Using Computerized Response Time Measurement For Detecting Secondary Distractions in Advertising

Citation:

Robert S. Owen, Kenneth R. Lord, and Martha C. Cooper (1995) ,"Using Computerized Response Time Measurement For Detecting Secondary Distractions in Advertising", in NA - Advances in Consumer Research Volume 22, eds. Frank R. Kardes and Mita Sujan, Provo, UT : Association for Consumer Research, Pages: 84-88.

Advances in Consumer Research Volume 22, 1995      Pages 84-88

USING COMPUTERIZED RESPONSE TIME MEASUREMENT FOR DETECTING SECONDARY DISTRACTIONS IN ADVERTISING

Robert S. Owen, SUNY Oswego

Kenneth R. Lord, SUNY Buffalo

Martha C. Cooper, Ohio State University

Use of the secondary task technique in the measurement of the distraction potential of prospective media environments is advocated and briefly discussed. Although there are a variety of secondary task methods that can be used to detect and measure the various attention related constructs, the current discussion focuses on the "RT-probe" technique, whereby the response time (latency) to a secondary task is taken as an indicator of mental attention devoted to a primary task. The proliferation of microcomputers brings this method within reach of most advertising researchers, albeit with some cautions and limitations that are discussed.

INTRODUCTION

The notion that the human processing system has an asymptotic limit in "capacity" can be used as the basis for the investigation of such related constructs and processes as "attention", "elaboration", and "mental effort." For example, television viewing could involve substantial elaboration as viewers anticipate a conflict resolution or relate an episode to their own prior experiences (Lord and Burnkrant 1993). As more attentional capacity is absorbed by the task of program elaboration, a viewer would likely become less aware of or responsive to other stimuli (e.g., advertisements) in the environment. The observation that a person is less aware of or responsive to other "secondary" stimuli can be taken as an indication that the "primary" task of product comparison is attracting or requiring a greater amount of "mental attention". A measure of the amount of deterioration in responsiveness to "secondary" stimuli can be taken as an indicator of the amount of mental attention that program involvement and elaboration attracts or requires.

This so-called "secondary task technique" has been implemented in a variety of ways. One interesting variation on the general idea of the secondary task technique has been used by Children's Television Workshop, producers of Sesame Street, to test television shows on 3- and 4-year old children (Waterman 1990). Children in day-care centers were shown segments of the TV show under test while slides were concurrently projected on a screen next to the TV display. The ability of a TV segment to capture the "attention" or "involvement" of these preschool viewers is taken as a function of the attention that is given to the secondary (slides) vs. primary (TV show) task. In this use of a secondary task technique, the observation of lesser attention to the secondary task is presumed to indicate that the primary task is absorbing, consuming, or in some way requiring greater viewer attention.

The focus of the present report is more specifically on the use of the "RT-probe" form of secondary task technique in the measurement of capacity-related constructs such as attention, elaboration, involvement, mental effort, and such. In using the RT-probe technique, changes in reaction times (RT) to a secondary task are taken to indicate changes in the use of attentional resources that are devoted to performance of the primary task. Commonly available microcomputers can be used to measure and record reaction times.

Although this method is not new, it has never been adequately described with regard to how it is implemented and, importantly, how it is limited. The present report, then, is a synthesis of issues associated with the measurement technique, the necessary physical instrumentation, and limitations of the physical instrumentation and of the theoretical basis of the technique. The general basis of this technique is first discussed, followed by discussion of how computers can be implemented in using this method, of some limitations and cautions in making these uses, and of some limitations of the secondary task technique itself.

THE SECONDARY TASK TECHNIQUE AND THE RT-PROBE

One of the more common operationalizations in the investigation of the various attention-related constructs and processes has involved the so-called "dual-task" and "secondary task" techniques. Dual-task and secondary task techniques generally presume a processing system of resource limitations: the consumption of processing capacity by one task will leave less capacity for the processing of a second concurrent task. When both tasks attempt to concurrently consume more capacity than is available, the performance of one or both tasks must suffer. This will presumably result in the observation of degraded task performance.

A frequently used secondary task in laboratory settings in recent years has been the "RT-probe", in which subjects must press a hand-held switch button in response to an occasional flash of light or an audible click or "beep" sound. A degradation (increase) in reaction time (RT) to an occasional beep (the secondary task) during, say, the reading of a message or the viewing of a television program (the primary task) is taken as a quantitative measure of an increase in attention or mental effort devoted to the primary task. Points at which RTs greatly increase above "normal", usually accompanied by many "misses" (no response to the beep) or "false alarms" (response where no beep occurred), are taken to indicate degraded secondary task performance and as an indication that the primary task is consuming attentional resources near the capacity threshold. This presumably provides an indication of, say, points within a communication that consume more attention or require more effort to process, or of, say, mental comparisons that require more mental attention or effort. Secondary task performance changes, then, are taken to function as a probe into attentional resource consumption by the primary task.

