Documentation:
Poletiek, E. H. (2002). Implicit learning of a recursive
rule in an artificial grammar. Acta Psychologica, 111, 323-335.
Summary
The ability of people
to learn structural knowledge by simply introducing them examples of these
structures has been brought out for many years.
Due to the fact that either studies or researchers advocates different
perspectives, which vary from unconscious acquisition of abstract knowledge to
conscious acquisition of segmental examples.
Thus, the aim of this research was to explore the learning ability of a
specific abstractive system through artificial grammar learning tasks, that's
recursive grammar. A series of
preparations were done before starting the task. First of all, an artificial grammar graphic was
adapted and extended for the satisfaction of recursion. Next, a large amount of exemplars that were
needed to present to participants for previewing and for testing were set
between nine to thirteen letters. Exemplars were created to include zero to three
self-embedding features, and they were made into fifty-nine correct and
incorrect structures equally. For the
tentative tasks, two parts which the first one was about to recognize the
feature of recursive rules by taking true and false question items selected on
the computer program and the other was post-interviewing of participants
strategies use during the tests were included.
Several hypotheses were proposed before testing. One, the recursive operation could be learned
in AGL tasks. Two, structures with zero
and two self-embedding levels were much easier to recognize. Three, structures with one and three levels
were more difficult to acquire. Four, generative
aspect of recursive rules could also be learned in the limited exposure of
examples. Yet, tests and interviews
results showed that participants memorize especially two or three letters to
answer questions. After revising the
hypotheses, the second tests were conducted on the premise that question items
had been reset, and self-embedding elements were advanced to five levels. While participants were taking the tests, hints
and clues about how many levels of the items they face were provided, and post
interviews were included as well. Even
though the second results were more close to what the research questions wanted
to discover, many limitations, such as the complexity of recursive rules, the
completeness of knowledge acquirement, and the relationships between examples
and recursive learning were revealed in the end of the research.
My Critique
After
reading this research, there are pros and cons that I am going to illustrate in
the below.
Speaking
of merits, I think, first of all, its introduction gives a very clear ideas
about the aim of this study, the abundance of literature reviews, the idea of
connection between artificial grammar task and recursive rules, the hypotheses,
and the two-times experiments. In other
words, the introduction of this research almost includes everything of the
whole research, except tests results, offering an integral previews. For another merit, preparation before
experiments are adequate, and the processes of experiments are explained
thoroughly and evidently. Also,
two-times of experiments are considered as the author's intention or
determination toward finding the proofs that recursion can be learned by AGL
tasks. Still another, testing results
and discussions for experiments are pointed out not only with accurate
statistics but also with insight explanation in the text level. In the end, limitations and suggestions for
further studies are presented in the end of the general discussion.
However,
the biggest problem of this research is the two-times processes of its
experiments conducting. According to the
title of this research and its introduction, the purpose is to make
participants learn recursive rules in AGL tasks under "IMPLICIT"
learning. To my understanding, implicit
learning is a little bit similar to inductive learning that is to generate or
to learn certain rules from many examples with that certain features. Then, in the two-times experiments, the
participants are shown exemplars on the screen in the first place. And, they are told to answer questions due to
what they have seen previously rather than provide examples as references for
them while answering. In the view of
this, I have my doubt on the necessity of second experiment conducting because
by memorization, recursive rules can be learned for sure, and the only thing to
concern is its time taking; in other words, the levels of repetition and reinforcement
even if the researcher found different participants and did in a parallel way. Moreover, if artificial grammar learning
model can be applied to real natural language learning uses is still another
central issue to solve. Other flows, may
be try to distinguish the literature reviews from the introduction to make the
whole part more pleasant, and move the acknowledgment to the very beginning of
the research would be better.
In
conclusion, the researcher's spirits toward study is what I really
appreciate. I think no matter what the
results would be, thoroughly preparations are needed in any cases.
沒有留言:
張貼留言