SUPERIOR COURT
OF THE
STATE OF DELAWARE
T. HENLEY GRAVES SUSSEX COU NTY C OUR THO USE
RESIDENT JUDGE 1 THE CIRCLE, SUITE 2
GEORGETOWN, DE 19947
(302) 856-5257
September 17, 2015
Caroline Brittingham, Esquire
Department of Justice
114 E. Market Street
Georgetown, DE 19947
Jerome Capone, Esquire
Office of the Public Defender
14 The Circle, 2nd Floor
Georgetown, DE 19947
RE: State vs. Richard Roundtree
Case No: 1412017602
Dear Counsel:
This is the Court’s written decision following its bench ruling issued on September 14, 2015,
that granted the State’s Motion to allow hypergeometric sampling of suspected drugs when a large
quantity is involved. The Court held a Daubert hearing over the course of two days.1 The issue
before the Court is narrow: whether hypergeometric sampling may be used by the Division of Safety
and Homeland Security, Division of Forensic Science (hereinafter, “DFS”) to test only a portion of
seized drugs in large volume cases.2
On September 11, 2015, the State presented two experts, John Mario, a forensic scientist
1
See Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 597 (1993) (holding the rules
of evidence assign to a trial judge “the task of ensuring that an expert’s testimony both rests on a
reliable foundation and is relevant to the task at hand.”). As will be discussed, infra, this case
ultimately did not involve a Daubert issue.
2
When a population contains nine or fewer items, the hypergeometric sampling plan does
not apply. See Hypergeometric Sampling Table, attached hereto as Exhibit A.
formerly with the Suffolk County Crime Laboratory of New York, and Dr. Bradley Venner, a
statistician with the Environmental Protection Agency. On September 14, 2015, the State presented
the testimony of Robyn Quinn, DFS’ forensic testing supervisor, and Melissa Newell, the chemist
who tested the drugs in this particular case. The defense argued that hypergeometric testing was not
as reliable as testing to weight but did not offer any witness testimony to support its argument.
After hearing the evidence, the Court concludes the matter before it is not a typical Daubert
issue, in that the issue is not whether to accept or reject a new technology or scientific theory. This
matter concerns the day-to-day operations in drug laboratories and the subsequent use of evidence
in a criminal trial. The question before the Court is: Does the hypergeometric method of sampling
in drug cases permit an inference upon which the jury can reasonably rely to satisfy the State’s
burden of proof? Thus, this issue is really a sufficiency of the evidence argument.
This case does not pertain to the chemical analysis of suspected drug evidence. Rather, the
question is how much of the seized drug evidence must be actually tested by the chemist.
Mr. Mario, a chemical forensic scientist, provided the Court with a brief overview of the field
of forensic drug testing. He testified that the laboratory with which he had been affiliated from 1981
through his retirement in 2013 uses hypergeometric testing. Mr. Mario was a member of the Core
Committee of the Scientific Working Group for the Analysis of Seized Drugs (“SWGDRUG”) from
1998 until he retired in 2013. He continues to consult with the Springfield Massachusetts Police
Department in a drug evidence audit.
Mr. Mario testified that SWGDRUG cannot publish standards, per se, but can make
recommendations to the international forensic community. SWGDRUG endorses statistical
sampling and supports the hypergeometric approach. The American Society for Testing and
2
Materials (“ASTM”) publishes standards. Their standards are reviewed every five years and updated
as needed. ASTM has adopted SWGDRUG’s recommendation on sampling. The International
Organization for Standardization/International Electrotechnical Commission (“IOS/IEC”) is,
according to Mr. Mario and not refuted by the defense, the leading standard for testing laboratories
and provides that laboratories must have a sampling plan and procedures in place.
Mr. Mario told the Court that a review of seventeen sampling plans from around the United
States, including two federal laboratories, revealed that all of them employ the hypergeometric
probability distribution approach when they conduct sampling.3 Among these laboratories are the
New York State Police Forensic Center, the Virginia Department of Forensic Science, the Texas
Department of Public Safety, the Massachusetts State Police, and the United States Drug
Enforcement Administration. The State presented ample evidence that the hypergeometric sampling
model is well entrenched in drug testing laboratories.
Dr. Venner testified that hypergeometric sampling is a reasonably relied upon probability
model that is used not only in testing drug samples, but also for all sorts of commercial sampling.
It is based upon solid mathematical and statistical grounds and was recognized centuries ago and has
since been refined.
