Final Post

Training Rat was an incredible experience! Not only was this my first time dealing with animals professionally, it was also my first time training using a proven, well documented method. Initially I was apprehensive; I was afraid of hurting Rat, of what Rat might do if she felt defensive, how aggressive the animals were, etc. But now after having experienced the rats first-hand, I would gladly do this as much as I can.

Training dogs (and attempting to train a friend's cat) is something that I have dealt with multiple times before, and I have done so without formal training of my own. Now that I have been educated in training and have real world experience, I feel confident that I can train just about anything, given a proper reinforcement and enough time.

Sniffy was very helpful, but in the end I had more problems with Rat, and that makes perfect sense. A program is designed to be perfect (although it may seem flawed to the user, it is still programmed to be perfect); a real animal is not. A real animal teaches you patience and how to overcome unforeseen obstacles. Rat did just that.

I have gained a lot from this experience, and I am immensely grateful for this opportunity! If I have learned a single thing from this first project, it is that I cannot wait for the second project. I am looking forward to the challenges facing me in the upcoming weeks to come! Rat, don't fail me now!

Sniffy the virtual rat vs. Rat the live rat

Training with Sniffy was an invaluable experience. I was able to learn about magazine training multiple times (as I restarted the program), shaping, and different reinforcement schedules. Similarities abounded but so too did differences. The major thing that stood out to me was the difference in timing between Sniffy and Rat.

Magazine training with Sniffy was painfully long, almost to the point of discouraging any hope of rapid progress. Maybe it was because of the expectations that Sniffy instilled, but magazine training with Rat was comparatively simple. It took almost no time at all, and within a few minutes (as opposed to nearly a hour), Rat was magazined trained.

Shaping was also comparatively simple, to a degree that is. Unlike Sniffy, who roamed all throughout the box and who would go back to the magazine to receive reinforcement when the "click" was heard, Rat stayed right by the one wall. She never wondered much at all, and it was easy to have to go to the bar. Unlike Sniffy, I never had to reinforce rearing behavior. With Rat, I would reinforce her when she was in the vicinity of the bar, then when she was at the bar, then when she sniffed the bar, when she touched the bar, when she had one paw on the bar, and two paws. It was an absolute breeze. Now here is when the major difference comes up: Rat WOULD NOT press the bar! She pressed it a few times by accident, but I couldn't get her to press the thing enough to elicit an association! It was simply painful watching her because I didn't want to extinct the behavior of being near or on the bar. Eventually Rat picked it up and quickly progressed.

The rate of progression for Rat was great, and I was constantly rewarded whenever she would press the bar or improve on her previous training session. Although she was quick to learn and quick to progress her learning, she was not even on the same planet as Sniffy. Within a few days of training, Sniffy was able to press the bar 50 times (nearly all in a row) with not noticeable PRP. I thought this was a little unrealistic, but if magazine training took a long time, maybe Rat would be even quicker, right? Totally wrong. Although Rat did max out the program at FR20, she was barely half way to FR50, like Sniffy.

I think Sniffy was an extremely helpful tool. But it was just that: a tool. Sniffy definitely was not an adequate replacement for an actual rat, and there were things built into the program that were completely unrealistic. That being said, I am extremely glad I was able to experiment with Sniffy before training a live rat. 

Sniffy Training

Training sniffy was definitely interesting. The graphs depicting Sniffy's associations with the sound and food, the bar sound, and the action strength of the desired behavior was extremely helpful in learning the steps for training and for grasping the notion of how training and learning occurs. One thing that I found frustrating, to say the least, was the required level of each bar before moving on to the next step. Although I can understand the principles behind why the programmers would have it that way (to truly mash it into our brains), it nevertheless was frustrating.

That being said, magazine training took much longer than I anticipated (nearly three quarters of an hour). Although it seemed like Sniffy had clearly built an association between the magazine and reinforcement, the program refused to believe it saying something along the lines of, "you're nearly there. Keep going and Sniffy will have a strong association so you can move on."Other than the time it took to magazine train Sniffy, the process was simple, and I was quite pleased by her progress (my own little reinforcement).

When it came to shaping, I was again shocked -- this time for the opposite reason. It took Sniffy only a few minutes before she was pressing the bar. The difference in times it took for magazine training to be completed and before her first bar press was striking and truly stunning. All I simply did was reinforce Sniffy when she was near the wall with the bar. I also reinforced her when she reared in the air, regardless of where she was in the box. Soon enough she was rearing only on the wall with the bar after I was more exclusive in my rewarding. After she pressed the bar a few times, she caught on extremely quickly, and from there on out, I only reinforced her when she pressed the bar. The ease of shaping and the speed at which she began pressing the bar gave me the false belief that Rat would be equally as easy.

Moving from CR (FR1) to FR2 was simple with almost no lag. I quickly went to FR3, FR5, FR10, etc. at an extremely fast rate of progress. Within a day or two (while also experimenting with different VRs), I had Sniffy doing an FR50. It quickly became like a game to me, seeing how I could manipulate Sniffy's behavior.

Although I did use the book as a guide, I took a much more organic approach to Sniffy. I would raise the frequency necessary to achieve a reward; I would change from FR to VR back to FR to FI to VI and repeat the process just to see how each schedule would effect behavior and how the cumulative record would reflect that. I ended up extincting Sniffy multiple times just for fun, only to bring her back to ratio schedules in the 40s or 50s.

Sniffy was not only a fun experience, but it was an extremely useful and interactive tool. I was able to get a grasp on how operant conditioning works and how different schedules at different rates (of time and ratio) influenced behavior. I am very glad that I worked with Sniffy as much as I did before training Rat.

