Researchers in Japan published a new study on the impact of radiation from the Fukushima nuclear power plant accident to the pale grass blue butterfly. (Fig. 1)

(updated again – 9:35 am; US MST; added a comparison chart of second generation abnormalities in adults; added link to Scientific Reports; correction last table. Updated Fukushima results Table 2 with correct total for adult butterflies.)

Researchers in Japan published a new study, “The biological impacts of the Fukushima nuclear accident on the pale grass blue butterly”  by Hiyama, A. et al, published by the Nature Publishing Group, a division of Macmillan Publishers Ltd (Scientific Reports – “Hosted on nature.com — the home of over 80 journals published by Nature Publishing Group and the destination for millions of scientists globally every month”). The study focuses on the impact of radiation from the Fukushima nuclear power plant accident to the pale grass blue butterfly. In reviewing their data it appears that their results do not fully support their claims.

Figure 1 (Nuclear Diner): From the report – the graph on the left shows % abnormalities vs distance from the reactor for the so-called F1 off-spring from the first generation butterflies taken from May 2011 samples and the figure on the right – is my plot using their data showing abnormalities for F1 first generation offspring as a function of radiation dose levels. The implication from their May 2011 samples for F1 first generation offspring is that the number of abnormalities do not increase with higher radiation dose rate. There does not appear to be any clear correlation between first generation off-spring and radiation dose rates.

Figure 2: Taken from the Japanese report showing % abnormalities vs dose rate for butterflies collected in September 2011. To the right is my plot showing their results for % abnormalities for butterflies collected in May 2011. While the graph from Sept 2011 shows a clear correlation between increasing dose rate and abnormalities, the data from May 2011 does not. Unfortunately, the Japanese authors neglected to include the data from May in the report.

As a lover of butterflies and a believer in nuclear power as an important source of very low carbon output, I was concerned when I read about a new study in Japan linking radiation from the Fukushima nuclear power plant accident and abnormalities in butterflies. I had to take another look –the article seems to imply that the rate of abnormalities is linked to radiation levels but their data clearly shows this is not the case.

As an engineer, I understand data and statistics, therefore I opted to move into the supplemental data provided at the end of the report. The study is based upon a collection of butterflies at several sites around Fukushima and in cities further away. In some cities, they collected butterflies from a single site, while at other cities butterflies were collected from multiple sites. The results are then averaged for each city. For each city, key information is provided including distance from the Fukushima nuclear power plant, ground radiation dose, and the number of male and female butterflies collected for the study. Butterflies were collected from multiple sites on May 13-18, 2011 and September 18-21, 2011, and then from Kobe on October 3-4, 2011.

In reviewing the data I found a number of issues:

1. Natural background – every location has some level of natural background radiation. The report fails to provide a baseline for any of the locations.

2. Sample size – looking at the results from the May collection in Table 2 it is surprising how high is the abnormality rate for adult butterflies. The report states that at Fukushima 20% of the wings from female butterflies have abnormalities. That sounds significant – yet what they neglect to say is that the sample size is only 5 female butterflies! In other words – a statistically insignificant sample! The radiation levels at Fukushima were 1/3 to 1/2 of those from two other sites in the study where the abnormalities in male and female wings were zero. This is poor data analysis and representation thereof, since the results were smeared into the overall abnormality results without providing qualifying statements. That is, the radiation level and damage to the wings did not correlate based on a dose rate. In other words, the report denotes “one” butterfly wing abnormality collected at Fukushima, yet there were no abnormalities for those butterflies collected at cities and locations where the radiation levels were significantly higher.

3. Motomiya, located 58.8 km from Fukushima, had roughly 3 times the radiation dose rate as compared to the Fukushima samples (3.09 and 2.73 μSv/hr vs 1.13 and 1.25 μSv/hr), but had no abnormalities out of the 11 butterflies collected in May – 2 female and 9 males. Yet as the report notes the “mortality rates of larvae, prepupae, and pupae and the abnormality rate of adults were high for Iwaki, Hirono, Motomiya, and Fukushima”. For Motomiya – that implies the offspring from two butterflies – neither of which had abnormalities. Overall the sample sizes for female butterflies are very low yet the authors make far reaching conclusions about their off-spring:

a. Fukushima – 5 females in May; 8 females in Sept (radiation levels around 1.2 May and 0.7 site 1 and 2.43 μSv/hr site 2)

b. Iwaki – 6 females in May; 22 females in Sept (radiation levels around 0.4 μSv/hr)

c. Hirono – 5 females in May; 6 females in May (radiation levels 1.3 μSv/hr in May and .85 μSv/hr in Sept)

d. Motomiya – 2 females in May; 9 females in September (radiation levels of around 2.9 μSv/hr in May and 1.6 in Sept)