To a degree this is much ado about nothing, or perhaps deja vu all over again because we
learned that the hypergeometric model was used in Delaware by the Office of the Chief Medical
Examiner (“OCME”) prior to its closure. I suspect it was such an imbedded practice, issues as to
its use never arose. The current notoriety of the method seems to stem from the objection to the
policy of the Pennsylvanian laboratory, NMS, of only testing nine or ten samples, regardless of the
3
One of the laboratories reviewed was Delaware’s DFS.
3
population of items seized.4 This is an acceptable procedure in Pennsylvania, but its drug laws are
different and, in fact, currently under revision. In Pennsylvania and some other jurisdictions, it is
permissible to make a general or arbitrary inference argument to the jury that, if all ten (10) samples
tested positive for heroin, it can infer the balance of the seized items would test positive for heroin.
In this case, the Court rejected this general inference approach by way of a bench ruling on August
12, 2015. The Court was not satisfied that a general inference satisfied a sufficiency of the evidence
threshold. Interestingly, having lost this initial battle, the State acknowledged in its written
submission to the Court that there are problems with the arbitrary inference approach.
The polar opposite of permitting a general or arbitrary inference would be to require the
prosecution actually test that amount of drugs required for conviction under the criminal statutes, i.e.,
that DFS test to the weight. This Court could only locate one jurisdiction, Illinois, that requires the
prosecution meet this strict standard.
The hypergeometric sampling methodology allows the testing laboratory to test a portion of
the seized drugs, and, based upon those test results, infer certain conclusions about the balance of
the untested seized drugs. It is a statistical model based upon a mathematical formula that produces
a statistical inference that, if a certain number of randomly selected samples are tested and all test
positive, then it is probable that most of the remaining items would likewise test positive if actually
tested.
The number of samples that must be tested has a direct statistical relationship as to the
conclusion on probability. In the drug testing world, the standard confidence and probability
percentages requires that the number of samples tested produce a 95% confidence level that at least
4
The drugs in this case were initially sent to NMS for testing.
4
90% of the remaining population is of the same composition as what was tested.
At the hearing, the Court also learned that a major driver for the use of hypergeometric
sampling model is not only the reliability of its probability inference, but also its practicality.
Testing to the weight requires much more time and State resources. The evidence showed that, in
Delaware, it would be impractical, if not impossible, to test to the weight without a huge influx of
money and resources into DFS.
Ms. Quinn, DFS’ laboratory manager, testified DFS is licensed by the American National
Standards Institute-American Society for Quality National Accreditation Board (“ANSI-ASQ”).
DFS also has ISO/IEC 17025 accreditation, which is issued by the ANSI-ASQ. As noted, supra, the
ISO/IEC 17025 standards require the laboratory have a sampling plan and procedures for sampling.
Ms. Quinn’s responsibilities include ensuring DFS’ forensic chemistry unit is in compliance with
ISO/IEC 17025.
Ms. Quinn was a laboratory manager at OCME and has been the laboratory manager since
DFS reopened under its new incarnation in January of 2015. When Ms. Quinn took her position, she
testified that she made changes to the sampling model to make it “more conservative.” In so doing,
she relied upon SWGDRUG’s recommendations and other laboratories’ policies. All of the
laboratories from which she sought input use hypergeometric testing.
Ms. Quinn testified that there are five chemists at the Delaware lab. There are only three Gas
Chromatography-Mass Spectrometry instruments (“GC-MS”), the machines utilized to identify
substances in a test sample. Two are currently “on line” and available for use.
5
In the present case, DFS received 389 bags of suspected heroin.5 The table for determining
the number of samples to be tested under the hypergeometric model requires 28 randomly selected
samples to be tested.6 Ms. Quinn testified that, if all 389 bags had to be tested, it would take
approximately thirty (30) hours of prep time plus ninety-seven (97) hours of machine time to
complete the GC-MS analysis. Although the instrument is programed to run on its own, when a
machine is in use it is not available for other testing.
The impracticality of testing to weight increases as the tier level for a drug offense increases.
For example, a tier five heroin offense requires that a defendant have five (5) grams or more of
heroin. Ms. Quinn testified that, based upon the average amount of heroin found in a bag of heroin
in this case (.004 grams), five (5) grams of heroin could be contained in as many as 1250 bags of
heroin. Ms. Quinn opined it would take one chemist close to 390 hours to test 1250 bags.
With this said, practicality and limited resources should not dictate the Court’s acceptance
of shortcuts. Thus, this Court did not accept the general or arbitrary inference theory first proposed
by the State.
