Sticklebacks

"Our results support the intriguing hypothesis that most stickleback evolution in fresh water occurs within the first few decades after invasion of a novel environment." — Lescak et al., PNAS, 2015¹

"The hypothesis does not address where or when an advantageous mutation arose" — Schluter and Conte on their own "transporter" model, PNAS, 2009²

Summary

The threespine stickleback, Gasterosteus aculeatus, is a finger-length fish that lives in the sea and breeds in rivers.  Sea-run fish carry a full row of bony side plates and a spiny pelvis.  Thousands of lakes and streams hold resident freshwater populations with few plates and, in some lakes, no pelvis.  Sticklebacks are the best-documented case of adaptation happening fast in a wild vertebrate.  This article asks what was actually watched, what was inferred, and how each genetic change scores under the site's gain, loss, modification or transfer taxonomy.

The findings, in order of strength:

1. Armour loss has been watched in real time, repeatedly. In Loberg Lake, Alaska, the low-plated form was absent in 1990 and made up 75% of the fish by 2001.³ In Cheney and Scout Lakes, stocked with about 3,000 sea-run fish each in 2009 and 2011, the freshwater Eda allele rose from under 1% to over 50% in eight years.^{4 4} Ponds lifted out of the sea by the 1964 Alaska earthquake hold freshwater-type fish that have diverged nearly as far as populations thought to be thousands of years old.¹
2. None of that change needed a new mutation. The freshwater Eda allele already sits in sea-run fish at 0.2% to 3.8%.⁵ In the three Alaskan lakes, 94% of the 344 regions that changed fastest were regions already known to differ between old freshwater and marine populations.⁴ No adaptive variant has been documented arising during any observed period. The study designs could not have detected one, so this is an absence of evidence, not a measured zero.
3. The armour change is a shift in allele frequency. Its cause is a weakened switch, not a deleted one. The freshwater Eda allele carries one T to G change in an enhancer that lowers its response to a growth signal in armour and leaves it working elsewhere.⁶ That scores as a modification, down.
4. The pelvis is lost by loss. Nine pelvic-reduced populations carry nine different deletions of one Pitx1 enhancer, overlapping in 488 base pairs.⁷ The region breaks easily because of a run of TG repeats.⁸ The deletions are independent and overlapping, not "the same deletions each time." Each scores as a loss.
5. Stickleback adaptation is not all loss. Freshwater benthic fish grew more teeth by raising expression of Bmp6,⁹ and several freshwater populations carry extra copies of Fads2, which lets them make an essential fatty acid their diet lacks.¹⁰ Both are inferred from comparisons, and neither adds unique functional sequence: one strengthens an existing switch, the other copies an existing gene.
6. The mainstream objection is strong, and the young-earth literature has not answered it. The standing variation had to come from somewhere. The Eda low and complete haplotypes differ at 1.9% of synonymous sites, more than two stickleback lineages dated at about 2 million years.⁵ A pedigree-measured stickleback mutation rate turns that difference into about 1.9 million generations.¹¹ Producing it by mutation in 4,400 years would need a rate about 430 times the measured one.¹² A young-earth model has to claim the difference was created, not accumulated. Answers in Genesis says exactly that, but no creationist source found tests it.

Verdict.  Sticklebacks are strong evidence for fast adaptation from variation that was already there, and that part fits the creation model of rapid speciation well.  They are weak evidence for the origin of anything new, because nothing new was seen arising.  But the young-earth reading has a hole of its own:  the reused variants look old by a measured clock, and the only published creationist answer, that they were created, has not been put to the test that could check it.

Part 1: What was watched

The membership rule of this series is that something must have been watched happening.  For sticklebacks, six things were.

