The Structure of Mathematical Language Skills During Preschool and its Relation to Numeracy Skills

• In Event: 3-227 - Relational Language and Math: Variation Across Types of Language and Math Activities

Integrative Statement

Introduction: Mathematical language—words and concepts necessary for children to learn and apply early mathematics skills such as more, few, before, near—is a strong predictor of early mathematics development (Toll & van Luit, 2014). Intervening on this skill has even been found to promote development of numeracy skills (Purpura, Napoli, Wehrspann, & Gold, 2017). Most often when measures of this construct are used, items are combined into a single measure, even though there appear to be distinct aspects of quantitative (e.g., more, fewer; Barner et al., 2009), spatial (e.g., near, above), and ordinal language (e.g., before, last; Pruden et al., 2011). The purpose of this study was to evaluate the factor structure of mathematical language and identify which component(s) are related to numeracy performance.

Hypotheses: It was hypothesized that a 2-factor structure of mathematical language—quantitative and spatial+ordinal—would emerge as the best fit to the data, but that only quantitative language would uniquely predict numeracy skills.

Study population: A total of 295 children participated in this study. Children were 3 to 5 years old (M = 4.45, SD = 0.58), 46.1% female, and 67.5% Caucasian. 40.7% of participants had parents with less than a college degree.

Method: Children were assessed in their preschools by trained project staff on a measure of mathematical language that included 35 items measuring quantitative, spatial, and ordinal mathematical language. A subsample of the participants (n = 124) were also assessed on measures of numeracy skills, IQ, and rapid automatized naming (RAN). The full sample was used to evaluate the factor structure of mathematical language skills through a series of confirmatory factor analyses at the item level. Using the subsample, a multivariate regression analysis was used, controlling for IQ, RAN, age, gender, and parental education, to evaluate which aspect(s) of mathematical language are significantly related to numeracy skills.

Results: To test the factor structure of mathematical language, three potential nested factor structures were compared: a 3-factor model (quantitative, spatial, ordinal), a 2-factor model (quantitative, spatial+ordinal), and a 1-factor model (quantitative+spatial+ordinal). Model fit indices are presented in Table 1. Although all three models fit the data well, chi-square difference tests indicated that the 3-factor and 2-factor models fit the data significantly better than the 1-factor model. However, there was not a significant difference between the 3-factor and the 2-factor model. Therefore, the more parsimonious model—the 2-factor model—was selected as the final model.
When quantitative and spatial mathematical language were included in the multivariate regression model to predict numeracy skills, accounting for covariates, quantitative language (β = .30, p < .001), but not spatial language (β = .04, p = .684) significantly predicted numeracy performance. These findings suggest that although quantitative and spatial mathematical language can be separated into distinct factors, only quantitative language is uniquely related to numeracy performance. Future work should investigate how both quantitative and spatial language are related to other aspects of mathematics performance such as geometry and spatial skills.

Authors

• David Purpura, Purdue University
  Presenting Author

• Caroline Byrd Hornburg, Purdue University
  Non-Presenting Author

• Sara Schmitt, Purdue University
  Non-Presenting Author