Two things about this breed are worth more than everything else written about it, and you can act on both before a puppy comes home. One is a $70 cheek swab for a gene called MDR1. The other is a question about the colour of both parents. Neither shows up in most breed descriptions, so we have put them ahead of the temperament paragraphs.
One note on the name first. The breed was not developed in Australia. Its US association says the dog we know today “developed exclusively in the United States”, named for Basque shepherds who arrived by way of Australia in the 1800s. The AKC recognised it on 1 September 1991 and moved it into the Herding Group on 1 January 1993.
Where the measured dogs and the breed club disagree
The panel above comes from weighed adult dogs, so it describes what walks into a vet’s waiting room.
The written standard is no help on weight, because it does not mention weight at all. It gives a preferred height of 20 to 23 inches for males and 18 to 21 for females, then adds that “Quality is not to be sacrificed in favor of size.”
The breed’s own association fills the gap, and its numbers run heavier than the scales do. USASA publishes 50 to 65 lb for males and 40 to 55 lb for females. Compare that with the panel. Club figures describe the dogs people show and breed from. Population data catches every dog, including the smaller ones.
That gap matters for one practical reason. A typical male sits only just inside our large size class, and a typical female sits below it. So this is a breed where the sexes can land in different size brackets, and the male-female difference is a real planning question rather than a footnote.
It matters again at feeding time. Large-breed puppy food in the US is defined around dogs reaching 70 lb or more as adults, and most Australian Shepherds finish well under that line. What the label actually promises is in large-breed puppy food and calcium.
Mini and toy Aussies
You will see both sold. Only one of them is a breed.
The AKC registers the Miniature American Shepherd, which finished 28th in 2025 registrations against 14th for the Australian Shepherd. It has its own standard and its own club, which became the AKC parent club in May 2011. The breed itself was recognised in 2015. It began in 1962, when a breeder started working with unusually small Australian Shepherds. It was called a Miniature Australian Shepherd for years, and the name was voted away.
So a “miniature Australian Shepherd” or “toy Australian Shepherd” is a sales description. It may be a Miniature American Shepherd under an older name, a small Australian Shepherd, or a cross. Our own dataset holds the two breeds as separate entries with different adult sizes, and nothing between them. If a seller cannot say which registry the parents sit in, our guide to predicting a mixed-breed puppy’s size is the more honest tool.
The gene test to run before your vet needs it
This is the strongest single item on this page.
MDR1, also written ABCB1, carries the instructions for a pump that sits in the barrier between blood and brain. The pump pushes certain drugs back out. A small deletion in the gene breaks it, and those drugs then reach the brain at levels they were never meant to.
The numbers, kept separate
Washington State University genotyped 1,421 Australian Shepherds between 2004 and 2007. The split:
| Result | Dogs | Share of the 1,421 |
|---|---|---|
| Two working copies | 754 | 53% |
| One broken copy | 525 | 37% |
| Two broken copies | 142 | 10% |
Add the last two rows and 47% of those dogs carried at least one broken copy. Count gene copies instead of dogs (809 broken ones out of 2,842) and you get an allele frequency of about 28%.
Both are correct, and they answer different questions. WSU’s own breed page rounds the first to about 50% for this breed, and it gets copied around as though it were the second. Whenever you meet an MDR1 percentage, check whether it counts dogs or gene copies.
One more caution about that sample. These were dogs whose owners paid to have them tested, mostly breeders. A separate study genotyped 286 dogs walking into a teaching hospital and found the broken copy in one of them. In the same paper’s testing-lab group, 120 Australian Shepherds gave an allele frequency of 32%. Same gene, very different rooms.
One copy is not a bystander
With many inherited diseases, a dog with one copy is a carrier, able to pass it on, unaffected itself. MDR1 does not work like that. WSU states that higher ivermectin doses “can cause toxicity in dogs that are heterozygous for the mutation (MDR1 mutant/normal)”, and that chemotherapy drugs and some tranquillisers need reduced doses in one-copy dogs too. A dog with two copies is more sensitive again.
So there are three results, not two. Your vet needs to know which one you have.
What the list actually says
Here the popular version of this story is wrong in a way that matters, and being wrong makes owners refuse useful medicines. WSU’s current problem-drug list opens with “Dogs with the MDR1 genetic mutation can become very ill or die if given certain medications,” then sorts them.
