Cat Nipping

So, because Rikki was super interested in the buttons and acting like he wanted to communicate, I decided to help him figure out how to actually press them. Starting in the third week of October, I began trying to target-train him, similar to this video. I started with a little glass container through which he could see the treat, so that he’d bat at it, and then bit by bit, worked on transferring that behavior to one of the starter buttons (that I had replaced on the actual board with the new electronic ones).

The combination of Rikki’s intense food motivation and his, shall we say, below-average ability to make cause-and-effect connections … did not put us in the path of success.

What’s supposed to happen with this kind of behavior shaping is that the animal associates a particular behavior with praise and/or treats, and then you make the tiniest shift in what behavior is rewarded, so that over a series of many micro-adjustments, you guide them toward the actual desired behavior.

It was slow going, but eventually I managed to get Rikki to understand that the treat had something to do with a paw. When all of his pacing and plaintive yelling and attempts to headbutt me (the cat version of “please!”) bore no fruit, he would belatedly remember to paw in the general direction of whatever was in front of him: button, tile, nearby grocery receipt … The paw would usually, but not always, make contact with the object.

Anytime I tried to ‘shape’ the behavior further — by only rewarding top-of-button touches, for example — he just rolled right into frantic extinction response. I moved a paw! Why you no treat?!? WHY??!? As far as he was concerned, I was standing there randomly doling out treats — or withholding them — for no reason. I had rerecorded the button with a meaningless BOOP sound, and I gave him ‘jackpots’ — an extra flurry of treats — whenever he accidentally pressed the button loud enough to make it BOOP, but he never paid that noise any attention at all.

I tried putting a laser-pointer dot on top of the button, but he wouldn’t reliably pay any attention to that, either, being so distracted by the imminent possibility of treats. So I just kept going with the micro-adjustments, despite his plaintive distress. Finally, there came a day when he seemed to be reliably flat-touching the top of the button, and making it BOOP maybe about a third of the time. Internally I celebrated our success … at last, we were almost there! But then I could never get that behavior again in subsequent sessions, except by accident.

Usually animals will notice when an accidental behavior has a positive result, and come back for more. Maybe he just needs more frequent examples, I thought. So I took four more of the old buttons, set all five to BOOP, and stuck them together on a tile. Then I lured him back and forth across the cluster until an accidental step triggered a BOOP. Jackpot! BOOP. Jackpot! BOOP. Jackpot! … No association. No hint of him ever trying to do it on purpose.

Maybe the problem is me, I thought. Maybe I’m just not accurate enough with the timing of my verbal responses. Right around this time I was reading a Karen Pryor book, Reaching the Animal Mind, and I ended up convinced by the evidence that using an actual clicker, as opposed to a marker word as I’ve always done before, is more accurate, more compelling, and gets faster results. So, I bought a clicker and switched to that instead.

… nnnnope. Not measurably better. Practice made me more accurate with the clicker, but still nothing was clicking for Rikki.

At this point I went back to a single button, and trying only for intentional presses. He still didn’t seem to be associating the clicker sound with the treat, which is the one response the whole system is predicated on. He also didn’t seem to notice that he only got treated when I gave a cue word first (“touch”, in this case). I ended up adding “sit” and swapping that out with “touch” in hopes of getting him to pay some attention to the human mouth noises.

Nothing I could find about operant conditioning or behavior shaping tells you what to do when the animal fails to make even the most basic cause-and-effect associations. All I had really succeeded in doing by starting up with the target training was to make Rikki desperate for treats (which had not previously been a part of his life), and prone to yelling and vaguely dabbing a paw at random things.

Meanwhile, in a fit of optimism (on the day I thought he’d mastered the full-paw top-of-button touch), I put a CATNIP button on the main board. Rikki loves catnip almost as much as treats, so I thought this would be a good modeling option that would not directly lead to obesity. But I’m afraid I overestimated my little dude.

Tashi, on the other hand, immediately became obsessed with the new button. All day long: CATNIP. CATNIP CATNIP. CATNIP.

