Sniffy the Virtual Rat Pro Version 2.0 is the most recent update to the Sniffy the Rat program modules. It has an extensive set of simulations related to Pavlovian and operant conditioning.
Sniffy, the virtual rat: Simulated operant conditioning. In Figure 1, Sniffy is exploring. In this exercise, Sniffy. Is trained to press the bar (although it is possible to train.
Many of the standard conditioning paradigms and phenomena are contained within this program. Pavlovian effects such as blocking, overshadowing, and stimulus competition are demonstrated in the data output. Operant manipulations allow shaping the behavior of the virtual rat, observing cumulative records of interval and ratio schedules, as well as conducting operant discrimination procedures. In both Pavlovian and operant paradigms, one can generate histograms that predict the degree to which various stimuli control the virtual rat's behavior. On-screen presentations may keep most users interested, but working with the generated data files sometimes can prove difficult. Overall, this program has a strong potential for facilitating the instruction in undergraduate conditioning courses, serving as an addendum to traditional undergraduate conditioning laboratories and as a supplement to the use of live animals.
Sniffy the Virtual Rat, Pro Version 2.0 may serve as a valuable tool in undergraduate teaching laboratories conducted concurrently with courses on learning, basic and applied behavior analysis, and research methods. The software demonstrates the effects of a wide variety of Pavlovian and operant manipulations on the behavior of a virtual rat. Installation is simple; however, software operation definitely requires reference to the user manual. Overall, I found the software to be useful for the teaching of conditioning principles.
User FriendlinessSoftware installation is easy because it runs with few prompts after inserting the disk into the CD drive. The software does not require much memory or hard drive space. System requirements for IBM-compatible computers include a minimum Pentium II processor, Windows 98, 32 MB RAM, an 8× CD drive, 16-bit sound card, and a 600 × 800-color display. System requirements for Macintosh computers minimally include a Power PC processor, Mac OS 8.6, 64 MB RAM, an 8× CD drive, and a 600 × 800-color display.User-friendliness, though adequate, certainly needs improvement.
The most user-friendly programs are those in which one can safely ignore the instruction manual. Sniffy the Virtual Rat was not such a program; designing and setting up experiments are often somewhat difficult, requiring careful attention to the manual's instructions. The intrusion into the manual of historical and theoretical information not directly related to operating the software sometimes makes using the program more difficult. Eliminating this material would significantly reduce the size of the manual. I shall address other usability issues subsequently in discussing details of program operation. Program EvaluationStarting the Sniffy the Virtual Rat program presents the user with an image of a virtual rat exploring its experimental chamber. Researchers familiar with rat behavior will recognize the common exploratory and grooming responses of the virtual rat.
This animation was produced by filming a real rat, selecting frames from the film and linking them together to form these images. In this way, the virtual rat exhibits species-typical behavior, lending a degree of realism to the program. However, the animation is somewhat choppy and could be improved if it contained more frames per second. Fixing this problem, however, may have to wait for faster processing speeds in commercially available computers.The Sniffy the Virtual Rat user interface looks much the same as other Windows® user interfaces. Pull-down windows for File and Edit are omnipresent in these programs. Joining them here are the program-specific pull-down windows Experiment, Windows (i.e., data presentation windows) and Help. New users may gravitate toward the File command that typically houses New File and Open File commands.
One accesses the Design Operant Conditioning Experiment and Design Classical Conditioning Experiment windows through the Experiment pull-down window. Getting started with an experiment will require most users to read the manual, as the control windows are not intuitive.
Designing classical conditioning experimentsOf all the commands in Sniffy the Virtual Rat, users will most likely apply Design Classical Conditioning Experiment and Design Operant Conditioning Experiment. The Design Classical Conditioning Experiment command allows one to set up an experiment in a stage, or phase, format.
For example, the virtual rat can be exposed to a series of unconditional stimulus (US)-only trials before instituting conditional stimulus (CS)-US trials. This procedure will produce data concordant with exposure to the separate stages of a US preexposure experiment.When designing an experiment, the user may choose CSs of low, medium, and high intensity values. The stimuli themselves can be a light, tone, or bell. The CS, however, is fixed at 30 s duration and a shock US at 1 s. In contrast, during higher-order or sensory preconditioning experiments, CS 1 is set at 45 s and CS 2 always occurs during the last 15 s of S 1.
