Reading time: Approximately 15 minutes

If you breed two Bettas together, can you predict what their offspring will look like?

Sometimes.

For certain traits, if we understand how the trait is inherited and know the genetic makeup of both parents, we can calculate the probability of particular genotypes appearing among their offspring.

For other traits, prediction becomes much more difficult.

Modern ornamental Betta splendens can carry a complicated mixture of genetic variants accumulated through generations of selective breeding, crossing between domestic lines and, historically, genetic contributions from related wild Betta species. Research has now identified major genetic regions associated with characteristics including colour, fin length, Doubletail, enlarged pectoral fins, body size and sex determination. It has also shown that other traits have much more complicated genetic architecture.

So a Punnett square can be an extremely useful tool.

But it is not a crystal ball.

Used correctly, it tells us about genetic probability.

Used incorrectly, it can give a breeder an entirely false sense of certainty.

This second article in the Castle Dawn Aquatics Betta Genetics & Selective Breeding Series explains how Punnett squares work, what breeders mean by P, F1 and F2 generations, how a simple Betta cross can be predicted, and most importantly, where those predictions begin to break down.

If terms such as genotype, phenotype, allele, homozygous and heterozygous are unfamiliar, we recommend reading Betta Genetics Explained: A Beginner's Guide to Genes, Traits and Inheritance before continuing.

Why do breeders try to predict a spawn?

Selective breeding is much easier when you have some idea what a pairing might produce.

Imagine that you have spent several generations developing a line and want to introduce a particular trait.

Without any knowledge of inheritance, you might breed two fish together simply because both look attractive and hope that something useful appears.

Occasionally that works.

It is not a particularly efficient breeding strategy.

Understanding inheritance allows a breeder to ask much more useful questions.

Could the desired trait appear in the first generation?

Could the first generation carry the trait without visibly expressing it?

Might two F1 siblings need to be bred together before the phenotype appears?

Could a parent that looks completely normal still carry the allele you need?

Would an outcross temporarily make your desired phenotype disappear?

These are genetic questions.

And for relatively simple inheritance patterns, a Punnett square can help answer them.

What is a Punnett square?

A Punnett square is a simple diagram used to show the possible combinations of alleles that offspring could inherit from their parents.

The concept is based on Mendelian segregation.

During the production of sperm and eggs, the two alleles an individual carries at a genetic locus separate, meaning an individual gamete receives one of them. At fertilisation, genetic material from the mother and father comes together again in the offspring.

Punnett squares show the possible combinations created by this process.

Modern genetics still uses these principles when discussing simple Mendelian inheritance, although we now understand many situations where inheritance is more complicated than the classic single-gene examples.

Many breeders use the Punnett square as a way of displaying potential offspring genotypes when the parental genotypes and inheritance pattern are known. 

The important word is possible.

A Punnett square does not tell us exactly which sperm will fertilise which egg.

It gives us probabilities.

The P generation, F1 and F2

Before constructing a breeding example, we need three pieces of terminology.

P generation

The P generation is the parental generation chosen as the starting point for the cross you are describing.

If you breed Male A to Female B and use that mating as the beginning of your breeding project, those two fish are your P generation.

F1 generation

Their offspring are the F1 generation.

F stands for filial, meaning offspring.

So F1 means the first filial generation produced from the designated parental cross.

F2 generation

If suitable F1 offspring are subsequently bred to one another, their offspring become the F2 generation.

The sequence can then continue:

F3, F4, F5 and beyond.

There is an important distinction here.

F2 does not simply mean "my second spawn".

If you repeat the original P-generation pairing, you have produced another F1 spawn.

F2 refers to the next filial generation descended from the F1 generation.

That distinction becomes particularly important when keeping proper breeding records.

Backcrosses are different

Breeders sometimes take an F1 offspring and breed it back to one of the original parents, or to a closely related individual representing that parental line.

This is a backcross.

A backcross is not the same thing as an F1 sibling cross producing F2.

This matters because the expected genetic outcome can be very different.

Later in this series, when we discuss creating a line, linebreeding and outcrossing, backcrossing will become an important tool.

For now, concentrate on understanding the basic P → F1 → F2 relationship.

A real Betta example: Doubletail

Rather than using imaginary flowers or peas, we can demonstrate simple inheritance using a genuine Betta phenotype.

Doubletail is particularly useful because modern genetic research has investigated its molecular basis.

A 2021 genomic study found Doubletail Bettas to be recessive homozygotes and mapped the responsible region to approximately 130 kb on linkage group 1. The researchers identified a regulatory deletion associated with altered expression of the developmental genes zic1 and zic4.

For this educational example, we will use simplified symbols:

ST = common Singletail allele

st = recessive Doubletail allele

These symbols are being used here for clarity. Betta hobby literature has used several different genetic symbols over the years, and there is no reason to assume every traditional notation system corresponds perfectly with modern molecular genetics.