4. Mixing apples and oranges- The charts look good, but only the ones that correlate to apparently predetermined conclusions are shown. For example:

Figure 1 – that includes my chart on right – shows that for F1 (offspring from the first females) offspring that there is no correlation between radiation dose rates and number of abnormalities. Therefore as the data did not correlate to radiation levels, the author’s chose instead to show the abnormalities as a function of the distance from the nuclear power plant as shown in the graph to the left.

The F1 abnormalities data for the May collection of butterflies have dose rates of:

  • 0.16 μSv/hr with abnormality rates of 39%;
  • dose rates of 0.53 μSv/hr with abnormalities of 65%, and
  • Fukushima with dose rates of 1.19% and abnormalities of 44%.

The data for the September collection is similar. There appears to be no correlation between increased dose rate and higher abnormalities in the F1 samples.The data appears to be random.

From the paper: “Somewhat unexpectedly, the half-eclosion time and half-pupation time were correlated with the distance from the NPP but not with the ground radiation dose (data not shown). The reason for this outcome is not apparent, but it may be that distance values are more resistant to measurement deviations. Dose measurements in a given collection site could sometimes vary more than one order of magnitude within a several-meter range due to uneven distribution of radionuclides in the field.”

In explaining the fact that their data does not correlate to radiation levels this would actually negate their other results saying there is a correlation. The authors note that the results vary as distance from the reactor rather than by dose rate but do not consider other possibilities related to the tsunami that happened concurrently. Could there have been damage to the butterfly population related to the storm or chemicals released in the storm that have nothing to do with radiation?

In reviewing the results for the F1 generation and observed abnormalities, it is important to note how small is the sample size. Beginning with Shiroishi – they have one female butterfly. One.

May F1 first generation abnormalities


# of female butterflies

Ground radiation dose μSv/hr

Ground radiation dose cal’d averaged μSv/hr

Total abnormality rate (%)








1.13, 1.25





3.0, 2.7










0.46, 0.6





0.3, 0.4





0.18, 0.14





0.16, 0.15, 0.17, 0.16



5. Figure 2 – to the left is actually Figure 4b from the report, which shows a nice correlation between ground radiation level and abnormality rate – but they don’t include the equivalent curve for the May data. Why? Possibly because it shows that the abnormalities are just as high, if not higher, for low radiation levels as for high radiation levels. I plotted their data and show the two curves side-by-side. While the September collection data shows a nice curve, the May collection data does not.

6. Also – the authors change one of the two collection sites in Fukushima in September to one with much higher radiation levels. The dose rates are averaged for the two sites. Most of the butterflies are from Site #1, where the radiation levels are 0.71 Sv/hr versus Site #2, which is 2.43 Sv/hr – yet they AVERAGE the two together to get their nice curve. If the authors were consistent, they would show the number of abnormalities at each site because the radiation levels are significantly different. But they include no qualifications or comments.

7. The authors claim that the radiation is impacting the second generation of butterflies, I put together a comparison of the ratio abnormal adults to total adults for several different cities with the radiation levels at the site. As you can see, the percentage of abnormalities for second generation were almost the same at Tsukuba as Fukushima yet the radiation levels at Tsukuba were close to background and varied slightly between May and Sept. 2011.

The background case, Kobe, had 19%** of their second generation adult population with abnormalities even they had what appears to be no radiation from the reactor.


