Curriculum Innovation, Testing and Evaluation: Proceedings of the 1st Annual JALT Pan-SIG Conference.
May 11-12, 2002. Kyoto, Japan: Kyoto Institute of Technology.

How C-tests work

by Ellen Head (St. Andrews/Momoyama Gakuin University)


Response to Paul Hackshaw's paper "Are the tests we give our students fair?"

This is a theoretical and speculative response to the paper given by Paul Hackshaw on piloting fair and reliable tests. In his presentation, Paul described the C-test and gave us an example to try, as well as a useful handout on C-tests from a website by Ulrich Raatz and Christine Klein-Braley of the Gerhard-Mercator-Universität Duisburg.
The C-test differs from a cloze test in that the initial half of each deleted word is given. There is some experimental evidence to show that the C-test has a good correlation with oral testing and is more reliable than cloze tests in which the initial portion of the missing word is not given. Why should this be so? Intuitively, I feel that giving the testee the first part of the missing word should make it easier to find the answer. Actually when we tried to do the example question, many of us had trouble finding the answer to at least one item. So why should C-tests be accurate? Perhaps the answer is to do with the way that words are stored in the brain.
The traditional model of word recognition involves the idea of a mental lexicon through which we search sequentially, rather as we look through a list of words in a paper dictionary (Forster's serial search model, 1976). In the early 1980s under the influence of artificial intelligence research, a rival model was developed, in which each word has a corresponding "feature counter" called a logogen. A logogen becomes activated when it gets a certain amount of evidence. For example, if the letter "t" is activated, all logogens which correspond to words containing "t" will increase. If the activation level manages to pass a threshold, the logogen "fires" and the word is "recognized". Evidence which increases the activation level can be either perceptual or contextual." (Morton's logogen model, described in Harley, 1998, p. 91).
The process at work when a student is reading a gapped text involves using contextual information. Giving the first half of the word, as the C-test does, adds an element of perceptual information. Hopefully, future research will be able to give an account of the neuro-linguistic processes which are at work in language testing. Currently, processing in the C-test is being researched by Dr Günther Sigott at the University of Klagenfurt. An extensive bibliography on the C-test is available at:

http://www.uni-duisburg.de/FB3/ANGLING/FORSCHUNG/HOWTODO.HTM

A final thought as to a source of texts which could be adapted as C-tests: student-written texts might be a nice source of material that is interesting to other students and is not too difficult.

References

Forster, K. I. (1976). Accessing the mental lexicon. In R. J. Wales & E. Walker. New approaches to language mechanisms, (pp. 257-276).

Harley, T. A. (1998). The psychology of language, from data to theory. Hove, UK: Psychology Press.



2002 Pan SIG-Proceedings: Topic Index Author Index Page Index Title Index Main Index
Complete Pan SIG-Proceedings: Topic Index Author Index Page Index Title Index Main Index

[ p. 168 ]

Last Next