Collecting and Analyzing Continuous Behavioral Data with the TRS-80 Model 100/102 Portable Laptop Computer (1989)
©1989 by Dallas Denny & James Fox
Source: Denny, Dallas, & Fox, James. (1989, Summer). Collecting and analyzing continuous behavioral data with the TRS-80 Model 100/102 portable laptop computer. Journal of Special Education Technology, 9(4), 183-189.
This article and the others in this issue of JSET was based upon a presentation Dr. Fox and I did as part of a symposium on electronic processing of behavioral data in Gatlinburg, Tennessee. Dr. Fox chaired the symposium and edited the resulting special issue.
Source: Denny, Dallas, & Fox, James. (1987). Collecting and analyzing sequential data with the TRS‑80 Model 100 and the Behavioral Observation Code Computerized (BOSCO). Twentieth Annual Gatlinburg Conference on Research and Theory in Mental Retardation and Developmental Disabilities, Gatlinburg, TN, 25‑28 March.
The behavioral data gathering and analysis program Behavioral Observation System, COmputerized is a program written in the BASIC programming language for the TRS-80 Models 100 and 102 portable computers. I am the sole author. It was used extensively by researchers at the Special Education Department at George Peabody College, which is a part of Vanderbilt University.
Guest Editor's Introduction
Source: Fox, James. (1987, Summer). Guest editor’s introduction. Journal of Special Education Technology, 9(4), 177.
Guest Editor’s Introduction
By James Fox
Vanderbilt University
Yogi Berra is reputed to have once remarked, “You can see a lot just by looking.” Although redundant or nonsensical on its surface, this statement is actually a nugget of truth for those of us who advocate and use direct observation in applied research and professional practice. As opposed to asking someone about their own or someone else’s behavior, one can discover much more useful information “just” by systematically looking and recording what is seen. Of course, there is much more to direct analysis of student behavior than simply looking. Operational definitions, standard recording procedures, observer agreement and accuracy checks, and materials to enable observers to apply definitions and procedures are Integral parts of “just looking.”
Interestingly, there have been relatively few changes in the technology of direct observation. Operationism is a long-standing concept in science. One of the basic and most frequently used procedures of direct observation, time sampling, can be traced back some 60 years to the pioneering work of Florence Goodenough and others from the golden age of child development research in the late 1920s and 1930s. In recent years, however, there has been increasing empirical interest in the evaluation of observational procedures such as the relative accuracy of time sampling versus whole and partial interval recording and the refinement of methods for the sequential analysis of behavior patterns.
This special issue of JSET addresses another area of observational technology: the development and use of laptop microcomputers and software as data collection devices. With the exception of certain very expensive and oftentimes unwieldy instruments, behavioral researchers have employed relatively simple recording devices—golf wrist counters, stopwatches, paper and pencil. Now the advances in microcomputer technology not only enable researchers to collect the most sensitive parameters of behavior (frequency, duration) at relatively low cost but to preserve the sequence of behavior-environment exchanges and quickly and efficiently analyze interobserver agreement and student behavior change. Not only are these hardware-software systems a methodological boon to researchers, they also have significant applications for practitioners such as school psychologists, consulting teachers, and classroom teachers themselves.
The first article, by Laurie Meltzer and Ted Hasselbring, provides an historical overview of microcomputer development and makes us aware of how far this technology has come in a relatively short time. In doing so that article also gives us an impression of how these technological advances can and will affect researchers and educators in the future. The four articles that follow can be viewed in two ways. On the one hand they can be used as a catalog of available hardware-software combinations from which one may select according to his or her research or application needs. One can also abstract from them a common basic technology that is still evolving and through its continued evolution promises to further reduce some of the difficulties of using human observers as transducers of behavioral events.
In closing I would like to personally thank each of the authors for their contributions to this issue. I would especially like to thank Herbert Rieth for his support and encouragement in putting this issue together and my colleague, Mary McEvoy, for her help in editing the manuscripts.
