Behavior and Its Causes: Reading #13

Copyright Sigrid Glenn, 1997; 2002.

Measurement in Science and in Practice

Measurement is critical to scientific understanding of how nature works. It is also critical to making a difference in the everyday world. It was not until the 20th century that scientists figured out a way to measure behavior so that they could begin to formulate behavioral principles that can be used to help humans make a difference in everyday behavior.

Because learning is something that goes on at the level of individual organisms, behavior analysts realized that study of behavioral processes would require studying the behavior of one organism at a time. Of course, what scientists are ultimately interested in is behavioral processes common to the learning of many different organisms of many different species - including humans. In other words, they were interested in formulating principles that applied to the behavior of all kinds of creatures. But behavioral processes operate at the level of individual organisms. In order to learn how behavioral processes operate, behavior must be studied as it changes in the repertoire of one individual.

The same thing is true of all processes that occur at the level of individual organisms. To find out how brains and cardiovascular systems work, their activities must be examined one organism at a time. As it turns out, there are many similarities in the brains of all humans and in all their cardiovascular systems. It even turns out that there are many similarities in the way brains and cardiovascular systems of all mammals - not just humans - work. This is also true of operant behavior. It works the same way in all humans and many other animals as well.

The necessity of examining the behavior of individual organisms made it hard for scientists not to focus on organisms themselves as the subject matter of interest. It took a long time for scientists to realize that the unit of analysis important to understanding behavior was not an organism but an operant. Not only did operant behavior have to be studied one organism at a time, but it also had to be studied one operant at a time in order to determine the ways in which the environment functions to build, maintain, and change operant behavior.

Once behavior analyst scientists formulated principles that describe the ways that operant behavior changes in relation to changes in its environment, those principles themselves became part of the behavioral environment of applied scientists and behavior-change practitioners. As part of their environment, the principles could affect the activities of applied scientists and practitioners, making it possible to base applications on known scientific principles of behavior for the first time in history.

Applied behavior analysts and behavior technologists or practitioners are interested in the content of particular operants. Their job is to bring about specific targeted changes in the behavioral repertoires of particular people. In order to do that, they need two kinds of knowledge. First they need to understand the principles that explain how behavior works; and second they need to define the particular behavior of interest and to identify the particular events that constitute the behavioral environment of the particular person they are trying to help.

1. Operant processes occur at the level of _______________ organisms.

2. Although behavior must be studied one organism at a time, the unit of analysis important to understanding behavior is not an organism, but a(n) _______________ .

3. Practitioners are interested in the content of _______________ operants of 4. a(n) _______________ human or other organism.

Operant Properties

Like any observable phenomena, the acts of an operant have physical dimensions or properties that can be measured. Some measurable dimensions or properties of operant acts are topography,duration, and latency. Consider the act of pressing a doorbell. The topography of the press is described by a behavioral definition. The definition specifies what will be counted as an instance of pressing the doorbell. In addition to topography, each press also has the dimension of duration which is the time it takes for the press to occur, from beginning to end. The latency of each hand raise is another dimension or property of the behavior that can be measured. Latency is the time it takes for the hand raise to begin after some other event occurs, for example after the instruction to "ring the doorbell" is given.

These dimensions or properties of operants are called variables because the numerical values of the properties of the re-occuring acts can (and do) vary from one occurrence to the next occurrence of the act. For example, if you were measuring the duration of Joe's hand raises, you might measure the first hand raise as having a duration of 2 seconds and the second a duration of 2.2 seconds and the third a duration of 1.8 seconds, etc. Joe's hand raises may turn out to have an average duration of 2 seconds.

Laboratory scientists have studied all of the dimensions of operant units that we will discuss in this course. The advantage of the laboratory is that scientists can arrange the environment so that a behavior stream is taken up almost completely by acts belonging to one operant unit...for example, a rat's lever pressing. The lab provides a way of isolating an operant unit so it can be observed as it changes over time when the behavioral environment is experimentally manipulated in various ways. That is how scientists discovered the ways that behavioral environments cause the properties of actions to have the numerical values they have and to change over time.

