Appendix — Chun v. New Jersey (No. 07-1562)
Supreme Court brief2008
Ask Donna
What actually matters in this document.
Text
la
APPENDIX A — OPINION OF THE SUPREME
COURT OF NEW JERSEY FILED MARCH 17, 2008
SUPREME COURT OF NEW JERSEY
A-96 September Term 2006
STATE OF NEW JERSEY,
Plaintiff-Appellant,
v.
JANE H. CHUN, DARIA L. DE CICCO, JAMES R.
HAUSLER, ANGEL MIRALDA, JEFFREY R. WOOD,
ANTHONY ANZANO, RAJ DESAI, PETER
LIEBERWIRTH, JEFFREY LING, HUSSAIN
NAWAZ, FREDERICK OGBUTOR, PETER
PIASECKI, LARA SLATER, CHRISTOPHER
SALKOWITZ, ELINA TIRADO, DAVID WALKER,
DAVID WHITMAN and JAIRO J. YATACO,
Defendants-Respondents,
and
MEHMET DEMIRELLI and
JEFFREY LOCASTRO,
Defendant,
and
DRAEGER SAFETY DIAGNOSTICS, INC.,
Intervenor.
2a
Appendix A
Argued April 5, 2007 — Remanded April 30, 2007
Master’s Report filed — November 8, 2007
Re-argued January 7, 2008 — Decided March 17, 2008
On certification to the Superior Court, Law Division,
Middlesex County.
JUSTICE HOENS delivered the opinion of the Court.
ROr 5 ERIE hv bee eeee dee es bel see hak 5a
I. Facts and Procedural History .......... 8a
A. Certification to this Court ......... 9a
eer 12a
II. Legislative Framework ................ 15a
IJI. How the Alcotest Works ............... Zla
A. Scientific and Physiological
PUES hod b a5 suaey hea eee 21a
1. Alcohol and Blood ............. 22a
2. Alcohol and Breath ............ 23a
3. Differences Between Blood and
hic eewe kek ieee 24a
3a
Appendix A
C. Test Administration and the Alcohol
oe es a 26a
IV. Findings of the Special Master ......... 34a
Pc ER kb ph eb ase eine ween 34a
B. Draeger’s Role in the Proceedings . 37a
C. Source Code Remand ............. 39a
ee errr ere rrr ree 40a
VI. Standards of Review .................. 43a
VII. Defendants’ Challenges to Scientific
eda peak ose uwae ee 45a
VIII. Disputed Findings and
eee errr 47a
A. Blood/Breath Ratio ............... 48a
B. Minimum Test Sample Criteria .... 52a
l. Scientific Data Concerning
og ere 53a
2. Equal Protection and Lowered
Breath Volume Requirement ... 56a
4a
Appendix A
3. Application to Pending
I oes 5k nose en eeaaes 63a
C. Breath Temperature Sensor ....... 64a
D. Acceptable Tolerance Analysis ..... 70a
1. Doubled Tolerance Range in
Firmware version 3.11 ......... Tla
4 Expert Testimony utah he aeaes 76a
3. Future Firmware Revisions .... 78a
4. Application to Pending
PIII 0 0 5 ho wae etaewes 80a
IX. Source Code Remand ................. 86a
A. EC Readings and Fuel Cell Drift
PT Se von dda nechnudsved cannes 87a
B. Weighted Averaging Algorithm .... 92a
C. Buffer Overflow Error ............. 94a
D. Catastrophic Error Detection ...... 99a
E. Overall Firmware Reliability ....... 10la
X. Additional Firmware Recommendations
Sa
Appendix A
- XI. Requirements Prior to the Admissibility
OF Aleotest HvViGONce 24... cece cece 105a
A. Confrontation Clause Implications .. 109a
B. Application of Crawford _ v.
eran pore ee eananpae 113a
1. Operator’s Qualifications. ...... 1l4a
2. Foundational Documents ....... 11l6a
8. Alcohol Influence Report
Pa rE a 12la
I ois ls Sea aia wigs a tere 126a
INTRODUCTION
For decades, this Court has recognized that certain
breath testing devices, commonly known as
breathalyzers, are scientifically reliable and accurate
instruments for determining blood alcohol concentration
(BAC)! and that drivers whose breathalyzer test results
demonstrate the requisite statutorily-imposed BAC are
1. Although the statute fixes limits in terms of BAC,
violations of the statute have been proven routinely through
analysis of breath and a conversion of breath alcohol
concentration (BrAC) into a BAC reading. See Sections III.A.
and VIIL.A., infra.
6a
Appendix A
guilty per se of driving while intoxicated (DWI).
Although the Legislature has from time to time reduced
the permissible BAC limits and has altered the penalties
for this offense, and although we have required
foundational proofs relating to the operation of the
breathalyzer device as a precondition for admission of
the breathalyzer test results into evidence, the accuracy
and reliability of the breathalyzer itself has remained
essentially unquestioned since our decision in Romano
v. Kimmelman, 96 N.J. 66, 474 A.2d 1 (1984).
Nevertheless, in the intervening years, the devices
have become technologically outdated, with the result
that replacement parts are no longer available and the
machines themselves, when they fail, cannot be repaired
or replaced with like equipment. Faced with an
increasingly difficult situation, the Attorney General’s
office began to consider alternate devices to use for
breath-testing purposes. That process led to the
decision by the Attorney General to select the Alcotest
7110 MKIII-C (the Alcotest).? Following its introduction
into service in a pilot program in Pennsauken, the use
of the Alcotest has been expanded to all but four of our
counties. Its use and its capabilities, as a means to
analyze breath samples with sufficient accuracy so that
2. Throughout this opinion, we will refer to the Alcotest
without specifying further the model number and we will
generally refer to the firmware without designating the version
utilized except in instances where the designation is important
for clarity. We intend to make no comments about other models
of the device or about the software used to operate any other
Aleotest model.
Ta
Appendix A
the results will be admissible into evidence to support a
conviction, withstood an initial challenge arising from
the Pennsauken program. Thereafter, the continued
expansion of use of the Alcotest around the state
resulted in a further challenge to its scientific reliability,
which has been the essential focus of our inquiry here.
In our effort to analyze the reliability of the Alcotest,
we have not only considered the questions concerning
the scientific challenges to the machine, but we have
also considered the underlying constitutional questions
about the permissibility of its use in the context of a per
se violation of the statute based solely on the results it
reports, together with such safeguards and foundational
requirements that will allow its admissibility in a DWI
prosecution. We have been aided enormously in this task
by the efforts of the Special Master for his analysis of
the voluminous record created during the extended
proceedings on remand.
In summary, we conclude that the Alcotest, utilizing
New Jersey Firmware version 3.11, is generally
scientifically reliable, but that certain modifications are
required in order to permit its results to be admissible
or to allow it to be utilized to prove a per se violation of
the statute. Some of these conditions upon admissibility
we impose as a matter of constitutional imperative,
others as a matter of addressing certain of the device’s
mechanical and technical shortcomings that were
revealed during the proceedings on remand. Within the
framework for admissibility that we here establish,
pending prosecutions should be able to proceed in an
orderly and uniform fashion.
8a
Appendix A
I. Facts and Procedural History
The matters that we have been called upon to
consider are both many and varied; even among those
issues on which the parties agree, we are required to
create mechanisms for addressing the uses of Alcotest
results generated in prosecutions undertaken prior to
this analysis.
The Alcotest is a breath-testing device,’
manufactured and marketed by Draeger Safety
Diagnostics Inc. (Draeger), which was first utilized in
New Jersey as part of a pilot project in Pennsauken.
The admissibility of the results derived from breath
testing by this device was first challenged in 2003. See
State v. Foley, 370 N.J. Super. 341, 851 A.2d 123 (Law
Div.2003). In a published decision addressing that
challenge, the Law Division judge concluded that the
device was generally scientifically reliable and that the
BAC readings it generates are therefore admissible as
proof of a per se violation of the drunk driving statute.
Id. at 345, 851 A.2d 123.
Following the decision in Foley, the State expanded
the use of the device to other municipalities, including
county-wide utilization in Middlesex County. At the same
time, in cooperation with State Police personnel charged
with overseeing the device’s implementation, see
3. To the extent that the technical manner in which the
device operates is germane to our analysis, we set it forth in
Section LIL.B., infra.
9a
Appendix A
N.J.A.C. 13:51-3.2, the manufacturer created revised
software for use in the device.‘
A. Certification to this Court
Defendants are twenty individuals who were
arrested in various municipalities in Middlesex County
and were charged with driving while intoxicated, see
N.J.S.A. 39:4-50. Each of these defendants challenged
the admissibility of results from the Alcotest in their
respective proceedings. The Law Division consolidated
all of these matters for consideration of the challenge
to the Alcotest. In response, the State filed a motion
seeking to have the court recognize the Foley opinion
as binding authority and apply its findings about the
scientific reliability of the device to all pending
prosecutions. The Law Division denied that motion and
stayed all DWI-related cases involving the Alcotest that
were then pending in Middlesex County.
The Appellate Division granted the State’s motion
for leave to appeal and remanded the matter to the Law
Division for a hearing regarding the admissibility of
Alcotest results. Before that hearing could proceed, this
Court certified the pending appeal pursuant to Rule
2:12-1, vacated the remand to the trial court, and instead
remanded the case to a Special Master, retired
Appellate Division Presiding Judge Michael Patrick
King. The Court ordered the Special Master to:
4. The technical alterations in the software, referred to as
firmware, some of which are significant to our evaluation of the
device, are explained in Section VIII.D.1, infra.
10a
Appendix A
1. Conduct a plenary hearing on the reliability
of Alcotest breath test instruments, including
consideration of the pertinent portions of the
record in State v. Foley, 370 N.J.Super. 341
1851 A.2d 123] (Law Div.2003), and the within
matters in the Superior Court, Law Division,
Middlesex County, together with such
additional expert testimony and arguments as
may be presented by the parties;
2. Determine whether the testimony
presented by the parties should be
supplemented by that of independent experts
selected by the Special Master;
3. Grant, in the Special Master’s discretion,
motions by appropriate entities seeking to
participate as amici curiae, said motions to
be filed with the Special Master within ten
days of the filing date of this Order;
4. Invite, in the Special Master’s discretion,
the participation of entities or persons as
amici curiae or, to the extent necessary in
the interests of justice, as intervenors to assist
the Special Master in the resolution of the
issues before him; and
5. Within thirty days of the completion of the
plenary hearing, file findings and conclusions
with the Clerk of the Court and
contemporaneously serve a copy on the
lla
Appendix A
parties and amic?z curiae, which service may
be effectuated by the posting of the report on
the Judiciary’s website.
Although we also vacated the Law Division’s stay of
all drunk driving cases then pending in Middlesex
County, we subsequently created a distinction among
pending prosecutions based upon the proofs and the
status of the charged individuals. Our January 10, 2006
Order therefore directed that all drunk driving
prosecutions, see N.J.S.A. 39:4-50, that did not involve
an Alcotest, and all cases of repeat offenders, should
proceed normally. As to repeat offenders who were
thereafter found guilty, we directed that the sentences
to be imposed on those defendants would be stayed only
if the conviction were based on the Alcotest results
alone. We ordered that first-offender cases involving the
Alcotest be tried “based on clinical evidence when
available, including but not limited to objective
observational evidence, as well as the relevant Alcotest
readings.” We further ordered that if a court found that
a first offender was guilty, it was required to articulate,
if possible, the alternate bases for the finding. We stayed
the execution of all first offenders’ sentences pending
resolution of this matter, except where public interest
required otherwise, and stayed all further requests for
Alcotest reliability hearings. Finally, we reiterated our
earlier Order authorizing conditional guilty pleas, see
R. 7:6-2(c), with a reservation of the right to appeal in
the event that we concluded that the Alcotest is not
reliable.
12a
Appendix A
The Association of Criminal Defense Lawyers of New
Jersey (ACDL) and the New Jersey State Bar
Association (NJSBA) were subsequently permitted to
participate as amici curiae in all of the remand and
appellate proceedings.
B. Remand Hearings
Shortly after being appointed to serve, the Special
Master issued a discovery order directing the State to
provide defendants with certain technical information
concerning the operation of the Alcotest device, followed
by an order directing the State to make several Alestest
machines available to defendants and the NJSBA. In
large part, the ensuing dispute about the disclosure of
the software used to operate the device, called firmware,
and the source codes needed for an analysis of that
software, caused significant disruption in the orderly
completion of the proceedings and eventually led to our
further remand for additional proceedings.
In short, however, the Special Master was advised
that Draeger considered the software and the source
code to be proprietary information and would not
disclose it. He proposed that counsel enter into a
standard protective order and invited Draeger, which
was not then a party, to intervene in the proceedings.
Draeger declined the Special Master’s invitation to
intervene. At the same time, Draeger refused to permit
the parties to review the software except under
extremely limited conditions and refused to disclose the
source code under any circumstances. As a result of this
l3a
Appendix A
impasse, the Special Master concluded that he could
utilize an adverse inference as to the reliability of the
device, but he proceeded with the hearings in the
absence of any participation by Draeger. Near the end
of the initial hearings, defendants and Draeger entered
into a letter agreement, which would have permitted
defendants to evaluate future changes to the software
in the event that the Alcotest was found to be
scientifically reliable.®
Following hearings that spanned four months, the
Special Master issued his findings and conclusions,
embodied in a report to this Court dated February 13,
2007. In that report, the details of which we address in
Section IVA., infra, the Special Master concluded that
the Alcotest is generally scientifically reliable, but he
recommended that several changes be incorporated
both prospectively and with respect to pending matters.
Thereafter, but prior to the time when we received briefs
on the merits and entertained oral argument, Draeger
moved for leave to intervene before this Court, which
motion we granted.
After the initial oral arguments on April 5, 2007,
including those offered by Draeger, we remanded the
matter to the Special Master again to allow defendants
an opportunity to conduct the analysis of the scurce
5. In som> respects, the parties disagree about the
continued need for ar. viability of the agreement, which they
referred to as Addendum A. We address future testing of
software revisions further below, see Section X, infra.
l4a
Appendix A
code that they had contended was essential to an
accurate determination of the reliability of the device.
State v. Chun, 191 N.J. 308, 309, 923 A.2d 226 (2007). In
doing so, we directed that the review be undertaken by
an independent softwarehouse, to be agreed upon by
Draeger and defendants, in order to preserve Draeger’s
proprietary interests. Jd. at 309-10, 923 A.2d 226.
The parties, however, were unable to agree on an
independent software house that would conduct the
source code analysis. Although our order authorized the
Special Master in that event to make the selection, he
believed he was not wel equipped to choose and he so
advised us. Therefore, this Court issued a supplemental
order allowing each of the parties, at its own expense,
to designate an independent software house to review
the source code. The supplemental order also provided
that the Special Master, at his discretion, could conduct
further hearings following his receipt and review of the
expert reports.
Draeger and defendants each designated a software
house to analyze the source code and report on its
reliability. Because the reports reached different
conclusions, the Special Master scheduled further
hearings. After ten additional days of testimony and two
days devoted to summations, the hearings were
completed on October 24, 2007. The Special Master
submitted his Supplemental Findings and Conclusions
to this Court on November 8, 2007. He concluded, in
summary, that the source code analysis did not alter his
original opinion that the Alcotest is scientifically reliable,
lSa
Appendix A
as to both its hardware and software elements. However,
he conditioned this conclusion on additional
recommendations, which supplemented those contained
in the initial report.
II. Legislative Framework
Our analysis of the issues surrounding the scientific
reliability of the Alcotest device and our consideration
of the Special Master’s recommendations must begin
with an understanding of the legislative framework that
bears upon drunk driving prosecutions. We turn, then,
to an explanation of the statutes governing the offenses
that we generally refer to as drunk driving, together
with an analysis of the relevant legislative history that
bears on the issues before us.
The Legislature has established that an individual
is guilty of driving while intoxicated if he or she
“operates a motor vehicle with a blood alcohol
concentration of [0].08 [percent] or more by weight of
alcohol in [his or her] blood.” N.J.S.A. 39:4-50(a). For
first offenders who have a BAC that is 0.10 percent or
greater, harsher penalties and higher fines apply. See
N.J.S.A. 39:4-50(a)(1). Subsequent offenses, as
measured by the 0.08 percent standard, are treated with
increasingly harsh penalties, including not only longer
periods of license suspension, but incarceration as well.