I am uncertain how the differences in stimulus arrangements affect the outcome of the virtual data. These differences may be irrelevant if the programming that generates the data is empirically consistent between basic acquisition and higher-order conditioning experiments. The user manual, however, does not address this issue and leaves one to guess whether the descriptions of the virtual experimental parameters are important with respect to the data generated.Using a conditioned emotional response (CER) method, the program simulates a wide variety of Pavlovian conditioning principles along with several procedures. These principles and procedures include acquisition, extinction, spontaneous recovery, CS and US intensity effects, blocking, overshadowing, overexpectation, inhibition, sensory preconditioning, higher-order conditioning, and background (context) conditioning. The user also may modify each of these procedures in various ways to examine different types of effects (e.g., changing the CS intensity during a blocking experiment). These many possible simulations can easily fill a semester with virtual experiments.Chapter 2 contains a section on the CER procedure.
The authors briefly discuss calculation of the suppression ratio and movement ratios (i.e., freezing) when bar pressing is not the dependent measure. When conducting these demonstrations, however, the software only programs for low, medium, or high intensity shocks instead of allowing the user to choose a virtual shock value in milliamperes. Moreover, the virtual rat responds in three categorically different ways depending on shock intensity (habituation to low intensity, sensitization to high intensity and neither habituation nor sensitization to medium intensity stimuli) rather than showing a range of behavioral variations.Another peculiarity of the program is the use of the variable-ratio 25 (VR 25) schedule as the operant baseline for CER experiments.
In fact, a variable-interval 60 seconds (VI 60 s) would be a more appropriate operant baseline in the CER demonstrations. Why the authors chose a VR 25 is not clear. Again, this may be an issue similar to that of the Pavlovian acquisition and higher-order conditioning differences. The description in the text may be irrelevant to the way the program actually calculates data output. The text does not address this issue, and one is left to guess whether the differences between VR 25 and VI 60 s are relevant to the data output from this program.The discussion of stimulus-stimulus (SS) and stimulus-response (SR) associations is rather bothersome. The program manual reads:As defaults, Sniffy associates the first stimulus with the second stimulus (learns an S-S association) whenever the shock US is used as the second stimulus, and associates the first stimulus with the response to the second stimulus (learns an S-R association) whenever a CS is used as the second stimulus. (p. 26, italics mine)Cited as references on this issue are.
I do not believe, however, that Mazur supports the associative explanation given herein. Such statements are unwarranted in a computer-simulation user manual; they are certainly not the most widely accepted views in the conditioning field. For example, the Mazur text has two paragraphs discussing the SS and SR argument without reference to an animal “associating” (Mazur, pp. 100–101). Designing operant conditioning experimentsThe Design Operant Conditioning Experiment command presents the options for creating an operant experiment.
The window gives options for fixed or variable schedules, specifying time-based schedules in seconds. Applying the Extinction function allows one to control the muting of the pellet dispenser or to apply punishment during the extinction process.
The program allows the user to specify one of several target responses such as bar pressing, face wiping, rolling over, and a few others.There is background information regarding the development and relevance of conditioning throughout the user manual. This is especially true in Chapter 9, the first of the operant conditioning chapters. Chapter 9 is devoted entirely to describing the history of operant psychology, a discussion that I believe unnecessary. In my opinion, the primary role of the Sniffy the Virtual Rat program is as a supplement to standard texts in conditioning.
The authors refer to Sniffy as an affordable way to “give students hands-on access to the main phenomena of classical and operant conditioning that courses on the psychology of learning typically discuss” (p. 1). Courses in learning, however, discuss the historical details of conditioning more thoroughly. Therefore, as a supplement to conditioning texts, this material is redundant.Chapter 10 begins a discussion of basic operant procedures. This chapter describes an automated shaping tutor for those users having difficulty applying successive approximations to a virtual rat. Using the shaping tutor function means the program automatically shapes the behavior of “rear up facing back.” Thus, a student can see the relations between the delivery of reinforcers and changes in current behaviors leading to the target response of rearing up and facing back.The operant conditioning sections of Sniffy the Virtual Rat automatically produce cumulative recordings of the virtual rat's responding. Cumulative records certainly have their place in operant analyses despite their nearly complete disappearance from the experimental literature. The program, however, unfortunately exports these data as a long string of garbage characters.
