Bleikr, Gulr, and the Categorization

of Color in Old Norse

Jackson Crawford, University of California, Berkeley

It is well known that human languages classify the experience of color in

various ways, but this is a matter of conceptualization, not of perception
(the physiological exceptions imposed on blind and color-blind speak-
ers excepted). For instance, an English speaker, a Russian speaker, and a
Welsh speaker will likely agree that a clear sky, a sapphire, and a blade of
grass are not identical in color. But in classifying these color perceptions,
their native languages will differ. In English, the color of both the clear
sky and the sapphire is classified as blue, though perhaps distinguishable
as light blue and dark blue, while in Russian, the distinction between light
blue (goluboj) and dark blue (sinij) is codified in two different categories;
goluboj and sinij are not merely shades of a larger color category but are
categorized differently just as English blue and green are. Meanwhile
Welsh categorizes the color of all three objects under one color category
(glas)—so English blue and green are both types of glas.
This phenomenon was recognized as early as the second century BC,
as witnessed in the Attic Nights of Aulus Gellius, where it is reported that
the philosopher Favorinus said:
[P]lura . . . sunt in sensibus oculorum quam in verbis vocibusque colorum
discrimina. Nam, ut alias eorum inconcinnitates omittamus, simplices isti
rufus et viridis colores singula quidam vocabula, multas autem species dif-
ferentis habent. Atque eam vocum inopiam in lingua magis Latina video
quam in Graeca. Quippe qui “rufus” color a rubore quidem appellatus est,
sed cum aliter rubeat ignis, aliter sanguis, aliter ostrum, aliter crocum, aliter
aurum, has singulas rufi varietates Latina oratio singulis propriisque vocabulis
non demonstrat omniaque ista significat una “roboris” appellatione, nisi
cum ex ipsis rebus vocabula colorum mutuatur et “igneum” aliquid dicit
et “flammeum” et “sanguineum” et “croceum” et “ostrinum” et “aureum.”

(More distinctions of colour are detected by the eye than are expressed by
words and terms. For leaving out of account other incongruities, your simple
colours, red (rufus) and green (viridis), have single names, but many differ-
ent shades. And that poverty in names I find more pronounced in Latin than
in Greek. For the colour red (rufus) does in fact get its name from redness,
but although fire is one kind of red, blood another, purple another, saffron
another, and gold still another, yet the Latin tongue does not indicate these

Journal of English and Germanic Philology—April

© 2016 by the Board of Trustees of the University of Illinois

JEGP 115_2 text.indd 239 1/27/16 12:21 PM

 240  Crawford
special varieties of red by separate and individual words, but includes them
all under the one term rubor, except in so far as it borrows names from the
things themselves, and calls anything “fiery,” “flaming,” “blood-red,” “saffron,”
“purple,” and “golden.”)1

The highest-level color categories of a human language—those that

are not perceived by speakers as variants of another color category—are
its basic color terms. This concept was named by Brent Berlin and Paul
Kay,2 who presented guidelines to help determine whether a color term
in a given language is basic or not. Their four primary expectations for
a basic color term are that it: 1) should be monolexemic, and therefore
not a compound or otherwise divisible into components, 2) should not be
considered a subclass of another color by speakers (indeed, all other color
terms in the language should be considered subclasses of the basic color
terms), 3) should not be restricted to only a limited class of referents, and
4) must be psychologically salient, that is, it must be readily educed and
used by speakers. Secondary (nonsemantic) criteria were also presented
for the evaluation of cases that these criteria might leave doubtful, and
additional criteria have been proposed by later researchers.3
It is important to note that these criteria were not intended as strict
requirements that all basic color terms in all human languages must sat-
isfy, but rather as guidelines to be adapted to the unique situation of each
language,4 as not all the suggested criteria are equally applicable to all
languages—for instance, in an agglutinative language in which all color
adjectives are derivative of nouns, it would be impossible to determine the
basic color terms if the criterion of monolexemicity were strictly adhered
to. Similarly, in the case of extinct languages such as Old Norse, for which
no native speakers are available, many of these guidelines must be modi-
fied or replaced by functionally similar criteria. In particular, psychological
salience must be indirectly assessed by such measures as relative frequency,
morphological productivity, and the degree and prominence of nonliteral
or idiomatic usage.5

1. Latin text and English translation is from John C. Rolfe, ed. and trans., The Attic Nights
of Aulus Gellius (New York: Putnam, 1927), pp. 210–13.
2. Brent Berlin and Paul Kay, Basic Color Terms: Their Universality and Evolution (Berkeley:
Univ. of California Press, 1969). See especially p. 6 for their earliest definition of a basic
color term.
3. See especially C. P. Biggam, The Semantics of Colour: A Historical Approach (Cambridge:
Cambridge Univ. Press, 2012), which contains the most up-to-date survey of the methods
for determining a basic color term, esp. pp. 21–44.
4. Berlin and Kay, Basic Color Terms, p. 6; Biggam, Semantics of Colour, p. 22.
5. Jackson Crawford, “Blåe og svarte augo: skipnaden av fargeuttrykk i norrønt,” Norsk
lingvistisk tidsskrift, 31.2 (2013), 147–64.

