Darwin's Finches

"The temporal pattern of change shows that reversals in the direction of selection do not necessarily return a population to its earlier phenotypic state." — Peter and Rosemary Grant, Science, 2002¹

"A question for future research is whether such change requires a new mutation, or whether a shuffling of existing variation is all that is necessary to produce just the right combination of genes in the formation of a new species." — Peter and Rosemary Grant, 40 Years of Evolution, 2014²

Summary

Peter and Rosemary Grant caught, measured and followed nearly every finch on Daphne Major, a 40-hectare island in the Galápagos, from 1973 to 2012.  It is the best long record of natural selection in a wild animal.  This article reports what it measured, what the genes behind it turned out to be, and how both sides of the origins debate have used it.  Changes are scored with the method of Measuring Biological Information.

1. Selection was real, strong and heritable. In the 1977 drought about 85 percent of medium ground finches died. Birds with deeper beaks survived, and their offspring had deeper beaks. Beak size is about 80 percent heritable. The response in the next generation matched the breeder's equation.
2. It reversed, more than once. After the 1982-83 El Niño, small birds were favoured. In the 2004-05 drought, small-beaked birds survived again, this time because a larger competitor finch had eaten the big seeds. The strongest reversal was driven by a competitor, not by the weather.
3. The birds did not end where they began. Mean beak size in 2001 was the same as in 1973. Body size was smaller and beaks were more pointed. After 2005, beak depth was smaller than in 1973. So the critics who said the beaks "went back to normal" were right about beak size until 2004, and wrong about the rest.
4. Every gene variant behind the changes is older than the species that carry it. The two main beak loci, ALX1 for shape and HMGA2 for size, have haplotypes that the mainstream dates at about 0.9 to 1 million years. Genome sequencing of 3,955 Daphne birds found that forty years of change were shifts in the frequency of these old variants. No new mutation was reported to have contributed.
5. Hybridisation, not mutation, refills the variation. The Grants estimate that interbreeding adds three or more orders of magnitude more genetic variance each generation than new mutation does.
6. One hybrid lineage became reproductively isolated in three generations. A single immigrant large cactus finch arrived in 1981 and first bred with a local medium ground finch in 1983.³ His descendants, the "Big Bird" lineage, now breed only with each other, kept apart by song and size. That is a real observed step toward a species. It made no new gene; it mixed two existing ones.
7. The "new species in 200 years" line was an extrapolation with a stated assumption. Peter Grant's 1991 estimate assumed no selection between droughts, and the same article said the 1977 change had been "approximately offset." The National Academy of Sciences printed the estimate in 1998 and 1999 without that caveat. Phillip Johnson was right to call out the omission, and wrong that there was "no directional change of any kind."
8. Creationists accept all of this as variation within one created kind. Young-earth biologists place Darwin's finches in a sparrow-finch kind of more than 1,000 species. One of them, Robert Carter, concedes that the beak variants are mutations and not created diversity. The gene blocks that carry those variants differ by about 200 times more than 4,400 years of mutation supplies at the rate measured by comparing bird parents with their chicks. The young-earth model's published answer to gaps of this kind is created heterozygosity: the founders of each kind carried several versions of each gene. Nobody has yet applied that answer to the finch blocks or tested it on them.
9. The ledger is tiny. About 8,000 medium ground finches hatched on Daphne in forty years. They carried about 170,000 new mutations between them, against about three billion possible single-letter changes. The study could not have been expected to catch a useful new mutation, and did not.

Verdict.  Daphne Major is the cleanest measurement of natural selection in this series and one of its weakest cases for new information.  Nothing was watched that added a new gene or a new structure.  That fits the young-earth claim that fast change runs on existing variation.  It does not settle where that variation came from.  The mainstream answer, that it accumulated over about a million years, is a clock inference, but the clock's rate matches the one measured by comparing bird parents with their chicks.  The young-earth answer, that the founders were created with much of it, has not yet been tested on these genes.

Part 1: What was watched

The island and the birds

Daphne Major is a volcanic cone in the middle of the Galápagos, 0.34 square kilometres, about 40 hectares.⁴  Four ground finches live or visit there.  The two residents in 1973 were the medium ground finch, Geospiza fortis, a seed-eater of about 18 grams, and the cactus finch, G. scandens, of about 20 grams.  The large ground finch, G. magnirostris, of about 30 grams, bred there only from late 1982.⁴

The Grants' method was simple and exhausting.  They banded nearly every bird, measured six traits on each adult, recorded rain and seed supply, and followed who bred with whom.  By 1981 they had marked more than 90 percent of the two resident species.³  A generation is about 4.5 years for fortis and 5.5 for scandens.¹  Forty years is therefore about nine generations of fortis.

Two numbers set the scale.  Between 1976 and 1991 the harmonic mean breeding population was 197 fortis and 94 scandens, and the genetically effective population was a quarter to a half of that.²  The census swung from about 1,500 fortis in 1973 to 83 in 2005.^{5 4}  This is a population of hundreds, not of billions.

1977: the drought

Only 24 mm of rain fell in the 1977 wet season, against 127 mm in 1976 and 137 mm in 1978.  The medium ground finches did not breed at all, and the population fell by 85 percent.⁶  Of 388 nestlings banded in 1976, one survived to 1978.⁶

Deaths were not random.  As the birds ate out the small soft seeds, what was left were the hard woody fruits of caltrop, Tribulus cistoides, which only large beaks can crack.  In May 1976, 17 percent of feeding was on medium and large seeds;  a year later it was 49 percent.⁶  Boag and Grant reported that survivors differed from the dead mainly "in body size and bill dimensions, particularly bill depth."⁶  Jonathan Weiner, writing from their data, gave the means in millimetres:  "The average fortis beak before the drought was 10.68 millimeters long and 9.42 deep.  The average beak of the fortis that survived the drought was 11.07 millimeters long and 9.96 deep."⁵  That is 0.54 mm, or about 6 percent.