See N.J.S.A. 39:4-50(a)(2), -50(a)(3).
l6a
Appendix A
As we have previously found, the primary purpose
behind our drunk driving laws is to remove intoxicated
drivers from our roadways and thereby “to curb the
senseless havoc and destruction” caused by them. State
v. Tischio, 107 N.J. 504, 512, 527 A.2d 388 (1987). We
have consistently construed these laws both broadly and
pragmatically to ensure that the Legislature’s intent is
effectuated. See id. at 513, 527 A.2d 388; State v.
Mulcahy, 107 N.J. 467, 479, 527 A.2d 368 (1987)
(concluding that turning on ignition is not required for
finding that person behind the wheel was in control of
and intended to operate vehicle); State v. Wright, 107
N.J. 488, 497, 527 A.2d 379 (1987) (concluding that
predicate of actual operation of vehicle is not required
for request that individual undergo breathalyzer
testing).
As part of the effort to rid our roads of drunk
drivers, the Legislature has sought over time to
streamline the process by which those charged with
DWI offenses are efficiently and successfully
prosecuted. See Tischio, supra, 107 N.J. at 514, 527 A.2d
388. Our current laws, as a result, can only be
interpreted correctly if they are viewed in the context
of this continuing evolution.
Our analysis begins in 1951, when, in order to
address growing difficulties and confusion surrounding
the evidentiary burden for establishing operation of a
vehicle “under the influence,” the Legislature enacted
N.J.S.A. 39:4-50.1. Tischio, supra, 107 N.J. at 514-15,
27 A.2d 388; see also State v. Protokowicz, 55 N.J.Super.
17a
Appendix A
598, 603, 151 A.2d 396 (App.Div.1959). This statute
provided that a 0.15 percent blood-alcohol level gave
rise to a presumption of intoxication for purposes of a
driving under the influence prosecution. Tschio, supra,
107 N.J. at 515, 527 A.2d 388. A blood-alcohol levei below
0.05 percent gave rise to a presumption of non-
intoxication, and a level between the two gave rise to no
presumption. 7d. at 515 n. 3, 527 A.2d 388. These
legislative presumptions were targeted at reducing the
evidence, specifically expert and other testimony, which
was otherwise needed to prove intoxication and convict
a drunk driver. 7d. at 515, 527 A.2d 388.
At that time, New Jersey’s 0.15 percent standard
was the most permissive in the country, see 2d. at 515-
16, 527 A.2d 388 (citing Motor Vehicle Study
Commission, Report to the Senate and the General
Assembly of 1975 (hereinafter “Report”), at 1385),
although the penalties imposed were “among the most
stringent.” Jd. at 515, 515 n. 4, 527 A.2d 388.
Nevertheless, studies revealed that most drivers were
impaired at BAC levels significantly lower than the
statutory presumption employed in the 1951 statute. /d.
at 516, 527 A.2d 388 (citing Report, swpra, at 141-42).
As a result, the Legislature amended N.J.S.A. 39:4-50.1,
in 1977, see L. 1977, c. 29, to lower the presumptive BAC
for intoxication purposes from 0.15 to 0.10 percent.
Tischio, supra, 107 N.J. at 516, 527 A.2d 388.
In 1983, the Legislature again amended the drunk
driving statutes to take into account “mounting
scientific findings,” to the effect that almost all drivers
18a
Appendix A
suffered reduced driving ability at a BAC of 0.10 percent.
Ibid. At the same time, the amended statute brought
the state into compliance with minimum federal grant
standards. L. 19838, c. 129; Assembly Judiciary, Law,
Public Safety & Defense Committee, Statement to
Assembly Committee Substitute for Senate Bill No. 1833
(Feb. 14, 1983). Significantly, the amended version of
N.J.S.A. 39:4-50 provided that a 0.10 percent BAC level
constituted a per se offense, instead of simply giving
rise to a presumption.
In 1990, the New Jersey Commercial Driver License
Act was enacted. L. 1990, c. 108. It created an even more
stringent standard to be applied to drivers of commercial
vehicles. It provides a penalty, in addition to any other
applicable penalties, of a one to three-year commercial
license suspension for commercial drivers caught driving
with a BAC level of 0.04 percent or greater. N.J.S.A.
39:3-10.13, -10.20(a)(1). The 0.04 percent BAC standard
for commercial drivers was enacted both to comply with
the federal standard in the Commercial Motor Vehicle
Safety Act of 1986, Pub.L. No. 99-570, 100 Stat. 3207
(1986) (codified at 49 U.S.C.A. § 31310), and in
recognition of the fact that significant impairment
occurred well below the otherwise applicable 0.10
percent BAC levels. See L. 1990, c. 103; Assembly
Appropriations Committee, Statement to Assembly Bill
No. 3258, at 23 (Oct. 1, 1990).
6. This change essentially engulfed the rule provided in
N.J.S.A. 39:4-50.1, which nonetheless remained in the statutes
until 1990, when it was repealed by L. 1990, c. 103, § 38.
19a
Appendix A
In 1992, the Legislature enacted an additional
drunk driving prohibition by creating a new per se
offense, which applies to drivers who are under the legal
drinking age. L. 1992, c. 189. This most recently-added
tier provides that any person under the age of twenty-
one who is caught driving with a BAC level above 0.01
percent faces a thirty to ninety-day license suspension,
in addition to community service requirements. See
N.J.S.A. 39:4-50.14. The statement attached to the
legislation explained that the bill was intended to
establish penalties for any driver under the age of
twenty-one who is “found to have consumed an alcoholic
beverage.” L. 1992, c. 189; Assembly Judiciary, Law &
Public Safety Committee, Statement to Assembly
Committee Substitute for Assembly Nos. 1447 & 1426
(June 1, 1992). The purpose of the enactment was two-
fold: “to deter younger drivers from drinking and
driving, and to establish an early detection and
treatment program for young people... .” Anthony
Impreveduto, et al., Statement to Assembly No. 1426
(May 14, 1992).
In 2003, the per se violation set forth in the statute
was further reduced. In order to comply with federal
highway funding requirements, the statutory standard
of 0.10 percent BAC was reduced to 0.08 percent BAC.
L. 2003, c. 314. At the same time, the amendment
created two separate, graduated penalties relevant to
prosecution for a first offense. As a result of this
legislative enactment, first time offenders with a BAC
level between 6.08 nercent and 0.10 percent are subject
to a three-month license suspension, but first time
20a
Appendix A
offenders with a BAC level of 0.10 percent or greater
are subject to a seven to twelve-month license
suspension. /bid.
In addition, throughout this time, penalties for
second and third offenders have become increasingly
harsh. See, e.g., L. 1995, c. 286 (registration revocation);
L. 1999, c. 417 (ignition interlock device installation);
L. 2003, c. 315 (Michael’s Law; imposing mandatory jail
time or inpatient rehabilitation program time for a third
or subsequent violation); L. 2004, c. 8 (increasing
penalties for refusal to submit to breath test).
Although when considered together, these statutory
enactments make plain the Legislature’s view that
drunk driving is not to be tolerated, the relationship
between this increasingly restrictive legislative scheme
and the new technology of the Alcotest, as compared to
the breathalyzer, requires us to re-examine much of our
earlier jurisprudence as part of our consideration of the
issues raised in this appeal.
In virtually all of these statutes, the Legislature has
utilized blood alcohol concentration, not breath alcohol
concentration, as its standard measure.’ Both the
breathalyzer and the Alcotest, however, test breath
samples and convert that analysis by mathematical
7. Although the commercial driving statute defines
“alcohol concentration” in terms of both blood and breath, see
N.J.S.A. 39:3-10.11, our focus here will be on the more commonly
applied articulation of blood alcohol.
2la
Appendix A
calculations to an expression of the subject’s presumed
blood alcohol concentration. The principle question,
then, is whether the Alcotest does so with sufficient
accuracy and reliability to permit the results to be
admitted in evidence in a DWI prosecution, or used as
the basis for a per se violation of the statute and,
therefore, a conviction.
Ill. How the Alcotest Works
The State seeks in this proceeding to establish that
the Alcotest is scientifically reliable to measure
defendants’ blood alcohol levels. We turn, then, to a
discussion of the physiological effects of alcohol on the
body, how the Alcotest measures the concentration of
alcohol in the breath and converts it to a measure of
blood alcohol levels, and the State’s proposed
procedures to ensure that the Alcotest functions
properly.
A. Scientific and Physiological Framework
Much of the scientific evidence in the record before
the Court is undisputed. In fact, the basic physiological
mechanisms on which all breath testing devices rely are
not themselves controversial. We set these scientific
propositions forth here, however, to provide the basis
for our analysis of the scientific matters that are in
dispute.
Be
22a
Appendix A
1. Alcohol and Blood’
Alcohol is ordinarily ingested orally and enters the
stomach where it is absorbed through the stomach’s
walls and intestines and is thereafter carried by the
blood through the liver to the heart. The heart pumps
the blood and, along with it, the alcohol, through the
body, including carrying it to the brain and the lungs.
Alcohol exerts its effects on an individual when the blood
containing the alcohol reaches the brain.
Absorption begins immediately once a person starts
drinking. The rate of absorption varies greatly from one
person to the next and can even vary in the same person
at different times. It depends on a wide variety of factors
including general health, recent food consumption,
physica! makeup, amount of alcohol consumed, weight,
and gender.
Elimination of alcohol also starts as soon as a person
begins to drink. Alcohol is eliminated through excretion
and metabolization, which occur when alcohol passes
through the liver and is broken down by enzymes and
dehydrogenates. When a person’s body is absorbing
alcohol faster than he or she is eliminating it, the
concentration of alcohol in the blood will continue to rise.
8. We draw thesc scientific descriptions from the testimony
in the record offered by Barry Logan, a board-certified forensic
toxicologist, and Patrick Harding, a biochemist who has also
previously testified in proceedings involving breath testing
devices. See State v. Downie, 117 N.J. 450, 454, 569 A.2d 242
(1990).
23a
Appendix A
This period of time is ordinarily referred to as the
absorptive phase. The concentration will reach its peak,
and it will achieve a plateau, at the time when elimination
and absorption are occurring at about the same rate.
When the person stops ingesting alcohol, or slows
down ingestion to the point where the body is
eliminating alcohol more quickly than absorbing it, the
body enters what has generally been referred to as the
post-absorptive phase. During this period of time, the
concentration of alcohol in the blood decreases.
2. Alcohol and Breath
The reported concentration of alcohol in any
particular person varies depending upon the source of
the test sample. An understanding of the relationship
of these potential test sample sources to BAC is
important to our analysis. Alcohol passes into the lungs,
through the walls of the air sacs, called alveoli. As it
does so, it mixes with the air that the person has inhaled.
When the person exhales, alcohol passes out of the body
as part of the breath.
An individual’s breathing pattern can influence the
amount of alcohol that appears in any particular breath.
In addition, the amount of alcohol in the breath sample
represented by a single act of exhalation will vary from
the beginning to the end. This is because the breath
actually comes from different parts of the body, from
the mouth to the deepest part of the lungs. Except for
the possible interference that would occur if the test
24a
Appendix A
subject had ingested alcohol so recently that residual
mouth alcohol were captured, the first part of the breath
comes from the mouth and throat where there is little
contact with the alcohol passing through the alveoli.
However, as the person continues to exhale, the expelled
air comes from deeper in the respiratory system, where
it contains alcohol that more closely represents the
amount passing through the lungs from the circulating
blood.
3. Differences Between Blood and Breath Tests
Our statute establishes the violation in terms of
blood, and not breath alcohol concentration. Although
testing an individual’s blood would presumably provide
more direct evidence of that person’s BAC, there are
obvious practical and logistical problems associated with
attempting to collect blood samples from suspected
drunk drivers routinely.
As a result, although because of our statute New
Jersey is considered to be a “blood state,” we have long
permitted BAC to be established through breath testing,
in which breath samples are tested and converted to
determine blood alcohol levels. Breath testing therefore
uses an indirect measure of BAC by calculating the
alcohol concentration in the breath (breath alcohol
concentration, or BrAC) and extrapolating to derive the
BAC using a blood/breath ratio. Breath testing has
become the preferred method for field testing because
it can be performed easily, is highly automated, does
not require scientific skill, and produces an immediate
result.
25a
Appendix A
B. Operation of the Alcotest
In light of the fact that breath testing always relies
on the extrapolation of BAC through testing of breath,
the precision with which any device evaluates BAC
through this method is critical to our consideration of
the admissibility of the device’s results. We turn then to
a description of the manner in which the Alcotest
operates.
The Alcotest, which is currently in use in seventeen
of our twenty-one counties,’ as well as in other states,
including Alabama and parts of New York, is a device
that purports to accurately measure the concentration
of alcohol from a human subject through breath testing.
The Alcotest is an embedded system, meaning that it is
a device with a specific purpose, and it relies on pre-
loaded software that the manufacturer refers to as
firmware.
The Alcotest uses both infrared (IR) technology and
electric chemical (EC) oxidation in a fuel cell to measure
breath alcohol concentration. The device therefore
produces two test results for each breath sample, one
derived from an IR reading and the other, by and large,
from an EC reading.
Although the precise mechanism by which these
tests are accomplished is net relevant to the issues
9. Only Bergen, Essex, Monmouth, and Hudson counties
do not currently use it.
26a
Appendix A
before us, the [IR chamber, also called a cuvette, captures
the breath sample and uses infrared energy to calculate
absorption cf the energy by the alcohol concentrated in
the chamber. IR technology has been available since the
1970’s or early 1980’s and scientists have concluded that
it is reliable. See, e.g., Foley, supra, 370 N.J.Super. at
350, 851 A.2d 123.
The EC, or fuel cell technology, uses a catalyst to
absorb alcohol and provide a second measurement” of
breath alcohol concentration from a small sample
captured from the cuvette. In the EC chamber, voltage
is applied to cause the catalytic reaction, which causes
any alcohol that is present to oxidize. As that occurs,
the oxidation process creates electricity, which is then
measured to determine the amount of alcohol interacting
with the fuel cell.
C. Test Administration and the
Alcohol Influence Report
The Alcotest reports the IR and EC readings on a
printout from the machine, referred to as the Alcohol
Influence Report (AIR)."' One of the claimed advantages
10. Draeger has consistently represented that the IR and
EC tests are “completely independent” as a basis for its claim
that the device is reliable. As our discussion of the fuel cell drift
algorithm, see Section IX.A., infra, explains, however, the
reported results of the two tests are not always independent.
11. To the extent relevant to our analysis, we describe the
specific details of the information reported on each AIR further,
see infra.
27a
Appendix A
of the Alcotest, as compared to the breathalyzer, is that
it is not operator-dependent, but performs its analysis
in accordance with a sequence through a computerized
program that gives visual prompts to the operator. We
turn, then, to a description of the manner in which the
device operates in practice in performing these
functions.
The actual administration of the test is performed
by one of the more than 5000 certified Alcotest operators
in New Jersey. When a person has been arrested, based
on probable cause that the person has been driving
while intoxicated, he or she is transported to the police
station to provide a sample for the Alcotest The
Alcotest, consisting of a keyboard, an external printer,
and the testing device itself, is positioned on a table near
where the test subject is seated.
Operators must wait twenty minutes before
collecting a sample to avoid overestimated readings due
to residual effects of mouth alcohol. The software is
programmed to prohibit operation of the device before
the passage of twenty minutes from the time entered
as the time of the arrest. Moreover, the operator must
observe the test subject for the required twenty-minute
period of time to ensure that no alcohol has entered the
person’s mouth while he or she is awaiting the start of
the testing sequence. In addition, if the arrestee
swallows anything or regurgitates, or if the operator
notices chewing gum or tobacco in the person’s mouth,
the operator is required to begin counting the twenty-
minute period anew.
28a
Appendix A
The Alcotest that is the focus of this matter utilizes
software developed in collaboration with the New Jersey
State Police and known as New Jersey Firmware
version 3.11.'* This software prompts the operator
through a specific testing sequence on each arrestee.
