1. To learn or not to learn: The growing paths of children’s phonological neighborhoods Yao Yao @berkeley.edu 2009-1-11 1YY @ LSA Annual Meeting 2009

2. Overview Background Data Study 1 Study 2 Study 3 Discussion 2YY @ LSA Annual Meeting 2009

3. Phonological neighborhood (PN) General idea – The lexicon can be viewed as a network of words, in which similar-sounding words are connected to form phonological neighborhoods. – Defining neighbors One-phoneme difference rule (Luce & Pisoni, 1998). 3YY @ LSA Annual Meeting 2009 cat kit mat cap map cape mop it

4. Evidence for PN Speech perception – Inhibitory effect (Luce & Pisoni, 1998) Speech production – Facilitative effect (Vitevitch, 2002) – Hyperarticulation (Munson & Solomon, 2004) 4 YY @ LSA Annual Meeting 2009

5. Children’s PN development Start small and sparse (Charles-Luce & Luce, 1990; Logan, 1992; Storkel, 2002) How do children’s phonological neighborhoods grow? – From the target language (TL) perspective Early acquired words are short in length, high in frequency and from dense neighborhoods in TL (Storkel, 2004) – From the local child language (CL) perspective Acquire similar-sounding words? Acquire dissimilar-sounding words? 5YY @ LSA Annual Meeting 2009

6. Puzzle Perceptual abilities – Infants at a very young age can perceive fine phonetic detail (Aslin, Jusczyk & Pisoni, 1998) Used in word learning ? – Children have difficulty in learning similar- sounding novel words (Stager & Werker, 1997) – Children are sensitive to the phonetic detail in the input. (Zamuner, 2006, 2009; Coady & Aslin, 2003) YY @ LSA Annual Meeting 20096

7. Two models LRM (Lexical Restructuring Model; Metsala & Walley, 1998) – Holistic lexical representations initially. More detailed phonemic representation is necessitated as the vocabulary grows. PRIMIR (Processing Rich Information from Multidimensional Interaction Representations; Werker & Curtin, 2005) – “phonetic detail is incorporated into early lexical representations” (Zamuner, 2009:7) YY @ LSA Annual Meeting 20097

8. Examining PN development If holistic representation… – Acquire dissimilar-sounding words – Avoid dense neighborhoods If detailed phonetic representation… – Acquire similar-sounding words – Form dense neighborhoods Influence from the TL YY @ LSA Annual Meeting 20098

9. Database From the Manchester corpus in CHILDES database (Theakston, et al., 2001; MacWhinney, 1991) – Two monolingual British children Joel [1;11- 2;10] “word learner“ (Ke & Yao, 2008) Ruth [2;0 - 2;11] “grammar learner“ (Ke & Yao, 2008) – 1-year longitudinal study – 34 recordings, 1hr each, approx. 1 per 10 days 9YY @ LSA Annual Meeting 2009

10. Stage division (Ke & Yao, 2008) – 0.5 increase in MLU (Mean Length of Utterance) per stage – 5 recording sessions – 40 days – Non-overlapping – Joel S1, S2, S3, S4a, S4b, S5 – Ruth S1, S2, S3a, S3b, S4, S5 10YY @ LSA Annual Meeting 2009

11. Network model Input –Spontaneous speech of the child –Phonetic transcription is obtained from the CELEX database. –No regular inflected forms or contracted forms if the base forms already exist Neighborhood definition –One-phoneme difference rule YY @ LSA Annual Meeting 200911

12. Preliminary counts 12YY @ LSA Annual Meeting 2009

13. Study 1: size and average density of 3- phoneme words in children’s networks Size 13YY @ LSA Annual Meeting 2009 Average density

14. Two children’s data overlayed (with Joel’s shifted by 2 stages) 14YY @ LSA Annual Meeting 2009 Joel is probably more advanced than Ruth by 2 stages. Study 1: size and average density of 3- phoneme words in children’s networks

15. Study 1: size and average density of 3- phoneme words in the local network Q: is the increase in density an artifact of the increase in lexicon size? Coady & Aslin (2003) – Calculate neighborhood density as a proportion of the entire lexicon, not in raw counts of neighbors. YY @ LSA Annual Meeting 200915

16. Study 1: size and average density of 3- phoneme words in the local network Neighborhood density relative to vocabulary size 16YY @ LSA Annual Meeting 2009

17. Study 1: Discussion Absolute number of neighbors increase over time When lexicon size is normalized, neighborhood density – Increases in Ruth’s early stages – Slightly decreases in Ruth’s late stages and all Joel’s stages 17YY @ LSA Annual Meeting 2009

18. Study 2: neighborhood density of the same 3-phoneme words in TL – Neighborhood densities in CL and TL are partially correlated (corr <0.4) 18YY @ LSA Annual Meeting 2009

19. Study 2: Discussion Average neighborhood density (in adult lexicon) – Increases in Ruth’s early stages – decreases in Ruth’s late stages and all Joel’s stages Joel’s development is probably two stages more advanced than Ruth 19YY @ LSA Annual Meeting 2009

20. Study 3 Assumptions – Word learning as a dynamic process Acquire words Lose words – Stage networks represent the lexicon of the child at that stage Q: What words are acquired, lost, and kept? Hi-density words or low-density words? YY @ LSA Annual Meeting 200920

21. Study 3: What words are acquired, lost and kept? New vs. old words – New words: words that don’t exist in the previous lexicon – Old words: words that already exist in the previous lexicon Lost vs. retained words – Lost words: words that don’t exist in the next lexicon – Retained words: words that still exist in the next lexicon Old/ Retained New Lost S1 lexicon S2 lexicon 21YY @ LSA Annual Meeting 2009

22. Study 3: What words are acquired, lost and kept? Average local density of 3-phoneme new vs. old words 22YY @ LSA Annual Meeting 2009

23. Study 3: What words are acquired, lost and kept? Average local density of 3-phoneme lost vs. retained words 23YY @ LSA Annual Meeting 2009

24. Study 3: What words are acquired, lost and kept? Before Avg. density = 2 24YY @ LSA Annual Meeting 2009 After Avg. density = (4+6)*2/7=2.85

25. Study 3: What words are acquired, lost and kept? Average density of 3-phoneme lost, new and old/retained words in the target language 25YY @ LSA Annual Meeting 2009

26. Study 3: Discussion In Ruth’s early stages – Newly-acquired words are structurally MORE important than existing words – Structurally-important words are MORE susceptible to being lost In Ruth’s late stages and all Joel’s stages – Newly-acquired words are structurally LESS important than existing words – Structurally-important words are LESS susceptible to being lost 26YY @ LSA Annual Meeting 2009

27. General Discussion In early stages of lexical acquisition – acquire words that sound similar to existing ones and increase local density – build up the backbone of the phonological network by adding important nodes – important words are also easier to lose After that – words that are acquired are less important – important words are less likely to be lost 27YY @ LSA Annual Meeting 2009

28. Implications Pressure to form/deform denser neighborhoods coexist in early stages Perceptual abilities – Probably used from the early stage of word learning PRIMIR model – Stabilize after a while YY @ LSA Annual Meeting 200928

29. Implications Critical mass hypothesis – after the lexicon exceeds a critical size, qualitative changes in linguistic performance and/or acquisition strategy will take place (Marchman & Bates, 1994) Critical size – ~200 words – cf. Vogel Sosa & Stoel-Gammon (2006) 29YY @ LSA Annual Meeting 2009

30. Complications Token frequency – Neighborhood density in CL is partially correlated with token frequency in child speech (cor < 0.45) and maternal input (cor <0.45). YY @ LSA Annual Meeting 200930

31. Caveats Sampling frequency – 5hrs recording spread over ~40 days – Is the stage lexicon a decent representation of the child’s productive lexicon? – Is the division of new/old, lost/retained words fair enough? Phonological network model – The use of dictionary pronunciation of words – Definition of neighbors – Different types of neighbors YY @ LSA Annual Meeting 200931

32. Acknowledgement Child subjects The Manchester Corpus CHILDES database Prof. Susanne Gahl Prof. Keith Johnson Audience at the Berkeley Phonology Phorum YY @ LSA Annual Meeting 200932

33. Selected references Aslin, R.N., Jusczyk, P.W. & Pisoni, D.B. (1998). Speech and auditory processing during infancy: constraints on and precursors to language. In D. Kuhn & R. Siegler (eds), Handbook of child psychology: cognition, perception and language, vol. 2, NY: Wiley. Charles-Luce, J. & Luce, P.A. (1990). Similarity neighborhoods of words in young children’s lexicons. Journal of Child Language 17, 205-215. Coady, J.A. & Aslin, R.N. (2003). Phonological neighborhoods in the developing lexicon. Journal of Child Language 30, 441-469. Logan, J.S. (1992). A computational analysis of young children’s lexicons (Tech. Rep. No. 8). Bloomington, IN: Speech Research Laboratory, Dept. of Psychology, Indiana Univ. Luce, P.A. & Pisoni, D.B. (1998). Recognizing spoken words: the neighborhood density activation model. Ear and Hearing 19, 1-36. MacWhinney, B. (1991). The Childes Project: Tools for Analyzing Talk. Lawrence Erlbaum Associates. Marchman, V.A. & Bates, E. (1994) Continuity in lexical and morphological development: a test of the critical-mass hypothesis. Journal of Child Language 21, 339-366. Metsala, J. L. & Walley, A. C. (1998). Spoken vocabulary growth and the segmental restructuring of lexical representations : Precursors to phonemic awareness and early reading ability. In J. L. Metsala & L. C. Ehri (eds), Word recognition in beginning literacy, 89–120. New York: Erlbaum. YY @ LSA Annual Meeting 200933

34. Selected references Munson, B. & Solomon, N.P. (2004). The effect of phonological neighborhood density on vowel articulation. Journal of Speech, Language, and Hearing Research 47(5), 1048-1058 Stager, C.L. & Werker, J.F. (1997). Infants listen for more phonetic detail in speech perception than in word-learning tasks. Nature 388, 381-382. Storkel, H.L. (2002). Restructuring of similarity neighbourhoods in the developing mental lexicon. Journal of Child Language 29, 251-274. Storkel, H.L. (2004). Do children acquire dense neighborhoods? An investigation of similarity neighborhoods in lexicla acquisition. Applied Psycholinguistics 25, 201-221. Theakston, A. L., Lieven, E. V. M., Pine, J. M., & Rowland, C. F. (2001). The role of performance limitations in the acquisition of verb-argument structure: an alternative account. Journal of Child Language 28, 127-152. Vitevitch,. (2002). The influence of phonological similarity neighborhoods on speech production. Journal of Experimental Psychology: Learning, Memory, and Cognition 28(4), 735–747. Vogel Sosa, A. & Stoel-Gammon, C. (2006). Patterns of intra-word phonological variability during the second year of life. Journal of Child Language 33, 31–50. Werker, J. F. & Curtin, S. (2005). PRIMIR: A developmental framework of infant speech processing. Language Learning and Development 1, 197–234. Zamuner, T. S. (2006). Sensitivity to word-final phonotactics in 9- to 16-month-old infants. Infancy 10, 77–95. YY @ LSA Annual Meeting 200934