DNA Testing Problems

Houston Police Department DNA/Serology Laboratory

Virginia Division of Forensic Sciences

North Carolina

Washington State



Failed DNA Controls in the Leon Winston Case (November 27, 2005)

by William C. Thompson

"Random sample controls" have been a distinctive feature of the Virginia Division of Forensic Science's protocol of DNA testing.  According to the lab's protocol:

A random control will be run with each extraction to serve as an extraction, amplification, and a typing gel migration control.  The random control is a previously analyzed convicted offender sample for which the identity of the sample is known to the Forensic Biology Program Manager and Laboratory Directors.  This sample serves as an internal laboratory control since the DNA profile is not known to the examiner/analyst and must be verified by the Forensic Biology Program Manager or the Laboratory Director prior to the sizing data being considered acceptable.  If a sample must be re-extracted a new sample must be extracted along with the sample.  (Section 1.1.12)

After developing a profile for the random sample control, the analyst searched the profile against the CODIS database to see if a "hit" could be obtained to the right person.  The "hit" confirmed the accuracy of the lab's procedures by showing that DNA from someone already in the database could reliably be typed and matched back to that same person.   

At least that was the theory.  Laboratory notes from the case of Virginia v. Leon Winston (see links below to copies of these notes) suggest that in practice the lab took a cavalier approach to the random sample controls.  Getting an incomplete profile, or even the wrong profile, apparently was not seen as a problem.    

If Your Control Fails, Substitute Another

On page 44 of the lab notes, the profile of one random control (STR2001-0057), as typed by analyst Nicole Harold, is shown in Lane 5.  No results were observed for TPOX and results for PentaD and CSF1PO are crossed out and marked "Inc."  Page 50 of the lab notes show that a second reader of the gel (initials "PJT") also marked the results for PentaD and CSF1PO "INC."   

Was this incomplete profile considered a problem?  Apparently not.  When Harold searched the profile against CODIS, she entered a genotype for CSF1PO even though both readers had indicated this locus was inconclusive.  A handwritten note on the State Match Results Report (lab notes page 51) says "CSF1PO results from another gel."   

The stated purpose of the random sample is to serve as a control for "extraction, amplification and gel migration."  To substitute data from another gel obviously defeats that purpose.  An analyst who says, in effect, "my control failed in this experiment, so I'll use data obtained on the control sample in another experiment" has failed to grasp an essential element of the scientific method.  That logic would fail to pass muster at a junior high school science fair.

If you are wondering what Harold would have done if she had obtained incorrect results from a random sample, rather than merely incomplete results, read on. 

That situation arose on a second gel run in the Winston case.  On page 102 of the lab notes, the profile of a second random control (STR2001-0046), as typed by Nicole Harold, is shown in lane 5.   The second reader's typing of the same sample is shown on page 110 of the lab notes, also in Lane 5.  Both Harold and the second reader reported a complete profile.  But the State Match Report (lab notes page 111) shows that Harold searched a different profile against CODIS.  For locus PentaD, the random sample was typed as a 9,12, but the "type" Harold searched against CODIS was a 12.  For locus CSF1PO, the random sample was typed as 12,12, but the genotype compared to CODIS was 11,12.  A handwritten annotation on the lab notes explains: "CSF1PO and PentaD results from other gel."  So Harold must somehow have known that the random sample had been typed incorrectly on the gel in the Winston case.  She responded to the failure of this scientific control by again substituting data from another gel (apparently from another case).  

Needless to say, by any conventional scientific standards the failure of a positive control requires re-running the assay or experiment in which the control failed.  In other words, the DNA testing in the Winston case should have been treated as invalid and re-run.  

Did the Independent Reviewers Notice This Problem?

The Winston case was one of 150 cases recently reviewed by an independent scientific panel appointed by the Governor of Virginia in the wake of the scandal over the Earl Washington case (see ASCLD-Lab Review).   The panel, chaired by Arthur Eisenberg, issued a report on September 12, 2005.  Although the independent panel's report discussed the Winston case, mentioning a problem with the statistical computations, it said nothing about the problems with the random sample controls.  This raises an interesting and important question.  Did the panel fail to notice these problems (perhaps because they did not receive complete documentation on the case)?  Did the panel notice the annotations about substituting results from other gels and conclude that this was not a problem?  Or did the panel fail to notice the problem, perhaps due to a cursory and inadequate review?  Inquiring minds want to know.

