The Origin of Multicellularity

"we have shown that the first crucial steps in the transition from unicellularity to multicellularity can evolve remarkably quickly under appropriate selective conditions"
William Ratcliff, Ford Denison, Mark Borrello and Michael Travisano, PNAS, 2012¹

"while simple multicellularity may be necessary for the evolution of complex multicellular organisms, it is not sufficient."
Lanyun Miao, Zongjun Yin, Andrew Knoll, Yuangao Qu and Maoyan Zhu, Science Advances, 2024²

Summary

Plants, animals, fungi and seaweeds are built of many cells that divide the work between them.  The standard account says they descend from single cells, and that the change happened many times.  This article splits that change into three steps and asks, for each one, what has been observed, what is inferred, and what is assumed.  The first step is cells that stay stuck together after they divide.  The second is a division of labour, where some cells give up reproducing so that others can, as a germ line.  The third is a body whose cells take on many different jobs under the control of networks of regulatory genes.  How animal body plans and nervous systems arose is a separate question, covered in The Origin of Nervous Systems and Animal Body Plans, and the timing of the first animals belongs to The Cambrian Explosion.

• Creationist and design writers have answered mostly the yeast and algae experiments. Michael Behe called it "the loss of two pre-existing abilities" and predicted that "functioning genes or regulatory regions were broken or degraded."³ Answers in Genesis wrote that "the information was in the genome all along."⁴ Shaun Doyle of Creation Ministries International argued that true multicellularity "requires the ex nihilo creation of an entirely new system of organisation."⁵ Ann Gauger of the Discovery Institute argued that the twelve-step story told for Volvox leaves out "the details necessary for this 12-step progression to occur."⁶ Jeffrey Tomkins of the Institute for Creation Research called the 2019 algae clusters "nothing more than a crude clumping together of individual cells."⁷ Doyle and Matthew Cserhati, also of Creation Ministries International, read the yeast and the algae as switching on "latent information" or "a pre-existing genetic mechanism."^{8 9} No response was found to the paper that identified the gene behind snowflake yeast, or to the genomes of the choanoflagellates, the single-celled relatives of animals.
• The first step happens in the laboratory, fast. Selecting brewer's yeast for settling quickly to the bottom of a tube produced clusters of attached cells, called snowflake yeast, in all 15 populations within 60 days.¹ The single-celled alga Chlamydomonas formed heritable clusters in 1 of 10 populations selected for settling,¹⁰ in 2 of 5 populations exposed to a predator,¹¹ and in 4 of 12 predator populations in a later run.¹²
• It came from breaking a gene, as Behe predicted. In five of ten populations checked, the snowflake form came from damaged copies of one gene, ACE2, which normally lets mother and daughter cells separate. Deleting it from the ancestor made snowflakes, and adding back one working copy turned an evolved strain single-celled again.¹³ In a second laboratory, ACE2 was mutated in 8 of 12 clumping clones, six of them by a mutation that cuts the protein short.¹⁴ Answers in Genesis was wrong, though, to say the ancestral yeast already formed snowflakes before the experiment. The paper reports the opposite.
• The clusters keep evolving, and stay undifferentiated. After 600 daily rounds of selection, about 3,000 generations, snowflake yeast grew from about 100 cells per cluster to about 450,000, more than a millimetre across.¹⁵ They got there with longer cells, doubled genomes and extra chromosomes. The authors still call snowflake yeast "a model of undifferentiated multicellularity."¹⁵
• No experiment has produced specialised cell types. The nearest thing is a minority of yeast cells, never more than 2%, that kill themselves at the points where a cluster breaks apart. The authors call this "functionally analogous to germ-soma differentiation," not differentiation itself.¹
• The experiments are too short to test the hard step either way. A snowflake population tries every possible single-letter mutation several times in two months, so a one-step loss like ACE2 is found at once. But on the mainstream timeline, the algae that came to have a germ line had roughly one million to ten million times as many generations as the longest experiment.¹⁶ The laboratory shows that the first step is cheap. It cannot show that the second is possible, and its failure to produce it shows nothing against it.
• Volvox was built mostly from old genes. Volvox carteri, a ball of about 2,000 body cells and 16 reproductive cells,¹⁷ has almost the same set of protein-coding genes as single-celled Chlamydomonas.¹⁷ Two Volvox genes needed for its development can be swapped for their Chlamydomonas versions and still work.¹⁷ One cannot: regA, the gene that keeps body cells from reproducing, "encodes an entirely novel combination of pre-existing and new protein domains."¹⁷ Among the related algae, sterile body cells arose "by different mechanisms in separate groups."¹⁸ The dates for these steps come from molecular clocks with borrowed calibrations, and published estimates differ by over 150 million years.
• Choanoflagellates carry animal parts, not the animal toolkit. The choanoflagellate Monosiga has at least 23 genes with cadherin domains, a kind of adhesion domain once thought to be animal-only, but no receptors or signals from three main animal signalling pathways, and it lacks many of the animal transcription factors.¹⁹ Mainstream authors conclude that animals needed new genes as well as reused ones.¹⁷
• Multicellularity is counted as arising many times, and complexity few. One review finds 45 separate lineages with simple multicellularity and expects the true total to be near a hundred.²⁰ Complex multicellularity, with communication between cells and tissues, is counted as arising "only six to seven times."² Mainstream authors read the many origins as showing the first step is easy. Design writers can read them as reuse of a common plan. Neither reading tests the hard step.
• The rocks give an order, not a pathway. Simple multicellular filaments appear in rocks dated about 1.6 billion years,² and algae with distinct holdfast and reproductive cells appear in rocks dated about a billion years,^{21 22} below the first animals. No fossil shows a population changing from one grade to the next.

The verdict.  Three questions get three answers.  That single cells can evolve into clusters that live and reproduce as units is demonstrated, and creationists who called it breakage were right about the gene and wrong if they meant nothing new arose.  That a cluster can evolve a germ line and a sterile body is unfinished on both sides:  living algae show a plausible series built mostly from reused genes, but the one new gene and the regulatory changes have not been watched or priced, and no experiment has run long enough to try.  That such a step leads on to bodies with many cell types under regulatory networks is undemonstrated, and the creationist argument is right that nobody has shown how it happened.

Part 1: The claim, and what creationists and design writers say

The claim.  Mainstream biology treats multicellularity as a series of grades.  Andrew Knoll and David Hewitt of Harvard put the central distinction this way:  "multicellularity per se, the transfer of individuality from one cell to a coordinated group of attached cells (Michod, 2007), was necessary for the emergence of complex multicellularity, but it was not sufficient."²³  Simple multicellular organisms are filaments, sheets, balls and colonies whose cells stick together but communicate little and differ little.  Complex multicellular organisms, in Knoll and Hewitt's words, "show evidence of cell-cell adhesion, intercellular communication (e.g., gap junctions, plasmodesmata), and cellular and tissue differentiation mediated by networks of regulatory genes."²³  The proposed route runs from single cells, to clusters, to clusters with a division of labour, to bodies.  Ratcliff and his colleagues, whose experiments are the centre of this article, write that the known transitions to complex multicellularity "occurred over millions of years" and that their own results show only "the first crucial steps."¹

Creationist and intelligent-design positions fall into four groups.  They are given here in their authors' words, and each is tested in the parts that follow.

