 All right everyone we're getting back to it and we're starting one of my favorite units where we talk about single subjects designs. Now when I say single subject designs what I mean is that these are the types of designs where typically we have few participants. Often we're talking about design where people are their own control group but that that name single subject can be a little bit misleading. I'm going to show you some designs where actually we're looking at the the behavior of multiple participants. So sometimes behavior analysts call single subject designs time series designs which perhaps is a little bit more accurate because we're looking at the behavior of a or a few participants over time. So when we're talking about experimental control let's go back for just a second. When we're talking about experimental control remember that's our confidence in the cause-effect relationship of our intervention that we we are confident that our intervention and only our intervention not some extraneous variable not some third variable was the cause of that change that we saw in the dependent variable. Another way of saying this is that the behavior changed when and only when the independent variable was put into place and I want to say that one more time because this is very important. Experimental control essentially means the behavior changes when and only when the independent variable is put into place. In this particular course we're going to talk about three major types of designs. The three that are covered by Miller are comparison designs sometimes known as AB designs, a reversal design or an ABAB design and a multiple baseline design. Now I am going to introduce you to a more advanced more complex experimental design in a moment but for the most part these are very very simple experimental designs that can show you whether or not you see a change in the dependent variable. On the very simplest level a comparison design and you'll have to forgive me here because the slides do get a little bit messed up we do lose some numbers on the Y axis or the left axis the one that goes up and down a comparison design shows you what happens to behavior in a baseline condition baseline is the condition where there is no experimental manipulation we're talking about behavior under naturalistic conditions so we see behavior before and after an intervention is put into place. So imagine that we're looking at the behavior here of this client and we want to see what the effect of the intervention is so maybe Janet wants to see if Johnny is going to read more books when she gives them a small reward every time he reads a new book. In this particular case we see that in the baseline condition the number of books read across time was probably one or excuse me zero then one in the next observation two in the third observation one one and then zero so we saw an increase and then it came back down that's the baseline condition and then in the reward condition we saw and I'm guessing on these numbers here six seven eight seven eight seven so an increase pretty powerful demonstration of the number of books being read increased is this change due to this reward or is it due to something else now in a comparison design it's very difficult to tell and a comparison design of all of the single subjects design is a very weak design the problem is that it shows you what happened before and after treatment and it can be really good because you can see what happens as a result of the treatment but remember that we need to demonstrate that the behavior changes when the treatment is put into place and only when the treatment is in place. A comparison design cannot rule out count funding variables because a comparison design cannot account for the fact that there could have been another variable that happened right at the start of treatment that we may not know about so for instance story time I am a woman of a certain age and when I was growing up when I was coming up through elementary school there was a program called book it right where if you read these little books and you kept track of all of your books then you could go in and you could trade in your book it menu you your teacher would sign off that you read your books I could trade that in and get a small pan pizza from Pizza Hut and God did I love that program because I like to read anyway but oh you read and then you get pizzas it's the best day ever in this particular case what if what happened was someone was trying to reward that client and at the same time that they decided they're going to do a reward program their teacher at school also said we're going to do the book it program that's the problem with comparison designs there's some weaknesses here we know that we saw the behavior increase but we cannot say that our intervention and only our intervention was the cause of that change because there's this thing called a time difference something else could have happened at the same time that could account for why that client was reading more books could be that this particular client met some people who likes reading and so they get together they hang out their friends they read books could be that they found a series of books that they actually enjoy maybe this person uh came up with like maybe they found the harry potter books that's a lot of books of course but you get the idea or maybe there was some other reward or some other variable in place so comparison designs they're really fantastic they're very promising and you're very confident there but a comparison design alone in a b design cannot rule out other variables time coincidences that could account for why that behavior increases if you want some experimental control if you want to demonstrate that your behavior changes when and only when treatment is put into place you need something like a reversal design or a multiple baseline design so with reversal designs you strengthen your confidence that your treatment and only your treatment caused that change because unlike the comparison design you actually take the treatment away so you have baseline naturalistic condition you have a treatment condition where your intervention is in place and then you remove treatment to see if the behavior will go back to baseline levels if when you take the treatment away you see that the behavior returns to baseline levels that's very compelling evidence that your intervention and only your intervention is the cause of a change so let me show you here's a comparison design so same again we've got books read zero one two one one zero then we put in the reward and we see six seven eight seven eight seven so pretty big pretty big increase there's no experimental control in this comparison design anything could have happened when that reward started but instead if we take away the treatment and we go back to a no reward baseline condition and we see here that the number of books read goes from seven or eight down to two three two one one we see it was low got high got low again that behavior changed or the number of books read changed when treatment was in place and only when treatment was in place because we see the treatment was removed and it went back down if i wanted more confidence i could put in another treatment condition so what i'm demonstrating here low data in a baseline condition high data in a treatment condition low data in a baseline condition high data and treatment condition low data again in a baseline condition and i could just keep doing this forever put it in take it out put it in take it out and see if there's a corresponding change in the behavior the whole reason that you do this kind of hokey pokey of experimental control is because you want to demonstrate that that behavior