Essentially, the process begins when the operator has
typed identifying information into the machine through
a series of questions and prompts. The device then starts
and automatically samples the room air to determine if
there are chemical interferents in the room. This is
known as a blank air test. Assuming that there are none,
the machine then uses its attached wet bath simulator
to heat a solution and produce a vapor sample from a
control test solution’® with a known alcohol
concentration of 0.10, which is then measured using IR
and EC technology. In order to be valid, the control test,
in accordance with currently-programmed firmware,
must produce results between 0.095 and 0.105. If the
results dc not identify the known sample within the
defined parameters, the device is programmed so that
12. The Alcotest that was the subject of the Law Division’s
findings and conclusions in Foley, supra, utilized an earlier
version of the software known as New Jersey Firmware version
3.8. A number of changes made to the software following the
court’s decision in Foley have become important to our analysis
as we will detail.
13. The record reflects that the control solution must be
changed after approximately twenty-five test sequences or
thirty days. The device prompts the operator when the solution
needs to be changed and generates a separate report evidencing
the results of contro! testing after each change in the solution.
29a
Appendix A
the test cannot proceed. If the machine is working
properly as demonstrated by the control test, then the
instrument performs a second blank air test, again using
room air to purge the test sample out of the chamber.
Assuming that the results of the control test are
within the established parameters, the instrument
prompts the operator through a message on the LED
screen to collect a breath sample. The operator then
attaches a new, disposable mouthpiece and removes cell
phones and portable electronic devices from the testing
area. The operator is required to read the following
instruction to the test subject: “I want you to take a
deep breath and blow into the mouthpiece with one long,
continuous breath. Continue to blow until I tell you to
stop. Do you understand these instructions?” The
arrestee then provides the first breath sample, which is
measured in the IR and EC chambers.
Lights on the LED screen and an audible sound alert
the operator when a breath sample which meets the
minimum fixed standards, comprised of four criteria,
has been provided. The operator then tells the subject
to stop and the instrument performs a third blank test
to purge the first breath sample. After a two-minute
lock-out period during which the device will not permit
another test, the instrument prompts the operator to
read the instruction again to the arrestee and collect
the second breath sample. The second sample is also
measured using the IR and EC technology. The second
sample is purged from the machine and the device
performs a fourth blank test using room air.
30a
Appendix A
If the measurements for the first breath test are
out of the accepted range of tolerance with the
measurements for the second breath test, the machine
prompts the operator to conduct a third breath test.
Depending on the relationship among the three tests,
the results are reported. The instrument then performs
a second control test with the known solution from the
simulator. Finally, the air is purged again and a final
blank test is performed.
The device gives the operator three minutes to
collect each sample. If that time expires without a
sample, the device will present the operator with three
options. The options are to terminate the test, report
that the person refused the test, or continue with the
test. If the officer opts to continue the test, the device
will purge itself and then prompt the operator to collect
another sample. The operator has a maximum of eleven
attempts to collect two breath samples. After the
eleventh failed test, the only two options permitted by
the device are to terminate testing or report refusal.’
14. Even if the officer types in the code for a refusai, he is
not required to issue a summons for refusal. Instead, the officer
may opt to start the test again and give the arrestee eleven
more attempts. Alternatively, the officer may decide to
terminate testing, without charging the test subject with
refusal. An operator will generally select this option if he or she
concludes that the subject has in fact attempted to comply but
is not capable of providing a sample that meets the minimum
test criteria.
3la
Appendix A
As currently configured by New Jersey Firmware
version 3.11, the software now being utilized, the device
will accept a sample only if it meets certain minimum
criteria that have been devised by the State.’° Once the
subject has provided an acceptable breath sample, the
machine prompts the operator, through a system of
lights on the LED screen and an audible beep, to tell
the subject that he or she may stcp. If any of these
minimum test criteria has not been met, the machine
will generate an error message and a report of how much
air was submitted. The machine then offers the operator
the option of giving the person another attempt or
asserting refusal.
The results of the test sequence are printed out from
the device in a sequentially numbered document
referred to as an AIR. The AIR contains the test
subject’s identifying information, date, time, and test
results for each stage of the procedure. Each AIR
includes a variety of other information relevant to the
test, including the serial number of the device used in
the test, dates of and file numbers for calibration and
linearity checks, and solution control lot and bottle
numbers. The operator must retain a copy of the AIR
and give a copy to the arrestee.
In the event that the administration of the test
resulted in errors because of, for example, insufficient
15. The legitimacy of some of these criteria are in issue in
this dispute. We need not explain them in detail here but will
do so in the context of our analysis of those criteria that have
given rise to a debate. See infra, Section VIII.B.
32a
Appendix A
breath volume or duration, the AIR will report those
errors and will not attempt to calculate the BAC from
an inadequate sample. Similarly, if the results of the
control test do not fall within the acceptable tolerance,
the device will produce an AIR that reports that the
test could not be accomplished because of an invalid
control test.
If the results are within the acceptable tolerance,
the AIR shows the BAC values for each IR and EC
reading for each of the tests to three decimal places.
The AIR then reports the final BAC test result, which
will be the lowest of the four acceptable readings, that
is, readings within acceptable tolerance, which the
device is programmed to truncate to two decimal places.
Truncating, as opposed to rounding, involves simply
reporting the first and second decimal places and
dropping the third. For example, by truncating, a
reading of 0.079 percent BAC would be reported as 0.07
and a reading of 0.089 percent BAC would be reported
as 0.08. The effect of truncating, as opposed to rounding,
is to under-report the concentration, to the benefit of
the arrestee.
By statute, the Legislature has designated the
Attorney General to create and implement a breath
testing program. See N.J.S.A. 39:4-50.3. The Attorney
General, in turn, has vested responsibility for carrying
out this command in the State Police. See N..J.A.C. 13:51-
3.2. The Alcotest program was designed and is overseen
by the Office of Forensic Sciences, a Division of the New
Jersey State Police. The director of the forensic
33a
Appendix A
laboratory, Dr. Thomas Brettell, together with other
forensic scientists in the Office assigned to the alcohol/
drug testing unit, conducted tests on a variety of breath
testing devices in an effort to select a successor to the
breathalyzer.
After the Alcotest was chosen, Brettell assisted in
the creation of the test criteria and provided other input
into the original programming and the updates to the
software that now is utilized in operating the device.
His office has collaborated with municipalities to train
Alcotest operators and to oversee certain aspects of the
program. State Police Sergeant Kevin Flanagan is the
field supervisor for five State Police coordinators, each
of whom monitors a geographic area. The coordinators
receive factory and classroom training from Draeger and
they, in turn, train the operators. Coordinators do not
perform any repairs, but they perform “black key”
functions, such as calibration and software uploads,
which are not done by other police personnel.
Calibration of the machines involves attaching the
machine to an external simulator which uses a variety
of solutions of known alcohol concentrations to create
vapors that approximate human breath. By exposing the
IR and EC mechanisms to these differing
concentrations, and by analyzing the device’s ability to
identify accurately each of those samples within the
acceptable range of tolerance, referred to as a linearity
test, the coordinator is able to ensure that the machine
is correctly calibrated. When coordinators undertake to
perform this calibration, currently on an annual basis,
34a
Appendix A
and other routine inspections, they also download the
device’s test information onto two compact discs.'* In
accordance with current State Police protocol, one of
these discs is kept in the local police department’s
evidence file and the other is held by the coordinator.”
IV. Findings of the Special Master
Following hearings that spanned four months and
included testimony from eleven fact and expert witnesses
called by the State and two experts offered by
defendants, the Special Master issued his first report
on February 13, 2007. Although there are some aspects
of that report and certain of the Special Master’s
recommendations that are not disputed by any of the
parties, much of the report and many of the
recommendations are challenged in this proceeding. As
a result, we briefly summarize the report and its findings
and recommendations before turning to our analysis of
the matters in dispute.
A. Initial Report
In short, the Special Master concluded that the
Alcotest in general is scientifically reliable, that it is
16. The record reflects that each device is capable of
storing the data from 1000 test results. Current State Police
protocol, however, requires the coordinators to download data
from each device before it exceeds 500 tests.
17. See Part IV, infra (Special Master’s Finding 7,
recommending creation of centralized database).
35a
Appendix A
superior to the breathalyzer because it relies less on
operator influence, and that the AIR it generates,
therefore, meets the test for admissibility in drunk
driving prosecutions in general. Notwithstanding that
conclusion, however, the Special Master offered a large
number of suggestions for modifications both as to the
future operation of the device and as to the use of the
extant AIRKs as evidence in pending prosecutions.
In his first report, the Special Master offered all of
the following specific findings and recommendations.”
He found that the use of the 2100 to 1 blood/breath ratio
is scientifically reliable (Special Master’s Finding 1(b));
he recommended that the AIR, solution change report
and calibration documents be amended to include a
listing of the temperature probe serial number and
value (Special Master’s Finding 2(a)); he reeommended
that the State be required to publish future firmware
revisions (Special Master’s Finding 2(b)); he
recommended that the State continue to lock the
firmware so that only Draeger and the coordinators
would be able to make changes to that software (Special
Master’s Finding 2(c)); he found that the AIR, which
reports all of the breath test results, rather than only
the final reported lowest result, should be admissible in
evidence (Special Master’s Finding 2(d)); he
recommended that the AIR be revised to identify the
reason that a particular defendant did not achieve a
18. We have elected to adopt, only for the sake of simplicity
and clarity, the numbering of the recommendations utilized by
the Special Master rather than to proceed with a sequential
enumeration.
36a
Appendix A
reportable result (Special Master’s Finding 2(e)); he
found that Firmware version 3.11 is itself scientifically
reliable and that future changes would not undermine
its current reliability (Special Master’s Finding 2(f)); he
concluded that the Alcotest is not operator dependent,
(Special Master’s Finding 2(g)), and that it is therefore
superior to the breathalyzer (Special Master’s Finding
8); he recommended that all defendants have access to
centrally collected data on their matters as well as to
redacted versions of information relating to breath tests
performed on other arrestees (Special Master’s Finding
2(h)); he recommended that the calibration, certification
and linearity reports be amended to include the serial
number of the digital temperature measuring system
utilized (Special Master’s Finding 2(i)); he found that
the State should be required to provide training for
defense counsel and their experts similar to that
provided to the certified operators (Special Master’s
Finding 2(j)); he found that the agreement between
Draeger and defendants regarding future testing of
firmware revisions should be enforced (Special Master’s
Finding 3); he concluded that the Alcotest is well
shielded against radio frequency interference (RFI)
(Special Master’s Finding 4); he recommended that
operators be required to testify about their
qualifications and the testing procedures utilized in any
proceeding relying on Alcotest results (Special Master’s
Finding 5(a)); he identified twelve foundational
documents that the State must provide in discovery,
which may be admitted into evidence without further
formal proofs, and reasoned that they must be admitted
into evidence in cases in which the defendant is not
37a
Appendix A
represented by counsel (Special Master’s Finding 5(b));
he concluded that the technical criteria for a minimum
breath sample utilized by the Alcotest are appropriate,
with the exception of the minimum breath volume as it
relates to women over sixty years of age (Special
Master’s Finding 6); he recommended that the State
create and maintain a centralized database of the
digitally recorded data (Special Master’s Finding 7); he
concluded that the State must commence use of the
Draeger breath temperature sensor and apply a
mathematical formula to account for the effect of
temperature to pending reported results (Special
Master’s Finding 9); and he recommended that the State
must reduce the acceptable tolerance for breath results
to a total range of ten percent in place of the currently
utilized calculation of a range of plus or minus ten
percent for future use of the device (Special Master’s
Finding 10).
B. Draeger’s Role in the Proceedings
During the first oral argument before this Court
following the Special Master’s release of his report and
recommendations, defendants argued that the entire
proceedings were tainted by the manner in which
defendants were required to proceed. They argued that
because Draeger had refused to make its source code
available for their inspection and for analysis by their
experts, the Court could have no confidence in the
reliability or accuracy of the device from a scientific
perspective. In short, they argued that the
manufacturer’s intransigence forced the Special Master
38a
Appendix A
and, by extension, this Court, to rely on “black box”
testing,’® when only a complete and thorough analysis
of the source code used to operate the device would
suffice for constitutional purposes.
Indeed, the refusal of Draeger to intervene
precluded the Special Master from permitting any
testing of the manner in which the device operates, and
required him to rely on tests that at best could only
demonstrate that the machine reliably appeared to be
able to identify correctly, or at least acceptably within
the established parameters, the alcohol concentration
of a known test sample. There is some logic to that
method of proceeding. If a breath testing device can,
reliably and consistently over time, correctly enalyze a
sample of known alcohol concentration, one might argue
that it matters little how the device is able to do so.
Notwithstanding the rather considerable force of that
logic, we were persuaded that, in light of the
constitutional dimension of the issues before us,
Draeger’s eventual election to intervene in this matter
afforded us the opportunity to permit defendants to
engage in the technical analysis of the source code that
they had asserted was so necessary to the adequate
protection of their rights.
19. “Black box” testing refers in this context to a method
of evaluating the reliability of the device by using known
concentrations to test whether the device accurately detects
those concentrations. It refers to testing that does not also
consider whether the mechanism by which the result is achieved
might be flawed.
39a
Appendix A
C. Source Code Remand
Following our order remanding the matter for
further analysis of the issues by means of the source
code evaluation by the two independent testing entities,
see Chun, supra, 191 N.J. at 309-10, 923 A.2d 226, the
Special Master entertained further testimony on the
issues. His supplemental report, dated November 8,
2007, included several additional recommendations, but
continued to adhere to his initial conclusion that the
device is scientifically reliable for use in pending and,
with modifications, future proceedings.
‘In summary, the Special Master found that a
mathematical algorithm that corrects for fuel cell drift
did not undermine the reliability of the results, but he
recommended that the machines be recalibrated every
six months rather than annually to afford more regular
opportunities to replace aging fuel cells; he found that
a specific buffer overflow error should be corrected in
future versions of the software and recommended that
in all pending matters in which a third test was
performed, that the AIR be excluded or recalculated
according to a corrective formula, described in the
record as the Shaffer formula; he recommended that
catastrophic error detection be re-enabled to stop and
restart the machine in the event that such an error
occurs; he recommended that the AIR should be
inadmissible in any case in which there is data missing
from it; he revised his initia! finding 5(b) to recommend
that the twelve foundational documents be produced in
discovery and be admissible in all cases, without regard
40a
Appendix A
to whether a particular defendant is represented by
counsel or not; he suggested that notice of any and all
proposed software revisions be provided to the NJSBA;
he recommended generally that defendants’ expert’s
suggestions for reorganizing and simplifying the source
code be considered for implementation, but declined to
mandate adherence to any specific design standard for
future software revisions; he concluded that a weighted
averaging algorithm in the code was an accurate
methodology that fairly aids in the measurement of
breath samples in a test subject; and he accepted the
testing method employed by the State’s expert and
rejected the hypothetical probability analysis raised by
defendants as being unnecessarily speculative.
V. Uncontested Issues
We begin our analysis with the observation that
some of the Special Master’s findings and
recommendations aave not been contested by any of the
parties. We will therefore limit our review of those
findings and recommendations to a consideration of
whether they are supported by sufficient credible
evidence in the record, see State v. Locurto, 157 N.J.
463, 472, 724 A.2d 234 (1999); State v. Johnson, 42 N.J.
146, 158-59, 199 A.2d 809 (1964), and, by extension,
whether we will adopt them as our own. With this
standard to guide us, we need only briefly address cach
of them. We do not, however, by the relative brevity of
the attention we here accord to these findings and
recommendations, intend to suggest that any of them
is unimportant to our overall evaluation of the support
Ala
Appendix A
in the record for the ultimate determination of the
scientific reliability of the device.
Certainly, there is adequate support in the record
for the Special Master’s finding that the Alcotest is not
as operator-dependent as was the breathalyzer. (Special
Master’s Findings 2(g), 8). Indeed, the testing sequence
we have described is almost entirely controlled and
prompted by the device and, with only a very few
exceptions, the operator is not able to influence the
manner in which the test is administered. Similarly,
there is ample support for the finding that the Alcotest
is well-shielded from the impact of any potential RFI
that might otherwise affect the reported results or limit
our confidence in the accuracy of the test results.
(Special Master’s Finding 4).
