Table of Contents
Description of Problem
Fitz is a two-year-old border collie, adopted three months ago. Previous owners reported that he was scared of “diesel truck” noises.
J. reports that Fitz is likely scared of the sounds of individual trucks on the road at the end of their country driveway. These are audible in the home to human ears but not loud. J. sees him freeze before she hears/notices the truck sound.
When he hears a scary sound, Fitz generally freezes, ears down, tail tucked, then hides under the bed. He seems to feel more secure under the bed in general.
He is at times scared to go outside, to the point that she has created a toileting arrangement for him inside for when he needs to potty at night. Fearfulness outside may be due to truck noises.
Individual truck noises originating on their road are audible at the house. Traffic can sometimes be heard from a busier road a fair distance away.
Fitz also hates riding in the car and appeared to react to vehicle noises when in the car.
- Amazon vans and other smaller vehicles
- White noise
- Brown noise from a noise machine
Possibly worrisome sounds
- Boiling oatmeal (probably because novel)
- Beeping of bread machine (probably because novel)
- Vacuum (generally normal dog response)
Scary but not top-tier scary sounds
- Two-stroke engine close to steady state—came out from under bed when it was quieter
- Boiling of teakettle (this might merit top-tier?)
- Diesel trucks
- Probably other large trucks
- Recorded truck sounds
- Pickup truck at the house
- Propane truck at the house (including beeping and backing)
- Car alarm
There are several possible aspects of truck sounds that might be triggering Fitz’ fear. We will need to identify them in order to create non-scary truck or trucklike sounds.
Among the possibilities are:
- The distinctive “rattle” of diesel truck engines
- Lower frequencies generated by diesel trucks
- Lower frequencies generated by any large trucks
- Acceleration noises of trucks (diesels have a distinctive sound)
- Vibrations from trucks
J. has tested Fitz with brown noise from a machine, to which he was indifferent. (Great!) She will next test him with brown noise from the YouTube video Super Deep Brown Noise generated on high-quality speakers.
She’ll start at a low volume and work up if Fitz is OK.
Result: no problem.
If Fitz is OK with brown noise at a moderate-to-high volume (what the humans can stand), this may mask some of the truck noises. Sounds used for masking have to include frequencies as low as, or lower than the problem sounds. Dogs can hear down to about 60 Hz, which is the low end for most good speakers. A good match. So brown noise that includes frequencies in that range may successfully mask truck noises from the road (but not in the driveway).
Fitz’ tolerance of brown noise may also give us a clue about what is scary about the truck noises. If he tolerates brown noise just fine, it may not be the low-frequency rumble of the trucks, or the brown noise may help us find a desensitization “path” to the particular rumble of trucks.
J. may experiment with creating a “cave” for Fitz on the upper floor of their house to emulate the “under the bed” environment.
Testing for Desensitization Starting Points
We will test Fitz with various truck sounds that I will filter to separate the variables listed above.
The first test will be the sound of a diesel engine with the lower frequencies filtered out. Fitz’ response will provide an indication, though not a certainty, as to whether it is the diesel sound that is bothering him rather than the low-frequency rumbles of any trucks. (Although the truck noise recording he reacted to could indicate that the trigger is trucks in general.)
The next tests will depend on the outcome of the first.
The following image shows the Fast Fourier Transform of a diesel truck sound played on a decent Bluetooth speaker with good bass response. The x-axis is frequency in Hz and the y-axis is sound pressure in decibels. The sound includes robust frequency components down to approximately 60 Hz.
The following image is the same sound played through the internal speaker of an iPad Air 2. Note that it can’t generate the lower frequencies characteristic of a truck. Anything below about 200 Hz is inaudible. This is typical of Apple handhelds (I haven’t tested other brands, but they are probably similar).
We probably won’t want to use a handheld to generate any of the test noises; I just wanted you to see the difference.
For the test noise, I applied a digital filter to the original sound, filtering out frequencies starting at 1500 Hz with a rolloff of 12 dB per octave. That sound looks like this when played on the speaker.
If you listen to the filtered sound on earbuds; make sure Fitz is not around. I found out the hard way that dogs can hear what comes out of earbuds from across the room at even a moderate volume.
