Wallace Milam and the heterogeneity of Sb

      Guinn’s entire interpretation of the five fragments depended on the two groups’ (representing two bullets) being analytically separate, which in turn requires that their difference in concentration (about 30%) be much larger than the uncertainty of each group (about 1% for each of the fragments). Obviously, 30% far exceeds 1%. But for some reason, Guinn published his results with the smallest possible analytical uncertainties attached. That 1% represented only the “counting standard deviation,” from summing the number of gamma-ray counts under each of the photopeaks of Sb in the samples. The details are not important, except to note that the full analytical uncertainties of the neutron-activation process are about 2–3 times larger, or 2%–3%. To this must be added the heterogeneities of Sb within the bullet’s lead core—the critical factor in the next stage of discussion. Guinn testified to the HSCA that according to his experience, the total uncertainties of his Sb data (including heterogeneities) were about six times the counting uncertainties, or about 6%. By this figure, the two groups of fragments remain clearly separated, and strong conclusions can be drawn from them. Guinn testified, to great acclaim from the HSCA and the media, that the fragments were “clearly distinguishable.”

Enter Wallace Milam
Fifteen years passed, and then Wallace Milam entered the picture. Wallace Milam is a high-school social-studies teacher from Dyersburg, Tennessee, a long-time Warren Commission critic who lectures widely on the subject. He loves to catch prominent figures saying things they can’t defend. [Recent targets include Gerald Posner ("Posner Follies") and Dr. John Lattimer ("The Great Thorburn Hoax"), articles about both of which can be found in David Starks's "The Assassination Web" at http://www.assassinationweb.com/issue4.htm .)] A few years ago, Milam turned his attention to the NAA data, and started writing articles accusing Guinn of duplicity at best and perjury at worst. He claimed that Guinn’s data show that MC bullets are actually much more heterogeneous than Guinn acknowledged, enough to invalidate his main conclusion about two separate groups. In Milam’s view, Guinn found only one group. Gone would the evidence for two and only two bullets. Gone would be the strong support for the single-bullet theory. Gone would be all the accolades heaped upon Guinn by Robert Blakey of the HSCA and many others. The Guinn who remained would be a charlatan in a scientist’s white coat. Never one to understate, Milam used phrases like “Guinn’s testimony is filled with inconsistencies and contradictions, and his laboratory data is shocking in both its quantity and its results,” “…he presented findings from his own laboratory tests which contradicted the very hypothesis which is the basis of his work: that fragments from the same Mannlicher bullet exhibit a high degree of homogeneity,” “…Dr. Guinn’s work represents, at best, an invalid scientific hypothesis based on inadequate research. At worst, it is a scientific charade meant to lend an aura of legitimacy to the single bullet theory,” and finally, “He raised serious questions about his own integrity.” Whew! [These quotes were taken from Milam's manuscript "Blakey's 'Linchpin': Dr. Guinn, Neutron Activation Analysis, and the Single Bullet Theory" of August 1994. This manuscript now appears at Starks's link given above, in two parts and seven appendices. Although it retains the message of the original 1994 document, some of the strongest statements have been deleted.]
Were Milam’s criticisms reasonable? My first reaction was to dismiss him as someone who was talking before thinking. But when I read his material carefully and checked Guinn’s data myself, I had to admit that Milam had found something that we all had missed. I tip my hat to Wallace Milam for daring to challenge the Guinn whom we all accepted uncritically because of his reputation.

