Welcome, toad overlords

Very cool study.

But just to play devils advocate a wee bit (without having read the paper), but how much of the trend in the first graph is actually produced by those three points? Then there is the r-square of 0.05 for the line in the third graph, so having longer legs explains one 20th of the variation observed (and again what is the affect of outliers).

There's the old statistician's test - put your fingers over those three points, and see what the correlation in the rest of them looks like. To me, it looks like nada.

And the change is heritable? Phenotypic selection in action, for sure.

It seems like I've heard of many cases of species being introduced to control pest populations, and it always seems to go wrong. Has it ever actually worked out as planned, I wonder?

And some snake heads are getting smaller so they can eat cane toads. When you look at the size of the toad in the picture you may think that a smaller head isn't going to help, but snakes with small heads apparently can't swollow a toad with enough toxin to kill.

Crows, being smarter than snakes, have shown behavioural adaption where they eat the less toxic bits of toads.

I would also guess that since toads are "over-toxic" for Australian predators, that is even if they produced much less toxin most things still wouldn't eat then, I think they might have lost some of their toxicity since their introduction to Australia. Well I hope so anyway.

Introduction of species to control pest species has worked, but in Australia generally only when introducing a species to control a species that was introduced to control species. One exception to this rule I can think of is the hardworking dung beetle which was introduced to control fly populations by burying the manure that livestock produce. They even have a species that is supposed to clean up after dogs in parks.

Snake heads are getting smaller so they don't eat mature, and hence more toxic, cane toads.

Ah yes, but they have yet to evolve a mechanism to defend themselves from that most fearsome of predators:

The one that boils them up and drinks the toad slime (cheap high anyone?).

The one that dries the toad skins and smokes them (Toad skin cigarette anyone?)

The one that dries the toads and sell the toxin for traditional chinese medicine (how about a Cane Toad Pill?).

Or the one that makes Souvenir stuffed cane toads?

A truly fearsome (and stupid) predator.

"They even have a species that is supposed to clean up after dogs in parks."

Here we call them dog owners. (It's a law.) I'm not so sure they like your terminology. :-)

I don't have online access to Nature, and I've just seen your presentation and the short note on the Nature site. I probably should wait, but I won't be able to get into a library in more than 2 weeks. (Based on this second hand information) I don't see how you can call this clear evidence of frogs adapting to increase their expansion rate. I see them adapting to become bigger and or faster. In the explanation of the paper they say that:

"Shine and his colleagues looked at preserved museum specimens and historical records and found that the toads have become 25% leggier and fivefold faster over a 60-year period."

Couldn't the increase in leg size be just a result of bigger/faster frogs being better at exploiting resources? I say that because their data shows that Frogs seem to become leggier everywhere, the regional differences (even if you find them significant) are much smaller than 25%.

Figure b) shows a very very weak trend, but curiously the first arriving toads do not have bigger legs, the median leg length of the first 30 arriving toads is nearly equal to the median length of the last 30 to arrive. This sort of contradicts the short note on the paper:

" They found that the first arrivals had legs stretching up to 45% of their total body length. Later waves of toads consisted of shorter-limbed kinfolk; frogs arriving a year later had legs measuring about 40% of their body length."

I think here they are talking about estimating the values from the linear fit, and that would be ok if they had given error estimates. The animal has about 24 cm, so a 5% difference means 12mm. From the plot you show, taking the straight line fit, the variation at the extremes is about 4 mm, not 12mm as claimed, with a 1sigma error that should be about +-5mm (based on percentiles computed from the plot you show). In no way this is significant or clear evidence of anything. In particular given that the population, on average, increased leg size 25% over time. Of course the REAL paper may explain this.

This paper (well the second-hand account I had from it) seems in fact to show that the driving force in these frogs evolution has nothing to do with expansion reasons. The regional differences are quite small. The simplest explanation for the increase in the expansion rate in figure d) is a side effect of having bigger/faster frogs. Probably the bigger or faster they get, the more things they can eat, and this holds even on already colonized territories.

This seems the second recent Nature paper about toads/frogs that tries to read too much into the data.

What they mentioned in the podcast but was not really emphasised in the paper was that in older populations leg length actually decreased over time - indicating some selection pressure against long legs in "sedentary" populations. This is quite an interesting example of bimodal selection.

The selection for snakes with smaller heads is interesting as well.

Ian B Gibson,
Many bio controls do go wrong and become invasive, but these were mainly in the past. Today we do extensive testing candidate biocontrols before releasing them into the environment to combat pests.
We have also learned from past experiments what make good and bad bio controls. Any vertibrate, social insect, or predatory snail can easily become invasive. On the other hand, predatory insects, parasitic insects, and diseases such as viruses can be very reliable.
In fact, when bio controls do become invasive today, it is most likely a result of it evolving.