I mean, I suppose I should have expected that. This was the first time I’d put a button on the board that wasn’t for him. And it was on the red tiles — apart from, but in the vicinity of, the super-important FOOD and KONG buttons. I let him sniff the dried catnip, hoping that would be uninteresting, and tried to explain with the limited buttons we had so far: CATNIP GRACIE RIKKI … NO CATNIP TASHI. And: CATNIP GRACIE RIKKI … KONG TASHI … TASHI WANT KONG HMM?

At this point, it helps to know that at no point have I seen, either from Tashi or any button-using pet on social media, a strong indication that word-order is a thing they find meaningful or salient, despite much human modelling being word-order dependent. As far as I can tell, words from pets tend to come in associative clusters with random order. I sometimes seriously wonder if dog thoughts — because they clearly do have thoughts! — are based on spatial associations or some other model that sounds-in-linear-time cannot reproduce, kind of like the aliens in the movie Arrival.

But anyway, over the course of about three days, Tashi’s CATNIPs stopped being mostly single-button presses and started becoming incorporated into larger clusters. A couple included KONG, which might either lend credence to my theory that he expected some kind of food-related yumminess from a button in that area, or else just indicate that he’d tracked my use of catnip in a cluster with KONG in my modelling, but those quickly diminished in favor of emotion clusters: CATNIP EXCITED. CATNIP FRUSTRATED HURT. And quite a few were in some combination with GRACIE and/or RIKKI, both with and without accompanying emotions. (Under the circumstances, the idea that CATNIP GRACIE RIKKI was Tashi asking me to give catnip to the cats was … beyond improbable. There is zero relationship basis for altruism, here.)

The cat names, like KONG, could be imitation of what I’d modelled … but that didn’t explain the emotions. And I was pretty sure he understood most of the five emotions on the board at that point — at least FRUSTRATED, EXCITED, and BORED. HURT was getting used a lot but often not in ways that made sense to me; MAD he only rarely touched — which actually increased my belief that the other four were being used with intent, because random or experimental presses would have been either roughly equal, or weighted toward the closest buttons — and MAD was very convenient, much moreso than EXCITED. Whether he didn’t understand MAD well or simply didn’t feel it often, I wasn’t certain.

Then, about five days after the addition of the CATNIP button, as I despaired of Rikki ever putting one and one together, Tashi pressed, not just CATNIP FRUSTRATED, but CATNIP FRUSTRATED MAD. Which … honestly, made me think he understood MAD just fine. He gave every indication of having escalated from simply being confused and frustrated by a button that didn’t do anything to being actively pissed off about it. I thought about human teenagers and the way that alcohol often takes on a heightened mystique when it is The One Thing You May Never Have, but when parents’ attitude is ‘here, have a sip’, they’re more likely to decide it’s awful-tasting and not worth it.

So I stopped trying to tell Tashi that catnip was for cats and started treating it like any other request — either giving him a tiny amount (which he licked up out of my hand like he thought it was good) or telling him CATNIP ALL DONE.

But … still the obsession with CATNIP continued, more or less unabated. And then a day and a half later, in the hour before dawn, I was awakened by a sequence of sounds, one familiar enough that I could envision the scenario with my eyes still closed:

Tashi sniffs invasively at Gracie’s butt. Gracie meows a complaint. Tashi noses back at Gracie harder. Gracie complains louder. More nose, more butt. Gracie gets fed up and starts hissing and swatting. This sends Tashi into play-fight mode and he starts biting at her. Then she gets really furious and starts fighting him for real, with all kinds of screeching as dog teeth make contact.

This whole escalation took much less than a minute. Jak was away on a trip and not here to be awakened, so I yelled from bed the command I’d been training mostly for this very situation: “Tashi, leave it!” — and then waited to see if it would work. … Sound of toenails against tile as Tashi moved off. Between five and ten minutes later, still lying exhausted in bed, I heard the first buttons of the morning: CATNIP HURT FRUSTRATED.