The change was heritable.  After correcting for the 20 percent of chicks fathered outside the pair, the Grants put heritability at 0.83 for beak size and 0.88 for beak shape, "exceptionally high values for wild bird populations."²  The 1978 generation was larger in all six traits, by 0.21 to 1.07 standard deviations.²  Weiner's figure for beak depth is "4 or 5 percent deeper than the beak of their ancestors before the drought."⁵

Boag and Grant already expected the other half of the story in 1981.  They wrote that "a different set of selection pressures operates when food is abundant and population size is increasing, thus giving rise to oscillating directional selection."⁶

1983 to 1986: the flood year and the reversal

In late 1982 an exceptionally strong El Niño reached the islands, and Daphne got 1,359 mm of rain.⁴  Small-seeded plants flourished.  In 1984 and 1985 small soft seeds made up 21 to 50 percent of seed biomass, against a previous maximum of 8 percent.²  When the dry years came, only 32 percent of adults survived from 1984 to 1987, and "small birds with small beaks had a strong selective advantage over the rest."²  Every selection differential was negative, where in 1976-77 every one had been positive.²  Lisle Gibbs and Peter Grant reported it in Nature in 1987 as "a reversal in the direction of selection following the opposite climatic extreme."⁷

2004 to 2005: the competitor

The third episode is the one that most complicates the textbook line "reversing with the weather."  The large ground finch had been breeding on Daphne since 1982 and reached 354 birds in 2003.  Little rain fell in 2003 (16 mm) and 2004 (25 mm).  Early in 2004 there were 150 large ground finches against 235 medium ones.  The large ground finches ate the caltrop that large-beaked medium ground finches had lived on in 1977.⁴

This time the small-beaked medium ground finches survived.  Average selection differentials over the six traits were 0.774 standard deviations for males and 0.649 for females, all against large size.⁴  The 1977 figures had been 0.642 and 0.668, all for large size.⁴  The rain was about the same both times.  The Grants named what was different:  "The conspicuous difference between these years was the number of G. magnirostris."⁴  The next generation's beaks were 0.70 standard deviations smaller than those of their parents before selection, inside the predicted range of 0.66 to 1.00.  "This is the strongest evolutionary change seen in the 33 years of the study."⁴  In their 2014 book, with offspring measured over 2006-08, the figure is 0.65 standard deviations.²

Did the finches end where they began?

The catalogue line for this article said the beaks were "reversing direction with the weather."  The record supports reversal.  It does not support a return to the start, and two of the three reversals were not about weather alone.

Grant and Grant compared the 1973 and 2001 samples directly.  Body size was smaller in both species.  "Mean beak size did not differ between years in either G. fortis (F1,166 = 0.004, P = 0.9480) or G. scandens," and "Beak shape differed between years in both species."¹  So for beak size alone, 1977's gain was cancelled by 2001.  Then the 2004-05 episode pushed beak size down past the starting value.  The Grants' summary of forty years is that "G. fortis became smaller over the whole period, and beaks became more pointed on average."²  Their figure of the whole community says that between 1973 and 2012 "fortis (F) became smaller in beak depth, and scandens (S) became shorter in beak length through a combination of selection and hybridization."²

Their own conclusion in 2002 was the sentence at the top of this article:  reversals "do not necessarily return a population to its earlier phenotypic state."¹  The honest short version is:  selection swung back and forth, beak size came back once and then overshot, and body size and shape drifted one way.  There was net change, and it was not in any one direction for forty years.

How often did it happen?  Body size was significantly selected in six of the years they could measure.²  Up to 2001 there were eight measurable evolutionary responses in fortis and seven in scandens.¹  Most years, selection was weak:  "usually less than 0.15 SD and rarely more than 0.50 SD," with medians of 0.03 to 0.06.¹  The famous episodes were the exceptions.

Part 2: The genes

What builds a finch beak

The first molecular work compared embryos of different species.  Arhat Abzhanov and colleagues found in 2004 that "the species with deeper, broader beaks relative to their length express Bmp4 in the mesenchyme of their beak prominences at higher levels and at earlier stages (a heterochronic shift)."  Adding extra BMP4 to chicken embryos made deeper, broader beaks.⁸  In 2006 the same group found that calmodulin, a calcium-signalling protein, "is expressed at higher levels in the long and pointed beaks of cactus finches," and raising it in chickens lengthened the upper beak.⁹  Both are ancient genes found in every vertebrate.  What differs among finch species is how much of each is made, and when.  Abzhanov wrote only that he was "tempted to speculate" that the cause lay in the gene's control regions.⁸

Genome sequencing then found the switches.  Sangeet Lamichhaney and colleagues sequenced 120 birds from every species in 2015.  They reported that "A 240 kilobase haplotype encompassing the ALX1 gene that encodes a transcription factor affecting craniofacial development is strongly associated with beak shape diversity across Darwin's finch species as well as within the medium ground finch (Geospiza fortis)."¹⁰  In 2016 they found a region of about 525 kb containing HMGA2, a growth gene, with two haplotypes tied to size.¹¹

A haplotype is a stretch of chromosome inherited as a block.  The "blunt" and "pointed" ALX1 blocks, and the "large" and "small" HMGA2 blocks, each differ at many sites.  Which of those differences does the work is not known.