Loberg Lake, 1982 to 2017

In 1982 the Alaska Department of Fish and Game poisoned Loberg Lake to clear it for sport fish.  Sea-run sticklebacks, probably from neighbouring Rabbit Slough, got in between 1983 and 1988.⁴  Michael Bell's group sampled the lake every year from 1990.  In 1990 the fish were "nearly monomorphic for the complete plate morph" and the low morph "was absent."  By 2001 complete morphs had fallen to 11% and low morphs had risen to 75%.³  By 2017 the armour-reduced form was about 95%, and the freshwater Eda allele was "approaching fixation in LB by 15 years."^{4 4}

Bell and colleagues drew the conclusion that "threespine stickleback adapt to freshwater habitats within decades after invasion from the ocean."³

Cheney and Scout Lakes, 2009 to 2017

To watch the start of the process, Bell's group released about 3,000 Rabbit Slough fish into each of two poisoned lakes with no outlet:  Cheney Lake in 2009 and Scout Lake in 2011.  Low-plated fish appeared "in the second and third generation after founding" and reached 20 to 30% by 2017.⁴  Garrett Roberts Kingman and colleagues sequenced a pool of 100 fish from every year, plus 200 of the founders' source population.⁴  They found 344 regions whose allele frequencies moved fast, and fitted an average selection coefficient of 0.30 to them.⁴  The Eda allele that marks low armour rose "from an initial frequency of <1% to over 50% within 8 years."⁴

This is the one genome-wide time series of wild vertebrates adapting from the first generation, and it is the centre of this article.

Barrett's ponds, 2006 to 2007

Rowan Barrett, Sean Rogers and Dolph Schluter measured selection on Eda directly.  They trapped about 35,000 fish at Oyster Lagoon, British Columbia, kept the 354 with incomplete armour, and found 182 of them heterozygous at Eda.¹³  On 1 June 2006 they put 45 or 46 heterozygotes into each of four ponds and sampled 50 offspring from each pond ten times over a year.¹³

Two selection coefficients came out, and they point in opposite directions.  From October 2006 to May 2007, the complete allele fell from 67% to 49%, a coefficient of 0.52 against it.¹³  But from July to October 2006, while the fish were still small, selection ran the other way at 0.50 against the low allele, "which offset the gains occurring later in life."¹³  Across the whole generation the low allele went from 50% at birth to 51.6% in breeding adults, and the first sample of their offspring stood at 57.0%, a difference with P = 0.061.¹³

That tension matters, because the single number 0.52 circulates on its own.  Lescak and colleagues write that "The strength of selection (s) on the region containing Eda" in transplanted fish "was found to be 0.52," without the opposing 0.50.¹  Barrett's own reading is more careful:  "countervailing selection pressures diminish the advantage of the low allele over the whole life span," and the fitness effects of Eda "do not seem to be determined solely by differences in lateral plate number."¹³  The honest figure for one generation is a small net gain for the low allele after two large and opposite pushes.  The same group's ponds also showed cold tolerance improving by 2.5 °C in three generations, a rate the authors call "among the most rapid to be observed in a natural population."^{14 14}

Lake Washington, 1957 to 2005

Evolution in lakes does not only strip armour.  In Lake Washington, Seattle, only 6% of sticklebacks were completely plated in 1968 to 1969.  By 2005 the figure was 49%, with 35% partially plated and 16% low.¹⁵  Jun Kitano and colleagues tie the change to Eda and to trout that could see their prey once the polluted lake cleared in the early 1970s.  They estimate a selection coefficient of 0.58 to 0.72 for the plated form between 1969 and 1976.¹⁵  The plate counts were watched through museum and survey samples.  The coefficients are model fits, not measurements like Barrett's.

The earthquake islands, 1964 to 2012

On 27 March 1964 the Great Alaska Earthquake lifted Middleton Island 3.4 m and turned sea floor into freshwater ponds.¹  Emily Lescak and colleagues genotyped more than 1,000 fish from three islands.  The freshwater fish came from the sea after 1964, more than once, and "have diverged phenotypically from oceanic ancestors to nearly the same extent as populations that were likely founded thousands of years ago."¹  The date of the ponds is known.  The colonisations themselves were not watched;  they are reconstructed from the genotypes.

What the watched cases share

Every case is a change in how common an existing form is.  In every case where genes were checked, the variant that rose was already present when the watching began.  That is the fact both sides have to explain.

