Phrase frequency, proficiency and grammaticality interact in non-native processing: Implications for theories of SLA

Kailen Shantz
  • Second language Research, October 2016, SAGE Publications
  • DOI: 10.1177/0267658316673403

Second language learning uses rule-based knowledge and fine-grained statistical knowledge

What is it about?

Broadly, there are two prominent classes of models for second language learning: rule-based models, which assume that grammar reflects a set of categorically violable rules, and usage-based models, which assume that grammar reflects fine-grained knowledge of the statistical properties of a language. In this paper, I pit the predictions of these approaches against one another by examining if and to what extent the sensitivity of second language learners to grammatical violations depends on the frequency of the phrase in which a violation occurs. The results are ultimately not fully compatible with usage-based or rule-based approaches to second language learning, suggesting instead that second language learning and sentence comprehension relies on both usage-based and rule-based knowledge and processes.

Why is it important?

In the field of second language acquisition, there is a strong tendency for researchers to work exclusively within the framework of either a usage-based or rule-based approach, without giving much consideration to what the other approach might predict. Consequently, there has been very little research that attempts to actually test these different approaches in a manner that might falsify one approach or elucidate the extent to which either approach can or cannot successfully capture the full range of linguistic knowledge and behavior observed in second language learners. Importantly, the results of this paper indicate that these approaches may be complementary rather than diametrically opposed, and demonstrates that there is much to be learned by simultaneously testing the predictions of both approaches.


Dr. Kailen Shantz (Author)
University of Illinois System

Writing this paper was interesting for me, in that it presented somewhat of a test to my integrity as a scientist. I was, admittedly, completely surprised by the results. My background and training has come largely from a usage-based perspective, and accordingly I was fully expecting the results to support usage-based approaches. When they did not, my first thoughts were that I must have done something wrong and that if I could figure out what I had done wrong, I would find the results I expected. At the same time, I was able to recognize that my thoughts were being driven by pre-existing biases, and that to do as I wanted would be to p-hack my way into the results that I wanted. Obviously, I ultimately did not take the p-hacking route, which I am quite proud of. But this gave me an interesting window into why p-hacked results are so prevalent, and into the importance of trying to be theoretically agnostic when conducting research.

Read Publication

The following have contributed to this page: Dr. Kailen Shantz