The main problem with empirical research is that it can only be 100% validated when you add a strong, statistical analysis of variance (ANOVA), or when you infuse the research with other quantitative and statistical research tools such as a chi-square. This is because empirical research is based entirely on your personal interpretation of the phenomena taking place at the time of your observation. As much as we try to maintain objectivity and not be biased, this is a very hard task to accomplish when it is time to process empirical data and then establish a correlation.
Now, a really unique situation happens in the field of linguistics, particularly educational linguistics and SLA. The variables found in studies made with ELLs always seem to be quite complicated. Most ELL research has to be repeated in a myriad of different contexts and scenarios because every empirical or quantitative data obtain will depend on these variables:
- IQ level
- developmental level
- lexile level
- language acquisition exposure level
- socioeconomic background
- communicative language usage in percentage/home
- communicative language usage in percentage/school-social
- language aptitude and social affect
And these are just some of the hundreds of variables to consider and parameters within which we must frame any investigation. Can the naked eye consider all of these variables with mere observation? It could be plausible, but statistically speaking you would need much more than just this. Therefore, the main problem with empirical research in L2 is the consideration that must be given to all of the social, academic, and developmental variables that can surface during a formal observation.