Elamur Research⌕
☰
home / papers / 2015 / Perceptual, conceptual, and frequency effects on error patterns in English color term acquisition

Perceptual, conceptual, and frequency effects on error patterns in English color term acquisition

Paper · 2015 · Barend Beekhuizen and Suzanne Stevenson

Summary

The paper asks why young English speakers make the specific color-naming mistakes they do, and which of three candidate influences best accounts for them: the perceptual geometry of color, how often languages around the world lump colors together (typological prevalence, used as a stand-in for cognitive naturalness), and how frequent each color word is. The authors simulate learning with word-learning models trained on generated pairs of a Munsell chip and an English basic color term. They vary how chips are represented: CIELab coordinates, a "conceptual" space derived by PCA from World Color Survey naming data, or both. They also vary how often each term appears in the input. Model predictions are compared with Bateman's (1915) elicitation data using a quantitative deviation score. Perceptual features alone reproduce all five observed error phenomena. Adding typological features or frequency brings no improvement, so frequency is effectively ruled out. The typological components turn out to track the perceptual dimensions closely. This matters for testing the Typological Prevalence Hypothesis in a domain where the perceptual structure is well understood.

Key points

  • Target data come from Bateman (1915). It covers 591 English-speaking children aged 6–11 who named eight "best example" chips. The responses use all eleven basic color terms.
  • Five phenomena had to be explained:
    • BLACK, WHITE, BLUE and RED are nearly error-free.
    • GREEN and YELLOW show early errors only, which vanish after age 6.
    • ORANGE has scattered errors that persist until 11.
    • PURPLE has persistent errors, mostly blue, persisting until 9.
    • The reverse error, calling BLUE "purple", does not occur.
  • Kay et al. (2009) supplied the naming data behind the conceptual space through their World Color Survey, which spans 330 Munsell chips, 110 languages, and roughly 24 speakers per language. It is built in three steps:
    • Compute a chip-by-chip distance matrix for each language from its naming counts.
    • Sum these matrices across languages.
    • Reduce with PCA, keeping the 4 components whose eigenvalue exceeds 1.
  • Each of the four PCA components correlates significantly with a CIELab dimension (L*, a*, b*). The typological signal therefore adds nothing beyond perception, and a conceptual-only model actually fits worse.
  • One limitation is that the model is "too good": it predicts no errors for 6 of the 8 stimuli, although children erred slightly on 4 of them.

Interesting ideas

  • I1. Meaning space from multilingual labeling: each language's naming of the same stimuli is turned into a distance matrix, the matrices are summed across languages, and PCA is run on the result. This yields a semantic space without anyone hand-designing features.
  • I2. Asymmetric errors as diagnostics: the fact that blue spreads to PURPLE but purple never spreads to BLUE shows that symmetric similarity alone cannot be the explanation.
  • I3. Frequency as a manipulable null hypothesis: term frequency is tested by changing how input items are generated, which makes it a baseline the data must reject.
  • I4. Redundancy check before claiming added value: the authors correlate the derived components with known dimensions to explain why a new feature set brings no gain.
  • I5. Scoring against observed error distributions: the model is judged by a numeric deviation from observed response patterns, rather than by eyeballing developmental trends.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (world): word senses or frame situations could be embedded the same way by how often translations in many languages share one label. This would give the ontology an explainable, data-derived similarity measure alongside WordNet relations.
  • I2 (learn): parser and tagger confusion matrices could be checked for direction, not just counts, to tell representation gaps apart from simple label closeness.
  • I3 (learn): as a negative control, retrain or evaluate with label frequencies flattened. This shows whether a learned behavior reflects real structure or merely corpus skew.
  • I4 (world): before adding a compiled knowledge source (e.g., Wikidata-derived features), its overlap with existing features could be measured so that redundant layers are not mistaken for useless ones.
  • I5 (answer): student readers could be scored on whether their wrong answers on school tests resemble typical human error profiles, not only on accuracy.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓University of TorontoToronto · Canada110 works · 179,061 citations🎓Leiden UniversityLeiden · Netherlands37 works · 1,408 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Arts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksPsychology · 1,649 worksExperimental and Cognitive Psychology · 930 worksDevelopmental and Educational Psychology · 523 works

🧭 Topics 1Research subjects and languages the work is about.English · 2,379 works

Cite

Cite as Barend Beekhuizen and Suzanne Stevenson. 2015. Perceptual, conceptual, and frequency effects on error patterns in English color term acquisition. Proceedings of the Sixth Workshop on Cognitive Aspects of Computational Language Learning.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/w15-2412

@inproceedings{beekhuizen2015perceptual,
  title = {{Perceptual, conceptual, and frequency effects on error patterns in English color term acquisition}},
  author = {Barend Beekhuizen and Suzanne Stevenson},
  booktitle = {Proceedings of the Sixth Workshop on Cognitive Aspects of Computational Language Learning},
  year = {2015},
  doi = {10.18653/v1/w15-2412},
  url = {https://doi.org/10.18653/v1/w15-2412}
}