Part 2: Armour plates and Eda

What Colosimo and colleagues measured

Pamela Colosimo, David Kingsley and colleagues published the key paper in 2005.⁵  It did five separate things, and they should be kept apart.

1. Mapped the armour gene. In a cross between a Japanese marine fish and a Paxton Lake freshwater fish, plate number mapped to Ectodysplasin (Eda), a signalling gene that in mammals is needed for hair, teeth and sweat glands. A mouse EDA transgene gave low-plated fish extra plates.
2. Sequenced the difference. Against the marine sequence, the freshwater allele carried "numerous noncoding changes" and four amino-acid changes, none "at sites that are highly conserved between mammals and fish."⁵
3. Showed the freshwater alleles are one family. Of 15 low-plated populations from across the Northern Hemisphere, every one but Nakagawa Creek in Japan carried alleles from one clade. Genes elsewhere in the genome did not group the low-plated populations together, so these are not one migrating lineage.
4. Measured the allele in the sea. In completely plated marine fish, the low allele turned up at 3.8% at the Navarro River, California (8 of 218 chromosomes), and at 0.2% at the Little Campbell River, British Columbia (1 of 604).⁵ Two later samples fall in the same range. O'Brown found the key freshwater base in 12 of 263 Alaskan sea-run fish, 2.3%.⁶ Barrett's Oyster Lagoon trapping implies at least about 0.26%.¹⁶
5. Estimated the allele's age. This step is an inference, and its method is set out below.

The conclusion follows from steps 3 and 4:  "worldwide evolution of the low-plated phenotype has occurred mainly by recurrent local selection for a family of alleles repeatedly brought into freshwater populations by marine founders."⁵  Lake populations did not invent low armour.  They inherited it from sea-run ancestors who carried it hidden, because heterozygotes are usually fully or partly plated.

How the age was estimated

Colosimo's team compared synonymous differences, the base changes that do not alter the protein and are assumed to build up at a steady rate.  The low and complete haplotypes differ at a fraction Ks = 0.01938 of synonymous sites.  That is more than the 0.01384 between Japanese Pacific and Sea of Japan sticklebacks, a pair the paper says was "thought to have diverged" about 2 million years ago "on the basis of genetic and geological estimates."  It is less than the 0.05875 between G. aculeatus and its sister species G. wheatlandi, dated at about ten million years "on the basis of the split between Pacific and Atlantic populations."  In the paper's words, "The common clade of low-plated alleles thus probably arose between 2 and 10 million years ago."⁵  The abstract rounds this to "more than two million years ago."⁵

The estimate is relative.  It depends on four things:

1. The dates of the two calibration pairs, which come from geology and biogeography.
2. A steady rate of synonymous change across the three comparisons.
3. Synonymous sites being close to neutral.
4. The difference having accumulated by mutation after the two haplotypes split, rather than being present from the start.

The third and fourth assumptions are where the two sides part, and Part 6 returns to them.  A second route to the age now exists that does not use either calibration pair:  a mutation rate measured in stickleback families.  Chaowei Zhang and colleagues sequenced 40 parent-offspring trios and counted 115 new mutations, a rate of 5.11 × 10⁻⁹ per base pair per generation.¹⁷  Dividing the Eda difference by twice that rate gives about 1.9 million generations, or 1.9 to 5.7 million years at one to three years a generation.¹¹  The two routes agree.

What the freshwater allele does

Natasha O'Brown and colleagues found the change that matters in 2015.  In hybrids carrying one allele of each kind, the freshwater allele was expressed "almost fourfold lower" in the developing flanks and spines.  It was not lower in the fins or the lower jaw.⁶  Every low-plated population tested shares one base change, T to G, in an enhancer downstream of the gene.  Putting that one change into the marine enhancer "strongly reduces expression in posterior armor plates," because it "reduces the sensitivity of an EDA enhancer to Wnt signaling, and alters expression in developing armor plates while preserving expression in other tissues."⁶

The same study revised Colosimo's one exception.  The Nakagawa Creek fish carry the same T to G change on an otherwise marine-like background.  The authors offer two explanations and do not choose between them:  the same mutation arose twice, or a tiny piece of the usual freshwater haplotype was carried in by recombination.⁶