The hypergeometric model for sampling may be a time saver, but it is a statistically and
mathematically sound method of providing the jury a statistically based and probability based
inference as to the untested items. It allows for a reasonable inference based upon a mathematical
and statistical model. Ultimately, the jury must decide whether the inference supports a conviction
or whether the State’s evidence falls short of its burden. For purposes of this motion, the Court finds
5
In point of fact, there were originally 399 bags in this case. However, ten of those were
tested by NMS. The remaining 389 bags were sent to DFS for testing.
6
See Exhibit A.
6
that the hypergeometric methodology of sampling provides sufficient evidence for the State to get
to the jury.7
The Court notes that the State’s expert witness, Mr. Mario, reviewed the State’s
hypergeometric testing protocol. That protocol tracks the standard operating procedure for the other
laboratories and agencies that use hypergeometric sampling.
In each future case where the State relies upon the laboratory’s sampling table attached to this
decision, the State must still satisfy the requirements for proper hypergeometrical sampling; to wit,
the State must show how a determination as to the homogeneity of the tested population was made
and how the tested samples were randomly selected. Homogeneity requires the chemist to inspect
visually each seized item to determine if the items are so similar in physical appearance that they
would be included in a single population. In the present case, Ms. Newell visually inspected the
exterior of all 389 bags (also known as bindles). Ms. Newell testified the bags were all of similar
appearance; in same size, blue bags; and stamped with the notation “Erruption” in the same color
ink. Ms. Newell stated that, had she found any bags that were not similar in appearance, she would
have segregated the dissimilar bags into a different population pool.
After the population has been determined, the samples to be tested must be randomly
selected. In this case, after counting the bags and making a homogeneity determination, Ms. Newell
placed the bags back into a common container and randomly selected the requisite twenty-eight, per
the hypergeometric sampling table. Dr. Venner testified that, in his expert opinion, this method of
7
See Ohio v. Gartrell, 24 N.E.3d 680, 707 (Ohio 2014) (holding evidence of
hypergeometric sampling is “sufficient as a matter of law to support a determination that the
entire substance recovered together and similarly packaged is the same controlled substance as
that tested”) (citation and internal quotation marks omitted).
7
random selection is an accepted practice. The twenty-eight samples were weighed and tested. All
tested positive for heroin. Therefore, the chemist concluded with 95% confidence the probability
was that 90% of the entire population of bags contained heroin. In conclusion, the hypergeometric
model provided September 17, 2015 a statistical probability inference as to what was in the bags;
i.e., heroin.
The Court notes Dr. Venner testified that, had one or two samples tested negative, the
hypergeometric model could still be used, but a different mathematical formula would be utilized,
presumably creating a different table. The Court’s decision only addresses the mathematical formula
used when all the tested samples test positive.8
In this decision, the Court accepts the validity of hypergeometric sampling. The Court
recognizes that if protocol is followed and all of the samples test positive, there exists a statistical
probability permitting an inference with a 95% confidence level that 90% of the population contains
the same substance as the tested samples. The Court will not require this statistical probability to
be proven again and again with expert testimony in the future.
The Court will, of course, permit the attorneys to argue their respective positions as to the
inferences that may be made from the use of hypergeometric testing because the jury ultimately
decides whether it is satisfied the State has met its burden of proof.
Hypergeometric sampling is used only for determining the probability of the entire population
containing the same substance. It tests each sample to determine one of two things: positive for
controlled substance or negative. Hypergeometric sampling is not used to determine the total weight;
8
As an aside, Ms. Quinn seemed baffled when the Court inquired about the potential
effect a negative result in the sampled group would have; in her experience, she had never
encountered such a scenario.
8
weight is determined by another process. The current lab does exactly what the OCME with regard
to making a total weight determination. In this case, Ms. Newell weighed the contents of the opened
and tested bags. She took an average of those twenty-eight bags and extrapolated the total weight
of the 389 bindles. Using this averaging process, a chemist does not open all 389 bags. The Court
suspects there are other methods of determining weight; for example, taking the total weight of all
389 bags and contents, then weighing out the average of the bag alone, which should be consistent,
and then subtracting the weight of 389 bags only from the total. Again, this method was not the
focus of the hearing because the hypergeometric model is only used to determine the probability
inference of identifying the substance.
The Motion to permit the use of hypergeometric sampling is granted.
IT IS SO ORDERED.
Very truly yours,
/s/ T. Henley Graves
T. Henley Graves
THG/ymp
pc: Prothonotary
Valerie Dunkle, Esquire
9