Darwin's fear was unjustified: Study suggests fossil record gaps not a major issue:
Fossils are used to reconstruct evolutionary history, but not all animals and plants become fossils and many fossils are destroyed before we can find them (e.g., the rocks that contain the fossils are destroyed by erosion). As a result, the fossil record has gaps and is incomplete, and we're missing data that we need to reconstruct evolutionary history.
Now, a team of sedimentologists and stratigraphers from the Netherlands and the UK have examined how this incompleteness influences the reconstruction of evolutionary history. To their surprise, they found that the incompleteness itself is actually not such a big issue.
"It's as if you are missing half of a movie. If you are missing the second half, you can't understand the story, but if you are missing every second frame, you can still follow the plot without problems.
"The regularity of the gaps, rather than the incompleteness itself, is what determines the reconstruction of evolutionary history," explains Niklas Hohmann of Utrecht University's Faculty of Geosciences, who led the study. "If a lot of data is missing, but the gaps are regular, we could still reconstruct evolutionary history without major problems, but if the gaps get too long and irregular, results are strongly biased."
The study is published in the journal BMC Ecology and Evolution.
Since Charles Darwin published his theory of evolution, the incompleteness of the fossil record has been considered problematic for reconstructing evolutionary history from fossils. Darwin feared that the gradual change that his theory predicted would not be recognizable in the fossil record due to all the gaps.
"Our results show that this fear is unjustified. We have a good understanding of where the gaps are, how long they are and what causes them. With this geological knowledge, we can reconstruct evolution hundreds of millions of years ago at an unprecedented temporal resolution," says Hohmann.
Journal Reference:
Hohmann, Niklas, Koelewijn, Joël R., Burgess, Peter, et al. Identification of the mode of evolution in incomplete carbonate successions [open], BMC Ecology and Evolution (DOI: 10.1186/s12862-024-02287-2)
(Score: 4, Interesting) by JoeMerchant on Tuesday September 03, @01:12PM
I would agree, seeing Twisters at 12 or even 8fps would still tell the whole story (although the question of "Did Han shoot first?" might come down to a single frame...)
However, I would also expect these gaps in the fossil record to be more chaotic than periodic. Missing every other 100 years here, every other 1000 years there, whoops: there goes a million, back to the every other 100, etc. Further, some of the most important events, evolutionarily, probably also coincide with some of the craziest shifts in sedimentation....
🌻🌻 [google.com]
(Score: 2) by VLM on Tuesday September 03, @09:35PM (5 children)
These guys should re-interpret digital music, might be funny from an EE perspective to see their take on sampling rates and the like.
I mean, if in some alternate history no EE telecom theory existed, I could see people getting all weird about digital recordings of pianos requiring full finite element analysis of all vibrating parts of the piano and individual analysis in real time of all the piano strings, which might get kind of complicated, but if you allow Shannon and sampling theory and all the cool kid stuff from last century, all you need is a 16 bit analog sample reliably a couple thousand times per second to get an excellent recording of a piano; sometimes things are simpler than they first appear.
And that's what they did with fossils.
(Score: 3, Interesting) by JoeMerchant on Tuesday September 03, @11:02PM (3 children)
I took all the classes, I know all the theories, I can quote the textbooks about "why a 1000 pole filter is useless"... then:
An M.D. who "didn't know any better" went ahead and applied a ludicrous digital filter to his data anyway - theories all said "there's no information in there to get, the sample rate is too low, etc."
All I know is: with his mega overkill filter, he got the results he was looking for (separating cardiac from pulmonary signals in high resolution measurement of torso cross sectional area) - his cardiac curves out of that filter very closely matched echo-doppler results, but without needing the ultrasound sensors.
I tried "downsampling" his filters to something sensible according to the textbooks, the signals were ... similar, but not as good.
This was _not_ the brain mapping researcher who had $5K audiophile speakers in his office (cast-offs after he got "better stuff" for his home). That guy didn't believe in mega-ultra-low oxy cables, but he certainly did believe in his exposed drivers and connoisseur developed ported resonant chambers. I would never say that there's "bang for the buck" in the high end audiophile gear, or wine, or any similar things, but there _is_ a difference.
🌻🌻 [google.com]
(Score: 3, Interesting) by Freeman on Wednesday September 04, @02:27PM (2 children)
Sometimes that difference is literally just noise that's been introduced into a perfectly good audio recording. They're just paying a premium for that "extra special bit of noise". Thanks guys, I lived through hearing the hiss of a cassette player while listening to things. It's Not A Feature, unless you're a salesman.
Joshua 1:9 "Be strong and of a good courage; be not afraid, neither be thou dismayed: for the Lord thy God is with thee"
(Score: 2) by JoeMerchant on Wednesday September 04, @03:34PM (1 child)
Subjective things are very tricky... many (most?) LP connoisseurs definitely had a soft spot for the EQ profile that studios put on their vinyl pressings to compensate for poor reproduction of the moving needle cartridges - I wonder if removal of that EQ for CDs was intentionally done to keep vinyl sales alive and potentially "premium"...
On the medical data, the "gold standard" was an optically processed ultrasound video - each frame measured the number of pixels "inside" the ventricle, our sensor wrapped the whole torso and so caught all kinds of stuff, mostly changing lung volumes, but when we hit that signal with the brick wall filter, we could get a very similar signal to the one coming from the ultrasound heart images - features in the time series were readily identifiable in both, and more muddied when the "pointless overkill" filter was reduced to something "just as good" according to all the textbooks.
🌻🌻 [google.com]
(Score: 1, Interesting) by Anonymous Coward on Wednesday September 04, @10:32PM
There's a parallel with AI, which can easily have 65 million free parameters ("poles") that are optimized just so to pull out, amplify and enhance an input signal. It's not quite the same, since the AI model can effectively pattern match to a high quality signal stored in its free parameter and output that if it's close to the original. Or in the case of analog, there's an extraordinary example of evolutionary circuit design (https://www.damninteresting.com/on-the-origin-of-circuits/ [damninteresting.com]) that exploits weird and wonderful effects unrelated to the components' intended purpose.
(Score: 0) by Anonymous Coward on Wednesday September 04, @04:06AM
The analogy isn't clear to me. They're not periodically measuring a single temporal signal (bandwidth-limited, of course, if you want to stick to Shannon-Nyquist). They've got many thousands of signals, some of them highly interrelated, with different gaps and gap durations. And even any individual signal (e.g., a fossil line) can be composed of many different signals since you aren't just dealing with a single change to a fossil at a time. I think if you want to stick to some kind of signal analysis analogy, I'd think of it more like compression or principle component analysis where you find that you can eliminate a lot of redundant or less significant information and still be able to do decent signal reconstruction, though I can also maybe talk myself into thinking of it like analyzing state machines (I might need to argue with myself over beer for that).