JEGP 115_2 text.indd 240 1/27/16 12:21 PM

 Categorization of Color   241

THE SEQUENCE OF ACQUISITION OF BASIC COLOR TERMS

Berlin and Kay are famous for their suggestion that human languages have
a restricted number of possible inventories of basic color terms, and that in-
ventories of fewer basic color terms will have more predictable constituents.
Based on their observations of color categorization in the ninety-eight lan-
guages they originally studied, they proposed that this synchronic predict-
ability—for instance, the fact that if a language had only three color terms,
they invariably were near in focus to English black, white, and red—cor-
responded to a predictable, universal, and cumulative diachronic sequence
in which basic color terms would emerge in any given language.
The form of this sequence as presented in Berlin and Kay 1969 has be-
come outdated, although it is often still cited in recent studies. The current
scheme6 for the first three stages of the evolution of basic color terms is
a gradual break-up of a Stage I system that distinguishes only light/warm
vs. dark/cool into a Stage V with at least six basic color terms (black, blue,
green, red, white, and yellow—the so-called Hering primaries). Stages I
and II are uniform (slashes join colors that fall into compositional color
categories together, for example, “red/yellow” is a color term that refers
both to colors that in English would be called “red” and to those that
would be called “yellow”):
Stage I:
light/warm (white/red/yellow)
dark/cool (black/green/blue)

Stage II:
white
red/yellow
black/green/blue

However, the intermediate Stages III and IV, during which these earlier
compositional color categories break down, allow much more variety than
in earlier models. The following diagrams demonstrate the possible vari-
ants of Stages III and IV,7 leading finally to the uniform Stage V. Note that
gray and brown may emerge at any point in this sequence after Stage I:

6. The most up-to-date version of Paul Kay’s theories (incorporating much earlier work
by Berlin, Biggam, Kay, and other scholars) is to be found in Paul Kay et al., “Color Naming
across Languages,” in Color Categories in Thought and Language, ed. C. L. Hardin and Luisa
Maffi (Cambridge: Cambridge Univ. Press, 1997), pp. 21–56, with some modifications in
Paul Kay and Luisa Maffi, “Color Appearance and the Emergence and Evolution of Basic
Color Lexicons,” American Anthropologist 101.4 (1999), 743–60.
7. Following Kay et al., “Color Naming,” pp. 31–33. The abbreviations of basic color terms
that appear in the names of the variants of Stages III and IV are standard in the works of

JEGP 115_2 text.indd 241 1/27/16 12:21 PM

 242  Crawford
Variants of Stage III: IIIG/Bu IIIBk/G/Bu IIIY/G/Bu
white white white
red/yellow red red
yellow yellow/green/blue
green/blue black
black black/green/blue

Variants of Stage IV: IVBk/Bu IVG/Bu IVY/G

white white white
red red red
yellow yellow yellow/green
green green/blue blue
black/blue black black

Stage V:
white
red
yellow
green
blue
black

While this theory predicts that a language will distinguish between red
and yellow if it distinguishes between blue and green, this is not obviously
the case in Old Norse, where there is no confusion between blue and green
but frequent confusion between red and yellow (from the perspective of a
speaker of Modern English—for instance in the description of gold, egg
yolks, and the sun as red). Indeed, it has not previously been satisfactorily
demonstrated that Old Norse has a basic color term for yellow, though
gulr has been claimed to fill this role.8
In research for this and related studies, I have excerpted all occur-
rences of color terms as well as of adjectives frequently associated with
color terms (such as fagr, døkkr, and ljóss), whether as a simplex or as part
of a compound, in the Poetic Edda,9 the Prose Edda,10 the Íslendinga sǫgur

