Nicole Fider, Louis Narens, Kimberly A. Jameson, and Natalia L. Komarova, "Quantitative approach for defining basic color terms and color category best exemplars," J. Opt. Soc. Am. A 34, 1285-1300 (2017)
A new method is presented that identifies basic color terms (BCTs) from
color-naming data. A function is defined that measures how well a term
is understood by a communicating population. BCTs are then separated
from other color terms by a threshold value applied to this function.
A new mathematical algorithm is proposed and analyzed for determining
the best exemplar associated with each BCT. Using data provided by the
World Color Survey, comparisons are made between the paper’s
methods and those from other studies. These comparisons show that the
paper’s new definition of “basicness” mostly
agrees with the typical definition found in the color categorization
literature, which was originally due to Kay and colleagues. The new
definition, unlike the typical one, has the advantage of not relying
on syntactic or semantic features of languages or color lexicons. This
permits the methodology developed to be generalizable and applied to
other category domains for which a construct of
“basicness” could have an important role.
You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Figure files are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Article tables are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Equations are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
The number of WCS color terms classified as BCTs,
“potentially basic” color terms (PBCTs), or
nonbasic color terms (NBCTs) according to the compared
B&K- and CS-definitions.
Table 2.
Brief Comparison of B&K-BCTs and CS-BCTs on
Eight Languages (Emphasized Previously in [36,42]) from the
WCSa
B&K
CS
Result
16: Buglere (Panama/Costa Rica)
6
6
20: Candoshi (Peru)
7
7
51: Kalam (Peru)
6
6
56: Konkomba (Ghana/Togo)
4
4
60: Kwerba (Indonesia)
4?
4
64: Martu Wangka (Brazil)
5?
5
74: Mura Piraha (Australia)
4
4
87: Siriono (Bolivia)
5
5
For each language, the number of B&K-BCTs is
listed, as well as the number of CS-BCTs. A mark
of () indicates that
our method identified terms that are
not B&K-basic; a mark of
() indicates that
our method did not identify
B&K-basic
terms as CS-basic. A question mark is used to
indicate B&K color terms that were
classified as “potentially
BCTs.”
Table 3.
Detailed Comparison of B&K-BCTs and CS-BCTs on
Eight Languages from the WCSa
Language 16: Buglere
B&K
CS
1. jutre
✓
✓
0.623
2. jere
✓
✓
0.537
3. dabe
✓
✓
0.528
4. moloin
✓
✓
0.532
5. lere
✓
✓
0.644
6. leren
✓
✓
0.383
Language 20: Candoshi
B&K
CS
1. borshi
✓
✓
0.764
2. chobiapi
✓
✓
0.725
4. kamachpa
✓
✓
0.404
5. kantsiripi
✓
✓
0.730
6. kavabana
✓
✓
0.723
12. pozani
×
✓
0.200
13. ptsiyaro
✓
✓
0.730
Language 51: Kalam
B&K
CS
1. muk
✓
✓
0.472
2. minj-kimemb
✓
✓
0.448
3. likan
✓
✓
0.584
4. tund
✓
✓
0.732
5. mosimb
✓
✓
0.177
9. walin
✓
✓
0.579
Language 56: Konkomba
B&K
CS
1. pipin
✓
✓
0.712
2. bombon
✓
✓
0.638
3. maman
✓
✓
0.732
4. yaankal
✓
✓
0.245
Language 60: Kwerba
B&K
CS
1. asiram
✓
✓
0.623
6. icem
✓
✓
0.618
11. kainanesenum
?
✓
0.182
17. nokonum
✓
✓
0.785
Language 64: Martu Wangka
B&K
CS
10. karntawarra
?
✓
0.189
25. maru-maru
✓
✓
0.633
26. miji-miji
✓
✓
0.621
38. piila-piila
✓
✓
0.317
48. yukuri-yukuri
✓
✓
0.574
Language 74: Mura Piraha
B&K
CS
1. bii sai
✓
✓
0.829
2. biopaiai
✓
✓
0.656
3. ahoasaaga
✓
✓
0.723
4. kobiai
✓
✓
0.768
Language 87: Siriono
B&K
CS
2. echo
✓
✓
0.521
4. eirei
✓
✓
0.748
7. erondei
✓
✓
0.530
8. eruba
✓
✓
0.495
9. eshi
✓
✓
0.721
For each language, the corresponding table shows
which color terms are B&K-basic and which
color terms are CS-basic; checkmarks denote
basicness, crosses denote nonbasicness, and
question marks denote ambiguity
(“potentially basic” terms). The
strengths of each color term are also provided,
rounded to the nearest thousandth. Color terms
which do not satisfy either definition and also
have strength less than 0.16 have been omitted for
brevity. The numbering assigned by the WCS Data
Archives is used for the languages and color
terms.
Table 4.
“CS-Basic” and CS-“Potentially
Basic” Color Terms That Do Not Have WCS
Focus-Choice Dataa
Language
Color Term
Strength of Color Term
B&K-Basic
3
15. pensaal
0.8161
yes
53
1. ikura
0.1993
no
53
2. iura
0.7741
yes
53
3. ilyby
0.8565
no
66
4. canga/cangu
0.7446
yes
78
7. istak
0.7799
yes
80
3. koomagi
0.2325
yes
80
10. wigium
0.2636
yes
97
5. matak
0.1993
yes
Languages and color terms are numbered according to the
WCS Data Archive.
