1 ANALYSIS OF RESPONSE FORCE IN TRANSLATIONAL AND APPLIED SET TINGS By GRIFFIN W. ROOKER A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010
2 2010 Griffin W. Rooker
3 To my family for all they have done to help me become who I am.
4 ACKNOWLEDGMENTS This research was supported in part by a grant from the Florida Agency on Persons with Disabilities. I especially thank Marc Branch his for thoughtful comments and suggestions throughout the course of the study and Angie Querim for the many hours she assisted in conducting sessions collecting data, a nd helping with various other aspects of the project. I sincerely thank John Gagnon, Scott Miller and Timothy Vollmer for their guidance and helpful suggestions on this and other projects. Finally, I would like to express appreciation to Brian Iwata for h is time, dedication, and support throughout this project and my career.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ...................................................................................................... 4 LIST OF TABLES ................................................................................................................ 7 LIST OF FIGURES .............................................................................................................. 8 ABSTRACT .......................................................................................................................... 9 CHAPTER 1 INTRODUCTION ........................................................................................................ 10 Basic and Applied Research on Response Force ..................................................... 10 Basic Research on Response Force ......................................................................... 11 Applied Research on Response Force ...................................................................... 15 2 STUDY 1: RESPONSE DIMENSIONS DURING REINFORCEMENT AND EXTINCTION .............................................................................................................. 18 Method ........................................................................................................................ 18 Subject, Settings, and Apparatus ........................................................................ 18 Response Measurement and Reliability .............................................................. 18 Preference Assessment ....................................................................................... 19 Procedures ........................................................................................................... 20 Results ........................................................................................................................ 20 3 STUDY 2: RESPONSE DIMENSIONS IN RATIO SCHEDULES OF REINFORCEMENT .................................................................................................... 25 Method ........................................................................................................................ 25 Subjects, Setting, and Apparatus ........................................................................ 25 Preference Assessment ....................................................................................... 26 Procedures ........................................................................................................... 26 Results ........................................................................................................................ 26 4 STUDY 3: RESPONSE DIMENSIONS IN BAND DISCRIMINATION ...................... 30 M ethod ........................................................................................................................ 30 Subjects, Setting, and Apparatus ........................................................................ 30 Procedures ........................................................................................................... 31 Results ........................................................................................................................ 31 5 STUDY 4: RESPONSE DIMENSIONS IN PROBLEM BEHAVIOR .......................... 35
6 Method ........................................................................................................................ 36 Subject and Setting .............................................................................................. 36 Functional Analysis .............................................................................................. 37 Video Modeling Analysis ...................................................................................... 38 Results ........................................................................................................................ 40 6 DISCUSSION .............................................................................................................. 45 LIST OF REFERENCES ................................................................................................... 57 BIOGRAPHICAL SKETCH ................................................................................................ 59
7 LIST OF TABLES Table page 2 -1 Subject characteristics ........................................................................................... 56
8 LIST OF FIGURES Figure page 2 -1 Response rate and peak response force under conditions of CRF and EXT fo r Evan, Jose, Karen, and Nick ............................................................................ 24 3 -1 Response rate and peak response force under conditions FR1 and FR10 reinforcement for Evan, Karen, Lisa, and Nick. .................................................... 29 4 -1 Response rate and peak response force under conditions under a highthreshold response -force requirement for Erin and Nick. ..................................... 34 5 -1 Response rate (left and right panels) and rate of arm rotation (right panel ) during the functional analysis of Billy and Evans aggression. ............................. 43 5 -2 Response rate (left and right panels) and rate of arm rotation (left panel) during the functional analysis of Regan and Erins property destrcution. ............ 44
9 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Phi losophy ANALYSIS OF RESPONSE FORCE IN TRANSLATIONAL AND APPLIED SET TINGS By Griffin W. Rooker August 2010 Chair: Brian A. Iwata Major: Psychology Research on operant behavior often focuses exclusively on response frequency because it is a convenient measure and easier to quantify than are changes in other dimensions such as topography (form) or force. However, force is a fundamental aspect of a response because some amount of force must occur for the response to be detected. In addition, force may be th e most relevant defining feature of some responses such as aggression or self -injury. This series of studies examined the relation between response rate (frequency) and response force in four translational contexts and one applied c ontext. The rate and for ce of a button press were measured under conditions of reinforcement and extinction (Study 1) and ratio schedules of reinforcement (Study 2) applied to response rate. The same measures were taken when an increasing requirement was applied to response force (Study 3) Finally, t he rate and force (or severity) of problem behavior were measured during the course of experimental (functional analysis) assessment in S tudy 4 Results showed positive correlations between rate and force under some conditions but neg ative correlations under others In addition, results indicated that the severity (force) of problem behavior decreases within the context of the assessment
10 CHAPTER 1 INTRODUCTION Basic and Applied Research on Response Force Several distinctive feature s define an operant response, including topography, force, and duration. However, force often is not measured but simply considered a static feature of the dependent variable Research on operant behavior focuses on changes in responding over time and most often on repeatability of a predefined response rather than changes in fundamental response properties. Response rate or frequency often has been identified as the primary measure in most behavioral research ( Cooper, Heron, Heward, 2008; Fester & Skinner, 1957; Michael, 1995) for several reasons. First, response rate is a convenient measure. It is relatively easy to count the number of responses that occur in a given period of time. Second, rate provides a quantitative index of change across time as a func tion of changing environmental conditions. Finally, rate allows for some variation in the dimensions of topography, force, and duration. That is, no change in definition is required when slight variations in the form of a response occur Nevertheless, fo rce is a fundamental characteristic of any response, and rate is sensitive to force only at a given threshold. That is, if a response does not reach a threshold, it is not considered to be a response at all. In this way, rate of response can be a limiting measure. For example, vocal responses that are whispered may escape detection, although they clearly are examples of speech. Similarly, rate measures are insensitive to changes in intensity when all response s occur above threshold. For example, increases or decreases in the intensity of self -injur ious behavior (SIB) cannot be detected in typical rate measures when responding is defined as any amount of
11 physical contact Thus, changes in response force per se may be important to understanding some behavior. Basic Research on Response Force Basic research on response force has focused mainly on lever pressing in rats to determine how variables that affect response rate also affect force and vice versa. Notterman (1959) for example, reinforced the occurrence of lever presses while measuring response force. He found that the force of rats responding on a lever during reinforcement decreased over time (to more closely approximate the threshold response) and that response force in extinction was highly variable. Similarly, Mintz (1962) delivered reinforcement for rats lever pressing on a fixed -ratio (FR 12) schedule. He examined peak response force (the highest force during a response) and found that response force increased throughout the inter -reinforcement interval until reinforcement was delivered. Studies such as t hese involved manipulat ion of response rate and observed correlated changes in response force. Other studies examined manipulation of response force directly. For example, Adair and Wright (1976) trained squirrel monkeys to regulate the air temperature in an experimental chamber by pulling levers connected to valves emitting warm and cool air. T he experimenters subsequently provided continuous cool air unless the subject pulled a lever to access w arm air and then incrementally added counterweight s to the chain. As weight increased, the interresponse time between chain pulls increased, and air temperature in the chamber decreased. These results were significant because effortful responses became mo re likely as the chamber got cooler. When the counterweights were not on the chain, more consistent temperatures and more frequent chain pulls occurred.