In this module we discuss ways that all operant behavior can be measured, whatever the content. We are interested in properties that all operants have in common and that can be observed to change when particular kinds of changes occur in the world outside those operants. We will also discuss some ways of measuring behavior as it occurs in the everyday world. Applied behavior analysts and behavior technologists usually cannot arrange the world so that the operant they are interested in can be isolated in a behavior stream. So they must figure out ways to measure it as it occurs -- distributed across time in a behavior stream in a constantly changing behavioral environment.

5. The acts that constitute an operant have many _______________ or properties that can be measured.

6. The properties (dimensions) of operant acts are called _______________ because the numerical values of any act's properties can (and do) vary from one occurrence to the next occurrence of the act.

7. Laboratory scientists arrange the _______________ that most of the behavior that occurs belongs to the same _______________ .

Ways to Measure the Behavior of Interest

Before we can measure behavior, we must identify or construct the behavior of interest (BOI). The BOI is a particular action that appears with some frequency in the behavior stream of some particular organism. If we work in a laboratory, we have the luxury of arranging the world so that the participating subject (call him Snowball the Rat) has a limited number of acts that can occur in that environment. Let us specify lever pressing as the BOI. We can also arrange the environment so that Snowball hasn't had food for some time and we can make food contingent on lever pressing. Given these arrangements, it is likely that we will see quite a bit of leverpressing, and not much other behavior, in Snowball's current behavior stream.

We can measure the lever pressing in many different ways. Before we can measure it, though, we must define what exactly will qualify as a "lever press". We need a behavioral definition. Let's define the BOI as "any activity of Snowball that has the effect of moving the lever downward far enough to operate an electric switch". So movement that operates the switch constitutes the acts that we are going to measure. Snowball is the organism whose lever pressing we are going to measure in order to begin learning how behavior works. If we learn how Snowball's behavior works, we can then see if some other rat's behavior works in the same way. And then we can see if the behavioral environment functions in the same way with respect to organisms of other species.

Measuring Operant Occurrences

We can begin measuring Snowball's behavior by finding a way to measure each occurrence of the BOI. Each time Snowball's pressing operates the switch, the press can be measured in many ways. As mentioned above, each lever press (and any other act) has a duration. For example, a particular lever press could have a duration of .1-sec, or.5-sec, or 1-sec, or 2-sec. The duration of a particular lever press may be measured more precisely or less precisely. For example, we could measure durations to 1/100 of a second, so the duration we called "2-sec" in the previous sentence may have been measured as 1.888-sec if we had a better measuring device.

Each lever press also has a latency, as does every act. The latency of an occurrence of a lever press is the length of time that passes between the occurrence of some previous event and the current lever press. For example, we could measure the time that passes between the onset of an external event, such a tone, and a press of the lever. A special kind of latency is time since the end of the previous lever press. This latency measure has the special name of "interresponse time" (IRT), so we will reserve the term "latency" for the time between the occurrence of an event external to behavior and the response that follows that event. Latencies and IRTs are recorded one by one.

Each press can also be measured in terms of its force - the pressure applied to the lever by the rat's movement. Consecutive lever presses can be recorded and a force value assigned to each lever press according to the amount of pressure applied to the lever. Duration, latency and force are three measures used most often to measure individual instances of an operant. Although there are other ways of measuring operant occurrences, they are often harder to accomplish - especially if we work outside the laboratory - so we won't discuss them here.

All of the measures of behavior that we have discussed so far are measures of occurrences of acts. All measurement of behavior begins with measurement of occurrences. But single occurrences of behavior are not what behavior analysts are mainly interested in. It is not usually very important that a particular lever press has a duration of 1-second, or a latency of 3 minutes. Behavior analysts are more interested in repeated acts that belong to a particular operant unit. They are especially interested in measures of operant units that allow the observer to know if the operant has changed as a result of an intervention.

