Using Multivariate Adaptive Regression Splines to Predict Lexical Characteristics' Influence on Word Learning in...

Area of Science:

• Child language acquisition
• Educational psychology
• Linguistics

Background:

• Identifying effective vocabulary targets and instructional sequences is challenging.
• Understanding factors influencing children's word learning is crucial for effective pedagogy.

Purpose of the Study:

• To investigate how lexical characteristics collectively influence children's word learning.
• To determine the relative importance of various word features in predicting word acquisition across different grade levels.

Main Methods:

• Secondary data analysis of 350 students (grades 1-3) from a vocabulary intervention study.
• Utilized multivariate adaptive regression splines (MARS) to model the influence of word frequency, concreteness, phonotactic probability, neighborhood density, and age of acquisition on learning 377 words.

Main Results:

• Age of acquisition emerged as the most significant predictor of word learning across all grade levels.
• The influence and importance of other lexical characteristics varied between the first-, second-, and third-grade models.
• MARS models demonstrated good fit, with low root-mean-square error and generalized cross-validation scores.

Conclusions:

• Lexical characteristics impact word learning differently depending on the child's grade level.
• Findings provide nuanced insights into school-aged children's vocabulary acquisition.
• Results can inform a framework for sequencing vocabulary targets based on predictive word features.