ChildPhon lab

This page has three parts:

• Introductory notes.
• In-class exercises.
• Take-home exercises. Due Nov 1.

Introductory notes

ChildPhon is a database system, developed by Yvan Rose, designed to assist in the analysis of the phonological aspects of child data.

Yvan has been kind enough to allow us to use some data that he has coded from the Smith's (1973) diary study of his son Amahl (a classic work in phonological acquisition).

ChildPhon requires FileMaker Pro (for which BU has a site license), and runs only on a Mac. It also requires a special (commercial) font called IPAKiel. Word is that it is soon going to be implemented in Java (cross-platform) and will no longer require the commercial font, but that day is still in the future. Keep an eye on Yvan's ChildPhon web page.

The files you need to do the lab can be downloaded here (use the same username and password as for the readings, if you have trouble, try Control-clicking on the link and saving it that way), but I cannot provide FileMaker Pro for you, so unless you have that, these files will not allow you to work on computers outside the lab. It is possible to download a free trial version of FileMaker 7 (claims to work for 30 days), so that might be an alternative option.

It might be also useful to look at the short PowerPoint introduction from class (PowerPoint, PDF).

In-class exercises

The task: Explore Amahl's use of l.

Find the words that end with l.

• On any of the pages (e.g., Transcription), click the Find button.
• We're not looking for words that are spelled with an l at the end, but rather for words that end in the sound l. So, click on the IPA Adult field and type *l] to find things that have anything ("*") followed by "l" followed by "]". Note: When you type *, it will not appear, but it is actually there nevertheless. This is, I believe, a font issue.
• Click the "Find" button on the left under the rolodex to initiate the search. You should see 34.
• Flip through them, get a sense for what they are like.

Export your result

It's rather annoying to have to try to flip through the individual cards, and particularly so if you want to print a list of what you found (ChildPhon at present seems only to allow you to print the whole card view for each word, resulting in at least 34 pages for 34 words found). So, we'll export the results and look at them in an external program.

• Click Export.
• Choose the tab-separated text file type, name the file (for example, "final-l.tab"), and save it someplace where you can find it.
• You will then be presented with a dialog box showing, on the left, a list of fields that ChildPhon is keeping track of and, on the right, a list of fields that will be in the file you are about to export. The defaults are not very useful, so first click Clear All (emptying the list on the right), then scroll to the bottom of the list on the left, and choose the following four fields (either by double-clicking on the name in the list on the left, or by selecting the name in the list on the left and clicking the "Move" button): Tape counter, Word orthography, Word IPA C, Word IPA A. Then click Export.
• Now, go find your file (final-l.tab) and double-click on it, which should bring up the file in TextEdit.
• The symbols are all garbled because it isn't showing the right font by default. Under the Format menu, in the Font submenu, choose Show Fonts. This will bring up a font selection window. Under the Edit menu, choose Select All. Then in the font selection window, change the font to IPA Kiel and change the font size to something bigger, like 14.
• Note: When you change the font to IPA Kiel, you'll find that the "tape counter" numbers disappear (or, rather, change into tiny diacritics). That's a font issue. For now, what you should do is remember what the last word from Stage 1 is in your list, and perhaps even add an extra line between the stage 1 words and the stage 29 words so you can easily tell which stage a given word comes from.

What does Amahl do with his final ls?

Take a look at what we have here. There are basically two things that Amahl does when a word ends in l in stage 1. What are they? (There are a couple of exceptions). Does he still do this in stage 29? How might you describe the situations in which the two different behaviors surface?

The task: Look at a couple of other changes, of th and f.

Find the words that have a th in them.

• This is tricky because it requires using the IPA symbol θ, which we can't type.
• So, find a word that has it. Thrush, for example.
• Click Find, type thrush in the Word field, click the Find button on the left.
• Now, select the θ, copy it (Command-C or Edit > Copy).
• Click Find again, and select the IPA Adult field (we're looking for cases where the target has the th sound). Type *, paste (Command-V or Edit > Paste), and type * again. Then click the Find button on the left.

Review the results

• Export the results, open the file in TextEdit, change the font, and look at what we have.
• Any ideas about what Amahl is doing when he tries to pronounce th? It's different in stage 1 and in stage 29, what does he do in each stage?