9. Before measuring behavior, it is necessary to _______________ the behavior of interest.

10. A short way (abbreviation) of writing or saying "behavior of interest" is _______________ .

11. Each occurrence of an act has many dimensions or properties, each of which can be _______________ .

Measuring Operant Units

Applied behavior analysts need to understand how environmental events are affecting behavior, when new behavior is acquired, how previously learned behavior is changing, and whether some behavior has ceased to occur. As it happens, most acts of humans are repeated many times. These acts are distributed across time in a person's behavior stream. Taken together they constitute an operant unit that itself has properties that can change over time. Even though an operant may change over time, it is still one and the same unit or we couldn't speak of it's "changing". Just as we think of any organism as the "same" organism even though it changes a great deal during a lifetime and each of its cells exists only a short while, we can think of an operant as the "same" operant even though each of its recurring acts exists only briefly in the behavior stream. Just as the organism is the sum total of its cells at any one time segment, so the operant is the sum total of its repeating acts during any one time segment. When we measure an operant, we are interested in properties of a series of acts that constitute the operant as a whole.

The most frequently used way to measure an operant is to count repetitions of an act as they occur. Snowball's first lever press is counted as "1"; if a second lever press occurs, it is counted as "2", etc. Let us say that we counted 10 lever presses as we observed Snowball in the laboratory on one occasion and 10 presses on a second occasion. This kind of measure is called a frequency count. What would it tell us about Snowball"s leverpressing? Not much that is very useful because we don't have enough information. What if we had observed Snowball for 10 minutes on one occasion and only 1 minute on the second occasion? Pressing 10 times in 10 minutes is different than pressing 10 times in 1 minute. Snowball's rate of leverpressing differed greatly on the two occasions even though we obtained the same frequency count. Although frequency can be misleading, rate turns out to be a very useful measure of operants.

To obtain Snowball's rate of leverpressing on each of the two occasions, we converted the frequency for each occasion into a rate measure by dividing the count we obtained on each occasion by the amount of time we spent observing and counting. Ten leverpresses in 10 minutes is a rate of 1 press per minute on the first occasion. Ten leverpresses in 1 minute is a rate of 10 per minute on the second occasion. By considering the period of time during which the acts were counted, we have a good way to compare the behavior on the first occasion to the behavior on the second occasion.

Sometimes behavior analysts just report frequency measures and do not convert them to rate. We can do that if we hold constant the time of the observation periods. If we always measure lever pressing for periods of 10 minutes, or 1 hour, or any other constant amount of time, then we can compare the behavior during one time segment to the behavior during another time segment. In effect, we are comparing rates but because the denominator (time) is constant, we can simply compare the numbers of the numerators (frequencies) for consecutive observation periods. Rate (count divided by time during which counting was done) has turned out to be a very useful measure of behavior, both for the science and the practice of behavior analysis.

There are other useful measures of operant units besides rate and we will give an example of two of them in the section below.

12. An operant's constituent acts are distributed across _______________ in a person's behavior stream.

13. Counting occurrences of an operant is one way to _______________ it.

14. To measure the rate of an operant, you count occurrences of an act and also keep a record of the _______________ during which you are counting.

15. Then you divide the _______________ by 16. the amount of _______________ during which you were counting occurrences.

Measuring Change in Operant Units

When we measure an operant in terms of its rate, we are grouping together all the acts that occur during a particular time segment and looking at them as a whole. When the whole is measured in terms of its rate, it is measured by counting the number of occurrences and dividing it by the amount of time during which we measured. The rate measure captures a property of an operant unit during any particular segment of time.

If the rate changes across time segments, we can say that the operant is changing.

Let's see how we might examine changes in an operant over time. Let's say we measured Joe's doorbell pressing by counting the presses for each year of the first 30 years of his life. Note that in measuring Joe's doorbell pressing, we measure two things. We count Joe's doorbell presses and we time how long we are doing the counting. At the end of a standard amount of time (1 year) we start a new time segment and begin a new count. If we simply counted the presses for 30 years, all we would know at the end of 30 years is that Joe's act of doorbell pressing occurred 3,473 times. We would have no way of knowing if Joe's doorbell pressing ebbed and flowed over time in Joe's behavior stream. By keeping track of the time that passes while we are counting, we can look at the rate of Joe's doorbell pressing during different time segments of equal length.