1.  The yeast experiment shows loss, not gain.  Michael Behe, the Lehigh University biochemist, wrote on his blog in January 2012 about the first snowflake-yeast paper:  "In fact, the results can be regarded as the loss of two pre-existing abilities: 1) the loss of the ability to separate from the mother cell during cell division; and 2) the loss of control of apoptosis.  The authors did not analyze the genetic changes that occurred in the cells, but I strongly suspect that if and when they do, they’ll discover that functioning genes or regulatory regions were broken or degraded."³  He closed:  "The apparently insurmountable problem for Darwinism is to build new systems."³  Apoptosis is programmed cell death, a cell killing itself on cue.

2.  Nothing new was made, because the ability was already there.  An unsigned Answers in Genesis news item of July 2011, written after the first conference report of the experiment, said:  "Thus, genetic information to become multicellular did not evolve in Ratcliff’s laboratory; the information was in the genome all along."⁴  It added:  "They had not evolved anything new genetically.  They had been selected for abilities they already had."⁴  Elizabeth Mitchell of Answers in Genesis, reviewing the published paper in January 2012, went further:  "They demonstrated the yeast had this characteristic behavior before beginning the experiment."²⁴  Her conclusion was that the yeast "did not evolve into anything new, just fluffier yeast."²⁴  Shaun Doyle of Creation Ministries International made the same case in 2013, in a page later revised to cover the long-term experiment:  "This is not the evolution of something completely new, but the activation of latent information that enables the yeast to survive in certain environments."⁸  Because only the oxygen-free populations grew large, he argued, "This suggests the changes were non-random and were part of a system designed to facilitate adaptation to new situations."⁸  Matthew Cserhati, writing in Journal of Creation in 2019 on the predator experiment in algae, concluded that "a pre-existing genetic mechanism had already been in place to adapt the algae to environmental cues, such as predation," and argued that the change came "much too rapidly for any kind of genetic mutation to cause the differences in phenotype."^{9 9}  He also placed Chlamydomonas, Gonium and Volvox in one created kind:  "The normally unicellular Chlamydomonas reinhardtii as well as the obligately multicellular Volvox carteri are both members of the same baramin."⁹

3.  Clumps and colonies are not true multicellularity.  Shaun Doyle, writing in Creation Ministries International's Journal of Creation in 2009, before the yeast work, set out requirements.  "The first requirement for multicellularity to emerge is that all the cells must contain the same genetic information."⁵  Beyond that, "Evolution requires more than a mere augmentation of an existing system for co-ordinated multicellularity to evolve; it requires the ex nihilo creation of an entirely new system of organisation to co-ordinate cells appropriately to form a multicellular individual."⁵  He held that programmed cell death is needed and "arguably irreducibly complex."⁵  He granted that Volvox shows "some specialisation" but argued that volvocine algae "cannot create new somatic cells, and are as a result unable to survive for very long as an organism."⁵  A 2016 post on the Discovery Institute's Evolution News made the same point about the colonial alga Gonium:  "Gonium pectorale is not a multicellular organism.  Rather, each cell is an identical organism and they just live together in colonies."²⁵  Jeffrey Tomkins of the Institute for Creation Research said the same of the 2019 predator experiment in algae:  "nothing more than a crude clumping together of individual cells had occurred.  A new multicellular organism was not created, nor was any real evolution observed."⁷  He suggested that the permanent clustering "likely meant that a loss of information had occurred,"⁷ and credited "the Creator that engineered the pre-programmed adaptability of these creatures."⁷

4.  The later steps are asserted, not shown, and the parts look designed.  Ann Gauger, a senior fellow at the Discovery Institute, wrote in 2015 that the evolutionary story of Volvox seems simple:  "Only one or two or genes might be necessary to get the cells to form clumps by incomplete cell divisions."⁶  That sentence describes the story as it is told, not her own view.  She then listed what the story leaves out:  "Missing from this story, though, are the details necessary for this 12-step progression to occur."⁶  Among the requirements she named were microtubules and the motor protein kinesin, which Volvox embryos need to turn themselves inside out:  "Neither microtubules or kinesin are present in bacteria so their origin must be accounted for."⁶  Her conclusion:  "Saying that something might have happened is not the same as showing that it actually could happen."⁶  The 2016 Evolution News post read the Gonium genome paper's account of reused genes as a sign of design:  "It sounds like front-loaded intelligent design."²⁵  Mitchell, writing on a 2016 study of an ancient protein, saw "a common design that God used in many kinds of animals and in man to make our multicellular existence possible."²⁶  Stephen Meyer, in Darwin's Doubt, argued that "transitions from a single cell to colonies of cells to complex animals represent significant—and in principle measurable—increases in genetic information,"²⁷ and measured them by genome size.

What was not found.  Behe's 2019 book Darwin Devolves does not discuss the yeast, algae or choanoflagellate work.  Searches of the Answers in Genesis, Creation Ministries International, Institute for Creation Research, Reasons to Believe and Evolution News sites found no response to the 2015 paper that identified the gene behind snowflake yeast, or to the choanoflagellate genomes.  The positions above are therefore mostly answers to the first experiment, and some of what they predicted has since been tested.

Part 2: Two grades, and how often each is said to have arisen

What separates the grades.  Knoll and Hewitt point to a physical difference.  In simple forms, "every cell in simple multicellular organisms is in direct contact with the external environment, at least during phases of the life cycle characterized by nutrient acquisition and active metabolism."²³  A filament or a hollow ball can feed every cell from outside.  A solid, three-dimensional body cannot, so its inner cells must be supplied, signalled and specialised.  That is why complex forms need communication between cells and regulated differentiation, and simple forms do not.  The same authors stress that the first grade does not lead automatically to the second:  "Complex multicellular organisms evolved from simple multicellular ancestors, but not all groups with simple multicells gave rise to complex descendants.  In fact, most did not."²³

How many times.  These counts are inferences from family trees of living organisms.  The organisms are observed.  That a trait arose separately in two branches is inferred from where the branches sit on a tree.

1. Simple multicellularity, at least 20 times and perhaps a hundred. A 2023 review by Ł. Lamża in Biological Reviews opens by saying that multicellularity evolved many times, "with most reviewers agreeing that it appeared at least 20 times in eukaryotes," and that no list with clear criteria had been published. Applying explicit criteria, it identifies "45 independent multicellular lineages," and estimates the true total as "likely close to a hundred."²⁰
2. Complex multicellularity, six or seven times. Miao and colleagues, with Knoll as a co-author, write that complex multicellularity "has arisen only six to seven times, all within the Eukarya."² Knoll and Hewitt list the groups: "animals, fungi (at least three times; Hewitt, 2007), green algae, red algae, and brown algae."²³
3. Within one small group of algae, at least twice. A 2023 study of 110 kinds of volvocine algae, the family that includes Chlamydomonas, Gonium and Volvox, finds that "multicellularity independently evolve at least twice in the volvocine algae."²⁸ Part 5 returns to this group.

