Empirical data in a scientific inquiry hide many things that cannot be uncovered until the field reaches a certain level of maturity. NASA scientists have collected seismic signals from the lunar surface from the 1970s, but they had to wait 40 years before they finally came to learn, thanks to the state-of-the-art technology of the computer analysis, that the moon has a liquid core like the earth. A classic example of applying ‘new and arguably better’ technology to the old data to obtain improved understanding of human language was the Empty Category Principle in the time of Government and Binding theory (Chomsky 1981) that provided principled account for such old data as that-trace effects, island constraints, etc.
The primary goal of this symposium is two-fold: (1) To see what we can uncover when we analyze the old data of psycholinguistics by using new or better methods of inferential statistics; and (2) To draw some consensus from the linguists' community that we must use best-possible statistical methods since advanced statistical analyses can reveal what we normally cannot see with basic statistics. Unlike theoretical linguists who always attempt to explain old and well-known phenomena with new and less-known frameworks, psycholinguists rarely apply new technology to the old data in the existing studies. What could we find out if we performed log-linear regression analyses of the multiple frequency tables in a journal article whose author carried out just as many Chi-square tests as the number of cross-tabulations? What could we find out if we transformed the raw data so that they could fulfill the crucial assumptions of certain statistical tests? What could we find out if we tested complicated causal models using Structural Equation Modeling for the data that used to be analyzed using basic regression methods? Papers presented in this symposium focus on applying new and less-known techniques to the old data in published works. The presenters analyze either their own data set or the raw data in earlier publications of other scholars. Notice that many papers already present the raw data in such formats as cross-tabulations in the published papers. Whatever data are chosen, we analyze them with new and less-known techniques to see what we could not see previously.
universal semantic features in the L2 acquisition of
This study investigates cognitive and affective factors in second language acquisition (SLA) using Structural Equation Modeling (SEM). Previous studies on the individual differences in SLA examined the role of learners’ cognitive abilities such as intelligence, language aptitude including working memory and phonological sensitivity, and affective variables such as motivation, anxiety and so on. Most of the studies looked at the relationship between each of such factors and the learners’ L2 proficiency using correlation analyses, factor analyses and regression analyses. However, such methods can only explain the direct relationships between dependent variable(s) and independent variable(s) without considering the direct effects among the variables, which fails to depict the full picture of L2 learning. On the other hand, SEM can reveal the direct and causal (based on a theory) relationships among the variables to show how the observed data set fit in a model of L2 learning based on the existing theory (Lewis & Vladeanu, 2006). Nevertheless, only a few attempts have been made so far (e.g., Sasaki (1993) with Japanese learners of English for English speaking; Winke (2013) with English learners of Chinese). In this study, we will review previous studies examining cognitive and affective factors in SLA and reexamine the data set using SEM in order to determine how well it fits in a model of L2 learning.
Lewis, M. B., & Vladeanu, M. (2006). What do we know about psycholinguistic effects? Quarterly Journal of
Winke, P. (2013). An investigation into second language aptitude for advanced Chinese language learning. The
Sasaki, M. (1993). Relationships among second language proficiency, foreign language aptitude, and intelligence: A structural equation modeling approach. Language Learning 43, 313–344.
This study examines the construct of learning ability in learners of Korean. Unlike previous KFL studies, the present study measures both cognitive and psychological factors, and interprets the whole factors as one dynamic system. The individual factors measured were sound discrimination ability, language analysis ability, working memory (WM), motivation, and anxiety. Both receptive and productive knowledge are measured as learners’ linguistic proficiency. 130 intermediate to high-intermediate learners of Korean from China, France, Japan, and the US take (i) phonological coding and language analysis test with PLAB, (ii) WM span with operation word span task, (iii) motivation level with Attitude-Motivation Test Battery, (iv) anxiety with Foreign Language Anxiety Scale survey, (v) receptive knowledge as vocabulary, grammar, reading, listening with optimized TOPIK test, (vi) productive knowledge as written and spoken language ability upon complexity, accuracy, and fluency. We’ll construct a structural equation model (SEM) to explain learners’ linguistic achievement. To explain the factors of Korean learning, receptive and productive knowledge scores were set as dependent variables, and period of learning, cognitive, and psychological factors as observed variables. We assume that the cognitive and psychological factors directly affect L2 proficiency mediated by period and psychological factors. Both significant and nonsignificant factors will be discussed, and, as conclusion, the model will explain the dynamic aspect of language acquisition in context of Korean language learning.
Native speakers’ introspective judgments of sentence
acceptability have been an essential tool for linguistic
research. However, this traditional method of data collection
has been criticized in various respects, and the crucial
criticism concerns its methodological ‘informality’; namely,
such judgment data cannot evade inter- and intra-speaker
variation, as it is not gathered by means of objective and
rigid procedures. A number of alternatives have been
suggested such as corpus data, but judgment data still
function as the primary source of empirical evidence in
syntactic research, due to its efficacy. Since the work of
In experimental syntax, one of the most useful and
frequently used ways to support or refute a theory is to
conduct acceptability judgment tests. Despite the
importance of acceptability judgment in theorizing, however,
few researchers in this field have seriously dealt with the
issue of how to analyze data from sentence judgments,
which are measured in Likert scales (e.g. from 1 to 5, or to
7). In fact, analysis of variance (ANOVA) has been taken
for granted by most previous works for the analysis of this
type of data