That distinction is important.

Cross 1: Singletail × Doubletail

Suppose we have a Singletail parent that is homozygous for the common allele:

ST/ST

and a Doubletail parent:

st/st

The Singletail parent can only contribute ST.

The Doubletail parent can only contribute st.

Every F1 offspring therefore receives:

ST/st

Under this simple recessive model, the F1 offspring would not display the Doubletail phenotype because they have only one copy of the recessive Doubletail allele.

But genetically they carry it.

This is the first major lesson Punnett squares give a breeder:

A trait can disappear visually in F1 without disappearing genetically.

If you looked only at phenotype, you might conclude that the Doubletail trait had been lost.

It has not.

It is present in every F1 offspring in this particular cross.

Why a carrier matters

The F1 fish now has two different alleles:

ST/st

It can pass ST to an offspring.

Or it can pass st.

This means the Doubletail-associated allele can travel silently through a line even when the fish carrying it has a Singletail phenotype.

That is why pedigree information is so useful.

Two Bettas can look like ordinary Singletails but have very different breeding value depending on what they carry.

Cross 2: F1 carrier × F1 carrier

Now imagine that we select a male and female from the F1 generation.

Both have the genotype:

ST/st

Each parent can produce gametes carrying either ST or st.

The possible offspring combinations are:

This produces an expected genotypic ratio of:

25% ST/ST

50% ST/st

25% st/st

Under a simple recessive inheritance model, that corresponds to an expected phenotype ratio of approximately:

75% Singletail

25% Doubletail

This classic 1:2:1 genotype ratio from two heterozygous parents is one of the most familiar results in Mendelian genetics.

But now we reach one of the most important sections of this entire article.

25% does not mean exactly 25% of every spawn

Imagine an F1 carrier cross produces 40 viable offspring.

The expected proportion of st/st offspring is 25%.

Twenty-five per cent of 40 is 10.

Does that mean your spawn must contain exactly 10 Doubletails?

No.

It means that under our simplified model, each offspring has the relevant probability of inheriting that genotype.

The actual number observed in one finite group can fluctuate around the expected value.

You might observe fewer.

You might observe more.

Over increasingly large numbers of offspring and repeated crosses under the same assumptions, observed frequencies would generally be expected to move closer to the theoretical probability.

This is the same reason that flipping a fair coin four times does not guarantee exactly two heads and two tails.

Probability describes likelihood, not a predetermined sequence.

In Mendelian breeding experiments, offspring counts can be treated statistically rather than as guaranteed fractions. A genotype with a 25% Mendelian probability does not have to make up exactly one quarter of a particular litter, clutch or spawn.

That distinction becomes particularly important with Bettas because breeders frequently work with a limited number of surviving fry rather than an infinite population.

Every offspring is another genetic event

One common misunderstanding is:

"I have already had three Singletails, so the fourth fry should be Doubletail."

That is not how the probability works.

Previous offspring do not normally "use up" the other possible genetic outcomes.

Under the simplified model, each fertilisation represents another opportunity for those alleles to combine.

The expected 25% is therefore not a quota that the spawn has to fill.

This is fundamental to understanding breeding percentages.

The Punnett square predicts genotype, not breeding quality

There is another limitation that is particularly important to Betta breeders.

Suppose our Punnett square predicts that approximately 25% of the offspring could be st/st Doubletails.

It tells us nothing about whether those Doubletails will be good examples of the phenotype.

A Punnett square does not tell us whether a particular fish will have:

excellent body structure,

balanced finnage,

good dorsal form,

strong growth,

desirable colour,

clean pattern,

good fertility,

or the overall quality you are trying to establish in your line.

It predicts allele combinations under a genetic model.

It does not replace breeder selection.

This is why producing the phenotype you want is only one stage of selective breeding.

Producing a high-quality and consistent line is a much larger project.

Genotype must be known, not guessed

Punnett squares become unreliable very quickly if the parental genotypes are wrong.

Consider two Singletail Bettas.

They look identical with respect to the Doubletail trait.

But one could be:

ST/ST

while the other could be:

ST/st

Phenotype alone does not necessarily tell you which fish is which.

If you assume a fish is ST/ST when it is actually ST/st, your prediction can change completely.

This is why breeder records matter.

Knowing that a fish had a Doubletail parent, or that Doubletails appeared among its siblings, gives you information that appearance alone cannot provide.

The same principle applies throughout selective breeding:

Phenotype gives you clues, pedigree gives you context, and offspring can reveal what the parent was carrying.

Sometimes offspring reveal the parental genotype

Imagine a Singletail fish of unknown genotype.