Your turn:

• Now, find the words that have a f in them.
• What usually happens to an f in Amahl's speech in stage 1?
• In stage 29?

The task: Explore Amahl use of puzzle, puddle, and pickle words.

Locate the puddle, puzzle, and pickle words.

• First, go to the Syllables page.
• Click Find, put apple in the Word field, and click the Find button on the left.
• Look at syllable 2 for the adult. This is what we're looking for, obstruent-vowel-l, where the vowel is a schwa. So, what we want to do is locate all of the words where the second syllable of the target is OVL. This will include all three kinds of words, pickle, puddle, and puzzle, but that's fine, we don't need to refine it further (we can do that by hand).
• So, find all of the words where the second syllable of the adult target is OVL. You should find 26 of them.

Review the results

• Export them and open them in TextEdit, fix the font, and look at the result.

In stage 1, sort out the puddle words.

• These are the words that R3 says should come out as puggle. They are characterized by having n, d, or t before the -el sound at the end. So, puddle of course, although that isn't one of the words you found in stage 1. For now, look only at the adult form when trying to sort them. We're looking for all of the words in which R3 should have applied. I found 5 of them.
• Now, look at the child pronunciations. How many of the times R3 should have applied did it apply? How does that compare with what Macken (1980) found?

In stage 29, sort out the pickle words.

• These are words that in previous stages were conflated with puddle words as a result of Amahl's output phonology (according to Smith), but by stage 29 R3 has mostly disappeared. So, what Smith expects is that the pickle words should come out as pickle while puddle words (which used to sound like pickle words) will now revert to puddle. That is, there should be a distinction between puddle words and pickle words. Specifiecally, these are words that have g or k before the -el sound at the end. Again, look only at the adult forms. I found 3-4 of them (depending on how you count).
• Now, look at the child pronunciations. How many pickle words came out as pickle, and how many were pulled along with the puddle words, coming out as puddle?

Unfortunately, there are no relevant puzzle words in our corpus to show Smith's puddle~puggle~puzzle in action, although in stage 1 they should have come out as puddle.

The task: Explore some of Amahl's other simplifications

Select stage one

• Click Find and type 1 in the Counter field, then hit Enter (or click the Find button on the left edge of the screen).
• This should tell you that 238 cards were found.

Look at what happens to SC clusters at stage one.

• Move to the syllable summary (SyllSumm) view and look at the SC row.
• Observe that there were 20 adult forms that have SC, but Amahl never hit the target.
• What did Amahl do instead?
• There are two other options, either he added something to break up the cluster, or he reduced the cluster by deleting one of the consonants.
• How many times did he reduce the cluster?
  • To find out, click Last Find (to bring the previous Find back) and then click on the Reduction dot in the SC row. Then, hit Enter (or the left-hand Find button).
• Given what you now see, how many times did he add something to break up the cluster? (Don't search for it, you can just tell).
• Flip through the results you found to get a feel for what Amahl is doing.
• Reducing a consonant cluster involves deleting one of the consonants. These are SC clusters (s+consonant). Of the reductions Amahl did, how many of them involved deleting C? How many of them involved deleting s?

How does that change in stage 29?

• Last Find, change Counter to 29, Enter/Find.
• Oops! Well, how many times did the adult target have an SC cluster?
  • Last Find, shift-click Reduction (to un-select it), and click the A dot in the left edge of the SC row, Enter/Find.
• Things seem to have changed between stages 1 and 29.

Does Amahl produce r's?

• r is traditionally a pretty hard sound to produce for kids. Wabbit and so forth.
• Let's search to see how well Amahl does with his rs.
• Move to the Consonants view.
• Find, (Child side) Manner, Liquid, Rhot, Find/Enter.
• Point one: Are there any?
• Point two: What's one generalization that covers all the ones you found?
  • Hint: Not from stage 1, but from...
• Point three: Look at what's happening for the first couple alphabetically.
• What seems to be the pattern? (In what way do A's forms deviate from the adult forms?)
• Interesting. Any speculations as to why this particular error was made?
• How systematic is this change? Look at all of the vowel-initial (target) words at Stage 29 and see what you see.
• Go to Word Edges screen.
• Find, Counter to 29, select V-initial Wd dot A, Find/Enter.
• Flip through the vowel-initial (target) words it found at stage 29. What can you say about the effect we saw with r just above?