- Avoid outright: loperamide, the drug in Imodium. At the doses used for diarrhoea, “this drug will cause neurological toxicity in dogs with the MDR1 mutation.”
- Ask before dosing: eight entries. WSU prints no dose for any of them and tells you to ask its pharmacologists instead. Acepromazine and butorphanol handle sedation and pain around surgery. Maropitant, sold as Cerenia, is the anti-sickness jab or tablet a vet reaches for when a dog is vomiting or travels badly. Of the eight it is far and away the one you are likeliest to be handed, so learn the name now. Then apomorphine, ondansetron, cyclosporine, grapiprant, and chemotherapy drugs including vincristine and doxorubicin.
- Cleared at label doses: the monthly heartworm and parasite products. The FDA determined that ivermectin as sold in Heartgard Plus is safe for MDR1-affected dogs at label doses, and milbemycin, moxidectin and selamectin carry the same finding. Several flea and tick products were tested by their makers and showed no problem.
Read that last group twice. “Aussies cannot have heartworm prevention” is false, and skipping prevention is its own risk.
Now read the limit on it, because the harm tracks the dose and not the bottle. WSU puts numbers on that. Heartworm prevention delivers ivermectin at 6 micrograms per kilogram. Treating mange takes 300 to 600, fifty to a hundred times as much. At the mange dose the drug “will cause neurological toxicity” in a two-copy dog and can do it in a one-copy dog. Livestock ivermectin is one route to that dose. A prescription for demodectic mange is another, and it arrives in a small-animal bottle from your own vet. Selamectin, milbemycin and moxidectin obey the same rule: cleared at the monthly preventive dose, dangerous well above it.
So the cleared sentence is narrow, and worth reading exactly. Monthly prevention at label dose is fine. For any other ivermectin product or dose, WSU’s list stops and refers you back to its own pharmacologists. Say the word “MDR1” before your dog is given any of it.
Getting it done
WSU’s laboratory runs the test for $70, from either a cheek swab or a blood sample. Budget more than a week for it. The lab posts results on a Wednesday afternoon for samples that reached the bench the Monday before, and that clock starts when your swab arrives, not when you post it.
Now the part buyers do not expect. The AKC publishes what each parent club asks breeders to screen, and for this breed that is hips, elbows and eyes. MDR1 is not on the list. The club’s own scheme does carry it, but filed as optional next to thyroid and Collie eye anomaly, while hips, elbows and eyes are the three a certificate turns on. Look up the Miniature American Shepherd on that same AKC page and MDR1 is sitting there.
That is not a scandal, and a certificate still tells you plenty. It does mean a fully certified Australian Shepherd may never have been MDR1 tested. Ask for the result by name.
Merle, and the mating to ask about
Merle is the mottled coat pattern the breed is famous for. The standard lists blue merle and red merle alongside solid black and red, “with no order of preference”, and allows blue, amber or brown eyes in any combination.
One merle copy gives you the pattern. Two copies (which is what breeding merle to merle risks) is where the measured harm sits.
Researchers ran hearing tests on 153 merle dogs across several breeds and split them by genotype:
| Group | Dogs | Deaf in one ear | Deaf in both |
|---|---|---|---|
| One merle copy | 113 | 3 (2.7%) | 1 (0.9%) |
| Two merle copies | 40 | 4 (10.0%) | 6 (15.0%) |
So 4 of 113 single-merle dogs had a hearing loss, against 10 of 40 with two copies. The authors found the link between genotype and hearing statistically solid, with a Fisher’s exact test P value below .0001.
Australian Shepherds were the second-largest group in that study, at 32 dogs. The breed-level numbers are far too small to quote as a rate: 27 single merles, one of them deaf in one ear, and 5 double merles, two deaf in both. Use the pooled figures instead.
Eyes go with ears. A 1981 study of merle Australian Shepherds found microphthalmia with coloboma, small eyes with gaps in the structures inside them, behaving “as an incompletely penetrant recessive trait”, and turning up more in merles carrying a lot of white than in merles carrying little.
Why the parents’ coats are not the whole answer
Ask about the parents’ colours. Then know the limit of the answer.
The merle gene is switched on by a piece of inserted DNA, and its length varies. A team measuring it in 181 dogs across 14 breeds, 40 of them Australian Shepherds, sorted the versions from short to long. The shortest are cryptic merle: present in the dog, invisible in the coat. A dog that looks solid can carry one and pass it on.
