Verbal and matched guise tests, and sense-relation networks
Now let’s look in more detail at the sort of data collection methods that are used in sociolinguistic research. In particular, these methods focus on collecting data about language attitudes.
Verbal guise test and matched guise test
Firstly, there is what is called a verbal guise test. This term (although it looks scary!) describes a pretty straightforward experiment design. In a verbal guise test, participants listen to different speakers reading out the same text. Most commonly, the recording(s) that participants listen to involve speakers with different regional accents reading aloud the same list of words. The participants’ job is to assess various qualities related to the voices that they hear.
In some cases, the participants might be asked to evaluate how friendly or intelligent a voice sounds. The idea here is to tap in to which pronunciations are linked with these somewhat abstract qualities. Participants might also be tasked with trying to identify the geographical region where they think the speaker is from. It might be of interest to the researcher to look at whether people who are more familiar with an accent have stronger views toward it.
One potential issue with this verbal guise approach is that different speakers with the same accent might produce speech that is slightly different - even if they are reading the same words out loud. As such, listeners might be assessing voices based on an individual speaker’s way of pronouncing words, rather than evaluating the accent more generally. To get around this issue with the verbal guise test, another type of tool was developed known as the matched guise test.
Instead of using different speakers, the matched guise test uses speech from the same speaker throughout the experiment and the speaker changes what accent they use. One way of collecting this sort of speech data would be to hire a voice actor who can portray different accents. By using this method, we can have better control over the more individualistic properties of people’s voices. Each listener’s assessment of the voices they hear is more likely to be based on the accent being portrayed. It is crucial in this method not to tell participants that they are only listening to one speaker throughout the experiment! Participants must believe that each voice is coming from a different speaker to conceal the experiment design.
Sense relation networks
Another recent method used for speech data collection in sociolinguistic research is called sense relation networks. The idea here is to give participants a few keywords. Then, the participants are asked to come up with words which mean the same thing as the keywords. Going back to our earlier bread rolls example, we know that participants will have different names for the same thing. In this data collection approach, participants are usually given time to complete this task independently, from a few days up to a week. This time allows each participant to think of a lot of words.
After completing the word list task, participants are then audio recorded in an interview about what they have written. Participants might be asked to talk about how they feel about specific words on their list, or what sort of factors influence their speech. Because the participant was given a while to complete the word list task, they have also had a chance to think about their own language use. This opportunity for self-reflection might also help the participant to feel more relaxed or better ‘prepared’ ahead of their interview.
Now you’ve been introduced to three new ways of collecting sociolinguistic data, take a moment to write down anything you feel is good or bad about these methods. If it helps, you could try thinking about what is good or bad about these methods for both the researcher and the participant. You’ll have an opportunity to discuss your thoughts in the next step.
© University of York