Using Sound Apps for Desensitization & Counterconditioning for Dogs

“What’s that noise and where’s it coming from?” Dogs’ hearing abilities are different from ours—a fact that is frequently and strangely unconsidered in the development of many audio products for dogs.

Dog trainers often recommend smartphone apps and YouTube videos for desensitizing and counterconditioning dogs who are afraid of specific noises. There are many apps designed for this, and they typically have recordings of a variety of sounds. However, the physics of sound production and the limitations of consumer audio present large problems for such use, problems substantial enough to prevent the success of many (most?) conditioning attempts.

This post describes the pitfalls of using sound apps and other recordings for this purpose. It maps out a strategy for sound desensitization that avoids most of the problems.

Why Many Audio Conditioning Products Fail

If quizzed, most people would likely guess that dogs have hearing abilities that are vastly superior to ours. In fact, it’s a mixed bag.

Humans can hear slightly lower frequencies than dogs can (Gelfand, 2010, p. 166). We can also locate sounds quite a bit better than they can (Mills, 1958), (Fay and Wilber, 1989, p. 519). But dogs are the big winners in the high frequency range—they can hear tones over about twice the frequency range that humans can (Gelfand, 2010, p. 166), (Heffner, 1983). Also, dogs can also hear sounds at a much lower volume level than humans can over most of our common audible range (Lipman & Grassi, 1942).

Yet the superior aspects of dogs’ hearing are rarely considered when we decide to use sound recordings in conditioning.

There are four major acoustical problems with using human sound devices to condition dogs.

  • The inability of smartphones to generate low frequencies, such as those present in thunder
  • The limited ability of even the best home audio systems to generate these low frequencies in high fidelity
  • The upper limit of the frequencies generated on all consumer audio
  • The effects of audio file compression on the fidelity of digital sound

There are other problems when using apps that can make or break attempts to positively condition a dog to sound.

  • Lack of functional assessment before attempting conditioning
  • The length of the sound samples used for conditioning
  • The assumption that lower volume always creates a lower intensity (less scary) stimulus

Some, but not all of the above problems can be addressed with do-it-yourself work and a good plan. But sound conditioning of dogs using recordings will always have some substantial limitations that can affect success.

What Is the Frequency?

Frequency is the aspect of sound that relates to the cycles of the sound waves per second. Cycles per second is expressed in units of hertz (Hz). I’ll refer a lot to low and high frequencies because they pose different challenges.

To help with the concept of frequency, think of a piano keyboard with the low notes on the left and the high notes on the right. The low notes have lower frequencies and the high notes have higher frequencies. Keep in mind that sound frequency goes much higher than the highest notes on a piano!

Common sounds with low frequencies include thunder, large fireworks demonstrations, industrial equipment, the crashing of ocean waves, the rumble of trains and aircraft, and large explosions. Common sounds with high frequencies include most birdsong, the squeaking of hinges, Dremels and other high-speed drills, referees’ whistles, and most digital beeps.

Motorized machinery generates sound frequencies that correspond to the rotation of the motor. These frequencies can be high like the dentist’s drill or low like aircraft. Motors can also vary in speed. For instance, when you hear a motorcycle accelerating, the frequency of the sound rises as the engine speeds up.

Humans can hear in a range of 20–20,000 Hz (Gelfand, 2010, p. 166), and dogs can hear in a range of 67–45,000 Hz (Heffner, 1983).

Some sounds don’t have a detectable pitch, meaning they include such a large number of frequencies that you can’t pick anything out and hum it. These are called broadband sounds. A clicker generates a broadband sound.

Missing Low Frequencies

…the better to hear you with!

Our human brains are great at filling in blanks in information and taking shortcuts. This makes it hard for us to realize what a bad job our handheld devices do in generating low-frequency sounds. Our dogs undoubtedly know, though.

Many people purchase sound apps in order to try to condition their dogs to thunder. The frequency range for rumbles of thunder is 5–220 Hz (Holmes, 1971). Handheld devices can’t properly generate those low frequencies. They generally have a functional lower output limit of about 400–500 Hertz. If you play a recording of thunder (or a jet engine, or ocean waves) on a handheld, the most significant part of the sound will be played at a vanishingly low volume or be entirely missing.

When performing desensitization, we aim to start with a version of the sound that doesn’t scare the dog, so using a handheld could possibly be a starting point. But you would have to fill in the missing low frequency sounds gradually as part of the desensitization process, which would mean using a different device after the first couple exposures anyway.