is changing when and only when treatment is put into place and again you need to demonstrate that the behavior is changing when treatment is put into place and it's changing only when that treatment is in place and you can get that kind of experimental control with a reversal design we also see in this reversal design remember we see the behavior before treatment we see what happens when we put the treatment into place and then we see the behavior change again when we remove the treatment this is fantastic because we're ruling out individual differences we're ruling out time differences so an individual difference is like a maturation difference maturation differences can be things like maybe a client learns to walk right and if you have a comparison design they're crawling they're crawling they're crawling and then suddenly they can walk if that's the same time as the start of your treatment you've got an individual difference there time differences are things that happen at the same time and external confounding variables just happen to be added because we can rule out both of those kinds of differences these types of designs time series designs like reversal design and i'm going to come back to multiple baseline design have very high internal validity internal validity is the extent to which we're confident that our treatment and only our treatment cause the behavior change unfortunately these designs don't have very good external validity which means that we can't necessarily generalize our results to other clients and just a quick note here when i talk about language here this is important the reversal in reversal designs refers to the fact that there's this third condition where we go back to the baseline condition so we're reversing the environment to baseline right we are reversing the environment to baseline conditions and i say that because sometimes when you take that treatment out it's possible the behavior maintains because it came into contact with something in the environment that's maintaining it that we don't control so there's no guarantee that when you take that treatment away that behavior is going to go back to baseline levels which is why the reversal has nothing to do with the behavior of the client the reversal has everything to do with what's going on in the environment the fact that behavior may not reverse to baseline levels is also the big weakness of a reversal design in a reversal design it's possible that some behaviors you either can't or won't be able to change so for instance if i teach someone social skills those social skills allow them to chat people up to make friends to hang out with people it's very unlikely that anything that i do to teach those social skills if i take it away they're going to keep using those social skills so the reversal design is really good for behaviors where it's very contingency dependent it's very controlled by the environment itself but these can be terrible your experimental control can be completely shot if the behavior is what we call irreversible which means it could maintain on its own or if it's something that you can't ethically remove like if i'm teaching someone safe needle sharing practices and i put treatment in i can't take away the safe needle sharing practices i can't give them dirty needles again to see if they're going to use dirty needles that's what a monster would do so in those circumstances when you cannot reverse you may want to use something like a time or excuse me like a multiple baseline design a multiple baseline design remember is is for when we have irreversible behaviors like you can't unlearn how to do something when the environment may be maintaining that behavior like in the social skills or if when removing the intervention would do harm to the participant and only a monster would do that the reason that we call this a multiple baseline design is because we have more than one baseline we're tracking multiple participants multiple behaviors or even multiple settings in a multiple baseline design so i'm showing you on screen here a demonstration that's a little bit woolly of three different participants on the y-axis the left axis we're tracking whatever the dependent variable is and we're tracking it across time and generally speaking we have low values in the baseline conditions and then as treatment is added and treatment is added at different times we see that when and only when that treatment is added do we see an increase in the response so let's take a second here in pause remember we want experimental control that means behavior changes when and only when the treatment is added so let me show you what that looks like here and i'm going to just do a little bit of drawing on screen and try to walk you through remember in the baseline conditions we have kind of low data and what we're seeing here is that when treatment goes into effect indicated here on screen after about five conditions or five observations excuse me for the first participant there's an automatic just a boop it just pops right up at about triples in the frequency of the behavior now when treatment started here for this participant we see an increase but remember we need to see when and only when so we see that treatment changes when or behavior changes when treatment went into effect here but we have to at the same time see that the behavior for this subsequent or other baselines remains relatively unchanged right so we see that behavior changes when treatment goes into effect but for the other baselines where treatment has not started we have to see that it remains unchanged and that's how we demonstrate experimental control behavior changes when treatment starts for the first participant but it remains unchanged for participants who haven't yet gone into treatment let me give you an example of what is not a multiple baseline design so this is an example of something that looks misleadingly like a multiple baseline but isn't a multiple baseline design has to have many critical features one you have to have multiple behaviors multiple settings multiple dependent variables that you're tracking but the second critical feature is that you have to stagger the start of treatment for one baseline and not the other and remember in a multiple baseline design what we want to see is that the behavior changes when and only when the treatment goes into effect so if you don't have a period of time where the baseline for one receives treatment but the baseline for the next hasn't yet started treatment if you can't see that it changed when and only when treatment begins you don't really have a multiple baseline design this is best described as two comparison designs because we don't have the staggered start of treatment so word of warning remember you have to see that treatment starts when and only when as i'm showing you here on screen we're seeing that it changes when treatment goes into effect and only when treatment goes into effect if you have something like this if you have something like this the treatment doesn't start at different times for each of the subsequent baselines and this is really just a trap remember multiple behaviors multiple people multiple settings and you have to stagger the start of treatment by at least one of those baselines by three observations so that we have time to see that the behavior changes when and only when treatment goes into effect why that magic number three well it's because three points make a trend and we want to demonstrate that there was a trend that the behavior remained unchanged let me know if you guys have any questions this stuff is pretty complex so stick around i'm going to show you one more advanced single subject design one of my favorites i'll see you guys next time