This study investigated possible gender bias on a vocabulary test, using a method suggested by Andrich and Hagquist to detect “real” differential item functioning (DIF). A total of 443 adult ESL learners completed all 228 items of the Peabody Picture Vocabulary Test (PPVT-IV). The 310 female and 133 male participants were assumed to be of equal competence, corresponding to levels B1 and B2 on the Common European Framework of Reference for Languages. Male participants outscored female participants, possibly due to the multiple-choice format and to the fact that most gender-biased questions favored men rather than women. Finally, our analysis process yielded only seven items out of 228 as showing gender DIF, which is much lower than the numbers reported in the literature for ESL tests. This low figure suggests that the high number of gender-related DIF items reported in previous research might be attributed to the use of DIF detecting methods that do not take into account artificial DIF stemming from the cross-contamination of test items.

Researchers are frequently confronted with unanswered questions or items on their questionnaires and tests, due to factors such as item difficulty, lack of testing time, or participant distraction. This paper first presents results from a poll confirming previous claims (Rietveld & van Hout, 2006; Schafer & Gra- ham, 2002) that data replacement and deletion methods are common in research. Language researchers declared that when faced with missing answers of the yes/no type (that translate into zero or one in data tables), the three most common solutions they adopt are to exclude the participant’s data from the analyses, to leave the square empty, or to fill in with zero, as for an incorrect answer. This study then examines the impact on Cronbach’s α of five types of data insertion, using simulated and actual data with various numbers of participants and missing percentages. Our analyses indicate that the three most common methods we identified among language researchers are the ones with the greatest impact n Cronbach's α coefficients; in other words, they are the least desirable solutions to the missing data problem. On the basis of our results, we make recommendations for language researchers concerning the best way to deal with missing data. Given that none of the most common simple methods works properly, we suggest that the missing data be replaced either by the item’s mean or by the participants’ overall mean to provide a better, more accurate image of the instrument’s internal consistency.

This study aims to assess the extent of crosslinguistic influence in English as the weaker language of unbalanced bilingual children, and to compare the extent of such influence to that reported in the second language acquisition (SLA) literature. Additionally, by comparing children from different L1 backgrounds, we aim to see if typological distance impacts crosslinguistic influence. We collected elicited speech samples from 16 Polish-English and 44 French-English children who have had dual language input from birth, but whose English is weaker mostly because it is absent outside the home environment. The crosslinguistic error rates (an average of 6%) are lower for our participants than averages found in SLA literature, but still considerably high. Although French- and Polish-dominant children present comparable error profiles, the extent of crosslinguistic influence tends to be greater in the case of French-English bilinguals than for Polish-English bilinguals, which may reflect the perceived distance between the languages.

L’indice de détection de patrons de réponses inappropriés l_z (Drasgow, Levine & Williams, 1985) a été appliqué à un test d’habileté en lecture en langue seconde de 64 items soumis à 171 étudiants universitaires. L’objectif était de confronter un rejet intuitif de données de recherche à une élimination suggérée par l_z. En outre, l_z a été mis à l’épreuve pour détecter 12 participants additionnels ayant répondu par pseudo-hasard. Les résultats suggèrent que, bien que l_z détecte efficacement des patrons de réponses aberrants pour de grands groupes et qu’il soit préférable à l’élimination intuitive, cet indice présente des limites pour l’analyse de plus petites matrices de données.