Which Rabbit Traits Respond to Selection: Heritability for Breeders

Growth and body weight respond to selection; litter size almost does not. A 2024 meta-analysis pooling 147 heritability estimates from 34 papers published between 1992 and 2022 put slaughter weight at h² = 0.18 (95% CI 0.14–0.23), litter size at birth at 0.11 (0.10–0.13) and litter size at weaning at 0.09 (0.07–0.10) — and individual herds have reported litter traits as low as 0.02–0.05.

That gap is the most useful thing quantitative genetics has to say to a rabbit breeder. Heritability is the share of visible variation in a trait that passes on as additive genetic value, so it measures how much of a culling decision actually reaches the next generation. At h² = 0.30, roughly a third of the superiority of the animals you keep carries forward. At h² = 0.05, nineteen twentieths of what you were ranking on was feed, season, cage position and luck.

Individual studies place the growth traits consistently above the litter traits. Body weight at eight weeks came out at 0.31 and marketing weight at 0.34 in a New Zealand White herd in Egypt; average daily gain sat between 0.21 and 0.27 across the Pannon lines at Kaposvár; daily gain from birth to 90 days reached 0.31 in Danish White rabbits. Litter traits in the same journal run from 0.02 to 0.19, clustering near 0.10.

• Litter size at birth — pooled h² 0.11 (95% CI 0.10–0.13), from 49 estimates
• Litter size at weaning — pooled h² 0.09 (0.07–0.10), from 33 estimates
• Litter weight at birth — pooled h² 0.11 (0.09–0.14)
• Litter weight at weaning — pooled h² 0.11 (0.07–0.14)
• Slaughter weight — pooled h² 0.18 (0.14–0.23), the highest of the five
• Body weight at 8 weeks — 0.31 in New Zealand Whites; marketing weight 0.34
• Average daily gain — 0.21–0.27 in the Pannon lines, 0.11 in a Spanish synthetic line
• Post-weaning daily gain, birth to 90 d — 0.31 in Danish White rabbits

Heritability is a ratio: additive genetic variance divided by total phenotypic variance, estimated in one population, in one environment, at one point in time. It answers a narrow question — if I rank these animals on what I can measure and keep the top ones, how much of that advantage will their offspring inherit? Everything else people read into the number is imported.

It does not describe an individual: there is no sense in which 31% of one rabbit's eight-week weight is genetic, only a statement about why animals in that herd differ. Nor does it transfer between herds — the same meta-analysis found heterogeneity indices of 90% for litter size at birth and 96% for litter weight at weaning, exactly as you would expect from a parameter that depends on local management. A review of genotype-by-environment interaction in rabbits and pigs makes the same point: heat, season and housing type change how genotypes rank, so animals should be evaluated in conditions resembling where they will work.

And low heritability does not mean genes are irrelevant. It means phenotype is a poor guide to genotype for that trait, so selection has to be indirect — through relatives, through repeated records on the same doe, through crossbreeding — rather than by looking at one doe's last litter and deciding. FAO's worked example for rabbits is explicit: because litter size is weakly heritable, selection should use the average of at least three successive kindlings, not one.

• h² is a property of a population and an environment, not of an animal
• A high h² means phenotype predicts breeding value well — nothing more
• A low h² does not mean the trait has no genetic basis
• Estimates from another country's herd are a guide, not your herd's number
• Traits with low h² need repeated records or family information to select on

Growth rewards selection in a way litter size never has. In a Spanish synthetic line selected for daily gain, the response was measured against a genuine control — embryos frozen at generations 3 and 4, thawed and raised alongside generation 10. The selected animals grew at 49.8 g/day against the control's 45.9, and reached 2350 g at slaughter against 2180 g. That is roughly 8% more growth, banked in seven generations, and it was verified rather than assumed.