12 The most complete study of response force was conducted by Notterman and Mintz (1965) who examined response rate and response force when force was manipulated both directly and indirectly When force was directly manipulated, the authors examined the effects of contingencies on force duration, band discrimination (responding within a range of force) d iscriminative stimuli, feedback, and proportional reinforcement (reinforcement proportional to force) When response rate was manipulated directly, the authors examined the effects of extinction, discriminative stimuli, deprivation, sequential ratio schedu les, and intensity of aversive stimulation. One example of a manipulation that directly affected response force was conducted i n the fourth study in which the authors examined band discrimination-differential responding within a range of parameters, i n this case, response forces capped at a high and a low level For example, when slowing a car to a stop, an individual typically use s a band of pressure on the brake to bring the car to a rest comfortably. Immediate and intense braking would cause pain to the individual and the avoidance of this consequence serves as a negative reinforcer. Thus a constant amount of moderate pressure applied to the brake stops the car gradually, although small deviations from the ideal amount of pressure still provide rei nforcement. Similarly, a rat can be trained to respond within a window of response force. The authors examined band responding in three groups of rats (six rats per group). The first group was trained to press a lever between 8 and 16 g for 20 sessions. Th e second group was trained to respond between 4 and 16 g for 7 sessions and then between 8 and 16 g for 13 sessions. The third group was trained to respond above 8 g for 20 sessions. During the reinforcement condition, stable responding was observed in all groups, with
13 responding within the band of response force for the first and second groups, and consistently around 20 g for the third group. During a subsequent extinction condition, a proportional increase was observed in response force with the highest force observed in the third group. An example of a manipulation that indirectly affected response force was conducted in the first study, in which the author s exposed 6 rats to c onditions of reinforcement and extinction. In addition to measuring response frequency, the authors collected data on the peak force of each response. T he rats received pellets on a continuous schedule of reinforcement (FR 1) for up to 50 pellets in the reinforcement condition The threshold to activate the magazine in the reinfor cement condition was 3 g. N o consequences were delivered for pressing the lever during extinction. The authors observed a decrease in response force and the variability of force over the course of the reinforcement conditions. In the extinction condition, both response force and the variability of force observed increased. A similar indirect manipulation of response force was seen i n the eighth experiment, in which the authors examined the sequential effects of fixed ratio ( FR ) reinforcement. The authors c onducted this study with three groups of five rats. T he authors exposed the three groups to three different FR schedules (12 days for each schedule). Each group began the procedure with five days of FR1, and two animals from each group ended the experiment with four days of FR1. The authors described the FR schedules in a slightly different manner than typical FR procedures. The number of responses between reinforcers for each animal was kept consistent across groups ; however, the number of reinforcers that could then be obtained changed across groups.
14 For example, the first FR schedule for the first group was FR 6(1). This meant that the rat received one opportunity to earn a reinforcer following six unreinforced responses which corresponds to a n FR 7 sche dule in current terminology The first FR schedule for the second group was FR 6(2). This meant that the rat received two opportunities to earn a reinforcer following six unreinforced responses or a tandem FR 7/FR1 schedule. The first FR schedule for the third group was FR 6(4). This meant that the rat received four opportunities to earn a reinforcer following six unreinforced responses or a tandem FR7/FR1/FR1/FR1 schedule. The sequence of FR schedules for all groups was 6, 12, and 24, and the tandem schedules were held consistent. The only distinction was the number of reinforcers that could then be earned by the different groups (1, 2, or 4). The authors found a characteristic response pattern for all FR schedules: a sharp decrease in response force imm ediately following reinforcement, and an acceleration in response force during nonreinforced responses immediately prior to the delivery of reinforcement. These results were similar to those obtained by Mintz (1962); however, the current study greatly enha nced the generality of these results. Morris (1966,1968) extended basic research on response force to human behavior using a spring-loaded lever press. The author examined response force in extinction following training at a criterion level of force. Mor ris included nonclinical subjects in the first study and both clinical and nonclinical subjects in the second study. He observed th at human responding on a lever was more variable and more forceful during extinction than during reinforcement phases. That i s, results were similar to those found by Notterman and Mintz (1965). It is important to note that in both nonhuman
15 (Notterman & Mintz, 1965) and human (Morris, 1966, 1968) studies, some form of training was conducted in the acquisition phase of the experi ment prior to extinction. Applied Research on Response Force Response force also has been used as the index of behavior change in some applied studies One particular response that has received attention is selective mutism (also called elective mutism and hysterical aphonia). Walton and Black (1959) described an early case history and treatment of one individuals aphonia through the use of a stimulant drug and social reinforcement. Wulbert, Nyman, and Snow (1973) increased the vocal output of an individual who had not spoken in 3 yr through the use of a stimulus fading procedure and contingency management. Stimulus fading involved the therapist slowly moving into a room with the subject and asking the subject a series of questions that required a oneword response, while the subjects mother slowly moved out of the room. The subject received a piece of candy for any verbal response that was audible. Although the authors presented the frequency ( number ) of responses that the individual emitted, the conting ency actually was placed on audible responses, which were a function of response force. A more direct measure of force was used by Jackson and Wallace (1974) who increased the subaudible vocal responses of one individual to audible vocal responses. Vocal output was measured by an audiometer and shaped into audible responses through a token system. The procedures were effective at increasing the volume of speech, as well as the length of the word (polysyllabic) spoken. In addition to increasing the volume of verbal behavior, applied research has examined decreasing the volume of verbal responses. Greene, Bailey, and Barber (1981) decreased the number and duration of noisy disruptions on a school bus. Noise
16 was measured by way of an audiometer, and the auth ors observed a decrease in noise after implementing reinforcement consisting of raffles, light display, and contingent music. However, because the authors reinforced the reduction of loud responses and not the occurrence of verbal responses at a lower level, it is unclear from the data presented whether the frequency of verbal behavior remained constant but was emitted with less force, or if there was a general decrease in all verbal behavior. More relevant to the applied aims of the present study is research on estimate s of the force of problem behavior and, more specifically, SIB. For example, Iwata, Pace, Kissel, Nau, and Faber (1990) developed the Self -Injury Trauma (SIT) Scale, an observational tool that permits objective measurement of the location, t ype, size, and severi ty of wounds produced by SIB. The authors evaluated the SIT scale by having pairs of raters score 50 clients who engaged in SIB and obtained a high percentage of agreement (94%) on wound size and severity. In a more recent study, Wilson, Iwata, and Bloom (in press) used computerized photographic analysis to precisely measure wound size, and then examined the relation between changes in the frequency of SIB and changes in wound size. Finally, Symons, Harper, McGrath, Breau, and Bodfish (2009) evaluated a checklist for identifying painful stimulation for individuals who were noncommunicative. They examined caregiver ratings of overt signs of discomfort (e.g., grimacing) from SIB. Here, the assumption was that increased painful reactions to injuries were related to the greater force of the response. Response force may be of interest because it may be the most relevant dimension of some responses. For example, self -injurious behavior (SIB) is defined by its outcome, which presumably is a function of response force That is, the severity or
17 force of the response determines whether it meets the criterion for SIB (may cause damage to self). Thus the force of SIB may be far more important than its rate. Consider the individual who engages in S IB by hitting the eye : O ne instance of forceful hitting may be sufficient to dislodge the retina. Newell and colleagues illustrated direct measurement of the force of problem behavior by examining joint movement during stereotypy (Newell, Incledon, Bodfis h, & Sprague, 1999) and SIB (Newell, Challis, Boros, & Bodfish, 2002). In both studies, the authors used the Peak Motus system to videotape individuals engaging in problem behavior The system then generated computerized model s of the subjects action to m easure limb movement. In the first study, the authors measured angular velocity to estimate the force of stereotypic movements; in the second study, the authors collected data on both limb movement for the action itself as well as corresponding body movements for SIB. That is, the authors were able to determine how the targeted body part moved based upon the self -injurious act (e.g., corresponding movement of the head for head -hitting). Following this, the authors then used a standard model to estimate the amount of force for each self -injurious blow. The overall purpose of the present studies was to conduct a series of related experiments on human response force. S tudy 1 and Study 2 examined effects on rate and force when variables that typical ly affect r esponse rate were manipulated. Study 3 examined effects on rate and force when variables that typically affect response force were manipulated. Finally, Study 4 examined the relation between the rate and force of problem behavior in the context of an asses sment of problem behavior.
