Summing up: how to make a good test
After receiving so much input in the previous videos, let us summarise the main points of how to make a reliable language production test.
- 4 seconds maximum: DES is not considered to have longer lasting effects than this.
- Nouns and verbs: the test should be triggering the production of both of the essential word classes.
- Ideally 50-100 items: per subtest you need a sufficient number to avoid repetition.
- No synonyms: an ideal item elicits only one word, not several possibilities
- 15% buffer: go with more items in the test phase than you are actually aiming for in case you need to throw some out.
- Run the test in a small group.
- Check that the length does not exhaust the participants. If so, shorten it.
- Correct the setup, if necessary.
- Throw out the items that do not meet the criteria (e.g. the 4 second rule)
- Redraw some of the pictures, if necessary.
- Only keep those items on the list that are named correctly by 90% of the participants.
- Once it runs smoothly in a small group, run your test in a bigger healthy population that meets the criteria of the patient group such as gender, age etc.
- Even after this run, you can throw out items that do not fit the criteria in the bigger group.
- Once it is proven, that the test triggers the desired results in a healthy population, test it in a big clinical group.
Only if it goes through all these steps, can a test be considered reliable and valuable for clinical assessment. This seems quite a long way, but it is the only way to be confident your test is working in a healthy population and can hence function as a comparison to the clinical group you want to test eventually. Moreover, this is the only way to ensure you are not wasting the doctors’ and patients’ time and the only way to ensure you are not misdiagnosing an impairment that is actually only due to some poor items.
© University of Groningen