Paul Kay and associated scholars: Bk (Black), Bu (Blue), G (Green), R (Red), W (White), Y
(Yellow). “Stage IIIG/Bu” is then to be read as the variant of Stage III wherein green and
blue are not distinguished.
8. In Kirsten Wolf, “Towards a Diachronic Analysis of Old Norse-Icelandic Color Terms: The
Cases of Green and Yellow,” Orð og tunga, 12 (2010), 109–30; as well as in Wolf, “Some Com-
ments on Old Norse-Icelandic Color Terms,” Arkiv för nordisk filologi, 121 (2006), 173–92;
and Georg C. Brückmann, Altwestnordische Farbsemantik (Munich: Herbert Utz Verlag, 2012).
9. Edda: Die Lieder des Codex Regius nebst verwandten Denkmälern I: Text, ed. Gustav Neckel,
5th ed., rev. Hans Kuhn (Heidelberg: Winter, 1983).
10. Snorri Sturluson, Edda: Prologue and Gylfaginning, ed. Anthony Faulkes (New York:
Oxford Univ. Press, 1982), Snorri Sturluson, Edda: Háttatal (New York: Oxford Univ. Press,
1991), Snorri Sturluson, Edda: Skáldskaparmál, ed. Anthony Faulkes (London: Univ. College
London, 1998).

JEGP 115_2 text.indd 242 1/27/16 12:21 PM

 Categorization of Color   243

and þættir,11 and Heimskringla.12 When “the texts excerpted” are referred
to in this study, it is these texts that are meant. I have also compared the
occurrences of each term gained from excerpting these texts to those
occurrences reported for each term in the Ordbog over det norrøne
prosasprog (ONP), <http://dataonp.hum.ku.dk/wordlist_d_menu.html>.

OLD NORSE GULR

Old Norse gulr, which gives rise to the basic color term for yellow in the
descendent West Scandinavian languages (Faroese and Icelandic gulur,
Norwegian gul, probably Norn *gol13), is exceedingly rare in classical Old
Norse texts. The word never occurs in Eddic poetry,14 and in the prose
of the Íslendinga sǫgur and Heimskringla considered together, it occurs
only seven times, five of which occurrences are in describing the color
of human hair.
At the deepest level reconstructible, gulr is formed from the putative
Indo-European root *ǵhel-, widely attested as a root for color terms, words
for bile and gold, and for the names of distinctively colored animals and
plants. While there are notable divergences toward other colors in the
Celtic, Albanian, and Balto-Slavic branches especially, the overwhelming
testimony of most branches of the family points to the color yellow or
green, and with the root’s frequent association with words that mean
“gold” or “shine,” it seems most likely to have been a root with a primary
meaning of “yellow.” Note that the probable presence of a color term in
an ancestor language does not necessarily imply the presence of that color
category in a descendent language, since basic color categories are (in

11. From the volumes in the Ìslenzk fornrit series (Reykjavík: Hið íslenzka fornritafélag):
Ljósvetninga saga, ed. Björn Sigfússon (1950); Vestfirðinga sǫgur, ed. Björn Þórólfsson and
Guðni Jónsson (1943); Laxdǿla saga, ed. Einar Ól. Sveinsson (1934); Vatnsdǿla saga, ed. Einar
Ól. Sveinsson (1939); Brennu-Njáls saga, ed. Einar Ól. Sveinsson (1954); Eyrbyggja saga, ed.
Einar Ól. Sveinsson and Matthías Þórðarson (1935); Grettis saga, ed. Guðni Jónsson (1936);
Kjalnesinga saga, ed. Jóhannes Halldórsson (1959); Austfirðinga sǫgur, ed. Jón Jóhannesson
(1950); Eyfirðinga sǫgur, ed. Jónas Kristjánsson (1956); Egils saga Skalla-Grímssonar, ed. Sig-
urður Nordal (1933); Borgfirðinga sǫgur, ed. Sigurður Nordal and Guðni Jónsson (1938);
Harðar saga, ed. Þórhallur Vilmundarson and Bjarni Vilhjálmsson (1991).
12. Heimskringla, vols. 1–3, ed. Bjarni Aðalbjarnarson (Reykjavík: Hið íslenzka fornritafélag,
1991).
13. Inferable from golgrav ‘a gutter in the byre for the cattle’s urine’, golmoget ‘dark-
coloured with lighter (yellowish, whitish) belly’. See these entries in Jakob Jakobsen, An
Etymological Dictionary of the Norn Language in Shetland (London: Nutt, 1928).
14. As noted already in one of the earliest studies of color terms in Old Norse: Arthur
Lawrensson, The Colour Sense in the Edda (London: Murray, 1882), pp. 737–38.