Lord and Burnkrant (1993), for example, used the RT-probe to find apparently "high involvement" and "low involvement" segments within a suspenseful "Alfred Hitchcock Presents" television program. Longer RTs to an occasional audible beep were taken as an indication of higher viewer "involvement" with particular points within the program story. After the RT-probe had identified the location of the more and less involving segments of the program story, commercials could then be positioned within these apparently high involvement or low involvement segments of the program story. The objective was to investigate differences in recall and attitude change associated with advertisements that were shown at points of high vs. low program involvement.

Lord, Burnkrant, and Owen (1989) had found that longer RTs continued throughout television advertisements embedded within high involvement segments, but remained normal throughout the same advertisements positioned within low involvement segments. These longer RTs were taken to indicate that processing resources were being consumed by program elaboration throughout advertisements positioned within high involvement program segments. Attitude and recall measures taken after the program viewing suggested that there apparently was some processing interference associated with commercials positioned within high involvement program segments.

Thorson and her colleagues (1985; 1987) have similarly used the RT-probe to investigate television viewer "attention". Moore, Hausknecht and Thamodaran (1986) have used the RT-probe in the investigation of "attention allocation" to the processing of compressed audio commercials. Britton and his colleagues have conducted a number of studies (1978, 1979, 1980, 1982) into the usage of "cognitive capacity" with differences in text complexity and structure. All of these uses of the RT-probe are based on the assumption that as the primary task consumes near the threshold of capacity, secondary task performance will be degraded due to a lack of sufficient reserve capacity.

USING COMPUTERS IN RT-PROBE MEASUREMENT

None of the above uses of the RT-probe secondary task technique required especially elaborate, specialized, or expensive equipment. Thorson et al. and Moore et al., for example, report using commonly available Apple II computers. Lord et al. used a Commodore C-64 computer. In the Lord studies, the built-in game ports were used to input both the beep sound stimuli from the video tape (recorded separately on one stereo track) as well as the hand-held button switches. The game port on the Commodore computer allows the input of analog signals (the beep sound) and allows as many as ten button switches to be attached simultaneously (i.e., ten subjects can be run in a single session). The button switches are simply substituted for the four direction switches and the fire button of a joystick. Note that the analog input does require some simple signal conditioningCit will not give good results if directly connected to the audio output of the video player.

Lord et al. also used the computer to dub the beep stimuli on the stimulus tape. The computer generated beep sounds (a pleasant "bing") that were randomly spaced between three and nine seconds apartCan average of 10 beeps per minute. These were recorded on one channel of the stereo audio track of the video tape. The reason that the spacing between the beeps was random, rather than maintained at a constant interval, was to minimize the possibility of automatism, i.e., so that subjects would not get acclimated to any particular pace in responding to the secondary task. The beeps were recorded on a stereo audio track separate from the program audio to simplify the task of detecting them with the computer equipment.

On playback to subjects, the computer was used to count beeps in the audio portion of the video program, thereby locating particular points in the program by beep number. The computer also measured the time between the onset of a beep and any button presses by subjects. The data that were saved, then, were the beep number, the subject identification number each time a new button press was detected (ten subjects could be run simultaneously), and the reaction time between the onset of the beep and each new button press.

Other commonly available, low cost microcomputers can similarly be used for these purposes. A low cost game port can easily be installed in an older, discarded IBM-PC, for instance. (Maybe the old beast has a clunky floppy drive and can't do Windows, but it can still make beep noises and can still count beeps and timer ticksCdon't throw it away yet!) One limitation that many potential users might encounter, however, is that writing the machine-level software to read these inputs and to take measurements of the time between the secondary stimulus input and the button press does require some specialized expertise.