Per Table 3 supplement: Comparison of 2nd generation F1; number of abnormal adults to total adults

Abnormal/Total May 2011 Collection

% May 2011 Collection

Radiation Levels May 2011 microSv/hr

Abnormal/Total Sept 2011 Collection

% Sept 2011 Collection

Radiation Levels Sept 2011 microSv/hr




1.13, 1.25



0.71, 2.43*




0.15, 0.16, 0.17







0.3, 0.42







0.46, 0.63



0.40, 0.42

Kobe (596 km from the nuclear power plant; background radiation Oct 2011)














*different site for Sept collection

**error in original calculation corrected 8/16/12.







In summary –

  • the increase in abnormalities for the butterflies that were captured in May 2011 do not correlate to increased radiation.
  • there is no relationship between increased abnormalities between F1 off-spring and increased radiation levels.
  • The number of abnormalities decreases with distance from the reactor, also the site of a tsunami, and yet no other possibilities were considered.

Cheers Susan


facebooktwittergoogle_plusredditpinterestlinkedinmailby feather
  1. Jim Greenidge says:

    And they could only find nuclear sins in innocent butterflies…

    Do you know Mark Twain mentioned two headed turtles and six legged toads and other such freak creatures? Which nuclear plant did he live near I wonder!
    It’s funny how these butterfly and bio studies always assume radiation’s at fault in an environment literally flooded by a massive tsunami laced with chemicals, sludge, sewage, pesticides, gasoline and God knows what else was swept inland with it. I hope the public is hep enough to know there are many other ways to induce mutations and discounting (by not even mentioning) the above situation suggests they’re only looking for a fast easy path to knock nuclear power with the radiation blame.Get WHO and other renown institutions backing it up and then I’ll put credence on such studies.

    James Greenidge
    Queens NY

  2. RachaelLudwick says:

    A few comments:

    * Science.com? I assume you meant nature.com, but why not just use an actual link?

    * You’re ignoring the feeding trials. The results and methods for that experiment are poorly presented and may not be any more significant (sample sizes much bigger though) but seems pretty solid. Once I understood that part I realized my kneejerk critical reaction wasn’t really fair. Feeding under controlled circumstances seems like it would have fewer or at least more complicated alternate explanations. Pollutants seem like they would be less of an issue for instance.

    * The issue of site selection seems bigger than just the radiation differences: for a bunch of the results they seemed to be grouping collections from two places in the same area. It seems like each individual collection point should be a sample, not “spot X and 100 yards east”. But I don’t know hardly anything about statistics (except very basic stuff) so perhaps it is valid to merge samples in this case (leaving aside the measured radiation differences).

    Though overall I think the main problem with this news isn’t really that the study is incredibly bad. Less than perfectly valid stuff is published all the time, then other researchers build on whatever seems actually good, etc. But the reporting on it far worse. If this study was about something less high profile, almost no one would bother to blog about all its methodological problems because no one would care. But is is high profile so there’s lots of repetitive news reports playing up dangers (or asking fairly speculative questions and even cracking Mothra jokes).

    Note: I blogged on this subject myself emphasizing the media aspect (but not linking as that seems kind of sketchy) and did link here as an example of actual critical examination. Specifically, it’sunfair to criticize you for ignoring the feeding trials in that post without letting you know. :)

  3. RachaelLudwick says:

    Also, I think you might be missing a bit more subtlety to what the study is trying to show. The explanation for this result is that the first voltine butterflies (the May collection) were exposed as overwintering larvae and are not themselves expected to be strongly abnormal. Instead, they passed on defects to their offspring. The September collection are third or fourth generation after the more strongly exposed ancestor. So really we should be graphing May radiation doses against September abnormality rates. The May radiation versus May abnormality curve isn’t necessarily interesting (by their argument) or even likely to show a similar curve.

  4. aligatorhardt says:

    Butterflies do not stay in one place, so the locations of the affected individuals may not be where the ancestors were exposed. The differences in radiation levels were small, so I am not surprised that a dose correlation with effects is not seen in the small number of subjects. There is no threshold for genetic effects, so level correlation may not be seen. This study is too small for the final word, but it is significant that mutated animals are now seen when before they were not. The fact that radiation is a strong mutagen has been known for many decades. Correlations between reactors and human birth defects have been shown many times, as well as the prevalence of cancer, another form of mutation.

    • Susan Voss says:

      You raise some really interesting insights and I agree, I just scratched the surface and there is much more that needs to be brought forward. I’d encourage anyone with additional insights to please post their comments.

  5. Bob Applebaum says:

    Rachael did a good job at illustrating how bad this post is. The study is very typical and mainstream science. Kobe wasn’t really part of the study, it was excluded because it wasn’t a butterfly source in May. One has to compare a particular population with the same population over time. Background radiation levels are just a rough guide since they are heterogenous and only show the gamma component.