JEGP 115_2 text.indd 243 1/27/16 12:21 PM

 244  Crawford

rare cases) lost, as the case of some Italian dialects shows,15 and in other
cases the basic color category may be retained while the term used for it
changes (as in English, where black has usurped the place of Old English
sweart).
The Old Norse noun gull (gold), together with its cognates in other
Germanic and Balto-Slavic languages, such as Old English gold, Gothic
gulþ, Old Church Slavic zlato, is to be traced to a derived zero-grade nomi-
nal formation *ǵhḷ-tó- from this root; it is, etymologically, a substance with
the property indicated by *ǵhel-, thus very likely “the yellow stuff.”
An adjective derived from the same root, *gelwaz, can be reconstructed
for Proto-West Germanic on the strength of Old English geolo, Old High
German gelo, Old Saxon gelu, all with the meaning yellow,16 but it does
not have attested cognates in East or North Germanic. Old Norse gulr,
which is frequently cited as a precise cognate with these West Germanic
terms,17 cannot in fact derive from *gelwaz, which would give Old Norse
*gølr, not gulr, by w-umlaut.18 There is therefore no color term for yellow
that is definitely reconstructible for Proto-Germanic, though it is more
likely than not that *gelwaz dates to that stage of the language’s develop-
ment, especially since *gel-waz is formed very similarly to firmly established
Proto-Germanic color terms such as *blē-waz ‘blue’19 and *grā-waz ‘gray’.
Gulr, which is restricted to North Germanic, may be a late, simple ad-
jective derived from the noun gull ‘gold’. Compare the similarly formed
adjective ljóss ‘light’ from the noun ljós; in both cases the uninflected
nominal root (gull-, ljós-) has simply been extended with the addition of
adjectival endings to form an adjective (with the expected syncope of the
medial *l in *gullr > gulr). The meaning in this case would be “with the
properties of gold.”
Yet in spite of its etymological association with gull, and its transparent
relationship with that word, gulr is never used to describe gold in Old
Norse. As mentioned above, five of its seven occurrences in the prose of
the Íslendinga sǫgur, Íslendinga þættir, and Heimskringla are in descriptions

15. See A. M. Kristol, “Color Systems in Southern Italy: A Case of Regression,” Language,
56 (1980), 137–45.
16. Frank Heidermanns, Etymologisches Wörterbuch der germanischen Primäradjektive (Berlin:
de Gruyter, 1993), pp. 240, 262.
17. E.g., in Earl R. Anderson, Folk-Taxonomies in Early English (Madison, NJ: Fairleigh
Dickinson Univ. Press, 2003).
18. Cf. Adolf Noreen, Altisländische und altnorwegische Grammatik, 4th ed. (Halle: Niemeyer
1923), p. 80.
19. Note that these glosses for reconstructed Proto-Germanic color terms should be taken
with grains of salt, showing etymological relationship rather than (necessarily) meaning.
Semantic drift and the replacement of vocabulary over time means that the precise seman-
tic value of a reconstructed color term in an unattested language is probably impossible to
establish, despite efforts such as that in Anderson’s Folk-Taxonomies to determine the basic
color terms of Proto-Indo-European and Proto-Germanic.

JEGP 115_2 text.indd 244 1/27/16 12:21 PM

 Categorization of Color   245

of human hair; the other two describe a lion sewn onto a jacket20 and the
fibers of an exposed human heart, in a list of four colors (grǿnn, gulr, hvítr,
rauðr) seen in these fibers.21 I consider it likely that in both of the latter
cases, the color term is deliberately chosen for its evocation of a certain
(probably blonde) shade of human hair.
The formula gult sem silki22 (g. as silk) describes human hair and paral-
lels formulations such as hvítr sem dript (white as driven snow), rauðr sem
blóð (red as blood), svartr sem tjara (black as tar), and the mysterious blár
sem hel (blue as the realm or goddess Hel). This suggests that silk might
have been a “best example” of gulr in thirteenth- century Iceland, but this
might refer to the material’s luster rather than to a particular hue, since
silken products in Old Norse literature are not necessarily chromatically
restricted; they may, for instance, be red: King Magnús the Good wears a
rauð silkiskyrta (red silken shirt) (Magnúss saga ins góða, 43.20). However,
silk is also used as the referent of comparison in describing hair color in
the expression fagrt sem silki (fair as silk) (Laxdǿla saga, 77.2), and the term
silkibleikr (silk-blonde?) occurs once as a descriptor of hair color as well
(Magnúss saga berfǿtts, 227.25). Since fagr is often used to describe hair,
and is frequently collocated with bjartr (bright), ljóss (light, shining), and
hvítr (white), the feature of silk being alluded to with its comparison to fagr,
bleikr, and gulr in describing the appearance of hair is probably the luster
of the hair and silk rather than a particular hue. Noting that undyed silk
is yellow,23 it may yet be the case that the defining characteristic of gulr in
early Old Norse was gold-like luminance or brightness rather than hue.24
I consider it more likely that gulr is in fact to be understood as a hue,
but that this hue was originally not yellow/blonde but rather reddish, con-
sidering that gold itself is always described as rauðr (red). Indeed a similar
situation may have prevailed in early Latin, where gold (aureum) was con-
sidered red, as suggested by a quotation from Varro’s De Lingua Latina:
Quod addit rutilare, est ab eodem colore: aurei enim rutili, et inde etiam mulieres
valde rufae rutilae dictae (As for his addition of rutilare “to be red,” that is
from the same colour; for rutuli is an expression for golden hair, and from
that also women with extremely red hair are called rutilae “Goldilocks”).25
The name Gullinkambi (golden-comb) for a rooster in Vǫluspá also suggests
that the color implied by gold, or adjectives derived from it such as gullinn,