James Fox
Collecting and Analyzing Continuous Behavioral Data with the TRS-80 Model 100/102 Portable Laptop Computer
By Dallas Denny
And James Fox
Department of Special Education
George Peabody College of Vanderbilt University
Abstract
Abstract
During observation sessions, discontinuous recording of behavioral events (time sampling and interval recording) introduces sampling error into the data. Continuous recording of data is methodologically superior, but makes greater demands on observers and often requires expensive and cumbersome equipment. There is a need for equipment and software that are easy to use and transportable and that integrate data collection and analysis functions. The Behavioral Observation System—COmputerized (BOSCO) was developed for the TRS-80 Model 100 and 102 portable laptop computers. BOSCO allows the computer to be used to (a) create and use an observation system, and (b) assess and calculate interobserver agreement. The computer’s built-in, real-time clock allows accurate timing of each entry to within one second. This article describes the features and functions of BOSCO, illustrating its application through brief descriptions of several studies in which it has been employed.
During observation sessions, discontinuous recordings of behavioral events (time sampling and interval recording) introduces sampling error into the data (Johnston & Pennypacker, 1980; Sackett, 1978). “Such sampling error is a major contributor to measurement-induced variability, which then becomes thoroughly confounded with treatment-imposed variability, severely compromising the accuracy and generality of experimental interpretations” (Johnston & Pennypacker, 1980, p. 149). Continuous recording of data is methodologically superior to discontinuous collection, but makes greater demands on observer energy and concentration and often requires expensive and cumbersome equipment (Sackett, 1978).
Currently there are portable electronic devices available that allow the collection of continuous observational data (Sackett, 1978). The OS-3 by Observational Systems, Inc., is an example of one such device (cf. Bakeman & Gottman, 1986). Data from these “electronic event recorders” typically is transferred to a desktop computer for editing and analysis. Also available are programs for desktop computers that permit data collection and detailed analyses of the data (e.g., Semmel & Frick, 1980), and others that calculate agreement between observers (MacLean, Tapp, & Johnson, 1985).
However, there is a need for software and hardware that can integrate these and other functions yet remain truly portable, reasonably priced, and flexible in their application. For example, observational data collection typically occurs in situations that require that observers and their equipment remain mobile. At the same time it is often desirable (sometimes necessary) to edit data or obtain results of analyses and checks of inter-observer agreement on-site without having to resort to a desktop computer, which may be many miles away from the actual research site. This may be particularly true in single-subject design studies because decisions about changes in experimental conditions are made on the basis of data trends rather than a priori criteria. When they do not allow relatively immediate access to the data, even electronic event recorders can serve to “distance” the experimenter from his or her data (Martin & Bateson, 1986). Inductive reasoning, a valuable tool of the behavioral researcher, is impeded when data are not readily available for inspection.
“Stand-alone” programs for portable laptop computers are beginning to appear, as witnessed by this special issue. Our purpose is to briefly describe and illustrate the use of one such integrated system, the Behavioral Observation System Computerized (BOSCO) and the Radio Shack TRS-80 Model 102, which we have been developing and using over the last several years. Its advantages and disadvantages over other, more traditional, paper-and-pencil and electronic devices will also be addressed.
The Computer
The TRS-80 Model 102 portable laptop computer (and its nearly identical predecessor, the Model 100) is an inexpensive battery-operated microcomputer that can be purchased and serviced at Radio Shack stores throughout the world. The Model 102 is available with up to 32K (32,000 bytes) of random access memory (RAM). Second-party suppliers offer plug-in devices that can increase RAM to one megabyte and more. The built-in software includes the BASIC programming language and text editing program. In addition, there are a built-in modem and telecommunications software that allow communication with other microcomputers either directly through the RS232 port or at a distance through telephone lines. The Model 102 can be easily interfaced with disk drives, cassette tape recorders, printers, bar code readers, and other peripheral devices.