The genes in the 2004-05 drought

The HMGA2 result is the only place in this story where selection was measured on a named gene in the wild.  The team typed 71 medium ground finches that lived through or died in the 2004-05 drought, 37 survivors and 34 dead.  Birds with two "large" copies died most.  The selection coefficient against them was 0.59, and the change in the gene's frequency "explains about 30% of the phenotypic shift in beak size due to natural selection."¹¹  The paper gives the error as ± 0.15 in one sentence and ± 0.14 in another.¹¹  A selection coefficient of 0.59 means those birds had 41 percent of the survival of the favoured type in that one episode.  The authors call it "higher than other examples of strong selection."¹¹

ALX1 did not matter that year, because beak shape was not selected:  "Survivors and nonsurvivors did not differ in beak shape; hence as expected ALX1 locus that regulates variation in beak shape (paper III) was not associated with survival," as Lamichhaney summarised in his thesis.¹²

Forty years in the genome

In 2023 Erik Enbody and colleagues sequenced 3,955 finches from Daphne, nearly every bird the Grants had blood from.  Six loci of large effect explain 45 percent of the variation in fortis beak size.  The largest, the HMGA2 region they call G03, explains 25 percent of beak size and 13 percent of body size, and "is most likely a supergene comprising four genes that contain multiple adaptive mutations."¹³  Over thirty years the frequencies at the six loci moved by 2 to 30 percent in fortis and 1 to 10 percent in scandens.¹³  In fortis, "Abrupt changes in allele frequencies at these loci in G. fortis resulted from strong natural selection during an extreme drought."¹³  In scandens the change was gradual and came from interbreeding with fortis.

Their one-sentence summary is exact:  "Allele frequencies at the loci fluctuate through directional natural selection and introgressive hybridization."¹³  The paper reports no mutation that arose during the study and contributed to the change.  It did not look for one, so the accurate statement is "none reported," not "none occurred."

How old the variants are

Every one of these variants is older than the species that carry it, by the mainstream clock.

The species themselves are younger.  Ground finches and tree finches split about 400,000 years ago,¹¹ and the medium, small and large ground finches later still.  So the haplotypes were already there when these species formed, and each species carries them in different proportions.  Carl-Johan Rubin and colleagues put it as a general rule:  "The reuse of ancestral genetic modules is a much faster route to adaptive change than the slow accumulation of adaptive de novo mutations."¹⁵

These ages are inferred, and how they were inferred matters.  The 2015 genome paper dated the branch points of the finch tree from the genome-wide differences between species, "assuming a mutation rate of 2.04 x 10-9 per base per year," after allowing for variation already present in the common ancestor.¹⁰  That rate came "from a comparison between a Darwin's finch and zebra finch," two lineages put about 25 million years apart, so it is a rate calibrated on an assumed date.^{10 16}  Mitochondrial DNA, dated with "the fossil-calibrated divergence rate 2.1% per million years for birds," gives similar dates, though it puts the deepest split at 1.4 million years against 0.9 million from the nuclear genome.^{10 10}  The HMGA2 age is not a separate calculation.  The two haplotypes split before the warbler finches branched off, and that branch point is the one dated at about a million years.¹¹  Rubin's 28 blocks were dated with the same rate, by dividing the differences between haplotypes by twice the rate.¹⁵

The rate can be checked against direct measurement, and it passes.  The 2015 paper's own table implies a generation of about five years, which turns 2.04 × 10⁻⁹ per year into about 1.02 × 10⁻⁸ per generation.¹⁰  Counting the new mutations in bird chicks that are absent from both parents gives an average of 1.01 × 10⁻⁸.^{17 18}  So the ages do not hang on a fossil calibration;  a measured rate gives the same answer.  What was watched is still only the frequency shift.  To make the age observed, someone would have had to watch the differences between the blocks accumulate.

The mainstream authors do not say the blocks were always there.  Rubin's team infers the opposite:  "The block structures at the 28 loci most likely reflect large-effect haplotypes composed of clusters of multiple causal variants that have accumulated during the evolution of Darwin's finches."¹⁵  That is, the blocks were built by mutation over the radiation, then sorted and swapped between species.  Nobody saw them being built.

Where the variation comes from

Heritability this high in populations of a few hundred birds is a puzzle.  The Grants note that it "should be in the region of 0.25 when mutation is the sole source of new genetic variation."²  Their answer is interbreeding.  Hybrids between fortis and its neighbours were rare but regular:  over twenty years, 45 fortis × small ground finch pairs and 8 fortis × cactus finch pairs, against 1,268 fortis pairs.²  From these they estimate that the genetic variance added by introgression "is three or more orders of magnitude greater than variation introduced by new mutations per generation."²  After the 1983 El Niño the hybrids were no longer at a disadvantage and some did better than the parents.²

So the fuel for forty years of selection came from old variants, kept alive and reshuffled by gene flow between species.  New mutation was not needed on this time scale, and nothing in the record shows it being used.

