views Page in Progress
DISCLAIMER: The information on this site is presented on an "as is" basis, and does not claim to present definitive information, make recommendation or to represent such official bodies such as the Standard Celeration Society. Users of this site agree that the owner of the site or authors of pages and articles are indemnified against legal liability
========================================================
An awesome discussion of SAFMEDS and the analysis.
4/26/07 I asked John a few newbie questions, and he, being the consummate teacher, did not give me a quick and facile answer, but rather a detailed and elegant discussion of SAFMEDS and instruction.
Regina Claypool-Frey
John W. Eshleman, Ed.D., BCBA
4/24/07
Question--Do you check answers for correctness during the timing? Unless you already know that you know the answer correctly, I can't see how not checking would keep from practicing an error, but ,conversely, it seems to really slow down the responding if you check after every card. It's more clear when you simply don't know the answer, because then it's just a go. Vihn, in the video, stopped during the timing to check the answer that he didn't know.
I do. I check cards as I do each one. See-say-flip-see-put is the complete stream. Doing a see-say I can get up to 60 per minute that way, and a see-think I can get over 100 per minute. So, does this represent a slowing down? You tell me.
If I don't know the answer, which technically is a "skip" not an error, I say "don't know." This shortens to "d'no."
Some people team up two or three people and one person checks for the accuracy of the other person. That's good, but it requires two or more people. What if you're alone and want to practice your cards?
Again, another variable: Keep text on fronts and backs short and brief. Use 1 word or short phrase. Do not use sentences or paragraphs on eithe side.
That way, when you turn the card over, you can simply "recognize" whether your response was correct or not, without having to "read" it. Again, with cards like this, turning them over and checking does not slow one down appreciably.
I used SAFMEDS for instruction, while others seem to use them to "test" a performance; but where the performance is mysteriously acquired somewhere other than the SAFMEDS timings. Where? How? How efficiently?
If you "practice" with the set of cards ahead of a timing, where you go through the set slowly, one at a time, and flip a card back and forth, then you are doing SASMEDS -- Say All Slow Minute Every Day Shuffled. But if you do not record this SASMEDS session, then it qualifies as unrecorded learning. You will not get a true picture of learning efficiency if you do not record such SASMEDS sessions and only use SAFMEDS to "test" or "assess" performance.
The other thing that I was curious about was, how "ALL" is the A in SAFMEDS? If you have 100 nouns to learn in x time, but only 7 are introduced each week, am I correct in assuming that the ALL refers to those 7 and then the 7+ whatever is introduced in subsequent weeks? It's a dumb question, but I doubt if it's original
The problem there is "introducing" only 7 new nouns per week. That's what Carl Binder would suggest as a fluency blocker. If you have 100 nouns, then the set of SAFMEDS would consist of 100 cards from the getgo. Now, yes, that conflicts with the tactic of "introducing" only 7 new ones per week. But, what do you want? Fluency? Or only 7 per week? How many can the kid handle? 7 new nouns per week sounds like "teacher knows best" to me, rather than "the child knows best."
Step back a second.
SAFMEDS represents a STRATEGY, not simply a technique. It's a strategy regarding teaching and learning, and aligns with a behavioral philsophical view of learning that suggests that much behavior is free operant, or should be treated as free operant instead of as controlled operant. If you think that life is controlled operant, then you will be toiling down in the pits of 7 new nouns per week when the child may be capable of all 100 of them from the get go using fluency-building and instructional design.
-- John
Nouns refer to people, places, and things, right?
OK, apply some instructional design to the task. This means make use of equivalence relations and generalization as much as possible.
E.g., one set of 100 nouns would have a picture, photo, or drawing on the front and the name of the thing on the back.
Another set of 100 nouns would be a separate deck of cards, and would have the noun on the front and a very brief definition on the back. This deck would be created to be "bi-directional." That is, you could run a timing where you see noun, say definition, and then reverse the cards and run a timing that is see definition and say noun. If we were teaching household objects, see "chair" say "sit in it," see "bed" say "sleep on it." Then in the reversals, see "sit in it," say "chair," and see "sleep on it," say "bed."
You could set up a timing where the cards are laid out on a table, and do a hear-point. Hear noun, point to its picture. This is more controlled operant, but I'd include it nonetheless. Hear "table," point to picture that shows a table.