Factors that influenced our selection of the Model 102 over other available portable microcomputers included the popularity and widespread software base, ready availability, comparatively low price, small size (12 x 8 x 1.75 in.), low weight (about 4 lbs), battery-powered operation, full-size keyboard and embedded key pad, large LCD screen, BASIC language, real-time clock and calendar, built-in modem and software, and ability to easily interface with a variety of equally portable peripheral devices.
The Software
BOSCO (the Behavioral Observational System—COmputerized) is a series of BASIC programs, SETUP, BOSCO, and RELY, that allow the Model 102 to be used, respectively, for: (a) the creation of an observation system, (b) collection and analysis of continuous observational data, and (c) calculation of interobserver agreement. Each program takes up between 6 and 12K of the memory of the Model 102 (total of 32K conventionally expanded). Once loaded from disk or tape, running the programs simply involves moving the cursor over the desired program listing on the main menu and pressing the ENTER key. The programs remain in memory until erased or until battery power is lost. The computer’s built-in nickel-cadmium battery preserves the memory even when the computer is shut off, and it will retain its charge for several weeks.
Using SETUP to Create an Observation Code
As with the other two programs, SETUP is loaded into the Model 102’s RAM from either disk or cassette tape. Figure 1 shows the main menu of SETUP. Through the “Define Behaviors” option, SETUP enables code numbers and brief labels to be assigned to as many as several hundred mutually exclusive behavioral categories. One simply types in the behavioral category label in response to the on-screen prompts. SETUP also allows system parameters to be set—bell on/off, dump file on/off, printer codes, and maximum number of observations/sessions. The coding system can be easily modified through this program also. Once a coding system is defined, it is saves as a TEXT (.DO) file in the computer’s RAM, using the “File Operations” option; SETUP then can be erased from the computer’s memory to create additional space for subsequent observation files.
Figure 1: BOSCO SETUP Program Menus (PDF)
Conducting an Observation with BOSCO
Once loaded into RAM, BOSCO is initiated by moving the cursor over the name BOSCO and pressing enter. On-screen prompts then direct the observer through several observation menu options. The range of options is shown in Figure 2, which presents the main BOSCO menu. The HEADER option allows the observer to enter descriptive information about the particular observation session—subject’s and observer’s names, date, time of the observation, etc. Actual observation is begun by selecting the ENTER DATA option from the menu. Observations are then entered by keystrokes, using the embedded key pad to enter the code number corresponding to the behavior observed. The code of the event becomes associated with the clock time of the observation. The event is assumed to continue (thereby accumulating clock time) until the code for the next event is entered or until a STOP code, created by the investigator under SETUP~ is entered. As currently structured BOSCO allows the recording of several hundred (299) behavioral events and their associated times within a single observation~ file. During observation the screen displays several important pieces of information in addition to the last code entered; an example of this screen and its information are presented in Figure 3. When the observation is completed (or the maximum number of events has been entered), it is necessary to save the data as a TEXT file (.DO) to the computer’s RAM storage area. Again, this is done by selecting the appropriate choice from the main BOSCO menu. Data files can later be loaded back into the program from this menu for purposes of examination, modification, or analysis. A maximum of 18 files can be stored in RAM before it is necessary to save them to a disk or tape. Of course, the number of files is also restricted by currently available RAM.
Figure 3: BOSCO Screen Display During Observation
Analyzing data with BOSCO and RELY
Data can be analyzed and presented on the LCD screen or a paper copy can be generated with the connection of a printer. As shown in Figure 4, there are three data analysis programs provided in BOSCO. Raw data analysis provides a printout of each code entry by (a) code number, (b) ordinal position in the data series, (c) time of onset, and (d) number of seconds each entry was “on.″ Frequency analysis shows for each code (a) its total frequency, (b) total time (or duration) in seconds, (c) mean time in seconds, and (d) total percentage of the observation time. Finally, Log 1 sequential analysis shows the frequency and percentage of occasions that each recorded code was (a) preceded and (b) followed by all other codes.