Part 3: Big Bird and the species count

One immigrant

In 1981 Trevor Price, netting young birds for the heritability study, caught a finch that did not belong.²  It weighed 29.7 grams, "more than 5 g heavier than any other G. fortis that had bred on the island."³  The Grants first took him for a hybrid medium ground finch from nearby Santa Cruz.  When his genome was sequenced in 2018, he turned out to be something else:  the founder male "was not a G. fortis × G. scandens hybrid as previously hypothesized (12), but rather a G. conirostris," a large cactus finch from Española, more than 100 km away.¹⁹

He bred with a local medium ground finch.  One of his sons bred with another.  After that, "all other matings occurred within this lineage, endogamously."¹⁹  A severe drought later cut the lineage to one brother and one sister, who bred with each other.³  All later generations descend from that pair.¹⁹  At its peak in 2010 the lineage had "eight breeding pairs and 36 individuals."¹⁹  The authors' headline:  "This example shows that reproductive isolation, which typically develops over hundreds of generations, can be established in only three."¹⁹

The lineage is visibly different.  Twenty of 24 measured members had wider beaks than any of the other 462 medium ground finches measured on the island from 2002 on.³  The males sing a song unlike that of the founder's home species, most likely because the founder copied a Daphne song imperfectly.¹⁹

What the isolation rests on

The isolation is by behaviour.  Finches choose mates by song, which sons learn from fathers, and by beak and body size, which they imprint on as chicks.  The 2018 paper names the barrier as "reproductive isolation based on learned song and morphology."¹⁹  No sterility or inviability is involved.  The 2009 paper said that an intrinsic barrier "is unlikely because it has not been detected among any of the six Geospiza species."³

Three limits follow from the authors' own words.  Isolation from the medium ground finch was observed.  Isolation from the founder's own species, the large cactus finch on Española, "is unknown because experiments have not been done there."¹⁹  The 2018 paper calls the lineage "an incipient hybrid species."¹⁹  And in 2009 the Grants wrote that "it is too early to tell whether reproductive isolation is transitory or likely to be enduring."³

They also wrote the sentence a young-earth reader will notice:  the observations "show how a barrier to interbreeding can arise behaviorally and without genetic change in sympatry."³  The barrier came from two existing species mixing, from song learning, and from inbreeding in a tiny population.  The general question of which speciation events have been watched is handled in Observed Speciation, which lists Big Bird as the one animal case followed from its first generation.

How many species are there?

Darwin's finches interbreed.  On Daphne about 2.8 percent of fortis males paired with small ground finches, and hybrids after 1983 survived as well as pure birds.^{2 2}  The 2015 genome paper found "extensive evidence for interspecific gene flow throughout the radiation" and that "Hybridization has given rise to species of mixed ancestry."¹⁰  The Grants' own summary is that "in this young radiation of Darwin's finches all species, apparently, still have their doors open."²

Two ornithologists took this further.  Bailey McKay and Robert Zink argued in 2015 that the six ground finches are one species caught in a cycle of divergence and collapse:  "the six putative ground finch species (genus Geospiza) of the Galápagos Islands represent a dramatic example of Sisyphean evolution that has been confused with the standard model of speciation."  They proposed a single species, "Darwin's ground finch (Geospiza magnirostris)."²⁰  Carlos Daniel Cadena and colleagues re-ran McKay and Zink's measurements with statistical mixture models and reported "negligible support for this claim relative to taxonomic hypotheses recognizing multiple species."²¹  The 2023 genome study takes a middle line:  the finches diverged ecologically "repeatedly and strongly but without the kind of genetic divergence that impairs fertility and viability."¹³

The genomes show how close the species are.  Two birds of different Darwin's finch species differ at about 2 to 3 DNA letters in every thousand.¹⁵  Two birds of the same species differ at up to about 2 in a thousand.  The 2015 genome paper's measure of diversity within a population, Watterson's θ, is 2.0 × 10⁻³ for the Daphne medium ground finch and runs from 0.3 to 2.2 × 10⁻³ across the finch populations sampled.  An outgroup population, the black-faced grassquit of Barbados, has 2.8 × 10⁻³, more than any finch population.¹⁰  So the species differ from one another only a little more than birds of one species differ among themselves.  Much of what separates them is old variation that they share and have sorted in different proportions, and the beak blocks are the clearest case:  "The origin of haplogroups at these loci, and possibly other loci that are associated with morphological variation, is older than the species splits that we have studied here."¹⁴  This is the pattern the one-kind reading expects.  It is also what the mainstream expects of a young radiation, so it does not decide between them.

The count today is 18 species, 17 in the Galápagos and one on Cocos Island.¹⁵  Textbooks of the 1990s said 13 or 14.  Whether the number is six, eighteen or one depends on how much interbreeding a "species" may have.  The birds are the same under every count.  What the count cannot do is carry the weight of "a new species in 200 years" without saying which definition it means.

Part 4: The radiation, which is inferred

Everything in this part is inference from genomes and arithmetic, not observation.

The molecular dates

The mainstream date for the whole radiation is about 1.5 million years:  "18 species of Darwin's finches evolved from a common ancestor that colonized the Galápagos ∼1.5 million years ago."¹⁴  Peter Grant's 1991 range was wider:  "the past one to five million years."²²  The dates rest on sequence differences divided by a rate, which matches the rate measured in bird pedigrees (Part 2).¹⁰  To observe a radiation, someone would have to watch one lineage become several over many thousands of generations.  Nobody has watched that happen in finches.

