Scientific Testimony: An Online Journal

People vs. Marshall - The Legal Story


PLEASE NOTE: If you wish to view some of the legal papers that were filed in this case, you need Adobe® Acrobat® Reader. If you do not have this program, you may download the free viewer. If you would like a copy of any of these documents in Microsoft Word format, you may contact the editor.

A prominent criminal defense lawyer from Los Angeles asked me to look at the DNA evidence against his client, a young man named Sammy Marshall. Marshall and a co-defendant were charged with kidnapping and rape. If convicted, Marshall faced life imprisonment. The prosecution's case against Marshall depended almost entirely on the DNA evidence, I was told.

Genetic Design, a commercial laboratory in North Carolina, had performed the DNA test using a procedure called RFLP analysis. [For background on RFLP analysis, see the DNA primer in TUTORIALS]. According to the laboratory report, a vaginal sample from the victim showed a DNA banding pattern consistent with two men: defendant Sammy Marshall and the co-defendant. In copies of the autorads I could clearly see a pattern consistent with the co-defendant. But I could not see the bands matching Marshall. And I thought I could see faint "bands" that did not match Marshall's. [Copies of the autorads are shown in the companion article to this one, Examiner Bias in Forensic RFLP Analysis].

The laboratory analyst had made felt-tip pen marks on the autorads to indicate where he thought he saw bands. These marks corresponded to Marshall's DNA profile, which was visible on the same autorad. I wondered whether the analyst's knowledge of Marshall's pattern had led him to "see" the same pattern in the vaginal sample.

Psychologists have documented people's tendency to see what is expected in ambiguous stimuli. The danger of "examiner bias" has led most academic scientists to use "blind" procedures when making subjective judgments. Forensic scientists, by contrast, seem to reject "blind" procedures. They like to know as much as possible about a case before making an interpretation. I began to suspect that "examiner bias" might have led to a false incrimination of Sammy Marshall. I agreed to enter the case as co-counsel for Marshall.

My initial strategy for defending Marshall was to challenge the admissibility of the DNA evidence. The legal arguments I advanced are set for in defendant's Motion to Exclude DNA Evidence. The motion concedes that forensic RFLP analysis, when performed correctly, is admissible under California law. It argues that the DNA evidence against Marshall should nevertheless be excluded because the laboratory failed to follow correct scientific procedures. A key claim is that the laboratory failed to use a reliable, objective methods for "scoring" the bands in the DNA profile and instead used subjective procedures that were biased against the defendant.

I also hoped that the prosecuting attorney might be persuaded to drop the DNA evidence against Marshall when he saw (or failed to see) the supposedly incriminating "bands." I spoke to the prosecutor and explained the problem of "examiner bias," but he was not persuaded. He contended that DNA tests are objective and involve no element of human judgment. He explained that the Marshall autorads were "scored" using a BioImage machine, an automated imaging system that was "completely objective because it is controlled by a computer."

If the scoring procedures were objective, I reasoned, then it should be possible to replicate Genetic Design's findings. I asked around and learned that Aimee Bakken, a molecular biologist at the University of Washington, had access to a BioImage machine like the one that Genetic Design had used to "score" the autorads. Professor Bakken generously offered to help. Using her machine, and copies of the autorads, she tried to replicate the findings of the forensic laboratory. She failed. In the vaginal lane she failed to detect some of the bands that supposedly incriminated Marshall and detected other bands that did not incriminate Marshall. I thought that her report might sway the prosecutor, but it did not.

The prosecutor contended that Professor Bakken's findings were meaningless because she had scored second-generation copies of the autorads that contained less detail than the originals. I suggested that the prosecutor send the original autorads to Professor Bakken and let her re-score those. But the prosecutor refused on grounds that the state's evidence can never be released to a "defense expert" who might lose it, destroy it, alter it, or engage in other nameless unethical acts with it. The original autorads must stay at Genetic Design, he said. I next asked if I could send Professor Bakken to Genetic Design so that she could re-score the original autorads there. The prosecutor refused this request as well because Genetic Design was unwilling to allow an outside expert to use its BioImage machine.

At this stage I decided it was time to litigate the issue. In my moving papers I argued that due process and the state discovery statute require the state to allow a defense expert to "re-score" the autorads at the forensic laboratory. The district attorney's response argued that there is no such legal requirement and that various disasters might befall the forensic laboratory if a "defense expert" were allowed to lay hands on the forensic laboratory's computer equipment. My reply contested the district attorney's assertions. After a rather brief hearing on the matter, Judge Mary Ann Murphy ruled against me and in favor of the prosecution.

Having lost the first round of the discovery battle, I tried a different approach. I had learned that the BioImage machine works by using a scanner to create a digital image of the autorads. This image is then "scored" by computer. My new plan was to obtain a copy of the digital image that was created at the forensic laboratory so that Professor Bakken could re-score it. If Professor Bakken failed to replicate the incriminating findings using the digital image, the problem could no longer be attributed to a bad copy. A digital image and its copies should be identical.

Obtaining the digital images proved to be a challenge. I asked Genetic Design to score the autorads a second time using the BioImage machine. I asked that the analyst refrain from any "operator overrides" of the machine's band-scoring determinations. I also asked that a copy of the digital image of each autorad be transferred to a floppy disk and sent to me. At first Genetic Design appeared to cooperate. The autorads were re-scored. Michael DeGuglielmo, the Director of Forensic Testing, reported that the BioImage device had confirmed the original findings which incriminated Sammy Marshall. But he said that the image files could not be copied to standard floppy disks, as I had requested, but required large capacity optical disks.

