Frequency and language processing

I will try to sum up what my second paper on the usage-based language acquisition approach is about:

Nick C. Ellis (2002). REFLECTIONS ON FREQUENCY EFFECTS IN LANGUAGE PROCESSING. Studies in Second Language Acquisition, 24, pp 297-339. doi:10.1017/S0272263102002140.

The paper was actually written in response to some feedback on a previous paper, which I haven't read, so I will most probably fail to understand it in depth, as I suspect it will descend even more to the nitty-gritty than the original paper did. Nevertheless, here goes my reading.

Ellis explains in the first point of his article that the previous paper dealt with the effect of frequency on implicit learning, and he stated that "(...) language learning is implicit learning.", although he also acknowledged the importance of noticing and explicit learning. He then goes on explaining the relationship between SLA (I assume it stands for second language acquisition, although it is not made clear in the text) and the neuroscience of learning and memory. Apparently, we have two sets of memory, which are independent from each other: explicit memory (where we consciously remember past experience) and implicit memory (a current stimulus relates to a past stimulus, with no conscious remembering). Ellis' prior article linked frequency effects with implicit memory. Implicit learning would work like a light drizzle, a bit of "learning without being aware of it", whereas explicit learning would be that sort of learning we would use to learn our home phone number by heart. So, new associations are best learned explicitly, and if they are to be acquired implicitly, they need many repetitions. Ellis claims that noticing (explicit) is important to learn formulae, but it is implicit learning which will give us things like figuring, generalisation, fluency and nativelike performance. So, explicit learning would give us the gross building blocks of L2, and implicit learning would do the fine polishing.

In the second point of his article, Ellis talks about what must and what need to be noticed. We need to notice the exceptions to general patterns or rules. He then goes on to explaining that frequency shapes language processing to some degree, as the occurrence of a certain word can actually be predicted on the basis of its past probability of occurrence, and not only that: a particular word is more likely to occur when other words that came along with it in a past sentence are present too. So, we tend to learn "chunks" of words, and that is why we give priority to the overall meaning of an idiom over the literal meanings of the words that it is composed of. Apparently, we need to notice the complex and ambiguous, but the more straight-forward can be learned implicitly. Thus, if we fail to notice consciously certain features of L2, no matter how frequent they are in the input received, we will not learn them.

The third point of the paper points out the most frequent situations where noticing fails. One would be related to grammatical particles and inflections which give redundant information, as they are accompanied by other elements in the sentence that offer the same information in a more outstanding way. In those cases, we need to draw the attention of the L2 learner to them. Another example has to do with nonnative L2 pronunciation, despite massive exposure. Ellis summarises two theories that explain why this happens and, in summary, proposes that offering L2 learners input that will exaggerate the difference between L1 and L2 sounds that are likely to be mixed helps learners notice and separate them. So, we need to give learners of L2 high contrast though input when they will be likely to ignore phonetic differences compared to L1. Transfer from L1 would be another reason to fail noticing. Regarding the fixed sequences of L2 acquisition that learners seem to follow, regardless of their L1, exposure to L2 and teaching methods, Ellis stresses the importance of frequency and salience in determining those sequences. Ellis also brings up an interesting outcome of previous studies, which shows that when the material to be learned is randomly structured, has a large set of variables and the important relationships are not obvious, explicit learning can be ineffective.

The fourth point in Ellis' article deals with attention and form-focused instruction. He claims that training L2 learners on where to place attention and how to perceive formal aspects of language learning can be improved.

The fifth point of his article talks about the role of frequency in both implicit and explicit learning. As a matter of fact, Ellis started to study frequency related to explicit learning. Ellis now explains that we seem to be unable to consciously access the data on frequency stored in our implicit memory, so we tend to make incorrect guesses when asked to do so.

The sixth point is on the sequences of acquisition, which Ellis believes to be formula - low scope pattern - construction. He explains that, unfortunately, no wide studies have been conducted in the development of L2, unlike the work of Tomasello and others on L1 acquisition by children. Despite evidence put forward by Bardovi-Harlig suggesting that Ellis' sequence would not explain all in L2 development, Ellis defends his view, stating that since frequency of occurrence leads to entrenchment, half of fluent native text is made by formulae. He also points out that much of the deviation from his default sequence noticed by Bardovi-Harlig could be explained by environmental and methodological factors that affect the L2 learner, and that the closer we get to the natural way in which children acquire language, the more his default sequence works. Studies conducted by Hakuta and Wong-Fillmore on child SLA would point in that direction.

The seventh point has to do with whether L1 and L2 are acquired the same way or not. While Bley-Vroman claims that both are guided by distinct cognitive principles, Ellis thinks that L1 and L2 acquisition are fundamentally the same. On the other hand, Bley-Vroman thinks that after learning a novel L2 construction, there is an strengthening mechanism to somehow "fix" it, which only works if we actually notice the frequency of that construction, whereas Ellis believes that strengthening occurs regardless of whether we notice that the construction is used or not.

The eighth point deals with grammatical categories, which are now considered to be much less defined than before. Thus, while many nouns have all of the behavioural characteristics of that category, many others don't; so, some nouns are more "nounlike" than others. Ellis supports the usage-based language acquisition approach, claiming that syntactic categories cannot be constructed in different languages following the same universal grammar, as generativists defend.

In the ninth point of the article, Ellis talks about the variationist perspectives on SLA, based on sociolinguistic research. He argues that, although psycholinguists and sociolinguists have different starting points, they converge in their outcomes.

The tenth point of the article feels like a sting, although I cannot understand what Ellis talks about, as I am unable to figure out whether he was accused of excessive parsimony by Eubank and Gregg, or it was for the lack of it. Whichever way the argument is, the three authors have clearly known each other for a long time and Ellis doesn't have very warm feelings towards them.

In the eleventh point of the article Ellis continues to whip Eubank and Gregg, describing their transition theory as autistic and self-contained acquisition theory, although he agrees that an acquisition theory is needed.

The last point of the article is devoted to wanna constructions, which generativists claim to illustrate the abstract syntactic properties of universal grammar (again, I assume UG stands for universal grammar). Ellis turns the generativist explanations around and ends up claiming that wanna forms, just like gonna forms, are actually arguments for his view.