The FBI tries NAA on the fragments

It took a push from the AEC
      It is interesting to compare the FBI’s zeal for spectroscopic analysis with their sluggishness toward neutron activation. The basic spectroscopic analyses were completed within twenty-four hours of the assassination because on 23 November 1963 J. Edgar Hoover sent a report to Dallas Police Chief Jesse E. Curry acknowledging receipt of the various pieces of evidence and stating “The lead metal of Q4 and Q5, Q9, Q14 and Q15 is similar to the lead of the core of the bullet fragment, Q2.”[1] By contrast, the FBI only got around to using NAA on these fragments the following May—more than five months later.[2] Why did they wait so long? No one knows. It is certain, though, that the FBI was prodded repeatedly by the Atomic Energy Commission beginning shortly after the assassination. The AEC was apparently very interested in being of service, and in demonstrating to the public what the powerful new technique of neutron activation could do. Correspondence unearthed by Harold Weisberg[3] reveals a series of letters between the AEC, the FBI, and the Warren Commission. The first letter, written 11 December 1964 (which I think is a misprint for 1963), was sent from Paul C. Aebersold, director of the AEC’s Division of Isotope Development, to Herbert J. Miller, chief of the FBI’s Criminal Division. According to Weisberg, Aebersold noted that for several weeks the AEC had been discussing with various people in the FBI “what additional light nuclear activation might shed” on the assassination, and that the AEC had first offered its assistance “within less than 24 hours of the assassination” (an interesting redundancy). Apparently the FBI wasn’t responding, for Aebersold stressed that “We believe it is not too late to outline what may yet be done.”
      Aebersold proposed trying to link the fragments to the unfired bullet found in Oswald’s rifle: “…it may be possible to determine by trace-element measurements whether the fatal bullets were of composition identical to that of the purportedly unfired shell.” He also thought NAA might link Oswald to the attempted assassination of General Walker: “If the same batch of ammunition was used in the sniper bullet fired at” resigned ultra-rightist General Edwin Walker, “the method might show a correlation.” Aebersold thought some of the clothing might be analyzed as well: “Other pieces of physical evidence in the case, such as clothing…might lend themselves to characterization by means of their trace-element levels.” He revealed how eager the AEC was to have a go at the Kennedy case: “…we wish to indicate our eagerness…Our work leads one to expect that the tremendous sensitivity of the activation analysis method is capable of providing useful information that may not be otherwise attainable.”
      The FBI and the AEC apparently began to cooperate in early January 1964. Shortly after 2 January 1964, AEC chairman Glenn T. Seaborg wrote to the Warren Commission that “…we are, however, in cooperation with the” FBI lab on “a very sensitive method of trace element analysis…may be of value in further corroborating evidence already in hand by the” FBI. “This work is being done at our Oak Ridge National Laboratory…”
      Again according to Weisberg, Rankin sent J. Edgar Hoover on 7 January 1964 a copy of Aebersold’s letter to Miller and asked Hoover whether he wanted to accept the AEC’s offer of help. Hoover replied that the FBI laboratory was well acquainted with NAA, and that they were already working with the AEC “applying this technique to certain phases.” He promised results as soon as they were available. Apparently this early cooperative work involved analyzing the paraffin casts from Oswald’s nitrate tests.
      There is conflicting evidence whether the AEC was analyzing any bullet fragments during January 1964. As noted above, Guinn stated that Gallagher analyzed them in May 1964.[4] According to Weisberg, FBI Agent John W. Kilty also swore to that effect.[5] But on 27 January 1964 Rankin told an executive session of the Warren Commission that the AEC was already analyzing the fragments[6]: 

      Now, the bullet fragments are now, part of them are now, with the Atomic Energy Commission, who are trying to determine by a new method, a process that they have, of whether they can relate them to various guns and the different parts, the fragments, whether they are a part of one of the bullets that was broken and came out in part through the neck, and just what particular assembly of bullet they were part of.
      They have had it for the better part of two and a-half weeks and we ought to get an answer.

      Weisberg concludes his review of correspondence between the agencies by noting that “the government now alleges” that this description by Rankin misled the Commission. That would be consistent with information on the 70 pages of raw NAA data obtained by Dr. John Nichols of the University of Kansas Medical School under a FOIA request and shared with Dr. Guinn, to the effect that the FBI/AEC analyzed the bullet fragments at Oak Ridge in May 1964.
      Whatever the original motivation for applying NAA to the JFK assassination, it eventually provided extremely strong evidence on two questions where reliable evidence was needed: the single-bullet theory and the source of the bullets and fragments. The NAA evidence on these questions is some of the strongest evidence of the entire assassination, if not the strongest. It solidifies several of the other lines of physical evidence into a powerful organic whole.

