Poster #187 - Gender Differences in Digit Dependence in Numerical Estimation

• In Event: 1-115 - Poster Session 03
  In Poster Session: PS 03 Section - Sex, Gender

Integrative Statement

A rich body of research has explored gender differences in numerical and mathematical abilities. While the findings have been mixed, a male advantage in overall accuracy has been observed on some symbolic numerical tasks such as number line estimation (NLE), which are widely used to explore numerical competence over development (e.g., Hutchison, Lyons, & Ansari, 2018). Though it is commonly assumed that the overall magnitudes of presented target numerals primarily drive NLE performance, recent work has shown that numbers with nearly identical magnitudes but different leftmost digits are estimated very differently (Lai, Zax, & Barth, 2018). For example, numbers like “302” are placed far to the right of numbers like “299” on a 0-1000 number line, although their magnitudes are essentially indistinguishable on a 0-1000 scale. Gender differences in this left digit effect (LDE) have not yet been explored.

To assess gender differences in NLE performance, we conducted pre-registered analyses on existing data (originally reported in Lai et al., 2018) from two 0-1000 NL tasks: a speeded task with children aged 8-10 and adults (N = 98) and a non-speeded task with children aged 7-11 and adults (N = 113). Overall accuracy was measured using percent absolute error (PAE) and the LDE was measured using difference scores of estimates of paired numerals falling on either side of “Hundreds” boundaries (e.g., 698 and 701).

Preregistered analyses revealed gender differences in performance at younger ages. On the speeded task, accuracy showed a gender and age interaction (F(3, 90) = 4.535, p = .005), with a female advantage at age 8 and a male advantage at age 9 (ps < .030). The same gender by age interaction was found for accuracy in the non-speeded task (F(5, 101) = 4.735, p = .001), with a male advantage at age 7 (p < .001) but no gender differences at other ages. No gender differences emerged for the LDE on either task. These results should be interpreted with caution due to small sample sizes when broken down by age/gender.

Exploratory analyses were conducted to investigate gender differences in children (of all ages) and adults respectively (with Bonferroni correction). Results from the speeded task revealed a stronger LDE for boys’ performance compared to girls’ (t(69) = 2.326, p = .046), and the reverse pattern in adults (t(41) = 2.422, p = .040). No gender differences in LDE were found on the non-speeded task, nor were they found in overall accuracy in children or adults on either the speeded or non-speeded task.

Further work is necessary to understand the sources of individual differences in accuracy, the LDE, and their relationship across development that may help explain the observed changes with age. Future work with additional data (i.e., larger sample size across age groups with an even gender distribution) is needed to more accurately assess gender differences in the various sources of error that underlie NLE performance.

Authors

• Katherine Gordon Williams, Wesleyan University
  Presenting Author

• Chenmu Xing, Wesleyan University
  Non-Presenting Author

• Taylar Clark, Wesleyan University
  Non-Presenting Author

• Esha Bhandari, Wesleyan University
  Non-Presenting Author

• Hilary Barth, Wesleyan University
  Non-Presenting Author