COMPUTER JIFFIES AND TICKS AS TIMING DEVICES

Most microcomputers provide user-accessible time keeping functions known as "jiffy clocks" or "tick counters". In HyperCard on a Macintosh, a "tick" is an interval of time of nominally [1/60]-second in length, and on a PC-DOS or MS-DOS machine, a "tick" is an interval of time of nominally [1/18.2]-second in length. Note that ticks are not implemented or used in the same way on all computers. On an IBM-PC, for instance, the tick is derived from a periodically occurring event called an "interrupt", and the frequency of the occurrence of this interrupt can be altered through software (that is, it is possible to write software to "speed up" this tick rate, thereby allowing time measurements of finer resolution: cf., Buhrer, Sparrer, and Weitkunat 1987; Heathcote 1988; Sargent and Shoemaker 1984, p. 263). On a Macintosh, the higher frequency tick is derived from an interrupt generated by the periodically occurring CRT display refresh (Lane and Ashby 1987), and therefore cannot be altered. These periodically occurring interrupts are used primarily as reminders to an operating system to perform certain operations, such as to check the keyboard for new information and to increment the tick counter by one count for time keeping functions. The value in this tick counter is used by many operating systems to update the time-of-day clock and can usually be accessed by the user. An exception is the Macintosh, which does not use the tick counter to maintain its time-of-day clock (Harvey 1988, p. 471).

There are three potential problems associated with using tick counters for timing measurements. The first is that the performance of some operations can require that other operations be temporarily suspended, including updating the tick counter. Disk drive operation, for instance, will interfere with the tick counter on the Macintosh (Harvey 1988, p.471; Kieley and Higgins 1989). Under normal uses, the loss of a tick here and there is of little consequence. For scientific measurements, however, the loss of a tick here and there is an unacceptable source of systematic error in a short-interval measurement. As long as one has a knowledge of the interrupt structure of the particular operating system being used, however, such sources of systematic error can usually be avoided. (E.g., simply avoid using the disk drive in the middle of taking a timing measurement!)

A second potential problem associated with using tick counters for timing measurements is that most users cannot instantly read the time in a tick counter. A line of program in a higher-level language (e.g., BASIC) which reads the tick register may take tens of milliseconds to execute, and this amount of time can vary within the same program, making higher-level languages unsuitable in many applications (Dlhopolsky 1983a, 1983b; Dorfman 1987; Emerson 1988; Flexser 1987; Grice 1981; Kieley and Higgins 1989; Mapou 1982; Rayfield 1981, 1982). This unknown amount of processing overhead can be an unacceptable source of systematic error if a short-interval measurement is required.

A third problem associated with using tick counters for timing measurements is that they sometimes lack the proper amount of resolution for the uses that are made of them. Importantly, to display, analyze, or report a unit of, say, [1/18.2]-seconds as 0.055 seconds is at best misleading, giving the false impression that the measurement instrument was capable of resolving to the level of at least one millisecond. The actual unit of resolution is 55 times more crude, and any analysis using the decimalized value looses meaning (cf., Owen and Cooper 1990). The Moore et al. (1986) study is noteworthy in that these investigators appropriately analyzed and reported Apple II sixtieth-second ticks directly in units expressed as ticks, without unnecessary conversion to decimal form.

As an aside, it is perhaps useful to note here that a typical fast reaction timeCwhen a person is in high arousal or a state of readiness to respond to a clear visual or audio stimulusCwill be on the order of magnitude of about 2-tenth-seconds with a hand-held button switch. That is, this is the sort of reaction times that would be typical in an RT-probe secondary task when enough "capacity" is available to respond to the secondary task. As the primary task requires processing resources at a level that approaches capacity limits, reaction times can be observed to degrade to well over a second, with the longer reaction times likely to be "false alarms"; these longer reaction times are also accompanied by many "misses".

Although the 2-tenth-second level of reaction time might appear to be well within the abilities of computer ticks to detect, interest is often in the difference between various reaction times, which can be at the hundredth-second order of magnitude. The needs of the measurement task therefore require special consideration to determine if a tick-based timer is appropriate for the particular measurement task. Also note that some advise an additional order of magnitude in taking measurements (e.g., Owen and Cooper 1991), i.e., differences of hundredth-second levels should be measured and analyzed at millisecond levels, although this is not a universally accepted caution.

KEYBOARD AS A RESPONSE DEVICE

The method used to gather subject responses to the secondary stimulus also requires some scrutiny. Note, for example, that the closure of a switch on an IBM-PC keyboard cannot be instantly registered by the main processor. Keyboard switches are not directly connected to the keyboard cable; a processor within the keyboard first interprets which key is depressed from the matrix of switches and sends a correspondingly coded series of pulses through the keyboard cable (see Glasco and Sargent [1983] for a very thorough discussion). On a PC running under PC-DOS or MS-DOS, the computer checks, or "polls", for a signal from the keyboard about 18.2 times per second. Although it is possible for a knowledgeable programmer to have the keyboard polled at a faster rate, the keyboard itself can still take tens of milliseconds to output a signal in response to a keypress, and this amount of time has been found to vary between keyboards (Graves and Bradley 1987). Substituting the PC keyboard as an input device if millisecond-level reaction time measures are required is at best hazardous, and more precise and faster response devices should be used instead for this purpose (cf., Crosbie 1989). The Macintosh functions in a similar manner, except that the keyboard is polled at a rate of nominally 60 times per second, and a keypress can be detected by the keyboard in about 6 milliseconds (Lane and Ashby 1987).