Obstacles

Although Rat was extremely compliant and very easy to work with, there were still some obstacles. Probably the biggest had to do with extinction. During training, Rat was doing so well that I decided to continue training for an extra two days. This in itself made it more difficult to extinct her in the recommended two days. An additional confound was the fact that I was "jackpotting" her by doubling or even tripling her reinforcement for a job well done. This increased the probability that she would perform the action (i.e., pressing the bar) more consistently and more completely, in that I would only jackpot her when she pressed the bar an exceptional amount of times in a row. Unfortunately, this made extinction even harder.

The compounding effects of an additional two training days and jackpotting good behavior during the last few days of training made extinction even harder (not only for Rat but also me to watch her hard work be destroyed). Her first day of extinction proved to be largely ineffective, as the rate at which she pressed the bar barely decreased. The second day's results provided similar results, although she definitely was not pressing the bar quite as much. It still was not satisfactory in my eyes; she still was associating the bar with reinforcement, and that association was still very strong. In order to overcome these obstacles, I had to add a third extinction day. All in all, extinction was a hard process to watch; Rat worked so hard and very quickly learned what she needed to. Having all of that work nullified was simply a shame.

Pictures!

Don't mind me. I'm just holding my Rat.

I wonder what she thinks when she's in The Box...

Trying to escape

Almost there! Just a bit more!

Data Analysis and Overview

Using SPSS, I was able to analyze a wide range of data to find descriptive statistics as well as discover trends and progressions of such trends.

Weight

Although Rat weighed 232g initially, her target weight was set at 87% of her body weight (202g). Once the food deprivation state was complete, Rat's weight (M = 202.76, SD = 3.45) was adequately kept on target with the proper amount of food given per day (M = 7.81, SD = 2.04). There was slight variability in weight change per day but very little average change (M = -.38, SD = 4.96). Although variability existed in higher amounts than are ideal, the weight change per day varied nearly equally both ways, leading to an overall average weight that was extremely close to the target weight, nullifying the concerns of variability.

Weight Change as a Function of Food Given

Performance

By excluding shaping sessions, it is possible to view Rat's performance (i.e., the number of times she pressed the bar) from FR1 to FR20 and through to the end of her third extinction session. Her performance (M = 224.07, SD = 89.59) was quite impressive, and her variance shows the amount of improvement that she showed over time with a range from 99 to 404 bar presses. Rat's performance was facilitated by the use of reinforcements over the length of the training session (M = 28.33, SD = 32.16) and ranged from zero to 135 reinforcements. When compared to the minimum number of bar presses, the high maximum number of bar presses shows how quickly she progressed (y = 8.91x + 152.75) with a decreasing number of reinforcements (y = -5.17x + 69.71).

Reinforcement as related to Bar Presses

Day 18 - Extinction

Time: 30:00
Protocol ID: 25 (Extinction)
Event 1 Total: 121
Reinforcement Total: 0
Goal: Under 100 bar presses
Goal Reached: No

05:00 - 19 : 19
10:00 - 70 : 51
15:00 - 73 : 3
20:00 - 77 : 4
25:00 - 81 : 4
30:00 - 121 : 40


1:30 - Rat initially went to the bar and had her paws on it without pressing down. She then wondered the box, sniffed the magazine (very briefly), and wondered a bit.

2:45 - First true pressing "session" (more than one at a time) of 5 presses. She is grooming more often and sniffing, but not chewing or scratching, the magazine.

5:00 - Second session of 8 presses; she is starting to chew and scratch the magazine again but with less effort than on Day 17.

8:00 - Sessions are consistently lasting around 5 presses. It seems as if her association between the bar and the reinforcement treat is decreasing, but she still strongly associates the magazine with reinforcement.

9:00 - Long session with ~10 presses, 5 second pause, ~10 presses

Time breakdown: 1/2 wondering, grooming, 1/4 standing still, 1/4 sniffing, scratching, or chewing magazine and pressing bar

12:45 - It took 2:45 for her to press once and only once; she immediately started grooming to calm herself.

14:45 - Another 2:00 passed before one and only one bar press occurred. She immediately went to check the magazine. After doing this again, she slowly wondered away.

16:30 - She is spending more time on the wall with the door looking out and rearing up than on the magazine-and-bar wall. It seems as if she is truly starting to become extinct - to have her slate wiped clean, if you will

Time breakdown: 3/4 looking at, sniffing, and rearing on the door, 1/8 grooming, 1/8 looking at, standing by, or sniffing the magazine (but no chewing or scratching). She seems to be forgetting her previous association between the magazine and reinforcement.

23:00 - Took ~8:15 before a single bar press followed by a sniff of the magazine and ~0:45 before another single bar press.

25:00 - Session lasted only 2 presses with ~5 seconds between press and only ~5 seconds of sniffing the magazine after each press.

26:15 - Two scratches of the magazine followed by a single bar press. She is no longer grooming herself (and hasn't been for the past ~10 min) after pressing the bar. She is no longer becoming frustrated after pressing the bar and apparently does not feel the need to groom to calm herself after each failed press.

29:00 - Session of 7, sniff; 4, sniff; 5, sniff; 13, sniff; 4, sniff; 6, sniff

With the exception of the burst at the end, I think you can fully consider Rat extinct. Again, with the exception of the last 3:45, she exhibited little regard toward the bar or magazine, and the amount of time she spend even at the magazine (whether it be sniffing, scratching, or chewing) decreased drastically.