20. Magnúss saga berfǿtts, chap. 24, in Heimskringla, ed. Bjarni Aðalbjarnarson, III, 235.
21. Fóstbrǿðra saga, chap. 24: see Vestfirðinga sǫgur, ed. Björn K. Þórólfsson and Guðni
Jónsson (Reykjavík: Hið íslenzka fornritafélag, 1943), p. 276.
22. Óláfs saga helga, chap. 102, and Ólafs saga kyrra, chap. 1, both in Heimskringla, ed.
Bjarni Aðalbjarnarson, II, 172, and III, 203, respectively.
23. Thor Ewing, Viking Clothing (Stroud: Tempus, 2006), p. 157.
24. Also alluded to as a possibility in Wolf, “Towards a Diachronic Analysis,” pp. 109–30.
25. Text and translation from Varro on the Latin Language, ed. and trans. Roland G. Kent
(Cambridge, MA: Harvard Univ. Press, 1938), p. 337.

JEGP 115_2 text.indd 245 1/27/16 12:21 PM

 246  Crawford

gullslitr, gulr, is reddish rather than yellowish. This leads to the next point,
the much-discussed problem of red gold in early Germanic languages.

THE BELABORED CASE OF RED GOLD

Gold is stereotypically rauðr in Old Norse, accounting for approximately

10 percent of all occurrences of the term in the texts excerpted (blood,
by comparison, accounts for another one-third). The association of the
color red with this material is pan-Germanic; the cognate color term rot
in Middle High German appears fifty-four times in the Nibelungenlied;
eighteen of these occurrences describe the color of gold.26 In Old English,
the cognate color term rēad is used more often to describe the color of
gold than the color of blood.27
Gold and golden objects are blóðrauðr (blood-red)28 and glóðrauðr
(ember-red),29 which may indicate purity of the substance rather than of
the hue. Gold is so stereotypically red that a compound gullroðinn (gold-
reddened) is attested.30
The collocation bleikt gull (yellow/pale gold?) does not occur in the texts
excerpted, but is (uncommonly) attested. One of its few occurrences, in
Rauðúlfs þáttr, is contrasted with desirable rautt gull, and thus probably
indicates an undesirable alloy of gold with some cheaper material.31
The red gold phenomenon has been discussed by many scholars, with one
of the most natural suggestions being that the gold in use in early medieval
northern Europe must have contained more copper than is normal in the
modern period. However, the finds of golden objects from this period do
not appear especially reddish to modern eyes. Earl Anderson suggests that
Old English rēad (and by extension its cognates in other old Germanic
languages such as Old Norse) had two focal points, one in the color of
blood and another in the color of gold. This may be true of Old English,
with which Anderson is primarily concerned, and where gold does indeed
outweigh blood as the principal referent of the color term rēad.32

26. Anderson, Folk-Taxonomies, p. 133.

27. Anderson, Folk-Taxonomies, p. 133.
28. Haralds saga Tryggvasonar, chap. 49, in Heimskringla, ed. Bjarni Aðalbjarnarson, I, 297.
29. All occurrences from the excerpted texts are in Eddic poetry, in the poems Fáfnismál,
Guðrúnarkviða ǫnnur, and Atlamál en grǿnlenzku. See Edda, ed. Neckel, pp. 182, 184, 224,
250.
30. ONP.
31. Rauðúlfs þáttr, ed. Anthony Faulkes (London: Univ. College London, 2011), p. 38.
32. The arguments in this paragraph, as well as others less persuasive—namely the nine-
teenth-century belief that medieval humans had poorer color vision—are summarized in
Anderson, Folk-Taxonomies, 130–41, which incorporates a revised and enlarged version of
Anderson, “The Semantic Problem of ‘Red Gold’ in Germanic,” English Studies, 81 (2000),
1–13.