As a general rule, older microcomputers in which the keyboard is contained in the same enclosure as the main processor do not operate in this way, allowing the keyboard to be sensed by the processor in real-time (i.e., usually within a few microseconds, discounting any mechanical latency of the switch itself). More specifically, the memory-mapped keyboard designs of many older microcomputers (e.g., Commodore C-64, PET, TRS-80, etc.) are better suited for high-resolution reaction time measurement. Due to the mechanical construction of the key switch, however, even these keyboards might sometimes (although very unlikely in advertising studies) not be suitable as response devices if millisecond-level response latency measurement is critical (Reed 1981). A variety of more appropriate or more desirable alternatives to the keyboard for real-time button-press detection exist (e.g., Creeger, Miller, and Paredes 1990). Recall, for example, that Lord and Burnkrant (1993) and Lord, Burnkrant, and Owen (1989) used external button switches that were plugged in to the game (joystick) port of a computer. Game ports for IBM-PC computers are readily available for well under $50.

SOME LIMITATIONS IN USING THE RT-PROBE TECHNIQUE

There are some important limitations to consider in using the RT-probe technique. First and foremost, one has to be careful with assumptions regarding the particular construct that is presumably measured by any secondary task technique. A general discussion of attention and the secondary task technique can be found in Lynch and Srull (1982); Owen (1991) outlines a variety of concerns and cautions regarding the attention-related constructs. The point here is that constructs such as attention, elaboration, involvement, mental effort, and such, all seem to be related to the notion of "capacity", but the full relationship and distinction between such constructs is not yet clear. The RT-probe technique merely detects when capacity is near the threshold of being "swamped", and conclusions regarding any of a variety of capacity-related constructs always require careful scrutiny. Nonetheless, the fact that the secondary task technique does appear to detect something should encourage us to use this technique as long as it allows us to make useful inferences regarding an attention-related construct (cf., Navon 1984; Owen 1991).

Additionally, there is evidence that the human processing system consists of more than just a simple, single "capacity" resource (cf., Owen 1991). The fact that subjects in high attentional load conditions will respond to a secondary task beep after, say, a one second delay does itself suggest that there exists an auditory resource that can independently attend to and store incoming audio stimuli until a general purpose, main processor can attend to this new input. Nonetheless, that one second delay can be used as evidence that some sort of general, global attentional resource is being used more heavily. The point is that one must always maintain caution regarding conclusions that can be drawn about the use of processing resources.

A second caution regards some of the limitations of microcomputer equipment that might be used in making RT-probe measurements. These limitations are not necessarily severe, but the limitations of any measurement instrument must be understood whenever it is used to take a measurement. For example, most timing measurements that use an IBM-PC platform make use of an internal timer that "ticks" at a rate of nominally 18.2 times per second. Unfortunately, many examples can be found which express such "ticks" in decimal form, incorrectly implying millisecond-level resolution where the true level of resolution is closer to about a tenth-second order of magnitude. For example, the "timer" function in MS-DOS QBasic and GW-Basic, commonly misused in writing timing routines, misleadingly returns a continuous value in decimal form, but a simple "print timer" loop will return values at increments of nominally [1/18.2]-seconds.

Such limitations in the abilities of the available hardware, however, do not necessarily restrict the ability to conduct RT-probe research. If the expertise exists, special programming techniques (which might require the use of machine language) can be used to obtain measurements at much finer levels of resolution. Lord and Burnkrant (1993) and Lord, Burnkrant, and Owen (1989), for example, used machine language and built-in timing hardware that enabled a resolution of much better than 100 microseconds. For human reaction-time measurements, however, such fine resolution is, in many cases, not necessary. Moore et al. (1986), for example, reported RT-probe reaction time measures at the level of sixtieth-second ticks; the resolution of these tick units is adequate and appropriate for the particular task that was reported.