Home sound systems, including some Bluetooth speakers, can do a better job. They usually generate frequencies down to 60 Hz. This is roughly the lower limit of dogs’ hearing, so it’s a good match. But even the best home system can’t approach the power and volume of actual thunder, and the sound is located inside your home instead of outside. Some dogs do not appear to connect recordings of thunder on even excellent sound equipment to the real thing, or they will respond to recordings with a lesser reaction (Dreschel & Granger, 2005).

In one study of thunder phobic dogs, the researchers brought their own professional quality sound system to each dog owner’s home; great mention is made that the sound system was large (Dreschel & Granger, 2005). This bulk indicates that they were serious about being able to generate low frequencies. In general, the larger the speakers, the better they are at generating low frequencies. The difference today is smaller than it was 15 years ago, however. Sound systems have improved a lot in recent years.

Some of the sound apps for dog training now instruct you to send the sound to a home sound system rather than using the speaker in the handheld. This is excellent advice for any sound. But the bottom line is that you will not always be able to emulate low frequencies well enough to function as desensitization for some dogs.

This image compares the magnitude of a recording of a roll of thunder played on an iPhone 7 vs. a home sound system (Altec Lansing speakers). The Blue Yeti microphone I used to capture the sound for analysis was the same distance from the speaker in each case.

This plot shows that the smartphone speakers don’t generate low frequency sounds effectively

The part of the plot in the oval is the approximate range of the rumbles of thunder. The navy blue line represents the sound generated by the smartphone in those frequencies. The red line was from the Altec Lansing home speakers. The speakers generate sound down to 60 Hz effectively (as per their specifications).

In contrast, the output of the phone is virtually inaudible below 300 Hz. Sounds below about 44 Hz in that frequency range are generally indistinguishable from indoor ambient sound.

Missing High Frequencies

All consumer audio equipment is designed for human ears. Our handhelds, computers, TVs, and sound systems put out sound only up to the frequency of 22,000 Hz. Humans can’t hear higher frequencies than that. But dogs can hear up to about 40,000 Hz. So again, the recordings are not high fidelity for dogs.

This is different from the thunder situation. The low frequencies of thunder are present in high quality recordings, but our equipment can’t perfectly generate them. With high frequencies, it’s not only a limitation of our speakers. The sounds in “dog frequencies” are not recorded in the first place.

It’s not that it can’t be done. Biologists and other scientists use special equipment that can record or play back sound in the ultrasound range. The recording device requires a higher sample rate (how often the sound is digitally measured) than consumer equipment and the speaker for playback requires a wider bandwidth for frequency response. 

How much do the missing high frequencies affect the fidelity of recorded sound for dogs? We can’t know for sure. But virtually all sounds include what are called harmonics or overtones. These are multiples of the original frequency into a higher range. Dogs can hear these in the range from 22,000 to 40,000 Hz, but they are never present in sound recordings made even by very high quality equipment.

Because of these missing frequencies, dogs with normal hearing will likely be able to discriminate between a natural sound and even the best recording of it. 

Other Sounds Missing Due To Compression

Digital audio files are large. Most files that are created to play on digital devices are saved in MP3 format. This format was created in the 1990s when digital storage was much more limited than it is today. Hence, MP3 files are compressed, meaning that some of the sound information is removed so they won’t be so large.

MP3 is termed a “lossy” compression because sound data is permanently lost through the compression. The compression algorithms are based on the capabilities of the human ear. Sounds we humans are unlikely to be able to hear are removed.

Some of these limitations may be shared by dogs. For instance, quieter sounds that are very close in time to a loud sudden sound are removed. We can’t hear those because of masking effects, and it’s probable that dogs can’t either, although there may be a difference in degree.

However, there are other limitations of the human ear that dogs do not share. For instance, our hearing is most sensitive in the range of about 2,000 to 5,000 Hz. So very quiet sounds that are pretty far outside that range will likely be eliminated by the compression algorithm. Dogs’ most sensitive range is higher than ours, so sounds they could hear are probably omitted from compressed recordings.

Keep in mind that dogs not only hear sounds that are higher than we can perceive, but they hear all high-pitched sounds at lower volumes than we do.

So the MP3 compression process is another reason that some sounds in dogs’ hearing range that would be present in a natural sound would be missing in a recording of it.

If you make your own recordings, there is an easy thing to do to prevent this problem. You can save your sound files in WAV or AIFF formats as discussed below. I haven’t seen a desensitization app that uses these formats, however.

Behavior Science Considerations

The problems I’ve discussed so far are caused by physics of sound and how it is recorded, compressed, and played.

The following cautions have to do with applying what we know about performing classical conditioning to sound without errors.