JEGP 115_2 text.indd 246 1/27/16 12:21 PM

 Categorization of Color   247

But for Old Norse, where the most numerous referent of rauðr is the
color of blood, I suggest that the single focal point of rauðr is in fact blood,
and that the consistency with which gold is referred to as rauðr in Old
Norse texts could instead reflect type modification, that is, the use of a
color term to classify rather than to describe colors.33 The use of color
terms as classificatory adjectives, outside of their expected semantic do-
mains, is not difficult to exemplify from modern languages. Red wine is
not often a color that is considered truly red in English, but is called red
to distinguish it from white wine, which similarly is not white. Similarly,
so-called “white,” “black,” and “red” people seldom come in hues as lively
as these adjectives might lead one to expect. We are acquainted with such
classificatory uses of color terms in modern languages, and it is likely that
Old Norse writers were accustomed to classificatory uses of color terms in
their language. Indeed, it is in the classification of economically or socially
important concepts and objects that type modification is most likely to be
seen, and type modification typically makes use of the most basic color
terms: namely, black, white, and red. For instance, Ralph Bolton records
that Peruvian potato farmers classify the colors of their potato cultivars
with the most basic color terms, most often black, white, and red, even
though their many types of potato are of numerous extremely various
colors.34
The consistent use of rauðr for gold and hvítr for silver may be a match-
ing scenario, contrasting two of the most precious metals by means of two
of the most basic color terms. Much as rauðr is consistently used for gold
in the texts excerpted for this study, silver likewise is almost always hvítr
in color (though occasionally grár, which like the use of bleikr for impure
gold, indicates silver of lower quality). In poetry there is one occurrence
each of the terms mjallhvítr and snæhvítr (both “snow-white”) used about
silver, but in both cases the compounding element participates in allit-
eration, so it is doubtful that snow and silver were actually regarded as
having very similar colors in normal descriptive terms. Rather, if silver is
classified as hvítr rather than described by it, then these words for snow
may communicate that the silver in question is extra pure rather than
literally extra white, in the same way that glóðrauðr and blóðrauðr appar-
ently communicate the purity of gold rather than the literal identity of its
color with that of embers or blood. There appear to be similar examples
of type modification in the color descriptions of precious metals in other
extinct Indo-European languages.35

33. Named and described in Biggam, Semantics of Colour, p. 210.

34. Ralph Bolton, “Black, White, and Red All Over: The Riddle of Color Term Salience,”
Ethnology, 17.3 (1978), 287–311.
35. Anderson, Folk-Taxonomies, pp. 91–92.

JEGP 115_2 text.indd 247 1/27/16 12:21 PM

 248  Crawford

If this is indeed a case of type modification, the phenomenon is probably

not limited to the color of precious metals in Old Norse; the use of blár and
svartr for human eyes in Old Norse is probably also type modification.36

OLD NORSE BLEIKR

In stark contrast to the rare and collocationally restricted gulr, Old Norse
bleikr is fairly common, the eighth-commonest color term in the language
(nearly as common as grǿnn, and much commoner than gulr) in the Old
Norse texts excerpted for this study.
Contrary to previous opinion, based on the discussion above, I suggest
that bleikr was the basic color term equivalent (or better: equivalently fo-
cused) to English yellow in Old Norse, despite the fact that both Modern
Icelandic and Modern Norwegian use descendants of gulr for this basic
color term. In the Old Norse texts excerpted, bleikr describes human hair,
human faces (when wounded, or in a state of fear or anger), arable fields,
birds’ feet, oxen and horses, the sun, the sea, dandelions, and corpses, all
of which are potentially yellow or pink in hue (see below for discussion
of how pink was probably categorized with yellow in Old Norse).

Table 1. Occurrences of color terms in the Old Norse texts studied.

frequency, relative to frequency of next term
word occurrences in texts (rounded to 2 decimal points after zero)
rauðr 441 2.44x commoner than:
svartr 181 1.14x commoner than:
hvítr 159 1.10x commoner than:
blár 145 1.45x commoner than:
grár 100 2.70x commoner than:
grǿnn 37 1.03x commoner than:
hárr 36 1.13x commoner than:
bleikr 32 1.19x commoner than:
brúnn 27 1.92x commoner than:
jarpr 14 2x commoner than:
gulr 7 1.17x commoner than:
gránn 6 1.5x commoner than:
rjóðr 4 1.33x commoner than:
hǫss 3 3x commoner than:
ámr 1
total 1213
Simple derivative verbs, such as causative bleikja, or inchoative roðna, are counted in the totals for the
adjectives from which they are derived since they provide unambiguous information on what referents
the adjectives might apply to. Occurrences of color (or color-like) adjectives as human and animal
names and bynames are also counted.

