Week 7 Update

This week, I got my first very rough look at some results, which is cool. I looked at the average PMLU scores for Quince’s produced speech, Quince’s comprehended speech, Reilley’s produced speech, and Reilley’s comprehended speech. Weirdly enough, all of these numbers were within about 0.2 of each other (with standard deviations of around 2.5), so right now it’s not looking good for the whole complexity-decides-which-words-kids-say hypothesis. But that’s ok, because 1. Dr. Ingram was explaining to me that we have a lot of work to do on the analysis of this part to fix it by eliminating outliers and stuff and 2. I have two other hypotheses to test before I have to start getting worried.  Also, this might maybe help me prepare for the AP Statistics mock that I’ll be taking soon. That should be interesting.

Some more stuff I’ve been doing at ASU involves typing out the IPA transcription of every single one of Quince’s produced words, fixing all the errors I’ve been finding in my spreadsheets, and performing more analysis on my data with semantic categories.

The transcriptions took forever since I had to prompt Quince to say the word into my phone so I could record it, listen to his pronunciation a ton of times, figure out the sounds that make up his pronunciation, translate those sounds to IPA, type them using this IPA symbol thing, and copy the final transcription into my spreadsheet. Even though this whole process was pretty tedious, it was definitely worth it. I always knew Quince had a limited range of sounds, but I never really realized that he essentially only has ONE vowel. His consonants are actually not even close to as restricted as his vowels are. For each word, he puts a consonant in front of, behind, or sandwiched between the ə sound, which is pronounced like the u in up, and then puts these consonant-vowel units together in different combinations. Dr. Ingram mentioned that vowel problems are very common in apraxia, so this definitely makes sense for Quince’s diagnosis. And the most exciting thing is that I don’t think anyone realized just how limited Quince’s vowels are, so if we focus on getting more vowel sounds in his speech therapy, we can hopefully improve his comprehensibility significantly.

Most of the errors I’ve been fixing in the spreadsheets have to do with excel auto-filling cells with things I typed in the first time, like replacing the word butter with butterfly because I wasn’t paying attention and had entered butterfly in an earlier cell. I also somehow managed to switch the produced and comprehended columns of words in some but not all of the spreadsheets, so I had to go through and check that all of those were in the right spot.

For the semantic analysis, I looked at a textbook that Dr. Ingram gave me (and also is the author of, so that’s cool). I broke my word lists into five categories: specific nominal, general nominal, action words, modifiers, and personal-social. I can go into more detail if anyone wants, but basically the most important things to know are that general nominals are most nouns and personal-socials are the words we use to relate to different states and social expressions like hi, bye, or thank you. There are two general kinds of kids, expressive and referential. Expressive kids tend to have a vocabulary that has a higher percentage of personal-social words than receptive kids, who tend to have more general nomials in their category.  We aren’t exactly sure why this is. Maybe it has something to do with personality, like more outgoing people using more personal-socials, or upbringing, if a child’s parents try to name everything they see to expand their child’s vocabulary or something. Anyways, I looked at my sample of Quince’s and Reilley’s vocabularies and calculated the proportions of each category in them. Right now, it looks like both Quince and Reilley fall somewhere in the middle of these two categories, but after Dr. Ingram helps me account for all the error and stuff this could easily change.

Week 6 Update

A couple of interesting things happened this week. First of all, I have been analyzing the CDI of another kid, Reilley. Dr. Ingram actually has three different CDIs of Reilley's, at 1;11, 2;2, and 2;5. (Everyone probably knows this already, but I didn't so I'll explain it anyways. 1;11 is a way to write the child's age. The first number is years and the second number is months, so 1;11 means Reilley was one year and eleven months old when the first CDI was done.) Even though Reilley was so much younger than Quince when this data was taken, they are at similar places in terms of vocabulary as Reilley's language developed normally. This allows me to see in what ways Quince's speech is unusual.

Right now, Dr. Ingram and I are trying to figure out why Quince and Reilley say some words but not others. For example, Quince says pajamas but not shirt. We have a few hypotheses for this right now: 1. The words that the children don't say are generally more complex in structure than the words they do, so they are avoiding the more complex, harder-to-produce words. 2. There are specific sounds within the words that the children don't say that are especially difficult for the children to produce.  3. The words that the children don't say are ones that they aren't exposed to or aren't interested in, so they simply don't need them. 

Personally, I think that it's a combination of these reasons, and probably some others as well. Off the top of my head, I can think of examples from Quince's CDI for all three. For the first one, Quince calls orangutans monkeys, even though he can make all the sounds within the word orangutan, because it's very complicated. For the second one, Quince uses the word pop instead of soda because he can't make the 's' sound. And for the third one, Quince never says the word brother because he doesn't have any. But still, I'm interested in seeing which reasons are more prominent than others.

