Chapter 2: The Study Of Learning And Behavior

Chapter 2: The Study of Learning and Behavior



Recall that Skinner said that “Behaviorism is not the science of human behavior; it is the philosophy of that science.”



So what is the science?



Behavior Analysis.

Behavior Analysis  

Two Branches Experimental Analysis of Behavior  



Focus on discovering basic principles of behavior Research done in a laboratory environment with both humans and nonhumans

Applied Behavior Analysis 



Focus on discovering and clarifying relations between socially significant behavior and its controlling environmental variables Research done in “real world” environments with humans

Behavioral Repertoire 

We want to study behavior. 



 





Q: How many behaviors are there? What types of behavior are there? A: It depends on the species and the individual. “Pigs don’t fly.” “Berry doesn’t dance.”

We must discover the behaviors that are possible or likely for a given species or individual. Behavioral Repertoire: a catalogue of behaviors that are characteristic of a

Behavioral Repertoire 

Alien Example



Are the behaviors we catalogued all that Fuzz can do? We don’t know. Behavior is contextual. We have only observed Fuzz in one context.



Operational Definitions 





To study specific behaviors (and their changes), we must define the behaviors. In behavior analysis (and science in general), we define variables in terms of how they are measured—i.e., by the operations that are used to measure a behavior. E.g.: When has a person learned to hit a golf ball?  

Ball is off the tee? Does a 30 yard “worm burner” count?



Some operational definitions of learning found in the literature…..

Errors: number errors made on a task. Decline in errors indicates learning. e.g., typing

Topography: the shape of a behavior. When the shape of a behavior resembles the shape of a model or goal, learning has occurred. (typing? Maybe “compare documents” with a test document)

Intensity: the strength of a behavior. Sometimes the goal in to increase intensity, sometimes to decrease intensity (e.g., learning to kiss). (typing? Maybe “lighter touch” is a goal)

Speed: the time it takes to perform a task or reach a goal. (Typing? Perhaps the time it takes to comple a document without errors)

Rates of Behavior [the importance of “per” (\)]frequencies or counts are often  Simple less informative than rates (X per Y, or X\Y) 

How many parking tickets have you had?    



5 5 in 10 years 5 this week I only misspelled 3 words on my typing test. How many words did you type? Uh, 6.

Rate is a frequency standardized (often by time)   

Miles per hour Responses per minute Errors per page

Rates

Cumulative Recorder: Measures rate of target behaviors (e.g., bar presses) over time. The slope of the curve indicates the rate. Greater steepness, greater Rate.

Research Designs in Learning 

Non-Experimental Research  





aka Descriptive or Correlational Researcher does not manipulate any variables, simply records behaviors and experiences to see if they are related (measure study habits and grades—see if they are related). Can’t infer causality with much certainty (e.g., Hans Eysenck on smoking and cancer; stressprone personality).

Experimental Research 



Researcher manipulates an environmental variable. Measures effect of the manipulation on behavior.

Experimental Research: Terminology Independent Variable 





Dependent Variable 



The variable that the researcher manipulates. The hypothesized cause of behavior change

The outcome variable (behavior) is expected to change as a result of the manipulation.

IV  DV

Between- vs. Within-Subjects Experimental Designs 

Between Subjects Experiments 

Involves at least two groups of subjects. 







Subjects randomly assigned to groups to make them equal on a variety of extraneous variables.

One group (Experimental Group) gets exposed to a manipulated variable (e.g., new teaching method; rewards for correct responses; electric shock for wrong responses). The other group (Control Group) is not exposed to the manipulation.

Compare the groups on the DV. Did exposure to the manipulation lead to differences between the groups?

Between-Subjects Experiments Manipulated

IV

Experiment al Group*

DV

Control Group*

DV

* Best if group membership determined by a random process

Between-Subjects Experiments New Study Method

IV

Experiment al Group*

DV

Control Group*

DV

(Study as Usual)

(Exam Scores)

(Exam Scores)

Compare the average exam scores of the experimental vs. control groups

* Best if group membership determined by a random process

Characteristics of BetweenDesigns Subjects Data in each group are pooled or 



averaged; you are comparing average change between the two groups. Individual variations around the average treated as “noise.” Researcher loses touch with individuals. Large samples and complex statistics are often needed to detect an effect of the manipulation.