Allowing for off-line analyses of these cumulative recordings would significantly increase the software's educational value. Moreover, a simple bar-press counter that shows a running tally of responding would enhance the program as well.Chapter 10 also contains a behavioral-repertoire observation exercise that helps students develop observational skills. In two conditions, one with a bar-press-trained virtual rat and another with an untrained virtual rat, a student may record the frequency of each of its eight responses (paw lifting, face touching, head lowering, sniffing, locomotion, bar pressing, eating, and drinking). This exercise can help teach observation techniques and allow comparisons of behavior change between a control and experimental virtual rat. Apart from simple frequency counts, students may use different observation techniques, such as whole or partial time sampling, while watching the same computer screen. This exercise can reveal how observation techniques influence interobserver reliability measures.Another function of Sniffy the Virtual Rat that I particularly like is that the virtual rat will acquire bar pressing if left alone in the experimental chamber.
The rate at which this occurs (as seen under the accelerated time function) can be compared to the time it takes for another virtual rat to acquire the same behavior after having been exposed to a magazine training procedure. I believe that such exercises can provide valuable homework assignments and fruitful discussion in the classroom.In Chapter 13, one learns how to program operant discrimination experiments. The steps to set up the procedure are relatively simple, although they still require the user manual to do so. Variations on S+ only, S− only, or S+/S− procedures may be studied. The DS Response Strength data output from these experiments are more user friendly than the cumulative recordings; however, the user must be careful to identify the appropriate data column for analysis given the poorly formatted column headers. Fortunately, this chapter has a brief description of how to export your response-strength data to a spreadsheet program, which should facilitate its use by the user.In Chapter 14, the program describes how to train the virtual rat to perform responses other than bar pressing. These responses include begging, rolling, and face wiping.
The authors clearly state that the demonstrations are not realistic because they violate the known empirical findings related to food reinforcement and food-elicited behavior. For example, the program allows one to use food reinforcers to develop increases in face-wiping behavior in the virtual rat. More realistically, such training would likely result in much goal-tending behavior and little face-wiping. Their assumptions, however, clearly are there to simplify the programming of the virtual rat. This simplification will not pose a problem for instructors aware of the program's idiosyncrasy or those familiar with the so-called biological constraints on learning.
The authors finish the user manual with an appendix devoted to managing Sniffy the Virtual Rat files as well as a glossary of terms. Both appendices are valuable, and I recommend reading the file management appendix before getting started with any experiment. Remaining commandsSome of the remaining commands include Export, which saves the data generated by the virtual rat into a text file for analysis. Copy Window Image allows one to copy images from the Sniffy the Virtual Rat program directly into word-processing documents, which should facilitate the preparation of student instructions. Remove Sniffy for Time-Out allows the virtual rat to be “removed” from the chamber for studies of spontaneous recovery. Accelerate Time allows the user to speed the rate of data collection in the program.
Without using this command, the user has to wait the full duration of an experiment to obtain results. Accelerate Time certainly saves the user's time; however, it may lead students to underestimate how long real experiments last. Both Pause and Resume will allow the user to take a break from a shaping experiment without the virtual rat emitting appropriate, but unreinforced, responses. The user manual contains a section entitled Quick Guide to Menus and Commands that explains how to operate all the software commands. AdvantagesThe Sniffy the Virtual Rat Pro Version 2.0 program conducts a wide range of simulations.
The program can offer one or more simulations relevant to most conditioning textbook chapters. For example, it simulates many of the most important Pavlovian topics covered in introductory conditioning courses. The simulations include acquisition, extinction, spontaneous recovery, stimulus-intensity effects, compound conditioning, blocking, overshadowing, overexpectation, inhibition, sensory preconditioning, higher-order conditioning, habituation, sensitization, and background conditioning.The program is less extensive in operant conditioning simulations.
