Both in sociolinguistics and in research on learner language (such as Second Language Acquisition Research) questions of keen interest are (1) how many vowel sounds a (group of) speaker(s) distinguish(es) - i.e. what their vowel phoneme inventory is - and (2) how they realise them. Information on this can gained from research methods probing speech production and speech perception. Methods that focus on speech production can be further distinguished based on the degree of control that the experimenter exerts. A very controlled design (e.g. a word list) holds many extraneous variables constant, but makes the task somewhat artificial in the sense of being far removed from ordinary communicative situations (e.g. speakers may get nervous and overly conscious of being the object of research, leading to hyper-correction). A very naturalistic research design relies on a task that is as close as possible to spontaneous, private conversations, but sacrifices a lot of control on extraneous factors.
A comprehensive research design for the elicitation of English vowels and consonants in various contexts and speech styles is provided by the Phonologie de l'Anglais Contemporain Project (PAC - Phonology of Contemporary English). The CAP research design has been used to elicit data from speakers of L1 varieties (e.g. British English) and institutionalised L2 varieties (e.g. Indian English) but may not in its entirety be useful for speakers of L2 leaner varieties, depending on their proficiency.
A less comprehensive and less time-consuming method that offers a high degree of control is to elicit vowels with a word list. Participants are simply asked to read a list of words. A drawback of this method is that speakers fall into a "list intonation" such as reading words in groups of four, in the patterns high-high-high-low. This can interfere with the reliability of this research instrument, such as through vowel lengthening in the final member of the group. This issue can be avoided by embedding words into a carrier phrase, such as "She said X again", where X is the target word.
A readily usable word list (with all targeted words inserted into a carrier phrase) is available on from a public website at the Open Science Foundation. The accompanying table lists the keywords, the lexical set they belong to and how the respective vowel is realised in British English.
A third option consists of having speakers read a text that contains phonemes of interest. This method is more naturalistic than reading a word list but provides more control than a free conversation. Moreover, speakers can then be asked to retell the story, which adds a semi-spontaneous speech condition to the repertoire of methods, while also maintaining a relatively high degree on the speech material (e.g. specific phonemes will occur in specific phonological environments in the retelling). This method permits the collection of semi-spontaneous data even from speakers who may not be proficient enough to sustain a regular conversation. Speakers who struggle with the task or leave out parts of the story that contain words that are of interest to the researcher can be prompted with questions by the researcher (e.g. 'And what happened then?', 'What did the mouse do then?').
A readily usable, short text is available from a public website at the Open Science Foundation. The accompanying table list the keywords, the lexical set they belong to and how the respective vowel is realised in British English.
Speakers with limited linguistic and/or cognitive capabilities pose particularly challenges to linguistic data collection. For example, learners of English in primary school typically have very limited proficiency in English and also have a fairly limited mental lexicon even in their L1. Instructional material for these learners will focus on words that denote concepts from the their daily lives (e.g. tooth, mouth, pen, dog, cat). Elicitation of speech data can be carried out by presenting cards with pictures of concepts and asking the speakers to name what they see in English. Only concepts and objects that are known to be familiar to this specific group of learners (e.g. because they feature in the textbooks this group uses) should be used.
Data collection should also involve the collection of some personal data on the participants. Variables such as age, gender and L1 background are often of interest to the researcher or should at least be taken into account. An example is provided on this website.