The parties agree, as well, about certain of the
Special Master’s recommendations for future revisions
in the firmware that will provide additional information
on the reported results that the device generates. For
example, the parties agree that the firmware should be
rewritten so that the AIR, solution change report, and
calibration documents include the temperature probe
serial number and probe value (Special Master’s Finding
2(a)); that if the particular test subject has not received
a reportable result, the AIR must include a statement
identifying why that occurred (Special Master’s Finding
2(e)); and that future calibration, certification and
linearity reports should include the serial number of the
Ertco-Hart digital temperature measuring system
utilized in performing those testing and maintenance
operations (Special Master’s Finding 2(i)).
Ra
Appendix A
As to each of these recommendations, there is
sufficient evidence in the record to support the
conclusion that the addition of this information for future
firmware revisions might be of some assistance to future
defendants. Notwithstandin.” our agreement that these
proposed alterations, to which the State has acceded,
might be beneficial, we discern no basis in the record
that suggests that any previously-generated report that
lacks these additional details is therefore insufficient
as a matter of proof of a per se violation. Rather, we
agree with the Special Master that updating the
firmware to provide this information in addition to that
which it already provides would merely be beneficial.
Similarly, the Special Master recommended, and the
parties by and large agrve, that the State should create
and maintain a centralized database of information
regularly uploaded through modem (Special Master’s
Finding 7), and that defendants should have access to
centrally collected and maintained data on their own
cases, as well as to the compiled scientific data on
matters involving others that has been redacted to
shield the personal information related to those other
individuals as appropriate (Special Master’s Finding
2(h)).”° Our review of the record satisfies us that there
20. The amicus NJSBA suggests that defendants should
have access to previously downloaded, centrally collected data.
We do not perceive this to be different from the Special Master’s
recommendation in this regard and the extent of the access to
be afforded to any litigant does not appear to be a matter in
dispute. In the absence of any suggestion in the record that
there is a genuine difference of agreement among the parties
on this matter, we see no need to address it further.
43a
Appendix A
is substantial, credible evidence that supports the
Special Master's recommendation concerning the
creation and maintenance of a regularly-updated
database, as well as his recommendation relating to
providing access to that data to defendants.
VI. Standards of Review
We turn, then, to the matters as to which the parties
are deeply divided. In part, our task is made more
complicated by the fact that some of the shortcomings
in the operation of the device can only be corrected with
respect to future uses of the machine, leaving,
potentially, doubt as to the validity of the previously-
generated AlRs which form the basis for prosecutions
stayed pending the outcome of these proceedings.
Moreover, our task has become further complicated by
the questions raised by the United States Supreme
Court’s recent Confrontation Clause”! cases, see
Crawford v. Washington, 541 U.S. 36, 124 S.Ct. 1354,
158 L.Ed.2d 177 (2004); Davis v. Washington, 547 U.S.
813, 126 S.Ct. 2266, 165 L.Ed.2d 224 (2006); cf Whorton
v. Bockting, _ U.S. _, 127 S.Ct. 1178, 167 L.Ed.2d 1
(2007), as to which we must proceed with great care
when the only “witness” confronting a defendant is a
machine.
21. Because the Crawford implications were not
thoroughly briefed in connection with our consideration of the
Special Master’s Initial or Supplemental Reports, we invited
the parties to submit additional briefs directed to these issues,
which we have considered.
sdo
Appendix A
We begin, as we must, with a brief review uf the
applicable principles of law governing admissibility of
novel scientific evidence. Admissibility of scientific test
results in a criminal trial is permitted only when those
tests are shown to be generally accepted, within the
relevant scientific community, to be reliable. See State
v. Harvey, 151 N.J. 117, 169-70, 699 A.2d 596 (1997)
(citing Frye v. United States, 293 F. 10138, 1014
(D.C.Cir.1923)); Romano, supra, 96 N.J. at 80, 474 A.2d
1; Johnson, supra, 42 N.J. at 170-71, 199 A.2d 809. That
is to say, the test must have a “sufficient scientific basis
to produce uniform and reasonably reliable results and
will contribute materially to the ascertainment of the
truth.” State v. Hurd, 86 N.J. 525, 586, 482 A.2d 86 (1981)
(quoting State v. Cary, 49 N.J. 348, 352, 230 A.2d 384
(1967)). As we have previously commented, however,
proof of general acceptance is often “elusive.” Harvey,
supra, 151 N.J. at 171, 699 A.2d 596.
Proof of general acceptance does not mean that
there must be complete agreement in the scientific
community about the techniques, methodology, or
procedures that underlie the scientific evidence. See
Romano, supra, 96 N.J. at 80, 474 A.2d 1. Even “the
possibility of error” does not mean that a particular
scientific device falls short of the required showing of
general acceptance. /bid. As we long ago recognized,
“{p]ractically every new scientific discovery has its
detractors and unbelievers, but neither unanimity of
opinion nor universal infallibility is required for judicial
acceptance of generally recognized matters.” Johnson,
supra, 42 N.J. at 171, 199 A.2d 809. Neither “complete
45a
Appendix A
agreement over the accuracy of the test [nor] the
exclusion of the possibility of error” is required. Harvey,
supra, 151 N.J. at 171, 699 A.2d 596.
Nevertheless, before we can conclude that scientific
test results are admissible in evidence, the proponent
of the scientific device must bear its burden to “clearly
establish” that,the device or the test meets the standard
of general acceptance as we have defined it. /d. at 170,
699 A.2d 596; see State v. Kelly, 97 N.J. 178, 209-11, 478
A.2d 364 (1984); State v. Cavallo, 88 N.J. 508, 521, 448
A.2d 1020 (1982).
VII. Defendants’ Challenges to
Scientific Reliability
Defendants raise three distinct sets of challenges
to the basic scientific reliability of the Alcotest. First,
they attack it on numerous traditional grounds relating
to scientific acceptance, not unlike the challenges raised
in Romano with regard to two breathalyzer models, by
contesting many of the Special Master’s findings and
recommendations. Second, defendants separately attack
the source code utilized to operate the device as being
so inherently flawed as to be independently lacking in
scientific reliability. Third, following the United States
Supreme Court’s lead in Crawford, defendants attack
the admissibility of documents generated by or in
connection with the device, which the Special Master
suggested be routinely admitted into evidence, as
violating their constitutional rights under the
Confrontation Clause.
46a
Appendix A
In addition, the State, although urging us to adopt
the Special Master’s conclusion about the general
scientific reliability of the device, argues that many of
his recommendations are unnecessary and that none of
them undermines the accuracy of any of the previously-
reported BAC results for any defendant. The State
therefore contends that the majority of the Special
Master’s recommendations are merely precatory, that
is, suggestions that the State may or may not elect to
adopt. Finally, the NJSBA, although in large part
agreeing with the Special Master’s findings and
conclusions, suggested a refinement to his
recommendation relating to minimum breath sample
criteria.
In reviewing the findings and conclusions set forth
by the Special Master in his report, we employ our
ordinary standards of review, considering them in the
same manner as we would the findings and conclusions
of a judge sitting as a finder of fact. We therefore accept
the fact findings to the extent that they are supported
by substantial credible evidence in the record, see
Locurto, supra, 157 N.J. at 472, 724 A.2d 234, but we
owe no particular deference to the legal! conclusions of
the Special Master, see Manalapan Realty, L.P v. Twp.
Comm. of Manalapan, 140 N.J. 366, 378, 658 A.2d 1230
(1995). With these standards in mind, we turn to our
analysis of the issues in dispute.
47a
Appendix A
VIII. Disputed Findings and Recommendations
We begin our discussion by more specifically
identifying the three categories of disputed findings and
recommendations. First, there are a number of disputes
about the criteria employed by the Alcotest to identify
an acceptable breath sample and convert the
measurement data into a reported result. This category
includes the Special Master’s recommendations on each
of the following matters: (a) the utilization of the 2100
to 1 blood/breath ratio (Special Master’s Finding 1(b));
(b) the minimum breath sample criteria (Special
Master’s Finding 6); (c) the requirement for the addition
of a breath temperature sensor (Special Master’s
Finding 9); and (d) the acceptable tolerance among test
results (Special Master’s Finding 10).
Second, there are a number of disputes arising from
the supplemental remand that relate to the firmware
and source code analysis. This category includes the
Special Master’s recommendations about each of the
following matters: (a) the fuel cell drift algorithm;
(b) the weighted averaging sequence; and (c) the
adequacy of the overall software design. In additicn,
although the parties agree on the need to revise the
firmware to address two shortcomings identified
through the source code analysis, namely, the buffer
overflow error and the disabling of the catastrophic
error detector, to the extent that these conceded errors
may have an impact on the reliability of AIR results
pending modification of the firmware, we are compelled
to address them as well.
48a
Appendix A
Finally, there are a number of issues that arise as a
result of the Special Master’s findings and
recommendations concerning foundational evidence
(Special Master’s Findings 5(a), 5(b)). This category
includes all of the following recommendations: (a) the
requirement for disclosure of foundational documents
as a prerequisite for admissibility of any Alcotest results;
(b) the required foundational documentary proofs at
trial; (c) the admissibility or uses of incomplete reports;
and (d) the constitutionally-required testimonial proofs.
We begin, then, with the disputed findings and
recommendations as they relate to the criteria employed
by the Alcotest for the collection of an adequate breath
sample and the creation of an acceptable and reportable
result.
A. Blood/Breatk Ratio
As we have previously noted, the drunk driving
statutes in New Jersey define the offense in terms of
BAC. In the majority of cases involving individuals
charged with these offenses, however, the particular
defendant has not undergone a blood test but instead
has submitted to a breath t.st. Modern breath testing
devices include an internal mechanism that collects an
acceptable breath sample and converts the alcohol
detected in the breath (BrAC) into a measure of the
person’s BAC.
Historically, breath testing devices convert from
BrAC to BAC by using a mathematical calculation based
49a
Appendix A
upon a scientifically accepted, judicially established
blood/breath ratio. The Alcotest utilizes a blood/breath
ratio of 2100 to 1, a ratio that this Court has previously
considered as a part of a challenge to the breathalyzer.
See Downie, supra, 117 N.J. at 460-63, 569 A.2d 242.
The Special Master concluded that the 2100 to 1
blood/breath ratio adopted by this Court in Downie and
utilized by the Alcotest remains a valid measuring
mechanism. He based this conclusion on the opinions of
three of the State’s experts and on a number of
published studies here and abroad relating to the
average, or mean, blood/breath ratio that he found to
be authoritative.” At the same time, the Special Master
rejected the opinions offered by two of the experts who
testified on behalf of the defendants. He found that the
analysis of one of these experts was filled with so many
errors that it could not be reliable, and he rejected as
flawed the assertion of the other defense expert that
the Alcotest actually does not test alveolar air.
Defendants nonetheless assert that the continued use
22. See, e.g., Allan R. Gainsford, et al., A Large-Scale Study
of the Relationship Between Blood and Breath Alcohol
Concentrations in New Zealand Drinking Drivers, 51 J.
Forensic Sci. 173 (2006); Alan Wayne Jones & Lars Andersson,
Variability of the Blood/Breath Alcohol Ratio in Drinking
Drivers, 41 J. Forensic Sci. 916 (1996). These studies appeared
in the Journal of Forensic Sciences, which our Appellate Division
has noted is an authoritative publication in the field of forensic
science. See State v. Miller, 64 N.J.Super. 262, 268-69, 165 A.2d
829 (App.Div.1960) (citing Journal of Forensic Sciences to
support reliability of breath test).
50a
Appendix A
of the 2100 to 1 ratio is not scientifically supported and
they urge us to reject any use of the Alcotest on this
basis.
The true focus of our analysis on this issue must be
on whether there has been any development in the
scientific community in the time since we decided
Downie that undermines our continued confidence in
the accuracy and validity of the conclusion we drew there
about the 2100 to 1 blood/breath ratio. Simply put, there
is not. Our review of the record demonstrates that the
arguments that we considered and rejected in Downie
have been raised anew, but there is no basis on which to
conclude that the continued utilization of this ratio is in
any way in error.
We reach this result for reasons similar to those that
we relied upon in Downie. First, we defer to the findings
of the Special Master concerning the credibility of the
expert witnesses who testified. See Locurto, supra, 157
N.J. at 471, 724 A.2d 234. In part, his credibility analysis
reflects the fact that one of defendants’ experts candidly
conceded that the use of this ratio generally tends to
underestimate blood alcohol, to the benefit of the test
subject.
Second, although there is some evidence that there
is a percentage of the population for whom the 2100 to
1 blood/breath ratio may actually overstate the presence
of blood alcohol, this evidence is not significantly
different from the record considered in Downie, supra,
117 N.J. at 460, 569 A.2d 242. Scientific studies
Sla
Appendix A
comparing actual blood alcohol content to breath-tested
alcohol content found only a minute number of
individuals for whom this ratio would have incorrectly
reported a result over the established legal limit for
driving while intoxicated. The percentage of individuals
for whom there may be an overestimation by use of this
ratio remains “extraordinarily small.” Jd. at 469, 569 A.2d
242.
Finally, defendants’ experts on this issue did not
produce any evidence to the effect that the ratio is
regarded by authorities in the field with even the
slightest suspicion or is otherwise subject to any
significant scientific challenge. Indeed, the
overwhelming evidence demonstrates that use of this
ratio tends to underestimate the actual BAC in the vast
majority of persons whose breath is tested. Although,
as in Downie, there may be a small number of individuals
who are disadvantaged by a device that uses the 2100
to 1 blood/breath ratio, there is sound scientific support
for its continued utilization.
We are confident, based on our review of the record
and our evaluation of the Special Master’s findings, that
there is sufficient credible evidence to support his
findings as to the continued validity of the 2100 to 1
blood/breath ratio. We therefore reject defendants’
challenge to its use and we adopt the Special Master’s
recommendation that it continue to be utilized in the
Alcotest.
52a
Appendix A
B. Minimum Test Sample Criteria
As we have explained, the Alcotest is programmed
to require that a test subject produce a breath sample
that meets four minimum criteria before the sample is
considered to be sufficient for purposes of deriving an
accurate test result. The Special Master recommended
approval, in general, of four minimum criteria for a
breath sample, which are: (1) minimum volume of 1.5
liters; (2) minimum blowing time of 4.5 seconds; (3)
minimum flow rate of 2.5 liters per minute; and (4) that
the IR measurement reading achieves a plateau (i.e.,
the breath alcohol does not differ by more than one
percent in 0.25 seconds). However, the Special Master
also found that there was credible evidence to support
lowering the minimum breath volume from 1.5 to 1.2
liters for women over the age of sixty. He recommended
that the State reprogram the device to reflect that
finding, but found no need to lower the minimum volume
for the general population.
Although both defendants and the State agreed with
these recommendations, the amicus NJSBA suggested
that the minimum breath volume be reduced to 1.2 liters
for all persons, so as to avoid a potential equal protection
challenge to the tests. Because no party has raised a
challenge to any of these criteria other than the
minimum required volume and because the Special
Master’s findings as to the other minimum criteria are
based on substantial credible evidence, we consider only
the minimum breath volume issue.
53a
Appendix A
1. Scientific Data Concerning Breath Volume
Breath alcohol concentration increases, in general,
as exhalation continues and deep alveolar air is expelled.
The rate of increase in alcohol concentration declines
as a person exhales, but the breath alcohol concentration
itself continues to increase until exhalation ends. The
record reflects that the minimum breath volume for the
Alcotest in New Jersey was fixed at 1.5 liters because
the State’s experts believe that this volume will exceed
the point after which most of the relatively rapid rise in
concentration has occurred and the average person is
in a fairly level part of the exhalation curve. In addition,
the State’s experts contend that 1.5 liters is the
minimum volume necessary for an accurate BAC
calculation because samples of lesser volume, in general,
do not include deep lung air.
At present, the most commonly used minimum
breath sample among the states is 1.5 liters. That
requirement, however, is not universal. Fo: example,
Alabama, where the Alcotest is currently in use, has
adopted a minimum sample requirement of 1.3 liters for
all test subjects. Moreover, although the experts
generally agreed that 1.5 liters is the optimal minimum,
some people may be incapable of providing that sample.
In particular, the record demonstrates that as
women age, they have an increasingly difficult time
producing a 1.5 liter breath sample. Data from Alabama
introduced during the proceedings shows that women
aged sixty to sixty-nine have more difficulty producing
54a
Appendix A
the 1.5 liter minimum requirement than their younger
counterparts. One of the State’s experts cited a study
from Germany* that demonstrated that women from age
sixty- to sixty-nine have an average breath volume of
1.4 liters, women seventy and over have an average of
1.3 liters, and women eighty and over have an average
volume of 1.2 liters. The German study included data
that demonstrates that men, regardless of age, were
capable of producing a sample of 1.5 liters. Indeed,
Brettell also conceded that his own study data confirmed
the accuracy of the assertion that older women were
the only ones unable to produce a sample of 1.5 liters.