18 CHAPTER 2 STUDY 1: RESPONSE DI MENSIONS DURING REIN FORCEMENT AND EXTINCTION Method Subject, Settings, and A pparatus All subjects for Studies 1, 2 and 3 were selected based on teacher reports that indicated they would be appropri ate for a study that required sitting at a desk. Four individuals diagnosed with developmental disabilities participated. Subject demographic information is listed in Table 2 1. All sessions were conducted in a designated area of a classroom in a special education school and were 5 min in duration During sessions, the apparatus was placed on a table approximately .25 m from the subject and a therapist was present to provide edible items. The apparatus was a Tekscan flexi -force load cell a force-resistance pad that was built into a switch press button. The cell was linked to a computer that monitored the force of each response by use of a software program developed by Tekscan (the ELF system). Thus each butt on press registered both the occurrence of a re sponse in real time and its force in lb per square in. The button emit ted an audible click when a sufficient amount of pressure was placed on the panel (.07 lb). T h e load cell was calibrated with standard weight s (5 and 25 lb) each week to ensure that it was recording response force accurately. Response Measurement and Reliability Following each session, mean response rate (responses per min or RPM) and mean peak response force (lb) was calculated from the recorded d ata. The occurrence of a response produced a different amount of force over the course of the response cycle and t he peak force was the greatest force exerted during the cycle In addition to
19 recording response rate and force from the load -cell data, observers recorded the frequency of button presses (defined as pressing the button with sufficient force to produce a click) and delivery of edible items (defined as placing an edible item on the plate). Interobserver agreement was assessed by having a second observer independently collect da ta during at least 25% of sessions. Proportional agreement percentages were calculated for each response by comparing the two observers recorded frequencies in each 10-s interval. The smaller number of responses was divided by the larger number of respons es in each interval with a disagreement, the fractions were summed across all intervals, and the total was added to the total number of agreement intervals in the session. The sum was divided by the total number of intervals in the session and multiplied b y 100% to yield reliability scores for each measure. Mean reliability scores were as follows: Evan, 90% for button presses (range, 77 % -100%) and 95.2% for reinforcers delivered (range, 90 % 100%); Jose, 95.9% for button presses (range, 85.6% 100%) and 95.7% for reinforcers delivered (range 86.7 % 100%); Karen, 94.5% for button presses (range, 80 % -100%) and 93.2% for reinforcers delivered (range, 83.3 % 100%); and Nick, 94.5 % for button presses (range, 80.6% 100%) and 93.2 % reinforcers delivered (range, 83.3 % -1 00 % ). Preference Assessment Prior to the experiment a paired-stimulus preference assessment (Fisher et al. 1992) was conducted to identify a highly preferred (selected on more than 80% of presentations) edible item. Nine edible items were compared, with two items presented on each trial, over the course of 36 randomized trials. In each trial, the subject was
20 asked to select one of two edible items until all items had been compared. Data collectors recorded data on the items that the subject selected. Pro cedures T wo conditions were conducted. In the reinforcement condition, the highly preferred edible item was delivered following each response on an FR1 (continuous reinforcement or CRF) schedule T he button was removed following delivery of an edible item until the item was consumed and then was returned to the same place in front of the subject. In the extinction (EXT) condition, no consequences were delivered for any response. These conditions were conducted in a reversal (ABAB) design with the exception that Nick experienced an additional reversal series because his initial results were not replicated. Results Figure 21 shows results of the analysis of reinforcement and EX T for Evan, Karen, Jose, and Nick All subjects response rates were stable or i ncreasing through out the FR1 condition. During extinction ( EXT ), one of two patterns was observed for response rate. Evans and Karens (top two panels) response rate showed an initial increase followed by a decrease to near zero, whereas Joses and Nicks (bottom two panels) responding decreased to near zero immediately. Jose responded seven times across two EXT conditions, and Nick responded four times across three EXT conditions. Data on response force during the initial FR1 condition showed relatively s table responding (Evan, Nick) or slightly decreasing trends (Karen and Jose). Three subjects showed an initial increase in response force during both EXT conditions (Evan, Karen, and Jose). Response force subsequently decreased across the phase as response frequency decreased to zero.
21 Data on response force during the FR1 and EXT conditions for Nick were somewhat different than for the other three subjects. A stable level of responding was observed in the initial FR1 condition. In EXT, there was no increas e in response force. In the second implementation of FR1, response force was higher in the initial FR1 condition but was still stable. In the second implementation of EXT, an increase in response force was observed during the first session. These results did not replicate those obtained in the first FR1 and EXT conditions. Therefore, an additional reversal was conducted. In the third implementation of FR1, response force was slightly decreasing and returned to the level previously observed in the first impl ementation of FR1. Following this, an increased level of response force was once again observed in the third implementation of EXT. T wo patterns of variation were seen between response rate and force during CRF. The first pattern was positive covariat ion; that is, the rate and peak force of button presses followed the same pattern (stability or variability) and fluctuated in the same direction (increased and decreased together) This pattern was observed for Evan across all four conditions, with the except ion of a few sessions immediately following a medication decrease. For example, in the first implementation of FR1, rate and peak force both occurred at a relatively stable level and seemed to fluctuate together. Similarly, when EXT was implemented, an inc reased variability was observed in both rate and peak force, and an immediate increase (and subsequent decrease) in RPM and force was observed. This effect was replicated in the second implementation of FR1 and EXT for Evan as well as in all four conditions for Karen, and more tenuously for the first three conditions for Nick.
22 The second pattern was negative covariation in which the rate and peak force of button presses followed a different pattern (stability vs. variability) and fluctuated in a different direction (increased and decreased separately) This pattern was observed with Jose. In his first implementation of FR1, rate increased across the course of the condition. By contrast, peak force decreased across the course of the condition. Although stabi lity was achieved in both RPM and force, the pattern observed was not similar: As one rate increased, force decreased. Similarly, in the first implementation of EXT, rate decreased immediately to near zero; however, peak force increased immediately. Response rate was stable throughout the condition, whereas force was variable, and, at least initially, force increased as response rate decreased. This effect was replicated (and more exaggerated) in the second implementation of FR1 and EXT for Jose, and was al so observed in the last three conditions for Nick. In multiple EXT conditions, peak force was higher than had been observed in the previous FR1 phase. An EXT burst has been described as an increase in the rate of a response in EXT over the last three sessi ons of the previous reinforcement phase (Lerman Iwata, & Wallace, 1999) An EXT burst in rate was observed in the response rate of two of the four subjects (Evan and Karen). B ecause response force is not often measured, there have been fewer reports of similar bursts in response force. Using the sam e definition for a burst in force that has been used for a burst in rate, an EXT burst in force was observed with every subject. In addition, an EXT burst in force occurred in every implementation of EXT with the except ion of the first implementation of EX T for Nick.
23 Another finding of minor interest was the effect on responding in FR1 with an intervening history of EXT. Three of the four subjects (Karen, Jose, and Nick) showed an increase in peak force in the initial sessions of the second implementation of FR1. As the FR1 condition continued, response force returned to levels similar to those observed in the initial (or previous, as in the case of Nick) FR1 condition.
24 Figure 21. Response rate and peak response force under conditions of CRF and EXT fo r Evan, Jose, Karen, and Nick
25 CHAPTER 3 STUDY 2: RESPONSE DI MENSIONS IN RATIO SC HEDULES OF REINFORCE MENT Study 1 demonstrated that response rate and peak response force may covary (either positively or negatively). However, that study assessed responding only in the presence or absence of reinforcement. As previously noted, both Mintz (1962) and Notterman and Mintz (1965) conducted basic studies that assessed response frequency and force in the context ratio reinforcement. Mintz, who used an FR12 schedul e found an interesting effect, in that there was an increase in response force under ratio schedules. This finding is counter -intuitive. Increasing the ratio of reinforcement has been shown to increase the rate of responses (Cooper, Heron, & Heward, 2007) Because more responses are required to obtain reinforcement, one might predict that a decrease in the amount of pressure per response would also occur. That is, because the organism must burn more calories by responding more to obtain reinforcement, the organism might decrease the amount of force per response to compensate for the additional calories burned. A decrease in response force as a function of an increased response frequency requirement has not been observed. Furthermore, the effect described by Mintz has not replicated in humans. Therefore, covariation between response rate and peak response force was examined in humans under FR1 and FR10 schedules. Method Subjects, Setting, and Apparatus Four individuals, three of who were subjects in Study 1 (Evan, Karen, and Nick) participated in this study (see demographic information in Table 21 ). The setting and apparatus were identical to those in Study 1, as were methods for collecting data.