A French divergent-selection experiment on 63-day body weight, with a cryopreserved control raised alongside generation 5, reported h² = 0.22 across 4754 animals and separated its high and low lines cleanly. FAO reaches the same conclusion from theory: mass selection on post-weaning growth progresses faster than on litter size, because the trait is more heritable and is measured in both sexes — which doubles the pool you select from.

Two cautions keep this honest. Early weights are not really the kit's own trait: maternal genetic effects accounted for 14–32% of phenotypic variance at weaning and 70 days in an Indian New Zealand White herd, and the common-litter effect explained 60% of variance in pre-weaning daily gain in Danish Whites. Weigh at weaning by all means, but select on later weights. And response is not infinite: a paternal line selected for post-weaning daily gain for 37 generations, checked against rederived vitrified embryos, had slowed to 0.113 g/day per generation between generations 19 and 37, which the authors read as exhaustion.

• Weaning weight (42 d) — h² 0.42, but 14–32% of variance is maternal
• Body weight 70 d — h² 0.40; body weight 135 d — h² 0.27
• 63-day body weight — h² 0.22 across 4754 animals in a divergent-selection trial
• Measured response — 49.8 vs 45.9 g/day and 2350 vs 2180 g against a frozen control
• Pre-weaning daily gain — 60% of variance came from the litter, not the individual
• After 37 generations, response fell to 0.113 g/day per generation

The pooled heritability of litter size at weaning is 0.09, and it falls further in real herds: a Brazilian closed population reported direct heritabilities of 0.02 to 0.05 for its reproductive and litter traits, and an Egyptian New Zealand White herd reported 0.05 for litter size at birth. Danish White rabbits managed 0.188 at birth and 0.081 at weaning. Whichever estimate you take, more than nine tenths of the difference between two does' litters is not additive genetic.

What that buys in practice has been measured. A maternal line selected for litter size at weaning was compared across nine generations using cryopreserved contemporaries, the cleanest design available: the direct response was 0.77 ± 0.27 more rabbits weaned per litter over the whole nine generations. That is real, statistically supported progress — and it is roughly a twelfth of a kit per generation, achieved by a research station with pedigree records, frozen controls and BLUP evaluation.

FAO's arithmetic agrees. With mass selection on the average of a doe's first three litters, and 25% of does kept, expected gain is 0.15 young weaned per generation; adding sisters' and dam's records lifts it to 0.19. Over ten generations that is about +1.5 kits — which FAO recommends to smallholders as worthwhile, and which is roughly a decade of work for one extra kit and a half. Repeatability tells you why: at 0.40 against a heritability near 0.10, most of what makes a doe consistently good is permanent environment and her own history, not something she transmits.

• Litter size at birth — 0.188 (Danish White), 0.11 (pooled), 0.05 (Egyptian NZW)
• Litter size at weaning — 0.081 (Danish White), 0.09 (pooled), 0.02–0.05 (Botucatu)
• Repeatability of litter size ~0.40 against h² ~0.10 — most consistency is not genetic
• Nine generations of dedicated selection: +0.77 ± 0.27 kits weaned per litter
• FAO expectation: +0.15 to +0.19 weaned per generation, about +1.5 over ten

Litter size is the product of two components — ovulation rate and prenatal survival — and both are dominated by things a breeder controls directly. A 2023 review of the evidence found that feed restriction and reduced daylight hours lower ovulation rate, while body condition, parity order and the doe's age raise it. Feed restriction and heat stress also cut embryonic and foetal survival, and good uterine vascularisation and available space in the uterine horn are what carry embryos to term.

The same review documents the most instructive failure in rabbit breeding. Ovulation rate was proposed as an indirect selection criterion precisely because it is more heritable than litter size itself. Selection worked — ovulation rate rose — but litter size did not follow, because oocyte quality deteriorated as ovulation rate climbed and prenatal mortality absorbed the gain. Selection for uterine capacity, the other obvious indirect route, has likewise failed to increase litter size.