When I sought information on the type of optical disks (so that I could find an expert with appropriate equipment to read them), I ran into a brick wall. DeGuglielmo suddenly refused to communicate with me except through the prosecutor. The prosecutor became uncooperative as well. I was getting nowhere with my efforts to obtain additional discovery and faced increasing pressure to take the case to trial with what I had.

Then I had a stroke of luck. I happened to attend a professional conference on forensic DNA testing. At the conference a number of commercial firms exhibited their products. One of the exhibiters was the BioImage Company, manufacturer of the machine used to score the Marshall autorads. I chatted with a representative of the company and mentioned my disappointment that image files produced by the machine could not be downloaded to standard floppy disks. The representative reacted with surprise. "Of course you can put the images on floppy disk," he declared. At that moment who should walk into the room but Michael DeGuglielmo, who had written a letter stating that floppy disks could not be used for that purpose. I summoned DeGuglielmo and explained to him (and various on-lookers) that he had been mistaken and that floppy disks would easily hold the image files. It was an "Annie Hall moment" in which the expert from the BioImage Company came forth to assure him that I was right about the suitability of floppy disks. DeGuglielmo seemed less than delighted to receive this information. He made several unpleasant comments about lawyers, but he agreed to send me the images on floppy disks.

Over a month later I still had not received the floppy disks. Genetic Design was also withholding other discovery materials that I had requested. Consequently, I filed a Motion to Compel Discovery in which I accused the DA and Genetic Design of stonewalling. The DA filed a response, denying the charge. In another stroke of luck, Judge Murphy was replaced by a new judge. At the hearing on the motion to compel, the new judge listened carefully to my arguments. He then signed a series of orders directing the prosecutor to produce the image files and various other items that had been withheld.

Yet the image files still were not produced. When I called Michael DeGuglielmo several weeks later to find out when I might receive them, he told me that the image files no longer existed; they had been erased from Genetic Design's hard drive. Based on this revelation, I drafted and filed a Motion for Sanctions for Intentional Destrucation of Evidence. On the day this motion was to be argued, the prosecution produced an affidavit from DeGuglielmo declaring (surprise, surprise) that he had been mistaken and that the image files had not been erased after all.

By this point I had become so deeply suspicious of Genetic Design and of Mr. DeGuglielmo that I doubted I could trust the image files he produced. What assurance did I have that the image files would accurately reflect the original autorads (beyond the word of an individual who had, in my opinion, proven less than reliable)? Fortunately, I had been able to persuade the new judge to order Genetic Design to score the autorads yet another time and to do it while I and a defense expert watched.

The trip to Genetic Design (to see the autorads scored) was a watershed for the Marshall case. I was accompanied by Professor William Shields of SUNY-Syracuse. Shields is a superb and impartial expert who has testified for both prosecution and defense in DNA cases. As Dr. Shields and I looked on, Michael Deguglielmo and an assistant "re-scored" the autorads. In order to detect the faint "bands" that incriminated Marshall, they had to increase the sensitivity of the BioImage machine to the point that it detected many other bands that did not match Marshall's. They then used "operator overrides" to cause the computer to ignore all of the faint "bands" except those that were adjacent to the felt-tipped pen marks that corresponded to Marshall's banding pattern. Some of the "bands" that were ignored had higher optical densities than the bands scored as matching Marshall's profile. When asked to justify the operator overrides, DeGuglielmo said he could "tell by looking" which "bands" detected by the computer were true bands and which were not. Thus, the claim that the autorads were scored "objectively by computer" was revealed to be a fraud.

Other problems became apparent as well. In some instances it appeared that the computer was detecting the felt tip pen marks adjacent to the lanes and counting those as bands.

The most serious problem, however, was that one of the bands detected and scored in the vaginal lane differed by over 9% in measured size from the corresponding band in Sammy Marshall's profile. By Genetic Design's own matching rules, this finding excludes him as the source of the band in the vaginal sample.

At this point, the prosecution offered Marshall a plea bargain. The rape charges would be dropped and he would plead guilty to charges related to the kidnapping. In return, he would be receive a prison term of approximately four years (beyond the time he had already served while awaiting trial). The alternative to this bargain was to proceed to trial where, if acquitted, he would go free and, if convicted, he would receive a life sentence. Marshall chose to accept the plea bargain.

Several lessons can be drawn from this story. First, it should be clear that forensic DNA test results need careful scrutiny. In this case, results that were portrayed in a laboratory report as powerfully incriminating turned out, on close examination, to be ambiguous or even exculpatory. The claim that this DNA test was "objective" was revealed as a fraud. The supposedly objective "band sizings" varied significantly depending on who was doing the computerized "scoring" (and who was watching while they did it). Whether the test results included or excluded Sammy Marshall as a rapist depended, ultimately, on a subjective judgment. And this judgment was made in the first instance by an analyst who was not "blind" to the expected result, who knew what pattern he was looking for, and who made sure he found it by engaging in unreported "overrides" of the computer's scoring determinations. The laboratory then did its best to cloak this "fudging" in a mantle of scientific objectivity. Indeed, the claim of scientific objectivity was used to help shield the test results from scrutiny that eventually revealed them to be anything but objective.