    The study authors then use lab-based external and internal radiation to show causality, ie., that radiation produces the effects seen.

    The mutations could be due to something else, but one would need to do a study like this to show causality. Just attacking a study because you don’t like its results is denialism.

  6. Susan Voss says:

    Per the report:
    We also collected fourth- or fifth-voltine adults on 18–21 September 2011 from 7 localities and on 3–4 October 2011 from one locality (Kobe) (Fig. 1a; Supplementary Table 1).

    We again collected Z. maha adults from the 7 localities (i.e., Fukushima, Motomiya, Hirono, Iwaki, Takahagi, Mito, and Tsukuba; 18–21 September 2011) and from Kobe (3–4 October 2011) (see Fig. 1a). They were probably the fourth- or fifth-voltine individuals.

    For the adults collected in September and October 2011, we selected both healthier (non-abnormal) and abnormal males and females from a collection locality and confined them in a single tank (2 males and 5 females for Kobe, 2 and 4 for Takahagi, 2 and 4 for Iwaki, 2 and 3 for Hirono, 2 and 5 for Motomiya, and 4 and 5 for Fukushima). Among these individuals, 7 adults from Kobe were all non-abnormal, but the adults from other localities included both non-abnormal and abnormal individuals. The females were allowed to lay eggs for 7 or 8 days.

    Table 1 Supplement:
    Kobe Distance from reactor 596 km; ground radiation 0.08 μSv/h (According to the Japanese Ministry of Education web site the maximum value at the Hyogo Prefect (Kobe) since published was – 0.065 μSv/h (http://mextrad1.blob.core.windows.net/page/28_Hyogo_en.html).

    Table 3 Supplement: Kobe F1 second generation from the adults collected – there were 43 abnormal adults out of 223 adults. 19% – why so high in the control group?

    I am interested in understanding the results – why do the number of abnormalities vary with distance from the reactor and not radiation level?
    What about other contaminants – chemicals – released and distributed in the air, ground, and water as a result of the tsunami?

  7. Susan Voss says:

    Concurrent events:

    Large earthquake

    Tsunami and rising waters

    Destruction of cities and towns

    Release and distribution of chemicals and other toxins

    High death rate of people and animals

    Accident at Fukushima and release of radioactivity (Wikipedia)

    Large oil fires – a 220,000 barrel/day oil refinery and a 145,000 barrel/day refinery set on fire

    Nuclear power was replaced by oil and natural gas – estimated as much as 300,000 barrels of oil per day (Wikipedia)

    If the only event was an accident at Fukushima, then yes, it would be easy to say – cause and effect.

    Could it have been something other than, or in combination with, the radiation?

  8. Bob Applebaum says:

    Different populations can have different mutation rates for all kinds of reasons (underlying genetic differences, exposure to different levels of environmental mutagens, viruses, etc.).

    That’s why in epidemiology you define a population and follow it over time (a closed-population study is preferred, but in this case it’s an open-population study due to the realities of butterfly migration). It’s hard enough to do epidemiology within a population, the confounders increase greatly when one tries to compare across different populations. That’s why the authors correctly ignored Kobe in making comparisons.

    In a reactor accident, there isn’t any uniform fallout/washout levels across an entire city. To assign one value to a city is to commit the ecological fallacy, and the authors avoid it. Instead, we expect in a general way that radiation levels will decrease in an overall way with distance. Also, generally, we know that the butterflies can migrate between cities (open populations). Aggregating the results helps to deal with that, which the authors do.

    The big picture is to look at the butterflies on the whole and then a particular population over time. The authors have shown that there is genetic damage to the butterflies. It varies generally with distance from the reactor, which suggests it’s due to Fukushima and not something else. They then do lab experiments with external and internal radiation and get similar effects. They conclude the most likely cause is in fact Fukushima. Look up the Bradford-Hill criteria for causality. They’ve addressed them pretty well.

    The mutations could still be largely due to something else, but unless you have a similar study regarding another agent, you’ve presented no evidence of anything else. As of today, the preponderence of the evidence is on their side.

