Linguistic diversity is a common and still growing phenomenon, and safeguarding this diversity is one of the most urgent challenges facing our world.  The number of spoken languages in the world is estimated at between 6,000 and 7,000. About 40% of the world’s population has one of the eight most common languages as its native language (Mandarin, Hindi, Spanish, English, Bengali, Portuguese, Arabic and Russian), and most of the languages (more than 4,000) are spoken by less than 2% of the world’s population. For many countries linguistic diversity is a domestic phenomenon. When different languages co-exist in one area, or when one person speaks more than one language, the impact of language and language related factors on psychological and educational assessment processes can be considerable.  The aim of this symposium is to show some of the most novel researches designed to improve assessment practices in linguistic diversity contexts.

The first paper which is presented by Paula Elosua  shows an integrative approach to deal with some language related factors affecting educational assessment in contexts of linguistic diversity.  Kadriye Ercikan from University of British Columbia addresses the question of to what extent the tests are accurate and meaningful measures of targeted constructs such as mathematics and science learning when students have limited proficiency in the language of the test. The third presentation given by Serge Lacroix, University of British Columbia, analyzes the impact of code-switching on tests scores coming from a variety of verbally mediated tests belonging to WISC, PPVT and WJ.  Alina von Davier (Educational Testing Service) presents an overview of the issues surrounding the  ways the linguistic diversity affects performance in game-based assessments. Finally Maria Elena Oliveri from Educational Testing Service analyze ways to evaluate, quantify, and minimize sources of potential construct-irrelevant variance in the assessment of linguistic minorities and the administration of assessments internationally along various test development stages

Sociolinguistic factors affecting educational outcomes in linguistic diversity contexts
Paula Elosua, University of the Basque Country, Spain

The potential factors affecting the assessment in linguistic diversity contexts and the approaches to deal with them can be described from different perspectives on multilingualism.  We consider four approaches: linguistic, psycholinguistic, sociolinguistic, and psychometric. In this integrative framework we revise the contributions of each of those areas from a theoretical and empirical point of view. The empirical research analyzes the impact of different language related factors on mathematic competence. The study was conducted in the Basque Autonomous Community, Spain, where Basque and Spanish are official languages. Based on population data (N=16580), the results show the effect of sociolinguistic factors on mathematic competence, and the risk of sociolinguistic bias when two or more language coexists. 

The impact of code-switching on cognitive test scores
Serge Lacroix, University of British Columbia, Canada

Code-switching is a common behavior among multilinguals. What would be the impact on test scores if responses were accepted in more than one language on measures of cognitive abilities?  This presentation will review a study conducted in French and English where students in the experimental group were given the option of responding in either languages. Following a strict procedure, they were also presented with test items in either language, based on their response pattern. The two groups of students (control-experimental) were compared on test results from a variety of verbally mediated tests from the WISC, PPVT and WJ. The control group was not presented with the opportunity to code-switch, although they did "naturally" while the experimental group was presented with the option without being told they had to code-switch. Results show that the experimental group obtained significantly higher results on most measures. They have also code-switched at a much higher rate than the control group. 

Validity of interpretation of mathematics and science scores for linguistic minority students
Kadriye Ercikan, University of British Columbia, Canada
Yue Chen,
Chris Campbell, University of Victoria, Canada

Students around the world are educated and take tests in a language different than their home language. This raises the question of to what extent the tests are accurate and meaningful measures of targeted constructs such as mathematics and science learning when students have limited proficiency in the language of the test.  The research reviews conceptual and empirical research on how language proficiency potentially affects performance on mathematics and science assessments and presents findings from empirical research.  The empirical research utilizes data from the Programme for International Student Assessment (PISA), which assesses 15 year olds’ reading literacy as well as their mathematical and scientific literacy. This presentation will review different ways language proficiency affects performance on assessments and present findings from empirical research regarding the degree to which language proficiency accounts for variations in mathematics and science scores and discuss the implications of these findings with respect to the how they affect comparability of scores for different language groups.

Facets of Fairness in Performance Assessments
Alina von Davier, Jessica Andrews, Thales Ricarte, Jiangang Hao, and Lei LiuEducational Testing Service, USA

There is societal need for developing assessments that are more in sync with the way people learn and work. These new assessments are more contextualized and are meant to be aligned with the learning materials--school curricula and/or students' life activities (playing games, for example). Many game- and simulation-based assessments (SGBAs) have been developed in the recent years as the medium of choice for performance assessments, some of them being available in several languages;  very few have been investigated for fairness, the impact of language and culture, and (longer term)transfer. Most recently, collaborative tasks have been included among the performance tasks in games and the fairness issues have been seen from yet another perspective.This presentation will review different ways language and cultural background affects performance on game-based assessments with collaborative tasks and provide findings from empirical research. In this presentation I will give an overview of  the issues surrounding the  ways the linguistic diversity affects performance in game-based assessments. The design of the performance test and of the target population are outlined first; then the psychometric considerations are presented. A real data example is presented; in a collaborative simulation-based assessment dyads of test takers solve a science problem together. The data were collected using crowdsourcing. The performance was analyzed across different background variables: gender, native language, country of birth, rural/urban upbringing, race. Results are discussed in terms of the propensity for different teams to exhibit a specific collaborative behavior.The rapid advances of technology change our life, our schools, and naturally, our assessments. The psychometric community has the obligation to ensure that the new assessments are appropriate for all test takers. This presentations attempts to formalize an approach to fairness for linguistic minorities in game-based tests with collaborative tasks.

A Validity Argument for the Fair and Valid Assessment of Linguistic Minorities
María Elena Oliveri and René Lawless, Educational Testing Service, USA

Globally, tests are administered to test takers who are not fully proficient in the test language.  Language proficiency however can impact students’ test responses often for construct-irrelevant reasons.  In assessment, valid score-based inferences require disentangling the effects of language proficiency from the assessed construct and not confounding it with test-takers’ ability on the assessed construct. Current efforts to increase test fairness for assessing linguistic minorities focus primarily on post hoc analyses consisting mainly of conducting differential item functioning analyses.  The existing approaches thus neglect considering validity along all test development stages.  Our objectives are to address this limitation and demonstrate the application of a validity framework that considers possible sources of construct-irrelevance occurring along key test development stages.  It also describes fairness threats arising along each of the six stages (domain definition, evaluation, generalization, explanation, extrapolation, and utilization) of Kane’s (2013) validity argument.  We collaborated with a team of content experts and evaluated two higher education assessments developed for a domestic (U.S.) population, which later were utilized internationally to make score-based inferences for multiple populations.  We conducted a literature review and evaluated a random set of items and ancillary materials from each assessment.  We developed recommendations for activities and considerations to enhance the assessments’ validity and fairness in support of principled international use.  This presentation helps address a paucity of research related to considering fairness and validity threats and possible sources of construct-irrelevant variance in the assessment of linguistic minorities and the administration of assessments internationally along various test development stages.  The results demonstrate ways to evaluate, quantify, and minimize sources of potential construct-irrelevant variance more systematically.  We conclude by providing recommendations to enhance the validity of assessing linguistic minorities along various test development stages.