Identifying Gifted English Learners: Reforging the Leaky Pipeline Into Programming and Services

• In Event: 42.076 - Talent Ignored: Investigating the Underidentification of English Learners for Gifted Programs

Abstract

Objectives or Purposes

Differential representation of English language learners (and other racial, ethnic, and income groups) is recognized as a glaring problem within gifted education. In this paper, we consider the implications of recent work on the identification process for EL students (c.f. McBee, Peters, & Waterman, 2014; McBee, Peters, & Miller, 2016; McBee, 2016; Card & Giuliano, 2015) and argue for changes to the identification process that may reduce these disparities

Perspective(s) or Theoretical Framework

Recent research has demonstrated the effect that the inappropriate use of multiple measures as well as the screening or nomination phase has on the size and relative diversity of gifted populations. Separate work from the field of economics demonstrated what when universal screening is applied along side a group-specific preference for ELs, their representation in gifted populations is substantially increased and underrepresentation is all but eliminated. There is also a growing research base suggesting that simply using non-traditional assessments, such as nonverbal ability tests, don’t serve to mitigate the problem of underrepresentation (Carman & Taylor, 2010; Giessman, Gambrell, and Stebbins, 2013; Matthews & Kirsch, 2011). New solutions for the effective identification of EL students are urgently needed.

Methods, Techniques, or Modes of Inquiry

We extend the quantitative theoretical analysis of identification systems, as presented in McBee, Peters, and Miller (2016), to the problem of identifying EL students. This analysis is leveraged on two key assumptions: (1) that classical test theory is a reasonably accurate model for test scores, true scores, and measurement error, and (2) that the true scores and observed scores follow a multivariate normal distribution. Under these two assumptions, conditional probabilities (representing system performance metrics such as sensitivity, false positive rate, and positive predictive value) can be calculated by numerical integration. Nomination and identification cutoffs can be represented as the limits of integration, while the reliability and validity of the assessments can be represented as values in the covariance matrix. Specific issues we will examine include the effectively high percentile cutoffs when traditional g/t cut scores are applied to disadvantaged subgroups, as well as the possibility of the disparate impact of poorly designed (but unbiased) nomination procedures.

Data Sources, Evidence, Objects, or Materials

Results of the theoretical analysis will be supported by empirical results from Card & Guiliano (2015) and from McBee (2006).

Results and/or Substantiated Conclusions or Warrants for Arguments/Point of View

Potential solutions include modifying (or eliminating, in the case of universal screening) the identification process, as well as adopting more inclusive standards for making placement decisions on the basis of multiple assessments. However, broadening or otherwise modifying identification procedures brings a corresponding need for changes to the services being provided. We will also address the implications that proactive identification measures carry, such as the fact that the newly-identified students will now have a wider range of learning readiness and needs that must be addressed through a wider range of learning readiness and needs that must be addressed through a wider range of services and supports.

Authors

• Michael S. Matthews, University of North Carolina - Charlotte

• Matthew McBee, East Tennessee State University

• Scott Joseph Peters, University of Wisconsin - Whitewater