26 Interobserver agreement was assessed during at least 25% o f sessions and was calculated as described previously. Mean reliability scores were as follows: Evan, 95% for button presses (range, 87 % -100%) and 93.9% (range, 81.7% 100%) for reinforcers delivered; Karen, 98.4% for button presses (range, 80 % -100%) and 90 .6% (range, 81.1 % -100%) for reinforcers delivered; Lisa, 95.4 (range, 83.9 % 100%) for button presses and 93.3% (range, 83.3 % 100%) for reinforcers delivered, and Nick, 99.3% (range, 83.1 % 99.2%) for button presses and 95% (range, 82.2 % -100%) for reinforcer s delivered. Preference Assessment Procedures identical to study 1 were used to identify a highly preferred (selected in greater than 80% of presentations) edible item. Procedures T wo conditions were implemented in a reversal design. Reinforcement (th e preferred edible item) was delivered for responding in both conditions; the only difference was the schedule in effect. In the FR1 condition, the edible item was delivered following each response. In the FR10 condition, the edible item was delivered foll owing every tenth response. Results Figure 31 shows the results of the comparison of FR1 and FR10 schedules for Evan, Karen, Lisa, and Nick. All subjects response rates were stable or increasing throughout the FR1 condition. During the FR10 condition, there was either an immediate (Karen and Lisa) or eventual ( Nathan and Evan) increase in all subjects response rate. This effect was replicated in both implementations of FR1 and FR10.
27 Data on response force during the initial and subsequent FR1 conditi ons showed relatively stable responding. Data on response force in the FR10 condition showed either a slight increase or no change in comparison to the FR1 condition. A slight increase in response force was observed under FR10 for three subjects (Karen, Ev an, and Nick) whereas no difference between the peak force was observed in FR1 and FR10 for one subject (Lisa). Lisa showed a slight increase in response force in the second FR10 condition; however her peak force subsequently decreased to levels similar to the previous condition. In addition, an increase in variability of peak force was observed for all su bjects in FR10. Positive covariation between rate and peak force was observed across all four conditions for Nick. In his first implementation of FR1, rate and peak force both occurred a relatively stable level and seemed to fluctuate together. Similarly, when FR10 was implemented, response rate increased by approximately 50%, and response force increased slightly. This effect was replicated in the second implementation of FR1 and FR10 and was more prominent in the second implementation of FR10. This positive covariation also was observed for Evan and Lisa to a lesser degree. Karens results were mixed. In her first implementation of FR1 and FR10, there was negative covariation between rate and peak force. As rate increased, peak force decreased. This effect also was replicated in the FR10. However, in the reversal of these conditions, there was positive covariation. That is, in her second implementation of FR1 and FR10, rate and peak force covaried positively. Another interesting finding was that response force was slightly greater in FR10 than FR1 for three of four subjects. Lisas results, with respect to force of responses in
28 FR1 and FR10, are unint erpretable. The final level of response force in FR10 was greater than the initial level in FR1 (but not the later observed level in FR1); however, an increasing trend was observed across all conditions.
29 Figure 31. Response rate and peak response for ce under conditions FR1 and FR10 reinforcement for Evan, Karen, Lisa, and Nick.
30 CHAPTER 4 STUDY 3 : RESPONSE DIMENSION S IN BAND DISCRIMINA TION Notterman and Mintz (1965) conducted a study in which rats were reinforced for responding within a particular ra nge (band) of force. Learning of this type may be of interest because everyday tasks often require an individual to respond with more or less force in order to obtain reinforcement. For example, many individuals with developmental disabilities are employed through piece meal work and often these tasks require a chain of motor responses. Here, the force of motor responses is important -enough force to put an item together, but not too much force as to break it Similar example s can be seen in a number of sp orts in which effective responding ( e.g., shooting a basketball, hitting a tennis ball, throwing a footb all ) occurs within a range of force. These controls of fine or gross motor responses are examples of band discriminations of response force required to obtain reinforcement Although Notterman and Mintz described discriminations within a band, an initial step in training such discriminations is to train subjects to respond above a criterion level of force. Therefore, the purpose of this study was to appl y reinforcement contingencies to responding above a given force level Method Subjects, Setting, and Apparatus Two individuals, Erin and Nick, diagnosed with developmental disabilities participated (see demographic information in Table 2 1 ). Nick previousl y participated in Study 1 and Study 2. S etting apparatus measurement, reliability assessment, and preference assessment procedures were identical to those of previous studies. In
31 addition, data from sessions were separated into responses above and below threshold (responses that did and did not produce reinforcement respectively ). Interobserver agreement was assessed by having a second observer independently collect data during at least 25% of sessions. Proportional agreement percentages were calculat ed for each response as described previously. Mean reliability scores were as follows: Erin, 87% (range, 83.1 % 93.3%) for button presses and 89.2% (range, 80% -98.3%) for reinforcer deliveries; and Nick, 88.2% (range, 82.2 % -100%) for button presses and 94.7% (range, 81.7 % -100%) for reinforcer deliveries. Procedures Following the preference assessment, two conditions were conducted. In the FR1 condition (A), the preferred edible item was delivered following each response. When a stable rate and force of resp onding were established, a criterion of a 100% increase in response force was determined based on the greatest mean response force from the last three sessions of that condition. In the highthreshold criterion (B), the highly preferred edible item was del ivered for respon ses that were above the threshold (100% increase), while no consequences were delivered for responses below threshold. Finally, the FR1 condition (A) was reimplemented. Conditions were implemented in a reversal design (ABA). Results Fig ure 51 shows results of the band discrimination and reinforcement procedure on response force and frequency. In the initial FR1 condition, response rate for above threshold responding (i.e., above .07 lbs) was stable for both subjects. No below threshold responses were observed for Nick and a lower number of sub-threshold responses than threshold responses were observed for Erin. When the highthreshold
32 requirement was implemented, a similar rate of above threshold responses was observed for both subjects In addition, sub -threshold responses decreased over the course of this condition for both subjects. The FR1 condition was then reimplemented for both, during which response rate was similar to that observed in the initial FR1 condition and no sub-thresho ld responses were observed. In the initial FR1 condition, the peak force was stable for both subjects. When the highthreshold requirement was introduced, peak force became more variable but, by the end of the condition, peak force was well above the force requirement. Peak force in the reimplementation of the FR1 condition was greater than previously observed in the initial FR1 condition. Positive covariation between rate and force was observed in all conditions for both subjects. In the FR1 condition st able levels of rate and force were observed for both subjects. In the high -threshold condition, above threshold responses covaried with peak force, both in the initial variability and later stability observed. In the second FR1 condition rate and force wer e both stable and covaried together; however, responses occurred at a higher force. The results obtained demonstrate that increases in response force can be obtained by providing reinforcement for an increased threshold requirement. The results obtained i n the reimplementation of FR1 are of particular interest. Following a history of providing reinforcement for increased response force, the removal of this requirement did not result in immediate decreases in response force. This finding may be explained by the presence or absence of a consequence. That is, when a subthreshold response occurred in the high-threshold condition, no item was provided. However, when the subject responds more forcefully than required in the FR1 condition, a reinforcer is still
33 p rovided. Therefore, delivery of reinforcement provides information about the required threshold in one condition but not in the other.