What the counts are used to show.  Matthew Herron and his colleagues, summing up their own and others' experiments, write that "Grosberg and Strathmann may have been right to call the evolution of multicellularity a 'minor major transition'."¹¹  The argument is that something that happened dozens of times cannot be hard.  Not every mainstream author agrees.  James Umen of the Donald Danforth Plant Science Center, reviewing the green algae in 2014, turned the same numbers around:  because complex forms arose from simple ones, "the odds of going from simple to complex multicellularity are quite high (approximately one in 12), whereas the initial transition from unicellular to simple multicellular seems far more difficult (one in thousands)."¹⁸  Those ratios are Umen's own rough arithmetic, with no measured number of single-celled lineages behind them.  They show that the same count can be read as "the first step is easy" or "the first step is rare."

A design reading treats repeated origins differently again.  Fazale Rana of Reasons to Believe, an old-earth ministry, argued in 2010 from the brown-alga genome that "The repeated origin of multicellularity in brown algae, plants, and animals defies an evolutionary explanation, particularly since it involves the independent origin of virtually the same biochemical systems."²⁹  His reasoning is that blind processes should not repeat themselves.  For a designer, using the same kind of adhesion protein or cell-cycle switch in separate groups is reuse of a good part, as Mitchell said of one animal protein.²⁶  Rana's argument would be stronger if it priced how often a blind process should repeat a solution.  Neither he nor his opponents have done that.  None of these readings is tested by the counts.  The counts say how often a grade appears on the tree.  They do not say what any single step required.

Part 3: Multicellularity in the laboratory

Every experiment below starts from single cells, applies a pressure that favours bigger particles, and watches what happens.  The selecting is done by the experimenter in some cases, by settling in a tube, and by a predator in others.

Snowflake yeast, 2012 to 2015

William Ratcliff and colleagues in Michael Travisano's laboratory at the University of Minnesota started from a single-celled diploid strain of brewer's yeast, Saccharomyces cerevisiae strain Y55.  Ten populations were grown "to stationary phase (∼10^9 cells/replicate population)" each day, allowed to settle, and only the bottom fraction was moved to fresh medium.¹  "After 60 transfers, all populations were dominated by roughly spherical snowflake-like phenotypes consisting of multiple attached cells."¹  Counting a second run of five populations, "Multicelled snowflake-phenotype yeast evolved in all 15 replicate populations, in two separate experiments, within 60 d of settling selection."¹

What the clusters could do was more than clump.  Cells tracked from single cells under the microscope each grew into a new cluster, and "aggregation was never seen," so each cluster is one cell's descendants.¹  Under settling selection the clusters had "a 34% fitness advantage over individual cells," and without it they appeared to "suffer a 10% fitness cost."¹  They had a juvenile phase:  "300 min of growth and numerous cell divisions are required before the cluster first reproduces."¹  And by 60 transfers a small share of cells were dying by apoptosis at the points where clusters break.  "The frequency of death in these central cells (76%) was far in excess of the random expectation (6%."¹  Those dead cells let smaller pieces break off.  Their frequency "never evolved beyond 2%."¹

The genetic cause came three years later.  Ratcliff and colleagues sequenced the gene ACE2 in one isolate from each of the ten populations.  Five carried mutations in it, two that cut the protein short and three that changed one amino acid in or next to the part that binds DNA.  "These mutations likely lead to non (or less)-functional ACE2p."¹³  Deleting ACE2 from the single-celled ancestor produced snowflakes, and putting one working copy back into an evolved strain "resulted in reversion to unicellularity."¹³  ACE2 is a transcription factor.  Seven of the ten genes most turned down in early snowflake yeast are ones it controls, and they are "involved in daughter cell separation, many acting directly to degrade the bud neck septum," the wall between mother and daughter.¹³  Without them the daughters stay attached.  The other five populations had no change in the protein-coding part of ACE2, and the paper does not say what changed in them.  The same paper adds that "Snowflake yeast evolve in as little as 7 days from a unicellular ancestor."¹³

A separate laboratory reached the same gene by a different route.  John Koschwanez, Kevin Foster and Andrew Murray at Harvard grew ten yeast populations on a low level of sucrose, a sugar that yeast break down outside the cell, so that cells sharing the products do better in clumps.  They used a strain with a raised mutation rate, "to increase the speed of adaptation."¹⁴  All ten populations became clumpy within 25 to 35 transfers.  "ACE2 was mutated in eight evolved clones; six of these mutations were nonsense mutations."¹⁴  A nonsense mutation turns a codon into a stop signal.  The authors' own summary:  "We suspect that most of the putative causal mutations are loss of function mutations."¹⁴

The long-term experiment, 2023 to 2025

Ratcliff's laboratory, now at the Georgia Institute of Technology, then started from snowflake yeast with ACE2 already deleted and asked how large the clusters could get.  Fifteen populations were selected for size once a day, five of them grown without oxygen.  After 600 rounds, which the authors count as about 3,000 generations,¹⁵ the oxygen-free clusters had grown from a mean radius of 16 to 434 micrometres.  "This corresponds to an estimated increase from ~100 cells per cluster to ~450,000."¹⁵  The cells became longer, with their length-to-width ratio rising "from ~1.2 to ~2.7," so that branches tangled and the clusters became more than 10,000 times tougher.¹⁵  The sequenced strains carried 123 mutations between them, with "genes of the cell cycle (29 / 123 mutations, p = 0.02)" over-represented.¹⁵  A 2025 paper on the same populations, run to 1,000 rounds and about 5,000 generations, found a second route.  "In all ten populations, tetraploidy had emerged and become dominant by 50 days," meaning that the cells had doubled their whole genome, and later gains and losses of single chromosomes "underpins the subsequent origins of macroscopic multicellular size."³⁰  A 2024 study of the same line traced the longer cells partly to lower production of a protein-folding helper, Hsp90.³¹  The three papers do not yet agree on how much each change contributed.