The 200-year extrapolation

The famous number comes from Peter Grant's 1991 article in Scientific American.  He and Trevor Price took the heritabilities and the strength of the 1977 selection and asked how many such events would turn a medium ground finch into a large ground finch, which is about 50 percent larger:

"The number is surprisingly small: about 20 selection events would have sufficed.  If droughts occur once a decade, on average, repeated directional selection at this rate with no selection in between droughts would transform one species into another within 200 years.  Even if the estimate is off by a factor of 10, the 2,000 years required for speciation is still very little time in relation to the hundreds of thousands of years the finches have been in the archipelago."²²

Weiner gives the same calculation, and adds that starting from a larger-beaked island "the change would take only a dozen droughts."⁵  Changing shape rather than size is slower:  turning a medium ground finch into a cactus finch "would take six times longer."²²

The calculation assumes three things, and the Grants' own data test each of them.

1. Every drought pushes the same way. The 2004-05 drought pushed the opposite way from 1977, because a competitor had arrived.⁴
2. Nothing pushes back between droughts. The 1984-85 episode did push back, and on the next page of the same 1991 article Grant wrote: "Effects of the droughts of 1977 and 1982 were approximately offset by selection in the opposite direction— toward smaller body size—in 1984-85."²²
3. A larger beak is a new species. A medium ground finch with a large ground finch's beak would still interbreed with its neighbours unless song and imprinting kept it apart. Big Bird shows that this can happen. It also shows that it depends on chance events, not on beak size alone.

Grant knew all this in 1991.  He wrote:  "Whether or not there is a net directional trend toward larger size, like an arrow through the oscillations, is unclear and could be determined only by a much longer study."²²  Twenty more years of study gave the answer:  there was no arrow for beak size.  The extrapolation shows that selection of the measured strength is strong enough to make species-sized differences quickly.  It does not show that it did.

The rate matters for both sides, and it points the same way for both.  If 20 strong droughts in one direction are enough, then a million years is far more than the finches needed.  The observed problem is not speed.  It is direction.

Part 5: The accusation

Darwin barely used the finches

The popular story says the finches inspired Darwin.  The historian Frank Sulloway showed in 1982 that this is a legend.  In his voyage notes Darwin recorded the gradation of beak shapes, in Sulloway's words, "almost as an afterthought," and offered no explanation for it.²³  He became an evolutionist only in 1837, after the voyage.  Sulloway's verdict:  "Contrary to the legend, Darwin's finches do not appear to have inspired his earliest theoretical views on evolution, even after he finally became an evolutionist in 1837; rather it was his evolutionary views that allowed him, retrospectively, to understand the complex case of the finches."²³  He also noted that "In the Origin of Species (1859) the Geospizinae go unmentioned."²³

This is a point about history, not biology.  It matters only to textbooks that still tell the legend.  The modern case rests on the Grants' measurements, which owe nothing to Darwin's notebooks.

What the National Academy printed

In 1998 the National Academy of Sciences published Teaching About Evolution and the Nature of Science for schoolteachers.  A boxed feature on page 19, "Ongoing Evolution Among Darwin's Finches," read:

"A particularly interesting example of contemporary evolution involves the 13 species of finches studied by Darwin on the Galapagos Islands, now known as Darwin's finches.  A research group led by Peter and Rosemary Grant of Princeton University has shown that a single year of drought on the islands can drive evolutionary changes in the finches.  Drought diminishes supplies of easily cracked nuts but permits the survival of plants that produce larger, tougher nuts.  Drought thus favors birds with strong, wide beaks that can break these tougher seeds, producing populations of birds with these traits.  The Grants have estimated that if droughts occur about once every 10 years on the islands, a new species of finch might arise in only about 200 years."²⁴

Its source note for the 200 years is Grant's 1991 article.²⁴  The box leaves out the two sentences on the following page of that article:  that the 1977 and 1982 effects "were approximately offset" by 1984-85, and that a net trend "is unclear."^{22 22}  It also leaves out Grant's stated assumption, "with no selection in between droughts."²²

The 1999 second edition of Science and Creationism reused the text almost word for word, with one change.  "A particularly interesting example of contemporary evolution" became "A particularly compelling example of speciation."²⁵  The rest, including the 200 years, is the same.²⁵  What the Grants had watched was a change in the average beak of one species.  Calling it speciation claimed more than the data showed.

Johnson and Wells, in their own words

Phillip Johnson, a Berkeley law professor and a founder of the intelligent-design movement, wrote in the Wall Street Journal on 16 August 1999:

"In 1977 a drought killed most of the finches, and the survivors had beaks slightly larger than before.  The probable explanation was that larger-beaked birds had an advantage in eating the last tough seeds that remained.  A few years later there was a flood, and after that the beak size went back to normal.  Nothing new had appeared, and there was no directional change of any kind."²⁶

He went on:  "To make the story look better, the National Academy of Sciences removed some facts in its 1998 booklet on Teaching About Evolution and the Nature of Science.  This version omits the flood year return-to-normal and encourages teachers to speculate that a "new species of finch" might arise in 200 years if the initial trend towards increased beak size continued indefinitely."  He compared it to "the sort of distortion that would land a stock promoter in jail."²⁶

Jonathan Wells, a biologist with a Berkeley PhD in molecular and cell biology, made the finches chapter 8 of Icons of Evolution in 2000.  His summary:  "The natural selection observed in the 1970s reversed direction in the following years, so there was no net evolutionary change."²⁷  He granted the core result:  the drought response "is direct evidence for natural selection in the wild.  In this limited sense, the finches provide evidence for Darwin's theory."²⁷  And he quoted the 1999 booklet's "a particularly compelling example" as a case of exaggeration.²⁷

