Design and validation of a rapid visual processing measure for screening reading difficulties in early childhood

Abstract

As the development of visual processing abilities is known to precede the development of reading abilities, identifying visual measures that reliably correlate with reading measures has tremendous theoretical and practical importance. A major challenge in addressing this question empirically is developing reliable behavioral measures for developmental studies, because most tasks are not equally reliable across different age groups and require iterative design changes to ensure that the measure reliably indexes the intended construct across the developmental span. Here, we present a series of studies that show how to iteratively modify a behavioral task by data-informed task changes and the use of item response theory to reduce task redundancies and develop a fun, fast, reliable, and easy-to-deploy web-based measure for K/1/2-graders in school settings. Our results show, in a large, diverse, and representative sample (N ~ 1,550), that the ability to rapidly encode visual information reliably correlates with reading outcomes at the end of the academic year. The developed measure offers the potential for use as an early screening tool to identify children at risk for developing future reading challenges.

Keywords: Early childhood; Early visual processing abilities; Item response theory; Measure development; Reading disabilities.

Fig. 1

Trial sequence in the partial report version of the MEP task. We modified two dimensions of the MEP task to achieve similar performance levels across different age ranges of the population: the encoding time of the string of letters and the number of elements in the string

Fig. 2

Task performance and encoding time. a How K/1 children perform in the MEP task with six-element strings and encoding times of 120, 240, and 480 ms; b Performance (d′) for each trial condition. Data show that 5–7-year-old children performed at chance, with an encoding time of 120 ms and string length of six letters. Performance increased above chance at longer encoding times, but the task was still very difficult. Error bars represent one standard error of the mean

Fig. 3

Task performance depends on string length and encoding time. a, b Overall task accuracy and d′ from Study 1 and Study 2. Including trials with four elements and cutting down the shortest encoding time of 120 ms seem to have improved performance in the MEP task in Study 2. c, d Task accuracy and d′ for each string length and its corresponding encoding time. Note that an encoding time of 480 ms did not improve task performance. Error bars represent one standard error of the mean

Fig. 4

Task performance with three string lengths and one encoding time. a, b Accuracy and d′ improvement across different cohorts. Spearman–Brown-corrected split-half reliability for each study is provided in b. With the addition of two-letter strings, we observe a significant improvement in performance without compromising task reliability. c, d Performance across each trial condition in Study 3

Fig. 5

MEP performance predicts reading skill. a Correlation between performance in the MEP task and KTEA standard scores. b Cross-sectional sample reported in previous work (reproduced from Ramamurthy et al., 2023). c–e Correlation between task performance in the two-, four-, and six-letter trials and KTEA standard scores

Fig. 6

Item response theory (IRT) analysis of MEP data. a Item difficulty as a function of trial sequence.b Item response functions for all three blocks of different string lengths. c Correlation between KTEA and ability estimates. d Correlation between performance in the d′ space and ability estimates from the IRT analysis

Fig. 7

MEP task performance in the K/1 population with letters and pseudo-letters. a Trial sequence in the pseudo-letter condition and performance across different string lengths in both versions of the MEP task. b Correlation between the letter and the pseudo-letter versions. c, d Correlation between task performance and reading outcome measured using the composite of three subtests of the KTEA assessment battery (see Methods for details)