None of this produced a second cell type.  The 2023 paper calls snowflake yeast "a model of undifferentiated multicellularity," and treats specialised cells as something larger size might make possible later, "creating opportunities for the evolution of cellular differentiation."¹⁵

Algae under settling and predators, 1998 to 2025

Yeast are fungi, a group that also contains many complex multicellular forms.²³  The green alga Chlamydomonas reinhardtii does not.  Ratcliff, Herron and colleagues selected ten outbred Chlamydomonas populations for settling every three days, "for a total of 73 transfers (~315 generations)."¹⁰  One population changed:  "only this single population had evolved to form multicellular clusters."¹⁰  Its clusters "can contain hundreds of cells" and reproduced through single swimming cells.¹⁰

Predators work better.  In 1998, Boraas and colleagues exposed the green alga Chlorella vulgaris to a single-celled predator.  As Herron's group describes it, this produced "small, heritably stable algal colonies," and they report that the change happened "within 100 generations."¹¹  The 1998 paper itself could not be obtained for this article.  In 2019 Herron and colleagues exposed five Chlamydomonas populations to the ciliate Paramecium tetraurelia, which eats single cells.  "After 50 weekly transfers (~750 generations), simple multicellular structures evolved in two of five predator-selected populations."¹¹  The clusters were small, with median sizes of about 1 to 6 cells depending on the strain, and some released propagules of up to 64 cells.¹¹  Single-celled strains suffered "~2.5-fold greater rates of predation compared to the multicellular strains."¹¹  The authors also noted that the clusters may come from a response Chlamydomonas already has, since it can form temporary clumps called palmelloids under stress:  "It may be, though, that this basis involves the co-option of a previously existing plastic response."¹¹  In 2025 a related team ran 24 populations from one common mix of ten strains.  "Heritable multicellular structures arose in 4 of 12 predation-selected populations," and "never in the settling selection populations."¹²  Every multicellular isolate "arose from one of two founders," so the starting genes mattered as much as the selection.¹²

What the experiments show

The pattern is consistent.  A pressure that rewards size produces clusters of attached, genetically identical cells within weeks to a year, in fungi and in algae.  In yeast the first step is the failure of daughter cells to separate, through a broken copy of ACE2.  The clusters can grow, toughen and acquire a juvenile phase and a cluster-level life cycle.  In a bacterial experiment by Hammerschmidt, Rainey and colleagues, selection on groups produced "a developmental switch" and groups whose fitness "became decoupled from the fitness of constituent cells."³²  But no experiment has produced a body with two or more kinds of cell that each do a different job.  The experimenters say so in their own terms:  the yeast's dying cells are "functionally analogous to germ-soma differentiation,"¹ and the alga clusters "recapitulated early steps hypothesized as leading to differentiated multicellularity."¹¹  The rest is inferred from living species, which is where Parts 5 and 6 go.

Part 4: Was it only a loss?

This site scores each mutation at the level of the molecule, in four classes set out in Measuring Biological Information.  A loss is a working coded element that stops working.  A gain is a coded element or control relationship that did not exist before.  A transfer is sequence taken from another lineage.  Everything else is a modification, marked as raising, lowering or shifting what it changes.  Scored that way, the experiments that found their genes look like this.

Behe's prediction came true.  In January 2012 he wrote that when the genes were found, "functioning genes or regulatory regions were broken or degraded."³  When genes were found, the first was a broken regulator, ACE2:  in a separate clumping experiment from another laboratory in 2013, and in Ratcliff's own populations in 2015.  No change in the table builds anything.  The first step to multicellularity in yeast fits the rule Behe had published in 2010:  "Break or blunt any functional coded element whose loss would yield a net fitness gain."³³  The one detail he overstated was the cost.  He wrote that the clusters were "10% less fit (that’s quite an amount)" without the selection,³ while the paper says only that they "appear" to pay that cost, at a level of confidence the authors did not call significant (P = 0.06).¹

Behe's second "loss" is not shown.  He called the dying cells "the loss of control of apoptosis."³  The paper agrees that apoptosis "was present in our ancestral unicellular strain," as Behe also said, but argues that it "rapidly evolved a new, co-opted function in our multicellular yeast with no obvious parallel in the unicellular ancestor."¹  A 2016 test from the same laboratory made single-celled versions of high-apoptosis snowflake strains by restoring ACE2, and found that they kept the higher death rate as single cells.  The authors conclude that apoptosis "is not caused by side effects of spatial structure, such as starvation or waste product accumulation," and is "a trait that co-evolves with large cluster size."³⁴  They also found that larger clusters hold more old cells, which may explain part of the pattern.  The gene behind the higher death rate has not been identified.  If it is a weakened brake on cell suicide, Behe's word "loss" will fit.  Until then the question is open.

Answers in Genesis was half right.  Its 2011 claim that "the information was in the genome all along" fits the genetics:  no new gene was made, and a yeast that cannot separate its daughters forms clusters because cells that divide stay attached unless something cuts them apart.  But Mitchell's 2012 claim that the authors "demonstrated the yeast had this characteristic behavior before beginning the experiment" is not in the paper.  The ancestor was single-celled.  The paper describes "a novel multicellular life history"¹ that "evolved in all 15 replicate populations,"¹ and the cluster shown forming cell by cell under the microscope came from evolved snowflake clusters, not from the ancestor.¹

Something new did arise, at the level of the cluster.  The new traits are a juvenile phase before a cluster can reproduce,¹ propagules smaller than half the parent,¹ a cluster size that is passed on faithfully, with a heritability of 0.84,¹³ and dying cells placed where the cluster breaks.¹  These traits came from losses and modifications of genes.  That is the same split this site found with the citrate mutation in Lenski's bacteria:  function and sequence are separate ledgers.  A cluster can gain a new way of living while its genome gains no new gene.

Tomkins's reading of the algae is half supported.  His article says that "In two of the five isolates, the algae seemed to permanently express the tendency to cluster together," and then that "In the genetically diverse populations exposed to the predator, this never happened."⁷  The second sentence contradicts the paper.  The predator-exposed populations were the genetically diverse ones, two of them evolved clusters, and the clusters "were not observed in any of the control populations," which had no predator.¹¹  It may be a slip for the controls.  His "just globs of algae"⁷ also undersells the result:  some strains "commonly formed stereotypic eight-celled clusters, with an apparent unicellular and tetrad life stage."¹¹  But his main suggestion, that the change was a loss, fits the authors' own guess.  They propose that the shift to permanent clusters "may have required only a change from facultative to obligate expression of the genes involved in palmelloid formation."¹¹  An optional response can become fixed by breaking the switch that turns it off, which would be a loss, or by other changes that would not.  No gene was identified, so the question is open.

Doyle's 2013 design inference does not follow.  He argued that because all the oxygen-free populations grew large and none of the others did, the change was "part of a system designed to facilitate adaptation."⁸  Replicate populations under the same pressure converging on the same result is what selection predicts.  An earlier experiment from the same laboratory found that clusters grew large without oxygen and at high oxygen, while intermediate oxygen held them back:  "the transition from an anaerobic to an aerobic world can strongly suppress the evolution of macroscopic multicellularity."³⁵  His figure of "10,000 times the volume" also understates the paper's "~2·10^4."[doyle-2013:volume] ¹⁵  His point that the yeast already had what clumping needs is fair:  a broken separation gene is all it took.