Here is a graph showing the frequency of Joe's doorbell pressing for each of the first 30 years of his life:Doorbell graph

In this case, time segments were all the same length so the frequency of pressing for each year was displayed on the graph. The length of time for each period of observation is somewhat arbitrary in this case - the frequencies could have been counted in 6-month time segments or in 1-month time segments as well as in 1-year time segments. By comparing the frequencies of different time segments, we can see the operant is undergoing change. The data in the graph shows several kinds of change in the operant over the 30 years of data collection. We can see from the graph that Joe pressed doorbells with increasing frequency during each of his first 9 years. During the next 9 years, his doorbell pressing was distributed fairly evenly - between about 40 and 50 doorbell presses each year. Then we saw that Joe's doorbell pressing increased in frequency to 400 presses during the year Joe was 19. For each of the next 6 years, Joe's doorbell pressing occurred between 380 and 550 times each year. Finally, we see that frequency decreased to 28 doorbell presses the year Joe turned 26 and his pressing continued to occur between 25 and 30 times a year for the next 4 years.

Although we have no information as to why Joe's doorbell pressing went through these rather dramatic changes, the data suggest that Joe's behavioral environment changed a great deal when he was 19 and that it changed again when he was 26. We may want to guess about what changes occurred in Joe's behavioral environment at those times, but since we can't return to the past, we will never know for sure what caused Joe's operant of doorbell pressing to change so dramatically at two different times in his first 30 years. In order to know for sure what causes any particular behavior change, we must use the experimental method. That is the topic of the next reading.

Sometimes behavior analysts find it useful to use measures of operant behavior other than rate. The way behavior is measured in behavior analysis depends on the property of behavior that is of interest to the person doing the measuring. If we are interested in increasing the length of time a child plays cooperatively with other children during the first 10 minutes of recess, then we would use duration as the measure of the behavior of interest. Once we have a behavioral definition of "cooperative play, we can measure with a stopwatch the total duration of time the child plays cooperatively during the first 10 minutes of recess each day. We can see if the duration is increasing, decreasing or remaining the same across days.

If we are interested in decreasing the latency between an instruction and a person's following the instruction, we can measure the time between an instruction and initiation of the instructed behavior. Because we are interested in the overall latency of "instruction following", and not the latency of following any particular instruction, we have to measure each instance of instruction-following and then find the average latency. The average latencies can then be compared across days to see if the behavior is changing.

Before leaving the topic of measurement, we will point out that we are usually interested in how to produce a measurable change in an operant. Is our friend returning our calls less or more often than in the past? Can we do something to decrease the latency of our son's coming to dinner when we call? Can we increase the frequency with which our daughter completes homework assignments? Can we decrease the duration of a child's temper tantrums? Can we increase the force with which we lift weights? In order to accomplish goals like these, we must first have a way to measure what is occurring. The measure we are ultimately interested in are almost always measures of an operant unit - not measures of single acts.

Look at the graph that shows Joe's doorbell pressing operant as it exists year by year over 30 time segments of one-year each.

17. What can you say about Joe's rate of doorbell pressing over the first 9 years?
a. was stable
b. increased steadily
c. decreased steadily
d. increased, then decreased

18. What can you say about Joe's doorbell pressing rates over the first 18 years?
a. was stable throughout
b. increased over the first 9 years and then decreased during the next 9 years
c. increased over the first 9 years and then remained stable for the next 9 years
d. increased steadily

19. If Joe spent some of his first 30 years working as a door-to-door salesman, what is likely to have been the first and last year of his sales career?
a. 19 and 25
b. 17 and 30 vc. 19 and 30
d. 24 and 30

20. The graph is designed to depict a(n) _______________ unit 21. as it exists each year over a period of _______________ years.

22. The operant, measured in terms of rate of occurrence, is seen to be changing during some time segments and remaining stable during other time segments. Between what 2 years was there the greatest change in the rate of Joe's doorbell pressing?
a. between 22 and 23
b. between 10 and 11
c. between 18 and 19
d. between 25 and 26

23. An operant unit is made up of a string of acts that may be widely distributed across _______________ .

Four Methods of Direct Observation (Ways to Measure Rate)

Applied behavior analysts often want to get a measure of an operant's rate. Sometimes they do that by observing the behavior continuously over some period of time and counting each occurrence. If they divide the number of occurrences by the amount of time they observed occurrences, they will have a measure of the operant's rate. This method of obtaining a rate measure is sometimes called event recording. When the BOI occurs, that event is counted.