The essence of Milam’s criticism
What Milam found was very simple. He noted that whereas Dr. Guinn said that the heterogeneity of Sb in MC bullets was 6% (recall the famous figure from above), his published data showed its heterogeneity to be far greater—so great that the two groups of fragments actually coalesced into one big, meaningless blob (my words). Guinn’s published data on heterogeneities consist of four samples from each of three MC bullets, from different production lots (see table in handouts). In the bullet from lot 6001, all four concentrations of Sb were quite similar, and had a standard deviation of 6% of the mean (just as Guinn had testified). But in the bullets from lots 6002 and 6003, one of the four samples differed greatly from the others. As a result, the standard deviations (heterogeneities) of those bullets were 36% and 29%, respectively. When the three standard deviations are combined, they average to 24%, a far cry indeed from the 6% that Guinn claimed. With a standard deviation of 24%, two groups of fragments that differ by only 30% are no longer separate—they overlap substantially.
How can someone claim a heterogeneity of 6% but publish 24%? Guinn’s 6% came from—he said—a larger set of analyses that he was then preparing to publish. But he didn’t show the data then, and he hasn’t published them in the 20+ years since. Guinn was not following the rules of science, which say that conclusions must be based on published, verifiable data. In effect, he was saying, “Trust me, fellas, it is 6%.” While trust has no real role in science, one can offer some latitude to a recognized expert, particularly in matters of little consequence. But the 6% was of huge consequence. It held the key to establishing two groups of fragments and thereby to supporting the single-bullet theory far more strongly than otherwise possible. It was, as Milam wrote, Blakey’s “linchpin” to the HSCA’s endorsing the single-bullet theory. And in the final analysis, that linchpin came down only to trust.
Is it 6% or 24%, two groups or one? The answer is not easy to find. The rest of this document shows that although it first appears strongly to be 24%, more-careful probing reduces it to 3% or so.
Consider first the unusual heterogeneities of MC bullets. They violate everything we think we know about normal materials. In well-mixed solutions containing impurities, larger samples ought to have smaller differences in composition because larger samples will afford more opportunities for regions of different composition to even out. WCC/MC bullets behave oppositely, however—their larger samples have the greater differences. Specifically, Sb in subfragments (masses of 1–40 mg) has heterogeneities of about 5% (FBI data), in fragments (10–50 mg) 9%, in Guinn’s quarter-bullets 8% (with the two outliers removed) and 24% with all 12 quarters counted, and in whole MC bullets (Guinn’s data) 90%. Why this backwards pattern?
The giveaway is Guinn’s data from his heterogeneity tests of the three bullets (Figure 18). The four fragments from bullet 6001 all fall reasonably close together, between about 1050 and 1250 ppm Sb. As noted above, however, each of the other bullets has three of the quarters close together and one outlier. In bullet 6002A, the outlier’s Sb is 358 ppm, nearly three times lower than the other three values of 880–980. In bullet 6003A, the outlier’s value is 667 ppm, about 1.5 times greater than the other three values of 363 to 441 ppm. Recall that samples of smaller size (fragments and subfragments) have Sb that is much less variable—5% to 9% only. It’s almost as though Sb either varies greatly or hardly at all—it varies greatly on the scales of entire bullets, but much less on smaller scales. This means that the vats of lead from which the WCC/MC bullets were made had not been completely mixed before the lead was poured into bullets.

Figure 18. Guinn’s results for antimony in quarters of three WCC/MC bullets.

    Such a situation is entirely plausible because the MC bullets were made by combining lead from various sources. At the Providence Conference in 1999, I suggested an analogy to making a marble cake at home. Into the bowl go the brown batter and the yellow batter. The circular mixing motion of the spoon produces whorls that remain distinct for a long time. Zones of brown adjoin zones of yellow. Within each zone the color is nearly constant; only right at the border does it change. The same basic situation in a vat of lead that began with streams of melted lead could easily produce zones with very different concentrations of impurities, invisible and consequently impossible to know when they were gone. The “width” of these zones need only be the order of the size of a bullet to produce the pattern of heterogeneities found in MC bullets. Many samples drawn from the vat (bullet-sized units poured into the mold) will contain some of each material, each better-mixed than the entire bullet would be mixed. Small fragments will generally be drawn wholly from one material, and would produce homogeneous subfragments. But one fragment may be very different from the other, and create the heterogeneities observed within the MC bullets.
    During the same discussion in 1999, W. Anthony Marsh, of Somerville, Mass., proposed an alternative, which he called the "nugget theory." A "nugget" was a place in the vat where the lead was not mixed thoroughly. As opposed to my extended zones, which presumably would have originated from streams of melted lead, Tony's nuggets would presumably represent melted lumps of lead that were not mixed with their surroundings. As seen below, Tony's idea probably represents reality better than mine, although both scenarios lead to zones of near-constant concentration that would be similar in size to each other.
    In a message of 19 March 2001, Larry Sturdivan, the HSCA guru on ballistics, offered a detailed scenario on how the Western Cartridge Company would have mixed its lead that was destined for Mannlicher-Carcano bullets. Here is his description, edited lightly:

      As lead is both valuable and hard to dispose of safely, the manufacturer will always recycle leftover core material from one manufacturing run into future bullets. Because the specifications for Mannlicher-Carcano bullets called for cores of unhardened lead, any leftover lead was candidate material. If a number of leftovers were tossed into the melting pot and melted without mixing, the distribution of silver, antimony, etc. would be multimodal. [Marsh's nuggets separated by my zones.] There would be a peak at each concentration represented by each hardened batch as well as different peaks in each batch of unhardened lead because of variation in the natural ores from which they were refined. With mixing, these peaks would broaden and move toward the average concentration. They would soon blend into a bimodal distribution with each peak proportional to the amount of hardened and unhardened lead introduced. With further mixing these two peaks would broaden and approach each other until they merged into a single, highly skewed peak. With perfect mixing the peak would become an impulse function at the average concentration. People do not realize how strong the law of diminishing returns is in this situation. Not only is perfect mixing impossible to achieve, but the amount of mixing required to reduce the heterogeneity by a fixed amount increases exponentially as the lead becomes better mixed. The Western Cartridge Company obviously stopped at an intermediate stage, probably as soon as they considered the batch to be considered "unhardened." Since the Mannlicher-Carcano specifications do not limit the concentration of antimony, this is an arbitrary point. You could not tell whether the actual distribution of antimony was bimodal or a skewed unimodal distribution without an excessive amount of testing. In unjacketed, semijacketed, and hollow-point bullets, the expansion properties are quite sensitive to the hardness of the lead, and most rifle bullets specs call for a specific hardness in the lead core. By contrast, WCC/MC bullets offered a rare opportunity to mix a lot of leftovers without having to keep track of the amount of antimony that would be present in the final batch.

    Figure 18a below shows a highly schematicized version of part of one of the WCC's vats of melted lead. Three "nuggets," each from a chunk of lead from a previous batch, are shown with concentrations of Sb denoted by C1, C2, and C3. Around the outside of each nugget are zones of decreasing or increasing concentration. Within each nugget, the concentrations will be nearly constant. The characteristic size of a nugget or a zone, while not known and certainly variable, will be on the order of the size of a bullet or a quarter, that is, millimeters to centimeters. A series of little particles drawn from inside a nugget or a zone would be nearly homogeneous in Sb (would have very similar concentrations), whereas the core of a bullet would be of a size that could fall completely within a zone or could be drawn from two zones (i.e., could be homogeneous or heterogeneous in Sb). These scenarios are shown in Figure 18a.

Figure 18a. Schematicized "nuggets" of lead with Sb concentrations C1, C2, and C3 within a vat of WCC/MC lead.

    Since Guinn’s data on heterogeneities are the only ones we have, we must base all our conclusions about separability of groups on them. What do his data show? That of 12 zones within bullets, only two differed markedly from the “pack.” The other ten zones varied from one another by only 8% on the average—about the same as Guinn’s anecdotal 6%. Thus for Mannlicher-Carcano bullets as a whole, 83% of the time (10 of 12 cases) we may expect bullets to produce highly reproducible fragments (heterogeneities of only 8%), and 17% of the time (2 of 12 cases) we will find much larger variations. Since variations of 8% will keep Guinn’s two groups of fragments separate to a very high confidence limit (between 95% and 99%), we may conclude that there is an 83% chance that the two groups are actually separate to the 95–99% confidence level, and a 17% chance that they are not separate.
The problem with this line of reasoning is that it is based on too few samples. Physicist Arthur Snyder has criticized it on these grounds (see his article on the subject), and he has a point.

Others join in trashing Guinn
Milam’s hue and cry against Guinn’s conclusions has been taken up by quite a number of others in the critical community, some quite vocally. One example is the article by Richard Bartholomew in JFK/Deep Politics Quarterly, Vol. I, #4, July 1966, pages 7–10. Bartholomew's message comes through clearly right  from its title, "Dial 'P' for Perjury," especially when linked with his sentences "Dr. Vincent P. Guinn's middle name is Perry. It may soon be 'Perjury.'" Bartholomew, who consistently refers here to nonconspiracists as "conspiracy deniers," perhaps in the derogatory sense of "holocaust deniers," follows up with statements such as "…one of the worst oversights committed by Warren Commission critics appears to be our failure to see that Dr. Vincent Perry Guinn committed perjury."
      Why all the strong words? Bartholomew sees bad stuff afoot. He begins by buying into the false story about Guinn's having previously worked for the FBI and the Warren Commission, a story that is debunked in the section "Vincent Guinn's neutron-activation analysis." He also bought the bogus argument that Guinn tested samples different from those run by the FBI. He also made a big deal over a few aliquots being discarded as radioactive waste, seeing something nefarious behind it: "Radioactive waste is not put at the curb on trash pickup day." He ends by questioning why all this has been allowed to "fall through the cracks of the major assassination literature."

      This piling-on attitude has in turn has swayed other critics on the sidelines. The result has been that Guinn and the NAA data have acquired a stained reputation that they do not deserve. The next section attempts to correct the situation.

Ahead to Problem of Tight Groupings
Back to Guinn's NAA
Back to NAA and the JFK Assassination
Back to NAA