When you are ready to test, play it at a low volume on the speakers. Be ready to turn it off if you see a bad reaction; no point in playing the whole 14 seconds.
The beauty of sound conditioning is that you can be sneaky with your setup to avoid predictors. After you’ve set up your cables and done some other sound tests over time, you can just be sitting innocently at your computer one day and play the sound anytime the conditions are right for Fitz. Just be careful about food handling; later when you do DS/CC you’ll want to randomly have the yummiest food out where he can smell it (but not get any) so the odor doesn’t become a predictor. You can start those habits now when doing tests.
If we get successfully to DS/CC and we know the sounds won’t scare him, I’ll build in some lead time to the sounds so you can press Play and then walk off, so you aren’t always sitting at your computer when the sound starts.
Test Sound Round 1: High-pass filtered diesel sound.
Results of Test Sound 1, 3/21/20
“He thought about it, then decided he wanted to go under the bed. He was convinced however to stay and eat cookies instead. (I tried hard not to pressure him, said COOKIES! and delivered a scatter, then ignored him and fed the other dogs.)”
If vibration is the most significant trigger, there is probably nothing I can do to help, but C. might be able to build Fitz a cave that is decoupled from the structure of the home enough to give him some relief.
Discussion after Test Results
I’ve identified two possible paths for sounds for Fitz.
- Start with a vehicle sound that he is not scared of. Have a “goal sound” of a truck that he is scared of. (We probably wouldn’t have to test this.) Cars and SUVs lack the lower frequencies of trucks in general as well as the specific sounds of diesels (usually). I can manipulate the initial sound toward the goal sound to create a set of sounds for desensitization with as many splits as necessary. When/if we get to the target sound and get a CER+, we could do the same thing, with probably fewer gradations, to a different truck sound.
- Start with a truck sound heavily masked with brown noise, such that the truck is inaudible. This would be the initial sound. Gradually bring up the volume to the truck and also “thin out” the brown noise to get to the target noise of the truck alone. The same as with the other option, we would then want to do it with a different truck noise, but could probably start farther along in the process.
If we continue, I’ll create a test sound for whichever path we choose. Once we have that, I’ll create the series of sounds, or at least the first part of it. I can also create a sound “between” two sounds in the series if we get an adverse response. Sound editing allows for very fine splits.
Pros, Cons, Considerations
- Always using the big speakers means that the response can get tied to that location/setup/speaker. It’s possible that deliberate exposures to different, real-life car sounds could help with that. It also wouldn’t hurt (might not help, but wouldn’t hurt) to play some benign car noises on other devices, even given the frequency response limitations.
- In either option, there’s a danger of making him scared of (respectively) more vehicle sounds, or brown noise in general. However, he does seem to have some flexibility/bounce back. We could stop at one scare or retreat to the less scary sound in the series to try one time.
- There’s a danger in not taking action as well, if his fear generalizes to vans, then passenger vehicles. Could possibly hold that off with CCing known, controllable real-life vehicle sounds. (Which seems like a good idea anyway. Wouldn’t hurt to separate out the car sound from his other fears of it.)
- There is a part of this for which I don’t have evidence, but maybe you know of some. Dogs discriminate sounds really well. E.g., some old Russian experiments showed that for individual pitches they had better discrimination than humans. Today we notice that they distinguish the family car sounds from “stranger” car sounds. But this means that as we do DS, if we dwell on any particular gradation long enough to get to a CER+, we **might** be diminishing the ability to change to the next one. With our sound series, we are trying simultaneously to get similarity and difference. When I did this with my own dog, I took her to about “neutral,” on the early sounds, then made the change. I did more iterations on each as I got closer to the original target sound. The optimum number of exposures to get a CER+ is smaller than we think (I’d have to look it up), then we get a plateau. If we “spend” lots of exposures on getting a wonderful, drooly response to an early iteration, I don’t know if that will lessen the response to the next iteration. In other words, I don’t know how generalization and discrimination interact in this situation. I’ll recheck the Rescorla Wagner stuff on my own time; there may be something in there. Also, I’ll check the one scholarly article that used varied pitch in DS. This issue does seem to “work out” with DS/CC to people.