34 Figure 4 1. Response rate and peak response force under conditions under a high threshold response-force requirement for Erin and Nick
35 CHAPTER 5 STUDY 4 : RESPONSE DIMENSION S IN PROBLEM BEHAVIO R Although several relations between response rate and force have been highlighted thus far, some of which may have clinical relevance, it is the assessment of problem behavior where force may be most useful as a measure of responding. As noted previously, force is probably the critical dimension of severity for problem behaviors such as aggression and SIB; however, almost all research on these behaviors has focused not on resp onse force but, rather, on response rate. Thus, it often is assumed that reductions in rate during treatment correspond to reductions in risk, but that is true only if there also is a reduction (or at least no change) in force. One particular area in whi ch the effect of rate changes on force may have immediate clinical impact is exposure of problem behavior to experimental conditions during assessment. Functional analysis (FA) methodology, as described by Iwata et al. (1982/1994) has become a standard me thod for identifying reinforcement contingencies that maintaining problem behavior. During an FA, consequences suspected of maintaining problem behavior are delivered following its occurrence in different test conditions, and a common finding is a higher r ate of problem behavior during one or more test conditions relative to a control condition. As a result, some authors have suggested that an FA, by definition, involves increased exposure to risk (Smith & Churchill, 2002). Whether rates of problem behavior during assessment actually are higher than those seen outside of the assessment context is unknown, but the concern in this study is the relation between rate and force during an FA. Basic research by Notterman (19 5 9 ) indicated that response force under continuous schedules of reinforcement (as in a standard FA) should decrease to the minimum force
36 required to obtain reinforcement, and this effect was observed generally in Study 1. For example, in both of Joses FR1 conditions, his response force decreased throughout the condition. The force (severity) of problem behavior has been assessed in applied settings (Newell, Incledon, Bodfish, & Sprague, 1999); however, force has not been assessed in the context of a standard assessment for problem behavior. Ther efore, research is needed to apply those procedures described by Newell et al. to assess the severity of problem behavior during the context of a FA. Method Subject and Setting Four individual s (Billy, Evan, Regan, and Erin) who were referred for the ass essment and treatment of aggression participated (see demographic information in Table 2 1 ). Evan participated in Study 1 and Study 2, and Erin participated in Study 3. Sessions were conducted in a therapy room at a public school for students with developm ental disabilities. Session materials included a table and two chairs, as well as materials appropriate to the FA condition. For example, in the attention and play conditions, moderately and highly preferred leisure items were present, respectively; in the demand condition, task materials were present. All sessions were 10 min in duration and were conducted 3 to 5 times per week. Items commonly found in the classroom (e.g., a clipboard) were placed around the room for Regina and Erin, so that the subjects h ad equal opportunity to engage in property destruction. No additional items were placed in the room for Brandon because his aggression usually took the form of throwing his shoes and clothing at the therapist. All sessions were videotaped in order to conduct a computer modeling analysis. Evan and Erin also participated in at least
37 one previous translational study. This was because we interested in observing the pattern of response force in an FR1 schedule for an arbitrary and problem behavior. Response Measurement and Reliability Trained observers collected data on the frequency of Billy and Evans aggression and Regan and Erins property destruction. Billys aggression was defined as grasping an object in one or both hands being thrown at he therapist (i .e., the object had to be throw n within .2 m of the therapist). Evans aggression was defined as forceful contact between Evans hand and the therapist (traveling from a distance of at least .2 m). Regans property destruction was defined as throwing objec ts more than .3 m (not directed at the therapist) or forceful contact between her palm and any object in the room (e.g., the wall). Erins property destruction was defined as throwing objects more than .3 m (not directed at the therapist). Interobserver agreement was assessed by having a second observer independently collect data during at least 25% of sessions. Proportional agreement percentages were calculated as described previously. Mean reliability s cores were as follows: Billy, 96.9% (range, 81.1 % 10 0%) for aggression; Evan, 99.1% (range, 96.7% 100%) for aggression; Regan, 97.9% (range, 95.8% 100) for property destruction; Erin, 91% (range, 96.7% 100%) for property destruction. Functional Analysis An FA as described by Iwata et al. (1982/1994) was co nducted to identify the source(s) of reinforcement for Billys, Evans, Reginas, and Erins problem behavior No interaction, Attention, Play, and Demand conditions were conducted in a multielement design with the order of conditions fixed as listed. In the No Interaction condition, the subject and a the rapist were present in the room, and no activities were present.
38 Throughout the session, the therapist never interacted with subject either prior to or following problem behavior. This served as a test for maintenance by automatic reinforcement. In th e Attention c ondition, the subject and the therapist were present in the room and a moderately preferred leisure item was present to which the subject had free access If the subject engaged in problem behavior, the therapist delivered a mild verbal reprimand (dont do that, you will hurt me), issued a statement of concern, and provided gentle physical contact. This served as a test for maintenance by social positive reinforcement. In the Play condition, the subject and the therapist were present in the room and a highly preferred leisure item was present to which subject had free access. T he ther apist provided brief statements to the subject related to the leisure item every 30 s or anytime the subject initi ated conversation. This served as the control condition. In the Demand condition, the subject and therapist were present in the room with various academic tasks. The therapist presented tasks continuously using a three step prompting procedure (vocal, model, physical prompt ). If the subject engaged in problem behavior the therapist removed the task materials and provided a 30 s break from tasks. This served as a test for maintenance by social negative reinforcement. Video Modeling Analysis All FA sess ions were videotaped for further analysis. Following each session, the video recording was divided in 40-s segments to limit the size when delacing The video was then delaced (each second of video was split into eight individual frames) and then uploaded to a kinetic movement computer program (HuMAN motion system) to generate computerized models and calculate the angular velocity of each response. Angular velocity is the speed at which a joint rotates in movement. Greater rotation per
39 second indicates a m ore forceful response. Angular velocities are a common means to assess the force of movement in biomechanics. (Nigg, MacIntosh, & Mester, 2000). We selected the HuMAN motion system rather than the Peak Motus system (used by Newell et al., 1999) because t he Peak Motus system is prohibitively expensive (approximately $7,000). In comparison, the HuMan system retails for around $40, allows users to enter their own video segments, provides training on use of the system, and generated the angular velocities required for this study. The use of HuMAN system also may make it more likely that clinicians and researchers would be able to replicate analyses of this sort. In order to create the computer models, two frames of action were selected from each instance of t he target response (i.e., each instance of aggression for Billy and Evan, and each instance of property destruction for Erin and Reagan). Because each second of video was split into eight frames, and individual responses varied in duration, several frames of each instance of the target response were available for analysis (e.g., from cocking the arm to releasing the item). Thus, we selected the frame that recorded the exact moment at which the response met the definition for the target behavior (i.e., the m oment at which a thrown object left the subjects or the subjects hand made contact with the therapist) as well as the frame immediately prior to that frame; the selection of these frames provided the best comparison of response force across instances. Th e HuMAN motion system used the selected frames to create a computer generated model of the angular velocity (mean degrees of rotation per second) at which specific joints (the shoulder and the elbow) were moving. In order to accomplish this, we selected sp ecific body parts to track over time. For the elbow angular velocity we
40 tracked the shoulder, elbow, and hand. For the shoulder angular velocity we tracked the hip, shoulder, and hand. Following each session, the mean angular velocity for each joint was calculated by totaling all of the velocities and dividing by the number of responses. Results Figure 5 1 shows the results of Billy and Evans FA. The left panels show the rate of aggression in the four assessment conditions. A higher rate of aggression wa s observed in the Attention condition, indicating that both Billys and Evans aggression was maintained by social positive reinforcement in the form of attention. The right panels again show the rate of aggression, but only in the Attention condition, plotted with the mean angular velocity of two joints (elbow and shoulder) of Billys and Evans aggression during Attention sessions. The angular velocity data show that the speed of the aggressive acts decreased through the course of the FA despite increases in the rate of aggression for both subjects For Billy, th ere was a slight increase in the force of the shoulder movement; however, there was also a sharp decrease in the force of elbow movements. By the final session, the force of elbow movements was less than 50% of the force in the initial Attention session Topographically, Billy initially was throwing items in an overhead fashion and using both his elbow and shoulder to generate the aggressive act. Through the course of the assessment, Billy continue d to use his shoulder to throw items, but decreased the degree to which he was flexing his elbow when engaging in aggressive acts. Because both the elbow and shoulder are involved in all throwing motions, these data indicate that the force of aggression de creased over the co urse of the assessment For Evan, both elbow and shoulder velocity increased initially and then decreased over the course
41 of the assessment. Topographically, Evan was initially slapping the therapist in an overhand motion that incorporat ed both his elbow and shoulder. As sessions progressed, Evan continued to use his elbow and shoulder; however, the degree to which he flexed these joints and the speed with which he moved the joints decreased. Because of this, the severity of his aggression decreased. Figure 52 shows the results of Regans and Erins FA. The left panels show the rate of property destruction in the four assessment conditions. A higher rate of property destruction was observed in the Demand condition, indicating that both R egans and Erins property destruction was maintained by social negative reinforcement in the form of escape from demands. The right panels again show rate of property destruction, but only in the Demand condition, plotted with the mean angular velocity o f two joints (elbow and shoulder) of Regans and Erins property destruction during the Demand sessions. The angular velocity data show that the speed of Regans property destruction decreased through the course of the FA despite increases in the rate of p roperty destruction However, this decrease in force was not as pronounced as for Billy and Evan. Topographically, Regan was throwing items in an overhead fashion and using both her elbow and shoulder to engage in property destruction through the course of the assessment and continued to use both the joints at a similar level, though the general range of action decreased. T hese data indic ate that the force of property destruction decreased over the course of the assessment. The force of responses did not de crease for Erin. The angular velocity of her joint movements increased in some sessions but did not increase significantly across the course of the assessment. At the same time, the frequency of her problem
42 behavior was variable. It should be noted that an gular velocities were generally low across Demand conditions for Erin even when a large number of responses occurred, and the severity of responding never increased to the point where any object was broken or damaged. Topographically, Erin threw items in b oth an overhead manner and in a pushing motion. In the 3rd and 5th sessions, she began to throw objects more forcefully in an overhead manner. Because of this, the severity of property destruction increased in these sessions; however, no real change in re sponding was observed.