Lack of Functional Assessment

Trainers and behavior consultants who help dogs with behavior problems perform functional assessments. They observe and take data to help them understand what is driving the problem behavior. In the case of fear, they analyze the situation in order to determine the root cause of the fear.

In the case of sound sensitivity, a dog may react because the sound has become a predictor of a fear exciting stimulus, as is the case with much doorbell reactivity. Or the dog may be responding to an intrinsic quality of the sound, as in the case of sound phobia. These are different fears that require different approaches. Sound phobia is a clinical condition that necessitates intervention. Many such dogs need medication in order to improve.

Trainers, working with veterinarians or veterinary behaviorists, can make these determinations. Consumers often can’t. And as the sound apps being marketed to consumers become more elaborate, pet owners who follow the directions have a good chance of worsening some dogs’ fears.

Jo is not impressed

For example, a newer sound app allows you to set up the app to play the sound randomly when you are not home for purposes of desensitization (without counterconditioning, although a mechanism for it is planned for the future). The instructions show an example of a dog’s doorbell reactivity going away through use of the app (although I’m skeptical that it could work long term). The app was programmed to play doorbell sounds randomly when the owner wasn’t home. This decoupled the doorbell as a predictor of strangers at the door.

This protocol would give any professional trainer pause. First, the cause of the reactivity, the dog’s fear of strangers, isn’t addressed at all. All things considered, that is not a humane or robust approach. The dog’s fear is left intact while the inconvenience of their barking at predictors is removed. Second, for a dog with a true sound phobia, playing a feared sound repeatedly when the owner isn’t home, even at a low volume (more on this below), could have ruinous results.

Apps that can play randomized, graduated sound exposures can be a good tool for trainers, as long as the trainers are aware of the physical limitations outlined in this post. They should not be marketed or recommended to consumers. Following the directions that come with the app could actually make a dog worse.

Length of the Sound Stimulus

Many noises in the apps are too long for effective desensitization and counterconditioning. Real-life thunder and fireworks both have an infinite array of sound variations. If you play a 20-second clip of either of these, there will be multiple sounds present and a sound phobic dog may react several times, not just once.

Classical delay conditioning, where the stimulus to be conditioned is present for several seconds, and the appetitive stimulus (usually food) is continually presented during that time, is said to be the most effective form of classical conditioning.

Delay conditioning would be appropriate to use for a continuous, homogeneous sound, such as a steady state (non-accelerating) motor. But fireworks and thunder are not continuous. They are sudden and chaotic. They consist of multiple stimuli that can be extremely varied.

To offer a visual analogy: if your dog reacts to other dogs and you seek to classically condition him, you might create a careful setup wherein another dog walks by at a non-scary distance and is in view for a period of, perhaps, 10–20 seconds. You would feed your dog constantly through that period. That is a duration exposure to one stimulus. (And you would try to use a calm decoy dog who doesn’t perform a whole lot of jumpy or loud behaviors!)

But for the first time out you would not take your dog to a dog show or a pet parade or an agility trial to watch 60 different dogs of all sizes and shapes coming and going and performing all sorts of different behaviors, even if you could get the distance right and the exposure was 10–20 seconds. That is the visual equivalent of the long sound clip of fireworks. There are far too many separate stimuli.

Also, if you play a longer clip, one lasting many minutes, as has been done in some sound studies, you are essentially performing simultaneous conditioning, a method known for its failure to create an association (Schwartz, 1989, p. 59). The fact that you started feeding one second after the sound started is not going to be significant if the thunder crashes and food keep coming for minutes on end. You have not created a predictor.

And if you are feeding the whole time but the scary sounds are intermittent, you are probably also performing reverse conditioning, where the food can come to predict the scary noise.

If you are working to habituate a non-fearful dog or a litter of puppies to certain noises, the longer sound clips are probably fine. They may even work for a dog with only mild fears of those noises. But the more fearful the dog is, and the closer the dog is to exhibiting a clinical noise phobia, the cleaner your training needs to be. To get the best conditioned response, you need a short, recognizable, brief stimulus. 

After you get a positive conditioned response to one firework noise, for instance, you can then start with a different firework noise. After you have done several, you may see generalization and you can use longer clips. But don’t start with the parade!

Assumptions about Volume

Most mammals have what is called an acoustic startle response. We experience fear and constrict certain muscles reflexively when we hear a loud, sudden noise. It’s natural for any dog to be startled by a sudden noise. It may be that dogs who have over-the-top responses to thunder and fireworks have startle responses so extreme as to become dysfunctional. For dogs who fall apart when they hear a sudden, loud sound such as thunder, it makes all the sense in the world to start conditioning at low volume, because this practice can remove the startle factor.

