1) Formulate a research question relating to a topic in phonetics. The simplest is simply to examine a phonological contrast -- eg, showing what the differences are between some minimal pairs.You are encouraged to investigate a language other than English, however make sure there are a sufficient number of speakers available to you. Try to develop a topic for which at least 3 true native speakers can be found.
2) Topics should be developed in consultation with the instructors before research is begun. Port will meet with each group a time or two to get your project idea sharp and to help suggest modifications that will make it easier to get good results. Consult with Port again as soon as you have a preliminary interpretation of your data and before you write it up.
3) An analysis of your data should be conducted using equipment in the Phonetics Lab. To keep your project manageable in size, try to use not more than 15 to 20 utterances per speaker, for a total of 100 to 200 utterances for measurement.
4) Your data should be presented in some appropriate statistical format, including relevant graphs or tables. It is nice to do some statistical tests, if you know how to do them, but it is not required. Hopefully the results will be so obvious, that statistical tests are not essential.
5) A thoughtfully-written, typed descriptions of your project should be handed in before the end of the semester. Your paper should have roughly 8 pages of text, plus tables and graphs (normally each on a separate page).
2) Make sure that the target word or words are placed in a phonetic environment in which they will be easy to measure. For example, it is easy to mark the boundary between [a] and [s]; it is NOT easy to mark the boundary between [a] and [j]. If you are measuring formants, it will be easier to do with a low fundamental frequencies (and thus easier with male speakers).
3) Embed your test words near the middle of a sentence rather than at the beginning or end. Words in sentence-initial and sentence-final position tend to have quite distinct and nontypical pronunciations. For example, it is difficult to mark the onset of an utterance-initial stop closure.
4) If your test sentences would reveal obviously to your subject what exactly you are looking at, you should probably include superfluous sentences so that your goal is not so obvious. This measure will help prevent your subject from exaggerating or biassing their productions in order to `help you out'. It's best not to tell them until the experiment is over what exactly you are looking for.
5) When speakers are presented with a list of sentences, their intonation often follows a predictable, list-reading pattern. Since their intonation is likely to be different at the very beginning and very end of the list, put irrelevant sentences in these places that you aren't going to measure.
6) Similarly, in order to average out any differences based on the location of a sentence within a list of sentences, it is a good idea to randomize the order of sentences for each subject. Or present the items in several random orders. This will hopefully even out any effects of position in the list and distribute the effect evenly across all the test items.
2) Although you do not need to do this for these course projects, keep in mind that in publishable research in this country, it is now expected that all subjects will sign an `Informed Consent' statement agreeing to participate in the experiment subject to a number of conditions that are spelled out. Normally the experimenter promises to either destroy the records after completion of the experiment or to preserve confidentiality. Subjects are assured they may leave the experiment at any time for any reason without penalty, and so forth.
3) Make sure you have tried out your task yourself. Decide just where you need little breaks and some bigger breaks from the task. You want the subjects to be comfortable. Think about whether they need time to pause, clear their throats, or even get a drink at some point.
4) While recording them, listen carefully to your subjects to make sure they are giving the kind of speech you want. Don't be afraid to tell them what to do, and ask them to repeat anything that is not done correctly. Sometimes you may need to demonstrate what you want them to do. Most people are pleased to be asked to serve as speakers of their language.
5) Have your speaker practice the reading task before you begin recording. This gives the speaker time to feel more comfortable, it gives you ample opportunity to make sure they are doing what you want (and it is a good opportunity to adjust recording levels, etc). It also helps to minimize speech errors or disfluencies (e.g., odd pauses or breaks).
6) Have your speaker produce more repetitions than you need. It is very time-consuming to have a speaker come in a second time, and you don't have to measure everything you record (though you should choose which ones to measure according to a criterion that you report -- eg, `the 2d through 7th repetition').
7) Make sure you are actually recording!
8) Leave plenty of space before and after your test items so that it will be easier to manipulate.
2) If you plan to come in at an odd hour--evenings, weekends, or very early morning-- make sure you get a lab key from Port or the Dept office - or have make arrangement to meet someone who has a key. The outside doors to Memorial Hall are normally unlocked on Saturday and Sunday and in the evening.
3) Please be VERY careful about security. The lab contains many items
that are very attractive to thieves.
February 23, 2003
RFP