If Your Control Fails, Discontinue It?

As of January 11, 2005 (more than two years after the Winston case was processed), the Virginia Division of Forensic Sciences decided to stop running random sample controls.  The memo announcing that these controls should be discontinued "in order to improve turnaround time for the analysis of casework" was signed by Jaffrey Ban, Forensic Biology Section Chief.  Ban was relieved of his duties as Technical Leader of the Virginia Division of Forensic Sciences DNA laboratories in May 2005, after the ASCLD-LAB Audit Report (dated April 9, 2005) faulted him for errors and misinterpretation of DNA testing in the Earl Washington case.  

Canaries and Controls

Scientific controls are like the proverbial canary in a coal mine.  A failed control, like a dead canary, indicates an unheathy state of affairs.  And discontinuing the use of controls, like banning canaries from the mine, is a way to hide rather than solve the underlying problem.

Case Materials from Virginia v. Leon Winston (Nov. 27, 2005)


Update of Lovitt Case  (July 9, 2005)

William C. Thompson

When asked about the state's decision to deny Lovitt's lawyers access to electronic STR data that could clarify whether he is really guilty, a spokesperson for the Governor told the Richmond Times-Dispatch that the conclusions of an outside review team were sufficient to address the issue.   This response confuses two issues.

One issue is whether the Virginia DFS laboratory properly conducted and interpreted the DNA tests in Lovitt's case.  The outside reviewers found no deviations from the lab's standard protocol and therefore found no problem with the lab's work.  I agree with that conclusion. 

The second issue is whether Lovitt received a fair trial, and specifically whether the DNA evidence was presented to Lovitt's jury in a misleading way.  Dr. Arthur Eisenberg of the review panel confirmed today that the outside review panel has not reviewed transcripts of the trial.    As explained above, I believe the trial was not fair because the jury did not receive a balanced account of the DNA evidence.

Because the outside review panel addressed the first issue, and not the second, the findings of the review panel do not address concerns about the fairness of Lovitt's trial. Hence, the Governor's staff is mistaken if they believe the findings of the review panel respond to the concerns raised here and elsewhere about the fairness of Lovitt's trial. 

In light of the continuing concerns about the fairness of Lovitt's trial, and the possibility that he might actually be innocent, it seems unconscionable for the state to continue denying Lovitt's lawyers access to electronic STR data that could clarify whether he is really guilty.

Virginia Refuses Condemned Man's Request for Access to Scientific Data That Might Prove His Innocence  

by William C. Thompson

July 7, 2005

Although Robin Lovitt is scheduled for execution on July 11, 2005, substantial questions still exist about his guilt or innocence--questions that might be resolved by further analysis of existing scientific data that the state has refused to disclose to Lovitt's lawyers.  

The facts of the case are nicely summarized in a recent editorial by Margaret Edds in the Virginian-Pilot.  The editorial laments the fact that two crucial items of biological evidence (blood stained scissors that were used to fatally stab the victim and the defendant's bloody clothing) are not available for additional DNA testing because they were illegally thrown away by a state official.  An initial round of DNA testing in 1999 by the Virginia Division of Forensic Science produced "inconclusive" results, but newer tests could be more definitive.  Although the editorial says these items were discarded by error, it now appears that the destruction of evidence was intentional misconduct.   

I recently prepared a declaration that makes three important points about the Lovitt case.

First, the DNA evidence against Lovitt was presented in a misleading way.  A highly problematic test result on the murder weapon (scissors) was presented to the jury in a manner that suggested it provided significant support for the proposition that Lovitt handled the murder weapon.  I believe most experts would view this result, which was used against Lovitt, as more exculpatory than incriminating.  