But it’s different for dogs who are scared of high-frequency beeps and whistles. These odd, specific fears are not necessarily related to a loud volume. I have observed that, with these dogs, starting at a quiet level can actually scare the dog more. Remember, dogs don’t locate sounds as well as humans do. It could be that the disembodied nature of some of these sounds is part of what causes fear. (Have you ever tried to locate which smoke alarm in a home is emitting the dreaded low battery chirp? Even for humans, it can be surprisingly difficult. And we are better at locating sounds.)

When lowering volume is ruled out as a method of providing a lower intensity version of a sound stimulus, virtually all apps for sound desensitization are rendered useless.


With apps that can do more and more for humans, it seems odd to suggest that in order to help your dog, you might have to invent your own helpful tools. But doing so can help you make recordings of better fidelity and more appropriate length, and if you or an acquaintance are at all tech savvy, you can also alter sounds in other ways besides volume.

  • Record sounds yourself using an application that can save the recordings in WAV or AIFF (uncompressed) formats. This eliminates one of the ways that recordings can sound different to dogs from real-life sounds. Newer smartphones are fine for this. Even though they can’t play back low-frequency sounds, they can record them.
  • Create short recordings of single sounds, especially for dogs with strong sound sensitivities. Or you can also purchase high-quality sounds. For instance, you could purchase a 20-second recording of a thunderstorm, and edit out one roll of thunder to use. But be sure that the file you purchase is uncompressed. I use the site Pond5, where I can buy appropriate, high-fidelity sounds for $3–7 apiece.
  • Play sounds for desensitization on the best sound system possible, especially if you are working with thunder, fireworks, or other sounds that include low frequencies. Be aware, though, that you can’t always successfully condition to those sounds.
  • For dogs who are afraid of high-pitched beeps, create a less scary version by changing the sound’s frequency or timbre rather than by lowering the volume. Generally, lowering the frequency works well. You will then need to create a set of sounds for graduated exposures. They should start at a non-scary frequency, then gradually work back up to the original sound. See the papers by Poppen and Desiderato listed in the references for validation of this approach.
Silly human says what?

There are several ways to change the frequency of a recorded sound. You can use video software that has good audio editing capabilities, the free computer application Audacity, or professional sound editing software. You can also generate beeps at different frequencies using a free function generator on the internet.

The one advantage of working with dogs who are afraid of such sounds is that the original sounds themselves are usually digitally generated, so when you create similar sounds the fidelity will be high. (In other words, when a dog is afraid of a smartphone noise, a smartphone is the perfect playback tool.)

This is not a project to be undertaken lightly, but it can be done if you have tech skills and a good ear. Be sure to use headphones and be at least one room away from your sound sensitive dog when you start working with recordings of beeps. My dog can hear high frequency beeps escaping from my earbuds from across a large room.

I create sound series for desensitization for dogs who are afraid of whistles, digital beeps, and some other sounds (not thunder or booming fireworks). You can check out the information on my Sound Sensitive Dogs site for more information on this service.

Sound Conditioning Is Not Always a Perfect Solution

With some dogs and some sounds, it will not be possible to play recordings that are similar enough to the natural sounds to be able to carry over a conditioned response. Thunder and fireworks will always present significant problems.

We want to believe there is always a training solution. But sometimes physics foils our plans and the gap between an artificially generated sound and the natural sound will be too high. In that case, masking, management, and medications will be the best help.


Desiderato, O. (1964). Generalization of conditioned suppression. Journal of Comparative and Physiological Psychology, 57(3), 434–437.

Dreschel, N. A., & Granger, D. A. (2005). Physiological and behavioral reactivity to stress in thunderstorm-phobic dogs and their caregivers. Applied Animal Behaviour Science, 95(3-4), 153-168.

Fay, R. R., & Wilber, L. A. (1989). Hearing in vertebrates: a psychophysics databook. Hill-Fay Associates.

Gelfand, S. (2010). Hearing: An introduction to psychological and physiological acoustics. Informa Healthcare.

Heffner, H. E. (1983). Hearing in large and small dogs: Absolute thresholds and size of the tympanic membrane. Behavioral Neuroscience97(2), 310.

Holmes, C. R., Brook, M., Krehbiel, P., & McCrory, R. (1971). On the power spectrum and mechanism of thunder. Journal of Geophysical Research, 76(9), 2106-2115.

Lipman, E. A., & Grassi, J. R. (1942). Comparative auditory sensitivity of man and dog. The American Journal of Psychology55(1), 84-89.

Mills, A. W. (1958). On the minimum audible angle. The Journal of the Acoustical Society of America30(4), 237-246.

