Word associations across early childhood: An early paradigmatic shift and reduction of variability

• In Event: 1-100 - The Role of Semantic Networks in Early Word-Learning

Integrative Statement

A classic finding is that between 6-9 years, children’s free associations for nouns and verbs shift from syntagmatic to paradigmatic relationships. (Nelson, 1977). Syntagmatic relationships are words that co-occur syntactically (summer-vacation); paradigmatic relationships are words from the same grammatical class (summer-winter). However, infant researchers have reliably found evidence for the activation of paradigmatic relationships by 24 months (Wojcik, 2017). Because of task differences and a lack of data for children aged 3-5, it is difficult to integrate this contradictory evidence regarding the emergence of paradigmatic relationships. By testing 3- to 8-year-olds in a toddler-friendly free association task, we aimed to reconcile these differences in order to understand the development of semantic networks across early childhood.

Method. Children (n=60, 3-8 years) and adults (n=60; 18-43 years) were instructed to respond to a prime word with the first word that came to mind. Each participant was tested with half of a 65 prime word list that was created by piloting the most frequent content words in the CHILDES database. Responses were recorded, transcribed, and tagged for whether they were a syntagmatic, paradigmatic, or other type of response (e.g., rhyme).

Results. While younger children (3-5 years) had more excluded trials (M=16%) than older children (6-8 years; 8.5%) or adults (2%), the exclusion rate was markedly low, suggesting that the task is appropriate for young children. Even the younger children gave an average of over 30% paradigmatic answers, and two analyses show that the syntagmatic-paradigmatic shift begins at around age five (Figures 1 & 2). Older children gave a higher proportion of paradigmatic responses than younger children (p=0.0017); adults gave a higher proportion of paradigmatic responses than older children (p=0.012. This and all further comparisons are Tukey post-hoc). Additionally, age significantly predicted proportion paradigmatic responses when treated as a continuous variable.
To investigate the variability in responses across development, an ANOVA compared a measure of how many different responses were given for each prime. There was a significant effect of age group, F(2, 170)=36.02, p<0.001, η2 = 0.30. The number of different responses was significantly lower for adults than for either older or younger children, ps <0.001. This result was supported by an ANOVA comparing the proportion of idiosyncratic answers (given by only one participant), which revealed a significant effect of age F(2, 193) = 49.85, p<0.001, η2 = 0.34 (younger children > older children > adults, ps<0.001).

With age-appropriate stimuli and a task that is engaging and easy to comprehend, this work is one of the first to collect free association data for children under five. Analyses revealed that the shift from syntagmatic to paradigmatic associations might occur several years earlier than previously thought, with children as young as three producing paradigmatic response to prime words. These findings suggest that, in line with infant studies, early semantic networks include paradigmatic relationships. Furthermore, younger children produce more idiosyncratic responses than older children or adults, suggesting that the study of early semantic networks should consider individual differences in word knowledge and input.

Authors

• Erica Wojcik, Skidmore College
  Presenting Author

• Padmapriya Kandhadai, Douglas College
  Non-Presenting Author