43 Figure 5 1. Response rate (left and right panels) and rate of arm rotation (right panel) during the f unctional analysis of Billy and Evans aggression.
44 Figure 5 2 Response rate (left and right panels) and rate of arm rotation (left panel) during the f unctional analysis of Regan and Erins property destrcution.
45 CHAPTER 6 DISCUSSION The present studies examined the relation between response rate and force in a series of translational contexts (Studies 1 -3) and in one applied co ntext (Study 4 ). Taken together, results of the studies identified several relations between response rate and force, which may be informative when teaching appropriate behaviors and treating problem behavior. Although the applied aspect of the present r esearch (Study 4) focused on force as it relates to problem behavior, force is equally relevant to the study of adaptive performance of clinical populations. For clinical populations, assessments of response force might be necessary for training individual sports (hitting the tennis ball at the appropriate speed, shooting a basketball, etc.), social skills (voice volume, handshake strength, etc.), and vocational tasks (piecemeal work, typing, etc.) A number of individuals with developmental disabilities are employed in piecemeal work or other activities that require fixed chains of behavior and varying levels of response force. Simplified chains of behavior could be designed for the benefit of those employed. That is, decreasing the number of movements to ob tain reinforcement, as well as determining and training an ideal amount of force during each action, might increase productivity. In this manner, errors made because the response force is either too high (breaking an item) or too low (insufficient force to complete the task) could be identified. Studies 1 and 2 examined the effects of manipulations that typically influence the rate of behavior. Specifically, Study 1 examined the rate and force of button press es under conditions of continuous (FR 1) reinfo rcement and extinction. Results of this study suggest that the relation between response rate and force is not straightforward.
46 In the FR 1 condition, some individuals (Evan and Karen, and in the second implementation of FR1 for Nick) response rate and fo rce covaried positively However, in the FR1 condition, some individuals (Jose and in the first and third implementations of FR1 for Nick) response rate and force covaried negatively. The negative covariation between response rate and force has been demon strated in a number of studies (Notterman, 1959; Notterman & Mintz, 1965), and that an organism would find the minimum response requirement to obtain reinf orcement seems intuitive. T his was not found for some subjects in Study 1 and Study 2, which is notab le. There are several possible explanations for this finding. The first is a species -specific difference. It is possible that humans as a species are less sensitive to identification of the minimum response force requirement than rats, although it is not c lear why this would be the case from an evolutionary standpoint. Additionally, some of the human subjects did seem sensitive to force as a contingency (Study 3); therefore, an explanation in personal learning history is more likely. In experiments where r ats have been used as subjects, these animals are often nave to the experimental situation prior to the experiment. Although none of the human subjects had a history of pressing a button to obtain reinforcement, the subjects were selected from a special education school. Thus, it is likely that in the subjects recent history, edible items may have been provided for academic tasks. These academic tasks most likely took the form of reinforcement provided for a rate of responding. For example, one might imag ine a scenario where the individual earns an edible item for completing math problems. Here, the number of problems completed, rather than the force with which the individual grips the pencil or presses down on the paper, is
47 reinforced. It is therefore lik ely that a history of receiving reinforcement exists for responding more frequently (contingency on frequency rather than force). Additionally, it is equally unlikely that reinforcement has been associated with increased response force, except on a very general level (i.e., sufficient pressure to make a mark with a pencil). Furthermore, when response force might be an important variable (e.g., increasing an subjects exercise behavior), it is likely that reinforcement is provided for a corollary measure (e.g., distance) because of ease of data collection. Therefore, it is possible that each subject had an idiosyncratic history of reinforcement associated with response force, which would explain the differences observed in covariation. This point was highlig hted by Nicks data, where at the start of the experiment he was relatively insensitive to the minimum response force contingency, but in the final FR1 phase he began to engage in less effort through the course of the condition. Here, response force decrea sed as has been observed with other species in FR1 schedules. It is also possible that human subjects are somewhat sensitive to contingencies placed on response force, but that differential EXT is a necessary component for subjects to attend to this featur e of their responding. That is, although subjects did not decrease response force in FR1 conditions, they did respond above the criterion when EXT was implemented for low levels of response force. In addition, the finding that EXT bursts of response force are more likely than EXT bursts of rate is significant and has direct implications for the clinical treatment of severe problem behavior. EXT is a common and effective procedure for reducing severe problem behavior exhibited by individuals diagnosed with developmental disabilities (Lerman & Iwata, 19 95 ). However, this procedure has also been associated
48 with undesirable side effects such as increases in the rate of problem behavior, extinction induced aggress ion, and emotional responding (Lerman & Iwata) To this list of potential side effect s, an increase in the severity of problem behavior exhibited by the individual should also be considered. The results obtained in Study 1, as well as results from the Notterman (1959) and Notterman and Mintz (1965) studies, demonstrated that an increase in response force is likely in EXT. This effect however, has never been demonstrated with problem behavior. This may be because of difficulties in measuring the severity of problem behavior ; the meth ods described in Study 4 provide a system through which the severity of problem behavior in treatment (e.g., EXT) might be measured. The addition of reinforcement procedures for appropriate behavior to an EXT procedure may make an EXT burst in frequency less likely. However, it is not clear that the addition of a reinforcement procedure for appropriate behavior would also decrease bursts in response force. In addition, bursts in the force of problem behavior may be more relevant as they may be more likely t o cause significant damage to the individual. The procedures described in Study 4 could examine this possibility as well. Notterman and Mintz (1965) suggested two explanations for an increase in response force during extinction, which they observed every time that extinction was implemented following FR1. One explanation was biological in nature: For evolutionary reasons, unsuccessful responding might be rendered more effective if it occurs more vigorously. The y related this explanation to the concept of frustration and provided the example of a squirrel attempting to drag a nut to storage. Here, the increase in response force has the biological advantage of eventually providing reinforcement. T he authors also discussed reinforcement as a function of for ce from a learning perspective
49 noting that in any experiment with animals, several dimensions of a response are in flux. Specific to the experiments conducted, rats engaged in several types of lever presses to obtain reinforcement which the authors desc ribed as more or less of a response -sampling repertoire that narrowed based on exposure to a contingency Applying t his explanation to results observed in reinforcement conditions, the authors noted that response force decreased because more forceful respo nses were not more effective in obtaining reinforcement. When rats subsequently were exposed to extinction, this sampling of response magnitudes was observed again. Thus the pattern of response force during acquisition and extinction was similar, but reversed. Study 2 examined the effect of ratio schedules of reinforcement on the rate and force of button press responses and results of this study also suggest ed that the relation between rate and force of a response is not straightforward. Results obtai ned during the FR10 condition were mixed. There was no difference or only a slight difference in response force for Lisa and Nick when a larger number of responses was required to obtain reinforcement. A larger difference in response force was observed for the other two subjects (Evan and Karen). In addition, these subjects also had decreasing levels of response force in the FR1 condition. As noted previously, an increase in response force under higher ratio schedules seems counterintuitive. As the number o f responses required to obtain reinforcement increases, response force might be expected to decrease because the organism is required to emit more behavior ; thus, reductions in force conserve energy However, this has not been observed in basic research. O nly two studies have measured response rate and force under ratio schedules (Notterman & Mintz, 1965; Mintz, 1962). Mintz conducted the more relevant