How the FBI did its NAA
      Some details of the FBI/AEC neutron-activation analysis were described by Guinn.[7] Apparently the FBI’s John Gallagher analyzed the samples himself at Oak Ridge, presumably under the guidance of experienced AEC personnel. Mortal Error, by Bonar Menninger, alleges something different, however. It states that Frank Dyer and Juel Emery of Oak Ridge analyzed the fragments in 1964.[8] Presumably both reports are right—Gallagher did the actual work with Dyer and Emery guiding him. Gallagher would have been actively involved in the analysis no matter what, since he was at Oak Ridge representing the interests of the FBI, which had ultimate responsibility for the samples and the data. Gallagher knew Dyer and Juel, because he had worked with them several months earlier analyzing Oswald’s paraffin casts by NAA.[9] Gallagher was also familiar with NAA, because in September and October 1962 he had attended a training course on the subject at Oak Ridge.[10]
      Guinn and Nichols later provided additional details about the FBI’s NAA work: “The FBI analyses were carried out during May 1964, using thermal-neutron fluxes of 5 to 7 x 10¹³ n cm^-2 s^-1 (Oak Ridge Research Reactor); irradiation times of 20 s and 2 min; decay times of 20 s, 2 to 3 h, 17 to 30 h, and 10 to 11 days; and counting times of 40, 200, and 300 s with a 4- x 4-in. Na(Tl) multichannel γ-ray spectrometer. Three elements were detected (Sb, Ag, and Cu), but only Sb and Ag were measured: Sb via the 564-keV γ of 2.80-day ¹²²Sb and via the 603-keV γ of 60.4-day ¹²⁴Sb, Ag via the 658-keV γ of 24.4-s ¹¹⁰Ag. Due to the poor resolution of the NaI(Tl) detectors, the 511-keV γ contribution of the 12.80-h ⁶⁴Cu had to be subtracted from the 564-keV γ peak of ¹²²Sb, or eliminated by decay and ¹²⁴Sb measured instead. All of the Dallas specimens were generally somewhat similar to one another in their Sb and Ag concentrations, but there was a wide spread in the values for individual samples and among the group of samples. For example, the 17 values obtained for various portions of the “Connally stretcher” bullet averaged 837-ppm Sb, but ranged all the way from 636 to 1135 ppm.”[11]
      For measuring silver, John Gallagher used the 24-second ¹¹⁰Ag isotope. For measuring antimony, he used two longer-lived isotopes, 2.8-day ¹²²Sb and 60-day ¹²⁴Sb. (A peak of ¹²⁴Sb is shown for the a sample of urban aerosol in Figure 2.) The detailed nuclear properties of these three radionuclides are shown in Table 10. Because of the difference in half-lives, each fragment was irradiated for short and long periods.

Table 10. Nuclear properties of the radionuclides used in the FBI’s NAA.

      The suite of bullets and fragments that the FBI analyzed by NAA is not quite the same as they analyzed by emission spectroscopy. Table 11 shows the NAA list, including the weights of the replicate analyses.

Table 11. Bullets and fragments analyzed by the FBI with NAA, including replicates.

      Note that the third column refers to the original total weight of the fragment as received by the FBI, not the weight remaining after analysis by emission spectrography. The latter weights are not available. This table shows that if Q4 and Q5 are considered as a single sample, each of the five samples was analyzed in 3–5 replicates. Also note that at most, one-third of each sample was used for this analysis. Generally, about one-quarter of the original sample was used.
      The FBI analyzed each of the replicate samples once for silver and four times for antimony. In his 1979 Analytical Chemistry article,[12] Guinn presented his recalculations of the FBI’s results, but in a table that only gives averages for the replicates. Guinn never mentioned replications, and referred to each set of replicates merely as “the sample.” Consequently, the reader thinks that the FBI analyzed only single samples.

The FBI’s results and conclusions
      Table 12 shows Guinn’s recalculated average FBI values (which agree with the FBI’s values, as far as I can tell). Again, recall that each concentration of antimony (Sb) in each of the four runs is itself actually the average of 3–5 measurements of separate subfragments, whereas each concentration of silver (Ag) is only a single value. Both kinds of data in this table look the same.

Table 12. The FBI’s results for silver and antimony in bullets and fragments 
(concentrations in ppm).

      To illustrate the range of individual values for antimony, Table 13 lists concentrations for the 21 subsamples that were measured in the FBI’s Run 4. These data are part of a tabulation of results appended to an internal FBI memo of 6 July 1964, from R. H. Jevons to a Mr. Conrad, that describes the FBI’s NAA results.[13] This important memo will be discussed in detail below. These results will be discussed more in the section on Guinn’s NAA.