Included are magazine training, reinforcement, punishment, secondary reinforcement, simple operant schedules, and discrimination. The program also allows the user to shape several different responses in the virtual rat. Files collect the data and generate graphs within the Sniffy the Virtual Rat program. Users also can access all but the cumulative record files with a text editor to obtain specific numerical information.Beyond the creation of data sets for analyses, theoretical questions are raised with Sniffy the Virtual Rat, as well. Stimulus-competition models describe well many of the above phenomena. In contrast, the program does not contain any simulations of those understood through comparator models of conditioning.
Exploring these differences may engage the students in considering theoretical questions about conditioned behavior. Adding comparator-type effects to Sniffy the Virtual Rat will give it wider appeal to those teaching conditioning. LimitationsAs part of the instruction on conditioning, the authors have added a Mind Window function to the virtual rat. The command Mind Window shows the expected level of stimulus associative value that develops in the virtual rat's “mind.” This function will be aversive to instructors loathe to present mentalistic ideas in their classrooms. Replacing the conceptual sloppiness of the common vernacular with the language of behaviorism is difficult enough in undergraduate conditioning courses.
That task is made all the more difficult when using a software program that lets you “see the mind” of the rat.The Sniffy the Virtual Rat program also claims to be sensitive to variations in intertrial intervals. The user manual states that,You specify the average time Interval Between Trials for the current stage by typing a number into the text boxThe shortest allowable average interval is 2 min; the longest 20 min. Remember that you are specifying the average interval between trials.
The actual intervals vary from trial to trial so that Sniffy cannot learn to anticipate when the next CS is going to occur. (p. 17)After conducting an experiment, one may save the generated data to text files through the Export command. In conducting my class, I expected to be able to use these files to combine data from different students to look at group functions and then to conduct statistics on those data. All students conducted the same Pavlovian experimental design on their own computer. Once the students submitted their data to the class pool, I found that the same data for all students was output for the CS and the background. I also checked this effect on my own computer and found that I generated the same data twice in a row as the students had first produced. Furthermore, this regularity of data showed up in the exported CS Response Strength text file and the exported Movement Ratio text file.
Both of these files revealed the exact same data and in the same format, negating the need for both files. This regularity in the exported data, therefore, eliminated the value of pooling data across virtual rats or conducting a statistical analysis as part of a classroom project. This result also called into question whether variations in intertrial intervals are affecting the processing of data sets, as described in the previous quotation.
It may be that variations in intertrial intervals do affect data output; however, if those variations are consistent across iterations and computers, then the data will not show variations in data output across virtual experiments. This issue is evident again when one tests for hysteresis effects.The Sniffy the Virtual Rat program does not show hysteresis effects and therefore may have limited applicability when used to assess single-subject experimental designs. To test the hysteresis effect, I exposed the virtual rat to 10 trials of Pavlovian tone–shock pairings three times, each time separated by 50 extinction trials. These 130 trials occurred in one session with the use of the Accelerate Time function. The results showed no indication of changing acquisition speed across CS-US training phases. In the interest of fairness, one may be asking too much from the Sniffy program to account for almost all significant behavioral phenomena.
Future versions of Sniffy the Virtual Rat, however, should contain such a hysteresis function so that students may explore single-subject manipulations constructively.Both the CS Response Strength and Movement Ratio text files contained columns of information for trial number, trial type, CS strength, background strength, and suppression ratio based upon the freezing response of the virtual rat. In contrast to the CS and background data mentioned previously, the suppression ratio data across students and across simulations revealed unique data sets.
Therefore, any user of Sniffy the Virtual Rat will need to rely on the suppression ratio data when attempting to conduct group statistics. I recommend that the authors change the way the simulations operate so that each virtual rat produces a unique set of data.Exported data are not user friendly. Column headers do not have the proper spacing to head their columns. Moreover, the files have markers, data and empty columns with unknown relation to the experiment conducted (see for a screen shot of a 20-trial classical conditioning acquisition data file). The user manual needs to explain the contents of the exported data files. I also found that the display of the virtual rat and its data to be unnecessarily small.