50 of these studies as it more closely approximated the procedures used in Study 2. Mintz examined FR12 sche dules of reinforcement in rats that were earning a sucrose solution. He fou nd that rats engaged in progressively more forceful responses until rei nforcement was obtained. Although these data were not presented, this pattern of responding would also increas e the mean peak force. Therefore, it is possible results obtained for Karen and Nick might be related to this effect Finally, three of the four subjects ( Evan, Karen, and Nick) showed increased variability in response force in the FR10 condition. Covariat ion (either positive or negative) between response force and response rate has been demonstrated in the previous Study 1 and this study in FR1 conditions. When positive covariation was identified, the force of response seemed to move with the response rate. This covariation also occurred in the FR10 condition. D emonstrations of how response rate and force change as a function of changes in the response ratio requirement to obtain reinforcement might shed light on several behavioral processes. An immediate increase in mean response force was observed for three subjects (Evan, Karen, and Nick) in the FR10 condition of Study 2 In this way, FR10 more closely approximated the effects of EXT than FR1, at least initially. The FR10 condition is similar to the EXT condition in the sense that initial responses in the schedule are not reinforced. Although reinforcement is eventually forthcoming in the FR10, initial responses go unreinforced. Therefore, it is likely that this schedule would increase response force for some individuals. A comparison of FR and fixed interval (FI ) schedules of reinforcement in Study 2 might have provided additional information. In FI schedules of reinforcement a lower response rate would have been expected. This lower rate then could be compared with
51 those results produced by an FR schedule of reinforcement. Furthermore, a comparison of FR and variable ratio (VR) schedules of reinforcement might have been interesting. In VR schedules of reinforcement, the delivery of the reinforcer is based on the average of the schedule value; reinforcer delivery would have been less predictable. This may have shed light on the underlying mechanisms for the increase in response force for some individuals. That is, the ordered increase in response force across the number of responses to obtain reinforcement may be disrupted when reinforcers are not delivered after a constant number of responses. Indeed, no research currently available has assessed response force in the context of VR schedules of reinforc ement and future research should assess this type of schedule. Study 3 examined the effects of reinforcing a specific band of response force. B oth subjects showed stable and moderate response rate s and low levels of peak force in FR1. When subjects were required to respond at the high-threshold, response rate (both below and above threshold) became more variable, and a significant increase in response force was observed Additionally, when the second minimum response FR1 condition was implemented, more fo rceful responding was observed for both subjects. These data may have clinical relevance to response acquisition. For example, measuring response force may be of applied value i n exercise. Sargeant et al. (1981) assessed exercise behavior through examining crank velocities on an exercise bike. A standard feature of many stationary exercise bikes is a watts reading. One could use the procedure described in Study 3 to increase energy output by reinforcing higher levels of exercise force and longer durations o f exercise. In this way, exercise
52 responding could be based on sustained caloric output rather than distance, a measure in which caloric output may fluctuate. One limitation o f Study 3 was in the design of the apparatus which ma d e an audible click when .07 lbs of pressure were applied to close the switch. This became problematic during reinforcement of higher levels of response force because in the initial CRF condition, the audible click was associated with forthcoming reinforcement. This was n o longe r the case in the high -threshold condition. That is, some response s that produced a click did not produce reinforcement. Morris (1968) used an apparatus that only produced reinforcement when t he lever was fully depressed. A spring controll ing the amount of pressure required to depress the lever could be adjusted to a threshold level. This apparatus was ideal because the signal that reinforcement was forthcoming was the act of depressing the lever. This was not the case in the present study. It is possible t hat some variation of response force observed in the highthreshold condition was related to the fact that a previously paired stimulus was no longer paired with reinforcement. F uture researchers assessing methods to increas e response force should use appa ratuses that provide either no exteroceptive cue that reinforcement is forthcoming, or a cue on ly when reinforcement is forthcoming. Another limitation of Study 3 was that the effects of EXT were not examined following training on high and low -threshold force requirements. Although the effects of EXT following high -threshold training would likely be identi cal to those observed in Study 1 (i.e., an EXT burst in rate and force) it is not clear what effect would be observed during EXT following low thresho ld training. Because the subject had a history of reinforced responding, an EXT burst might occur. However, the most immediate
53 history would be reinforcement at a low level of force, which may engender responses of minimal force when reinforcement is no lo nger forthcoming. Additional translational research is also required in examining the choice between responses when the parameters of reinforcement (rate, quality, immediacy, and magnitude) are held constant and only the amount of force required to engag e in the behavior differs. However, no cohesive definition of response effort exists in the field of behavior analysis. For example, in a basic study, Baum (1982) examined the distance that pigeons traveled to press either of two buttons associated with concurrent VI schedules. The requirement for the VI schedules increased every time the organism obtained reinforcement on each button. In addition, the length of a partition between two buttons was increased such t hat switching between the two buttons was more or less effort ful depending on th e partitions length. The author characterized switching as related to the effort to obtain reinforcement. In an applied study, Ringdahl et al. (2009) examined the effects of training three types of communication respons es for three subjects with problem behavior maintained by access to tangible items. The authors trained the subjects to engage in three alternative responses (a switch press, a hand sign, & a card). During training, the authors collected data on the number of trials to acquire each response. The authors defined proficiency as the number of training trials in which the subject correctly emitted the communication response. The authors found that when subjects were highly proficient with a communication system this communicative response was effective at treatment. In this study, effort was conceptualized as the subjects proficiency with each response. These two studies demonstrate that response effort has come to define a number of disparate
54 manipulations in behavior analysis. Therefore, the use of a load cell, which produces a quantitative measure of effort (in the amount of response force), is ideal for assessing the role of effort in choice on a translational level. Study 4 evaluated a computer modeling approach to the measurement of severity of problem behavior in the context of assessment. Results of this study were significant in two ways. First, the study demonstrated a method for conducting a more through analysis of problem behavior than is obtained in clinical settings. Although time consuming, the technology could be helpful in quantifying risk during assessment or success/failure during treatment when problem behavior occurs at varying intensities. For e xample, initial probes of a treatment could be conducted to determine whether alternative or supplementary procedures are needed to minimize the risk of potentially severe problem behavior. Study 4 also demonstrated that problem behavior did not become more severe over the course of the assessm ent of problem behavior. The force of Billys, Regans, and Evans problem behavior actually decreased over the course of the assessment, even while their rate of problem increased. These results correspond with previous findings for rats lever responses exposed an FR1 schedule of reinforcement (Notterman, 1959). The force of Erins problem behavior remained stable over the course of the assessment, even while the rate of her problem behavior was variable. Thus, increase s in the frequency of individuals problem behavior per se during assessment do not necessarily increase risk of injury. Most research on response force indicates that response force decreases in FR1 schedules (Notterman, 1959; Notterman & Mintz 1965). All of the experiments
55 conducted i n the present study included at least one FR1 condition, providing a total of 24 FR1 implementations. Of the 24 FR1 conditions, a decreasing trend in force was observed in 12 implementations, and a stable trend was observed in the other 12. Two subjects (E rin and Evan) participated in at least one translational and one applied study, both of which included the implementation of an FR1 schedule. Erin showed a stable trend in all FR1 conditions in Study 3 and Study 4. Evan showed a decreasing trend in three o f the four implementations of FR1 in Study 1 and Study 2 and a decreasing trend throughout Study 4. In these two cases it appears that the results of FR1 implementations conducted with an arbitrary response predicted the results when FR1 was implemented wi th problem behavior. In summary, results of the present studies suggest that the force of a response, which often is taken for granted in both basic and applied research, deserves further study as a dimension of behavior that it important in its own right Furthermore, the technology exists to study this relation in different settings. Relations between the rate and force of human responding need to be examined more thoroughly such that better predictions can be made of changes in force during both respons e acquisition and reduction procedures. Equally important, attempts should be made to measure response force more directly during the course of assessment and intervention when changes in force may be either beneficial or problematic.