The same team reported that a single copy of the longest version seemed to predispose a dog to hearing and sight problems too. They argued for testing every dog in a merle breed, so pairings get chosen on genotype rather than on what the coats look like.
Practical version: two visibly merle parents is a clear flag. Two solid parents is reassuring but not proof, and a length-based merle test closes the gap.
What one is actually like to live with
Start with the finding that limits every sentence written about breed character.
A 2022 study surveyed 18,385 dogs and genotyped 2,155. Its headline is that breed accounts for roughly a tenth of the behavioural differences between individual dogs. The authors put it at “just 9%”. Nine parts in ten of what makes your dog your dog sit outside the breed label.
But this is a breed where the exception is interesting.
Biddability, and the number that gets misquoted
Biddability means how readily a dog takes direction. It is one of the traits where breeding has left a mark, and the same study put heritability, the share of the differences between dogs traceable to genes, at 30.5% ± 8.5.
Then the authors corrected for breed ancestry, and it fell to 20.0% ± 8.8. That was the largest drop of any trait they measured.
Quoting only the first figure inverts the point. Part of what looks like a genetic effect is really breed membership doing the work. What survives the correction is still real: they write that for traits like biddability, “knowing breed ancestry can make behavioral predictions somewhat more accurate.” Somewhat is the operative word. Even after the correction, four fifths of the difference between dogs traces to something other than genes.
The survey result people will not expect
A survey of 33 breeds using the C-BARQ owner questionnaire scored 177 Australian Shepherds. The breed cleared the study’s threshold of 45 dogs, and sits in its online sample rather than its breed-club sample. The table below counts dogs hitting the top of the scale, a score of 4, on at least one question in a category. A 4 means snapping, biting or trying to bite.
| Directed at | Australian Shepherds | Share |
|---|---|---|
| Unfamiliar people | 11 of 177 | 6.2% |
| Their own household | 1 of 177 | 0.6% |
| Other dogs | 26 of 177 | 14.7% |
| Another dog at home | 10 of 155 | 6.5% |
The stranger figure is the surprise. In the same table German Shepherds came in at 4.5% and Rottweilers at 4.8%, both lower than this breed.
Three things to hold alongside it. This was owner report, not observation. The online sample recruited itself, and the same questionnaire gave different answers when the paper reached owners through breed clubs instead. How you find people moves the result. And in every row above, at least 85% of the 177 dogs scored below the top of the scale.
It fits the breed standard better than it fits the marketing. The standard asks for a dog who “may be somewhat reserved in initial meetings” and tells judges that any “shyness, fear or aggression is to be severely penalized.” Reserve with strangers is written into the breed. A dog bred to keep watch over stock is not built to greet every visitor.
Growing up
The months when training stops working
Somewhere in the second half of the first year, plenty of owners find their puppy has stopped hearing them.
The best study of this ran two groups of guide dogs. Carers rated 285 of them for trainability at 5, 8 and 12 months and marked them down hardest at 8. A separate 93 sat the hands-on version, a sit command at 5 and 8 months only: slower to obey at 8, and only when their own carer asked. A stranger got a better response, not a worse one. Note the samples: golden retrievers and Labradors in the sit test, German Shepherds added in the rated group. No Australian Shepherds in either. We are borrowing the finding, and saying so.
A larger UK cohort supports the timing without naming breeds either. Owner reports of unwanted behaviour peaked at 12 months, where 513 of 1,219 owners (42.1%) had something to report, then fell back. Barking, jumping up, lead-pulling and ignoring recall led the list.
Joints, and an honest borrowing
Keep a growing Australian Shepherd lean. The evidence for that is strong, and it was collected in a heavier breed.
Forty-eight Labrador puppies were split into pairs by litter, one of each pair fed 25% less than the other from eight weeks. By two years, hip dysplasia had been diagnosed in 7 of the 24 fed less and 16 of the 24 fed freely. Following the same dogs for life, hip osteoarthritis appeared in half the restricted group against 83% of the others, and arrived at a median of 12 years rather than 6.
Two caveats we will not skip. The comparison group ate as much as it liked, so this is about the cost of overfeeding, not a case for underfeeding a lean dog. And an Australian Shepherd is much lighter than a Labrador, so the size of the effect here is unknown. More on that in rapid growth and joint problems.
Neutering looks different here
For the big working breeds, delaying neutering is usually argued on joints. In this breed that argument mostly is not there.