  9. Bob Applebaum says:

    Most of the stuff on your list is not mutagenic. Of those that are, why would the degree of effect vary with distance from the nuclear plant? The authors have shown that radiation induces the observed effects (reversibility).

    Mutagens certainly work with each other (ie, the DNA damage accumulates regardless of the number of types of mutagenic agents). So, it could have been something other than or in combination with the radiation. But if you’re RATIOnal than you ratio your belief to the strength of evidence. There’s no EVIDENCE of anything else which explains the observations as well as radiation.

  10. Guest says:

    “There’s no EVIDENCE of anything else which explains the observations as well as radiation.”

    They also weren’t looking for any OTHER explanation. There is no evidence to explicity link these mutations to radiation either. Your argument is pretty telling when you excuse the sloppiness of this research by stating that plenty of other accepted research is poorly done so we should just excuse this outright as well. This is compounded by insinuating a person isn’t rational if they don’t perform some “belief ratio”. Concerning the “strength of the evidence”, that is exactly what you start asking us to make exceptions for! Don’t tell us to overlook the methods and then say people aren’t rational for not accepting them based upon these overlooked problems. It doesn’t matter what one’s views are on LNT and the effects of low dose radiation, poor research should be called out, NOT excused. The answer isn’t lowering the bar, it’s having a standard and pointing out when it isn’t met. Which would you rather your arguments be supported by? Research that is solid or poorly done and shaky at best? Perhaps you’d prefer the challenge of looking like you require work that is not rigorously done to support yourself, personally I wouldn’t choose a standing that makes it look necessary for me to avoid solid scientific research to make my points.

  11. Jaro Franta says:

    “Most of the stuff on your list is not mutagenic.”

    OK, let’s take a closer look at just what the stuff is, as was done in this NIH publication:

    “Chemical Aftermath: Contamination and Cleanup Following the Tohoku Earthquake and Tsunami”

    Specifically, a list of the chemicals, their known effects, and main locations along the east coast of Honshu is provided here:

    A couple of quotes from the NIH web post:
    “In the weeks following the Tohoku disaster, Toxic Watch Network, a Tokyo-based nonprofit organization, combed the PRTR data to get a general idea of the chemicals that may have been onsite at affected facilities. The resulting list includes acrylamide, asbestos, benzene, bisphenol A, bromomethane (methyl bromide), cadmium, chromium compounds, chloroform, chlorodifluoromethane, ethylene glycol, dioxins, formaldehyde, lead, mercury, toluene, and xylene (see map, p. A292).22 Many of these compounds are respiratory hazards, neurotoxicants, and/or carcinogens. Many are potentially acutely toxic. Some are also environmentally persistent, which raises potential issues of long-term contamination, particularly to local soil and water.”

    “The Kashima coastal industrial zone is, according to Japanese accounts, home to the largest number of petrochemical industrial complexes in Japan.14 Other facilities here are a Mitsubishi Gas Chemical plant whose products include hydrogen peroxide and polycarbonates; an Adeka Company plant that produces chlorinated inorganic chemicals, flame retardants, caustic soda, and other chemicals along with oil-based food products; and a Mitsubishi Chemical plant that is Japan’s largest ethylene production site. Facilities reporting damage include the Shin-Etsu Chemical Plant, a polyvinyl chloride (PVC) factory (Shin-Etsu’s optical fiber plant in Fukushima was also damaged); Sumitomo Metal, where the earthquake toppled equipment and triggered gas tank fires; and an Asahi Glass plant that manufactures caustic soda, propylene oxide, fluorocarbon resin, and other chemicals, along with various types of flat glass used in building construction.”

    Assigning blame to low-level radiation in a background of such a chemical soup, based on a ridiculously low sample number, seem disingenuous to say the least !

  12. RachaelLudwick says:

    I’m not really going to disagree that there are some major problems with both the May/Sept collection comparisons and the breeding experiments from those collections. I was pretty disappointed to see all the media stories on this emphasize that part of the study since it had such obvious problems. Other major problems are things like previous radiation levels, previous abnormality rates, etc. But really given the small sample sizes I’m not sure it’s worth worrying about these other confounders.

    But, this is science and mediocre papers get published but that it’s still not right to dismiss the entire thing. We should be critically analyzing all of it. The feeding study with gathered leaves *also* might have some problems but it seems to be something worth pursuing more. Unfortunately the public is just going to think this is a slam dunk towards there being evidence for major harm from Fukushima … and other scientists will move on and find more or less evidence for this paper’s ideas.