Table 5.
Comparison of Four Algorithms for Identifying Category Best
Exemplars
Indicates uncommon events (0.75% of the cases) where
some computed data yielded centers of modal maps that
were not observed to correspond to the categorical
color terms typically used to name color space
regions.
The number of WCS color terms classified as BCTs,
“potentially basic” color terms (PBCTs), or
nonbasic color terms (NBCTs) according to the compared
B&K- and CS-definitions.
Table 2.
Brief Comparison of B&K-BCTs and CS-BCTs on
Eight Languages (Emphasized Previously in [36,42]) from the
WCSa
B&K
CS
Result
16: Buglere (Panama/Costa Rica)
6
6
20: Candoshi (Peru)
7
7
51: Kalam (Peru)
6
6
56: Konkomba (Ghana/Togo)
4
4
60: Kwerba (Indonesia)
4?
4
64: Martu Wangka (Brazil)
5?
5
74: Mura Piraha (Australia)
4
4
87: Siriono (Bolivia)
5
5
For each language, the number of B&K-BCTs is
listed, as well as the number of CS-BCTs. A mark
of () indicates that
our method identified terms that are
not B&K-basic; a mark of
() indicates that
our method did not identify
B&K-basic
terms as CS-basic. A question mark is used to
indicate B&K color terms that were
classified as “potentially
BCTs.”
Table 3.
Detailed Comparison of B&K-BCTs and CS-BCTs on
Eight Languages from the WCSa
Language 16: Buglere
B&K
CS
1. jutre
✓
✓
0.623
2. jere
✓
✓
0.537
3. dabe
✓
✓
0.528
4. moloin
✓
✓
0.532
5. lere
✓
✓
0.644
6. leren
✓
✓
0.383
Language 20: Candoshi
B&K
CS
1. borshi
✓
✓
0.764
2. chobiapi
✓
✓
0.725
4. kamachpa
✓
✓
0.404
5. kantsiripi
✓
✓
0.730
6. kavabana
✓
✓
0.723
12. pozani
×
✓
0.200
13. ptsiyaro
✓
✓
0.730
Language 51: Kalam
B&K
CS
1. muk
✓
✓
0.472
2. minj-kimemb
✓
✓
0.448
3. likan
✓
✓
0.584
4. tund
✓
✓
0.732
5. mosimb
✓
✓
0.177
9. walin
✓
✓
0.579
Language 56: Konkomba
B&K
CS
1. pipin
✓
✓
0.712
2. bombon
✓
✓
0.638
3. maman
✓
✓
0.732
4. yaankal
✓
✓
0.245
Language 60: Kwerba
B&K
CS
1. asiram
✓
✓
0.623
6. icem
✓
✓
0.618
11. kainanesenum
?
✓
0.182
17. nokonum
✓
✓
0.785
Language 64: Martu Wangka
B&K
CS
10. karntawarra
?
✓
0.189
25. maru-maru
✓
✓
0.633
26. miji-miji
✓
✓
0.621
38. piila-piila
✓
✓
0.317
48. yukuri-yukuri
✓
✓
0.574
Language 74: Mura Piraha
B&K
CS
1. bii sai
✓
✓
0.829
2. biopaiai
✓
✓
0.656
3. ahoasaaga
✓
✓
0.723
4. kobiai
✓
✓
0.768
Language 87: Siriono
B&K
CS
2. echo
✓
✓
0.521
4. eirei
✓
✓
0.748
7. erondei
✓
✓
0.530
8. eruba
✓
✓
0.495
9. eshi
✓
✓
0.721
For each language, the corresponding table shows
which color terms are B&K-basic and which
color terms are CS-basic; checkmarks denote
basicness, crosses denote nonbasicness, and
question marks denote ambiguity
(“potentially basic” terms). The
strengths of each color term are also provided,
rounded to the nearest thousandth. Color terms
which do not satisfy either definition and also
have strength less than 0.16 have been omitted for
brevity. The numbering assigned by the WCS Data
Archives is used for the languages and color
terms.
Table 4.
“CS-Basic” and CS-“Potentially
Basic” Color Terms That Do Not Have WCS
Focus-Choice Dataa
Language
Color Term
Strength of Color Term
B&K-Basic
3
15. pensaal
0.8161
yes
53
1. ikura
0.1993
no
53
2. iura
0.7741
yes
53
3. ilyby
0.8565
no
66
4. canga/cangu
0.7446
yes
78
7. istak
0.7799
yes
80
3. koomagi
0.2325
yes
80
10. wigium
0.2636
yes
97
5. matak
0.1993
yes
Languages and color terms are numbered according to the
WCS Data Archive.
Table 5.
Comparison of Four Algorithms for Identifying Category Best
Exemplars
Indicates uncommon events (0.75% of the cases) where
some computed data yielded centers of modal maps that
were not observed to correspond to the categorical
color terms typically used to name color space
regions.