Cserhati's speed argument fails, and his "pre-existing mechanism" may be right.  He held that the algal clusters appeared "much too rapidly for any kind of genetic mutation."⁹  That can be priced.  Chlamydomonas makes about 0.1 new mutations per cell division,³⁶ and a predator population in the 2019 experiment made at least about 140 million cell divisions over 50 weeks.¹⁶  That gives about 15 million new mutations per population, enough to knock out any one named gene dozens of times.  A single loss is well within reach.  A particular single-letter change, by contrast, arises in only about one population in twenty, which fits clusters appearing in some populations and not others.  The 2025 follow-up adds that every multicellular isolate "arose from one of two founders,"¹² so variation already present in the founders mattered.  That fits Cserhati's view that the ability was there before the experiment, in the plain sense that the algae could already form temporary clumps and some strains carried variants that made it easier.  His statement that "Para-multicellular structures appeared in two of the five predator-selected populations and one of the control populations"⁹ counts the short-lived clusters that every dividing Chlamydomonas forms.  The paper describes that stage in the ancestor, "the dividing parent cluster is a transient multicellular stage," and does not count it as evolved multicellularity.¹¹

Doyle's requirements were partly met and partly wrong.  His first requirement, that all the cells "contain the same genetic information," is met:  every snowflake cluster grows from one cell.¹  His claim that programmed cell death is a complex system that multicellularity needs, and that design must supply, does not fit the yeast.  The ancestor already had apoptosis as a single cell, and the first clusters formed before apoptosis played any role:  in "the first multicelled genotypes to evolve (14–28 transfers), the frequency of apoptotic cells" did not track settling speed.¹  His larger claim, that true multicellularity needs "an entirely new system of organisation," is about the second and third steps.  The yeast do not reach them, so they neither confirm nor refute it.

Part 5: The volvocine algae, a living series

The best-studied case of the second step is a family of freshwater green algae.  At one end is Chlamydomonas, a single cell with two flagella.  Gonium "ranges from 8 to 32 cells arranged in a flat or slightly curved plate."  Astrephomene has 32 to 64 cells, "with 2–4 sterile somatic cells at the posterior pole."  The genera of the Volvocaceae range "from 16 cells up to 50,000 cells."³⁷  Volvox carteri has about 2,000 small body cells that swim but never divide, and about 16 large reproductive cells inside the sphere that make the next generation.¹⁷  This is a germ line and a sterile body, in an organism whose cells look like Chlamydomonas.

What is observed and what is inferred

The species are observed.  That they form a series of ancestors is inferred.  In 2005 D. L. Kirk listed twelve "important ways in which the developmental repertoire of an ancestral unicell similar to modern C. reinhardtii was modified to produce first a small colonial organism like Gonium."³⁸  Matthew Herron and Richard Michod, mapping those steps on a family tree, number them:  "(1) incomplete cytokinesis, (2) partial inversion, (3) rotation of basal bodies, (4) establishment of organismal polarity, (5) transformation of cell wall into ECM, (6) genetic control of cell number, (7) complete inversion, (8) increased volume of ECM, (9) partial differentiation (soma), (10) complete differentiation (specialized germ), (11) asymmetric division, (12) bifurcated cell division program."³⁹  ECM is the extracellular matrix, the jelly the cells sit in.  Inversion is the embryo turning itself inside out so its flagella face outward.  Umen notes that "Six of the 12 innovations outlined by Kirk (2005) occur in the initial multicellular transition from Chlamydomonas to Gonium."¹⁸  The first step, then, is the largest, and it has no living intermediate.

The tree does not line up neatly with the series.  Herron and colleagues summarise:  "phylogeny does not strictly mirror ontogeny; some traits have multiple independent origins and reversals."³⁷  A 2023 analysis of 110 species found that colonies arose at least twice in the family, not once.²⁸  Sterile body cells arose more than once, and in Volvox the germ-soma split was "established by different mechanisms in separate groups."¹⁸  On the mainstream reading, a germ line evolved several times in one family.  On a design reading, the same outcome was reached by different means.  Neither reading is tested by the tree alone.

The dates are models, not measurements.  There are no fossils of these algae old enough to date the steps.  Kirk wrote in 2005 that "It is estimated that these organisms all shared a common ancestor about 50 +/- 20 MYA."³⁸  Herron's group put the split between Volvox and its single-celled relatives at "≈234 Ma (95% Bayesian credibility interval 209–260 Ma)," and noted that "all of these changes were previously thought to have occurred in the last 50–75 million years."³⁷  Their estimate rests on "a secondary calibration," a date borrowed from another molecular-clock study, and they defend clocks because "in cases in which no relevant fossil record exists they are the only methods available."³⁷  The 2023 study puts the two origins of colonies in the Triassic to Jurassic, again from calibrations borrowed from another clock study:  "In the absence of reliable fossil records, secondary calibration is an efficient approach for time estimation."²⁸  The agreement of the 2009 and 2023 dates is not two independent measurements.

What the genomes show

Few new genes.  The Volvox genome, sequenced in 2010, is about 138 million letters long, close to that of Chlamydomonas.  The authors found that "Despite fundamental differences in organismal complexity and life history, the two species have similar protein-coding potentials, and few species-specific protein-coding gene predictions."¹⁷  "Only 32 EST-supported Volvox gene models lack detectable homologs in Chlamydomonas or other species."¹⁷  What did grow were existing families that build the jelly around the cells:  pherophorins, "49 versus 27 members," and the matrix-cutting enzymes called VMPs, "42 versus 8 members."¹⁷  The authors conclude that the developmental innovations of Volvox "did not involve major changes in the ancestral protein repertoire."¹⁷  The genome of Gonium, published in 2016, found "a maximum of 180–357 genes" tied to the origin of colonies across the family, and transcription factors in about the same number in all three species.⁴⁰

Old genes in new jobs.  Two Volvox developmental genes can be swapped for their Chlamydomonas versions and still work.  glsA, needed for the uneven divisions that set apart the germ cells, can be replaced by Chlamydomonas GAR1, "even though there is no known asymmetric division in the Chlamydomonas life cycle."  invA, the kinesin motor the embryo needs to turn inside out, "can be replaced by its Chlamydomonas ortholog, IAR1."¹⁷  The Gonium team showed that a single cell-cycle gene can push Chlamydomonas toward colonies:  "expression of the Gonium retinoblastoma cell cycle regulator in unicellular Chlamydomonas causes it to become colonial."⁴⁰

One genuinely new gene.  The exception is regA, the gene that keeps the body cells from reproducing.  "Chlamydomonas lacks any recognizable ortholog of the regA gene of Volvox," and "regA encodes an entirely novel combination of pre-existing and new protein domains, of which only the sequence of the presumed DNA-binding domain can be traced back to its Chlamydomonad ancestry."¹⁷  Chlamydomonas does carry a related gene of the same family,²³ and Nedelcu and Michod argue that the sterile body "evolved via the co-option of a life-history gene whose expression in the unicellular ancestor was conditioned on an environmental cue," a gene that in a single cell pauses reproduction under stress.⁴¹  That is a hypothesis about where the family came from.  Nobody has shown the steps from a stress gene to regA, and other Volvox lineages separated germ from body by different means.¹⁸