A full screen option would facilitate watching the virtual rat's behavior and the ongoing collection of data. A Movement Ratio data file generated from a 20-trial classical conditioning response-acquisition procedure.Note the misalignment of headers to columns and the two rightmost columns with contents of unknown relevance to the demonstration.Apart from the issues of data output, users cannot control the duration of an operant experiment. Operant experiments cease only after filling 10 cumulative-record files or applying the Remove Sniffy for Time-Out command.
Apart from that function, the timeout command experimentally isolates the periods in which experimental stimuli are presented to the virtual rat in spontaneous recovery experiments. It also is applicable in obtaining breaks from using the software. Concluding operant experiments after filling 10 cumulative-record files indicates the session terminates based upon the amount of time it takes to fill these records. Users cannot program the experiment to end after a chosen number of responses or consequences. This command is unlike that in the Design Classical Conditioning Experiment interface that requires users to specify the maximum number of trials in the experiment. Conditioning and Research Methods LaboratoriesMost certainly, Sniffy the Virtual Rat can play a large part in an introductory conditioning course. There are enough simulation exercises to occupy a full academic semester.
Research methods courses also may benefit from the use of the program. For example, the program can be used to assess interobserver reliabilities and their calculation. Other possibilities include evaluating two-group studies in which each student combines suppression ratio data from their virtual rat. Applied Behavior Analysis CoursesThis software also may serve applied behavior analysis (ABA) courses through its demonstrations of operant behavior. Sniffy the Virtual Rat's programming allows one to simulate the shaping of four primary responses (bar pressing, begging, rolling, and face wiping). I believe this shaping function was the original purpose of this software.
Students conducting the shaping exercises are fully engaged in watching the virtual rat and deciding when to deliver virtual reinforcers. Because the program covers basic operant phenomena and allows for the comparison of response–observation techniques, I believe it is well suited for an ABA course. ConclusionThe Sniffy the Virtual Rat program contains most of the phenomena treated in an introductory conditioning textbook. Thus, an instructor can use the program throughout the semester to demonstrate a wide variety of conditioning phenomena, including the collection and analysis of data on these topics. Data usable for statistical tests are limited to suppression ratios; however, all data can serve as the basis for theoretical discussions about behavior.Future versions of Sniffy the Virtual Rat should add a stochastic process to the CS and background data output from the program.
Such changes would encourage the use of descriptive and inferential statistics with the data, serving a pedagogical connection to other experimental psychology courses. Other additions should include simulations for fading, matching, and the temporal control of behavior. These are such basic phenomena that I strongly urge their inclusion in a subsequent edition of the program.The program also needs to improve its user friendliness. A more intuitive set of controls and a full screen option for the display of data and the virtual rat will improve the user's experience.
Furthermore, a reduction in text devoted to nonsoftware issues would significantly reduce the size of the user manual.With respect to the shaping simulations, it would be helpful to see output correlating the program's optimum times for virtual reinforcement and the user's actual responses. Such an upgrade will give users some feedback as to how well they shaped the behavior of the virtual rat. In addition, I suggest that the authors consider a virtual human in a vocational or educational setting. Students may use that simulation to train the virtual subject to perform more on-task behavior, at higher accuracies, etc. Such a simulation would mesh well with applied behavior analysis coursework as well as introductory conditioning courses.The criticisms described herein should not detract any potential user from purchasing this software.
It is well worth the list price of $30.95 and will likely support college-level teaching on conditioned behavior for many years to come. The principles of learning and behavior (4 th ed.) Pacific Grove, CA: Brooks/Cole; 1998. Gibbon J, Balsam P. Spreading association in time. In: Locurto C.M, Terrace H.S, Gibbon J, editors. Autoshaping and conditioning theory. New York: Academic Press; 1981.
Graham J, Alloway T, Krames L. Sniffy, the virtual rat: Simulated operant conditioning. Behavior Research Methods, Instruments, & Computers.
1994; 26:134–141. Mazur J.E. Learning and behavior (4 th Ed.) Upper Saddle River, NJ: Prentice Hall; 1998. Rescorla R.A, Wagner A.R. A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement and nonreinforcement.
In: Black A.H, Prokasy W.F, editors. Classical conditioning II: Current research and theory. New York: Appleton-Century-Crofts; 1972.