- It’s hard to tell with the info we have from Fitz so far regarding generalization. There’s the smaller engine issue. On the other hand, he’s OK with passenger vehicles and vans (yay)! Could it be that habituation has helped with that? We can’t know it, of course, but that does seem like a plus with regard to trying DS/CC, along with the other things you have described that he has learned to handle better.
- Disclaimer: I think we are all on the same page with this, but there’s a possibility that, even if we don’t scare him, we won’t make much progress. There are so many variables. There are some dogs, both from anecdotes and found in studies, that do not associate sounds from speakers with “real life” sounds. I think that came out in a fireworks study. But we do know that Fitz didn’t like the recorded truck noises, so that gives me hope that the association can carry over the other direction.
So there are my thoughts, for now. See what you and N. think, and tell me if you want to proceed. If you want to proceed, we can discuss which path to take.
Decision on Proceeding with Test Sounds
J. and N. chose the first option I described, morphing from an acceptable vehicle sound into various truck sounds. J. will make recordings of their vehicles.
Recording Automobile Sounds
The goal is to get recordings for both vehicles of the following:
- steady state (idling)
- revving the engine
- car approaching the house.
- Practice with the audio recorder a few times, recording and saving files. Figure out how to export as a WAV and how it can be shared. (E.g., email, Dropbox)
- Choose a quiet time for the car recording.
- Do this away from Fitz if possible (because of the revving, just in case).
- Put your phone on a tripod or have a person hold it about 15 feet from the car. Turn the recording app on.
- Get in the car, close the door, wait a second or two, then start the car.
- Let the car run (idling) for 15 seconds or so.
- Rev the engine a few times. Let’s say once with a pause, then twice in a row.
- Done. Turn off the car, then end the recording. Save the whole recording.
- Do the same for the other car.
This will give us the idling and the revving.
Then do the same thing with each of the cars approaching. It doesn’t matter how far ahead of time you turn the recording on (this can be a psychological barrier to get over, but audio files are smaller than video, so it won’t get out of hand if it’s too long). If it captures you walking to the car, starting the car, driving away, and driving back again, that’s all fine. We want the “car approaching” part but I can edit it out of the whole file. During this whole recording, don’t drive closer than 15 feet to the mic on the way back.
Background sounds will be tricky. I can’t easily edit them out because of the wide bandwidth of the auto sounds. Birdsong is probably OK if it’s unavoidable, but some of it would likely stay in the recording.
If, say, someone starts hammering (just an example) during the recording, best start over. Or if it’s during the idling, and it stops, you can just record longer.
Please let me know if I left anything unclear.
Test Sounds Round 2
Here are the edited test sounds. I faded them in so they wouldn’t be abrupt. The recording of the Echo idling is shorter than the CRV because there were several seconds where there were some thumps in the Echo recording, which I think may have been from wind hitting the mic.
I threw in the fifth one that has a little pastiche of both cars revving. I would play that only if he was perfectly fine with the other revving recordings.
We will test approaching cars after you can get those recordings made.
CRV idling. Response: considered running away, changed his mind when I shut it off and tossed him treats
Echo idling. Response: interesting, a little tense
CRV revving. Didn’t do because of response to Echo revving.
Echo revving. Response: very tense, very interested, did not run
Multiple revs (only if he sails through the others). Didn’t do.
Test Sound Round 3
Echo idling with frequencies below 1500 Hz attenuated. Response: no problem!
Get another recording of the Echo (steady state idle and a slow rev). Use the steady state alone, first. Filter it as above (Test Sound Round 3.) Create a series of sounds from that sound that adds back in the lower frequencies gradually until we have reached the unfiltered recording of the Echo.
Desiderato, O. (1964). Generalization of conditioned suppression. Journal of Comparative and Physiological Psychology, 57(3), 434–437. (on order)
10 young (5½ mo.) and 10 old (9 mo.) rats trained to lever press for food on a VI schedule were trained to suppress to a tone of either 3,500 cps or 670 cps. Generalization of the CER was measured in the presence of tones lower than 3,500 cps or higher than 670 cps. The results indicated that the CER will generalize along an auditory continuum without prior discrimination training: amount of suppression was inversely related to the frequency difference between CS and test stimulus. Older rats showed flatter gradients than younger rats, after both groups had been trained to suppress to a 3,500-cps. tone, but the groups did not differ following training with a 670-cps. tone.
