 Hey everyone, welcome back. It's Professor Howard. So in this topic, topic number four, we're going to be talking about the methods of observation. And I have provided some links to psych core videos for you to watch. The psych core guys are great. They summarize the content really, really quickly and cleanly. But I want to make sure that we spend a little bit of time expanding on these methods of observation. So definitely start by watching their videos because it's going to give you a brief introduction. And then come on back where we're going to break this down. And I'm going to give you more detailed in depth information about these methods of observation. So just as a quick review, since of course you've already watched those psych core videos, remember that in behavior analysis, in Miller's textbook, at least, we're talking about four methods of observation. And we're going to break out one of these a little bit so that we're talking about more than four, we're going to actually talk about five. The four that I'm talking about here, of course, are outcome recording where you're looking at a tangible, physical result of the behavior once the behavior is done. We're talking about event recording or discrete instances of behavior as they happen. So an instance of behavior is just whether or not it happened as it happened. And this doesn't work as well for behavior that's less discrete, something that can last for a long time or a little bit of time and it doesn't look the same or exactly the same every single time. Event recording is really good for stuff like it's got a clear on and off point. Like you can tell in this video how many times I do a tongue pop, like I'm known for this kind of noise that I make when I'm recording videos. And you could go back through and actually count the number of times that I tongue pop in this video, as opposed to in this video, how long am I talking for? And I'm intentionally stopping to talk here so that you can see that there are going to be some times that I'm actually talking, talk, talk, talk, talk, talk. And there's going to be some times in this video when I'm not talking or when we have brief moments of silence. That kind of talking behavior is less discrete. For those less discrete behaviors, we typically use time-based measurement that can be interval recording. Now Miller just describes this as interval recording. We're going to break it down further into whole interval and partial interval. Interval recording means that you're just observing within those intervals to see whether or not behavior is happening, whether or not I'm talking in this video. And time-sample recording is when you pre-designate a period of time to actually check for that behavior. It can be very consistent or not, which I'll get into in a little bit, but you're, again, looking in discontinuous intervals, you're not actually watching the whole time and you're just checking to see whether or not that discontinuous or non-discreet behavior is happening. So let's go through these in more detail. I'm going to begin with outcome recording. As the site core fellow said, this is hands down one of the easiest methods of observation to observe if you have a physical tangible product. So for instance, if you are going into observe, you just have to make sure that you're observing after the behavior actually occurs and you're looking for physically some product that has been produced by whether or not the behavior happened. So for instance, you might in the wintertime check your child's elbows to see if they're kind of ashy. And if they're really dry and ashy, that would be evidence that they had not, in fact, used any kind of lotion when they were finished getting ready. Or you might go into my kitchen and see whether or not my dishwasher has been loaded. The outcome of me loading the dishwasher is the dishes are actually in the dishwasher. So you're not actually seeing the client do the behavior, you're looking for a product of the behavior. There are some pros and cons to this. So like we said, super easiest method of observation hands down very, very easy to use. But it's not appropriate for every situation. For instance, we know that outcome recording can be really bad if you're doing something like trying to observe someone's private events. Because for the most part, having a private event like feeling depressed one day doesn't actually leave any permanent product. You might look for other evidence like maybe there are a lot of tissues in the trash, because the person cried a lot or maybe you're looking to see whether or not they're wearing the same clothes they did the day before. But for the most part, you can't actually see and observe that private state that the person was experiencing on that day. So first of all, that's challenging. Second of all, you can't use methods of observation like outcome recording where you have to go and find the physical tangible product if the results are in a location that you cannot get to. So for instance, we know that people have walked on the moon. We have some video evidence of people walking on the moon. But you and I could never go there unless Elon Musk gets his way and we actually can go to outer space because it's so cost prohibitive for us to get there. And there are going to be some times when you're working with a client that you cannot get to the physical location where the behavior happens. Maybe you're doing safe sex education and you're looking to see whether or not your client is having safe sex, whether they're using a condom. You can't necessarily go to the client's home and like dig through their trash to look for evidence of condom use. It could be that it's really difficult for you to get to wherever it is. Maybe it's in a different country or in a different planet, or maybe