A separate program, RELY, examines and calculates interobserver agreement between the files of two independent observers. This program must be loaded separately from BOSCO, and both observation files for comparison must be loaded into the same computer. As with BOSCO and SETUP, RELY is initiated by first placing the cursor over RELY on the main menu and pressing enter. A series of on-screen prompts then asks that the names of the two observation files be entered and, next, that the user specify the window of time over which each entry will be compared. This agreement window is adjustable by the user from one second to several hours. Calculation of observer agreement is done using the point-by-point agreement formula (Kazdin, 1982). A print-out of each code entered by the primary observer is obtained, indicating agreement or disagreement with the second observer for that entry; an overall summary percentage of agreement for that session is also calculated and displayed or printed.
An option of the BOSCO program allows the separate recording of all keystrokes in a “dump” file. Each time the Enter key is pressed, the time is recorded. In this way, it is possible to accurately record and time events that would not be possible to record within the framework of the defined coding system.
Figure 4: BOSCO Data Analysis Subprogram
Applications of BOSCO and the Model 101/102 Computer
Throughout its evolution BOSCO has been applied in a number of research studies at Peabody College. Next, we will briefly describe several of these studies to illustrate BOSCO’s actual and potential applications.
One focus of our research group has been the social development and peer interactions of handicapped children. Social delays may consist of a lack of initiations, inappropriate or ineffective initiations, lack of responsiveness to another’s initiations, skill deficits in maintaining interactions once they have begun, or some combination of these problems. Thus, we needed an observation system that would allow us to identify such things as who initiated an interaction (the handicapped subject or a peer), the type of initiation used, whether or not initiations received a response, and how often they led to some sustained interaction between the children.
Although paper-and-pencil data collection methods had been used in prior studies (Fox, Shores, Lindeman, & Strain, 1986; Hendrickson, Strain, Tremblay, & Shores, 1982; Tremblay, Strain, Hendrickson, & Shores, 1981), there were several problems facing the continued use of these techniques. For example, paper-and-pencil methods required observers to periodically shift their attention from the subject to the paper and pencil in order to record behavioral events. This, combined with the time it took to actually write down coding symbols, resulted in potential gaps in the observed behavioral flow and in loss of data for analysis. Also, paper-and-pencil procedures seem less efficient in actual recording of continuous sequential observation. Without concurrent use of other recording devices (e.g., stop watches), there was no way to directly and efficiently record both the frequency and the duration of social behaviors and interactions, Finally, paper-and-pencil measurement procedures seem less efficient in the resultant data analyses. The tabulation of daily frequencies of behaviors, their duration, and the proportion of time certain initiations resulted in social responses and interactions were integral to our research questions; yet such analyses were very labor-intensive and time-consuming when done by hand. Use of BOSCO and the Model 102 obviated these difficulties.
In a recent study (Savelle & Fox, 1988) BOSCO was used to record and analyze the effects of different types of initiations on autistic children’s social responsiveness. Nonhandicapped age peers were taught to increase both reciprocal (e.g., sharing and verbal invitations to play) and nonreciprocal (statements, vocal attention-getting behaviors) initiations toward two autistic subjects. Changes in social responsiveness were evaluated in three ways: (a) whether or not the autistic subjects made an initial positive response to either type of initiation, (b) if and for how long the children continued to interact following an initiation and response, and (c) whether or not the autistic subjects themselves began to initiate interactions. The observation component of BOSCO permitted us to more accurately record each of these parameters of interaction and their sequential relationships, while its data analysis component increased the speed with which we could make data-based decisions in this single-subject design study. Subsequently, we put BOSCO to similar use in a descriptive analysis of the differences and similarities of mentally retarded and non-handicapped children’s interactions with their normally developing siblings (Savelle, Fox, & Phillips, 1987).