Having sat down and read the online version, I have to say I'm not convinced. Sure it's a nice story which fits in with everybodies assumptions, but the stats are dodgy! I don't care what the P values are.

Look at the the first graph. There is the main cluster then there are three pivotal points. The regression is just not convincing.

Graph b. There's no obvious outliers, but there is quite a bit of scatter, and just from eyeballing it, I have this urge to fit a quadratic curve. Just for giggles.

Graph c. r2 = 0.05. 'nuff said. What about the other 95% of the variation?

Graph d. Interesting, looks a lot like an exponential growth rate. So how long do these things live? Are they territorial, is the population constrained to expand in some directions? How does this look if instead of km increase, we plot percentage increase in occupied area size?

The text of the artical talks about the correlation co-efficents rather than r-square. Admitedly this is a brief communication in nature but the coefficents for the slopes in the graphs and standard errors would be nice.

While I think the idea and the story are cool, I don't think the stats is convincing. Maybe I'm just maladjusted.

DrYak, that was the impression people had from toads in the Cairns area, but that goes against this paper (or what I could get from it), they claim there was an increase of 25% in leg size (about 3 cm) since the frogs first arrived. If some toads were decreasing you would expect a huge difference in figure c, clearly much higher than the fluctuations. All populations have nearly the same average size (by the way I was assuming the unit in the figures was mm, but what are the two characters before? how is relative leg length computed?).

About the snakes, does anyone have a reference? There were a series of claims about a Freshwater snake (Tropidonophis mairi) being immune to these toads but I don't know of any recent research on the subject.

Is there are reason why shorter legs would be favored? I'm assuming that stabilizing selection in the ancestral population favored the shorter form. Any explanations as to why?

"Sometimes I call him Cane. Sometimes I call him Toad. Sometimes I call his Dairy Queen."

Or was it, "Sometimes I call him Greeny, sometimes I
call him Red, sometimes I call him Cane but mostly I call him Dairy Queen."

Here is a popular article on snakes adapting due to the presence of cane toads from our esteemed ABC (Australian Broadcasting Association).

http://www.abc.net.au/science/news/enviro/EnviroRepublish_1250708.htm

ABC - Australian Broadcasting Corporation, not Association! Sorry! (Crikey, just imagine what I'm going to be like when I'm actually old and senile.)

I actually heard a really neat story on slashdot claiming that local birds of prey had actually learned to flip the toads over and eat them belly first... thus avoiding the toxins. Don't have confirmation, but it sure sounds like a neat little trick (supposedly one of those behaviors that originated somewhere with one set of birds and then spread elsewhere)

Yep, crows flip them over in Australia, and aparently so do kites. Of course, these birds are diurnal, so it's mostly road kill they are going to eat.

This reminds me of one of my favorite nature documentaries, which, I am surprised, no one has mentioned yet:

"Cane Toads: An Unnatural History" (imdb; amazon)

If you haven't seen this quirky and highly entertaining film, find it in your nearest video library or other source, for it is well worth the effort - especially the shower scene! As the blurb on the cover puts it, "If Monty Python produced a National Geographic Special, it would be Cane Toads!"

One of many pests to be introduced to Australia was the cactus (prickley pear)plant, brought as a garden ornament and inevitably escaping. It laid waste to vast areas of the country. It was successfully controlled eventually with the Cactoblastis moth, So far as I know the moth has not attacked anything else in Australia - when I was taught about it in school it was regarded as our savior. So in answer to Ian Gibson's question, I think that was a biocontrol that worked.

Having just done a quick Google on it though, I note that it is now advancing across Florida and other southern states, much like "our" cane toads. Maybe a research topic there? Are their wings getting longer?

I live in a rural area in northern Australia that is currently right on the frontline of the advancing cane toad population, and they are a freakin disaster. I killed my first one the other night, a large adult. (Christ, they are ugly. At least they are easy to hate.) Besides killing or outcompeting native species, they are also LOUD and are drowning out the often subtle and always beautiful natural nocturnal symphony of mainly insects and frogs.

It is true that some predator birds, such as crows, turn the toads over and eat the poison free underside, but it is not clear that this "behavioural adaption" is recently acquired or due to the cane toads. It is quite possible that this behaviour existed prior to the introduction of cane toads.

"I think they might have lost some of their toxicity since their introduction to Australia." Ronald Brak

Interesting possibility, do you have any references?