I'm testing these hypotheses by looking at the data from the CDIs in different ways. To test complexity, I’m calculating the PMLU scores for each of the words. Dr. Ingram actually invented the PMLU score as a way to measure word complexity based on word shape. Each vowel unit (not individual letter) is assigned one point, and each consonant gets two. The higher the score, the more complex the word is. I haven’t started on the specific sounds yet, but I’m probably going to try to identify patterns of particularly prevalent sounds and conspicuous absences in the list of produced words. For the situational part, I’m dividing the words into categories such as foods, clothing items, actions, etc. I’m not really going to be able to look at the exact details, like Quince not saying brother because he doesn’t have one, because I don’t have these details for the other subjects.

Something I'm realizing about this field is that it's a lot more subjective than what I'm used to in the research I’ve done, which is mostly physics/engineering kind of stuff. First of all, how can we define exactly which words the child understands and which the child says (which is what these CDIs are based on)? You can try to test comprehension by putting objects in front of the child and asking them to identify one, but a correct response could be produced through random guessing, or process of elimination using already known objects. And where do you draw the line for ability to produce a word? If the child can repeat it? If the child can respond to a question with it? If the child can label something with it? If the child can use it spontaneously without any prompting? 

This uncertainty is very frustrating for me. I can’t tell what criteria were used to fill out Reilley’s CDIs, so the data might not even be comparable. But I guess it’s good for me to get used to doing this kind of thing too? 

Finally, I’ve been trying to record a lot of Quince’s speech, and I noticed something very interesting. When I record what he says and play it back for him, he listens and tries to correct his own recording. The corrections are clearer and more coherent. This is very exciting for us, as it presents a new technique we might be able to incorporate into his treatment.

Here is an example recording I took of my mom working with Quince. In it, my mom is trying to get Quince to say "mow." On the third try, he manages to get the right vowel sound, and then says "I did it." (I repeated "I did it" afterwards to translate for my mom, because she didn't recognize what he said at first.)

Quince's speech exercise (this is a link, you can click on it)

My mom: Mow
Quince: Ma-me
My mom: Mow
Quince: Ma-me-me
My mom: Mow
Quince: Ma-mo, I dee ee
Me: I did it, ok 

Week 5 Update

I took my spring break last week.

Week 4 Update

I'm going to make this update pretty short because I already posted something this week, my project hasn't changed much, and I'm exhausted. Mostly, I've just been going through the dissertation I mentioned before, working on analyzing the CDI data, and studying more phonetics. I got really tired of going through all of the words and individually assigning them a PMLU score, so I've been working on writing a program that will do it for me. So far, I've been completely and utterly unsuccessful with that.

One interesting thing that I figured out this week happened when I found an old photo album filled with pictures of Quince as a baby (birth to about six months). I decided to try and get Quince to talk about stuff in these pictures, because he's really tired of the ones we are using now and actually so am I so I wanted to change things up. He got really excited and was able to focus on the new pictures for a lot longer, probably about ten minutes straight. When I asked him who the infant in the pictures was, he said 'baby' (actually, he pronounces this more buh-buh, or sometimes buh-bee if it's especially good), and he also could identify 'Daddy', 'Casey', 'Mommy', and 'Kiele'. Elke and Mary have changed so much in the past five years that he couldn't recognize them. Clearly, he is getting bored with being quizzed on the same things over and over, and actually seems to be doing better when I switch things around and introduce new stuff.

On a completely separate note, Rachel gave me permission to add this picture to my post, since I haven't had any for the past couple. It's not really related but I guess it adds color or something (see Rachel's blog at rachel-lincoln.blogspot.com for more color-related things).

Article Link

Since a couple people seem to be interested, here is a link to the article I mentioned in my previous post. For everyone who is too lazy to open it, I'll summarize it.

https://drive.google.com/open?id=0BzYV_fxDrzE_aXRmTUtTQVQ3WUE

Gibson, the first author, observed her son (who had language delays) from when he was about two and a half to about age four. She kept a detailed log of all the words he comprehended, and later, all the words he spoke or signed. Along with Ingram, she analyzed this data, and found that even though the onset of language comprehension and production was significantly delayed (Graeme, the subject, was nearly three when he understood his 50th word, while in a study by Benedict of developing children, the mean age at which the 50th word was understood was about one), the rate of acquisition once certain milestones were hit were relatively normal. Specifically, Graeme underwent "word spurts" similar to those experienced by most children that were merely occurred later than usual.

Dr. Ingram informed me that today, Graeme would most likely be diagnosed with autism spectrum disorder, just like Quince. This paper was published in 1983, when far fewer children were diagnosed with the disorder. However, I don't believe that Graeme's process of language acquisition is very similar to Quince's, which demonstrates the diversity within the autism community. Quince's comprehension is far more advanced than his production, and his production rate is far slower and more steady than Graeme's spurts of progress. This is probably because Quince doesn't struggle with the concept of producing words to communicate and label objects, but rather the physical production of the words themselves, especially in muscle coordination.

Even though the article isn't particularly applicable to Quince specifically, it was still interesting to read and very informative. Right now, I am working my way through Gibson's dissertation, which is over 400 pages of analysis, background, and further research centering on the same diary study analyzed by the previously-described article. This dissertation is even more interesting than the paper; written almost thirty years later, it is significantly more recent and relevant to the field.