The creationist claims, tested here

• Doyle's "entirely new system of organisation" overstates what the genomes show. Volvox has the protein set of a single-celled alga, two key developmental genes that work when swapped for single-celled versions, and one new gene assembled partly from old parts. That is not an entirely new system. It is also not nothing: regA is new, and no step-by-step origin for it has been published.
• Gauger's kinesin objection does not apply. She wrote that "Neither microtubules or kinesin are present in bacteria so their origin must be accounted for."⁶ The volvocine story does not start from bacteria. It starts from Chlamydomonas, a cell that already has a microtubule skeleton. Its basal bodies, Umen writes, "nucleate the flagella and organize the internal microtubule cytoskeleton,"¹⁸ and the kinesin Volvox uses to turn inside out can be replaced by the Chlamydomonas version.¹⁷ The origin of microtubules belongs to the origin of eukaryotic cells, covered in Endosymbiosis and the Eukaryotic Cell.
• Gauger's main complaint is fair. The Gonium authors themselves write that the twelve steps were "thought to involve at least 12 steps (Supplementary Fig. 1) though the genetic basis of these steps remains enigmatic."⁴⁰ Most of the twelve have no named gene, no measured advantage and no observed origin.
• The Evolution News post on Gonium was partly right. It was right that the paper does not show that "Gene duplication forms the basis of subsequent multicellular innovations," a sentence the paper states as a conclusion.⁴⁰ It was wrong to say the authors "put the Gonium pectorale versions of the genes back into Chlamydomonas": they put in one gene.²⁵ Its claim that Gonium "is not a multicellular organism" is a matter of definition. Gonium is undifferentiated, which is what the paper says.
• Cserhati's created kind puts the step inside a young-earth history. He grouped Chlamydomonas, Gonium and Volvox into one baramin, a created kind, by how many gene families they share.⁹ Members of a kind share descent, so on his model a germ line was either gained or lost inside this kind within a few thousand years. He prefers loss: "It is also possible that the unicellular algae are derived from multicellular forms. This implies devolution and not evolution, which is a much easier process to explain."⁹ A gain would mean the hard step happening thousands of times faster than the mainstream proposes. A loss fits Behe's rule, and it makes a prediction: a single-celled alga that lost a germ line a few thousand years ago would be expected to carry recognisable remains of regA and of the expanded pherophorin and VMP families. The Volvox genome paper reports that "Chlamydomonas lacks any recognizable ortholog of the regA gene,"¹⁷ which counts against the loss route unless the genes were deleted cleanly. The family trees also place the colonial species inside groups of single-celled ones, with colonies arising at least twice,²⁸ which a loss model would have to reinterpret. And clustering by shared gene families groups close relatives under either view, so it cannot decide between them. Still, Cserhati's model is the one creationist proposal found that makes the second step a testable question, and its test has not been run.
• Meyer's measure, applied here, finds little. Meyer proposed that the step from single cells to colonies brings "in principle measurable" increases in genetic information, and used genome size as the measure.²⁷ For the best-studied colonies the measure has been taken. Volvox, with two cell types and a germ line, has about as many genes as Chlamydomonas.¹⁷ Whatever the step required, it was not mainly more protein-coding genes. It was new control of old ones, plus regA, and nobody has measured how much new information that control takes.

Part 6: Choanoflagellates and the animal toolkit

Choanoflagellates are single-celled filter feeders with a collar of fine projections around one flagellum.  On mainstream trees they are, in the words of one group, "one of the closest living relatives of animals."⁴²  Some live alone.  Others form colonies.  Salpingoeca rosetta forms ball-shaped rosettes, and its colonies grow by cell division, not by cells gathering:  "rosette colonies develop from slow swimmers by serial cell division."⁴³  The same species "differentiates into at least five distinct cell types, including three solitary cell types (slow swimmers, fast swimmers, and thecate cells) and two colonial forms (rosettes and chains)."⁴³  These are stages of a life cycle that one cell line passes through, not different cells side by side in one body.  The authors call the species "a model system for studying simple multicellularity in choanoflagellates."⁴³

What their genomes hold.  The 2008 genome of Monosiga brevicollis, sequenced by Nicole King's group at Berkeley, "contains approximately 9,200 intron-rich genes, including a number that encode cell adhesion and signalling protein domains that are otherwise restricted to metazoans."¹⁹  There are "At least 23 M. brevicollis genes" with cadherin domains, about 120 tyrosine kinase domains, and "78 protein domains that are exclusive to choanoflagellates and metazoans."¹⁹  The same paper lists what is missing.  "In contrast, components of most other intercellular signalling pathways, as well as many of the diverse transcription factors that comprise the developmental toolkit of modern animals, are absent," and "no receptors or ligands were identified from the NHR (nuclear hormone receptor), WNT and TGF-β signalling pathways."¹⁹  Those three pathways are among the main signals animal embryos use to tell cells what to become.  The 2013 genome of Salpingoeca rosetta has 11,628 genes, of which "at least 57% were present in the Uropisthokont," the shared ancestor of animals and fungi on the mainstream tree, and "31% are apparently unique to S. rosetta."⁴⁴

What switches colonies on.  In S. rosetta, colonies form when the cells sense a bacterium.  The paper says "rosette colony development is induced by the prey bacterium Algoriphagus machipongonensis," through a fatty molecule, RIF-1, with "extraordinary potency (femtomolar":  it works at a millionth of a billionth of a mole per litre.⁴²  The first gene shown to matter for rosettes "maps to a predicted C-type lectin," a sugar-binding protein of a kind animals use in adhesion and immunity.  The authors call it "the first link between genotype and phenotype in choanoflagellates."⁴⁵  A colonial species found in 2019, Choanoeca flexa, forms cup-shaped sheets that flip inside out when the light changes.  In the version available for this article, the authors write that it does so "without observable spatial cell differentiation."⁴⁶  Coordinated movement of a sheet of cells, then, does not by itself require specialised cells.

How the mainstream reads this.  Xavier Grau-Bové and colleagues argue that adhesion, signalling and transcription-factor families "were already poised to be co-opted for multicellularity when animals emerged."⁴⁷  Other authors are clear that reuse is not the whole story.  Richter and King write:  "Although many genes involved in development, body patterning, immunity, and cell-type specification evolved in the animal stem lineage or after animal origins, several gene families critical for cell adhesion, signaling, and gene regulation predate the origin of animals."⁴⁸  The Volvox genome team drew the contrast directly:  key animal transcription factors and signalling molecules "are absent from the closest unicellular relatives of animals," which suggests "that animal multicellularity also involved protein-coding innovation."¹⁷  Jordi Paps and Peter Holland put a number on it in 2018.  Comparing 62 genomes, they found that "the ancestral metazoan genome included a remarkable 1189 Novel HG," gene families present in the first animals and in no outgroup, "a threefold increase compared to novelties in the previous nodes."⁴⁹  Casey Luskin of the Discovery Institute read this as proof that "thousands of new genes would have been necessary for the origin of animals."⁵⁰  The paper counts families, not genes, and says each family "can contain multiple genes," so the number of genes is larger than 1,189 but is not given.  Its abstract also says that animals "co-opted numerous unicellular genes in their transition to multicellularity."⁴⁹  Both new families and reused ones are part of the mainstream account of animals, and that question belongs to the sibling article.