36. Crawford, “Blåe og svarte augo.”

JEGP 115_2 text.indd 248 1/27/16 12:21 PM

 Categorization of Color   249

The identification of bleikr rather than gulr as the basic color term fo-
cused on yellow is supported by the much greater preponderance of occur-
rences of bleikr than of gulr in Old Norse, as well as the much broader range
of referents whose color is described by bleikr than by gulr. It should be
borne in mind, however, that color labels can go in and out of a language
without leaving a trace, even as the category survives—as happened with
Old English hæwen ‘blue’, which has neither certain Germanic forebears
nor later English descendants, or in the western Romance languages with
the shift away from Latin albus to the Germanic borrowing *blankaz to
indicate “white” (Spanish blanco, French blanc, etc37). Consider also the
case of the Salishan languages, in which only six to nine color roots are
reconstructed for Proto-Salishan, but more than eighty appear in the
descendent Salishan languages.38
To exemplify the loss of color terms from within the history of West
Norse itself, it is notable that three other terms that are reasonably com-
mon in the description of human skin or hair color in Old Norse are
absent from Modern Norwegian and/or Icelandic: hárr, jarpr, and rjóðr.

A HISTORICAL PERSPECTIVE

While bleikr appears to be a basic color term in Old Norse, it also strikes
one as “less” basic or, more likely, newer to the basic color terms of Old
Norse than the other six basic color terms, blár, grár, grǿnn, hvítr, rauðr,
and svartr.39 Unlike the other six, it does not participate in any obvious
type-modification systems, is not as frequently contrasted with other basic
color terms, and there are convincing fossils of a time before the distinc-
tion of yellow from red—especially the use of rauða (literally “the red”)
for the yolk of an egg, whose name in English is indeed etymologically
“the yellow” and attested already in Old English as geolc. I believe that this
situation, and the use of bleikr vis-à-vis gulr in Old Norse, can be explained
by the following historical scenario.
Robert MacLaury40 notes that “widely dispersed languages encompass
a broad band of middle brightness color with a single category.” I postu-
late that this was the situation in Proto-Germanic, and that derivatives of

37. Anderson, Folk-Taxonomies, pp. 41–45

38. M. Dale Kinkade, “Proto-Salishan Colors,” in In Honor of Mary Haas, ed. William Shipley
(New York: Mouton de Gruyter, 1988), pp. 443–66.
39. The question of what color terms are basic in Old Norse is discussed in detail in Jack-
son Crawford, “The Historical Development of Basic Color Terms in Old Norse-Icelandic”
(PhD diss., Univ. of Wisconsin–Madison, 2014); as well as in Crawford, “Blåe og svarte augo.”
For an alternative view, see Wolf, “Some Comments on Old Norse-Icelandic Color Terms.”
40. Robert E. MacLaury, “From Brightness to Hue: An Explanatory Model of Color Cat-
egory Evolution [and Comments and Reply,” Current Anthropology, 33.2 (1992), 155.

JEGP 115_2 text.indd 249 1/27/16 12:21 PM

 250  Crawford

the Indo-European root *bhel- ‘bright, shining’ (from which blár/*blēwaz

and bleikr/*blaikaz are both derived) were used for this “middle bright-
ness color.” The Indo-European root *bhel- gives rise to an extraordinarily
wide range of color and color-like terms in the various Indo-European
languages, and so it is unlikely that its derivatives were associated with
particular hues at the earliest stage of Proto-Germanic. Before the Proto-
Germanic color classification system took its first steps toward encoding
a greater number of specific hues, derivatives of *bhel- probably referred
generally to items that were chromatic (i.e., not describable as black,
white, or gray) but not red.
The differentiation of blár/*blēwaz from bleikr/*blaikaz, and their emer-
gence as basic color terms probably proceeded later from the identifica-
tion of *blēwaz with the “cool” colors (blue and green) and *blaikaz with
the “warm” colors (other than highly salient red). Based on its salience
and semantic clarity41 in the Old Norse system, *blēwaz was probably the
first to become a basic color term but originally had focal points in both
blue and green—such a macrocolor is called grue and is quite common
in the world’s languages. Indeed such a hypothetical prehistoric system
for Proto-Germanic, Stage IIIG/Bu, is robustly attested in living languages42
and was probably similar to that of its related, neighboring Proto-Celtic
language, which, on the evidence of the Old Irish and Welsh color term
systems, probably incorporated blue, gray, and green under one color
term (derived, however, from the root *ghel-). From the semantic portfolio
of this early grue macrocolor *blēwaz, the uniquely Germanic neologism
*grōnjaz (green—a word formed from the same root as grow) subsequently
split off at such an early time as to have left no trace in the attested Ger-
manic languages of its former inclusion in a grue macrocolor; this was
Stage IVR/Y, a stage that likely reflects the situation in Proto-Germanic,
as the descendants of all six Proto-Germanic color terms postulated for
this stage (*blēwaz, *grēwaz, *grōnjaz, *hwītaz, *rauðaz, *swartaz) are basic
color terms in Old Norse and are almost certainly basic color terms in
Old English and Old High German (they are highly salient in these lan-
guages, and with the exception of the replacement of *swartaz with black
in English, their descendent terms continue to be the equivalent basic
color terms in their modern descendent languages).
In this earlier system, the term *blaikaz likely described MacLaury’s
“broad band of middle brightness color” (or better: broad band of middle-
reflective color), not so reflective as white, not achromatic as gray, and
not as saturated as red, but middle-reflective, chromatic, and warm (i.e.,