How the Eda change scores

The site's scoring procedure asks four questions in order.  Nothing arrived from another species.  No element stopped working:  the enhancer still drives expression in other tissues, and anterior plates still form.  No new element appeared.  So the T to G change is a modification, down.  The four amino-acid differences are modifications whose effect has not been measured.  Behe's scheme would reach the same verdict, since his modification-of-function covers a mutation that does "something else—either tweak an old FCT in a way that leaves it still working or affect some noncoded feature of a cell."¹⁸

Two claims on opposite sides fail this check.  Creation Ministries International wrote that "the mutated (i.e. broken) genetic switch simply prevents expression of a gene."¹⁹  It does not.  It weakens one input to one enhancer.  On the other side, Gregory Lang and Amber Rice cited Eda against Behe:  "none of the observed Eda mutations were predicted to damage the function of the protein."²⁰  That is true, and it is not a gain.  Lower expression in armour is a reduction of an existing function.

What was watched in the lakes is simpler still.  An allele that already existed rose in frequency.  No sequence was gained or lost during the observation.

Part 3: The pelvis and Pitx1

Nine different deletions

Many freshwater sticklebacks have lost the pelvic girdle and its spines.  In crosses, most of that loss maps to the region of Pitx1, a gene that shapes the hind limb in vertebrates and is lethal when knocked out in mice.  Yingguang Frank Chan and colleagues found in 2010 that the gene itself is intact.  What is missing is a stretch of DNA upstream that switches Pitx1 on in the developing pelvis.⁷

Paxton Lake benthic fish carry a 1,868 base-pair deletion that removes this pelvic enhancer.  The Paxton version drove no pelvic expression in any of 8 transgenic fish.  Bear Paw Lake fish carry a different deletion of 757 base pairs, and Hump Lake fish a third of 973.  "The three different deletions in PAXB, BEPA and HUMP overlap in a 488bp region."⁷  A wider survey of 13 pelvic-reduced and 21 pelvic-complete populations found deletions in 9 of the reduced ones and none of the complete ones:  "nine different haplotypes with different staggered deletions, each consistently seen within a pelvic-reduced population."⁷  Putting the marine enhancer and Pitx1 back into pelvic-reduced fish restored pelvic structures.⁷  Four reduced populations had no detectable deletion, and three of those look as if they lost the pelvis another way.⁷

The catalogue once described this as "the same deletions each time."  That is wrong.  The deletions are different in size and endpoints, each is private to its own lake, and they overlap on one target.  Chan's team calls them "independent deletion mutations fixed in different pelvic-reduced stickleback populations."⁷  The Eda story is the opposite:  one old allele reused many times.

Why the same place breaks

Kathleen Xie and colleagues asked in 2019 why the same enhancer keeps being deleted.  In yeast artificial chromosomes, test chromosomes broke at a background rate of 3.37 per million divisions.  Chromosomes carrying the marine pelvic enhancer "broke ~25–50 times more frequently," a rate "even higher than previously analyzed human fragile sites."  Versions from pelvic-reduced lakes broke at the background rate, "suggesting that natural Pel mutations remove breakage-prone regions."⁸  The cause is about 170 base pairs of TG repeats.⁸

Their model is the key point for the ledger.  "Population genetic modeling indicates that new mutations occurring at the low rates of typical single nucleotide changes (~10⁻⁹) would rarely arise at a particular locus in postglacial stickleback populations, while mutations occurring at elevated rates (~10⁻⁵ for fragile sites) would arise often."⁸  So the Pitx1 deletions are the one class of stickleback adaptive change that the mainstream literature treats as new mutations arising inside young lake populations.  The breakage rates were measured in yeast and monkey cells.  The deletions in wild fish were not watched arising.