References

27 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Elsa Jaffe Bartlett. 1978. The acquisition of the meaning of colour ters: a study of lexcal development. pages 89–108. find
  2. W. G. Bateman. 1915. The Naming of Colors by Children the Binet Test. The Pedagogical Seminary, 22(4):469–486, December. ↗
  3. Barend Beekhuizen, Afsaneh Fazly, and Suzanne Stevenson. 2014. Learning Meaning without Primitives: Typology Predicts Developmental Patterns. In Proceedings of the 36th annual meeting of the Cognitive Science Society. find
  4. Brent Berlin and Paul Kay. 1969. Basic color terms: Their universality and evolution. University of California Press, Berkeley, CA. find
  5. Melissa Bowerman. 1993. Typological perspectives on language acquisition: Do crosslinguistic patterns predict development? In Eve V. Clark, editor, Proceedings of the Twenty-fifth Annual Child Language Research Forum, pages 7–15, Stanford, CA. CSLI Publications. find
  6. Eve V. Clark. 1973. What’s in a word? On the child’s acquisition of semantics in his first language. In Timothy E. Moore, editor, Cognitive Development and the Acquisition of Language, pages 65– 110. New York: Academic Press. find
  7. Ian R. L. Davies, Greville Corbett, Harry McGurk, and David Jerrett. 1994. A developmental study of he acquisition of colour terms in Setswana. Journal of Child Language, 21:693–712. find
  8. Ian R. L. Davies, Greville Corbett, Harry McGurk, and Catriona MacDermid. 1998. A developmental study of the acquisition of Russian colous terms. Journal of Child Language, 25:395–417. find
  9. Afsaneh Fazly, Afra Alishahi, and Suzanne Stevenson. 2010. A probabilistic computational model of cross-situational word learning. Cognitive Science, 34(6):1017–1063. ↗
  10. Dedre Gentner and Melissa Bowerman. 2009. Why some spatial semantic categories are harder to learn than others. The Typological Prevalence Hypothesis. In J. Guo, E. Lieven, N. Budwig, S. Ervin-Tripp, K. Nakamura, and S. Ozcaliskan, editors, Crosslinguistic approaches to the psychology of language. Research in the tradition of Dan Isaac Slobin, chapter 34, pages 465–480. Psychology Press, New York, NY. find
  11. Dedre Gentner. 1982. Why Nouns are Learned Before Verbs : Linguistic Relativity versus Natural Partitioning. In Stan Kuczaj, editor, Language Development. Volume 2: Language, Thought, and Culture, volume 2, chapter 11, pages 301–334. Lawrence Erlbaum Associates, Hillsale, New Jersey. find
  12. Judith C Goodman, Philip S Dale, and Ping Li. 2008. Does frequency count? Parental input and the acquisition of vocabulary. Journal of child language, 35(3):515–31, August. find
  13. Insa Gülzow and Natalia Gagarina, editors. 2007. Frequency Effects in Language Acquisition. Defining the Limits of Frequency as an Explanatory Concept. De Gruyter Mouton, Berlin. find
  14. Sara Harkness. 1973. Universal Aspects of Learning Color Codes: A Study in Two Cultures. Ethos, pages 175–200. ↗
  15. H. Hotelling. 1933. Analysis of a complex of statistical variables into principal components. Journal of Educational Psychology, 24:417–441,498–520. ↗
  16. Z.M. Istomina. 1960. Perception and naming of color in early childhood. Izvestiia Akademii Pedagogich-eskikh, 113:37–45. find
  17. Paul Kay, Brent Berlin, Luisa Maffi, William R. Merrifield, and Richard Cook. 2009. World Color Survey. CSLI Publications, Stanford, CA. find
  18. Stephen C. Levinson, Sergio Meira, The Language Group, and Cognition. 2003. ’Natural Concepts’ in the Spatial Topological Domain – Adpositional Meanings in Crosslinguistic Perspective: An Exercise in Semantic Typology. Language, 79(3):485– 516. ↗
  19. Asifa Majid, Fiona Jordan, and Michael Dunn. 2015. Semantic systems in closely related languages. Language Sciences, 49:1–18. ↗
  20. Aida Nematzadeh, Afsaneh Fazly, and Suzanne Stevenson. 2012. A computational model of memory, attention, and word learning. In Proceedings of the Third Workshop on Cognitive Modeling and Computational Linguistics. find
  21. Robert M Nosofsky. 1987. Attention and Learning Processes in the Identification and Categorization of Integral Stimuli. Journal of Experimental Psychology, 13(1):87–108. ↗
  22. Nicola J. Pitchford and Kathy J. Mullen. 2003. The development of conceptual colour categories in preschool children: Influence of perceptual organization. Visual Cognition, 10(1):51–57. find
  23. Terry Regier, Paul Kay, and Naveen Khetarpal. 2007. Color naming reflects optimal partitions of color space. PNAS, 104:1436–1441. ↗
  24. Debi Roberson, Jules Davidoff, Ian R L Davies, and Laura R Shapiro. 2004. The development of color categories in two languages: a longitudinal study. Journal of experimental psychology. General, 133(4):554–71, December. ↗
  25. Nancy N. Soja. 1994. Young Children’s Concept of Color and Its Relation to the Acquisition of Color Words. Child Development, 65:918–937. ↗
  26. Anna L. Theakston, Elena V.M. Lieven, Julian M. Pine, and Caroline M. Rowland. 2001. The role of performance limitations in the acquisition of verb-argument structure: An alternative account. Journal of Child Languages, pages 127–152. ↗
  27. Daniel Yurovsky, Katie Wagner, David Barner, and Michael C. Frank. 2015. Signatures of Domain-General Categorization Mechanisms in Color Word Learning. In Proceedings CogSci. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/beekhuizen-2015-perceptual-conceptual-frequency-effects-error/