Content uploaded by Iver Iversen
Author content
All content in this area was uploaded by Iver Iversen on Dec 04, 2017
Content may be subject to copyright.
Behavior and Philosophy, 39/40, 303-307 (2011/2012). © 2011 Cambridge Center for Behavioral Studies
303
COMMENTARY ON CYBERRAT
Iver H. Iversen
University of North Florida
ABSTRACT: This commentary on the development of CyberRat points out that 1)
CyberRat is an excellent educational alternative to a live rat in cases where instruction of
basic operant conditioning principles cannot be carried out with live animals due to a lack
of laboratory facilities, 2) CyberRat simulates a live rat very nicely as long as one expects
no more than demonstrations of basic operant behavior principles (i.e., CyberRat is not
suited for research into operant behavior), 3) neither a Kantorian interbehavioral analysis
nor a Skinnerian functional analysis is sufficient for CyberRat to work, yet a combination
of both types of analysis is in fact necessary for CyberRat to emit an adequate and realistic
flow of operant behavior interceded by other (non-reinforced) behavior, 4) CyberRat has
developed to the point where it certainly provides a near perfect illusion of being a single
animal that quite realistically demonstrates basic operant conditioning phenomena
embedded in a flow of natural behaviors.
Key words: CyberRat, Kantor, Skinner, behavioral interdependence, computer simulation,
operant behavior
Roger Ray (2011/2012) deserves considerable credit for developing CyberRat,
which provides an excellent demonstration of basic principles of operant
conditioning in a close-to-real video and data format. Operant behavior (lever
pressing) can be shaped with CyberRat and maintained under different schedules
of reinforcement and extinguished as well as put under discriminative control.
CyberRat is highly interactive as the user can modify the rat’s behavior via
controls that can be clicked on the computer screen (e.g., reinforcer delivery and
discriminative stimulus on/off). In addition, the operant behavior changes are
documented visually in traditional formats such as cumulative records of bar
pressing and tabulated averages of response rates. As such, CyberRat is an
excellent educational alternative to a live rat in cases where instruction of basic
operant conditioning principles cannot be carried out with live animals due to a
lack of laboratory facilities. In addition to educating about traditional findings
from operant conditioning experiments, CyberRat also illustrates that operant
behavior occurs not in isolation but embedded within a stream of additional natural
non-reinforced behaviors such as grooming, exploring the chamber, sniffing at the
lever, visiting the food tray, etc. Indeed, CyberRat is a compilation of actual video
of three live rats, and the viewing experience is, for the most part, akin to that of
looking at video of a live rat. To put it simply, CyberRat simulates a live rat very
nicely as long as one expects no more than demonstrations of basic operant
AUTHOR’S NOTE: Please address correspondence to the author at:iiversen@unf.edu
IVERSEN
304
behavior principles. CyberRat cannot be made to do something that the three live
rats that form the video data base were not trained to do (i.e., you cannot shape a
novel behavior other than lever pressing). Thus, CyberRat is not suited for
research into operant behavior but serves well for demonstration purposes. The
monograph provides an interesting and compelling narrative of the development of
CyberRat along with descriptions of its theoretical underpinnings and validation
methods for simulation authenticity or fidelity along the lines of various Turing-
like tests.
CyberRat encompasses more than meets the eye as it is based on an approach
that is a hybrid of Skinner’s experimental analysis of behavior (Skinner, 1938) and
Kantor’s interbehavioral psychology (Kantor, 1959). Ray argues that the methods
of experimental analysis of behavior by themselves do not provide a sufficient
basis for development of CyberRat. The reason for this argument is that Skinner’s
analysis does not deal with other behaviors that intercede between the operant
responses. Yet these behaviors are visible when one watches CyberRat in action,
and given their electronic nature, they must be computer-selected based on some
model. Ray suggests that there would be no way from a Skinnerian analysis alone
to develop a model for how such other behaviors actually intercede between the
operant responses, which behavior occurs when, how such behaviors follow each
other, etc. For CyberRat to switch realistically among operant behavior and other
behaviors, the computer algorithm that searches the video clips that form the basis
of CyberRat’s behavior transitions must be based on some model of behavior flow
and transition. Ray has based such a model on further extensions of his
comprehensive research on structural hierarchies of interbehaviors (e.g., Ray &
Brown, 1975, 1976). Carefully recorded behaviors (by visual observation) of live
rats were analyzed in kinematic flow diagrams that capture conditional
probabilities of behavior changes. These probabilities of changing from one form
of behavior to another are used for the computer to select among the video clips.