How the Pitx1 change scores

A coded element that worked, the pelvic enhancer, stopped working, and nothing new took its place.  That is a loss under the site's procedure and under Behe's.  It is also adaptive:  Chan found reduced diversity and an excess of derived alleles centred on the enhancer in pelvic-reduced lakes, the signature of selection.⁷  The sibling article Measuring Biological Information scores the same case the same way.  The fish gained fitness by losing a part.  Both facts are true, and neither cancels the other.

Part 4: The genome-wide picture, and the search for a gain

Old variants, mostly regulatory

Felicity Jones and colleagues sequenced a reference genome and 20 more fish from around the Northern Hemisphere in 2012.  About 0.5% of the genome groups fish by habitat rather than by geography:  242 regions by either of two methods, 147 by both.²¹  Three are chromosome inversions.  Of the 64 regions with the strongest evidence, 41% lie entirely outside genes, 17% change amino acids, and 43% contain genes but no habitat-specific amino-acid change.²¹  The authors conclude that "regulatory changes appear to predominate."²¹

"Regulatory" here means "not an amino-acid change."  It does not mean deleted.  Most of these regions are alternative versions of the same stretch of DNA, differing at many single bases.  The catalogue's phrase for this article, adaptive change "that works by removing regulatory DNA rather than adding any," fits Pitx1 and does not fit the genome as a whole.  Answers in Genesis summarised Jones fairly as "about 80% of the genetic differences arise from alterations in regulatory DNA and about 20% in actual coding DNA."²²

Thomas Nelson and William Cresko dated these regions in 2018.  Using the ninespine stickleback as an outgroup and fixing the split between the two genera at 15 million years, they found the habitat-linked variation "averages six million years old, nearly twice the genome‐wide average."^{23 23}  In some regions, marine and freshwater threespine sticklebacks "are as divergent as threespine and ninespine stickleback, which are classified into separate genera."²³  Roberts Kingman's team, using three calibration species, put the reused regions at 1,600 thousand years against 700 thousand for the rest of the genome, "the majority (161 of 209)" at over a million years.^{4 4}  The two papers' absolute ages differ severalfold.  Both find the adaptive regions about 1.6 to 2.3 times older than the genomic background.²⁴

Candidates for a gain

The series rule is to look hard for a gain before saying there is none.  Four candidates turn up.

More teeth.  Freshwater benthic sticklebacks from Paxton Lake have more throat teeth than marine fish.  Phillip Cleves and colleagues mapped the largest effect to Bmp6, found no amino-acid differences, and found the freshwater allele expressed more strongly in developing tooth plates.^{9 9}  They present it as a rare case of "how morphological features may be gained in natural species."  At the level of the organism it is a gain of structures.  At the level of the molecule, an existing gene is switched on more strongly:  modification, up.  The causal bases were not identified in that paper.

More copies of a fat gene.  Freshwater food is poor in DHA, a fatty acid fish need.  Asano Ishikawa and colleagues found that several freshwater populations carry extra copies of Fads2, a gene for making DHA, duplicated on the X chromosome, and that "fish with higher Fads2 copy number better synthesized DHA when fed with DHA-deficient diets."¹⁰  This is the clearest adaptive gain of function in the stickleback record.  Under the scoring procedure an element "present in more copies" is a modification, up.  It adds no unique functional sequence, because a duplicate is a copy.  Behe scores duplication the same way:  "no new functional coded element was gained or lost, just copied."¹⁸  The 2019 Science paper that reported the duplications was not obtained;  the summary here is the authors' own, from 2021.

Haplotypes that grew.  Roberts Kingman's team found that older reused regions are larger, and that the oldest bases sit at their centres.  Their reading:  "alleles originally based on an initial beneficial mutation accumulating additional linked favorable mutations, snowballing over time to form a finely tuned haplotype with multiple adaptive changes."⁴  If that is right, these haplotypes were built by mutation and selection, and they would be real gains of sequence-specific DNA.  It is inferred from age patterns over a million years.  Nothing like it was watched.

The Eda T to G itself.  On the mainstream account it arose by mutation once, or twice, long ago.  It scores as a modification that lowers activity, not a gain.

So nothing in the stickleback record is a newly arisen gain, if "newly arisen" means watched.  If it means inferred, the best candidates are the Fads2 copies, which are copies, and the snowballing haplotypes, which are a reconstruction.