you're just not allowed to be in the place where the behavior happens. So outcome recording is really not good for that approach to observation. Like this is really very difficult to use when the results are in a place that you can't get to and you can't measure. The third is that sometimes we can capture results of other behavior. So what I have here on screen is a picture of someone washing windows. Imagine I own a business and I want to make sure the windows are clean and I have two employees, Jeff and John, and it's Jeff's job to wash the windows. But they'll both get in trouble if they're not clean. So even though it's Jeff's job to wash the windows, John, seeing that Jeff isn't doing it, washes the windows. So now I've got clean windows, which is great. That's what I want as a business owner. But it was Jeff's job to do it. And someone else did it. And so we have this kind of confounding variable there, where what I could be looking at is not the product of the behavior that I wanted, which was Jeff washing the windows. Rather, it's a result of an alternative response, someone else doing the behavior. And there are disadvantages to that. If I create a contingency for Jeff, where Jeff gets a bonus for doing all of his job tasks, but really it's John doing the work, I can't reinforce Jeff properly. We'll come back to confounds and other challenges later in the course. And then finally, sometimes if behavior and their results don't last long enough, we can't get to them. So for instance, we know that in prehistoric times that there probably were quite a lot of cave painters. There were a lot of people creating art, but we don't have records of a lot of that because it's been destroyed or ravaged. It's been lost through time. Same again, if you have a client that you're asking to do a particular task, clean their room, but what if they share a room with a younger brother or sister and they clean the room perfectly, but then the younger brother and sister comes in and destroys it, makes a huge mess. And so the product, the cleanliness, didn't last long enough for you to capture it. Those are some disadvantages of outcome recording. When you can use it, it's really great, but there's a lot of reasons why you might not be able to. Turning to event recording, remember that this is just counting those discrete instances of behavior as they happen. So what I'm going to show you on screen here is this is the behavior as it's happening. These little check marks is when the behavior is occurring. Maybe this is me raising my hand in a lecture class. The method has to be used for behaviors that are discrete and uniform in length, so they have to kind of look the same every single time. Remember my tongue pop behavior sounds the same. It's roughly the same length and it looks the same every time. Raising one's hand relatively discrete and it looks the same every time. And what I'm doing here is just count the behavior as it happens. What I'm showing you on screen is one, two, three, four, five, six, seven, eight, nine, ten, eleven check marks. So in this session I would say eleven instances of the behavior happened. When you're recording from session to session, of course you just want to make sure that your sessions are the same length, so you don't have that confound of like one session was super long and one session was super short, so it looks as if the date is a little funny, but in this case we're just counting the behaviors that have a clear on and off point as they happen. Eleven total events occurred here. I might do this in a classroom. Maybe I'm interested in seeing how many of the students in this class speak at any point. Speaking or opening one's mouth is relatively discrete. I'm really just counting the number of students who are doing it, not how much or how little or maybe the number of times they raise their hands in a classroom. Now there's a number of reasons why this particular method of observation could be inaccurate. In this case we know that behavior that occurs at a super, super high rate. Imagine we've got all these kids in the classroom. There's 15 or 20 kids in the classroom and then suddenly three students at the same time start talking. Well as an observer I'm just human and I may lose that. I may not accurately capture what's going on there. We can also have observer error when we have really long sessions, so if I have one observer like you see here on screen sitting in a classroom with students for eight hours in a day of course they're going to start to make some mistakes. So if you're using event recording there are some some barriers there to this method and it's especially prominent when you have a super high rate behavior. Now switching to interval recording, remember interval recording is we're watching in periods of time to see whether or not a behavior is happening. An interval or time-based measurement systems are really really good to use when the behavior is non-uniform. In the video so far I've not actually said the same thing again and again and again and again. If you were to if we were to make a drinking game out of this of course and you were to have a a sip of coffee every time I use the word data right we'll do that in a future video I'm sure you'll love it. But in a future video if I were to say count the number of times I say the word data that's discrete data the word itself is really really discrete on and off. But observing whether or not I'm speaking during the video is less uniform because I could be saying things like data or interval recording or method of observation or and none of those terms are exactly alike and so it's much harder to account for those. So when we're using a non-uniform behavior you want to try to use the the method of observation that's going to work best and those tend to be time