Lightbrown et al (2003)

The New Brunswick E.S.L. Study & the power of comprehensible input

Eric Hermann shared a really cool article on Ben’s about E.S.L. language acquisition in New Brunswick. The study compared classes who basically got to do a load of free voluntary reading (and/or listening) for five years from Grade 3 to Grade 8 with classes who had direct instruction in writing etc. The results are that while after two years the free-reading/listening kids did a well as the others on all measures (comprehension, speaking, writing), but 8th grade the structured lesson kids (well, one batch of them) had pulled ahead.

The study is worth reading– and taking a hard look at the numbers– because it can suggest a number of things which seem to contradict Krashen and others which support his views that language is acquired only via loads of comprehensible input.

a) It seems that, in terms of writing, some guided practice and feedback makes a difference when kids get to beyond-beginner levels. There is a lot of speculation about why this is, but it basically boils down to this: acquisition (via comprehension only) seems to have certain limits. For example, when a student hears the Spanish “me gustan los tacos,” s/he can figure out one way or the other that “me gustan” means “I like.” The problem is that once learners “get” what “gusta” means, they are less inclined to focus in on other stuff. The –n on gustan is used for plurals (liking more than one object). You need to know what gusta means to communicate or understand; you don’t need to know (or use) the -n rule for this, so naturally it is much later acquired.

Kids and second-language learners do this all the time. The way English kids pick up negation like this, as VanPatten and Gross remind us:

“At first negation is simply NO followed by a phrase. Then NO moves inside the phrase and “don’t” is also used in an unanalyzed way. Then the idea of “-n’t” attaching to modals in an unanalyzed way. Finally the whole system of auxiliary verbs and the correct usage of “not” and of contractions.”

So, basically, the “most important stuff”– meaning-based– is acquired first; the later stuff is window dressing. (This is one of the reasons why the third-person -s ending is late acquired in English…it’s there but basically unnecessary). In this study, it is noteworthy that the direct-instruction kids did somewhat better in writing than the acquisition-via-comprehension-only kids. It is possible that this happened because people needed to have their attention consciously focused on some aspects of writing that they would otherwise not naturally pick up on right away. It is also possible that– as a number of researchers have argued, defying Krashen– that the production system and the compehension system in the brain operate in an “at odds” way sometimes, and must be trained differently. While input drives pretty much everything to do with remembering etc, (there is no argument among SLA researchers about this) it is possible that output needs some training.

b) There are some design flaws in the study. E.g. group RG1 was assigned homework; there were therefore different treatments (in terms of time) between the control and the experimental groups when the only difference should have been what was taught. In addition, while the RG1 group was assigned homework, the experimental groups were not allowed to take reading/listening home with them, further amplifying the differences ebwteen them. One of the experimental group teachers also started grammar teaching three days a week in Grade 8. It is possible that the two teachng styles– read on yoru own, or listen to teacher– conflicted.

c) The study could very well support Krashen’s comprehensible input hypothesis in that it is amazing how well the experimental kids did with no instruction. One of the traditional groups scored basically as well as the experimental groups; one control (RG1) scored somewhat better. The study seems to suggest that free voluntary reading (and/or listening) works about 90% as well as a teacher…but without a teacher). This suggests to me that, in terms of aquiring language, what the teacher does is really very minimal compared to what reading (or listening) do. This research broadly reflects other research on literacy, which notes that the teacher accounts for 10-20% of kids’ achievement in class, while other factors– parental literay, wealth, etc– account for much more.

d) Broadly, what I get out of this– loads of input for beginners; more specific writing feedback as they get older is needed– is what Blaine Ray and Susan Gross have recommended for years.

The moral of the story? Comprehensible input– via stories, reading, listening, etc– works incredibly well. Students may do well with feedback (in writing, for writing) to improve their writing. Teacher effects on writing quality are significant but small. My recomendation? Help your Level 2 and up kids with writing: give feedback (simple feedback) and model the sticking points in stories.