The scored record

"New sequence" counts bases changed by a mutation that are not a copy, following Measuring Biological Information.  The two observed genetic rows have zero because nothing new arose while anyone watched.

Part 5: The ledger

The series runs the same ledger for every organism:  individuals times generations times mutation rate gives the number of new mutations the population tried.  Then it lists every documented adaptive gain.  For sticklebacks the ledger can be run for the observed lakes, and it gives a clear answer about why standing variation matters.  It cannot be run for output.

Mutations tried in the watched lakes

The arithmetic is in the note.²⁵  The populations surely grew past their founding size, so the true number of mutations tried is larger, perhaps by one or two orders of magnitude.  That does not change the conclusion.  A specific single-base change, such as the Eda T to G, had about one chance in 7,000 of arising anew in these lakes in the time observed.  It rose from under 1% to over 50% in eight years because it was already there.  Xie's team makes the same argument for postglacial lakes.⁸  Both sides agree on this point.  Fast adaptation in small lakes runs on standing variation, except at a fragile site, where a deletion could appear about once in a decade.

Adaptive variants that arose during observation

The zeros need a caution.  A new mutation starts as one copy among thousands.  Pooled sequencing of 100 fish cannot see anything below a few percent, and a new allele could not climb that high in a decade unless its advantage were enormous.  These studies were designed to follow old variants, not to find new ones.  So the right entry is "none documented," not "none happened."

Gains that cannot be priced

The candidate gains in Part 4, the Bmp6 teeth, the Fads2 copies and the snowballing haplotypes, happened before anyone was watching and in populations of unknown size.  They cannot be put on this ledger at all.  The series treats such gains as a reason the ledger is a lower bound on what evolution produced, which cuts against the young-earth side.  Here the effect is small, because none of the three adds unique functional sequence in the observed period, but the principle stands.

Unique functional sequence per organism-generation

This cell is left empty, for two reasons.  The numerator from the watched lakes is zero, but it is not a measured zero, for the reason above.  The inferred numerator, the sequence built into old haplotypes over a million years or more, has never been counted in functional nucleotides.  The EcoPeaks are regions tens of thousands of bases long, and the bases that matter inside them are known for only one, the Eda T to G.  No denominator is offered either.  The long-term effective size of sea-run sticklebacks is about 160,000 by the standard formula,^{23 26} but the census population is far larger and unmeasured.  An invented denominator would sink the argument faster than a missing one.

Part 6: What each side makes of it

The creationist and design reading

Creation Ministries International's 2009 article is the fullest.  Its central claim is that "the so-called “rapid evolution” did not result in the addition of any new features but rather the reduction or even complete loss of existing features—bony plates and pelvic spines."¹⁹  Answers in Genesis goes further in 2012:  "The “ancient genes” the fish seems to switch on and off are the genetic information God provided in order to enable this fish to adapt to various environments."²²  And in 2010 it used the pond experiments for the young-earth timescale:  "This three-year example of stickleback “evolution” is a soft echo of the adaptation and speciation that would have occurred in the centuries after the Ark landed."²⁷  The Institute for Creation Research's current model, Continuous Environmental Tracking, holds that "sensors trigger an organism’s pre-programmed logic to select pre-existing information designed to respond to the changed condition," and cites sticklebacks as a case of traits "produced “on the fly” by changes to the “instruction code”" of DNA.^{28 28}

On speed and on standing variation, this reading is right, and the mainstream papers say the same.  Lescak's team concluded that "most stickleback evolution in fresh water occurs within the first few decades."¹  Roberts Kingman's team calls "the ancient SGV" (standing genetic variation) "the primary genomic feature enabling extremely fast evolution."⁴  A post-Flood model needs freshwater forms to appear within centuries.  Sticklebacks show them appearing within decades.

