to lose all moisture, is
known as mamara < para "desiccation." A third opposition, dividing the
two already suggested, is that of deep, unfading, indelible, and hence
often more desired material as against pale, weak, faded, bleached, o r
"colorless" substance, a distinction contrasting mabi:ru and marara with
malagti and malatuy. 1 2 . .

MABl:RU Spectral opposition MARARA

Indelible Indelible

Weak Weak

MALATUY Intensity opposition MALAGTI

Figure 7

We have then a system of cultural units - lightness, darkness, wetness,

dryness - which are expressed by four fundamental colours; these colours are,
in tum, four cultural units expressed by four linguistic terms. This double
organization of the content depends, as does any organization of this kind, on a
system of disjunctions: it represents a structure.Just as a "mouse", within a
semantic space concerning rodents, is everything which is not a "rat", and vice
versa, so the pertinent content space of malatuy is determined by its northem
borderline beyond which there is marara, and its southern borderline, below
which there is mabi:ru.
Geopolitically speaking, Holland is a negative concept: it is the class of all
12
Ibid., p . 342.
 HOW CCLTLTRE co;-.;rJITJO;-.;s THE COLOl'RS \',E SE<: 171

points adjacent to, but not, Germany. Bc1¢um or the �orth Sea. The same
principle hol? � for all other geopolitical c,--prcssion s such as Gcrma�y or ltaly or
the SoYict l..m10n. In any system, whether gcopohncal __ _
or chromatic or lcx1cal,
units arc defin.d not in themselYes but in terms of opposition and position in
relation to other units. There can be no units without a system. The different
waYs in which cultures make the continuum of colours pertinent, thereby
caiegorizing and identi�·ing hues or chromatic units, correspond to different
content systems. This semiotic phenomenon is not independent of perception
and discrimination ability; it interacts with these phenomena and frequently
o\"crwhelms them.
Consider again our diagram of colours (figure 8): it takes into account not
onlv differences in the organization of content among different contemporary
cultures, but also depicts different leYels of complexity within a given culture as
well as differences through the ages. It represents a reduced model of a
tri-dimensional system of differences.
At this point we can probably tackle .\ulus Gellius' puzzle. Rome, in the
second century A.D., was a very crowded crossroads of many cultures. The
Empire controlled Europe from Spain to the Rhine, from England to North
Africa and the Middle East. All these cultures, with their own chromatic
sensitivities, were present in the Roman crucible. Diachronically speaking,
Aulus Gellius was trying to put together the codes of at least two centuries of
Latin literature and, synchronically speaking, the codes of different non-Latin
cultures. Gellius must have been considering diYerse and possibly contrasting
cultural segmentations of the chromatic field. This would e"J>lain the
contradictions in his analysis and the chromatic uneasiness felt by the modem
reader. His colour-show is not a coherent one: we seem to be watching a
flickering TV screen, with something wrong in the electronic circuits, where
tints mix up and the same face shifts, in the space of a few seconds, from yellow
to orange or green. Determined by his cultural information, Gellius cannot
trust to his personal perceptions, if any, and appears eager to see gold as red as
fire, and saffron as yellow as the greenish shade of a blue horse.
We do not know and we shall neYer know how Gellius really perceived his
Vmwelt; unfortunately, our only evidence of what he saw and thought is what he
said. I suspect that he was prisoner of his cultural mish-mash.
- - Yet it also seems to me (but obYiously this h)J>Othesis should be tested on
more texts) that Latin poets were less sensitive to clear-cut spectral oppositions
or gradations, and more sensitive to slight mixtures of spectrally distant hues. In
. other words, they were not interested in pigments but in perceptual effects due
to the combined action of light, surfaces, the nature and purposes of objects.
Thus a sword can be fufoa as jasper because the poet sees the red of the blood it
may spill. That is why such descriptions remind us more of certain paintings of
-.fr-anz Marc or of the early Kandinsky than of a scientific chromatic polyhedron.
A... a decadent man of culture, Gellius tends to interpret poetic creativity and
.,.--- irr,cntion as a socially accepted code and is not interested in the relationships
- "'1lich colours had with other content oppositions in different cultural systems.
It would be interesting to transform a given Latin chromatic system, that of
t
 mµ. Average English I Latin I Hanun6o Level 1 I Hanun6o Level 2

r
800-650 Red
Marara
640-590 Orange (dry)
f/ �� I. -�
� ..
Flavus

.!S I �
580-550 Yellow
jl i
�:i
Malatuy

,�,�
(fresh)
540-490 Green

·I 480-460 I Blue I \ Glaucus

l Mabi:ru
(rotten)
1 I
450-440 Indigo I
I
\
'
I
I I I - I '

:s� I a
430-390 I Violet I

Figure 8
\ I �
....�...
I�:j I
 •
HOW CULTURE COl\'DITIOl\"S THE COLOURS WE SEE 1 73
Virgil for example, into a structure more or less like the one I proposed for the
Hanun6o system, where the names of hues must be associated to opposition
between dark and light (also in psychological and moral sense), euphoric and
dvsphoric, excitation and calm, and so on. The names of colours, taken in
{hemselv�!;,, have no precise chromatic content: they must be viewed within the
general context of many interacting semiotic systems.