Poppen, R. (1970). Counterconditioning of Conditioned Suppression in Rats. Psychological Reports, 27(2), 659–671.
In the Poppen article, what they call “toleration” corresponds to what we would call desensitization. The rats were subjected to conditioned suppression with a 3700 Hz tone. Then they were exposed to a much lower tone (400 Hz) that was gradually raised in 5 increments back to 3700 Hz. The change to the next highest frequency was done when the suppression was 90% extinguished. (We don’t have a comparable criterion.)
What they call “interference” corresponds to what we would call counterconditioning. The rats that “unlearned” the conditioned suppression the fastest were the ones who got simultaneous toleration and interference protocols, i.e., DS/CC.
Here’s my take. I think in the article, a “trial” is one exposure, not a group of exposures. Table 1 says that Group D (the DS/CC group) ranged from 22 to 67 trials to reach criteria (which I think was 90% reduction of the suppression). The mean was 44 trials. These were spread over 6 frequencies. Figure 3 showed that the mean ranged from 5–9 trials at each frequency. This could give a ballpark figure for the number of exposure of each sound in the series. (The first series will not by frequency, but by unmasking. Subsequent series to get us down to truck frequency will be by frequency.) Figure 3 also indicates that it takes more trials as you get closer to the original sound.
It’s cool that one of the articles in the literature that shows the success of DS/CC used the frequency approach for desensitization.
The number of exposures is not my role here, but I wanted to show you this article in case you want to consult it.
Onine tone generator, in case you want to listen to the frequency series used for the Poppen rat study. Just don’t do it around Fitz or any sound sensitive dog, even with earbuds. That’s a much wider range of sound than we will be covering.
Comments on Sound Series 1: Echo Idling
All sounds have 60 seconds of lead time and a 0.5-second fade-in. They are approximately 8 seconds long. They are in order of neutral to previously bothersome/scary. The descending cutoff frequencies represent more and more of the low frequencies included. All filters have a rolloff of 12 dB per octave (steep).
I made more splits near the end since that’s when the lowest frequencies are rolling in the fastest.
Keep in mind that we are not varying on frequency (as in the rat studies and the work I did with my own dog); we are gradually allowing in more of the low frequencies that exist in a complex sound. After you complete this, we will probably use frequency more to work into the truck range.
Proposed Plan for Echo Sounds
Per discussion between J. and Eileen, the changes between the first few sounds are very small. I.e., I may have done too many splits. One way to handle this would be to skip every other sound for the early part of the series. That would mean using sounds 1, 3, 5, 7, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19. I will leave the sounds available in case they are needed. I will mark the “skippable” ones as such in the titles.
I’m working on a chart that will show how the sounds are changing throughout the series where the changes will be easier to see than in the images I’ve provided. Then J. and N. can take a look. before making a decision.
Sound Series 1: Echo Idling
1. Echo idling HPF 1600 Hz cutoff
2. Echo 1400 Hz cutoff CAN SKIP
3. Echo 1200 Hz cutoff
4. Echo 1100 Hz cutoff CAN SKIP
5. Echo 1000 Hz cutoff
6. Echo 900 Hz cutoff CAN SKIP
7. Echo 800 Hz cutoff
8. Echo 700 Hz cutoff CAN SKIP
9. Echo 600 Hz cutoff
10. Echo 500 Hz cutoff
11. Echo 400 Hz cutoff
12. Echo 300 Hz cutoff
13. Echo 250 Hz cutoff
14. Echo 200 Hz cutoff
15. Echo 150 Hz cutoff
16. Echo 100 Hz cutoff
17. Echo 75 Hz cutoff
18. Echo unfiltered (GOAL!)
Progress Checklist for Echo Sounds
The Engine Hum of the Echo
Here’s an image of the frequencies of the Echo idling, unfiltered. I’ve circled and labeled the area that I filtered out to make it less scary.
This is interesting and heartening for Fitz. All of those peaks are overtones of the 45.5 Hz fundamental that the motor generates. At first it is surprising that the fundamental frequency of a sedan would be such a low frequency. But the fundamental relates to the speed of rotation of the engine, so the fundamental for a truck wouldn’t necessarily be lower than that of a car. Trucks can rumble for other reasons, though. (Can’t wait to analyze the slow rev!)