  13. Will Boisvert says:

    Good post, but you’ve barely scratched the surface of the problems with the butterfly paper.

    (1) One is the reliance on assessing very subtle “abnormalities”—things like the pattern of spots on the wing, tiny discrepancies in the size and hue of spots, etc. These abnormalities are often in the eye of the beholder.

    The authors themselves say so in the Methods section. They had two people check each butterfly for abnormalities, but “disagreement often arose from overlooking apparent abnormalities. Thus when two persons disagreed, they discussed discrepancies together and reached agreement on those issues.”

    That sounds like they did two independent assessments, but in fact it’s only one. Whichever observer is the better arguer—or has more academic clout—will win the negotiation over whether a feature is or is not an abnormality. The way it should work is to have two truly independent assessments by people who don’t communicate with each other. Anything that’s not flagged by both should simply be ignored, because if two expert entomologists don’t agree on whether a given feature is an abnormality, then it’s not an abnormality.

    The fact that the paper uses such subtle and imponderable phenomena as its data means that it must be rife with subjective judgments rather than hard evidence. One way to deal with that problem is to do the assessments blind, that is, without the assessor knowing which test- or control-group specimens came from. (That’s why human clinical trials are routinely double-blinded.) The authors don’t say whether they blinded the assessment, which leads me to conclude they did not, because if they went to the trouble of establishing rigorous blinding procedures, they probably would have said so in their otherwise elaborate methods section.

    Given the subjectivity of the data and the apparent lack of proper blinding, the possibility of observational bias distorting the paper’s findings has to be considered.

    (2)The experiments that exposed butterflies to external radiation in the laboratory are of dubious relevance.

    The external radiation exposures in the laboratory were at doses many times higher than butterflies in the field could have seen. The highest external air-dose rate at a field collection site was 3.09 uSv/hour. Over a butterfly’s one-month lifetime that amounts to a total dose of 2.3 mSv. Over the roughly 6 months between the tsunami and mid-September the collective population of butterflies was exposed to that radiation level, that could have amounted to 13.8 mSv (actually less because of decay). But in the data presented, the laboratory butterflies were exposed to radiation levels of 55 and 125 mSv, that is 24 to 54 times higher than any field butterfly received, and 4 to 9 times higher than the collective field population accumulated over 6 months. Moreover, the dose rates in the data presented were 0.20 and 0.32 mSv/hr—that is, 65 to 104 times higher than the highest dose rates in the field. Mutation rates are known to increase with higher dose rates as well as higher total dose. So, these very high laboratory radiation exposures are unrealistic models for the mutations that wild butterflies might suffer from much lower external radiation levels in the field.

    (3) The feeding trials in which laboratory butterflies were fed radioactive leaves from various sites close to the Fukushima plants are completely uncontrolled, introduce possible confounding variables and do not correlate well with the alleged effects.

    The right way to do this experiment is to take the same kind of plants, raise them under identical conditions in the same soil, add varying amounts of radio-cesium to them and feed them to test groups of butterfly larvae to see what the effects on mutation and survival rates are. (You might either spray cesium on grown plants or grow the plants in soil with varying amounts of radio-cesium.) That would be a rigorously controlled experiment.

    The wrong way to do the experiment is the way the researchers did it—to go out in the field, at points separated by hundreds of miles and collect leaves from wild plants that just happen to have varying amounts of radio-cesium in them. That protocol potentially introduces all manner of confounding variables into the experiment that can effect the mutation and survival rates of the test butterflies that feed on the leaves. Soils in one locale might lack an essential nutrient that butterflies need to thrive. Leaves in another locale might have been sprayed with an insecticide recently, or some other mutagenic chemical. Maybe there’s a drought where the most radioactive leaves are collected, making them unpalatable, while the control leaves from Ube, hundreds of miles away, are luscious from ample rainfall. The protocol is therefore uncontrolled, and its results are meaningless.

    In any event, the survival rates don’t correlate very well with internal radiation exposures. The data in the paper’s table 5d show that the survival rates for BFs fed on leaves from Iitate montane area are substantially higher than those for leaves from Fukushima and Iitate flatland, even though Iitate montane leaves are 3-4 times more radioactive than Fukushima or Iitate flatland leaves. (Supplemental table 8.)