Table 13. Individual determinations of antimony in the FBI’s Run 4

*Original weight was 3.34 mg. Its surface was cleaned for a second time
 and the resulting fragment of 1.16 mg was reanalyzed, with the result shown here.
**Guinn’s table in Analytical Chemistry gives an erroneous value
 of 54 ppm for this standard deviation.

      Note that the resulting standard deviations of antimony’s concentration are quite small—only about 5%. This standard deviation is a very important number, for it represents the actual measured variability of antimony in the specimens analyzed, i.e., the actual heterogeneity of the bullets and fragments from the assassination plus the analytical uncertainty. Thus the heterogeneity must be smaller than 5%. The small observed heterogeneity plays a pivotal role in justifying the separation of the two groups, as discussed in a later section.

Public vs. private conclusions
      It seems that the FBI held separate public and private conclusions on the results of their NAA analyses. On the one hand, they left the impression that they could not group the fragments because they were too similar in composition, as implied by J. Edgar Hoover’s letter of 8 July 1964 shown above. With respect to the NAA analyses, its key sentence is:

      While minor variations in composition were found by this method, these were not considered sufficient to permit positively differentiating among the larger bullet fragments and thus positively determining from which of the larger bullet fragments any given small lead fragment may have come.

      Guinn later reinforced this impression when he described the systematic error that spread the data for each fragments over a wide range and blurred the fragments into one another. But a careful reading of the full passage from which the above segment was drawn shows, however, that the letter is more specific than that—it states only that the [two] larger fragments could not be differentiated (stretcher and front seat—Q1 and Q2). It does not say whether the smaller fragments could be differentiated. Figure 4 below reveals, though, that the smaller fragments, with the possible exception of Q9, are just as differentiable as the larger fragments are. Hoover’s remark thus appears odd. But no matter how these paragraphs are read, their clear message is that the FBI was unwilling to identify groups of fragments or origins of individual fragments.
      The plot thickens when we realize that the FBI did indeed take its interpretation well beyond Hoover’s letter. Hoover drew his letter from text recommended by R. H. Jevons of the FBI, included as an attachment to the abovementioned memo of 6 July 1964 to Mr. Conrad of the FBI. The purpose of the memo was to report the FBI’s NAA results and their interpretation of them. The text of the memo contains only two paragraphs, the first of which became Hoover’s letter. The second paragraph is key here. It amplifies the first paragraph by noting that the FBI had indeed found two groups of fragments—the very groups that are obvious from the plot—but that they considered them probabilistic rather than “positive.” Because of the extreme importance of this memo, we reproduce it in its entirety:

      As previously reported to the Commission, certain small lead fragments uncovered in connection with this matter were analyzed spectrographically to determine whether they could be associated with one or more of the larger bullet fragments and no significant differences were found within the sensitivity of the spectrographic method. Because of the higher sensitivity of neutron activation analysis certain of the small lead fragments were subjected to neutron activation analyses and comparisons with the larger bullet fragments. While minor variations in composition were found by this method, these were not considered sufficient to permit positively differentiating among the larger bullet fragments and thus positively identifying from which of the larger bullet fragments any given small lead fragment may have come.
      Generally speaking, the small lead metal particles fell into two categories. In one category fell the small lead fragment from the arm of Governor Connally and the lead from the almost intact bullet recovered from the stretcher. In the other category fell the lead particles from the head of President Kennedy, the lead fragments from the rear floor of the car and the lead from the mutilated bullet fragment from the front seat cushion. While there is a probability that the fragment from [next words apparently blacked out; probably “the Governor’s arm”] came from the whole bullet rather then [sic] from the mutilated bullet and that the fragments from the President’s head and from the floor of the car could have come from the mutilated bullet fragment from the front seat cushion, it remains a probability and does not permit a positive finding or statement that any given small lead fragment did in fact come from one of the bullets to the exclusion of the others. Accordingly, the proposed letter to the Commission reports that the findings do not permit a positive determination as to the particular origin of any of the small lead metal fragments. A tabulation of the results of the analyses is attached hereto. A letter is attached for forwarding to the President’s Commission.

    The FBI had even graphed the concentrations of antimony in the five fragments and shown them with confidence limits of 95% and 99%. The graph (courtesy of NARA and Mr. W. A. Marsh), shown as Figure 3, removes any doubt that the FBI's NAA results group the particles into two groups or that the FBI knew this. Q1 and Q9 group clearly, as do the remaining Q2, Q4,5, and Q14.