So the trait that resists selection is the one most responsive to husbandry. A doe in good condition, on unrestricted feed, in a shed that does not overheat, will out-produce a genetically identical doe kept badly by a margin far larger than a decade of selection would deliver — which is an argument for recording management alongside litters, not against records.

• Raise ovulation rate — good body condition, higher parity, adequate daylight
• Lower it — feed restriction, short photoperiod
• Reduce prenatal survival — feed restriction, heat stress
• Selection for ovulation rate raised the trait but not litter size
• Selection for uterine capacity also failed to increase litter size

If litter size will not respond to selection, it will respond to crossing. A full diallel between four Spanish maternal lines, analysed over 34 546 parities from 7111 does on four farms, produced heterosis of up to 1.05 kits total born and 1.11 kits born alive in the best cross, and 0.90, 0.70 and 0.58 extra kits weaned in three others. Set that against 0.77 kits weaned bought with nine generations of within-line selection, and the commercial logic of the crossbred doe is immediate: one mating decision delivers what a decade of culling delivers.

Heterosis is a maternal and reproductive phenomenon, not a growth one. In the growth analysis of the same diallel, maternal heterosis was generally negative — the AH cross lost 53 g of body weight at 43 days and 3.5 g/day of gain between 28 and 42 days — which the authors attribute to the positive heterosis for litter size: more kits on the same doe means less milk each. A separate Spanish trial on post-weaning growth, feed intake and feed conversion found neither heterosis nor maternal effects significant for any trait.

Hence the standard commercial structure: crossbred does for litter size and mothering, a selected paternal line for growth, and every kit from the terminal mating going to market rather than back into the herd. FAO describes the same family of schemes for smallholders — a two-breed cross with the buck bought in, a three-breed cross using a crossbred female, or rotational crossing, which lets a breeder produce his own replacement females and buy only bucks.

• Two-breed terminal cross — local does × bought-in meat buck, all progeny sold
• Three-breed cross — an AB crossbred doe mated to a C-line terminal buck
• Rotational crossing — alternate sire breeds; you keep breeding your own does
• Synthetic strain — cross, then breed within, keeping half the original heterosis
• Expect heterosis in litter traits, not in post-weaning growth

Genetic gain per year is accuracy times selection intensity times the genetic standard deviation, all divided by the generation interval — and three of those four terms are set by how a rabbitry operates. Accuracy comes from the quality of your records and whether you use relatives' information; intensity from how small a fraction of candidates you keep; the interval from how old parents are, on average, when their replacements are born.

FAO's worked scheme keeps 25% of does as dams of daughters, giving a selection intensity of 1.27, and 10% as dams of sons, giving 1.75 — and notes that a generation interval of 10 to 12 months is practicable at that rate. Real herds sit close to this: an 18-year pedigree analysis of a closed New Zealand White population measured a mean generation interval of 1.49 years. Halve the interval and you double annual gain for the same selection pressure, which is why commercial nucleus units breed replacement does early.

Replacement rate is what makes intensity possible, and rabbits are unusually well supplied. Monitored production units cull an average of 112% of the female stock per year, rising to 126% in the most productive, with monthly rates of 8–10% not uncommon. You already replace the whole doe herd annually; the only question is whether the replacements are daughters of your best-recorded does or simply the ones that happened to be available — decided entirely by whether you know which doe produced which young rabbit.

• Keep 25% of does as dams of daughters — selection intensity i = 1.27
• Keep 10% as dams of sons — i = 1.75
• Generation interval of 10–12 months is achievable at those rates
• A real closed herd averaged a 1.49-year generation interval over 18 years
• Commercial female culling runs 112% per year, 126% in the best units

None of the above is available to a breeder who cannot say which kit came from which doe. FAO's performance-control list begins with individual identification: every rabbit tattooed or ear-tagged at weaning, with a number long enough not to repeat. Then a genealogical card carrying the animal's own number and those of its sire and dam — without it no relative's information can be used, and low-heritability traits become unselectable.

