Published on May 18th, 20170
Q&A with the Winner of the 2017 Editor’s Award
The annual SAC Editor’s Award recognizes the best article published in the Canadian Journal of Speech-Language Pathology and Audiology each year. The 2017 recipient of the award is Elin Thordardottir, PhD, for her paper “Long Versus Short Language Samples: A Clinical Procedure for French Language Assessment.”
Dr. Thordardottir was kind enough to answer a few questions about what led her to write the paper and the clinical applications of her research:
How did you become interested in speech-language pathology?
I was studying Icelandic and linguistics at the University of Iceland, but I didn’t think I would find a job in linguistics. So I looked for another way to do something language-related that would lead to a job. I came across [speech-language pathology] and I thought it would be a good fit because it’s related to linguistics but it does come with a professional certification and it does have a specific job attached to it so I thought I’d have better job prospects.
What led you into research?
When I started studying speech-language pathology, I got interested in audiology. I took courses in audiology and I found that really cool – I also have an MS in audiology and that is what I thought I’d do. I thought I’d be a clinical audiologist. But I got into writing my Master’s thesis, which involved the development of the MacArthur-Bates CDI in Icelandic. It kind of went from there. I was told by some of my teachers that I’d be good at research and why wouldn’t I just do a PhD? I don’t think I really thought about that before, but that’s kind of how that came to be. I guess I really like that aspect of the field.
How did you come to focus on the topic of language sample length, with an emphasis on sampling in French?
I studied speech-language pathology at the University of Wisconsin, which is where the SALT software was developed. This software was designed to be used for language sample analysis for clinical purposes and it has norms in it. I got very good at language sampling and I also came to understand very deeply how you do it and how you use it. I can’t imagine doing clinical work without it.
When I came to Montreal, I started doing a lot of research in French. If you really want to evaluate language, spontaneous language sampling is an obligatory part of that evaluation. I’ve shown in research on French language sampling that it gives results that do not correlate with other measures, so it says something unique, it gives a different perspective. I think that you have to have this as part of your evaluation battery. Because we didn’t have norms or a systematic way of analyzing language samples in French, or in Quebec French, I started developing such a procedure a long time ago just because it’s a tool that I have to have for my research.
Language sampling is something that is very hard to do clinically. If you take a fairly long sample, transcribe and analyze the whole thing, it really takes a lot of time. At the same time, why would I present that if nobody is going to use it? So I wanted to make it more user-friendly and to make the procedure more clinically feasible.
In this study, I was looking at how short you can make a language sample and still get a lot of benefit from it. My idea was: I’ll cut my sample in two, and I’ll cut it again and again and I’ll see what’s the shortest we can go while still maintaining accurate results. What I present in the article is a shortcut procedure that lets you collect a short sample and analyze it in a simple way, but then use it to derive much more complex information based on the data presented.
How did you go about conducting this research? Did you run into any unforeseen challenges?
Well, when I first started to adapt language sampling techniques to French, I didn’t just take what they do in English and do the same thing in French. I took the general idea. And I adapted that, taking into account the structure of French. The coding of grammatical morphemes is a lot more complicated in French than it is in English because there’s more morphology in French. The procedure took years to develop because we would code things a certain way, we would test it, we would rethink it and then we would go back and recode everything to change the method. It took quite a while, there’s a lot of questions that come up. You have to decide how to make the coding complex enough that it captures relevant things, but you also have to keep it simple enough that it can be done reliably.
So it’s something that was quite a few years in the making, but for this study it was more that I thought: I have all of these language samples accumulated from studies that have all been done the same way – why don’t I pool them all together? I hired a research assistant to enter them all into one big SPSS sheet. In that process I did not run into difficulty. The difficulties had already been dealt with during the French language coding process. There were some issues in how to assign an MLU group when this assignment can change depending on which sample length is used. These are methodological questions that are discussed in the article.
What was the most surprising part of your research?
To me, what I found really spectacular when I looked at the results was how systematic the relationship between samples of different lengths is. It has been debated in the literature how long a language sample has to be. In some studies, samples of 200 or more utterances are used. In normative samples, the length is usually 100. To me, what was really surprising was just how low you can go. And to see, as reported in the graphs of this article, just how stable the measures of MLU and vocabulary diversity are. Even though you have more words in a longer sample, the relationship is just so predictable and systematic. And not even just that. For the morphology, even though in a longer sample you’re going to hear more types of different morphemes, the kinds of morphemes you hear in a shorter sample can more or less be predicted from the longer sample also because the differences are quite systematic. That I found quite surprising – how much you preserve even though you go with a short sample.
What do you see as the next step? How would you like to see this research applied?
First and foremost, I hope this simplified procedure will lead to language sampling being adopted as a routine part of the clinical assessment of children who speak French. It is an essential part of the assessment protocol, not so much for diagnosis, but more for the planning of treatment and evaluation of its outcomes. It would add so much to the quality of French language assessment if systematic language sample analysis were added to the standard protocol. I think the method I present makes that possible. But I think that this also has a much broader applicability and theoretical implications. It´s important to understand how language sample results are affected by factors such as length and context. Other researchers and I have looked at language samples across contexts such as conversation and narration, across different languages and samples produced by monolingual and bilingual children. The more we understand the impact of such factors, the better we can develop reliable methods of assessing both the input presented to children and their output.
What message do you have to other S-LPs who want to conduct research on clinical practices?
Research directed at improving clinical practice is incredibly important, it is something that we have to do to advance the field. There’s a lot of fundamental research that doesn’t really become useful to us in any immediate sense, so in order to make our clinical practice more evidence-based, we have to also research clinical practices directly. A message that I have is that this kind of research is important – I think sometimes it isn’t viewed that way, but it is. And also, when you do things that may look more clinical, there is always a theoretical implication. Clinical research really does advance theoretical questions.
I’m really happy that this article was selected because it’s the fruit of a lot of work. But I’m also happy just to see this type of work acknowledged as important.
Elin Thordardottir was born in Iceland and grew up there and in France. She studied Icelandic and linguistics at the University of Iceland, and speech-language pathology and audiology at the University of Wisconsin-Madison. Since graduating in 1998, she has been a professor at McGill University in Montreal. Concurrently, she has been an active researcher at the Academy of Reykjavik in Iceland and a part time lecturer at the University of Iceland and at the University of Alberta. Elin’s research is in the area of language development and language disorders in monolingual children speaking English, French, Icelandic and in bilingual children and second language learners in Canada and Iceland. Her work has had a strong cross-linguistic and multicultural component, focusing on the underlying nature of bilingual development and language impairment, as well addressing clinical research issues in assessment and intervention efficacy. She is the author of several language assessment measures targeting French and Icelandic, and bilingual assessment. She is frequently invited to give continuing education workshops to clinical audiences all over the world. She has also had a leading role in two European multi-country research projects (Cost Actions) focusing on bilingualism and childhood language impairment.