Apparently, some 1800 clips of varied length (i.e., 1–15 seconds) and behavior
composition taken from three albino rats (presumably of the same size) and
exposed to the same experimental conditions form the actual database that
CyberRat operates on. For example, as the user gives a reinforcer to CyberRat after
it sniffs the right side of the lever, in an attempt to shape lever pressing, the
computer program will search for the next video clip based on what was the most
likely behavior to occur next for the live rats as they received a reinforcer after
sniffing the right side of the lever. So, an interbehavioral analysis was necessary to
give CyberRat a realistic flow of behaviors. On the other hand, the interbehavioral
analysis is not sufficient as the conditioning aspects of behavior obviously stem
from a Skinnerian analysis. In essence, neither an interbehavioral analysis nor a
Skinnerian analysis is sufficient for CyberRat to work. But both types of analysis
are in fact necessary for CyberRat to emit an adequate and realistic flow of operant
behavior interceded by other (non-reinforced) behavior.
Skinnerian and Kantorian analyses are sometimes set up against each other as
being incompatible. Skinner’s analysis is a functional analysis with a focus on
independent and dependent variables in experimental arrangements with actual
COMMENTARY ON CYBERRAT
305
manipulations of variables while Kantor’s analysis is descriptive and focuses on
determining conditional probabilities of behavior. For example, Ray writes:
For Kantor, functional relations were a two-way process of reciprocated
influence, or mutual implications, that logically rule out a distinction between
independent and dependent variables. Instead, all contributing components in
the psychological field were seen as being interdependent. (p. 213, emphasis in
the original)
However, to say that two behaviors, A and B, are interdependent quite literally
means that A influences B and that B influences A. For such influences to be
determined empirically it may not be sufficient just to measure the behaviors and
perform kinematic analyses. Interdependence can, in fact, be determined
experimentally with separate manipulations of A and B as independent and
dependent variables. For example, the length of post-reinforcement pauses in
operant behavior maintained under fixed-ratio schedules of reinforcement can be
manipulated in various ways and be shown to influence the extent of other non-
reinforced behavior (e.g., water-drinking after food reinforcement in rats); on the
other hand, the extent of such non-reinforced behavior (e.g., length of drinking
bouts) can be manipulated in and of itself and be shown to influence the length
post-reinforcement pauses (Henton & Iversen, 1978; Iversen, 1976). Thus, an
experimental analysis can show that the same behavior (in this case drinking) can
be both a dependent variable and an independent variable in different experimental
settings, and its relation to the operant response can be characterized as
interdependence. So, a Kantorian focus on interbehaviors can actually enrich and
encourage an experimental approach to behavioral interdependence rather than
deny or question it.
Progress in science and technology is often rooted in methods of validation
based on experimental approaches. In contrast, Ray argues in favor of purely
descriptive methods in psychological inquiry because they may more realistically
represent interactions among many variables. It is true for some large-scale
scientific domains (e.g., forecasting of weather patterns and ocean currents) that
experimental manipulation is not possible except on a small scale in the laboratory.
Ray includes modern brain- and neurosciences in the fold of areas where
description is said to supersede experimental manipulation. An example is
functional brain imaging (fMRI) with its complex measures of brain activity in
conjunction with performance on various cognitive tasks. However, some recent
advances in neuroscience have moved beyond description and now focus on
manipulation of fMRI activity based on experimental approaches such as giving
participants feedback on their ongoing fMRI patterns during sessions (e.g., Rota et
al., 2009); definitions of dependent and independent variables are a necessity for
such studies. Similarly, complex electroencephalogram (EEG) recordings, which
traditionally served as descriptive or diagnostic measures of brain activity, can now
be modulated based on feedback and can even be shown to submit to operant
conditioning (e.g., Birbaumer, 2006 ; Iversen et al., 2008).