CONCLUDING REMARKS

The secondary task technique has seen more extensive use in engineering and cognitive psychology than in advertising. Greater use of the secondary task technique in advertising research is advocated, and it is hoped that the "how to" nature of the present report will encourage greater use in both experimental research and in field practice. The secondary task technique can be used in the study of a number of constructs, processes, and issues of relevance to advertising effectiveness, such as television program involvement, the attention-engaging capacity of a particular advertisement, and the quantity of "processing resources" that are required for such objectives as brand awareness, claim recall, message elaboration, or persuasion.

The RT-probe is a special form of secondary task that holds appeal in advertising research for several reasons. One feature of the RT-probe is that it can provide continuous yet relatively unobtrusive measures of attention consumption. That is, it enables the advertiser to pinpoint a particular segment of an advertisement or program/editorial environment in which greater or lesser attentional resources are being utilized. Used in conjunction with post-task paper and pencil instruments, the RT-probe can provide an objective measure of attention-related constructs and processes to assist in validation of subjective measures. The RT-probe is a form of secondary task which provides data in numerical form, enabling common statistical sorts of inferences.

Several cautions regarding the underlying theoretical assumptions of the RT-probe and the use of computers in response time measurement have been discussed. Although particular limitations in theory and in the level of software/hardware expertise of the research team must be important considerations in applying the RT-probe, the secondary task is a well established and respected method outside of the marketing and advertising disciplines. Dual-task and secondary task methods have been used effectively by experimental psychologists for over a century (e.g., Jastrow 1892; Welch 1898), and the RT-probe technique has been in use for at least two decades (e.g., Posner and Boise 1971). This technique has seen recent use in advertising and provides great promise in field and experimental work in the study of consumer advertising processing.

REFERENCES

Britton, Bruce K. and Abraham Tesser (1982), "Effects of Prior Knowledge on Use of Cognitive Capacity in Three Complex Cognitive Tasks," Journal of Verbal Learning and Verbal Behavior, 21, 421-436.

Britton, Bruce K., Timothy S. Holdredge, Cheryl Curry, and Robert D. Westbrook (1979), "Use of Cognitive Capacity in Reading Identical Texts with Different Amounts of Discourse Level Meaning," Journal of Experimental Psychololgy: Human Learning and Memory, 5, 262-270.

Britton, Bruce K., Robert D. Westbrook, and Timothy S. Holdredge (1978), "Reading and Cognitive Capacity Usage: Effects of Text Difficulty," Journal of Experimental Psychology: Human Learning and Memory, 4, 582-591.

Britton, Bruce K., Robbie Zeigler, and Robert D. Westbrook (1980), "Use of Capacity in Reading Easy and Difficult Text: Two Tests of an Allocation of Attention Hypothesis," Journal of Reading Behavior, 12,23-30.

Buhrer, Michael, Bernd Sparrer, and Rolf Weitkunat (1987), "Interval Timing Routines for the IBM PC/XT/AT Microcomputer Family," Behavior Research Methods, Instruments, & Computers, 19(3), 327-334.

Creeger, Carl P., Kevin F. Miller, and David R. Paredes (1990), "Micromanaging Time: Measuring and Controlling Timing Errors in Computer-Controlled Experiments," Behavior ResearchMethods, Instruments, & Computers, 22(1), 34-79.

Crosbie, John (1989), "A Simple Turbo Pascal 4.0 Program for Millisecond Timing on the IBM PC/XT/AT," Behavior Research Methods, Instruments, & Computers, 21(3), 408-413.

Dlhopolsky, Joseph G. (1983a), "Limitations of High-Level Microcomputer Languages in Software Designed for Psychological Experimentation," Behavior Research Methods & Instrumentation, 15(4), 459-464.

Dlhopolsky, Joseph G. (1983b), "Machine Language Millisecond Timers for the Z-80 Microprocessor," Behavior Research Methods & Instrumentation, 15(5), 511-520.

Dorfman, David (1987), "Software Timing of Events in Cognitive Psychology Experiments," Behavior Research Methods, Instruments, & Computers, 19(2), 185-190.

Emerson, Phillip L. (1988), "TIMEX: A Simple IBM AT C Language Timer with Extended Resolution," Behavior Research Methods, Instruments, & Computers, 20(6), 566-572.

Flexser, Arthur J. (1987), "Accurate Display Timing Using an Enhanced BASIC for the Tandy Color Computer," Behavior Research Methods, Instruments, & Computers, 19(5), 457-459.

Glasco, David B. and Murray Sargent III (1983), "Using IBM's Marvelous Keyboard," Byte 9(5), 402-415.