How the creationist and design writers read it.  None of the writers in Part 1 discusses the choanoflagellate genomes.  Meyer's paragraph on choanoflagellates asks how much more complex a sponge or trilobite is than a single cell, and answers with genome sizes and cell-type counts, not with these data.²⁷  The closest creationist treatment is Brian Thomas of the Institute for Creation Research, writing on the 2010 sponge genome:  "Sponges and humans share so many genes not because they share an evolutionary history, but because they share the same Designer."⁵¹  The choanoflagellate data fit that reading and the mainstream reading equally well, because both expect shared parts.  What separates them is the missing regulatory toolkit.  On the mainstream account it had to be built on the line to animals, and how it was built is the question taken up in The Origin of Nervous Systems and Animal Body Plans.  On the design account it was supplied.  Neither account has been tested by building one.

Part 7: Pricing the experiments

The experiments can be priced in two ways:  by how many mutations they tried, and by how long they ran compared with the history they are meant to model.  The calculation is in research/pricing.ts beside this article, and every input is listed in its note.¹⁶

How many mutations.  Brewer's yeast makes about 1.67 new single-letter mutations per ten billion letters per generation.⁵²  Its genome is about 12.5 million letters,³⁶ and the experimental strains carry two copies, so each cell division adds about 0.004 new mutations.  A population that regrows to a billion cells every day¹ therefore makes about 250 million new mutations in 60 days.  The genome offers about 37 million different single-letter changes, so each one arises about seven times in each population.  Over the 1,000 days of the long-term experiment the figure is about a hundred times.

The first numbers explain why every population found the snowflake form.  ACE2 can be broken by a stop codon at many places along its length, and every one of those changes turns up several times within weeks.  A one-step loss of this kind is close to certain in any population this size.  The last column shows the limit.  A change that needs two named letters to change together, neither useful alone, is expected about once in a billion 60-day experiments, and about once in a hundred million of the long kind.  The experiments can find any single step, and any series of steps that are each favoured as they come.  They cannot find a pair that must arrive together.

How long.  The longest experiment ran about 5,000 yeast generations.³⁰  The predator experiment in algae ran about 750 generations by the authors' count, and about 195 if the generations are counted from the dilution alone.^{11 16}  On the molecular clock of Herron and colleagues, colonies arose in the volvocine algae by 223 million years ago, and sterile body cells in the line leading to Pleodorina and most Volvox "between 76.6 and 116 Ma."³⁷  That leaves about 107 to 146 million years between the two.  How often these algae divide in the wild is not measured.  Allowing anything from one division a week to one a day, that interval holds about 6 billion to 50 billion generations.  It is about one million to ten million times the longest yeast experiment.  The young-earth window is much shorter.  Cserhati places Chlamydomonas, Gonium and Volvox in one created kind, so on his model whatever separates them happened within a few thousand years.  At the same rates, 4,500 to 6,000 years hold about 230,000 to 2.2 million generations.  That is still 50 to 400 times the longest experiment, so the laboratory cannot test his model either, though it comes closer.

What the prices mean.

1. The laboratory shows that the first step is cheap. It needs one loss, which any population of a billion cells reaches in weeks. That supports the mainstream claim that clumping is easy, and it supports Behe's claim that the easy adaptations are losses.
2. The laboratory has not run long enough to test the second step. Its failure to produce a germ line is not evidence that a germ line cannot evolve. A mainstream biologist is right to say so, and a creationist who cites the experiments as showing that only clumping is possible is claiming more than they show.
3. The same arithmetic cuts the other way. The mainstream account of Volvox rests on tens of millions of years, on molecular clocks with borrowed calibrations, and on steps that nobody has watched. Saying that time was available is not the same as showing what happened in it. That is Gauger's point, and it stands.
4. What would settle the second step is not more years but a measurement: which changes separate Chlamydomonas from a germ-soma alga, how many must arrive together, and how large an advantage each one gives on its own. The genomes say the changes are mostly in control, not in new proteins, apart from regA. Nobody has counted them.

Part 8: What the rocks show, as order

The fossil record gives the order in which grades of organisation first appear in the rocks.  It does not show any population changing from one grade to the next.  The dates below are the authors' labels for the layers.  How they map onto a young-earth history is a separate question that this article does not settle.

1. Simple filaments first. Miao and colleagues describe cellular filaments of Qingshania magnifica, with cells up to 190 micrometres wide and possible spores inside some cells, from rocks dated about 1,635 million years. They write that their result "supports the hypothesis that simple multicellularity arose early in eukaryotic history, as much as a billion years before complex multicellular organisms diversified in the oceans."² Older claims, such as centimetre-sized structures from Gabon dated about 2.1 billion years, remain disputed: Miao and colleagues list them among structures "whose biological origins are uncertain."^{2 53}
2. Differentiated algae later. Bangiomorpha pubescens, a red alga from arctic Canada, has "A differentiated basal holdfast structure" and separate spore and gamete formation, and Nicholas Butterfield called it "the first occurrence of complex multicellularity in the fossil record."⁵⁴ Its first appearance was later dated to "1.047 +0.013/–0.017 Ga," about 150 million years younger than first thought.²¹ Proterocladus antiquus, a green alga from rocks dated about a billion years, has "multicellularity and cell differentiation," including a holdfast and distinct resting cells.²²
3. Animals last. Animal fossils and the timing of their appearance are covered in The Cambrian Explosion.

Miao and colleagues draw the lesson themselves:  "While simple multicellular eukaryotes evolved early, complex multicellular organisms did not appear for another billion years," so simple multicellularity "may be necessary for the evolution of complex multicellular organisms, it is not sufficient."²  On any timescale, the rocks hold a long run of layers with simple multicellular forms beneath the first differentiated ones.  A young-earth account of these rocks has to explain that order as well as the mainstream account does.  The volvocine algae add nothing here, because, as Herron and colleagues say, no relevant fossil record exists for their steps.³⁷

Objections and responses

"Multicellularity evolved in a test tube in 60 days.  The question is settled."  The first step is settled.  Single cells became clusters that live and reproduce as units, in every yeast population and in some algal ones.  The experimenters themselves claim only "the first crucial steps."¹  No experiment has produced a second cell type, and the authors of the longest one still call their yeast "a model of undifferentiated multicellularity."¹⁵

"Five thousand generations is nothing.  Nobody expects a germ line that fast."  That is correct, and Part 7 says so:  the mainstream timeline gives the algae a million or more times as long.  But it means the hard step rests on inference from living species and molecular clocks, not on observation.  A step that has not been watched has not been demonstrated, however much time was available for it.

"Volvox has almost no new genes, so the step to a germ line must have been easy."  It shows the step did not need many new protein-coding genes.  It does not show the step was easy.  regA is new,¹⁷ the twelve steps have mostly unknown genetic bases,⁴⁰ and half of them fall at the first step from a single cell to Gonium, for which no living intermediate exists.¹⁸  How many regulatory changes the step took, and whether any had to arrive together, is not known.