Check on Rule 1

Smith's Rule 1 says that a nasal consonant is deleted before any voiceless consonant. Let's see if he's right, and see how much coverage that rule has.

• Go to the Syllables view.
• Find, Counter to 1, and type *NO* in the Adult row under Detailed String.
  • Why *NO*? Rule one is talking about a nasal consonant before a voiceless consonant. Generally only obstruents are going to be voiceless. What *NO* means is "any number of characters or letters, then a nasal followed by an obstruent, then any number of further characters or letters"; * is a "wildcard."
  • Why Detailed String? We need the detailed string because it differentiates between the different types of consonants -- we need to distinguish nasals from non-nasals.
• How many of the 19 that you found do what Rule 1 leads you to expect? How many don't?
  • Note: b, d, g are voiced; p, t, k are unvoiced.
• Of the ones that didn't fall under Rule 1 (that is, for those that Rule 1 didn't make a prediction about at all), what happened?
  • (What you just came up with, probably, is pretty close to Smith's Rule 2)

Take-home exercises

Your task: Run through the following further ChildPhon exercises. Turn in answers to things underlined in red.

Now, we'll check on Codas, first in stage one, then in stage 29

• Go back to all of stage 1
  • Find, set Counter to 1, Enter/Find.
• Coda also seems to have a pretty low success rate at stage 1.
• Figure out what Amahl does with codas using the same kind of procedure as above.
  • That is to say, we know how often he hits the target. How often does he delete the coda? Add a coda? Add a vowel?
• Ok, that's pretty clear. Now see how he's doing at stage 29.
  • Find, set Counter to 29, Enter/Find.
• You will find that Amahl doesn't always hit the target at stage 29 with codas. Find the word(s) where Amahl doesn't hit the target.
• Do you think Amahl has codas down by stage 29? What do you observe about the failure(s)? Why did the counting work this way?
• Kids do tend to drop syllables. Amahl did a few times, check out away, escape, banana (stage 1).
  • In away, ChildPhon suggests that what's happening here is that Amahl realizes the underlying syllable a as we: (presumably inserting a glide for an onset, altering the vowel quality slightly, and lengthening it), and then deletes the second syllable wei.
  • What do you think? Does it seem like there's a more straightforward (and correct) way to describe what happened here? What is it?
• Any guesses as to which syllables are the ones most vulnerable to being dropped? (based on these three examples and/or intuition)

Does Amahl produce l's?

• l is a bit like r, but l is perhaps slightly easier (at least, it appears sooner).
• Let's search to see how well Amahl does with his ls, first in stage 1.
• Move to the Consonants view.
• Find, Counter to 1, then (Child side) Manner, Liquid, Lat, Find/Enter.
• Point one: Are there any occurrences of l at stage 1?
  • Recall, there were no rs.
• Point two: There's something a bit odd about all of the words that have ls in them (and we're looking at all of the words from stage 1 that have ls in them, remember). What do you notice about the other parts of these words?
  • Hint: In particular the other consonants...

Check out the harmony

There's another observation we can make about Amahl's pronunciation.

• Go to the Harmony view
• Find the forms from stage 1 where there are two consonants (separated by an arbitrary distance) with the place of articulation of the first one being labial, and of the second one being dorsal.
  • Find, Counter to 1, POA agreement (Adult's form) Lab...Dor, Find/Enter.
• Scan through these. Any systematic errors? There seem to be a couple; Rule 2 accounts for a couple, and A uses w in a couple of strange places. But, watch the dots in the POA agreement box as you scan through -- for the most part, the child form still has a Lab...Dor sequence when the adult form does.
• Ok, now, look for the sequence coronal following (eventually) by dorsal.
  • Find, Counter to 1, POA agreement (Adult's form) Cor...Dor, Find/Enter.
• (From watching the dots in the POA agreement box as you scan through them) how many of the child forms have a Cor...Dor sequence?
• In all of the cases you found (with adult Cor...Dor sequence), how does Amahl pronounce coronal consonants followed by dorsal ones?
  • Ignore dog. We of course don't want to suppose that Amahl, trying to say dog, came out with wowo. Presumably he was using some kind of dog noise as a label and trying to say that, not trying to say "dog".
  • Running and swing also don't fit so well into the generalization.

That's it.