The mainstream reply

The National Center for Science Education answered Wells in 2003, chapter by chapter, in a reply written by Alan Gishlick.  On the reversal, he wrote that oscillation is what evolution predicts:  "Evolutionary theory would predict that if climate oscillates, morphology would oscillate as well.  The finches fit the predicted pattern."  On the 200 years, he defended extrapolation as ordinary science, like working out a comet's orbit from a few sightings:  "In the case of the finches, climate varied and the extrapolations changed."²⁸  He added that "no one would expect speciation to occur on that scale, and the Grants never claimed to expect it either," and that "Whether the species are merging or diverging is unimportant because both divergence and merging are forms of long-term evolutionary change."²⁸

The reply is right on its main point:  selection that reverses with the weather is still selection, and Wells granted as much.  It never mentions the National Academy booklets or the 200 years, which were the centre of Johnson's and Wells's complaint.  It accepts Wells's premise, writing "just as the finch beaks return to pre-drought levels," which the later data overturned for body size and beak shape.  It says that in wet years "the beak size is more varied," where the Grants measured selection for smaller birds in 1984-85.^{28 22}  And its comet analogy cuts against the booklets.  An orbit can be projected because gravity does not change, and Grant's own article had said the forces on the finches did.

Which parts were fair

The critics were writing in 1999 and 2000 with the data then published.  The two decades since have made Johnson's "back to normal" wrong for beak size and have confirmed his "nothing new" for genes.

Part 6: The creationist reading

In their own words

Young-earth creationists do not deny any of the Daphne results.  They claim them.  Carl Wieland wrote in Creation magazine in 1992 that Darwin's explanation, that the finches "are all the offspring of an original pair of finches, and that natural selection is responsible for the differences," is "the explanation now held by most modern creationists."²⁹  He read Grant's rates as support:  "it is real, observed evidence that such (downhill) adaptive formation of several species from the one created kind can easily take place in a few centuries.  It doesn't need millions of years."²⁹

Jean Lightner's baraminology study of 2010 set the size of the kind.  From recorded hybrids between bird families, she found "a large sparrow-finch monobaramin that includes over 1,000 species," which includes the tanager family where Darwin's finches now sit.  She dated its diversification to after "the severe genetic bottleneck at the time of the Flood less than 5,000 years ago."³⁰  She also wrote that "much more work is necessary to identify the genetic and environmental mechanisms responsible for generating this tremendous intrabaraminic diversity from one Ark kind."³⁰

The three main organisations say the same thing in different words.  Answers in Genesis:  "As creationists, we fully accept the fact that adaptation/natural selection can occur rapidly.  In fact, such processes (and perhaps other genetic factors) would have occurred rapidly after the Flood, producing variation within the animal kinds."³¹  The Institute for Creation Research, on Big Bird:  "It only took three generations to establish this new finch species," and "the new finch species remains fully finch."³²  Creation Ministries International, in 2023, reviewing the genome work:  "all finches and sparrows appear to belong to one baramin."³³

On where the variation came from, the creationist writers split.  Wieland (1992) and Jonathan Sarfati (2002) said the changes were "choosing from what is already there" and "selection of existing genetic information."^{29 34}  Brian Thomas of ICR called the standing variation "a design element."³²  Robert Carter, a biologist writing for CMI in 2023, said the opposite about the beak genes themselves:  "When a genetic variant only appears in one animal, one species, or a small group of species within a baramin, we can be fairly certain that this is a new mutation, not created diversity.  Thus, the genetic variants that affect beak size and shape among these birds are clearly mutations."³³

Errors in the creationist sources

Checked against the papers they cite:

• Wieland (1992) and Sarfati (2002) cite Grant's Scientific American article as pages 60-65. It is on pages 82-87.^{29 22} Wieland, like the National Academy, reports the 200 years without Grant's caveat that the 1977 change had been "approximately offset." Sarfati does mention the reversal.³⁴
• Georgia Purdom of Answers in Genesis wrote in 2006 that the opposite direction of selection in 1977 and 2004 was something "the Grants do not have an explanation" for.³¹ The paper she was answering gives the explanation in one line: the difference was the large ground finch.⁴ She also suggested the large ground finch "had lost the genetic information to make a smaller beak" because its beak did not change. The paper reports that its mortality "was apparently not selective," so no change was expected.^{31 4}
• Thomas (2017) describes the Big Bird founder as "a male hybrid between Geospiza fortis and G. scandens, as later determined by genetic testing."³² The genetic testing in the paper he cites found that he was not that hybrid, but a large cactus finch.¹⁹ Carter (2023) repeats a version of the same error and says the new species appeared "in a single breeding season." The paper says three generations.^{33 19}

None of these errors changes the creationist argument.  They show that the organisations' articles were written from press reports and abstracts more than from the papers.

What each side gets right

The creationist writers are right that the whole Daphne record is change within a group of birds that interbreed, that it ran on variation already present, and that it made no new gene or organ.  The mainstream papers say the same:  Enbody's "allele frequencies ... fluctuate," Rubin's "ancestral genetic modules," the Grants' "without genetic change."^{13 15 3}  They are right that speciation, in the sense of a new breeding group, is fast and is part of their model.

The mainstream side is right that selection is strong, heritable and predictable from the breeder's equation;  that the variants selected are old by any clock that uses a measured mutation rate;  and that a lineage can become reproductively isolated in a few generations.  It is also right that the finch record, taken alone, is no argument against common descent of the finches.  The creationists agree:  they put the finches in one kind.

Does a young earth need faster change here?