Based on this data and the expert opinions offered
during the hearing, the Special Master recommended
that the minimum breath sample be fixed at 1.5 liters
for all test subjects except for women over the age of
sixty. He suggested that the device be reprogrammed
to require women over the age of sixty to provide a 1.2
liter minimum sample for a valid test result. Although
defendants and the State agreed with these
recommendations, the NJSBA suggests that this Court
should instead require that the minimum required
sample volume for all subjects be reduced from 1.5 to
1.2 liters in order to avoid a future potential equal
protection challenge.
23. Although it is not entirely clear, it appears that the
study, a copy of which was marked in evidence, is only available
as an unpublished manuscript. See G. Schoknecht & B. Stock,
The Technical Concept for Evidential Breath Testing in
Germany 1 (1995)(unpublished manuscript, Institute of
Biophysics).
55a
Appendix A
There is substantial credible evidence in the record
to support the Special Master’s findings and
recommendations concerning the required minimum
breath sample volume. The assertion by the NJSBA that
adopting a different standard for women over the age
of sixty than we apply to all other test subjects might
give rise to an equal protection challenge, however,
requires our careful consideration.
The minimum breath volume is significant, in and
of itself, because the Alcotest is programmed to
determine whether the four minimum criteria have been
met in a precise order, the first of which is the volume
analysis. A sample that falls short of the currently
required 1.5 liter volume measurement will be found to
be unacceptable. In that event, the Alcotest will report
the amount of air delivered and will display an error
message which reads: “minimum volume not achieved.”
The Alcotest permits up to eleven attempts to collect
two breath samples, after which, the only options that
the device offers are “terminate” or “refusal.” If the
operator chooses terminate, the Alcotest will reset and
the subject can then be given the opportunity for eleven
more attempts. If the operator chooses “refusal,” the
test sequence ends, but the operator is not required to
issue a summons for refusal. N.J.S.A. 39:4-50.4a.
Charging an arrestee with refusal remains largely within
the officer’s discretion. See generally State v. Widmaier,
157 N.J. 475, 724 A.2d 241 (1999).
Although an Alcotest operator has several options
if the device reports that the test sample is inadequate,
56a
Appendix A
the fact remains that one of them, refusal, carries with
it the possibility of severe sanctions. See N.J.S.A. 39:4-
50.4a. In the face of abundant evidence in the record
that there is an identifiable group in the test population
who may be physiologically incapable of complying, the
risk of permitting the device to reject samples from
members of that group and, by extension, authorizing
the issuance of a summons for refusal, is unjust.
By the same token, however, if the machine were
reprogrammed to accept the lowered volume from a
woman of the appropriate age, even if she could produce
the ordinarily required higher volume but attempted
to limit her breath output to avoid producing the deep
lung air needed for the most accurate analysis, the
machine would reject the sample because it would not
achieve the plateau. It is therefore clear that lowering
the volume for this identifiable group of test subjects
will not, in reality, afford them any advantage over
others. The constitutional question raised by the
NJSBA, however, also requires us to consider whether
it will disadvantage the other individuals required to
take the test.
2. Equal Protection and Lowered Breath
Volume Requirement
Lowering the minimum breath volume for women
over sixty implicates both age and gender classifications
and requires us to consider a potential challenge
brought pursuant to both the federal and state
constitutions. Because these standards are different and
57a
Appendix A
because the decision-making paradigm is different in
the federal and state courts, we address them in turn.
The Equal Protection Clause of the United States
Constitution mandates that no state shall “deny to any
person within its jurisdiction the equal protection of the
laws.” U.S. Const. amend. XIV, § 1. The Equal Protection
Clause “is essentially a direction that all persons
similarly situated should be treated alike.” City of
Cleburne v. Cleburne Living Ctr, Inc., 473 U.S. 432,
439, 105 S.C:. 3249, 3254, 87 L.Ed.2d 313, 320 (1985).
The federal equal protection analysis looks to the
characteristics of the impacted protected class or the
nature of the right being affected by the government
action.
The federal test used to evaluate an age-based
challenge is concerned with whether “the age
classification in question is rationally related to a
legitimate state interest. The rationality commanded by
the Equal Protection Clause does not require States to
match age distinctions and the legitimate interests they
serve with razorlike precision.” Kimel v. Fla. Bd. of
Regents, 528U.S. 62, 83, 120 S.Ct. 631, 646, 145 L.Ed.2d
522, 542 (2000). On the other hand, if the government
distinguishes between males and females, the
classificatior is subject to a heightened scrutiny. Nev.
Dep't of Hunan Res. v. Hibbs, 5388 U.S. 721, 728, 123
S.Ct. 1972, .978, 155 L.Ed 2d 953, 963 (2003). For a
gender classification to survive this scrutiny, the
government “must show ‘at least that the [challenged]
classificationserves ‘important governmental objectives
58a
Appendix A
and that the discriminatory means employed’ are
‘substantially related to the achievement of those
objectives.”” United States v. Virginia, 518 U.S. 515,
533, 116 S.Ct. 2264, 2275, 185 L.Ed.2d 735, 751 (1996)
(alteration in original) (quoting Miss. Univ. for Women
v. Hogan, 458 U.S. 718, 724, 102 S.Ct. 3331, 3336, 73
L.Ed.2d 1090, 1098 (1982) (quoting Wengler v. Druggists
Mut. Ins. Co., 446 U.S. 142, 150, 100 S.Ct. 1540, 1545, 64
L.Ed.2d 107, 114 (1980))).
Unlike its federal counterpart, the New Jersey
Constitution does not contain an equal protection
clause. Instead, we have found that “[a] concept of equal
protection is implicit in Art. I, par. 1 of the 1947 New
Jersey Constitution. ...” McKenney v. Byrne, 82 N.J.
304, 316, 412 A.2d 1041 (1980). Therefore, even though
Article I, paragraph 1 of our Constitution does not
include the phrase “equal protection,” “it is well settled
law that the expansive language of that provision is the
source for [this] fundamental constitutional guarantee
[ J.” Sojourner A. v. N.J. Dep’t of Human Servs., 177
N.J. 318, 332, 828 A.2d 306 (2003).
“Although conceptually similar, the right under the
State Constitution can in some situations be broader
than the right conferred by the Equal Protection
Clause.” Doe v. Poritz, 142 N.J.1, 94, 662 A.2d 367 (1995).
Indeed, we have held that our Constitution provides
“analogous or superior protections to our citizens” in
the context of equal protection. Peper v. Princeton Univ.
Bd. of Trs., 17 N.J. 55, 79, 389 A.2d 465 (1978).
59a
Appendix A
[Wjhere an important personal right is
affected by governmental action, this Court
often requires the public authority to
demonstrate a greater “public need” than is
traditionally required in construing the
federal constitution. Specifically, it must be
shown that there is an “appropriate
governmental interest suitably furthered by
the differential treatment.”
[Taxpayers Ass’n of Weymouth Twp. v.
Weymouth Twp., 80 N.J. 6, 48, 364 A.2d 1016
(1976) (citing Collingswood v. Ringgold, 66
N.J. 350, 370, 331 A.2d 262 (1975)).]
In considering equal protection-based challenges,
we have not followed the traditional equal protection
paradigm of the federal courts, which focuses rigidly on
the status of a particular protected class or the
fundamental nature of the implicated right. Instead,
when analyzing equal protection challenges under New
Jersey’s Constitution, we have applied a balancing test
that weighs the “nature of the affected right, the extent
to which the governmental restriction intrudes upon it,
and the public need for the restriction.” Caviglia v.
Royal Tours of Am., 178 N.J. 460, 473, 842 A.2d 125
(2004) (quoting Greenberg v. Kimmelman, 99 N.J. 552,
567, 494 A.2d 294 (1985)).
Finally, in addressing equal protection challenges
raised in the context of the exercise of police power, we
have held that “{t}he constitutional principles of due
60a
Appendix A
process and equal protection demand that the exercise
of the power be devoid of unreason and arbitrariness,
and the means selected for the fulfillment of the policy
bear a real and substantial relation to that end.”
Katobimar Realty Co. v. Webster, 20 N.J. 114, 123, 118
A.2d 824 (1955).
There are, in theory, two potential equal protection
challenges to the adoption of a different minimum
volume standard for women over the age of sixty. First,
one could argue that the lowered volume allows testing
of a smaller sample of shallower depth and therefore
results in a lower BAC reading. As to this challenge, it
is undisputed that the device will not accept a sample
that has not reached a plateau. An older woman who is
capable of producing a greater volume of air but does
not do so can be identified by her failure to meet the
plateau. Therefore, we can be certain that all test
subjects, regardless of age or gender, will only achieve
a valid sample when the deeper lung air is included.
Second, one could argue that the differentiation
permits older women who produce a sample with a
volume between 1.2 and 1.5 liters to avoid being charged
with refusal but exposes both younger women and all
men who provide samples of the same volume to be
prosecuted with that offense. The record on which the
differentiation between the test groups is based,
however, demonstrates that the older women, and only
the older women, may be physically incapable of
producing the larger sample.
6la
Appendix A
The right to equal protection does not require us to
scrutinize gender distinctions that are based on real
physiological differences to the same extent we would
scrutinize those distinctions when they are based on
archaic, invidious stereotypes about men and women.
See State v. Vogt, 341 N.J.Super. 407, 418, 775 A.2d 551
(App. Div.2001) (recognizing that “(t]he Equal Protection
Clause ...does not demand that things that are different
in fact be treated the same in law, nor that a state
pretend that there are no physiological differences
between men and women”). Similarly, the federal courts
have recognized that not all sex-based differentiations
are actionable. For example, in the employment context
some “standards that appropriately differentiate
between the genders are not facially discriminatory.”
Jespersen v. Harrah’s Operating Co., 444 F.3d 1104,
1109-10 (9th Cir.2006); see Healey v. Southwood
Psychiatric Hosp., 78 F.3d 128, 182 (3d Cir.1996)
(recognizing that gender may, in certain defined
circumstances, be a bona fide occupational qualification
for employment).
Applying the principles we have derived from both
the federal and state constitutional analyses, we discern
no meritorious ground for an equal protection challenge
to the proposed two-tiered approach for minimum
breath sample volume, regardless of which level of
scrutiny we apply. Viewed against our flexible approach
to equal protection challenges as derived from
Article I, paragraph 1 of our Constitution, the system
survives the constitutional challenge. The governmental
policy of achieving accurate breath samples as part of
62a
Appendix A
law enforcement’s role in ridding our roads of drunk
drivers is appropriately coupled with the authority to
prosecute for refusal. The proposed two-tiered system
for minimum breath volume, however, is neither
unreasonable nor arbitrary for it advances these goals
without holding the identified class, older women, to a
standard that they cannot meet. In this manner, the
policy goals are fulfilled through “means... [that] bear
a real and substantial relation to that end.” Katobimar,
supra, 20 N.J. at 128, 118 A.2d 824.
Similarly, under either the rational relationship test
applicable to age-based classifications, or the heightened
level of scrutiny applied to gender-based classifications
under the federal constitution, the lowered requirement
for women over sixty passes constitutional muster. The
policy goals we have identified for our state
constitutional analysis are, in federal parlance,
“important governmental objectives,” see Hibbs, supra,
538 U.S. at 728-29, 123 S.Ct. at 1978, 155 L.Ed.2d at
963. The selection of the two tiers for this aspect of the
test requirements is both rationally related to those
goals and “substantially related” to their achievement.
Ibid.
Notwithstanding the concern voiced by the NJSBA,
there is no scientific or other ground in the record to
direct that the minimum volume be lowered for all test
subjects. On the contrary, there is ample support for
the Special Master’s two-tiered approach and we discern
no equal protection violation in lowering the required
breath volume to 1.2 liters for women over the age of
sixty.
63a
Appendix A
3. Application to Pending Prosecutions
Our conclusion that the firmware must be revised
to accept a minimum breath volume sample of 1.2 liters
from women over the age of sixty requires us to consider
the impact of this directive for pending prosecutions.
We presume that there may be women who meet this
criteria and whose prosecutions have been stayed
pending our decision on these issues. For the sake of
completeness of our analysis, we address briefly the
possible factual scenarios relevant to these defendants.
First, there may be defendants who attempted but failed
to achieve a sufficient volume for an acceptable sample.
These individuals will be readily identified by an AIR
with a breath volume error message. Obviously, proof
of the charge of drunk driving for these women can only
be based on observational proofs because there will be
no reportable BAC results in an AIR.
The significance of the lowered breath sample
volume, however, rests less in the evidence utilized to
support a charge of drunk driving and more in its
relationship to a charge of refusal. In light of the
scientific evidence that we have found to be persuasive,
in the absence of some other evidence that supports the
conclusion that any such individual was capable of
providing an appropriate sample, by volume, we must
assume that she was unable to do so. For these
individuals, then, an AIR demonstrating insufficient
breath volume may not be used as proof on a charge of
refusal. On the other hand, if the AIR demonstrates that
a woman over the age of sixty was able to provide at
64a
Appendix A
least one sample that was deemed to be sufficient for
purposes of the 1.5 liter volume requirement, but she
failed to do so on a subsequent attempt, the AIR
demonstrating those facts may be utilized as evidence,
albeit not conclusive proof, in support of a refusal
charge.
C. Breath Temperature Sensor
The Special Master also recommended that in the
future the State acquire and utilize a breath
temperature sensor device separately marketed by
Draeger,” and that, in the interim, all previously
reported results be reduced by 6.58 percent to account
for breath variations in individuals tested. (Special
Master’s Finding 9). This recommendation was based
on the Special Master’s factual findings about breath
temperature.
We are compelled to reject this recommendation
because there is insufficient support in the record for
the factual findings on which it is based. In particular,
the Special Master found that “[mJost breath analyzers
24. There are several temperature devices related to the
Alcotest. One, which is an integral part of each device, and the
report of which is included on the AIR, heats the simulator
solution in the control test both in the device and, by extension,
in the calibration process. Another heats the breath tube, but
not the subject’s actual breath sample, to prevent condensation.
The device that is the focus of this recommendation, is an
optional device that tests the temperature of the actual breath
sample and reports it.
65a
Appendix A
used in the United States operate on the assumption
that the temperature of an expired breath sample is 34
degrees C[elsius],” but that “[rJecent scientific research
supports the proposition that the temperature of an
expired breath sample is actuaily almost 35 degrees
C(elsius].” He then found that BrAC increases by 6.58
percent for each degree above thirty-four degrees
Celsius, and reasoned that all BAC results should be
reduced by 6.58 percent to ensure their accuracy and
that the optional breath temperature sensor should be
used in the future. He noted, in support of his
recommendation, that the State of Alabama requires
reduction of all breath results from the Alcotest by this
percentage.
Although defendants and the NJSBA urge this Court
to adopt this finding and recommendation, in part based
on the assertion that the most relevant scientific
community is Alabama, the State argues that Alabama’s
program is an aberration and that this recommendation
is both unsupported and unsound.
We are persuaded to agree with the State for both
evidentiary and practical reasons.” First, the record
2 We reject, however, the State’s suggestion that a
measuring device that might more accurately determine BAC
and serve as a basis for a per se prosecution is an “option” that
falls within the sole discretion of the State in performing its
prosecutorial function. Rather, to the extent that the State seeks
to utilize a device, like the Alcotest, to prove a per se violation
of the statute, we think it abundantly plain that the decision as
to the accuracy of any innovation for proof purposes, consistent
with our Constitution, is ours to make.
66a
Appendix A
reflects that the generally accepted average
temperature for human breath is 34 degrees Celsius.
Only one study, performed in Alabama and therefore
relevant for that jurisdiction’s purposes, concluded that
the average breath temperature is closer to 35 degrees
Celsius. At best, then, there is a debate about average
breath temperature. In fact, however, there is no
support in the record for the Special Master’s
assumption that a rise in breath temperature increases
BrAC.