IVERSEN
306
CyberRat is validated based on several separate Turing tests that seek to
demonstrate how well CyberRat simulates a live rat under similar experimental
conditions. Ray argues that the cumulative records and tabulated overall operant
data generated by CyberRat show great fidelity to the operant behaviors generated
by live rats under the same experimental conditions. That is indeed what would
need to be the case for CyberRat to serve as a computerized demonstration of basic
operant conditioning principles in lieu of live rats. The experience of watching
CyberRat perform on the video display also has fidelity to the performance of a
live rat. However, the interpretation of how well the experience of watching the
video compilation of CyberRat matches the experience of watching a live rat under
the same experimental conditions clearly depends on how much perfectionism the
viewer seeks. Ray is clearly aware of the “jumps” that sometimes occur on the
computer screen video as CyberRat suddenly moves from one corner of the box to
the other. The smoothing of the video and diminishing of “jumps” progressed with
additional behavioral analyses beyond those of Ray and Brown (1975, 1976) to
improve and refine the conditional probabilities of switching among behaviors. It
seems to be exactly at this juncture that Ray believes that the interbehavioral
approach is superior to the Skinnerian approach because it is only by enhancing the
interbehavioral analysis technique that the model for behavioral transitions can be
improved so that video clip selections become smoother and less “jumpy” and
therefore more realistic and lifelike. Ray deserves credit for basing the
performance improvements of CyberRat on algorithms derived from performance
of live rats as opposed to arbitrary video “smoothing” techniques, tempting as that
may have been. However, to base the algorithms for CyberRat on statistical
averages and standard deviations of empirical kinematic data across several live
animals actually makes CyberRat fail another test of authenticity, namely that of
depicting control of the behavior of a single animal. Skinner’s experimental
analysis of behavior became famous and influential, resulting in substantial clinical
and educational applications precisely because it provided a system and a
methodology for functional analysis at the level of the single organism (e.g.,
Iversen, 2013). Along the same lines, there is one hitherto unmentioned aspect of
CyberRat that fails the Turing test, yet this failure is entirely invisible to the
observer. Because the video basis for CyberRat is three live albino rats, one cannot
distinguish the rats on the video. So a transition from one behavior (clip) to another
may also be a shift from one rat to another. Had the three live rats been dyed red,
green, and blue, then CyberRat would show random color changes that obviously
would render CyberRat highly unnatural. The fact that CyberRat is based on the
performance of identical, yet different, live rats is not a detriment but may be a
necessity for a realistic educational experience for the viewer. In conclusion,
CyberRat has developed to the point where it certainly provides a near-perfect
illusion of being a single animal that quite realistically demonstrates basic operant
conditioning phenomena embedded in a flow of natural behaviors.
COMMENTARY ON CYBERRAT
307
References
Birbaumer, N. (2006). Breaking the silence: Brain–computer interfaces (BCI) for
communication and motor control. Psychophysiology, 43, 517-532.
Henton, W. W., & Iversen, I. H. (1978). Classical conditioning and operant conditioning: A
response pattern analysis. New York, NY: Springer-Verlag.
Iversen, I. H. (1976). Interactions between reinforced responses and collateral responses.
The Psychological Record, 26, 399-413.
Iversen, I. H. (2013). Single-case research methods: An overview. In G. J. Madden (Ed.),
APA handbook of behavior analysis: Vol 1. Methods and principles (pp. 3-32).
Washington D.C.: American Psychological Association.
Iversen I. H., Ghanayim N., Kübler A., Neumann N., Birbaumer N., & Kaiser J. (2008). A
brain–computer interface tool to assess cognitive functions in completely paralyzed
patients with amyotrophic lateral sclerosis. Clinical Neurophysiology, 119, 2214-2223.
Kantor, J. R. (1959). Interbehavioral psychology (2nd ed). Granville, OH: Principia Press.
Ray, R. D. (2011/2012). CyberRat, interbehavioral systems analysis, and a “Turing test”
trilogy. Behavior and Philosophy, 39/40, 203-301.
Ray, R. D., & Brown, D. A. (1975). A systems approach to behavior. The Psychological
Record, 25, 459-478.
Ray, R. D., & Brown, D. A. (1976). The behavioral specificity of simulation: A systems
approach to procedural distinction of classical and instrumental conditioning.
Pavlovian Journal of Biological Sciences, 11, 3-23.
Rota, G., Sitaram, R., Veit, R., Erb, M., Weiskopf, N., Dogil, G., & Birbaumer, N. (2009).
Self-regulation of regional cortical activity using real-time fMRI: The right inferior
frontal gyrus and linguistic processing. Human Brain Mapping, 30, 1605-1614.
Skinner, B. F. (1938). The behavior of organisms: An experimental analysis. New York:
Appleton-Century-Crofts.