Are we, in any sense; freer than Gellius from the armour of our culture? We arc
animals who can discriminate colours, but we are, above all, cultural animals.
Human societies do not only speak of colours, but also with colours. We
frequently use colours as semiotic devices: we communicate with flags, traffic
lights, road signs, various kinds of emblems.
Now a socio-semiotic study of national flags 1 3 remarks that national flags
make use of only seven colours: red, blue, green, yellow, orange, black and
white. For physical reasons, the proportion of these colours is as follows:

Colour combinations Per cent

Red/white/blue 1 6.8
Red/white 9 .5
Red/yellowIgreen 7.3
Red/white/green 6.6
Red/white/green/black 6.6
Blue/white 6.o
Red/yellow/blue 5.8

58.6

Orange, hardly distinguishable from red, is rarely used. What counts in the
perception of a flag is categorization, not discrimination. If we were to look up
the flags of the Scandinavian countries, we would realize that the blue of the
Swedish · and Finnish flags (which is light) is different from the blue of the
Icelandic and Norwegian ones (which is dark). Now look at Sweden's yellow
cross on a light blue field - there is not a flag in the world with a yellow cross on
a dark blue background, and for good reason. Everyone would recognize such a
flag as the symbol for Sweden. (And, thinking of Norway's dark blue cross on a
red field, a flag with a light blue cross on red would similarly be recognized as
Norway's symbol.) In national flags, categorization overwhelms discrimination.

n Sasha R. Weitman, "National Flags ", Semiotica, vol. 8, no. 4 (1 973) pp. 328-67.
174 UMBERTO ECO

This simplification exists not only for reasons of easier perception: such "easier
perception" is supported by a preYious cultural coding by virtue of which
certain colours form a clear-cut system of oppositional units which are, in tum,
clearly correlated with another system concerning values or abstract ideas.
, . In the study on national flags I have been referring to, it is interesting to
check the symbolic values assigned by different countries to the same colour.
Red, for example, symbolizes bravery, blood and courage in many countries
(Afghanistan, Austria, Italy, Bulgaria, Burundi, Chile, Ecuador, etc.), but it also
represents animals in Bolivia, faith in Ethiopia, soil in Dahomey. White, almost
universally, stands for peace, hope and purity, but in Congo Kinshasha, hope is
represented by blue which, for the majority of countries, stands for sky, sea and
rivers. The colours of national flags are not colours: physical pigments; they are
expressions correlated to cultural units, and as such are strongly categorized.
But the real problem is not - or not only - that our discrimination ability is
limited to few colours. It is that the system of basic values to be ex-pressed by
colours is a limited one. The nature of these values (hope, peace, and so on) is
irrelevant: what counts is the structural architecture of their basic oppositions,
which must be clear.
One should remark that a greater variety of colours exists, or existed, in
heraldry . But heraldry represents a case of an elaborated code for a cultivated
minority able to discriminate more colours and associate more refined names to
different hues, as well as memorize numerous aristocratic stocks.
The same strong categorization is at work in traffic lights and road signals. A
traffic light can work and transmit its orders irrespective of the shade of green,
red or yellow that, in terms of wavelengths, it emits. One would certainly stop at
a traffic light with an orange light on, and continue moving even though the
green light were a shade of blue. (Note that in the traffic light code, the
signification of colours is reinforced by the position of the lights, which reduces
the relevance of hues - and helps the colour-blind.) In any case, here too, in
traffic regulation, people can only recognize a limited system of obligations. I do
not think it is possible to found a system of communication on a subtle
discrimination between colours too close to each other in the spectrum. This
may seem strange since, as I have said, we potentially have a great capacity for
discrimination, and with ten million colours it would be interesting to compose
a language more rich and powerful than the verbal one, based as it is upon no
more than forty phonemes. But the phonemes of verbal language are, in fact, a
reasonable reduction of the great variety of possible sounds that our phonatory
apparatus can produce. The seven colours of flags and signals are probably the
most a human culture can recognize - by a general agreement as to categorizable
expressive entities. This agreement has come about, probably, because verbal
language has shaped our average sensitivity according to the macroscopic
segmentation represented by the seven colours of the rainbow which is a
Western conventional way of segmentation. The agreement has also come
about because average verbal language, with its polysemy, works better for
common people when many names stand synonymously for few basic concepts,
 HOW CCLTLTRE co:-.DITIO'.'\S THE COLOCRS WE SEE 1 75
rath er than the opposite, when few names stand homonymously for thousands
of concepts.
The fact that a painter (think of Paul �lee) can recognize and name more
colours, the fact that verbal language itself is able not only to designate hundreds
of nuances, but also describe unheard-of tints by examples, periphrases and
poetic ingenuity - all this represents a series of cases of elaborated codes. It is
common to every society to have members able to escape the determination of
the rules, to propose new rules, to behave beyond the rules.
In everyday life, our reactivity to colour demonstrates a sort of inner and
profound solidarity between semiotic systems . Just as language is determined by
the way in which society sets up systems of values, things and ideas, so our
chromatic perception is determined by language. You may look up your flags
again: suppose there is a football match between Italy and Holland. One will
distinguish the Dutch flag from the Italian one, even tho�gh the red of either of
them, or of both, were looking orange. If, on the contrary, the match were
between Italy and Ireland, the Italian flag would be characterized by a dark red,
since white, green and orange are the Irish colours.
If one wants to oppose, for shorthand purposes, a Mondrian to a Kandinsky,
Mondrian would be recognizable even though its reds were more or less
orange, but in the course of an aesthetic discourse on Mondrian, and in judging
the correctness of an art book's reproduction, one should spend much careful
analysis in discriminating the better and more faithful colour among Mondrian
reproductions.
Thus the artistic activity, be it the poetry of Virgil or the research on
pigments by Mondrian, works against social codes and collective categorization,
in order to produce a more refined social consciousness of our cultural way of
defining contents.
Ifpeople are eager to fight for a red, white and blue flag, then people must be
ready to die even though its red, due to the action of atmospheric factors, has
become pinkish. Only artists are ready to spend their lives imagining (to quote
-James Joyce) "an opening flower breaking in full crimson and unfolding and
fading to palest rose, leaf by leaf and wave of light by wave of light, flooding all
the heavens with its soft flushes, every flush deeper than the other".