    4) Discrepancies between May-collected and September-collected butterfly abnormalities are suspect.

    One important point to remember is that the abnormality rate in the May-caught butterflies is completely ordinary, (Table 1) 12.4 % for 7 locales and 13.2 % for ten locales, all with elevated radiation. By comparison the abnormality rate in radiation-free Kobe is 13% (Supplementary Table 7), and for radiation-free controls in the laboratory it is 16.7 %.

    In other words, the radiation hasn’t caused any extra abnormalities at all in the May butterflies. The authors’ real argument is 1) that there are hidden germline mutations in those May butterflies that will show up in their lab-bred offspring, and 2) that their wild offspring will accumulate more germline mutations that will produce elevated abnormality rates in September butterflies and their lab-bred offspring.

    Their data do seem to show pronounced increases in September abnormality rates compared with May, but these are doubtful. To add to the raft of possible confounding variables, including assessment bias, you have to add possible seasonal variations—maybe butterflies are simply more stressed, sickly and damaged in the wilting fall than in the blossoming springtime. That could account for the seeming accretion of abnoramlities over months, which the authors attribute to accumulating radiation-induced germline mutations for no good reason.

    There are also supposedly much higher rates of abnormality in the September F1 offspring than in the May F1 offspring. But this could be due to a selection bias by the researchers. According to the Methods section, to breed the May F1, “we selected healthier and vigorous males and females from specified collection localities,” although a few displayed “mild abnormalities.” But to breed the September F1, the researchers “selected both healthier (non-abnormal) and abnormal males and females from a collection locality.” The September Kobe breeders, the no-radiation control group, were all non-abnormal, but the other September F1s were bred from “both non-abnormal and abnormal individuals.”

    In other words, for the May F1 they made an effort to select the healthiest breeders with only a few “mild abnormalities” included, while for the September F1 they just threw in both healthy and abnormal breeders without any apparent selectivity—except for the Kobe control group, where they deliberately excluded all abnormals! So the result that September butterflies seem to spawn more abnormals than May butterflies may well just be selection bias, not any extra germline mutations accumulated over time in the field.

    That’s a pretty severe problem since the paper’s whole point is that the butterflies accumulate germline mutations over time because of radiation in the field. But there are other weaknesses with that thesis. For example, the mutations in the May F1 don’t actually seem to be in the germline since the subsequent mutations in the May F2 generation, the F1’s offspring, are usually not the same mutations as in the F1. Germline mutations should be passed along from parent butterflies to their offspring, but in fact that rarely happens; offspring usually have abnormalities that are different from the ones afflicting their parents. (Table 3b.) That means that F1 and F2 abnormalities are more likely developmental mutations rather than germline mutations—but there is no radiation in the laboratory to cause developmental mutations.

    In short, this is a really lousy paper. It’s a shame that the press is touting it.

  14. Aaron Barmer says:

    A very unscientific question, nerds, from a person reading anything he can to quell his fear: when are you harm deniers not already living there planning to move to Japan (with your families) to directly and personally collect data for research of the aftermath of these accidents?

  15. Susan Voss says:

    Interesting post by folks at MIT:
    Recommended reading:

    The butterfly mutation study, unfortunately, has some rather unscientific facts not made clear through the media:

    The study had small numbers of samples from arbitrary locations between the May and September samplings.

    It was stated that there was a correlation between radiation exposure and abnormalities, but given with potential errors far higher than accepted in statistics to make any statement. (“p” should be less than 0.05, here, it was 0.13) This is not good science or good research.

    The study picked only a few of these to breed (ranging between 1 and 6 of each sex, usually 2-3) to form the first generation (“F1″) including those with abnormalities.

    For the second generation (“F2″), the study only used 1 “significantly aberrant” female with a couple males from a single locality.

    So, the astounding [sic] results of breeding abnormal butterflies is that there are more abnormalities in the next generation.
    Here is the good news: regardless of whether higher radiation has increased abnormalities, nature is accustomed to radiation and weeds out detrimental mutations and keeps the beneficial mutations–the species ends up growing stronger. On top of that, radiation levels decay and within a couple decades radiation from Cesium 137 will be more negligible.

Leave a Reply