based so we're talking about the interval whole and partial interval and we're talking about time sampling and you're really just seeing at the point when we're observing did the behavior happen or was the behavior not happening right a perfect example of this is when kids have a tantrum. Tantrum and kids are the best because you know sometimes it can be like and it's just kind of an outburst and other times it can go on for minutes or if they've got really good constitution like if it's a kid who's well-practiced they could tantrum for a half an hour and if we're doing a session there it's going to look differently maybe that tantrum was like 30 seconds and that counts as a tantrum session versus the kid who's having the tantrum for 15, 20, 30 minutes that's a very different tantrum than the first one but if we used event recording they would look exactly the same. So event recording the way that you use this method of observation is you take a session length and you break it down into smaller intervals. So notice that this is the session and I'm breaking it down what I'm showing you on screen is that unit of time is being broken into little bins of time and I'm going to see whether or not in that first period of time did the behavior happen or not in the second period did the behavior happen or not if it happens put a check mark if it doesn't happen leave it blank so on and so forth. So I'm going to continue going here and this particular client that I was observing engaged in the behavior during session one or excuse me during interval one interval three, interval seven, interval eleven and interval 13 of 14 intervals. So then I can break that down and I see that they're they're engaging in the target behavior during five of the intervals and I could even break that down even further you know five of 15 intervals here we're looking at a total duration of the behavior of about 33 percent. Does that make sense? Okay so interval recording there are two more specific kinds of interval recording partial interval and whole interval recording and like they described in the video uh the site core videos the difference between partial interval and whole interval recording is for how long the behavior was happening within each one of those little sections right within each one of those intervals so if I say partial interval recording what I'm creating is a behavioral definition that says if the target behavior happened at all even for the tiniest little sliver of a section if it happened at all score it as if it happened during that interval regardless of the length of the interval right so in this particular case these 15 intervals and I'm showing you on screen here if the behavior happened even briefly like at all just a tiny bit then you you score it as if it happened whereas with whole interval recording you can only count that the behavior happened if it lasted for the whole interval like the name suggests so if we're taking this session and breaking it down into 15 intervals that behavior would have had to have been happening at the start of the interval and it would have to have continued through until the end of the interval for you to count that yes it happened and as you might expect you you result you get a very different result in terms of how often the behavior is happening and how accurate that is to the true measure of the behavior based on whether you're using whole interval or partial interval recording let me tell you some of the inaccuracies with interval recording and I'll tell you more about our measurement artifacts in a little bit okay so returning to interval recording I want to talk now about the difference between partial interval recording with an example on screen what I'm showing you here are 15 sessions or 15 sections excuse me intervals within a session now remember that with partial interval recording if the behavior happens at all within that interval you record it as occurring so what I'm showing you on screen here these green bars that occur underneath each of the intervals tell you the true measure the true duration of behavior so this first one that I'm showing you starts begins partway through interval two and extends through interval three then I record the behavior actually happened during those two intervals the next is in interval four so I score that it happened during interval four even though it's happening for only a small portion then the next starts interval eight through eleven I record the behavior is happening during those four intervals and then finally we've got a couple in interval 13 14 and 15 so in total we have behavior occurring during one two three four five six seven eight nine ten intervals or approximately 66% of these intervals even though if you were to have used a true measure of the behavior meaning that I start the stopwatch when the behavior begins and I end the stopwatch when the behavior ends we would be looking at a true total duration of less than 50% for this behavior so we see the partial interval recording over estimates behavior but we use this because it's actually easier to use partial interval recording than it is to measure the true value of the behavior excuse my cat she's got a little sneeze we measure this because it's easier than the true value the true duration start and stop time in much easier to use in clinical practice now I can increase the accuracy if I were to make smaller intervals I could make them as tiny as possible and just say yes it's happening during this one second and the next second and the next second and so on but again the smaller your intervals you will be more precise but it's going to be much more effortful to use this method of observation when we are deciding on which method of observation to use we always have to kind of balance the ease of use or