In Part 1 of this blog series on Rethinking Data-Collection in Speech Therapy I tried to hammer home the ideas that we can collect data effectively without letting it control our sessions.
And in Part 2 of this series, I talked about how one way we can do that is by collecting data intermittently, where we designate a few sessions in a year to focus on data collection, and during the others – we set the data sheets aside and focus on therapy.
And you may be thinking…OKAY…that kind of makes sense…but I live in the REAL world where I HAVE to collect data on a regular basis. Like, every. single. session., to be exact.
I get it. You’ve got administrators, bosses, insurance companies, the state, parents, IEPS, teachers, progress reports, and a million other reasons why you feel like you need a TON of data. You may even love collecting data yourself *gasp*.
A child walks into your room chatting to their friend, completely oblivious to why they’re there (at least without a reminder). If you were to count the number of correct R productions, it would be a big, fat zero.
You hang your head for a minute, get their attention and begin with your session.
“Remember Anton, we’re working on your R at the beginning of words. Can you remind me what you know about where your tongue should be for the R sound? Yup, that’s right. Now repeat after me and say these words using your good R sound.”
And let’s say he’s 30% accurate.
“Great job! (feedback feedback feedback) Now try this word…(feedback), and this one (feedback)…” You get the idea.
So now, if we collect data, because we’re good at our jobs, it’s more like 50%, right?
And we’re going to continue to use our therapy techniques over the course of our session, until (in many cases) by the end the child is pretty accurate. Maybe not always, but a lot of the time.
I’m always trying to find that spot where a child can practice their sound pretty accurately so they get lots of CORRECT practice. And that means my levels of input may change, but I really do want most of my students to be practicing their sound CORRECTLY a large majority of the time.
So if I continued to collect data over the whole session, the child looks like they’re 75% accurate by the end of the session.
Do you see the problem?
The percentage of the data I now have is more representative of how LONG I collected data for than it is of their skill level. If I’d have collected only on the first 10 reps, I’d have one number, but if I collected over 100, I’d have a different one. Neither are “inaccurate”,” but which one is “right?”
The more data I collect, the higher the percentage will be and the less sensitive it would be to change.
Not to mention, a difference in data from one session to the next may have more to do with how long we sat there than anything else.
10 solid data points collected under the same conditions > 100 data points collected under a variety of conditions.
Therapy is dynamic. And data doesn’t play well with changing conditions.
kiwi speech
In therapy we use a ton of different techniques to elicit the sounds we want to hear, provide support for the language we want a child to use or understand, and these are not always exactly the same from day to day.
So when we collect data daily, as we target skills in a variety of ways, our data often represents what we as the SLPs are doing rather than what our students are doing. And that’s not the point.
Okay, I get it. So what do I do?
You collect daily data on reduced trials.
Or, as my friend Mandi over at Panda Speech calls it – data sampling.
I’ve also heard it called data probes.
All of these names work.
You collect data on just a few trials under controlled conditions at the beginning of your session and then move on with your day.
It looks like this:
A child walks in, and I have our activity or stimuli set up.
If necessary, I explain our activity.
Before getting into teaching, prompting, models, or anything related to their target skill (let’s say it’s artic*), I have them say the words from the activity. *If this were an expressive language activity I’d find a semi-structured way to probe the target structures at the highest level of independence possible. What matters most is that you’re consistent from session to session and that you define the conditions in your notes. I don’t give feedback, I don’t correct, and I don’t give hints. I just listen and take data. I do this for 10-ish responses, then I put my data sheet aside and get to work.
Then, I start diving in there to elicit the best responses possible. I give feedback, I give models, and I do whatever it takes to get that student to practice their skills and improve. And yes, by the end of the session, they’re probably better at it. That’s literally my job. But what I’m really interested in is how much of what they learned sticks. And the cool thing is – I’ll find out in the data I collect in the next session.
One Comment
Comments are closed.