If you put this on a computer, you could do matching to sample, and put the name on the screen and have the person point and click on the picture that goes with it; you'd have 4 or 5 distractor photos that randomize position. This would be a reversal of that first set above.
You'd set up additional sets to test generalization. E.g., if you have a card with a table picture on the front and the noun is 'tab;e' on the back. Then in the next set, you'd have a picture of a different table on the front. That way you'd begin testing for generalization.
Alternatively, you'd have within the same set several examples of each of the objects. E.g., pictures of 3 different tables, 3 different chairs, 3 different sofas, 3 different vacuum cleaners, etc. This deck might run up to 300 cards (which is why it'd begin to be useful to put all this on the computer and work in that modality.).
I'd run these sessions concurrently rather than sequentially. I.e., do a timing with each set every day, rather than work on 1 set and finish it before starting the next set.
So, you do a see-say, see picture, say its name. Then right away, see word, say its brief definition. Then turn this stack over and see definition and say word. Then spread out the photo cards and hear word, point to photo. That'd be 4 timings. It would involve two sets of minimally 100 cards (200 cards in total). If you had several examples of each object in the photo cards set, then you might have closer to 400 cards total, 300 in the photos deck and 100 in the words-only deck.
The way I conduct timings, I start the timer and the timer is stopped when the kid stops. I.e., if you have 100 cards, the kid does all 100 of them and you measure the time it takes to do that. The number is always a constant. The time it takes varies.
Otherwise, from session to session the kid will not contact certain cards, and those will go unpracticed an unassessed. That introduces a lot of inefficiency. Also, the timings are the instruction, because you'd turn the cards over during the timing and check. If you do SASMEDS (slow timings separate from "test" timings, then count and chart these separately. But SASMEDS will introduce a lot of inefficiency to the process.
Errors? Study the errors right away after the timing. Time the errors instruction. How much time is spent correcting the errors? Also, you might want to study (and record and separately chart) the errors before the next timed session.
Always let the kid re-start if they start, get flustered, and quit. Re-starts are ok. This isn't a sport like a 100 meter dash. It may help to model the performance, too. And, model it more than just the one time.
You could continue with the instructional design:
A-B Condition = see photo, say noun.
B-C Condition = see noun, say definition.
A-C Condition = see photo, say definition (you would not teach this relation but test to see if it occurs).
C-B Condition = see definition, say noun.
B-A Condition = see noun, select photo (point, touch, etc.).
C-A Condition = see definition, select photo (point, touch, etc.)
D-A Condition = hear noun, select photo (point, touch, etc.)
D-C Condition = hear noun, say definition.
C-D Condition = hear definition, say noun.
Other Considerations:
B-B Condition = See noun, say noun (textual verbal behavior; "reading").
A-A Condition = See photo, select matching photo of same concept.
D-D Condition = Hear noun, say same noun back (echoic verbal behavior).
Not all of these relations would need to be directly taught. That's the benefit of stimulus equivalence. Logically, if you learn an A to B relation, and then learn a B to C relation, then you should already have a high likelihood, without being taught, of doing an A to C relation fairly well. Look up "stimulus equivalence," "equivalence relations," "conditional discriminations," and "matching-to-sample" in the behavioral literature.
With a couple hundred cards, 100 with photos on the front, and 100 with the noun on front, you could cover those above conditions.
Also, by bringing in Skinner's analysis of verbal behavior, you set up success with other repertoires. Such as reading the noun, pronouncing it, hearing it, selecting it, and so on.
If you set up a deck of a couple hundred more cards with variations on the fronts, then you minimally set up success for generalization. You would test for generalization sequentially (i.e., run through the other conditional relations listed above first; though you could test for it periodically within the course as well.)
By LOGICALLY organizing and arranging the instructional design, and taking into account a broad spectrum of human verbal behavior repertoires and what we know about generalization and equivalence, as well as fluency, you could very likely succeed with teaching those 100 nouns up front, rather than waste a considerable amount of time introducing only 7 more per week.
Moreover, if you chart the above conditions, then you will see whether or not there is learning going on. You will also soon discover the "hard" cards and the "easy" ones. So will the kid.
So, you could stack the deck with more hard cards or more easy cards, depending on the objective. If the objective is to get the kid feeling more success and encouraging him, then add some more easy cards. If you want to focus on learning the hard cards and getting the kid accurate and fluent on those, then stack the deck with more hard cards. This isn't playing cards, so stacking the deck is permitted. -- John
2/8/08