That fundamental shows low amplitude in part because of the limitation of my speaker, but if I do a pure FFT of the digital file, it does come out lower. In any case, it **should** be below Fitz’ hearing range.
On the next page I provide images of some of the different splits in the above sounds in case you want to skip some. They are very similar for the first few, then do start to change a little more rapidly toward the end.
Not for Fitz! For Humans Only
This is a recording of 3 seconds of each sound chained together from most filtered to the original unfiltered sound, in case you are interested. You won’t hear much change until about 15-24 seconds in. You’ll hear some repetitive higher noises, probably bird song. They sound rhythmic because I used the same chunk of the file for every iteration. I cut that section out of the sounds in the series.
Next Steps After Echo
These are the things we know he doesn’t like:
- Low-frequency engine noises (not terrible with Echo, but not great; ready to run away at idling sound of CRV)
- Any revving (bad) Diesel rattle even with engine noise removed (he handled it with cookies)
We discussed in Messenger that after he is good with the Echo, we would go to steady-state truck engine, then tackle the revving/acceleration/deceleration. It turns out that this happens way more often than I thought. (Parking, stopping, turning.)
Truck Sounds for J. to Check Out—Not for Fitz To Hear
I would choose one of the first three. Our goal is to graduate to a truck but not super scary, so no gear changes, probably no acceleration, certainly no revving. But also want to make it different enough from the final Echo recording that we are progressing. See what you think; maybe consult C.
After you let me know your thoughts and we agree on one, I will take a chunk out of it and do the sound series.
Truck Sample Candidate (Still on Headphones!)
We agreed to try truck sound #1 with the truck driving at a steady state. Here is a sample of that.
And following is what it sounds like filtered. It is odd, because there were some road sounds and wind. I could smooth it, but that would be more steps. If J. thinks he will be OK with this, she can try (with cheese).
Truck Test Sound (with 60 seconds leading silence)
Fitz’ response, 5/14/20: “He alerted to it, thought about it (he ALWAYS thinks about it), decided it was dangerous, slunk a few steps away, I tossed cheese, he stayed to think some more, I turned it off, he ate cheese.”
J. and I have agreed this is not a good reaction. We will try another “truck” moving recording that I will find, I’ll try to get one with less road noise/rattling, since I can’t edit that out at the same time as editing out low frequencies. If that doesn’t work, we’ll go to an idling recording.
J reports that Fitz has had a setback “where he became super anxious again. Modified his meds which helped but he’s not quite back to where he was.”
Truck Test Sound #2 (with 60 seconds leading silence)
This is another moving truck sound with less road noise (Pond5 8896371). I compared to the Echo sound Fitz is currently listening to (#15) and set a 400 Hz cutoff frequency on the new truck sound for the test. J. can do an exposure when she feels ready.
Response to Truck Test Sound #2
J. says: “He didn’t want to eat right away, wanted to orient to it. But he didn’t leave the room, in fact he went up to the speaker to check it out. So, definitely scarier than where we are with the Echo (#16) but not as scary as the last truck noise. When the noise stopped he immediately ate the cheese that I had scattered all around him.”
We agreed I’d create a more filtered version of this sound.
Truck Test Sound #3 (with 60 seconds leading silence)
This comes from the same source sound as #2 (Moving truck Pond5 8896371). I have set a cutoff frequency at 1000 Hz, up from 400 Hz.
If Fitz isn’t moderately comfortable with this version, I may take a different tack. One possibility would be to play the Echo and some version of the new truck sound at the same time, but the latter at a very, very low volume. Gradually raise the volume so they are equal, then gradually fade out the Echo. Those are long series, though. I made something similar the other day and it was 25 sounds. Plus we still might have some steps to go back to the unfiltered truck, if I filtered it to begin with. The other possibility, which I will probably experiment with first, is to just change the pitch of the whole sound. One doesn’t usually do that with broadband sounds, but there’s no law against it.
But here’s crossing my fingers that he’s fine with this one.
Outcome of Test Sound #3 (September 2020)
J. reports that Fitz did fine. “Lots of ear wiggling at it but he was otherwise unfazed.”
Here is sound series #2. The first sound is the same as the test sound #3 above.