For phenotype, no.  The measured episodes are more than fast enough.  If twenty 1977-strength droughts would turn a medium ground finch into a large one, then 4,400 years holds about 440 droughts at one a decade, the rate Grant assumed.  The young-earth model needs direction, as the mainstream one does.

For species, not by any measure that has been observed.  Lightner's sparrow-finch kind holds more than 1,000 species, all descended, on her reading, from the birds on the Ark.³⁰  Going from one lineage to 1,000 takes about ten doublings, so the model needs each lineage to split about once every 440 years, or about 90 finch generations.³⁵  The Big Bird lineage became isolated from its neighbours in three generations, and Grant's own estimate for turning one finch species into another by selection was 200 to 2,000 years.^{19 22}  The young-earth timetable sits inside that range.  The skeptic has two replies.  Grant's estimate assumed selection kept one direction, and Part 4 shows it did not.  And forming a breeding group is not keeping it.  McKay and Zink's "Sisyphean" cycle and the Grants' "doors open" describe species that merge again.^{20 2}  The young-earth model needs most of its new groups to last, and nobody has shown whether they would.  The mainstream's slower speciation rate comes from its molecular dates, which are the next problem.

For genotype, yes, and by a large factor.  The mainstream ages of the HMGA2 and ALX1 haplotypes, about 0.9 to 1 million years, rest on the mutation rate checked in Part 2, which matches the rate measured in bird pedigrees.  In 4,400 years at that rate, two copies of a stretch of DNA would come to differ at about 2 letters in 100,000.  The beak blocks differ by far more.  The HMGA2 region alone holds 1,327 sites where large and small birds differ strongly, against about 9 differences that 4,400 years of mutation would put in a stretch that long.^{11 36}  By the dates, the shortfall is a factor of about 200 to 230;  by the count of strongly differing sites alone, which leaves out weaker differences, it is at least 140.  The problem is not special to the beak genes.  The whole genomes of different finch species differ at 2 to 3 letters per thousand, about 110 to 170 times what 4,400 years supplies.^{15 36}

The young-earth literature has a published answer to gaps of this kind.  Nathaniel Jeanson and Jason Lisle argued in 2016 that comparing mitochondrial and nuclear clocks "necessitates the existence of created nuclear DNA heterozygosity within the 'kinds' of the Creation week."³⁷  On their model, "diploid individuals were created heterozygous," and later processes "distributed and/or added to the original created genetic diversity."³⁷  A kind on the Ark could have had up to 14 individuals and so up to 28 versions of each gene.³⁷  Two deep haplotypes at a beak locus are well within that.

That answer does not contradict Carter as directly as it first seems.  His sentence is about "the genetic variants that affect beak size and shape," the few letters that do the work.³³  Lamichhaney's team found that only one of the 1,327 strongly differing HMGA2 sites changes a protein, so the working changes are probably a few switches in gene control.¹¹  The clock counts the thousands of other differences.  A model in which the blocks' backgrounds were created and a few beak-changing mutations arose on them after the Flood fits both Carter's words and the measured rate.  No creationist has written this out for the finches, so it is a reading the model allows, not a claim anyone has defended in print.

The skeptic has a strong reply.  Most of the letters inside the blocks look new, not inherited from the kind.  At the most differentiated sites of the 28 size blocks, 67 percent of the large ground finch's main alleles, 1,273 of 1,914, are "derived relative to outgroups," that is, absent from two related Caribbean birds, the Lesser Antillean bullfinch and the black-faced grassquit.¹⁵  At HMGA2, the small-bird haplotype carries the derived letter at 13 of 17 differing sites in highly conserved DNA.¹¹  By Carter's own rule, letters found only in a small group within a kind are mutations, and these are letters throughout the blocks, not only the few that do the work.  A created-heterozygosity answer has to explain why the kind's other members lack them.  Loss by drift in the outgroup lineages could do it, and two outgroups out of a thousand species is a small sample, but nobody has shown it.  The other ways out are a mutation rate about 200 times higher after the Flood than any measured in birds, or fewer differences than the papers report, and the differences are counted, not inferred.  This remains the clearest open problem for the young-earth reading of Darwin's finches.  It now has a test:  if the beak blocks are created diversity, versions of them should turn up in other members of the sparrow-finch kind.

Part 7: How it scores, and the ledger

Scoring each change

The procedure of Measuring Biological Information asks, in order:  did the sequence come from another lineage (transfer);  did a working element stop working (loss);  does something now work that did not exist in this genome (gain);  otherwise, modification.  A change with no mutation behind it, such as a shift in the frequency of alleles already present, is not scored.  The same rule was used for the Cichlid Radiations and Sticklebacks.

No row is a gain.  No row is a loss either.  The ALX1 and HMGA2 cells are left empty because nobody knows which of the many differences within each block do the work, so there is no honest count of the nucleotides that matter.

The ledger

The series asks how many new mutations a population explored while it was watched, and what it bought.³⁸

The study explored about 170,000 new mutations, so any particular single-letter change had about one chance in 19,000 of turning up even once.³⁸  The finches could not have been expected to find and spread a new beneficial mutation in that time, and the study was not designed to look for one.  This row belongs beside the stickleback lakes in Benchmarking Evolution:  about the same number of new mutations, and the same result, fast adaptation from old variants.

That cuts both ways.  The finches are no evidence that new information is hard to make, because the population was far too small to test it.  They are strong evidence that visible, heritable, adaptive change in an animal does not need new mutations on a human time scale.  For the benchmarking argument, which is about how many births it takes to find a new function, Daphne Major tells almost nothing.  For the young-earth claim that fast adaptation runs on variation already present, it is the best field case there is.