In a third and somewhat different study (Ragland & Fox, 1988) we were concerned with experimentally analyzing the relationship between particular teacher behaviors and an autistic child’s self-injurious behavior (SIB)-head slapping/hitting. The student’s SIB had been persistent and increasing for some time, yet it was irregular enough that teaching staff had been unable to precisely determine what environmental events reliably preceded the SIB. Informal observation and interviews with the classroom teacher were used to develop a multicategory observation system of teacher and student behaviors and a series of experimental probe conditions (e.g., social attention, task presentation, no demand-free play), some of which were hypothesized to be functionally related to self-injury. Observations using BOSCO were conducted during these probe conditions and during a naturally occurring school activity that had sometimes been associated with the student’s SIB. Use of the sequential recording and analysis capabilities of BOSCO in conjunction with the probe conditions allowed us to pinpoint a particular task, shoe tying, and a series of teacher behaviors (presentation of the shoes, verbal instructions for and physical prompting of shoe tying) that reliably elicited SIB. In addition, the observational data indicated a strong relationship between the events identified in the probe conditions as eliciting SIB and those associated with SIB in the natural environment. This analysis in turn led to development of a set of intervention tactics designed to reduce SIB for this student.
Future Applications and Summary
In each of the applications described here, BOSCO and the Model 102 represent distinct advantages over traditional data collection methods such as paper and pencil. These advantages include increased flexibility in the creation and modification of an observation system. preservation of the sequence of subject-environment interactions, simultaneous and efficient recording of both frequency and duration of events, and more rapid and extensive analyses (e.g., conditional probabilities) of the resultant observations.
There are, of course, drawbacks to this and other similar computer-based observation systems. We have had our share of equipment failures— weakened or dead batteries forcing cancellation of observations, misaligned disk drive heads or bad disks causing loss of data. Human factors also enter into the “disadvantages” of the system. For example, there is initially more start-up time in observer training in order to familiarize observers with operation of the computer itself and, in some cases, to desensitize those observers who are computer-phobic. Even when both computer and observational training have been completed, there have still been observer errors resulting in loss of data, e.g., inadvertent erasure of a file before it has been saved to the computer’s memory or to disk. However, for the most part, these problems have counterparts in the more traditional observational procedures. With the proper management their likelihood of occurrence and impact can be reduced.
When compared to other electronic data collection instruments, the programmability of the Model 102 computer gives it a flexibility not available in “dedicated” devices such as the Datamyte and OS-3. Because it uses a relatively small portion of the Model 102’s memory, it will be possible to enhance BOSCO or other observation programs to include additional recording, data analytic, and print-out capabilities. Other future plans for BOSCO are several. They include the development of a bar-code reading program to eliminate the time and distraction of keystroke entry of observation codes. Also in the planning stages is a program for both on-screen and printout graphing of certain behavior parameters, e.g., session frequency and duration of behavior codes. The BOSCO user can thus be increasingly freed from dependency upon desktop and mainframe computers for most typical data analyses. Even if the Model 102 is used as a stand-alone data collection and analysis tool, the internal telecommunications package of the computer will allow data transfer to other computers running other programs and further expand its uses. It would be instructive, for instance, to transfer BOSCO files to an Apple II + or IIE computer for analysis with a decision-making program such as AIMSTAR (Hasselbring & Hamlett, 1984). With such modifications, BOSCO and the Model 102 can emerge not only as a research tool but as an efficient and effective package for practitioners such as teacher trainers, consulting teachers, and school psychologists, perhaps increasing the ease and accuracy of data-based educational assessment and programming. In the future, perhaps a system of programs that generate compatible files will allow “generalization across machines” and across programs.” Such generality, unfortunately, is often lacking today.