Sometimes event recording cannot be accomplished so that good measures of the BOI can be obtained. So applied behavior analysts have devised other ways to get measures of rate. The other ways are actually estimates of the rate, but research has shown that they can be very good estimates. Very often, a good estimate of the rate of an operant is sufficient. The point is to measure behavior accurately enough to be able to tell if the behavior changes when an intervention is tried. Lesson 3 in your textbook, which you read for this module, is about event recording as well as methods of observation that yield estimates of rate measures. In addition to the direct measure of frequency accomplished by counting occurrences of an act, three other methods of direct observation are discussed; outcome recording, interval recording, and time sample recording.

In outcome recording, the behavior itself does not have to be observed but the product or outcome of the behavior is observed. Teachers often use outcome recording as a measure of student behavior. If a teacher counts 19 correct answers on a student's math test, the answers on the test are assumed to be the outcome or product of a particular learner's math behavior. Of course, it is possible that the answers are products of behavior other than that which the teacher wants to measure. For example, the answers rather than being the outcome or product of the student's mathematical acts, the answers may be the outcome of the behavior of reading the answers off a cheat sheet. Or it could be the case that the answers are the outcome of another student's mathematical acts and not those of the student whose name is on the paper. Despite the possibility that the outcomes counted may not always be the outcomes of the behavior of interest, outcome recording is a very useful way to measure some behavior.

Interval recording is another indirect measure of rate. It is most useful when instances of the behavior vary a great deal in duration. For example, if you were measuring crying, and sometimes an individual cried for 3 seconds, sometimes a minute or more and sometimes an hour at a time without stopping, counting occurrences of crying (event recording) could be very misleading. On one day, you may count 2 instances but the individual was crying for a total of 40 minutes, whereas on another day you could count 10 instances but the individual was crying for a total of 1 minute. In one kind of interval recording (partial-interval), the observation period is divided into intervals of equal length and if the behavior occurs at any time during an interval, a mark is made (for example a +). Each interval that passes with none of the BOI occurring is recorded with another mark (for example a - ). At the end of the observation period, the observer tallies the number of intervals during which the behavior occurred any time during the interval and the number of intervals during which none of the behavior occurred. Intervals are usually quite short in this method, often about 10 seconds. Observation periods are often about one hour. In another kind of interval recording (whole-interval), the behavior must be occurring throughout the entire interval to be marked as having occurred. This kind of recording is often used when the goal is to increase the duration of behavior that needs to be sustained for prolonged periods--for example, studying or playing cooperatively.

The last indirect measure of rate we will discuss is called time sample recording, or time sampling. The observation period is divided into equal intervals and brief observations are made at the end of each interval. If the behavior is occurring at the time of an observation, the interval is marked one way; if the behavior is not occurring at the time of an observation, the interval is marked another way. This kind of recording method allows the observer to do other things in addition to observing the BOI during the observation period.

24. The method of direct observation that counts each occurrence of an act at the time it occurs is called _______________ recording.

25. The method of direct observation that divides the observation period into intervals during which the observer makes a mark in each interval indicating behavior occurred or didn't is called _______________ recording.

26. The method of direct observation in which the observer counts a product of the behavior of interest is called _______________ recording

27. The method of direct observation in which the observer divides the observation period into intervals and observes briefly at the end of each interval is called _______________ recording.


As you can see, measuring an operant unit can be difficult. Because the unit is made up of a string of acts, which are sometimes widely distributed in time, we must always measure the individual occurrences in some way and also measure the time during which we are measuring those individual occurrences. Discovering this way of measuring behavior was one of the great scientific discoveries of all time because it allowed scientists to begin to understand the processes that cause changes in behavior during an organism's lifetime. The importance of measuring operant units in a way that allows us to observe behavior changing over time is not yet fully appreciated. As you can see, it is not particularly easy to understand. (It is also hard to explain!) The exercises for this week will help you understand it better.