The creationist sources also make errors a skeptic will find.  CMI attributes armour loss to Pitx1:  "Specifically, it affects expression of the Pitx gene."¹⁹  Armour is Eda;  Pitx1 is the pelvis.  CMI and Answers in Genesis both give 200 founders for Barrett's ponds;  the paper had 182.^{19 29}  Brian Thomas of ICR wrote that the sequences Jones compared "were very similar between all the fishes studied, showing that they have been resistant to mutational changes,"³⁰ when Jones had found the opposite.  In 2010 Thomas described the cold-tolerant fish as descendants of "rare individuals" and then called "the reference to "selection" superfluous."³¹  Survival of rare existing variants is what selection means.

The mainstream objection, at full strength

Standing variation is not free.  Every allele that a lake population reuses had to arise by mutation at some earlier time, in some population.  Schluter and Conte's transporter model explains how freshwater alleles are kept in the sea and carried into new lakes, and they say plainly that it "does not address where or when an advantageous mutation arose."²  Their answer to the "when" is that "the original low-armor mutation occurred long ago."²  The measured numbers back them.  The Eda haplotypes differ by more than a pair of populations dated at about 2 million years.  The habitat-linked regions are older than the genome around them.  A measured family mutation rate gives about 1.9 million generations for Eda.  And measuring the rate directly made stickleback dates older, not younger:  the family rate is "3.1 times smaller than the commonly used substitution rate," which moved the threespine-ninespine split from 5.6 to 17.5 million years.^{32 32}  This is the opposite of the human mitochondrial case, where family rates run faster than phylogenetic ones and creationists have used the gap to shorten dates.

On this view, "no new mutation in the lakes" is exactly what the mainstream predicts.  The new mutations happened over millions of years.  The lakes show only the harvest.

What the age depends on

The pedigree route to the Eda age rests on four inputs:  the difference between the haplotypes, 0.01938 per synonymous site;  the family mutation rate, 5.11 × 10⁻⁹;  the generation time, one to three years;  and the assumption that the difference built up by mutation after the haplotypes split.  The first two are measurements.  The third moves the answer threefold and no more.  The fourth is the whole question.

What a young-earth model has to claim

A model with about 4,400 years since the Flood has two options.

Option 1:  the difference arose by mutation, faster.  Producing a 1.9% synonymous difference between two lineages in 4,400 one-year generations needs a rate of about 2.2 × 10⁻⁶ per base per generation, 431 times the measured rate, and 1,293 times with three-year generations.¹²  Applied to the whole genome, that is about 2,000 new mutations per fish per generation instead of about 4.7.  No creationist source found proposes this.  The numbers are here for scale only.

Option 2:  the difference was created.  This is what Answers in Genesis says:  the ancient genes are "the genetic information God provided."²²  On this option, the created stickleback kind carried both marine and freshwater haplotypes from the start, and the ages are artefacts of reading created differences as accumulated mutations.  This option is coherent.  It is also untested.

What would test each mechanism

Continuous Environmental Tracking has been tested at Eda, and failed there.  If fish changed their own DNA "on the fly," genotypes would shift toward the freshwater allele within individuals or at conception.  Barrett's F1 fish were born at the Mendelian 1:2:1 ratio in all four ponds.¹³  The frequency then moved because some genotypes died faster than others.  Roberts Kingman's team found the low-armour phenotype tracking the square of the allele frequency, the ordinary signature of a recessive allele.⁴  CET may still apply to traits that are plastic.  Armour plates are not one of them.

Created versus accumulated differences can be told apart.  Mutations leave a fingerprint.  Some kinds of base change are far commoner than others, and Zhang's team has now measured the stickleback pattern in families.¹⁷  O'Brown's team noted that the Eda T to G is a transversion, "one of the least common types of changes seen in large scale studies of spontaneous germ-line mutations."⁶  If the thousands of differences between marine and freshwater haplotypes carry the same fingerprint as new mutations, that supports accumulation.  A created set of differences has no reason to carry it.  The public genomes from Jones and Roberts Kingman would allow the test.  No study found has run it, on either side.