It could be:

ST/ST

or:

ST/st

If that fish produces Doubletail offspring when bred to a genetically appropriate partner, the offspring provide evidence that the unknown parent carried the recessive allele.

Historically, breeders have used breeding results in exactly this way to infer hidden genetic information.

This is sometimes called a test cross when a cross is deliberately designed to help establish genotype.

Modern molecular genetics can directly identify some variants where the causal variant and suitable testing methods are known.

Most hobby Betta breeders, however, still rely heavily on pedigree, phenotype and actual spawn results.

Why the F2 generation can be so revealing

The F1 generation is sometimes visually disappointing to a breeder.

That does not necessarily mean the cross failed.

Imagine crossing two lines because you want to combine traits that have never previously existed together.

The F1 fish inherit material from both lines, but some desired recessive phenotypes may disappear.

Breed appropriate F1 individuals together, and genetic recombination can produce new combinations in F2.

This is why breeders frequently talk about F2 as an important generation.

Traits that were hidden in F1 may reappear.

New combinations can emerge.

Variation can increase dramatically.

But so can undesirable combinations.

F2 is therefore often where the breeder has to do a considerable amount of selection.

F1 does not mean "better" and F2 does not mean "more advanced"

The generation numbers describe ancestry.

They are not quality rankings.

An F5 Betta is not automatically better than an F1 Betta.

A poorly managed breeding programme can become worse over five generations.

A carefully chosen F1 cross can produce excellent fish.

Selective breeding depends on what is being selected and why, not simply how many generations have passed.

This is another reason breeding records should contain much more than generation numbers.

What happens when you try to predict two traits?

So far, we have followed one genetic locus.

Real Betta breeding rarely stays that simple.

Suppose a breeder is simultaneously interested in:

fin type,

colour,

pattern,

body structure,

scale characteristics,

and perhaps several additional traits.

Each may have a different genetic architecture.

If two simple genes are on different chromosomes or sufficiently unlinked, classical Mendelian models can treat their assortment independently. But genes that are physically linked can be inherited together more often than expected under simple independent assortment.

The number of possible genetic combinations also increases rapidly as more loci are considered.

A simple one-locus heterozygous cross gives four boxes.

A conventional two-locus heterozygous cross gives sixteen.

Add more genetic loci and the theoretical combinations multiply rapidly.

This is one reason breeding towards several characteristics simultaneously can require many offspring and several generations of selection.

The 6.25% problem

Here is a useful way to understand how probabilities compound.

Imagine two independent recessive traits.

Suppose a particular cross gives a 25% probability of producing the desired genotype for Trait A.

And independently, there is a 25% probability of producing the desired genotype for Trait B.

If those events are genuinely independent, the probability of an offspring receiving both combinations would be:

25% × 25% = 6.25%

So even before considering quality, form, sex, fertility or any other selection criterion, relatively few offspring may have the exact combination you are looking for.

Add a third independent 25% requirement and the theoretical probability falls further.

This helps explain why serious selective breeding often requires growing out significant numbers of offspring.

It is not simply about producing fry.

It is about producing enough variation to give selection something to work with.

But Betta traits are not all simple Mendelian traits

This is where we need to be particularly careful.

Punnett squares work beautifully when the assumptions are appropriate.

Nature is under no obligation to keep every ornamental Betta characteristic simple.

Modern genomic studies have shown both ends of the spectrum.

Doubletail, Elephant Ear or Dumbo, albino and fin-spot phenotypes have been associated with major-effect loci.

Other characteristics are clearly more complicated.

Research involving 727 Bettas found multiple genomic associations affecting colour patterns, body size, long fins and behaviour. Mosaic colour pattern showed associations across multiple genomic regions, while aggression displayed a polygenic signal involving several neural-system-related genes.

Therefore:

A simple Punnett square is appropriate for some questions.

It is not a universal model for every colour, pattern, body shape or behavioural characteristic in a modern Betta.

A note about older Betta genetics literature

This is also where source checking becomes important.

Older Betta breeding books and hobby literature contain enormous amounts of useful observational knowledge.

They were often written before today's genomic studies existed.

For example, some books present a Doubletail inheritance model and genetic notation based on the breeder understanding available at the time, including discussion of visible effects in presumed carriers.

Modern molecular work has since mapped Doubletail and described the Doubletail phenotype as a recessive homozygous condition associated with a regulatory region affecting zic1/zic4.

That does not make older breeder literature worthless.

It demonstrates why the responsible approach is to combine:

historical breeding observations + controlled crosses + modern genomic evidence.

Throughout this Castle Dawn Aquatics series, when newer research changes or refines an older breeder model, we will say so.

Why predicted ratios may differ from the fry you actually raise

Random sampling is only one reason.

The ratios you observe among adult or sale-sized Bettas can also differ from the theoretical ratio expected at fertilisation because several biological stages occur between conception and adulthood.