Within-Subjects Designs 





No separate groups of subjects: each subject will serve as “their own control.” Each subject is assessed before the manipulation (baseline measures), then the IV is applied, and each subject is measure again. Determine whether there was change in behavior from baseline to the next measurement period. 

 

Within-subjects designs can involve large groups (but fewer are needed on statistical grounds). Data can be averaged and tested statistically. But within-Ss designs can also be applied to individuals.

Within-Subjects Experiments

DV Before

All Subjects (maybe just 1)

DV After

IV Manipulated

All subjects measure on DV before manipulation, then again after manipulation. Can use large groups and average data, but fewer subjects needed.

Behavior Analytic “Research Culture” 

Behavior analysts are very fond of within-subjects research designs 

  



Often present data for only a few subjects (sometimes only 1 subject). Seldom average data across subjects. Graphical displays preferred. They dislike complex statistical analyses, preferring graphical data displays. In these preferences they are imitating Skinner.

Skinner on methods: 

“Young psychologists should learn how to work with single organisms rather than with large groups”



“Most of the experiments described in The Behavior of Organism were done with groups of four rats. A fairy common reaction was that such groups were too small. Keller, in defense of the book, countered with the charge that groups of four were too big.”



“For more than a generation . . .our graduate schools . . . Have taught statistics in lieu of scientific method.”



“What statisticians call experimental design . . . Usually generates a much more intimate acquaintance with a calculating machine than with a behaving organism.”



JOURNAL OF APPLIED BEHAVIOR ANALYSIS 2008, 41, 131–135 NUMBER 1 (SPRING 2008)



TRAINING TO INCREASE SAFE TRAY CARRYING AMONG COCKTAIL SERVERS MEGAN D. SCHERRER AND DAVID A. WILDER FLORIDA INSTITUTE OF TECHNOLOGY

 



Simple AB design Baseline

Treatment

Minutes spent on assignments

18 16 14 12 10

A

8

B

6 4 2 0 1

2

3

4

5

6

7

8

9

10

11

Sessions (time)

Why several baseline and treatment period measures? To look for trends, not just levels. [e.g., on board)

12

13

14

15

16

17

Problem with AB Design 



Something other than the experimental manipulation might have caused the change. Confound—an extraneous variable (something other than the manipulation) that actually causes the DV to change. 

Whole Language vs. Phonetic (e.g. from book, p.43)

ABA Reversal Design Baseline

Treatment

Withdraw Intervention Baseline

Minutes spent on assignments

18 16 14 12 10

A

8

B

A

6 4 2 0 1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Sessions (time)

ABAB Reversal Design Baseline

Treatment

Baseline

Treatment

Minutes spent on assignments

18 16 14 12 10 8 6 4 2 0 1

3

5

7

9

11

13

15

17

19

Sessions (time)

21

23

25

27

29

31

33

Problems with Reversal Designs Not useful ABA, ABAB,if the etc.learning is irreversible 



(i.e., doesn’t go back to baseline after manipulation stopped) Ethical issues in withdrawing manipulation 

e.g., if the DV is head banging in a child, and the manipulation decreases headbanging, is it ethical to remove the manipulation?

Multiple Baseline Designs Conduct more than one set of trials. Vary the duration of baseline periods. Aim is to show that the learning occurs after the baseline period, regardless of duration of the baseline period. Shows that your manipulation Is “in control” of the bahavior.

Multiple baseline design 

3 types 

Across participants 



Across behaviors 



Same settings and behavior, but different participants Same participant and setting, but different behaviors (or the same behavior measured in different ways)

Across settings 

Same participants and behavior, but different settings

Carol

Homeroom

Homeroom

Joe

Minutes on Task

e.g. time reading

Bob

Across participa nts

Homeroom Sessions

Time sitting Time being quiet

Homeroom

Min. In Seat Min. w/ No Talkouts

Time reading

Bob

e.g.

Min. on Task

Across behavio rs

Homeroom

Homeroom Sessions

Math

Bob

Bob

Reading

Minutes on Task

e.g., time reading in different classes

Homeroom

Across Setting s

Bob Sessions



JOURNAL OF APPLIED BEHAVIOR ANALYSIS 2008, 41, 131–135 NUMBER 1 (SPRING 2008)



TRAINING TO INCREASE SAFE TRAY CARRYING AMONG COCKTAIL SERVERS MEGAN D. SCHERRER AND DAVID A. WILDER FLORIDA INSTITUTE OF TECHNOLOGY

 





Across individuals multiple baseline design