Mature toads may well be more toxic than immature ones, but this isn't saying much, as the eggs, tadpoles, and immature toads are still pretty toxic.

'Cane Toads: An Unnatural History' is worth seeing. It has a great theme song.

BTW, my handle (Spotted Quoll) is one of the species being decimated by the toad. A SQ is a marsupial carnivor that serves a similiar ecological role to a cat, hence the SQ's common name of 'native cat'.

And, the cane toad is not the only serious introduced pest we have up here, just the most well-known. For example, there are several introduced pasture grasses that are starting to cause big problems, with no end in sight. Northern Australia a pretty good place to live with oodles of natural heritage and wilderness, but it is definitely not a pristine environment.

With regards to cane toads losing some of their toxicity, I'm afraid I don't have any references. This was just speculation born from my undisciplined thought processes. I've speculated that it might be possible because a cane toad that put less energy into producing toxin might have a breeding advantage over other toads while still being protected from most predators. But if this has actually occured or not I have no idea. The ones in Queensland are still pretty toxic.

i think the first graph (r2=.34) looks pretty good. this kind of field data is bound to be super noisy. the others are a tough sell. they're measuring what's probably a rather small and variable difference, but that can still be very noticeable and selectable. i mean, dutch people are on average less than 1% taller than americans, but when you go there it's like "holy shit everyone is so tall," and natural selection is more observant than people.

that might be a really stupid example that does not parallel this case very well, but i'm tired and going home...

While it's good to be critical, one thing to keep in mind here is what we'd expect the data to look like. That r2 of 0.05 sounds tiny, but as an indicator of selective pressure it looks large. And since selection is a process that filters on natural variation, we'd expect the data to be incredibly noisy. I think the difficulty of extracting a signal from all that variation should be very difficult, and it's impressive that they're seeing any signs of it at all.

You know, before I even read the comments, I said to myself "Australian herpetology, must be a paper by Shine" and lo and behold, it is. What I wouldn't give for an eighth of that guy's pub count...

Anyhow, I'm not really convinced of the stats themselves. Aside from some pretty low slopes and high errors, capability =/= behavior. I can turn sommersaults, but that's not a normal component of my behavior.

Additionally, long legs might have alternative explanations, such as those toads who can make longer sudden jumps are less likely to be flattened by cars.

What I'd be convinced by is combination of laboratory performance testing on first-wave vs subsequent individuals for max speed, aerobic scope, and net cost of transport, as well as radiotelemetry studies showing that longer-legged individuals actually *do* move faster and farther.

Ron Brak: I doubt the cane toad has had enough time (generations) or selection pressure in Australia to reduce its toxicity.

The lack of time is also why I doubt that the behavioural adaption by predator birds is due to cane toads. At least one widespread native frog species (Litoria caerulea, the Green Tree Frog) can make predator birds very sick. (I know because I nearly lost a prized drake when it ate one. Litoria caerulea certainly behaves as if it has no serious predators and shows little fear of any other creature.) So the behavioural adaption in native predator birds may well have preceeded the cane toad's arrival in Oz.

This evening's local TV news reported the findings of this paper (Phillips BL, et al 2006) and included pictures of a toad with a tiny radio transmitter belt with comments about this is for radiotelemetry tracking. Don't know if any tracking data was in the Phillips paper, or is not yet published, or is work underway, etc.

"I don't care what the P values are"?
Look, this is what sampling statistics are for! Phenotypic variation is extensive and complex. Only a very small probabilistic effect on fitness is required for selection to cause directional shifts in mean population phenotypes. Sure, if you cover up the three shortest-legged data the trend looks weaker (though still there, to my eye)...but they are real data, not ignorable "outliers." Even if the story is one of truncation selection against the relatively shortest-legged individuals (note that "relative" variation is shown, centered on zero, which suggests the [proper] use of allometric residuals), the effect would be an increase in average relative leg length in the population.
And these are in fact questions of correlation, not regression, so r, not r^2 is the relevant statistic.
And yeah, Rick Shine's lab is insanely productive. All his pdfs are available from his website, btw.

Not my idea of "clear, empirical evidence of evolution shaping a population." It may be simply an example of the environment shaping the distribution and proportion of genetic characteristics. If the average height of the human population in the world today is greater than it was a generation ago, is this directly attributable to a mutation? I don't think so. Granted, humans aren't part of the same sort of struggle for survival that frogs are. But even if we could show that taller humans were more likely to live longer or more likely to produce more children, this would be no proof that any mutation had occurred between my parents' time and mine. Positing mutations in a study like this is quite a leap in logic. Can we pinpoint when or where this supposed mutation (or mutations) took place? What evidence do we have that the genetic potential for longer or shorter legs wasn't there in the first place?