Eggs may fail to develop.

Embryos may not hatch.

Some fry may grow poorly.

Some individuals may die.

Certain genotypes may potentially affect survival or development.

Phenotypes may also be incorrectly identified at a young age or become clearer as the fish matures.

Therefore, a Punnett square predicts genetic combinations based on its assumptions.

The group of fish you eventually evaluate months later has also passed through development and survival.

Those are not necessarily the same statistical population.

Penetrance and variable expression

Another complication is that inheriting a genetic variant does not always guarantee an identical visible result in every individual.

In genetics, penetrance refers to whether a genotype produces its associated phenotype at all.

Expressivity describes how strongly or in what manner that phenotype is expressed.

Environmental conditions, modifier genes and interactions elsewhere in the genome can influence what you actually see.

This is another reason breeders should be cautious when trying to reverse-engineer an entire genotype from appearance.

A photograph can tell you a lot.

It cannot tell you everything.

What about Halfmoon, marble, koi and complex colours?

This is where simplistic online genetics charts often become misleading.

You should not assume that because AA × aa can be represented neatly in four boxes, every Betta characteristic can be predicted the same way.

Some traits have major-effect loci.

Others involve multiple genetic regions.

Some traditional breeder inheritance models remain useful but have not been fully validated at the molecular level.

Patterns such as modern mosaic phenotypes can have a polygenic basis. Modern genomic research found multiple associated loci for mosaic colouration rather than one simple universal mosaic switch.

That is particularly important when discussing modern koi and fancy lines.

Two fish that look similar can have different genetic backgrounds.

Commercial names describe phenotypes and breeding traditions.

They do not necessarily describe one precise genotype.

We will explore this in much greater detail in the dedicated colour and marble genetics articles later in this series.

Can you predict the sex ratio?

Sex determination in ornamental Bettas is another good example of why modern research matters.

Genomic studies have identified dmrt1 as an important sex-determination locus in domesticated ornamental Bettas. Research has also found that sex determination in the wider B. splendens complex is more complicated than a simple universal mechanism shared by all related populations and species.

For the practical breeder, this means we should avoid casually treating every aspect of sex inheritance as perfectly predictable from a basic classroom Punnett square.

Again, genetics gives us increasingly powerful information.

It does not justify pretending biology is simpler than it is.

Punnett squares are most useful when you ask a narrow question

A Punnett square becomes most useful when the breeder asks something specific.

For example:

"If both parents carry this known recessive allele, what genotypes could their offspring inherit at this locus?"

That is a good question.

A much poorer question is:

"Here are two photographs of Bettas. Tell me exactly what their entire spawn will look like."

There is simply too much missing information.

The best genetic predictions begin with good information.

What a Punnett square can tell a Betta breeder

For a suitably understood genetic trait, it can help identify:

possible offspring genotypes,

expected genotype probabilities,

which offspring may be carriers,

whether a recessive phenotype could appear,

whether an F1 generation may need to be bred again to recover a phenotype,

and which crosses might be useful for a longer breeding objective.

What it cannot tell you is equally important.

It cannot guarantee exact numbers in a spawn.

It cannot identify unknown parental genotypes by magic.

It cannot predict complicated polygenic traits using a one-gene model.

And it cannot tell you which offspring will be the best breeding fish.

A breeder should think in probabilities, not promises

Perhaps the most useful change genetics can make to a breeder's thinking is replacing certainty with probability.

Instead of:

"This pairing will produce 25% Doubletails."

think:

"Under this inheritance model, each offspring has an expected 25% probability of inheriting the Doubletail genotype."

Instead of:

"These parents will produce good Halfmoons."

ask:

"What evidence do I have about the genetic background, phenotype, siblings and previous offspring of these parents?"

And instead of:

"The Punnett square says this should happen."

ask:

"Are the assumptions behind my Punnett square actually correct?"

That is the difference between using genetics as a breeder's tool and treating a breeding chart as a guarantee.

Where selective breeding begins

A Punnett square can help you understand what may be genetically possible.

The breeder then has to decide what is worth keeping.

Suppose your F2 spawn produces the desired genetic phenotype.

You may still have considerable variation in body shape, finnage, colour intensity, pattern, vigour and overall quality.

You now have to select.

That is where genetics begins to turn into selective breeding.

And it leads directly to the next major stage of our series.

Next in the Castle Dawn Aquatics Betta Genetics & Selective Breeding Series

How to Create Your Own Betta Line: A Selective Breeding Guide

In Article 3, we move beyond predicting a single cross.

We will look at how a breeder defines a goal, selects foundation fish, evaluates F1 and later generations, decides which offspring should move forward, maintains more than one branch of a line, and gradually works towards consistency.

This is where the long-term work begins.