The doe and litter records are next: service dates with the buck's identity, the palpation result, kindling date, the doe's parity, and the number born live and stillborn at the first nest inspection. Then weaning date, number weaned, weaned litter weight and — where growth is a selection objective — individual weaning weights, plus the number and individual weights at 70 days. Buck cards carry service dates, does served, palpation outcomes and live and stillborn counts, so each buck's conception rate and the prolificacy of his does can be read off.

Every one of those items exists because a calculation needs it. Average daily gain needs two weights and the two dates between them. Distinguishing a small litter from a well-grown one that lost kits needs born-alive and weaned counted separately. Comparing does on numerical productivity needs the interval between first and nth kindling, which needs the dates. This is why rabbit selection is a paperwork problem before it is a genetics problem, and why keeping the herd's identity, parentage, weights and dates in one register — a doe card and buck card system, or an application like RabbitBreeder that computes the intervals and gains for you — is the precondition rather than the polish.

• Individual identity — tattoo or ear tag applied at weaning
• Parentage — the animal's number plus its sire's and dam's numbers
• Service date and buck identity for every mating; palpation result
• Kindling date, parity, born alive and stillborn at first nest check
• Weaning date, number weaned, weaned litter weight, individual weights
• Number and individual weights at 70 days where growth is selected

Inbreeding accrues at a rate inversely proportional to effective population size: ΔF = 1/(2Ne), where Ne is not your animal count but the size of an idealised herd that would drift as fast as yours. Wright's approximation, Ne = 4NmNf/(Nm+Nf), shows why the scarcer sex dominates, and because some mass selection is always present FAO advises multiplying the result by 0.7. Its breeding-programme guidance sets the target plainly: keep effective population size above 50.

Published figures for small breeding units make the arithmetic concrete. With three bucks and nine does under random selection, inbreeding runs at 5.6% per generation (Ne 8.9); careful control of each animal's contribution brings that to 2.9% (Ne 17.2). Five bucks and 25 does give 3.0% falling to 1.7%; six bucks and 18 does give 2.8% falling to 1.4%; ten bucks and 50 does give 1.5% falling to 0.8% (Ne 62.5). The lesson is not the exact figures but their shape: adding bucks and equalising their contributions does more than adding does.

What happens when nobody watches is documented. A New Zealand White herd closed for 18 years finished with a mean inbreeding coefficient of 13.23%, the whole population inbred, inbreeding rising 1.94–3.98% per generation, four ancestors supplying half the gene pool and a realised effective population size of 14.93 — which the authors called critical. The cost is measurable: in a Brazilian herd each 10% increase in the dam's inbreeding coefficient cost 0.805 kits born alive, 0.589 weaned and 211 g of litter weaning weight.

The remedies are unglamorous. Buy bucks in rather than breeding your own indefinitely; avoid full-sib, half-sib and parent-offspring matings; let each buck leave roughly one son so no male's line swamps the herd; and rotate breeding groups on a fixed scheme rather than mating whoever is convenient. The Pannon programme credits its circular mating scheme with keeping inbreeding rising slowly across three breeds. All of it depends on knowing relatedness before the mating, which is what a maintained pedigree — on cards or in a herd register such as RabbitBreeder — exists to tell you.

• 3 bucks, 9 does — ΔF 5.6% per generation (Ne 8.9), or 2.9% if contributions are controlled
• 5 bucks, 25 does — 3.0% (Ne 16.7), or 1.7% (Ne 29.4)
• 6 bucks, 18 does — 2.8% (Ne 17.9), or 1.4% (Ne 35.7)
• 10 bucks, 50 does — 1.5% (Ne 33.3), or 0.8% (Ne 62.5)
• FAO target — effective population size above 50
• Cost of ignoring it — −0.805 born alive per 10% of dam inbreeding