Notwithstanding that, some of the experts conceded
that a one-degree Celsius increase in breath
temperature could theoretically produce a 5.5 to 6.8
percent increase in BrAC, assuming that all other
variables remained constant. Accordingly, a one-degree
Fahrenheit increase in breath temperature could
theoretically cause the BrAC to rise by 3.8 percent.
There is, however, no evidence in the record that this
theoretical increase translates into an inaccurately
elevated BAC result.
Moreover, all of the experts agreed that even a
theoretical possibility of a link would not alter the
reported BAC readings in practice. That is, if a person
with a normal temperature submitted a breath sample
with a 0.07 percent BAC, that person’s breath test would
be read as being over 0.08 percent BAC only if he had a
2.5 degree-Celsius or 4.5 degree-Fahrenheit increase in
body temperature. There is no evidence in the record
from which we can conclude that there is any risk that
any individuals with such an elevated temperature are
67a
Appendix A
even being tested. There is also no evidence in the
record to support the finding that the average breath
temperature exceeds 34 Celsius or that an elevation of
the breath temperature, in and of itself, results in an
elevated BAC reading.
Second, to the extent that there might be a
relationship between the breath temperature of the
subject submitting the sample and BAC, there is
significant evidence in the record to support the finding
that an independent device to measure that
temperature or to reduce the results to account for it”
would be redundant. The device as currently configured
incorporates two methods that account for any possible
overestimation of the BAC reading that an elevated
breath temperature might theoretically cause, and they
operate to the benefit of the person being tested. Both
the truncation of results and the use of the 2100 to 1
blood/breath ratio, a ratio that in part takes
temperature into account, effectively underestimate the
calculation to the advantage of the test subject.
The debate about the effect of temperature is not
new. It was presented specifically in Foley, supra, and
in part in Downie, supra. The trial court in Foley, supra,
26. The record reflects that the Alcotest with the added
breath temperature device does not actually recalculate BAC
to account for elevations in breath temperature. Instead, in
Alabama, the sensor reports breath temperature and if it is
shown to be elevated above 34 degrees Celsius, the court reduces
the reported BAC results by a factor of 6.58 percent for every
degree.
68a
Appendix A
analyzing virtually the same factual assertions as are
included in this record, concluded that, apart from a test
subject suffering from a very high fever, the natural
variation of temperature was subsumed within the
variability of the blood/breath ratio. 370 N.J.Super. at
355, 851 A.2d 123. As that court recognized:
The factor of 2100 to 1 was developed by doing
studies on persons in the field including both
arrested subjects and research subjects. The
breath temperature of all these subjects
varied. Therefore, the 2100 to 1 ratio already
subsumes within it the variation in breath
temperature of the general population.
[Tbid. |
We, too, have previously considered the relationship,
in general, between temperature and the blood/breath
ratio, see Downie, supra, 117 N.J. at 462-63, 569 A.2d
242. We there concluded that the utilization of the 2100
to 1 ratio adequately accounts for any small impact that
a particular subject’s elevated temperature might
potentially have on the result.
Our review of the record convinces us that the
Alcotest BAC reading would not be made more accurate
by the addition of the breath temperature sensor or by
the across-the-board reduction of all values by 6.58
percent to account for the theoretical temperature
factor as suggested by the Special Master. More to the
point, perhaps, we reach our conclusion for practical
69a
Appendix A
reasons as well. The unrebutted evidence in the record
convincingly demonstrates that requiring the addition
of the breath temperature sensors would result in an
unreasonable maintenance burden to the program. In
fact, the record includes detailed descriptions of the
added steps, equipment, time and personnel that are
necessary simply to maintain and calibrate the
temperature sensors.”’ That added practical and
logistical burden on the State and the municipalities in
New Jersey, while perhaps not prohibitive, is
unreasonable in light of the scant basis in the record
that might support requiring the sensor.
Our evaluation of the evidence therefore leads us to
reject the Special Master’s recommendation concerning
utilization of a breath temperature sensor or reduction
in BAC results by a 6.58 percent factor as unsupported
by the factual record and unnecessary. Rather, we are
persuaded that the effect of breath temperature on BAC
is theoretical at best, and that the effect, if any, is
ameliorated because the Alcotest uses both truncation
and the 2100 to 1 blood/breath ratio to calculate BAC.
Because both of these safeguards effectively
underestimate BAC, any additional subtraction to
account for temperature is redundant and unnecessary.
27. Because of the equipment needed to do so, the
temperature sensors cannot be maintained or calibrated on-
site. Instead, the equipment must be taken out of service and
moved to a central location for these purposes, resulting in the
need for arrestees to be transported to an adjoining
municipality for testing while the equipment is undergoing
routine maintenance.
70a
Appendix A
We therefore reject the Special Master’s finding and
recommendations concerning the breath sensor and a
6.58 percent compensating reduction.
D. Acceptable Tolerance Analysis
The Special Master recommended that the firmware
be revised to correct the acceptable tolerance among
the reported results so as to permit results to be
accepted if they are within plus or minus 0.005 percent
BAC or plus or minus five percent of the mean for the
four readings, whichever is greater. (Special Master’s
Finding 10). Although the State does not dispute the
need to correct future firmware versions, both the
recommendation of the Special Master as to the
acceptable tolerance range and the effect of this
determination upon pending cases require our analysis.
The acceptable tolerance question raises a variety
of concerns, including its implications for the validity of
any particular test result, our confidence in the accuracy
and reliability of a specific Alcotest unit, the need for
performance of a third test on any particular test
subject, and the appropriate method by which to assess
tolerance in light of changes to the quantification of the
per se violation in recent years. We address each of these
difficult issues in turn.
Tila
Appendix A
1. Doubled Tolerance Range in
Firmware version 3.11
Tolerance is the range of any set of measurements
that is accepted as being representative of a true
reading. Precision and accuracy can be ensured by
requiring the application of a narrow range for tolerance.
Conversely, the wider the acceptable tolerance between
reported results, the lower our confidence in the
accuracy of any of the reported results. Therefore, for
purposes of permitting any device to be utilized for proof
of a per se violation of the statute, the acceptable
tolerance is of fundamental importance.
As a matter of historical perspective, we first
considered the question of acceptable tolerance ranges
in Romano, supra. There, as a part of our evaluation of
whether the test results obtained from two breathalyzer
models which might have been affected by radio
frequency interference (RFI) could be admissible, we
accepted the 0.01 percent BAC standard as a
scientifically reliable tolerance range, based on the
opinions of two experts who so opined, see Romano,
supra, 96 N.J. at 86, 474 A.2d 1. At the time, the statute
created a per se offense for any person whose BAC was
0.10 percent or greater, see id. at 78, 474 A.2d 1. As we
articulated the tolerance analysis in Romano,
“admissibility is satisfactorily established . . . [i]f the
breathalyzer results consist of two tests or readings
within a tolerance of 0.01 percent of each other... .”
Id. at 87-88, 474 A.2d 1. The point, of course, was that if
a breathalyzer that might be influenced by RFI could
72a
Appendix A
nevertheless read two separate breath samples with
results within this range, we would presume those
results were unaffected by external influences and,
therefore, valid.
After our decision in Romano, the 0.01 percent BAC
tolerance range became the benchmark against which
all breathalyzer results, not just those from RFI-
susceptible models, were tested for general reliability
and accuracy. In Downie, we again referred to the 0.01
percent BAC tolerance range as a benchmark for
reporting accurate results. See Downie, supra, 117 N.J.
at 455, 569 A.2d 242. Although we did not independently
evaluate the continuing validity of that tolerance range,
we adhered to it as a part of our evaluation of the overall
scientific accuracy and reliability of the breathalyzer.
Indeed, we have never departed from that standard and
have not previously been called upon to consider any
different articulation of that accepted range of tolerance.
Prior to the trial court’s decision in Foley, the
tolerance range for the Alcotest was fixed by the
software to be 0.01 percent BAC or a range of ten
percent for all samples. That range was determined by
Brettell when the Alcotest program was first devised.
The range, however, was tested by reference to the
arithmetic mean, the effect of which halves the
expression of the range. In addressing the challenge to
the tolerance as being inconsistent with Romano, the
court in Foley described the tolerance as fixed in the
Alcotest in somewhat different terms. The Foley court
explained that our long-accepted standard of a required
73a
Appendix A
tolerance of 0.01 percent BAC between two breath
samples was the “strictest standard in the United
States,” and concluded that, as applied to the four
results derived by Alcotest, “the additional parameter
of +10 [percent] is within the tolerance considered
acceptable for reliable results by the scientific
community.” Foley, supra, 370 N.J.Super. at 357, 851
A.2d 123.
In so articulating the tolerance range, however, the
court did not simply re-articulate a long-accepted
tolerance, expressing it as a percentage rather than an
absolute. Nor did it accurately express the tolerance
used by the device, an earlier version of software known
as Firmware version 3.8, in which the tolerance was
expressed in alternate terms. Rather, the court,
inadvertently, we think, endorsed a tolerance range that
effectively doublea that which we have allowed.
There are several considerations arising from this
expanded tolerance that are now before us. First, the
use of a percentage tolerance range tends to permit
readings at higher levels that are wide of the previously
accepted 0.01 percent BAC standard. This might lead
to results that are, in and of themselves suspicious in
terms of their intrinsic reliability. That is to say, although
for purposes of guilt, it might not matter whether we
accepted two test results that were within ten percent
but beyond 0.01 percent BAC of each other, those results
might raise a concern about the overall reliability of the
particular machine. Second, however, use of an absolute
rather than a percentage might arguably disadvantage
74a
Appendix A
subjects whose test results are at the lower end of the
range by accepting test results that are, by percentage,
more widely separated and that would be rejected as
out of tolerance were a percentage analysis applied.
Third, in some measure the amendments to the
statute and the creation of new per se offenses, not
extant when we considered the acceptable tolerance in
Romano and Downie, makes our evaluation of this issue
more complex. In the abstract, tested against a statute
that only utilized one per se test for drunkenness,
namely, 0.10 percent BAC, our acceptance of the single
test for acceptable tolerance was well supported in the
scientific record. The question, in light of the lowered
per se limits now in force, is what we should demand in
terms of precision to demonstrate accuracy and support
admissibility.
Taking into account these considerations, we turn
to an evaluation of the evidence in the record concerning
tolerance and its significance. At present, assuming the
subject has provided an otherwise acceptable sample,
the Alcotest reports the EC and IR results of the first
sample. The device is programmed to accept the EC and
IR test results from a second sample only if those results
are within its programmed tolerance of the EC and IR
results from the first breath sample. If the second-
sample results are not within the tolerance, the Alcotest
will record the results, but require a third sample.
For Firmware version 3.8, used in the Alcotest
program at issue in Foley, Brettell testified that he set
75a
Appendix A
the tolerance in accordance with the breathalyzer
tolerance expressed in Downie. He interpreted the
Downie standard to mean that two breath tests had to
be within 0.01 percent BAC of each other when the mean
BAC measured below 0.10 percent BAC, which was the
per se level when Downie was decided. Brettell testified
that, notwithstanding the fact that the Court never
varied from the 0.01 percent BAC standard, he assumed
we intended a tolerance of ten percent for BAC values
above 0.10 percent BAC. Therefore, Firmware version
3.8 was programmed to accept the second breath test if
there was no more than 0.01 percent BAC or ten
percent between the highest and lowest readings.
Notwithstanding Brettell’s acknowledgment that he
knew that the Foley statement about tolerance was
mathematically incorrect, he concedes that following the
decision in Foley, the State directed Draeger to
reprogram the device so as to take advantage of that
far wider, effectively doubled, range for tolerance. He
explained that he did so to make the test conform with
programs in other states and to address criticism of the
relative frequency with which the device in Foley rejected
results for being out of tolerance and required the
administration of a third test. Brettell believed that
taking advantage of the court-sanctioned wider
tolerance would alleviate a similar challenge in the
future. The State concedes that Firmware version 3.11
76a
Appendix A
did precisely that, creating a range of either plus ten
percent or minus ten percent of the mean, for a doubled
tolerance.”
2. Expert Testimony
Although New Jersey, prior to the introduction of
Firmware version 3.11, in compliance with our decision
in Romano and Downie, adhered to the 0.01 percent
BAC tolerance standard, there is no general agreement
among the states as to what standard is acceptable.
Many states other than New Jersey utilize the 0.01
percent BAC tolerance standard as well, but the
National Safety Council, for example, recommends a
tolerance of no more than 0.02 between the highest and
lowest readings.
One of the State’s witnesses, Rod Gullberg, testified
about his previously published conclusions on tolerance
measurement. He opined, therefore, that the Firmware
version 3.11 tolerance is too broad. See R.G. Gullberg,
Determining an Appropriate Standard for Duplicate
28. There is, in addition, a further distinction that is a
subtle one. Using a range, whether expressed in absolute or
percentage-based terms, when comparing two numbers as in
Romano, is not the same as expressing the same range for
tolerance among four numbers as evaluated against their
arithmetic mean. Technically, Firmware version 3.11 is a good
deal more sophisticated in its measure for tolerance. In fact the
device tests tolerance as the greater of plus or minus ten percent
of the mean of all four results or plus or minus 0.01 percent
BAC of that mean.
77a
Appendix A
Breath Test Agreement, 39 Can. Soc’y Forensic Sci. J.
15, 23 (2006). Instead, he recommended using plus or
minus five percent of the mean of the four tests. He
estimated that if the firmware were changed to utilize
this tolerance, the number of people who wouid have to
submit additional samples would increase by
approximately five percent. That estimate is mirrored
by a comparison of the data from Pennsauken, in which
Firmware version 3.8 was used, with the data from
Middlesex County, in which Firmware version 3.11, with
its doubled tolerance, was used.
Another of the State’s witnesses, Hansueli Ryser,
explained that if New Jersey used a tolerance of plus or
minus 0.005 percent BAC, or plus or minus five percent,
of the mean, whichever is greater, then for mean
measurements below 0.10 percent BAC, the acceptable
tolerance would be plus or minus 0.005 percent BAC.
As an example, if a person had a mean alcohol
concentration of 0.08 percent BAC, the tests would be
in tolerance if they fell between 0.075 and 0.085 percent
BAC.” For mean concentrations above 0.10 percent BAC,
the relevant tolerance would be plus or minus five
percent.
29. The significance of tolerance, as this example
demonstrates, is related to the truncation procedure. A test
subject with results ranging from 0.075 to 0.085 would not be
guilty of the per se violation because the machine must report
the lowest truncated value, here 0.07 percent BAC. By the same
token, however, the machine could not reject these results and
subject the individual to a third test, with a potentially higher
BAC.
78a
Appendix A
Brettell testified that he planned to “revisit” the
tolerance because it had caused “so much litigation.”
He testified that the 0.02 percent BAC National Safety
Council recommendation might be the easiest to adopt,
but he preferred the use of a combination of a set value
and a percentage because the percentage would account
for scientifically defensible wider tolerance at very high
values. Overall he favored® plus or minus 0.005 percent
BAC from the mean or plus or minus five percent of the
mean, whichever was greater.
3. Future Firmware Revisions
Although we have never considered the use of a
tolerance other than the absolute 0.01 authorized in
Romano, intervening legislative enactments require us
to address the continuing validity of that standard. At
the time that we decided the question of acceptable
tolerance in Romano, there was but one per se standard
for drunk driving prosecutions, namely, the 0.10 percent
BAC. Since that time, however, the Legislature has
reduced that per se limit to 0.08 percent BAC, while
maintaining the 0.10 percent BAC standard for
enhanced punishment.” The issue is what measure of
30. We are constrained to observe that, for purposes of
assessing scientific accuracy and therefore admissibility in
evidence as proof of a per se violation, “ease” and “simplicity”
are irrelevant. Similarly, a test based on whatever the current
director of the program “favors” is unlikely to withstand
scrutiny.
31. In addition, the separately-adopted per se limits that
apply to commercial drivers (0.04 percent BAC) and ind: viduals
(Cont'd)
79a
Appendix A
tolerance comports with scientifically reliable, and
therefore admissible, results.
Expressing the tolerance in terms of the greater of
the absolute or a percentage of deviation from the mean
authorizes, in effect, a wider range of tolerance at the
higher readings. There is, in this record, evidence that
demonstrates to our satisfaction that at the higher
readings, all measures of BAC are somewhat less precise
than they are at the lower ranges. As a result, the wider
tolerance expressed by a percentage deviation from the
mean applied to the upper ranges of possible readings
does not suggest that the device is not working properly.