41. The notion that blár is an ambiguous color term that may encode both blue and black
is refuted in Crawford, “Historical Development,” esp. pp. 44–47.
42. For extremely numerous analogues, see Paul Kay, Brent Berlin, Luisa Maffi, William
R. Merrifield, and Richard Cook, The World Color Survey (Stanford: CSLI Publications, 2009).

JEGP 115_2 text.indd 250 1/27/16 12:21 PM

 Categorization of Color   251

usable for oranges, yellows, and browns, but not for blues, greens, and
violets). Such a middle-reflective, chromatic category will, according to
MacLaury, most likely become a hue category after this transitional phase
has passed, and in such a case is especially likely to be tied to the hues
yellow and/or green.43 This, I suggest, is what happened in Old Norse,
with bleikr emerging as a basic color term focused near English yellow but
also still capable of being used (as it was in the earlier system) for any
nonred warm colors that were contrasted with red, especially the pale
pinkish hue of a frightened human face. The use of bleikr to indicate the
color of impure gold (as opposed to pure, rauðr gold) is probably another
reflection of the earlier system’s organization, when the relationship of
bleikr to rauðr was such that bleikr might indicate any nonoptimal red (i.e.,
any other warm colors), while rauðr remained the only warm basic color
term.
While it is typologically uncommon for the dissolution of blue/green
to precede the dissolution of red/yellow,44 a parallel may be seen in the
emergence of most of the basic color terms that appear latest in the his-
torical development of most languages—brown, orange, pink, and often
purple—from division of the semantic range of red. Red is “the unmarked
pole of the hue dimension”45 and therefore something of a “default” for
color descriptions in the warm parts of the spectrum. It has also been
noted that all shades of red perceivable by the human eye contain some
yellow,46 and that the two colors border one another on the spectrum of
visible light. Based on these considerations, it is unsurprising that the
two colors might be grouped together under one term. The very fact that
Old Norse uses a different adjective than West Germanic *gelwaz to form
its basic color term for yellow, while all other basic color terms in Old
Norse are cognate to highly salient color terms in the early West Germanic
languages, also suggests that North Germanic and West Germanic both
codified yellow as a basic color term separately, at a late stage following
the division of these two branches. It is also possible that an earlier basic
color term for yellow, perhaps cognate with West Germanic *gelwaz, was
lost in the history of North Germanic, though the loss of a color category
is rare and should not be presumed without direct evidence.
Moving forward from the color-classification system of classical liter-
ary Old Norse (such as that of the texts excerpted for this study), it may
be observed that bleikr loses much of its semantic portfolio over time,

43. MacLaury, “From Brightness to Hue,” p. 159.

44. Kay et al., World Color Survey, pp. 31–41.
45. Stanlely R. Witkowski and Cecil H. Brown, “An Explanation of Color Nomenclature
Universals,” American Anthropologist, 79.1 (1977), 53.
46. Robert E. MacLaury, “Color-Category Evolution and Shuswap Yellow-with-Green,”
American Anthropologist, 89.1 (1987), 113.

JEGP 115_2 text.indd 251 1/27/16 12:21 PM

 252  Crawford

becoming restricted to nonoptimal warm colors, while its former cen-

tral point of “yellow” is usurped in the descendent languages (possibly
already in postclassical Old Norse47) by the less semantically ambiguous
gulr, which in classical Old Norse had been collocationally restricted to
hair color. Yet for classical literary Old Norse, it appears that this was not
yet the case and that gulr was a specialized term for a hair color (possibly
blonde or reddish), while bleikr was a basic color term focused on yellow
but usable also for other nonred warm colors such as pink.

47. As suggested by its greatly increased frequency in late Riddara sǫgur. See Wolf, “Towards
a Diachronic Analysis,” pp. 117–20.

JEGP 115_2 text.indd 252 1/27/16 12:21 PM