how practical it is for us to use this in clinical practice versus the accuracy of the data that we're going to get now partial interval recording can overestimate the rate of the behavior or the true the value of the behavior it can tell you it's happening more than it actually is even more than the true value what we would get if we had a start and stop time that's a measurement artifact whole interval recording can also underestimate the rate of the behavior so again same 15 intervals remember with whole interval recording I only count those behaviors happening if it began prior to the start of the interval and continued throughout the interval to end after the interval is finished same 15 intervals on screen here remember that my green bar shows that it started part way through two and extended through three but it did not actually fill any one of those intervals so I don't score that behavior is happening in the next we have a little bit in session five but not the full interval and then finally in let's see one two three four five six seven eight in nine and ten the behavior started in part way through interval eight and extends through interval nine and interval ten and then into eleven but it only fully encompassed intervals nine and ten and then even with the small little bits of behavior that we have happening in session or in excuse me intervals fourteen and fifteen and twelve they're very brief so the behavior is not scored as happening because it did not begin before and end after the interval was finished so in this case you see that when you compare it to before using whole interval recording we would record that the behavior only happened during two of fifteen intervals or about 13.3 percent of the time remember that the true value of this behavior the true duration was around or a little bit under 50 percent so this is a very different result that we're getting with whole interval recording versus partial interval recording. You want to use the method of observation that's going to be best for the behavior and we can see that for some behaviors like these ones that are relatively brief in duration it may not be ideal to use whole interval recording if it's a brief behavior because you're never going to capture. Moving on to another method of observation that has a lot of flaws time sample recording is really really really good when you maybe have to observe multiple clients or for whatever reason you cannot give your full attention to a single client during the observation. Time sample recording is observing in discontinuous intervals so we're going to watch for a little bit and then we're going to look away our attention is going to go somewhere else and during that little period of time that discontinuous observation we see whether or not the behavior occurs. Again this just like with the tantrums and just like with time sampling or excuse me interval and partial and whole interval recording we want to use this when behavior is non-uniform so it doesn't have a clear on and off point. If behavior is uniform use a different method of observation use event recording but you can do time sampling in one of two ways. You can either record consistently and set a pattern so for instance maybe you observe the first 15 seconds of each minute to see whether or not the behavior occurs. What I'm showing you on here are those same 15 intervals from before but in this case all I'm going to do is observe for the first 15 seconds. The other 45 seconds of each minute I'm not observing the client so I only score the behavior if it occurs during that first 15 seconds. If the behavior occurred for instance right here I would score it if it occurs during that little window when I'm going to watch the client but if it occurs at any other point if it occurs from second 16 through second 59 do not score the behaviors happening. Again score it if it occurs during that little interval when you're going to observe but do not score it if it occurs outside of your observation window. So time sampling can be really really good when you have lots of clients to observe when you have to switch maybe from client one to client two to client three to client four it can be very good to use when you have limited resources so maybe you can't dedicate a lot of time to just watching a single client you could set up a time when you're going to say maybe I'll check in on my learner every five minutes or so and it can be very good when for instance you don't really want the person who also has to record behavior to spend all of their time looking at the client for instance this is very often used with teachers because teachers they have a job to do right they have to teach in the classroom we can't make them our data observers because they simply don't have the time and resources to do that. Another way that you can do time sample recording is rather than being very consistent and saying you're going to observe for the first bit of time out of a larger interval maybe you're a little bit more random this could be more like a parent checking in on child to see if they're doing their homework the parent doesn't necessarily have a timer that says you know what every three minutes I'm going to go see if junior is working on math maybe they come in every four minutes and then eight minutes and then two minutes and so on and so forth but they're not necessarily scheduled another way that this can another way that time sampling excuse me occurs is like I said when you're alternating between clients so what I'm showing you on screen here is a colored display of switching between different clients we've got a larger session that we're breaking up into time blocks or intervals and maybe I observe client one and then switch to client two then go back to client one then switch to client