At the lower readings, in contrast, a deviation outside
of the tolerance limit we have traditionally required most
assuredly will raise a question about the functioning of
the particular device.
Our evaluation of the record compels us to conclude
that, even in light of the lowered overall per se limit
adopted since Romano, the continued use of the absolute
0.01 percent BAC standard, coupled with the use of a
like range of tolerance expressed as a percentage
deviation from the mean, is both scientifically
appropriate and consistent with our understanding of
the intention of the Legislature in adopting these per
se limits.
(Cont'd)
under the legal drinking age (0.01 percent BAC) are entirely
new. Although the effect of the absolute measure of tolerance
might have less validity when applied to these separate offenses,
it is not challenged here and we do not address it.
80a
Appendix A
To the extent that Firmware version 3.11 took
advantage of an explanation of the tolerance range in
Foley that inadvertently doubled the permissible range,
however, it cannot be sustained. We therefore direct that
for future firmware revisions, the device be programmed
to fix the tolerance range to be plus or minus 0.005
percent BAC from the mean or plus or minus five
percent of the mean, whichever is greater, in order to
ensure scientifically accurate, admissible test results.
4. Application to Pending Prosecutions
Our inquiry, however, cannot end there. There is
stark evidence in the record, based on a comparison of
the data from the Pennsauken program, in which the
device with Firmware version 3.8 and the appropriate
tolerance was utilized, with the data collected in
Middlesex County, using Firmware version 3.11 and its
doubled range, that the intervening expansion of the
tolerance range resulted in tests being deemed
acceptable by the device that cannot meet the tolerance
range we have required. In fact, the data demonstrates
that precisely the effect that Brettell desired, namely,
reducing the frequency of out of tolerance readings that
required third samples, was achieved to the point of
apparent elimination. The Special Master, while
recommending that the software be revised for future
uses to reflect his analysis of acceptable tolerance
ranges, did not regard the State’s adoption of a different
and widely expanded tolerance to be problematical for
pending prosecutions. The State urges us to adopt this
finding that the doubled tolerance had no effect on any
defendant’s substantive rights. We disagree.
Sla
Appendix A
The simple fact is that the tolerance range is a critical
component in our conclusion that this or any other
device correctly and accurately measures breath alcohol
and converts that data into a scientifically reliable,
accurate BAC analysis. Our acceptance of those results
for purposes of supporting, without more, a criminal
conviction, must be based on our conclusion that the
results are reliable and accurate. The use of a doubled
tolerance, however, deprived some percentage of test
subjects of a third, and perhaps dispositive, test. At the
same time, it undermines our confidence in the accuracy
of the reports of those tests that fall outside of the range
that we have demanded be utilized as a prerequisite for
scientific accuracy and that undergirds admissibility in
a criminal proceeding.
It is easy enough to identify those individuals for
whom a third test should have been given. To be sure, if
we had the third test data for those defendants, some
of them would achieve a result within the authorized
tolerance and thus be shown to have violated the per se
limits. But just as surely, there may be others for whom
a third test would have yielded a result still further out
of range so as to, perhaps, call the accuracy of the
particular machine into question. And it is even possible
that there might be a defendant for whom a third test
would result in a reading that would meet the test for
tolerance but would exonerate that individual.
The suggestion that we permit those test results
that are outside of the range for tolerance to be utilized
for purposes of a per se conviction unfortunately is,
82a
Appendix A
simply put, unacceptable. Zealousness in ridding our
roads of drunk drivers cannot overcome our ordinary
notions of fairness to those accused of these offenses.
Therefore, we are constrained to direct not only that
future firmware updates utilize the tolerance
computation that we have concluded is acceptable, but
that all pending prosecutions include an evaluation of
whether the two reported test results exceeded this
acceptable tolerance.
Any AIR that reports results from tests of only two
breath samples, therefore, must be analyzed to
determine whether its results are within our accepted
tolerance by use of a mathematical calculation. The
appropriate calculation for this purpose will consist of
applying the following formula: (a) add the IR and EC
results given for the first breath sample to the IR and
EC results for the second breath sample; (b) divide the
sum calculated in (a) by 4 to derive the arithmetic mean;
(c) compute the upper limit of tolerance by taking the
larger value of the mean multiplied by 1.05 or the mean
plus 0.005 percent BAC; (d) compute the lower limit of
tolerance by taking the smaller of the value of the mean
multiplied by 0.95 or the mean minus 0.005 percent BAC;
(e) if all of the IR and EC results of the two samples fall
within the upper and lower limits of the tolerance range,
the AIR is valid, but if any of the results fall outside of
the tolerance range, the AIR is not valid.
Although we have prepared a worksheet that is
attached to the order that accompanies this opinion for
use in all prosecutions pending reprogramming of the
83a
Appendix A
device, two examples will, we think, illustrate the way in
which the formula should be utilized in practice to
differentiate between an AIR that reports results within
tolerance and one that does not. If, for example, a
defendant’s first breath test sample yielded an IR result
of 0.100 percent BAC and an EC result of 0.101 percent
BAC, and the second sample yielded an IR result of
0.104 percent BAC and an EC result of 0.103 percent
BAC, the calculations would be performed as follows:
(a) first all four of the results (two IR and two
EC) would be added, in this example, 0.100 +
0.101 + 0.104 + 0.103 = 0.408;
(b) next, the arithmetic mean would be
derived by dividing that sum by four, 0.408 / 4
= 0.102;
(c) then the upper limit of acceptable
tolerance must be determined by comparing
the two methods for computing the range,
namely, the use of the absolute or the
percentage. This is done by computing each
separately and selecting the greater of the
two. In this example, the computation would
yield the following options: (0.102 x 1.05 =
0.1071) OR (0.102 + 0.005 = 0.1070). Because
the greater of these is 0.1071, that will be the
correct upper tolerance limit;
(d) next, the lower limit of acceptable
tolerance must be derived by comparing the
84a
Appendix A
two methods for computing the range, again,
by using the absolute and the percentage
calculations. This is done by computing each
separately and selecting the lesser of the two.
In this example, the computation would yield
the following options: (0.102 x 0.95 = 0.0969)
OR (0.102-0.005 = 0.0970). Because the lesser
of these is 0.0969, that will be the correct lower
tolerance limit; and
(e) finally, by comparing all four of the
reported test sample results (0.100, 0.101,
0.104, 0.103) against this accepted tolerance
range of 0.0969 to 0.1071, it becomes plain that,
in this example, the AJR is valid because all
four test results fall within the accepted
tolerance range.
Because the Firmware version 3.11 utilized a
doubled tolerance range, there will be AIRs that will
not meet the test for tolerance that we have deemed to
be permissible. We therefore provide a further example
to illustrate the calculations relating to an AIR that
would be out of tolerance under this standard and,
therefore, inadmissible in a prosecution. If, for example,
a defendant’s first breath test sample yielded an IR
result of 0.089 percent BAC and an EC result of 0.080
percent BAC, and the second sample yielded an IR
result of 0.091 percent BAC and an EC result of 0.084
percent BAC, the calculations, which would be
85a
Appendix A
performed in the same manner, would yield a different
outcome, as follows:
(a) first, all four of the results (two IR and
two EC) would be added, in this example, 0.089
+ 0.080 + 0.091 + 0.084 = 0.344;
(b) next, the arithmetic mean would be
derived by dividing that sum by four, 0.344 / 4
= 0.086;
(c) then the upper limit of acceptable
tolerance must be determined by comparing
the two methods for computing the range,
namely, the use of the absolute or the
percentage. This is done by computing each
separately and selecting the greater of the
two. In this example, the computation would
yield the following options: (0.086 x 1.05 =
0.0903) OR (0.086 + 0.005 = 0.0910). Because
the greater of these is 0.0910, that will be the
correct upper tolerance limit;
(d) next, the lower limit of acceptable
tolerance must be derived by comparing the
two methods for computing the range, again,
by using the absolute and the percentage
calculations. This is done by computing each
separately and selecting the lesser of the two.
In this example, the computation would yield
the following options: (0.086 x 0.95 = 0.0817)
OR (0.085-0.005 = 0.0810). Because the lesser
86a
Appendix A
of these is 0.0810, that will be the correct lower
tolerance limit; and
(e) finally, by comparing all four of the
reported test sample results (0.089, 0.080,
0.091, 0.084) against this accepted tolerance
range of 0.0810 to 0.0910, it becomes plain that,
in this example, the AJR is invalid because
the first breath sample’s EC result (0.080) does
not fall within the accepted tolerance range.
The use in Firmware version 3.11 of the doubled
tolerance range, which we have rejected, requires that
all AIRs that report results of only two breath samples
be tested for validity against the tolerance range we
have accepted. Therefore, in all prosecutions stayed by
our January 10, 2006 Order, the State shall review the
BAC results as reported in the AIR and shall calculate
whether those results fall within tolerance, and the court
shall review those calculations and make them a part of
the record. In those cases in which this review reveals
that the results fall outside of the acceptable tolerance,
the AIR cannot be deemed to be sufficiently scientifically
reliable to be admissible and it shall not be admitted
into evidence as proof of a per se violation.
IX. Source Code Remand
We turn, then, to a series of issues that arose
following the supplemental remand for evaluation of the
source code. Not aii of the firmware issues we must
address are disputed, but our review of the record has
87a
Appendix A
identified issues that bear on the extent and manner in
which existing AIR results may be utilized in pending
prosecutions. We begin with the software- based
questions that are in dispute.
A. EC Readings and Fuel Cell Drift Algorithm
One of the most controversial findings that came out
of the second remand proceedings, during which the
parties were afforded the opportunity to undertake an
analysis of the source code that is the heart of the
operation of the Alcotest device, related to the EC
readings. During the proceedings, the Special Master
summoned Brian Shaffer, a Draeger employee
responsible for the code and for implementing changes
to the New Jersey Firmware since the Foley decision,
to testify. Near the end of his testimony, Shaffer
revealed that Firmware version 3.11 utilizes a
compensating algorithm to account, in part, for a
phenomenon known as fuel cell drift.
As Shaffer explained it, the EC reading is obtained
by passing an electrical current through a small sample
of the breath that has otherwise been captured for IR
testing in the cuvette. The fuel cell that creates the
electrical charge reacts in the presence of alcohol. The
reaction of the fuel cell can be represented graphically
as a curve and the percentage of alcohol in the breath
is measured by calculating the area under the curve
mathematically. As fuel cells age, the area under the
curve that expresses the same breath alcohol content
is unchanged, but the shape of the curve itself changes
88a
Appendix A
from a high sharp peak to a longer, flatter one. As a
matter of mathematical computation, the area being
measured is the same even though the curves, were they
plotted graphically, would appear to be different when
observed visually. |
According to Shaffer, the flattening of the curve is
caused by the aging of the fuel cell, which reacts more
slowly and with less intensity to the same amount of
alcohol than when the fuel cell is new. This phenomenon,
known as fuel cell drift, does not actually alter the
accuracy of the EC measurement. However, because the
fuel cell begins to react more slowly to the presence of
alcohol as it ages, a portion of the area under the curve
that is the basis for the alcohol measurement is not
captured during the time when the Alcotest EC data is
collected. Instead, a portion of the end of the curve is,
in essence, cut off, resulting in a lower than accurate
measurement.
Because fuel cell drift is a known scientific
phenomenon that would otherwise result in an
inaccurate underreporting of the percentage of alcohol
in the test subject’s breath, Draeger added a
compensating algorithm into the firmware. The EC fuel
cell drift algorithm, therefore, is intended to capture a
portion of the missing data and, in theory, create a more
accurate result as the reported EC reading. The
algorithm, however, does not attempt to quantify the
missing area under the curve per se, but instead
attempts to compensate in part for the lack of complete
data arising from the EC measurement. In the event
89a
Appendix A
that fuel cell drift is detected during the control test,
the algorithm mathematically increases the EC reading
that is reported by up to twenty-five percent of the
difference between the IR and EC readings from the
tests of the subsequent breath samples.
The compensating algorithm is not routinely applied,
but only functions if the appropriate preconditions are
met. The device, in performing the control test,
compares the EC and the IR readings and accurately
reports those results. Because the control test utilizes
a known test solution to ensure that the device is
functioning properly and that it accurately reads a
solution of a known percentage of alcohol, fuel cell drift
can be detected from the control test’s results. If the
device detects drift, the algorithm will adjust the EC
measurement standard, which, in turn, will slightly
increase the reported EC results for the test subject’s
breath sample to account for ‘he fuel cell drift.
The discovery of the EC fuel cell drift algorithm in
the source code prompted the Special Master to conclude
that more frequent re-calibration of the devices with
replacement of fuel cells that had become “depleted”
would reduce reliance on the EC fuel cell drift algorithm
and, therefore, increase the accuracy of the readings.
The State objects to this proposal as unnecessary and
burdensome, arguing that its current program of annual
calibration is sufficient.
Defendants, on the other hand, raise several
challenges to this EC algorithm, both in theory and in
90a
Appendix A
practice. First, they argue that it demonstrates that
Draeger’s claim that the device uses two completely
independent measurements for breath alcohol is false.
Second, they argue that it demonstrates that the device
is simply not accurate in any sense. Third, they argue
that the algorithm, which they attack as having been
hidden from them throughout the initial remand
proceedings, is evidence that the software may be
utilizing other hidden mechanisms that might inflate
readings so that the accuracy of the results can never
be reliable.
We do not share either the State’s or defendants’
concerns. The record reflects that a semi-annual
inspection and recalibration program recommended by
the Special Master is consistent with the manufacturer’s
recommendations. At the same time, it provides a useful
safeguard by affording a more regular opportunity to
evaluate and replace aging fuel cells. We discern no
reason to permit the State to continue to adhere to its
program of annual recalibration, particularly in light of
the concerns raised as to the utilization of a
compensating algorithm in the interim.
However, we do not find merit in defendants’
concerns about the EC algorithm or its use. There is
sound scientific evidence that supports the conclusion
that fuel cells begin to age as soon as they are put into
service and that fuel cell drift is inevitable. But there is
equally ample support for the proposition that even as
the intensity of the peak demonstrated by the EC
evaluation of the sample diminishes over time, the
Ola
Appendix A
reactive effect overall (that is, the area under the curve
being calculated) does not. Instead, the time within
which the test is performed simply truncates the EC
reading before all of the otherwise appropriate data can
be generated. Theoretically, one could, perhaps,
program the machine to calculate the missing area
based on a presumed regularly-shaped curve. Although
that might even be a more accurate method of supplying
the missing data, it would not, in the end, be as
advantageous to defendants as is the minor upward
adjustment that the algorithm effects. Indeed, because
the device will not generate a result that can be utilized
if the readings are out of tolerance, the algorithm alters
the EC result in an amount that, we are confident,
cannot fairly be seen as convicting the innocent.
Nor do we consider the fact that the algorithm was
unknown until Shaffer revealed it or the fact that neither
of the independent experts who evaluated the source
code recognized its existence to be indicative of any
broader shortcoming in the firmware. Two reasons
support this result. First, in “black box” testing, the
machine performed accurately by demonstrating the
ability to identify the percentage of alcohol in known
solutions within the applicable tolerance parameters.
Were there a fundamental defect in the source code, one
would expect that the machine would not be able to
perform in this fashion. Second, the evidence in the
record demonstrates that the EC reading is not always
less than the IR reading either during control tests or
in actual testing. If, as defendants fear, the EC is always
being artificially inflated to approximate, if not
92a
Appendix A
absolutely match, the IR, one would expect to find only
results in which the IR was the higher reading. That,
however, is simply not the case, as there are numerous
examples of readings from both actual and control tests
in which the EC reading is higher than the IR. We cannot
therefore conclude that the source code includes hidden
commands to artificially inflate the EC to raise it to the
level of the IR.
Finally, however, defendants argue that the
existence of the EC algorithm calls into question all of
the testimony received during the original remand
proceedings. They point out that several witnesses
referred to the fact that the Alcotest uses two
independent testing methods as proof of its superiority
and as support for their opinions that the device is
scientifically reliable and accurate. They further point
to Draeger’s representations to the State that this
technology made the device superior to others which
was essentially accepted by the Special Master.