Gains that cannot be priced

One change here matters and has no nucleotide count:  the Big Bird lineage's isolation.  A new breeding group formed, and the ledger reads zero, because it came from mixing two species and from learned song.  The Observed Speciation article makes the same point about polyploid plants.  The count of new sequence is a lower bound on what evolution produces.  That weakens the young-earth side:  a zero on this ledger does not mean nothing biologically new happened.

Objections and responses

"The finches went back to normal, so nothing evolved."  Beak size came back by 2001, then went below its 1973 value after 2005.  Body size fell and beak shape became more pointed, and neither returned.^{1 2}  The gene frequencies changed and were measured.¹³  Evolution in the sense of heritable change in a population happened, in more than one direction.

"Selection only sorts what is there;  it cannot create."  That is an accurate description of Daphne Major.  It is not a finding that new variation cannot arise, because the population was far too small to test that (Part 7).  The mainstream view is that the old haplotypes were themselves built by mutation over a million years.¹⁵  That is inferred, and so is the young-earth alternative.

"The Grants proved a new species can form in 200 years."  They estimated that one could, if drought-driven selection kept one direction with nothing pushing back.²²  Forty years of data showed it did not keep one direction.²²  The estimate shows the potential speed of selection, not a finished species.

"Big Bird is a new species, so macroevolution has been observed."  It is a new breeding group, isolated by learned song and size, from two existing species, and its isolation from one parent species is untested.¹⁹  It shows how quickly a barrier can form.  It made no new gene.  The creationist organisations accept it as a new species within a kind.³²

"The haplotype ages assume a million-year clock, so they beg the question."  The ages use a rate, and the rate can be checked.  The one the papers used is within about 10 percent of the rate counted directly in bird parents and chicks (Part 2).¹⁸  The differences between the blocks are counted, not assumed:  1,327 strongly differing sites in the HMGA2 block alone, against about 9 that 4,400 years of mutation would supply.^{11 36}  A young-earth answer has to say where those differences came from.  Created heterozygosity is the model's published answer, and it fits Carter's claim if only the few working letters are mutations.^{37 33}  But most letters in the blocks are absent from the two outgroups studied, which by Carter's rule makes them mutations.  Until versions of the blocks turn up elsewhere in the kind, the answer is untested (Part 6).¹⁵

"A thousand species from one pair in 4,400 years is impossibly fast."  It needs each lineage to split about once every 440 years, about 90 finch generations.³⁵  The Big Bird lineage became isolated in three generations, and Grant estimated 200 to 2,000 years for selection to turn one finch species into another.^{19 22}  The speed of isolation is not the young-earth model's problem.  Whether new groups last, and where their genetic differences came from, are (Part 6).

"The finches hybridise, so they are one species, and there was never a radiation."  McKay and Zink argued this for the ground finches;  a reanalysis of their data found clear support for several species.^{20 21}  Either way the birds differ in beak and body as they do.  The species count changes the label, not the biology.

"Darwin was inspired by the finches, so they are the heart of evolutionary theory."  He was not, and they are not in the Origin.^{23 23}  Their modern value comes from the Grants' fieldwork.

Issues and unanswered questions

1. The main text of Lamichhaney et al. 2015 was not obtained; its supplementary text and Extended Data tables were. The full texts of Gibbs and Grant 1987, Abzhanov et al. 2006 and McKay and Zink 2015 were not obtained. The ALX1 age is taken from the same group's later paper and the first author's thesis.
2. The mutation rate behind the ages has been read, and it matches measured bird rates (Part 2). What is still missing is a full count of the differences between the two HMGA2 haplotypes, rather than only the 1,327 strongly differing sites, and any count at all for the ALX1 block. Those counts would turn the factors in Part 6 into direct measurements.
3. Nobody has looked for the HMGA2 or ALX1 haplotypes, or close versions of them, in other members of Lightner's sparrow-finch kind beyond the two outgroups. That search would test the created-heterozygosity reading in Part 6 directly: shared deep haplotypes would support it, and their absence across many relatives would count against it.
4. The Grants rank the 1984-85 episode differently in 2006 ("weaker") and 2014 (one of the two strongest).⁴ The magnitude of that reversal was not read in Gibbs and Grant 1987.
5. Nobody has searched the 3,955 Daphne genomes for new mutations that arose during the study and had an effect. Until someone does, "none reported" is the honest phrase.
6. The birth count in the ledger is scaled from 21 seasons and is a floor.

Verdict

Darwin's finches on Daphne Major are the cleanest field measurement of natural selection in this series.  Selection was strong, heritable, repeated and predictable, and it changed the birds in both directions and left them different from where they began.  It is also one of the series' weakest cases for new biological information.  Every change that was watched was a shift in the frequency of old variants or a transfer between species, and no new gene or structure was reported.  Critics were right that the National Academy dropped Grant's own caveat, and wrong that nothing changed.  Young-earth writers are right that this is fast variation within one kind, and the first step of speciation, a new breeding group, took three generations here, faster than their model needs.  Their model still owes an account of the beak haplotypes, which differ about 200 times more than 4,400 years of mutation at the measured bird rate supplies.  Its published answer, created heterozygosity, fits if only the few working letters are mutations, as Carter's words allow.  It has not been tested on these genes, and the first outgroup comparisons lean against it.

Sources