Second, the DNA test results suggest that the blood stains on Lovitt's jacket, which the prosecutor attributed to the victim, could not have been from the victim and were actually from Lovitt himself.   Although the test results on the jacket were deemed “inconclusive” by the Division of Forensic Sciences, these results appear to be  far more complete and revealing than the test results that were used to link Lovitt to the murder weapon.  I believe the government applied a double standard in this case.  The extremely weak and problematic results that linked Lovitt to the murder weapon were presented to the jury and used as a basis for arguing that he was the killer.  But the jury was never told about the relatively stronger and more convincing results that supported Lovitt’s innocence by showing that the blood on his jacket did not come from the victim, as the prosecutor had argued, but came from Lovitt himself. 

Third, and most important, despite the loss of the original biological sample, it may still be possible to draw more definitive conclusions about the DNA on the scissors and the jacket through re-analysis of electronic data collected by the state lab in its initial round of testing.  The existing electronic data could be re-analyzed using new analytic methods in order to clarify the test results.  Unfortunately, the state has repeatedly refused the requests of Lovitt's lawyers for copies of these electronic files.  The state appears intent on proceeding with Lovitt's execution without giving Lovitt's lawyers an opportunity to examine evidence that conceivably could support his claim of innocence.   This hardly seems fair.

I invite others to review the materials that I reviewed in order to see whether they agree with my assessment of the case.  I have made available here everything that I reviewed, including:

As mentioned, I am very interested in hearing views of others on this case, particularly DNA experts.  Click here to e-mail your comments.  I'll post the most interesting comments here (whether they agree with me or not).   


Other Errors in DNA Testing

March 20, 2003

News of the exoneration of Josiah Sutton has raised questions about other cases in which DNA evidence has produced false incriminations.  In a recent article in the Journal of Forensic Sciences (January 2003), William Thompson, Franco Taroni, and Colin Aitken discuss several cases where such errors occurred.  We provide documentation on some of those errors here.


DNA Evidence Sent Wrong Man to Prison

March 11, 2003

The Houston Police Department acknowledged yesterday that new DNA tests appear to exonerate a man serving a 25 year sentence for rape.  Josiah Sutton was convicted in 1999 based largely on DNA tests performed by the Houston Police Crime Laboratory.  Yesterday's announcement confirms the position of UC Irvine Professor William Thompson, who told several news organizations in January that he though Sutton was probably innocent (see story below).  It also vindicates forensic scientist Elizabeth Johnson, a long-time critic of the Houston Police Department Lab, who has complained about incompetence and bias in the lab's work.  

Background on Houston Crime Laboratory Scandal

March 7, 2003

Last Fall, several DNA experts went public with accusations that the DNA/Serology Unit of the Houston Police Department Crime Laboratory was doing grossly incompetent work and presenting findings in a misleading manner designed to unfairly help prosecutors obtain convictions.

The story was broken by Houston television station KHOU.  See news accounts and streaming video on these accusations at the KHOU Website.

In response to these accusations, the Houston Police Department asked the Texas Department of Public Safety (DPS) to audit the DNA/Serology Unit.  The audit report, released in January, confirmed serious inadequacies in the laboratory's procedures, including routine failure to run essential scientific controls, failure to take adequate measures to prevent contamination of samples, failure to adequately document work performed and results obtained, and routine failure to follow correct procedures for computing statistical frequencies.  Click here to see a copy of the audit report.  

In January, UC Irvine Professor William Thompson told KHOU, the Houston Chronicle, and the New York Times, that he thought the HPD Crime Laboratory had falsely incriminated an innocent man in a 1999 rape case.  Key documents related to the case of Texas v. Josiah Sutton can be located through the following links:

Thompson's allegations led the Houston District Attorney to send remaining evidence from the case to a private laboratory for retest.  Results are expected Monday, March 10th.  

RFLP Match and Binning Procedures at Cellmark Diagnostics: Historic Summary and Current Practices

A critique of Cellmark's binning procedures, with supporting documents. [.pdf file]

Impact of Contamination Upon Reliability of DQA Typing: Lessons from the O.J. Simpson Case

Link to article describing Gerdes' pivotal study of the LAPD DNA lab. Raises important issues about susceptibility of PCR-based tests to contamination and about the adequacy of the standard controls designed to detect contamination.

Top of Page