Henry sez:
"What I'd be convinced by is combination of laboratory performance testing on first-wave vs subsequent individuals for max speed, aerobic scope, and net cost of transport, as well as radiotelemetry studies showing that longer-legged individuals actually *do* move faster and farther"
look again...Figure a shows precisely those latter data (it is not max sprint speed, it's actual 3-d movements).
Max speed, aerobic scope and COT would be interesting data for getting at the physiological underpinnings of the observed patterns...I'd also add a test of locomotor endurance/stamina capacity. And, to nitpick, in this ecological context gross/total COT would be more relevant than NCOT.

Madhu - BTW, Rowdy was definitely referencing the "Cane Toads" documentary, a few posts above yours. He was quoting the little girl with the pet toad that was roughly half the size she was. When I clicked on the comments for this story, I just knew this film would come up sooner rather than later. At my alma mater, it's traditionally watched at an end-of-final-exams party in the bio department. I'm sure it's a similar situation at many other schools. (I was an English major, but sought out the video from the library and introduced a few non-science friends to it as well. We still reference it occasionally in conversation, over ten years later.)

That's another abstraction that weakens the data more than reality probably warrants. The first graph shows that mobility isn't entirely a function of leg length, although leg length contributes significantly. In the third graph, they are using only leg length as a proxy for mobility.

The first graph shows that mobility isn't entirely a function of leg length, although leg length contributes significantly. In the third graph, they are using only leg length as a proxy for mobility.

One of the very first exercises we got the students to do for a bio-stats course I TA'd was to get them to fit linear regressions to various data sets. Each of these data sets gave the same answers, had the same coeffients. The moral of the story? The importance of graphing the data first and asking the question: "Does it make sense to put a linear regression through these data points?"

Looking at that first graph, it's quite a stretch. I'd hate to have any of my results leveraged by THREE points.

Sure, if you cover up the three shortest-legged data the trend looks weaker (though still there, to my eye)...but they are real data, not ignorable "outliers."

Those three points probably are real data. This is a paper in nature after all.
Let me try and put this a little differently. The cluster consists of a large number of points. Look at the variation within the cluster.

Now look at the three points.

Two of them aren't markedly outside of the variation obeserved in the large cluster right? For this graph to be convincing they need more data points for those toads with relatively shorter legs. Otherwise it just looks like an effect due to small sample size. Admitedly it's an effect that makes sense, but they still need more data points for that graph to be convincing.

I'm still interested to see what would happen if quadratic curves were plotted to graphs b and c. What happens if it's not just a linear realtionship. What if shock horro the curve were to fit better to the data?

"I don't care what the P values are"?
Look, this is what sampling statistics are for!

The P values in the paper are for the correlation coeffiecients. Now repeat after me "correlation is not casusation".

They are interesting I suppose, but as I'd much rather see coefficents and se for the slopes of the regressions as those are presented in the figures.

I just got the paper. It's simply too short and the authors had no way to include models and discuss scenarios. I would be more comfortable with extra points in figure d). Particularly since that figure shows that something must have changed in the frog population during the 90s. The sudden increase in the last ten years is really impressive. I dont't buy the explanation it all as to do with leg increase. Was there a climate change? did the frogs reach wetter areas? I understand that in such a short paper authors have to focus too much and can not discuss things in depth. That's the reason why I always become frustrated with Nature papers. Probably there will be a PhD thesis on this and we will get juicy lengthy papers and not the "condensed science" that Nature gives us.

I would like to see the equivalent to figure d) but with frog size from toads collected during the last 60 years. In the short (well, not really, it's almost as big as the article) note accompanying the paper there is a reference to a 25% increase in frog size which is not in the research article.

The biggest problem with that paper is actually the 4th figure.

The speed of movement seems to me to correlate pretty well with the type of habitat the toads were moving through -- they sped up as they got into the flat, warm, fairly wet areas around the Gulf of Carpentaria, and really got cracking through Kakadu which is just the same only more so. The southern limit doesn's show any such zooming tendency, or they'd be in Sydney by now.

A few points raised by other commenters:

-- Yes, the toads are getting less toxic. The data is in Ben Phillip's thesis, but AFAIK hasn't been published yet; this is a source of irritation to me, because I want to cite it.

-- There has been been behavioural adaptation by a number of species to safely feed on toads. You're talking about the likes of crows, and those suckers are smart. I've seen toad carcasses that were obviously killed by one blow to the back, then flipped over and disembowled. Other, stupider species like Eastern water dragons manage to co-exist with toads, so they must have at least figured out to leave them alone.

