3.3. Critical assessment of the methods

3.3. Critical Assessment of the Methods

Limitations

My mini-project was limited by my lack of statistical expertise and instruments. Therefore, I could not determine the individual influence of multiple factors. This meant, in turn, that I could not explain certain differences between the participants’ ratings and the corpus probabilities. Neither could I account for variation between individual participants, but had to rely on the relatively high number of participants to balance those differences. Moreover, I had to replace the factor verb by verb sense. All in all, my mini-project points out some tendencies, but cannot by any means be as precise as a more comprehensive approach.

100-split task

The 100-split task poses several chances and risks. On the positive side, it allows for a greater differentiation than yes – no questions. Even when it is transformed into a ‘10-split task’ as in my study, participants are given a range of choices, and the method still yields comparable figures that can easily be processed by statistical programmes and Excel. Moreover, no technical equipment is needed. Furthermore, the task grants a high degree of control: unlike in the sentence completion task, for example, in which participants do not necessarily continue the sentence with a dative construction, all results obtained with this task are usable. As all participants deal with the same sentences, a huge amount of comparable data is gathered. Finally, the method makes it possible to use real-life sentences from a corpus.

A drawback of the method is that it requires the participants to express their intuitions about language in figures, i.e. to measure something that can hardly be grasped. Furthermore, the participants know that their results will later be analysed. According to the observer’s paradox, this may influence their choices. Moreover, rating a given sentence is not the same as producing a sentence. The fact that a participant rates two alternative datives with 5 – 5, for example, does not automatically mean that he4 would produce each alternative equally often if he uttered this sentence several times in real life. But as the results from Bresnan and Ford’s experiment and the corpus model are in line, those constraints obviously do not have too strong an influence.

Convergent evidence

By drawing on Bresnan’s and Ford’s corpus probabilities and conducting a psycholinguistic experiment, I used convergent evidence, too. Especially since the experiment and the corpus probabilities were based on the same dataset, I was able to compare the results. Generally, although a multi-method approach requires more time and knowledge and is thus not always applicable, it enables the researcher to look at a complex phenomenon from different angles.

4  I am going to use masculine forms when I refer to a participant in general. This is meant to include female participants as well.

Created with the Personal Edition of HelpNDoc: What is a Help Authoring tool?