Figure 3.  The FBI's graph of the concentration of antimony in the five basic fragments.

      In other words, the FBI had categorized the fragments into two groups: the Governor’s arm and the stretcher versus the President’s head, the rear floor, and the front seat. These are just the groups seen in Figure 4 below. Of course it could not be stated with certainty that a small fragment from one group originated from the bullet in that group—the very nature of the analytical and matching procedure precluded that. (Recall from above that equality of composition does not equal identity of origin.) Jevons’s memo and Hoover’s letter were thus correct on a technicality—the NAA results did not “positively determine” origins because they could not. In other words, their strong-looking statement says exactly nothing.
      The FBI should have realized this limitation of similarity measurements before they began the neutron activation, because it is a fundamental principle of any chemical matching technique. If they realized it only later, they are less competent than generally supposed. If they realized it from the beginning, they appear to have used it as an excuse to avoid commenting on the fragments. Either explanation raises uncomfortable questions about the FBI. Maybe they really didn’t want to say anything unless they could be certain of it. But if so, why did they start down a road that led only to uncertain results? Because the AEC pushed them too much to resist?
      No matter how the FBI’s silence is viewed, they highly misled the Commission by failing to report their preliminary groupings. They also gave themselves an undeserved black eye by leaving the impression that they couldn’t handle the neutron activation. The FBI clearly knew a lot more about those fragments than they let on. Why did they hide it?

The FBI’s NAA was far better than portrayed
      Over the years, the distinct impression has been created that the FBI’s 1964 neutron-activation analysis was preliminary and indeterminate, and that Guinn’s 1977 analysis was thorough and decisive. From my initial readings on the subject, I certainly came to believe this. But I eventually came to realize that the FBI’s analysis was much better than generally appreciated, and that Guinn’s analysis was less thorough than portrayed. In many ways that mattered most, the FBI did it better than Guinn.
      To be sure, the FBI was forced to use the lower-resolution NaI(Tl) gamma-ray detectors for their 1964 analysis because the superior Ge(Li) detectors were not yet available in quantity. But that didn’t significantly degrade their results.
      To be sure, the FBI couldn’t or wouldn’t take the time for proper background studies on WCC/MC bullets. For whatever reason, Gallagher analyzed only two other WCC/MC bullets, one from each of two production lots. He analyzed no bullets of any other types. But that was no hindrance, because the two bullets he chose were just as different as Guinn later found.
      To be sure, the FBI limited themselves to measuring silver and antimony in the fragments. They detected copper but did not quantify it, and they failed to explore the potential of other elements available from neutron activation. But silver and antimony are the two most useful elements that can be determined in bullets by NAA, so that was no hindrance.
      The FBI did the important stuff right—all fragments analyzed in replicate, results graphed, confidence limits calculated, fragments tentatively classified into two groups—but then they incomprehensibly refused to make any of this public. They had the right answer in their grasp, but stopped just short of clinching it. Then they sealed away their work so securely that only now, thirty years later, do we realize that it holds the key to much of the assassination. For thirty long years, dedicated assassination researchers have suffered needlessly because the FBI refused to tell what they found in 1964. Why did they turn suddenly silent? Why did they make it almost impossible for future students of the case to piece together their pioneering NAA work and its great significance? What was so important as to cause someone in authority to hide all their findings? To me, this is one of the biggest mysteries of the assassination.

Guinn finds a systematic error
      Now back to the results in Table 12. In Analytical Chemistry, Guinn described how he obtained the FBI’s results, how he recalculated them, and how he interpreted them in the light of his results. He correctly noted that (1) the data for silver are not sufficient to distinguish groups of fragments; (2) if the four sets of data on antimony are lumped together, groups of fragments cannot be distinguished; but (3) if the antimony results from each run are plotted separately, two groups of fragments are seen consistently: Q1/Q9 versus Q2/Q4,5/Q14. Those groups agree with Guinn’s results, with the single-bullet theory, and with several other aspects of the validated physical evidence on the assassination.
      Figure 4 shows the FBI’s four sets of results for antimony in the fragments. Note how nearly parallel the results from the four runs are. Guinn was almost certainly correct in ascribing these differences to some systematic error of preparing the standards, irradiating, positioning the samples during counting, or calculating the results. Thirty years after the fact, though, there is no way to know exactly where the errors crept in.

Figure 4. The FBI’s four sets of results for antimony in the JFK fragments.

      Guinn also correctly stressed that these systematic errors don’t matter terribly, for once identified, they can be compensated for. From Figure 4 it is obvious that in each run, fragments Q1 and Q9 (stretcher bullet and Connally’s wrist) grouped together at high concentrations, and fragments Q2, Q4,5, and Q14 (front seat, JFK brain, rear carpet) grouped together at lower concentrations. We consider the statistics of these groups later—for now it is enough to note their existence.