Their references are outdated too! :-)

Not really the place for a statistical argument but (knee jerks here)...must...reply...
>"In the third graph, they are using only leg length as a proxy for mobility."<

Nah. They are (explicitly) testing a prediction of the hypothesis that 'the invasion process has been assisted by the evolution of improved dispersal ability.' Fig. c shows that the vanguard of dispersers (a measure of realized, ecologically relevant mobility) tend to have longer-than-average legs.

>"Does it make sense to put a linear regression through these data points? Looking at that first graph, it's quite a stretch."<

I agree, it doesn't make sense to put any sort of regression through those points--the question being asked is one of correlation, not regression.

>"I'd hate to have any of my results leveraged by THREE points."<

Good luck with the field work then. Those data (Fig. 1) were hard-won, every one of them, by radiotracking individual toads over three days. Three individuals really did have relatively short legs. Those three animals really did move below-average distances. And there's near-equal leverage from the two most peripatetic puddlejumpers, both of which had longer-than-average legs.

>"For this graph to be convincing they need more data points for those toads with relatively shorter legs. Otherwise it just looks like an effect due to small sample size."<

It's a random sample. You get what you get. Sample size is accounted for in calculating r and its associated P.

>"I'm still interested to see what would happen if quadratic curves were plotted to graphs b and c. What happens if it's not just a linear realtionship. What if shock horror the curve were to fit better to the data?"<

What if? In my view it would be just as wrong as fitting the linear regression...the questions being asked call for interpretation of the probablilities associated with the correlation coefficients. If you wanted to predict the y-axis values from the x-axis values (thereby making an assumption of dependence and independence, i.e. causation), you could go ahead and fit your quadratic--but to answer the question being asked you'd only be interested in the linear component anyway.

>"The P values in the paper are for the correlation coeffiecients."<

yep. Correctly so.

>"Now repeat after me"<

bite me

>"correlation is not casusation"<

And regression is?? To repeat: these figures are presented as tests of three predictions generated by the hypothesis that 'the invasion process has been assisted by the evolution of improved dispersal ability.' These predictions are couched in terms of probablilistic correlations, and the mechanistic causation is largely irrelevant. Cause and effect for each relationship can now be assessed by further, focused observations or experiments.

>"I'd much rather see coefficents and se for the slopes of the regressions as those are presented in the figures."<
They should not have been.

>"The sudden increase in the last ten years is really impressive. I dont't buy the explanation it all has to do with leg increase."<

I sincerely doubt the authors would argue that leg length explains 'all' of the increase.

I think the authors have made an interesting case and collected several nice, independent, consilient datasets in the field that support it. Their conclusions follow from their data. It's presented as a "brief communication." It ought to be heuristic. Not bad, IMO.

(knee jerks here)

:)

I think it is an interesting paper, their story makes intuitive sense, but I also think that they are stretching with their stats.

You're right that the linear regression isn't going to show causation any more than the correlations are. The benefits from the linear regression are that it allows us to make some predictions about how speed will increase with increases in relative leg length. Sure life's not as simple as a linear regression, but I think that the coefficient for the slope and especially the standard errors associated with them should be able to tell us something about what is going on and what kind of precision we can get from the data set. I'm inherently suspicious of data presented without standard errors or confidence intervals. Especially when ther eis as much scatter amongst the data as they have.

To repeat: these figures are presented as tests of three predictions generated by the hypothesis that 'the invasion process has been assisted by the evolution of improved dispersal ability.'

Graph b is the most convincing of the four graphs presented. Looking at that one I can be believe that toads with relatively longer legs get to unoccupied areas earlier. I just don't find the other three all that convincing. They need to show that the frogs with relatively shorter legs are on average slower, they don't really enough data to be convincing. What data they have is indicative of a trend, but there is a strong argument to be made that two of those shorter legged frogs aren't moving any slower than some of the longer legged.
Ok so the field work is tough.
Bugger.

Thanks for telling us about toads decreasing toxicity, Chris. It's nice to know that I sometimes have ideas that don't entirely owe their existance to having been extracted from my behind.

I am not a scientist or a statistician, so maybe there is something I am missing in these graphs or maybe it's just because I haven't read the whole paper, but can anyone explain how a difference of tenths of a millimeter in leg length amounts to a toad being "25% leggier" than its predecessors? Judging from the photo here, this is a rather large toad. And, by the way, how do you measure a toad's leg length to an accuracy of hundredths of millimeters?