Term Project, Part 2, Production data

Ling H286, Autumn 2007, Ohio State University)

Copyright © 2007 by the class and Mary E. Beckman

0. Due date.

This part of the term project involves making and interpreting vowel charts for each different member of your group and for any other (groups of) speaker(s) whose vowel space(s) you want to examine. This will involve making measurements of first and second formant values in most of the h ____ d words that you recorded in class on October 1. Over the next few days, you need to fill in a TextGrid and make a Table of formant measurements for each of the members of your group and give them all to Mary to collate. These tables will be due by the end of class on Wednesday, October 24. Your group presentation will be in class on Wednesday, October 31, and your personal annotated copy of the presented material will due that day as well.

1. The recordings.

Here is the list of h ____ d words that you should measure, in the order in which you produced them for the recording:

  1. heed
  2. hid
  3. hayed (This is the past tense of hay -- i.e., to make grass into hay. It should rhyme with made.)
  4. head
  5. had
  6. hod (This is a wooden trough that rests on the shoulder which brick-layers use for carrying bricks up a scaffolding. It has the same vowel as in cot and should rhyme with sod.)
  7. HUD (This is the acronym for the U.S. Department of Housing and Urban Development, and it rhymes with bud.)
  8. hawed (This is the past tense of haw, as in He hemmed and hawed. and it should rhyme with sawed.)
  9. hoed (This is the past tense of hoe and should rhyme with mode.)
  10. who'd (This is the contracted form of who would and it rhymes with sued.)

I have made TextGrid objects for you to use in marking off the words and their vowels, with the vowels in the words hod and hawed at least marked off for everyone. You can get these TextGrids and the wave files from the recordings directory on the class web page. There is a huge zip file that includes all of the files, if you want to download them in one go.

2. The measurements.

Work with your group to measure all of the formants in your groups recordings. A good way to do this would be for each of you to mark off and measure your own vowels. To do this, begin by reading your wav file and TextGrid file into Praat.

Highlight the Sound object (e.g., Sound Alexa, if you are Alexa) and the TextGrid object (e.g., TextGrid Alexa) together and open them in an editor window using the "Edit" command button that will appear to the right of the "Objects:" list window when you select these two files togehter. You are now ready to start marking off where each target word is in the file and where the vowel in it starts and ends. Use the following ASCII approximations to the IPA symbiols as the labels on the vowel tier:

  1. i -- for the [i] in heed
  2. I -- for the [ɪ] hid
  3. e -- for the [e] in hayed
  4. E -- for the [ɛ] in head
  5. a -- for the [æ] in had
  6. A -- for the [ɑ] in hod
  7. v -- for the [ʌ] in HUD
  8. O -- for the [ɔ] in hawed
  9. o -- for the [o] in hoed
  10. u -- for the [u] in who'd

Each time you stop work on this part of the task, be sure to write your TextGrid to a file so that you do not lose your work. Do this by using the "Write TextGrid to Text File..." command under the "File" pulldown menu at the top of the edit window. The next time you start working on the file, you can then read the TextGrid file into Praat as well as the wav file.

When you have finished marking all of the vowels, make a Table object and put the formant values in it. You can use the script makeVowelTable.Praat script to set up the parts of the table that don't involve actual measurements. We will go over this script in class and show you how to adjust the formant settings if the formant values you see don't make sense.

Your TextGrid file and the Table file were due by end of class on Wednesday, October 24. You can still ask questions in class if you find difficult cases where you weren't sure how to mark the beginning or end of the vowel, or where it wasn't completely clear which formant values to take.

3. The analyses.

All of the Table files and many of the TextGrid files are now available in the directory termProjectPart2, so you can now download them all and start making and interpreting the vowel charts. That directory also contains the script classVowelCharts.R that we wrote together in class on October 29, comparing Joe's and Chanelle's vowel spaces. You can use the code there to figure out how to make plots that you may want to use to answer the following sets of questions that were suggested in the first set of reports from the perception experiment:

  1. How do people who differentiate the vowels in hod and hawed differentiate them? Does everyone who differentiates make the difference in the same way?
  2. Do people who identified as not making a distinction produce subtle differences that are like the differences that (some of) the distinguishers use?
  3. What adjustments do we have to make in our analyses to adjust for the fact that the men in the class have longer vocal tracts than the women?
p.s. If you think you might want to make histograms of the first and second formant values for all of the distinguishers and then do a t-test, as we did with the Hillenbrand et al. data in class on October 31, you can start with the code in the file classVowelsHistograms.R in the same directory with the other files.