The juxtaposition gave me a lightbulb moment: Tashi has been using CATNIP to mean ‘cat’. Of course he knows the word ‘cat’; we say it to him and each other all the time. It’s even in my soundboard plan, but not a button I’d prioritized. It might even be my own fault: he might have taken CATNIP GRACIE RIKKI not to mean “catnip is for Gracie and Rikki” but as confirmation that “catnip is Gracie and Rikki”.

So CATNIP FRUSTRATED MAD would have meant he’s frustrated and mad at a cat (or vice versa, because I do say “Gracie mad” a lot, but more likely talking about his own feelings). About ten minutes after that, I’m trying to pee while he’s dancing around me in the bathroom and nipping and generally being a nuisance, so I said, “what Tashi want, hmm? what words?” — trying to get him to go off and use the board while I get my contacts on.

He trotted off and: CATNIP HURT. Okay, clearly he’s still bothered about the interaction. And yeah, from the sound of things, there was probably mutual pain involved.

I mean, I felt bad for him. Poor thing is so frustrated by cats being cats and not dogs! They hurt! They don’t play right!

A little later, I looked back over the recent button record. The night before, while I was watching television, the incomprehensible CATNIP RIKKI SETTLE, which I basically ignored. Yes, I remember noting at the time that Rikki had just settled on the back of my armchair behind my head; it was the inclusion of CATNIP that threw me. CATNIP EXCITED HURT RIKKI TASHI CATNIP BORED BORED BORED HURT FOOD MAD — I don’t pretend to understand what that means exactly, but if you read CATNIP as “cat”, it almost sounds like a reasonable story, from a dog who keeps trying to play with (or eep, chase and eat?) cats and failing.

Most compellingly, although I couldn’t be certain of the timing in hindsight, it sure looked like OUTSIDE CATNIP EXCITED — which he pressed twice the evening before, almost an hour apart — would have lined up with Gracie managing (unbeknownst to me) to pry open one of the screen doors and let herself and Rikki outside. (The resurgence of Gracie’s Escape Artistry is a whole other story.)

It was enough to convince me, so one of my first tasks that morning (this was November fifth; life has been crazy and I am no closer to catching up) was to retrieve the blue tile from my Rikki training attempts and add PERSON, DOG, and CAT buttons to the board. I also rerecorded the CATNIP button to say “kittynip”, in hopes of resetting its meaning. When I demoed them for Tashi, he had zero reaction to the changed KITTYNIP, but went full-on stare-at-speaker head-tilt for each of the new buttons, which almost certainly means he knows all three words.

Afterward he switched between CAT, the cats’ names, and KITTYNIP, but KITTYNIP got more use, I think because the red tiles are more convenient to his most common approach direction than the blue tiles. A couple of days after that, I made another call: I gave up on trying to make a joint board for both cats and dogs — pulled all the cat-specific words from the board plan — and removed the KITTY/CATNIP button entirely.

If Gracie ever decides to give it a go, I’ll readjust, but I don’t think Rikki has a snowball’s chance of ever using the big board. Lately he’s started going over to the main board and gnawing on the buttons in his frustration. My plan — once I can get another spare tile — is to make a little mini-board with a couple of my old non-connected buttons and keep it on the kitchen counter to redirect Rikki toward. CATNIP and PRACTICE, probably (“practice” is what I call the clicker training sessions, and it’s a way he can earn treats). If he learns those two, great, I can add a couple more, like PLAY. If not, that’s fine too. But I’m afraid a giant board full of buttons, almost all of which have no concrete result, is doomed to perpetually confuse and frustrate him.

Since then, Tashi has used CAT a few times, but more often he uses their individual names, just one or both together. (He has a lot of Feelings about the cats.) He uses PERSON sometimes, too (maybe when he wants me or Jak but doesn’t care which?), but DOG only once that I’m aware of, to refer to himself. I expect that ratio to change dramatically if he figures out that he can talk about his friend Chewie that way …