A second systematic error
      While examining the FBI’s NAA data in Table 13, I thought I noticed an inverse relation between concentrations of antimony and weights of fragments: heavy fragments had systematically lower antimony than light fragments did. To check this out, I made several plots of concentration versus weight, and confirmed this effect.
      The simplest of the plots is Figure 5. Here the average concentration of antimony in each of the five basic fragments (as reported by Guinn in Analytical Chemistry) is plotted against the total mass of the fragment, for each of the FBI’s four runs. This is the way that a mass effect would be sought most directly from Table 12. (Individual runs had to be plotted separately so that systematic differences between the runs would not interfere with detecting a mass effect.) An inverse mass effect is clearly seen, most regularly for Runs 1 and 2, but for the others as well. This figure shows that the two fragments in the high-antimony group (Q1 and Q9) are just the ones with the smallest mass. If this figure were all we knew about the system, we might conclude that the two groups of fragments were just an artifact of the weights of fragments comprising them. If the concentration of antimony is plotted against the average mass of subfragments in a fragment (the total mass of the fragment divided by the number of subfragments comprising it), a similar pattern is seen (Figure 6), but with the positions of Q4,5 and Q14 reversed. Again, the groupings would appear to be artifactual.

Figure 5. Average antimony in the five basic fragments vs. the total mass of the fragments.

Figure 6. Average antimony in the basic fragments vs. the average mass per fragment analyzed.

      The groups are not an artifact. This can be seen from Figure 7, which shows the concentrations of antimony in the 21 individual subfragments versus the mass of the subfragments. The FBI’s two groups of fragments are plotted separately here, because they had different mean concentrations of antimony.

Figure 7. Antimony in subfragments vs. mass of subfragments, all data.

      Figure 5 reveals that both groups of fragments show a regular inverse trend of concentration with mass, with the high-antimony group (Q1, Q9) falling systematically above the low-antimony group. For both groups, the highest concentrations are associated with fragments whose mass is less than about three milligrams. For masses of fragments greater than about five milligrams, the effect of mass is very small. The largest effect is seen for the Q1/Q9 group, because it has more fragments of smaller mass than the low-antimony group does.
      Could the difference in average antimony between groups be explained simply by one having more light fragments that the other, and therefore a higher average antimony? If so, the two groups would be apparent rather than real—an artifact of mass of fragments rather than any inherent characteristic of the lead in the bullets. Under this scenario, no fragments could be differentiated by mass, and all hope of supporting the single-bullet theory chemically would collapse.
      The answer is a clear “no,” because the two groups of fragments fall on two parallel lines, not one line. Note that with increasing mass of fragments, the line for the Q1/Q9 group remains clearly above the line for the other group. The groups maintain a separation of 100–150 ppm of antimony wherever they can be compared, which is from about 1–7 milligrams of mass. Thus the groups are clearly different, even though they both show the mass effect.
      Lastly, Figure 8 examines whether a mass effect is displayed by the subfragments within each fragment. Like Figure 7, it plots the antimony in each subfragment, but separately for each of the five large fragments. Although the plot is a bit messy, an inverse effect of mass can be seen for Q1, Q4,5, and Q2. Q9 shows no effect because the masses of its three particles are all about the same; Q 14 simply shows no effect. Thus the inverse mass effect is seen in three of the four particles whose masses vary enough to make an effect detectable.

Figure 8. Antimony in subfragments vs. mass of subfragments, each fragment plotted separately.

      What caused the mass effect? I don’t believe it represents a fundamental property of the fragments because I can’t think of anything that would cause it. The most likely explanation would be some artifact of the FBI’s analysis, some problem that made small amounts of antimony seem disproportionately large. An obvious candidate would be an improper background subtraction in calculating the area of the photopeaks. If insufficient background counts on either side of the peak were subtracted from the gross area of the peak, large peaks would remain largely unaffected but small peaks would appear too large, just the effect observed here. Of course, it is nearly impossible to determine whether this explanation is correct, now that so many years have elapsed since the original analysis.
      Does this inverse mass effect matter? Not really. Figures 7 and 8 show that both basic groups of fragments show the effect, and that they clearly differ by 100–150 ppm of antimony. Thus with or without the mass effect, the same groupings of fragments are found. The inverse mass effect does not prohibit us from differentiating samples Q1 and Q9 from samples Q2, Q4,5, and Q14.

Ahead to Guinn's NAA
Back to NAA Offers Better Prospects
Back to NAA and the JFK Assassination
Back to NAA