Although the use of this algorithm certainly undercuts
the accuracy of the marketing claims made by Draeger,
it does not, in and of itself, alter the support in the
record for the conclusion by the Special Master about
the general scientific reliability of the device.
B. Weighted Averaging Algorithm
During the supplemental remand proceedings,
source code analysis revealed the use of a calculation
referred to as the weighted averaging algorithm. In
short, this algorithm relates to the manner in which the
93a
Appendix A
IR result is calculated. This technology measures the
effect of breath alcohol on an infrared signal. In order
to calculate the result, the device is programmed to
calculate a reading every quarter of a second, based on
measurements taken every 8.192 milliseconds. The
reported IR result is then computed by means of the
algorithm, which places proportionately greater weight
on the later measurements than on the earlier ones. In
operation, the algorithm directs that the first two
readings are averaged, and that value is averaged
further with each successive reading. The effect is that
the measurement is calculated to place greater and
greater weight on the readings taken as the sample of
breath continues.
Defendants attack the use of this methodology as
scientifically unsound. They point out, correctly, that it
is neither an average nor technically even a weighted
average. They further assert, however, that the use of
this algorithm is evidence of a scientifically unsound
device operated by inherently flawed software. We do
not find merit in these arguments. To be sure, the
calculation is not an average in accordance with the strict
mathematical definition. It is, however, in a more general
sense, a calculation designed to accord greater weight
to that part of the breath sample that enters the cuvette
at the end. In doing so, it gives greater weight to the
breath that, inevitably, includes the deepest air drawn
from the lung. It therefore focuses the analysis on the
portion of the breath sample that most accurately
represents the subject’s BAC. In this manner, the
weighted averaging algorithm seeks to achieve a more
94a
Appendix A
accurate result. We discern nothing in defendants’
attacks on this weighted averaging algorithm that
persuades us that it is inherently flawed or that it leads
to an inaccurate measurement of BAC.
C. Buffer Overflow Error
During the proceedings on remand, Draeger’s
expert, Bruce Geller, identified a significant flaw in the
program’s source code that, in limited circumstances,
can lead to an inaccurate reported BAC test result.
Following Geller’s testimony, Draeger’s programmer,
Shaffer, disputed many of the conclusions proffered by
defendants’ experts, but he acknowledged and
explained the buffer overflow defect, admitting that he
was responsible for the inclusion of this error in the code.
The buffer overflow error is only relevant when a
test subject, based on the IR and EC results of the first
two breath samples that fall outside of the accepted
tolerance, is given a third test.* Whenever that occurs,
there are six results (an IR and EC value for each test)
that must be evaluated. According to Shaffer, an array
of temporary variables is declared in order to calculate
32. Although the frequency with which this error occurs
has not been quantified with precision, the experience in
Pennsauken suggests that, absent the State’s adoption of an
expanded tolerance level between the two initial sets of results
in NJ Firmware version 3.11, and its resultant diminution of
third test results in Middlesex, one might estimate that as many
as five percent of all test subjects would ordinarily have results
that would require a third breath sample.
95a
Appendix A
the blood alcohol level from the six readings. The
available Alcotest array, as currently programmed,
however, is only large enough for four readings, and
therefore does not hold the second and third EC values.
For purposes of this calculation, the third EC value is
stored, accurately, in a previous memory location, but
the second EC value is altered because of the buffer
overflow error. In a situation in which there are six
readings, if the second EC result is the lowest value,
the device will effectively overlook it and the calculated
BAC level will be incorrectly reported instead as the
next lowest of the six readings.
Shaffer testified that although the buffer overflow
error must be corrected,* the previously recorded AIRs
correctly display the values for each of the six readings.
According to Shaffer, the only error on the AIR will be
its report of the BAC. Whether the buffer overflow error
affected the reported result, however, is not immediately
obvious from looking at the AIR. Instead, a set of
calculations, referred to by all of the parties as the
Shaffer formula, must be employed to determine
whether the buffer overflow error occurred, and, if so,
what the proper BAC should have been.
The Special Master, finding Shaffer’s candor to be
impressive and his testimony “completely reliable,”
concluded that the buffer overflow error is a “real” one
that must be corrected. Pending any corrective action,
33. He explained that he has not done so because of the
pendency of this litigation.
96a
Appendix A
he recommended that the use of all AIRs that report
three breath samples either be prohibited as a basis for
prosecution or, in the alternative, that Shaffer’s
corrective formula be applied.
Defendants argue that the Special Master’s solution
is inadequate because of the effect of the widened
tolerance which led to fewer third tests, but they do not
otherwise suggest that either of his proposed
alternatives is inappropriate for AIRs reporting third
test resulis. The State, although conceding that the
error is one that must be corrected, argues that there
is no basis on which to discard previously reported
results in light of the ability of the courts to apply the
corrective formula to the reported results.
There is no doubt in the record that the Firmware
version 3.11 source code includes an error, which may
cause the BAC to be incorrectly reported in cases when
a third breath sample has been taken.* The record,
however, makes clear that the error does not in any way
alter the accuracy of the reported results for each test
34. We recognize, of course, the force of defendants’
argument that the severity of the impact of this error has been
masked by the State’s unilateral decision to double the tolerance
and therefore to reduce the circumstances in which a third test
would be permitted. Our decision to address the extant third
test result cases does not in any way, in our view, alter the
separate manner in which we have elected to deal with the
problem presented to us by the increased tolerance range. We
instead have addressed that aspect of the record separately, see
Section VIII.D., supra.
97a
Appendix A
of each breath sample, but instead lies in the manner in
which the device reads and evaluates that data to
calculate the lowest BAC, which then may be both
inaccurately calculated and reported.
Were we without confidence in the accuracy of the
individually reported results, we would be constrained
to agree with the Special Master’s suggestion that we
reject all of the tests in which a third sample was taken.
However, in light of the fact that there is no evidence in
the record on which to conclude that the six readings
will themselves be inaccurate, we find no ground on which
to order a resolution so drastic.
Instead, we conclude that each AIR that includes
three breath tests will be admissible as evidence of an
accurate BAC reading only after application of the
Shaffer formula® to ensure the correct calculation of the
lowest possible result and reading. We do so, however,
with two added cautions heretofore unspoken. First, a
third sample is taken only when the four readings from
the first two samples are outside of the accepted range
of tolerance. As a result, where there are three breath
samples, the device does not simply identify and report
the lowest of the six reported readings. Instead, it must
35. Although referred to as the Shaffer formula, the
mechanism for the correct determination of whether a buffer
overflow error has occurred and, if so, the calculation of the
correct BAC is embodied in a worksheet that we have revised
to apply the correct tolerance range and have appended as
Worksheet B to the Order that accompanies this opinion for use
in all proceedings pending revision of Firmware version 3.11.
98a
Appendix A
first evaluate the six readings to determine which of the
samples fall within the accepted tolerance and then
determine, through truncation, which is the lowest
acceptable reported result. Calculating the correct
result in the face of the buffer overflow error is therefore
not a matter of visually inspecting the reported results
and selecting the lowest of them. Rather, the use of the
formula is required to ensure that the apparently lowest
result is also the lowest acceptable one in accordance
with the tolerance range.** Second, we note that in
devising the formula, Shaffer continued to utilize the
tolerance calculation reflecting the doubled range.
Because we have rejected that range as unacceptable,
we have revised the formula, in the form of a worksheet,
and have appended it to the Order that accompanies
this opinion in its corrected format.
The use of this methodology, however, will require
that, pending appropriate correction to the firmware,
each AIR with three test sample results must be
separately reviewed and that calculations must be
performed and verified for accuracy in accordance with
6. As an example, if the results on test one were IR = 0.030
and EC = 0.031 and the results on test two were IR = 0.085 and
EC = 0.088, and the results on the third test were IR = 0.091
aud EC = 0.092, the latter tests are within tolerance of one
another, but neither of the latter tests is in tolerance with the
first. The correct BAC result, therefore, would be 0.08 even
though the results of the first test were far lower. We use this
example by way of explanation and only to highlight the need
for applying the formula.
99a
Appendix A
Shaffer’s formula.*’ We therefore direct that the State
undertake to review all such AIRs, perform the
calculations to identify the correct BAC in accordance
with the Shaffer formula as we have adopted it, and
provide that data to the court in which each matter is
pending. We further direct that the calculations be made
a part of the evidence in any prosecution, pending
correction of the firmware, to facilitate appellate review.
D. Catastrophic Error Detection
Following the remand for source code analysis, the
Special Master also recommended that the machine’s
catastrophic error detection device be re-enabled. He
based his recommendation on his findings that the
Alcotest’s ability to detect catastrophic errors, which
was included in the original source code, had been
disabled from use in Firmware version 3.11 and that, if
utilized, it would ensure that the device would shut down
if it encountered such an error. Although defendants
agree with the recommendation that this device be
enabled in future software updates, they argue that the
implications of the unilateral decision of the
manufacturer to disable this feature and the use of the
Alcotest without this error detection capability must
undermine any confidence in any of the results reported.
The State, although disagreeing with both the
significance of the decision to disable this detection
37. The calculations based on the formula, which is set forth
in atable, appended to the Order that accompanies this opinion,
should be included as part of the record to facilitate further
review.
100a
Appendix A
device and with the impact it might have had on any
readings by the machine, agrees that the firmware will
be revised to re-enable catastrophic error detection.
Our review of the record demonstrates that there
is ample support for the findings and recommendations
of the Special Master concerning this aspect of the
source code. The witnesses were in general agreement
that the absence of an operational catastrophic error
detection device is not optimal, and they candidly
conceded that in the interim, and based on these
proceedings, the feature has been re-enabled for use in
other jurisdictions. Notwithstanding that general
agreement, the experts disagreed about how the
machine would respond if it encountered a catastrophic
error.
Defendants’ expert suggested that the machine
might under those circumstances create an inaccurate
AIR, although he could not explain, even theoretically,
how it would do so. Apart from that rather speculative
opinion, the experts agreed that the machine would
most likely enter an endless loop of non-productive
analysis and become unresponsive. Because there is no
credible evidence in this record that an Alcotest machine
that encounters a catastrophic error would create, in
reaction thereto, an incorrect AIR, we discern no basis
on which to conclude that any of the previously-
generated AIRs might represent a test in which the
machine encountered an error of this magnitude and
reacted by recording an inaccurate series of test results.
10la
Appendix A
Rather, we direct that the State arrange to have the
software corrected to re-enable the catastrophic error
detection feature.
E. Overall Firmware Reliability
As part of the analysis during the supplemental
remand proceedings, defendants’ expert opined that his
evaluation of the source code revealed thousands of
programming errors. He criticized the source code on
multiple levels, arguing that the style utilized is
outdated, that the reliance on global variables leaves
too much room for executional errors, and that the
program lacks adherence to any recognizable design
criteria. In short, he opined that there are so many, and
so great a variety of shortcomings in the source code
and the programming methodology that we should
conclude it is too flawed to be relied upon to generate
accurate test results.
The State and Draeger disagree. They assert that
most of the programming flaws that defendants’ expert
identified are simply stylistic programming preferences
and that they do not, in fact, represent errors in theory
or in reality. They urge us not to be misled into
concluding that the source code is inadequate for
purposes of scientific reliability.
Our consideration of this matter need not be
extended. In actuality, few aspects of the firmware
required our analysis and fewer still require our
intervention. Of the four major issues—the EC fuel cell
102a
Appendix A
drift algorithm, the buffer overflow error, the disabled
catastrophic error detention device, and the weighted
averaging algorithm — only the buffer overflow error
is capable of producing an erroneous AIR. Two of the
challenged features, the EC fuel cell drift algorithm and
weighted averaging algorithm, we have concluded,
contrary to defendants’ assertions, are scientifically
sound. The last of these, the catastrophic error detection
device, we have concluded should not have been disabled
but its absence was incapable of producing an inaccurate
AIR.
Our evaluation of the exhaustive record relating to
the source code leaves us confident that its errors have
been revealed. Based on that record, we do not share
defendants’ larger concerns that it is likely to generate
inaccurate results simply because, from a source code
writer’s viewpoint, it is complex or prolix. There being
no evidence in the record that these asserted
shortcomings are anything more than stylistic,
theoretical challenges, we decline defendants’ invitation
to require that the firmware comply with any specific
programming standards as unnecessary at this time.
X. Additional Firmware Recommendations
We next turn to a variety of issues arising from the
Special Master’s recommendations that require our
attention. Some of these matters are not in significant
dispute, but our consideration of each of them is essential
to ensuring that the Alcotest remains in compliance with
our directives. Most of these matters relate to the
103a
Appendix A
recommendations of the Special Master concerning the
future revisions to the firmware, but some we
independently deem to be necessary based on our review
and analysis of this record. None, however, requires
significant analysis or detail.
The Special Master included a recommendation
(Special Master’s Finding 2(c)) that the firmware be
locked so that only the manufacturer or the coordinators
would be able to make changes to the firmware.
Although defendants and the State agreed with this
recommendation, the NJSBA argued that it would be
more appropriate if only the manufacturer had the
ability to make changes to the firmware. In light of
defendants’ continued criticism of the programming
style and lack of rigorous programming standards used
in the source code, we are firmly convinced that the pool
of individuals who are able to make alterations to the
firmware should be reduced rather than expanded. Our
concern for uniformity in the firmware compels us to
direct that the firmware be locked so that only the
manufacturer will be able to make changes to it, which
changes may then be downloaded by the coordinators.
Further, considering the numerous changes that we
have directed be incorporated into the Alcotest in order
to ensure that the device is scientifically reliable and as
a prerequisite for admissibility of its BAC readings in
the future, we anticipate that our courts will encounter
AIRs from devices that utilize different versions of the
firmware. We therefore direct, for ease of analysis, that
the device be programmed so that on all future AIR
104a
Appendix A
printouts, the firmware version then being utilized by
the device is reported.
The Special Master also recommended that the
State should be required to publish future firmware
revisions and that notice of all such future revisions
should be given to the public in general and to the
amicus NJSBA in particular. Our analysis of this record
demonstrates, however, that this recommendation may
be insufficient. In our view, merely requiring that the
parties or the NJSBA receive notice of future revisions
will not be sufficient to ensure that the device is not
reprogrammed in a manner inconsistent with producing
accurate and reliable results that will be admissible in
DWI prosecutions. We therefore have concluded that
this required notice, to the parties, the public and the
amicus NJSBA, of the future firmware revisions must
be sufficiently specific to identify the proposed changes
in a manner that affords notice in compliance with due
process. A generic notice to the effect that the firmware
has been revised, in light of some of the previous
alterations that we today correct, will not suffice.*
The Special Master also recommended that the
State provide Alcotest training for defense attorneys
and their experts similar to that provided for operators
38. We note that the parties asked this Court to appoint an
independent software house to be responsible for any future
reviews of the Alcotest source code. We decline to do so at this
time, and will determine that issue should there be a challenge
in the future to the scientific reliability of the Alcotest based on
future firmware revisions.
105a
Appendix A
and coordinators. The State, understandably, objected
to this recommendation and urges us to reject it.
Although we reject it in part, defense attorneys should
not be left without any means of learning about the
device or its operation. Rather, we deem it to be in the
interests of justice that some form of training be made
available to defense attorneys to enable them to better
prepare to represent their clients. However, we agree
that the State should not be burdened with this
responsibility. We therefore direct that Draeger make
Alcotest training, substantially similar to that provided
to Alcotest operators and coordinators, available to
licensed New Jersey attorneys and their designated
experts. The training shall be offered at regular
intervals and at locations within the State of New Jersey,
at a reasonable cost to those who attend.
Xl. Requirements Prior to the Admissibility
of Alcotest Evidence
Our analysis of the general scientific reliability of
the Alcotest is grounded, in part, on our expectation
that there will be proof that the particular device that
has generated an AIR being offered into evidence was
in good working order and that the operator of the
device was appropriately qualified to administer the test.
This requirement that the test results be supported by
foundational proofs for admissibility has been part of
our jurisprudence since we decided Romano. There we
demanded that, as a precondition for admissibility of
the results of a breathalyzer, the Stat~ was required to
establish that: (1) the device was in working order and
106a
Appendix A
had been inspected according to procedure; (2) the
operator was certified;
This text is long and has been trimmed here. Open the source document for the complete record.
This is a copy of a public record, reproduced as it was published. It is not legal advice, and it may not be the version a court would rely on. Check the official source before you cite it.