"It is only loss of function.  Nothing new was made."  At the level of the gene, the first step is a loss, as Behe predicted.  At the level of the organism, the clusters have traits the ancestor lacked:  a juvenile phase, small propagules, a heritable cluster size and dying cells placed where clusters break.^{1 13}  This site keeps function and sequence in separate ledgers, so both statements stand.  The long-term experiment also includes changes that are not losses, such as a doubled genome and lowered production of a folding protein.^{30 31}  Those are modifications, not new genes.

"The experimenters did the selecting.  It is selective breeding, not evolution."  Answers in Genesis made this point:  "In botany and animal husbandry, this process is called selective breeding."⁴  The selection in the yeast work is a physical filter, faster sinking, which the experimenter applies but does not steer toward any particular design.  In the algae work a predator did the selecting, the same pressure that acts in ponds.  The objection is right that the conditions were chosen.  It does not change what the populations did under them.

"Yeast came from multicellular ancestors, so they only reawakened an old ability."  Answers in Genesis wrote that "perhaps the reawakening of a dormant ability" would be observed.⁴  On the mainstream tree yeast are fungi, a group with many multicellular members, so the point has force for yeast.  It has less for Chlamydomonas, which, Herron's group notes, "has no multicellular ancestors."¹¹  Even there, the clusters may reuse the palmelloid response the alga already has.¹¹  In both organisms, the first step uses abilities already present, which is what both sides should expect of a one-step change.

"Choanoflagellates carry animal genes, which proves common descent."  Shared parts are expected by common descent and by common design alike.  Thomas of the Institute for Creation Research says shared sponge and human genes reflect "the same Designer."⁵¹  What tests the two accounts is what is missing:  the animal signalling pathways and transcription factors, and how they arose.  That is the subject of the sibling article on animal body plans.

"Multicellularity arising 45 or more times is too improbable for an unguided process."  For simple multicellularity the opposite is true.  If the first step is a single loss that a billion cells find in weeks, many separate origins are what one would expect.  The stronger form of the argument concerns complex multicellularity, counted at six or seven origins.²  Nobody on either side has priced how likely those origins were.

"Intermediates would die before they could reproduce."  Howard Glicksman argued on Evolution News that "It is as if evolutionary biologists don’t take death into account."⁵⁵  For the first step the experiments answer him:  the clusters survived, reproduced and outcompeted single cells by 34% under the conditions that favoured them.¹  For the later steps, where cells inside a body must be fed and signalled, the objection has not been tested either way.

"The verdict on the second step should be 'demonstrated', since the volvocine series shows every stage."  The series shows living species of every grade, and genomes that share almost all their genes.  It does not show the transitions.  The order of the steps is reconstructed with "multiple independent origins and reversals,"³⁷ the dates come from borrowed calibrations that disagree by over 150 million years,^{38 37} and the key new gene has no published step-by-step origin.  That is a strong case for common ancestry of the algae and a weak case for any particular mechanism.  Hence "unfinished".

"The verdict on the first step should not credit evolution, since Behe called it."  Behe predicted the mechanism, a broken gene, and he was right.  He did not predict that the clusters would fail to evolve further, and they have kept changing for 1,000 days.³⁰  Calling the first step demonstrated does not concede the later steps.

Issues and unanswered questions

1. Sources not read. The 1998 Chlorella experiment of Boraas and colleagues is described here only through Herron's group, and Ratcliff's 2012 paper calls its predator a "small-mouthed ciliate" while the title says a flagellate.¹ Grosberg and Strathmann's 2007 review and Knoll's 2011 review in Annual Review of Earth and Planetary Sciences were not reachable, and Knoll and Hewitt's 2011 chapter stands in for the second. Kirk's 2005 paper, Kirk's 1999 regA paper, Nedelcu and Michod 2006 and Richter and King 2013 were read in abstract only. The 2012 yeast paper's supplementary methods, which hold its generation count, could not be fetched.
2. Generation counts. Herron's group gives about 750 generations for 50 weekly transfers at a 1-in-15 dilution, which gives about 195 by itself. The group's later paper uses about 15 generations per transfer,¹² but neither paper says how that was measured.
3. Inconsistencies in the long-term yeast papers. The 2023 paper says the clusters became "~2·10^4" times larger in its abstract and "~2·10^5" in its introduction; the radii it reports give the first. The 2023 paper credits point mutations in cell-cycle genes, and the 2025 paper credits genome doubling and chromosome changes. How much each contributed is not settled.
4. The genetics of the yeast's dying cells is unknown. Whether the higher rate of cell suicide is a lost brake, as Behe said, or something else, waits on the gene.
5. Nobody has priced the second step. The number of regulatory changes between Chlamydomonas and a germ-soma alga, whether any must arrive together, and the advantage of each, have not been measured. Until they are, neither side can say whether the step is easy or hard.
6. The choanoflagellate colony that moves without specialised cells is quoted from the preprint of Brunet and colleagues. The published version in Science was not read.
7. Inputs to the pricing that are assumptions. The cell count per day in the long-term experiment is taken from the 2012 experiment, and the number of divisions per year of wild algae is an assumed range. Both are labelled in research/pricing.ts.
8. Cserhati's created kind has not been tested. If Chlamydomonas lost a germ line within a few thousand years, its genome should hold recognisable remains of regA and of the Volvox-expanded gene families. A search of the Chlamydomonas genome for such remains, and an estimate of how fast deleted genes vanish without trace, would test it. Neither has been done. If instead the kind gained its germ line after creation, the hard step happened thousands of times faster than the mainstream proposes, and that would need its own explanation.
9. No creationist response was found to the identification of ACE2, to the genome-doubling paper of 2025, or to the choanoflagellate genomes. Doyle's revised 2013 page discusses the long-term yeast through its 2021 preprint.
10. Sibling article. Whether the animal toolkit could have been built step by step belongs to The Origin of Nervous Systems and Animal Body Plans, which should take up Paps and Holland's count of new gene families and Luskin's reading of it.

Verdict

Cells that stick together:  demonstrated.  Selection for size produced clusters of attached, genetically identical cells in every yeast population within 60 days, and in a minority of algal populations within a few hundred generations.  In yeast the cause is a broken gene, which Behe predicted before it was found, and the creationist argument is right about that.  It is wrong where it says the ancestors already formed clusters, and it says too little when it calls the result nothing new:  the clusters gained a life cycle the single cells did not have.

A germ line and a sterile body:  unfinished on both sides.  No experiment has produced one, and none has run for more than a millionth of the time the mainstream account allows.  The living volvocine algae show every grade, built mostly from genes the single-celled Chlamydomonas already has, with old genes put to new jobs.  They also show one new gene, regA, and a set of regulatory changes nobody has counted.  Gauger is right that the twelve steps are asserted more than shown.  Doyle's 2009 claim that they require an entirely new system is wrong.

Bodies with many cell types under regulatory networks:  undemonstrated, and the creationist argument is right.  No experiment approaches it and no fossil shows it happening.  The choanoflagellates show that many animal parts are older than animals on the mainstream tree, and also that the animal regulatory toolkit is missing from them.  How that toolkit arose is taken up in The Origin of Nervous Systems and Animal Body Plans.

Sources