Poppen, R. (1970). Counterconditioning of Conditioned Suppression in Rats. Psychological Reports, 27(2), 659–671.

Schwartz, B. (1989). Psychology of learning and behavior. WW Norton & Co.

Photo Credits

  • Jo the black pug copyright Blanche Axton
  • All other photos copyright Eileen Anderson

Thank you to Whole Dog Journal, who originally published this article in 2020.

Copyright 2020 Eileen Anderson

9 thoughts on “Using Sound Apps for Desensitization & Counterconditioning for Dogs

  1. Brilliant post. So much ‘dog stuff’ in consumer land is far from evidence based,and that is so frustrating. Thanks for your great work, it’s awesome.

  2. Very interesting, thanks! My dog reacts to all kinds of noises, like barking dogs, truck brakes squeaking, car horns and voices over the fence. I think it is because they happen so randomly during the day. I am never quick enough to countercondition the real thing, and they all happen outside of my house! So I dont think playing a recording at home will work, do you?

    1. IT would be super hard since they are random and unpredictable, not to mention coming from outside, so they won’t sound the same if you play them inside. My first impression would be that recordings would be tough to use in this situation. Have you talked to your vet about meds?

  3. Wow, Eileen, that was a lot of research and terrific explaining! What are your thoughts on typical advice to play baby sounds to prep dog for addition of a baby? I have always advised against this practice because, as you mention above, the duration of real baby crying is likely to be much longer than any expectant parent is willing to “practice” ahead of time but also because it seems counterproductive to introduce a disembodied distress sound that may actually sensitize the dog and then bring home a baby that sounds like that. At least with a real life baby, it is clear where the sound is coming from. What do you think?

    1. Hi Madeline! I’m sorry to be so behind on comments. Most of us know from experience that most dogs can tell the difference between a “canned” voice coming out of a speaker and the real thing. I do hear of the occasional dog who responds to their person’s voice over a speaker with apparent recognition, but it seems to be the exception. (None of this is data-driven, just speculation born of experience.) Also, on another level, they can tell the difference between individual human voices. So I’m not sure that a recording of a crying baby would “carry over.” If you think the distress carries over (I have no opinion/experience so I defer to you on this) I probably wouldn’t do it.

      But if I had a dog who was not sound phobic, I would be introducing them to the new sounds that are going to be happening, especially in the nursery, but with the dog in whatever part of the house they will be in. (I’m assuming you recommend against people having their dog in the nursery with a newborn!) So from their own hangout, the dog will hear cupboards being opened and closed, things being unwrapped, any mechanical or electronic gizmos being run. I would probably do this without counterconditioning; just add these noises to the dog’s environment because it is something you can do ahead of time, keep an eye on them to make sure they aren’t bothered, and allow them to habituate. Countercondition if they are bothered. Ahh, and the baby monitor! It doesn’t even have to be baby noises, but I’d be letting the dog experience that there is going to be a new electronic sound source in the house. So an adult can have it on in the nursery and “broadcast” some baby care (not baby) noises including adult speech.

      If such a thing could be had, I would consider a playlist of various infant noises if it didn’t center on crying. In the classification of “might help, can’t hurt” if you thought that could be the case. And again, as a habituation attempt, since it would be too long for CC. But I think I’d be concentrating more on the actual real-life sounds that will be new to the dog and that can be demoed ahead of time.

      Is this in keeping with your experience, practice, and philosophy? Because if I have said anything ridiculous please correct me, since I’m one of the least child-experienced people around!

      Thanks for a great question!


  4. Thank you so much for your series of articles on sound sensitivity – they have been really useful in working out the best ways to help my two dogs. Once I stopped looking for high tech solutions (a noise blocking crate; the perfect desensitising cd) and concentrated on the dogs and what I had to hand things went much better. Bang-Bang treats on walks have made a big difference to their reactions to the occasional sound of shooting (one happy result of Covid has been far fewer pheasant shoots round here), and Sophy now asks to have the small cloakroom set up with comfy bed and the fan running when it is thundery. Unfortunately she still wants my company, and the only place to sit is the loo seat, which gets very hard after a while. I must remember a cushion for human comfort, too…!

  5. Eileen – Thanks for putting together a Sound Sensitivity site, and sharing your acoustical engineering skills. It looks like a great resource. We may be calling on you – our problem being that we often don’t know what (presumably) auditory signals Rowan is responding to. Ah, the joys of an environmentally sensitive border collie…

    1. It’s really hard when you can’t tell what is bothering them! I’m sorry you are going through this. Drop me a line anytime.

Comments are closed.

Theme: Overlay by Kaira Extra Text
Cape Town, South Africa
%d bloggers like this: