Detail of contribution
Auteur: Jongsup JUN
Titre:
Multi-Variate Interaction in Children's Acquisition of English Derivational Morphology
Abstract/Résumé: In this study, we propose a regression model that entertains both linguistic and usage-based factors to explain English-speaking children’s acquisition of derivational morphology. We analyzed children’s production of derived words in a million-word corpus taken from the CHILDES (MacWhinney & Snow, 1990) database. The data came from the voluntary production of derived words by 469 individuals at age 3-10. We used Baayen’s (1993) hapax-conditioned degree of productivity (=PROD) as a function of both linguistic and usage-based factors in the analysis. The linguistic factors include the POS of the root and its derivative, semantic transparency (=S), and phonological transparency (=PH). Besides, we used four distinct input frequency measures as usage-based factors: How often the root of each derived word occurs in caregivers’ speech (=M1); How often each derived word occurs in caregivers’ speech (=M2); How many types of derived word each root generates in caregivers’ speech (=M3); and How many types of derived word each affix generates in caregivers’ speech (=M4). We conducted hierarchical multiple regression analyses using the natural log of PROD as a dependent variable and POS, S, PH, M1-4 and age (=A) as independent factors. Among these factors, POS, M1, and M2 were not significant, and were eliminated from the model. We obtained (1) as the final regression equation (R^2=.349). (1) ln(PROD) = -7.784+0.084*A-0.574*S+0.36*PH-0.373*M3+0.015*M4 The results show that the family frequencies of the root and the affix play the most important role in the model, and that semantic and phonological transparencies as well as age show limited effects on the total variation of ln(PROD). In particular, a single principle in (1) explains children’s performance on derivational morphology for all age groups. We can view the regression equation in (1) as a grammatical generalization that constrains the productivity of a morphological process. For instance, children produce 'dusty' more easily than 'electricity' at age 3-4 because 'electricity' has lower M3 and M4 values than 'dusty'. We propose that the acquisition process be the expansion of the (functional) domain from high-frequency words like 'dusty' to low-frequency words like 'electricity' as children grow up. In short, English-speaking children's acquisition of derivational morphology results from multi-variate interactions of both linguistic and usage-based factors.
Baayen, H., (1993). On frequency, transparency and productivity. In: Booij, G., van Marle, J. (Eds.), Yearbook of Morphology 1992. Kluwer, Dordrecht, pp. 181-208.
MacWhinney, B., and C. E. Snow. (1990). The child language data exchange system: an update. Journal of Child Language 17, 457-72.
Titre:
Multi-Variate Interaction in Children's Acquisition of English Derivational Morphology
Abstract/Résumé: In this study, we propose a regression model that entertains both linguistic and usage-based factors to explain English-speaking children’s acquisition of derivational morphology. We analyzed children’s production of derived words in a million-word corpus taken from the CHILDES (MacWhinney & Snow, 1990) database. The data came from the voluntary production of derived words by 469 individuals at age 3-10. We used Baayen’s (1993) hapax-conditioned degree of productivity (=PROD) as a function of both linguistic and usage-based factors in the analysis. The linguistic factors include the POS of the root and its derivative, semantic transparency (=S), and phonological transparency (=PH). Besides, we used four distinct input frequency measures as usage-based factors: How often the root of each derived word occurs in caregivers’ speech (=M1); How often each derived word occurs in caregivers’ speech (=M2); How many types of derived word each root generates in caregivers’ speech (=M3); and How many types of derived word each affix generates in caregivers’ speech (=M4). We conducted hierarchical multiple regression analyses using the natural log of PROD as a dependent variable and POS, S, PH, M1-4 and age (=A) as independent factors. Among these factors, POS, M1, and M2 were not significant, and were eliminated from the model. We obtained (1) as the final regression equation (R^2=.349). (1) ln(PROD) = -7.784+0.084*A-0.574*S+0.36*PH-0.373*M3+0.015*M4 The results show that the family frequencies of the root and the affix play the most important role in the model, and that semantic and phonological transparencies as well as age show limited effects on the total variation of ln(PROD). In particular, a single principle in (1) explains children’s performance on derivational morphology for all age groups. We can view the regression equation in (1) as a grammatical generalization that constrains the productivity of a morphological process. For instance, children produce 'dusty' more easily than 'electricity' at age 3-4 because 'electricity' has lower M3 and M4 values than 'dusty'. We propose that the acquisition process be the expansion of the (functional) domain from high-frequency words like 'dusty' to low-frequency words like 'electricity' as children grow up. In short, English-speaking children's acquisition of derivational morphology results from multi-variate interactions of both linguistic and usage-based factors.