One of the early technological advances in the science of human behavior was the development and application of direct observational measurement procedures by Florence Goodenough and other child development researchers some 60 years ago. With the exception of certain unfortunate and unscientific regressions, these methods have become a staple in the analysis of children’s behavioral development. However, as we have noted, there are both methodological and practical limitations to these sampling procedures as originally conceived and, frequently, still practiced. Direct observational assessment continues to be both underused and incorrectly used by professionals (e.g., psychologists, consultants, teachers) in field settings. We would argue that BOSCO and other similar hardware-software packages are a step, if not several steps, in the direction of remedying many problems in the assessment and analysis of behavior-environment interactions. Through the continued refinement, adaptation, and field-testing of these packages we can reach a point at which sophisticated behavioral observation can be as user-friendly and informative to teachers, parents, and program staff as it is now becoming for researchers.
References
Bakeman, R. & Gottman, J.M. (1986). Observing interaction: An introduction to sequential analysis. New York: Cambridge University Press.
Fox, J.J., Shores, R,E., Lindeman, D., & Strain, P. (1986). The effects of response dependent fading procedures in developing and maintaining social initiations of withdrawn preschool children. Journal of Abnormal Child Psychology, 14, 387-396.
Hasselbring, T.S., & Hamlett, C. (1984). Planning and managing instruction: Computer-based decision making. Teaching Exceptional Children, 16, 248-252.
Hendrickson, J.M., Strain, P., Tremblay, A., & Shores, R.E. (1981). Functional effects of peer social initiations on the interactions of behaviorally handicapped children. Behavior Modification, 6, 323~353.
Johnston, J.M., & Pennypacker, H.S. (1960). Strategies and tactics of human behavioral research. Hillsdale, NJ: Lawrence Erlbaum.
Kazdin, A.E. (1982). Single-case research designs: Methods for clinical and applied settings (pp. 53-56). New York: Oxford University Press.
MacLean, W., Tapp, J., & Johnson, W. (1985). Alternate methods and software for calculating interobserver agreement for continuous observation data. Journal of Psychopathology and Behavioral Assessment, 7, 65-73.
Martin, P., & Bateson, P. (1986). Measuring behavior: An introductory guide. Cambridge: Cambridge University Press.
Ragland, E,U., & Fox, J.J. (1988, November). Analogue probes as valid predictors of aberrant behavior in the natural environment. A paper presented at the 12th Annual TECBD National Conference on Severe Behavior Disorders of Children and Youth, Tempe, AZ.
Sackett, G.P. (1978). Measurement in observational research. In G.P. Sackett (Ed), Observing behavior. Volume 2: Data collection and analysis methods. Baltimore: University Park Press.
Savelle, S., & Fox, J.J. (1988). Differential effects of training In two classes of social initiations on the positive responses and extended interactions of preschool-age autistic children and their nonhandicapped peers. Monograph in Behavioral Disorders: Severe Behavior Disorders of Children and Youth, 11, 75-86.
Saville, S., Fox, J., & Phillips, M. (1987, March). The role of parents and siblings in handicapped children’s social development: Initial findings and future perspective. Paper presented at the Gatlinburg Conference on Research and Theory in Mental Retardation and Developmental Disabilities, Gatlinburg, TN.
Semmel, M.S., & Frick, T. (1980). CARTLO: Computer-Assisted Research into Teaching-Learning Outcomes. Bloomington: Indiana University, School of Education, Center for Innovation in Teaching the Handicapped.
Tremblay, A., Strain, R., Hendrickson, J.M., & Shores, R.E. (1981). Social interactions of normally developing preschool children: Using normative data for subject and target behavior selection. Behavior Modification, 5, 237.253.
Preparation of this manuscript was supported in part by Grant #7-pO1-155051 from the National Institute of Child Health and Human Development and by Grant #GCO&302929 from the U.S. Department of Education, Office of Special Education and Rehabilitative Services. The authors wish to thank Wilma Davis for her assistance in typing this manuscript.