Graves, Roger and Ron Bradley (1987), "Millisecond Interval Timer and Auditory Reaction Time Programs for the IBM PC," Behavior Research Methods, Instruments, & Computers, 19(1), 30-35.

Grice, G. Robert (1981), "Accurate Reaction Time Research with the TRS-80 Microcomputer," Behavior Research Methods & Instrumentation, 13(5), 674-676.

Harvey, Greg (1988), Understanding HyperCard. San Fransisco: Sybex.

Heathcote, Andrew (1988), "Screen Control and Timing Routines for the IBM Microcomputer Family Using a High-Level Language," Behavior Research Methods, Instruments, & Computers, 20(3), 289-297.

Jastrow, J. (1892), "The Interference of Mental Processes," The American Journal of Psychology, 4, 219-223.

Kieley, James M. and Timothy Higgins (1989), "Precision Timing Options for the Apple Macintosh Family of Computers," Behavior Research Methods, Instruments, & Computers, 21(2), 259-264.

Lane, David and Brad Ashby (1987), "PsychLib: A Library of Machine Language Routines for Controlling Psychology

Experiments on the Apple Macintosh Computer," Behavior Research Methods, Instruments, & Computers, 19(2), 246-248.

Lord, Kenneth R. and Robert E. Burnkrant (1993), "Attention versus Distraction: The Interactive Effect of Program Involvement and Attentional Devices on Commercial Processing," Journal of Advertising, 22(March), 47-60.

Lord, Kenneth R., Robert E. Burnkrant and Robert S. Owen (1989), "An Experimental Comparison of Self-Report and Response Time Measures of Information Processing," in AMA Educators' Proceedings: Enhancing Knowledge Development in Marketing, Paul Bloom, Russ Winer, Harold Kassarjian, Debra Scammon, Bart Weitz, Robert Spekman, Vijay Mahajan, Michael Levy, eds., Chicago: American Marketing Association, 196-200.

Lynch, John G. and Thomas K. Srull (1982), "Memory and Attentional Factors in Consumer Choice: Concepts and Research Methods," Journal of Consumer Research, 9, 18-37.

Mapoou, Robert L. (1982), "Tachistoscopic Timing on the TRS-80," Behavior Research Methods & Instrumentation, 14(6), 534-538.

Moore, Danny L., Douglas Housknecht, and Kanchana Thamodara (1986), "Time Compression Response Opportunity in Persuasion," Journal of Consumer Research, 13(June), 85-99.

Navon, David (1984), "Resources - A Theoretical Soupstone?" Psychological Review, 91, 216-234.

Owen, Robert S. (1991), "Clarifying the Simple Assumption of the Secondary Task Technique," Advances in Consumer Research, 18, 552-557.

Owen, Robert S. and Martha C. Cooper (1990), "Some Definitions and Issues in Measurement and Instrumentation," Working Paper Series of the College of Business #WPS 90-49. Ohio State University.

Posner, Michael I. and Stephen J. Boies (1971), "The Components of Attention," Psychological Review, 78(5), 391-408.

Rayfield, Frederick (1981), "Controlling Behavior Experiments with BASIC on 6502-Based Microcomputers," Behavior Research Methods & Instrumentation, 13(6), 735-740.

Rayfield, Frederick (1982), "Experimental Control and Data Acquisition with BASIC in the Apple Computer," Behavior Research Methods & Instrumentation, 14(4), 409-411.

Reed, Adam V. (1981), "On Choosing an Inexpensive Microcomputer for the Experimental Psychology Laboratory," Behavior Research Methods and Instrumentation, 12(6), 607-613.

Sargent, Murray III and Richard L. Shoemaker (1984), The IBM Personal Computer from the Inside Out, Reading, Mass: Addison-Wesley Publishing Company.

Thorson, Esther, Byron Reeves, and Joan Schleuder (1985), "Message Complexity and Attention to Television," Communication Research, 12(4), 427-454.

Thorson, Esther, Byron Reeves and Joan Schleuder (1987), "Attention to Local and Global Complexity in Television Messages," In Communication Yearbook 10, Ed. Margaret McLaughlin. Newbury Park, CA: Sage Publications, 366-383.

Waterman, Robert H. Jr. (1990), Adhocracy: The Power to Change, location unlisted: Whittle Direct Books.

Welch, Jeanette. C. (1898), "On the Measurement of Mental Activity through Muscular Activity and the Determination of a Constant of Attention," American Journal of Physiology, 1, 283-306.

----------------------------------------