The age of the reused variants can be checked against the kind.  If the stickleback family was created as one kind, front-loaded haplotypes should be shared across its genera.  Nelson and Cresko note that genetic mapping of armour traits points "either to independent genetic architectures in the two species" or to repeated mutation at one locus.  That leans against a shared created pool, though it rests on a small number of mapping studies.²³

Objections and responses

"Sticklebacks are evolution in action.  Why call it anything else?"  They are, in the population-genetic sense of a change in allele frequency.  Answers in Genesis concedes the definition:  "the stickleback population’s loss of armor was evolution, and now the reappearance of the armor is evolution."³³  What was watched is the spread of existing alleles.  It is not evidence for how those alleles began.

"The standing variation had to come from somewhere, and the dates say millions of years."  This is the strongest objection, and Part 6 states it at full strength.  The young-earth answer is that the variation was created.  That answer is coherent but has not been tested, and a test is available.  Until someone compares the marine-freshwater differences with the measured mutation fingerprint, this objection stands unanswered.

"The Pitx1 deletions show evolution happening by new mutation, right now."  They show new mutation happening easily at a fragile site, and every such mutation is a loss.  They were inferred from lake populations, not watched.  They fit the claim that losses are cheap and gains are expensive.  They do not show new function arising.

"Creationists say it is all loss, but more teeth and extra Fads2 copies are gains."  At the level of the fish, yes, and this article says so.  At the level of the molecule, one is a stronger existing switch and the other is a copy.  Neither adds unique functional sequence.  CMI's line that there is "no new genetic information here"¹⁹ is right about the Pitx1 enhancer and wrong as a description of stickleback adaptation as a whole.

"Barrett measured a selection coefficient of 0.52.  That is enormous."  It was one half of a pair.  The low allele was selected against at 0.50 earlier in life.  Across one generation the net change was small.  The large figure describes part of a life cycle, not a generation.

"Fast change within decades proves the young-earth timescale."  It shows the timescale is not needed for this change.  It says nothing about the age of the variants that changed, and the one measured clock available says those are old.  The speed of the lake change and the age of the alleles are separate questions.

"The mutation rate may have been higher in the past."  It may have varied, and Zhang's team says so.  But the required rate is 300 to 1,300 times today's.  At that rate each fish would carry about 2,000 new mutations per generation.  A proposal of that size needs its own evidence, and none has been offered.

"Old dates assume common descent."  The pedigree route assumes only that the two Eda haplotypes share an ancestor and that their differences arose by mutation.  Within one interbreeding species, common ancestry of two versions of one gene is not in dispute.  The second assumption is the real point at issue.

Issues and unanswered questions

• Loberg Lake population size and Bell 2004 full text. Only the abstract was read. The ledger uses Cheney and Scout founders instead.
• Ishikawa et al. 2019 (Science 364:886) not obtained. The Fads2 account rests on the authors' 2021 summary. Whether the extra copies sit under new control, which would make them a gain by copy rather than a modification, is not known here.
• The mutation-fingerprint test has not been run. It would decide between created and accumulated haplotype differences. It needs published genomes and the spectrum from Zhang 2025.
• What the 6% of fast-moving regions outside known old regions are. Roberts Kingman does not say they are new. Nor does it say they are not.
• Behe has not written on sticklebacks. Darwin Devolves does not mention them. The Behe scores in the table are this article's application of his published definitions.
• The generation time of sea-run sticklebacks. Zhang assumed three years. Lake fish often breed at one. The Eda age moves threefold with it.

Verdict

Sticklebacks are a strong case for rapid adaptation from standing variation, and a weak case for the origin of new function.  Everything watched happening was the rise of alleles already present.  In the watched lakes, no adaptive mutation was seen arising, though the studies could not have seen one.  Of the inferred changes, the pelvis was lost by deletions, the armour by a weakened switch, and the two gains of function are a stronger switch and a copied gene.  That record supports the claim that fast adaptation mostly spends existing information.

The young-earth side should not overclaim it.  The variants being spent look old by a measured mutation rate, not only by an assumed calibration.  The one published creationist answer, that the differences were created, is coherent.  It has not been put to the test that could separate it from accumulation.  Until it is, this is the main open weakness of the creation model of rapid speciation in its best example.

Sources