Selecting Test Participants
From SurveyWiki
| Administer the Pilot Test | |
|---|---|
| Before Testing Participants | |
| Testing Participants |
There are some differences between the selection criteria for participants in the pilot test and those in the final test. The suggestions below are not exhaustive, and the rationale for each may not be the only one.
In both cases, you will need:
- At least ten (10) people – this number is the minimum necessary to give your results meaning, that is, to establish statistical significance. This number may be reduced for the pilot test if ten people are not available.
- Both men and women – this practice ensures that if a sex-based, in-group variety has inadvertently been used for the text, the results will show it. Unless you have valid reasons for focusing on one sex, there should be equal numbers of both.
When selecting subjects for the pilot test, look for people who:
- Speak, as their first language, the speech variety of the text used in the pilot test – this restriction is a general confirmation that the person can verify the text and questions.
- Were born in the local community – being native to the area means that a person will be a good judge of a text that is supposed to have come from that area.
- Lived nearly their entire lives in the local community – this condition limits the amount of exposure to other varieties the subjects have had through travel.
When selecting subjects for the final tests, look for people who:
- Have not lived in other areas for any extended time – a long time away may have changed the person's ability in the test point language and therefore his or her ability to understand related varieties.
- Have had little or no contact with the speech varieties you will be testing at that particular test point – such contact gives a person an undesirable "advantage" in being able to understand the texts.
You may also need:
- People chosen at random – usually, statistical random sampling is only needed when a RTT is used to test bilingualism.
- Other criteria that your entity or the team leader has deemed necessary.
A bit of additional advice that was obtained the hard way:
- Limit the number of texts – if you have several texts to play in one place, there may be too many for each person to sit through. Choose more subjects than the usual ten so you can spread out the tests. For example, instead of playing six stories to each of ten people, choose another five people and have each of the fifteen listen to four stories.
- Investigate gender roles – there may be a correlation between gender and exposure to other lects. For example, if the women in a group do not typically travel, then they will tend to have less exposure to and understanding of other lects. The opposite will be true of the men. On the other hand, if the sellers in area markets are mostly women, then they will have more exposure.
- Check the sociological factors that may influence sampling. For example, if you find that a group practices clan exogamy, then you would know not to test the married women, since they come from another area. (However, such women may provide other useful information related to language use, bidialectalism, perceived inherent intelligibility, etc.)
- Question the frequency and duration of travel and be careful about how the terms you use are defined locally. In Tanzania, people reported that they had always "lived" in their hometown, even though they had "traveled" to another place and "stayed" there for five years!