One US veterinary hospital’s records cover 440 Australian Shepherds: 93 intact males, 135 neutered males, 76 intact females and 136 spayed females. Joint disorders turned up in 3% of intact males and 4% of intact females, with no clear rise after neutering at any age. Cancers ran 9% in intact males against 1% in intact females. Mammary cancer appeared in none of the intact females and in 8% of those spayed between two and eight years old.
The authors’ guidance for this breed reads: “Lacking a noticeable occurrence of increased joint disorders or cancers in neutered males, those wishing to neuter should decide on the appropriate age.” Take the figures to your own vet; the sample is one hospital’s caseload, not the country’s.
Over a lifetime
What follows is a list to raise at a check-up, not a list to act on at home.
Seizures are the condition with real breed-specific research behind them. German, Austrian and Swiss neurologists followed 50 Australian Shepherds with idiopathic epilepsy (seizures with no other cause found) against 50 unaffected ones. The first seizure came at a median age of 2.5 years, with a range of 0.3 to 5.0. Poor seizure control “is evident in 56% of epileptic ASs”, and poor control went with shorter survival. Two details are worth carrying to a vet. Onset was earlier in non-merle dogs, at a median 1.8 years against 2.8 years in merles. And control was unrelated to MDR1 genotype, which retires a theory that circulates in the breed. These were recruited cases, so the study says nothing about how common epilepsy is here.
Inherited variants are common enough that testing is routine. A Slovak laboratory reported results from 2,595 Australian Shepherds sent in by European owners between 2012 and 2022:
| Variant | Dogs tested | One copy | Two copies |
|---|---|---|---|
| Hereditary cataract (HSF4) | 1,641 | 382 (23.3%) | 0 |
| Collie eye anomaly (NHEJ1) | 1,503 | 284 (18.9%) | 4 (0.3%) |
| Progressive retinal atrophy (prcd) | 1,454 | 40 (2.7%) | 3 (0.2%) |
| Degenerative myelopathy (SOD1) | 722 | 160 (22.2%) | 5 (0.7%) |
These were paying customers, mostly breeding stock, so read them as a picture of tested dogs rather than of the breed as a whole. The first three are eye conditions. Degenerative myelopathy is the odd one out (a slow wasting of the spinal cord), and two copies marks a dog at risk rather than a dog with the disease.
The cataract variant deserves a note, because “one copy” means something unusual here. It is dominant with incomplete penetrance: one copy is enough to raise risk, and not every dog with it goes on to develop cataracts. UC Davis puts dogs carrying the variant at 17 times the odds of developing cataracts in both eyes. Higher odds is not the same as high likelihood, and a gene result is not a diagnosis. Only an eye exam is that.
Hips and elbows are screened, and the published rates cannot be trusted. The parent club requires both X-rays for a health certificate, which is the useful part. The rate is the useless part. Where sending films in is voluntary, owners of dogs with bad hips often do not bother. Somebody measured that: researchers followed X-rays taken on 200 Golden Retrievers and 140 Rottweilers and tracked which reached the registry. Normal hips were 8.2 times as likely to be sent in. Nobody has repeated that in this breed, so we can tell you the direction of the error but not its size.
What we could not verify
A lifespan figure for the breed. The largest UK dataset, 584,734 dogs, gives 12.5 years across all dogs, 12.7 for purebreds and 12.8 for purebreds with ordinary head proportions. The per-breed rows sit in a supplementary table we could not open, and this breed is not named in the main paper. The breed’s own institute reports about 11 years from a 2017 survey, with no sample size published. The 12 to 15 years quoted everywhere else traces to nothing we could find.
Exercise in minutes. Not in the standard, not in any paper we opened. The breed association says Aussies “need a great deal of exercise on a daily basis” and a job to do, which is informed opinion from people who know the dogs. It is not a measurement, and nobody has tested what happens without it.
A hip or elbow dysplasia rate. See above. Better to hand you the screening question than a number we cannot defend.
A breed puppy weight chart. No published growth curve exists for Australian Shepherds. The table on this page comes from our own model using this breed’s measured adult weights, which makes it checkable, not a record of what real puppies weighed.
Everyday illness rates. The UK primary-care research programme that has published breed papers on Labradors, German Shepherds and others has not done one here. So we have genetics and screening, but not the “what does the vet actually write down” picture those papers give.