two I as the observer I'm watching for the whole time but from the perspective of the person being observed it's a discontinuous observation because I'm only actually observing client number one during intervals one and three and five and so on so it's actually a discontinuous observation of each client's behavior even though I the observer in watching them the whole time when you're trying to determine this here if you see switching between clients you're very likely looking at time sample recording now the biggest limitation of time sample recording is you are absolutely going to miss instances of behavior so what I'm showing you on here this one is definitely going to underestimate behavior because keep in mind that there are going to be periods of time when the behavior is occurring but you're just not seeing it because you're not observing what I'm showing you on screen again is that accurate or that representation of 15 intervals with the inner time 16 seconds through 59 seconds blocked out in some of these instances we see that the true measure of the behavior was quite a bit longer but we would actually only capture of these one two three four five six kind of spells of behavior we would only capture the behavior one two three four times or so because if it happens during that blackout period when we're not looking we don't count it is actually happening time sample recording is another one that underestimates the true value of behavior just like whole interval recording and then wrapping up here that we've talked a lot about the different methods of observation but what I want you to remember is we want to try to capture the most accurate representation of data and flawed observations can occur for lots of different reasons if we choose a wrong method of observation or a very insensitive method of observation for instance if I choose whole interval recording when the behavior is actually really brief in true duration then I'm going to miss a lot of it so picking a bad method of observation for the target behavior is one of the reasons why behavior may be inaccurate or the observations may not be accurate another reason that behavior or the observations that we have can be flawed is because we didn't measure at the right time for instance if I want to check eating behavior in a client that I need to increase or change the rate of their behavior I shouldn't look at them when they're in math class I should probably be looking at their behavior when they're in lunch so it does require some work to understand the client where the client's coming from to make sure that you're actually measuring the behavior at a time and place when it would likely happen this also related to that can explain why sometimes you have two different people in a learner's life that report very different instances or very different like characteristics of that person so maybe the person is really off task in math class but they're they're a rock star in English class and you get those two teachers together and they have no idea why the other person would think that about the learner because that's not what they've seen keep in mind that there's those are very different contacts you might occasion different kinds of behavior from the learner last you can have flawed observations just because the observations were really flawed meaning that if you're choosing learner if you're choosing observers that are not well trained they could be using the behavioral definition poorly maybe the behavioral definition is super flawed maybe the environment in which they're observing is chaotic and it's very difficult to see what the learner is actually doing so your observations your data are really flawed and there's nothing worse than having flawed data like if you take nothing else away data is the most important thing that you are going to need to understand whether or not the behavior is happening so you really really have to invest heavily in understanding whether your observations were valid and whether they were appropriate and you need to fix that method of observation before you even consider moving forward and actually doing an intervention if you have flawed data you're really just wasting your time so get the data right get a strong method of observation choose the right one for the type of behavior happening measure at the right time have a good behavioral definition and train people to observe behavior before you just set them loose to observe behavior because if you come in with flawed data and you don't know if what you did was effective you just wasted time and resources and potentially expose someone to a really ineffective form of treatment that they didn't need to be exposed to and then one last thing bear in mind that again we've got to compare and contrast the values of the method of observation I wish that it was as easy as saying oh just use this one method every time but no method of observation is perfect there's no single method of observation that's going to be best for all behaviors you have to understand that every method of observation has strengths and weaknesses so you want to be comparing how easy is this to use to how accurate is the data really really accurate data is fantastic but not if it's going to be data that no reasonable person can collect because nobody will collect it and vice versa you can have really easy method of methods of observation but if the data that you're collecting is trash and is not an accurate representation of the behavior that you're interested in then that's just as bad try to find that sweet spot of ease and accuracy in your data so that you can be confident in what you're concluding I look forward to hearing if you guys have any questions otherwise make sure you check out the site core videos and I'll see you guys next time