56 Table 2 1. Subject c haracteristics Name Age Classification Communication Skills Definition of Target Behaviors Evan 8 Autism No verbal or symbolic language Study 1 & 2: Depressing a button so a click is heard Study 4: F orceful contact between Evans hand and the therapist traveling from a distance of at least 6 inches Karen 14 Autism & Mental Retardation PEC sentence strip Depressing a button so a click is heard Jose 14 Autism & Mental Retardation Three to four word sentences Depressing a button so a click is heard N ick 19 Autism No impairment Depressing a button so a click is heard Lisa 22 Autism No impairment Depressing a button so a click is heard Leo 15 Autism & Mental Retardation Three to four word sentences Depressing a button so a click is heard Regan 14 Autism & Mental Retardation One word sign language T hrowing objects more than 3 inches (not directed at the therapist) or making forceful contact between her palm and any object in the room Erin 15 Down S yndrome One word sign language Study 3: Depressi ng a button so a click is heard Study 4: Throwing objects more than 3 inches (not directed at the therapist) Billy 11 Autism One word sign language Throwing an object within .5 m of the therapist
57 LIST OF REFERENCES Adair, E. R., & Wright, B. A. (1976). Behavioral thermoregulation in the squirrel monkey when response effort is varied. Journal of Comparative and Physiological Psychology, 90 179184. Bailey J. S, & Burch M. R. (2002). Research methods in applied behavior analysis. Thousand Oaks, CA: Sage; 2002. Baum, W. M. (1982). Choice, changeover, and travel. Journal of the Experimental Analysis of Behavior, 38, 3549. Cooper, J. O., Heron, T. E., & Heward, W. L. (2007). Applied Behavior Analysis. 2nd Edition. Upper Saddle River, New Jersey: Pearson Ferster, C. B., & Skinner, B. F (1957). Schedules of reinforcement New York: Appleton Century -Crofts. Greene B. F., Bailey, J. S., & Barber, F. Analysis and reduction of disruptive behavior on school buses. Journal of Applied Behavior Analysis 14, 177-192. Iwata B. A., Dorsey M. F., Slifer K. J., Bauman K. E., & Richman G. S. (1994). Toward a functional analysis of self -injury. Journal of Applied Behavior Analysis 27, 197 209. Iwata, B. A., Pace, G. M., Kissel, R. C., Nau, P. A., & Farber, J. M (1990). The Self Injury Trauma (SIT) Scale: a method for quantifying surface tissue damage caused by self -injurious behavior. Journal of Applied Behavior Analysis 23, 99110. Jackson, D. A., & Wallace, R. F. (1974) The modification and generalization of voice loudness in a fifteenyear old retarded girl, Journal of Applied Behavior Analysis, 7 461 471. Lerman, D. C., & Iwata, B. A. (1995). Prevalence of the extinction burst and its attenuation during treatment. Journal of Applied Behavior Analys is, 28 9394. Lerman, D. C., Iwata, B. A., & Wallace, M. D. (1999). Side effects of extinction: Prevalence of bursting and aggression during the treatment of self -injurious behavior. Journal of Applied Behavior Analysis, 32 1 8. Morris, J. P. (1968). Changes in response force during acquisition and extinction in retarded children. American Journal of Mental Deficiency 73, 384390. Michael, J. (1995). What every student of behavior analysis ought to learn: A system for classifying the multiple effec ts of behavioral variables. The Behavior Analyst 18, 273-284.
58 Newell, K. M., Incledon, T. Bodfish, J. W., & Sprague, R. L. (1999). Variability of stereotypic body -rocking in adults with mental retardation. American Journal on Mental Retardation, 104, 279-288. Newell, K. M., Challis, J. H., Boros R. L., & Bodfish, J. W. (1999). Further evidence on the dynamics of self -injurious behaviors: Impact forces and limb motions. American Journal on Mental Retardation 107, 6068. Nigg, B. M., MacIntosh, B. R., & Mester, J. ( 2000 ). Biomechanics and biology of movement Champaign, IL: Human Kinetics. Notterman, J. M. (1959). Force emission during bar pressing. Journal of Experimental Psychology, 58, 341347. Notterman, J. M. & Mintz, D. E. (1965). Dynamics of R esponse. United States: John Wiley & Sons, Inc., U.S.A. Ringdahl, J. E., Falcomata, T. S., Christensen, T. J., Bass -Ringdahl, S. M., Lentz, A., Dutt, A., & Schuh -Claus, J. (2009). Evaluation of pre-treatment assessment to select mand topographies for func tional communication training. Research in Developmental Disabilities 30, 330-341. Symons, F. J., Harper, V. N., McGrath, P. J., Breau, L. M., & Bodfish, J. W. (2009). Evidence of increased nonverbal behavioral signs of pain in adults with neurodevelopmental disorders and chronic self -injury Research in Developmental Disabilities 30, 521-528. Walton, D. & Black, D. A. (1959). The application of modern learning theory to the treatment of chronic hysterical aphonia. Journal of Psychosomatic Research 3, 303-311. Wilson, D. M., Iwata, B. A., & Bloom, S. E. (in press). Evaluation of a computer assisted technique for measuring injury severity. Journal of Applied Behavior Analysis. Wulbert, M., Nyman, B. A., Snow D. & Owen, Y. (1973). The efficacy of st imulus fading and contingency management in the treatment of elective mutism: A case study, Journal of Applied Behavior Analysis, 6, 435 441.
59 BIOGRAPHICAL SKETCH Griffin completed his Bachelor of Arts degree at Emory University in 2001 then took a position as a direct care staff at the New England Center for Children in Southboro, MA. The re he assessed and treated severe problem behaviors, such as self -injury, aggression, and proper ty destruction. At this time, he also received a Masters of Science degree with Dr. Eileen Roscoe. In 2006, he came to the University of Florida to pursue a doctoral degree in psychology specializing in behavior analysis. During his graduate training, he has been involved in research projects on asses sing data analysis methods, behavioral assessment methods for problem behavior, acquisition procedures and physiological c orrelates to problem behavior. He also served as a behavior analyst for 14 individuals (overseeing and implementing behavior management plans), provided behavioral services to students and teachers within a special education setting, and served as teaching assistant and primary instructor for introductory courses in applied behavior analysis and a laboratory class in behavior analysis. Following graduation, he became a post doctorate fellow at Johns Hopkins University working in the Neurobehavioral Unit of the Kennedy Krieger Institute.