U.S. v. HEBSHIE

tarted on the first floor, then the stairs down to the basement would not show the signs of fire damage that they did.

"Let's draw a plane meaning a plane that's floor level. Anything below that is in the basement, and we don't expect the heat from the fire to go down in the basement. The ladder is charred. . . . [T]hat heat can only come from the basement." Id. at 115:3-11.

Domingos reiterated his view that he did not bother to take photographs in the basement because there were no signs of fire there. Id. at 380:2-381:2. He conceded that he never mentioned his observations of the basement in any report and that the basement stairwell was absent from the government's floor plan of the store introduced at trial.

When Domingos was asked about the stairwell's location at the evidentiary hearing, he placed it in the wrong area. Id. at 408:4-17. It was behind the partition in the back office area.Id. at 522:6-14.

Agent Marquardt, a long-time ATF agent and fire investigator, testified for the government, discounting the basement hypothesis based on Myers' (the first-in firefighter) observations. If the fire had begun in the basement, it would have continued to burn because Myers did not extinguish any flames in the basement. Id. at 464:13-15. Marquardt also testified that if there had been fire in the basement, "heavy heat and smoke" would have been "pouring up through that stairwell into that back area," yet Myers did not report seeing any fire back there. Id. at 465:4-7. Further, the "floor should have collapsed into the basement." Id. at 466:15-17.

(2) Building Construction and Burn Patterns

Lentini, the defense expert, testified that, in fact, the left-hand wall was not the area of greatest damage, precisely because the walls in this area were still intact. Id. at 50:21-24. Indeed, he asserted that the government experts' theory of the fire's movement was scientifically impossible. If the fire entered the pipe chase, it would go up, where it had a clear space not five feet to the left through the wall's vertical studs, as Domingos suggested. Id. at 36:-3-13.

Lentini also testified that Domingos' theory and testimony with respect to the movement of the fire was based on a faulty understanding of fire dynamics. Id. at 90:2-9. First, the wall paneling next to the supposed point of origin was quarter-inch plywood paneling, which burns extremely easily. See Gov't Trial Ex. 1-14; Evid. Hr'g Ex. 4, 5. If the fire had started there, the paneling would have been burned all the way through. Instead, it was only partially burned. Evid. Hr'g. Tr. 36:23-37:10. Second, the government experts concluded that the fire had started on the first floor and then burned upward, based on the "V-pattern." According to Lentini, there was not much of a "V-pattern;" the fire did not interact at all with a shelf just above the alleged point of origin. See Gov't Trial Ex. 1-14; Evid. Hr'g Ex. 4, 5. Typically, a fire would burn upward and then spread out when it hit the shelf; photographs suggested that that did not occur. Evid. Hr'g Tr. 37:10-15. The government's theory would have required the fire to burn up near the alleged point of origin, bypass the shelf, then fall down again near the pipe chase. According to Lentini, it was far more likely that the fire started in the basement, came up through the wall, and then broke out of the wall into the store, a conclusion he contended was reflected in the V-pattern where the wall had burned. Id. at 38:19-39:11. Gov't Trial Ex. 1-10; Evid. Hr'g Ex. 4.

Domingos stated that the "heat source had come as a result of heat banking from this type of chase banking in behind the tin wall radiating and igniting . . ." Trial Tr. vol. 3, 118:5-7.

Titus testified that rebuttal witness Miller's conclusion that the grate would have turned "bright red" if exposed to fire was also scientifically incorrect. Evid. Hr'g Tr. 276:17-277:6.

(3) Thermal Imaging

Lentini disagreed with Myers' interpretation of the thermal imaging "hot spots." If the fire began along the left-hand wall of the store, the thermal image camera would show a "bright spot where the flame was," then the image would darken as it moved farther from that point. Evid. Hr'g. Tr. 47:2-7. Instead, there were white spots all around. The only way for all four walls to show hot spots would be if the entire room were on fire, an observation consistent with a basement fire traveling through the walls. Id. at 46:5-47:7; see also id. at 289:9-12 (Titus testifying) ("[T]he only way for the fire to get into the walls and generate those hot spots is to be able to move up from a lower area where it has access to those channels between the wall studs.").

b. Arson Evidence

(1) Canine Evidence

To the Spinales, Titus indicated great concerns about the use of accelerant-detection canines in the scientific community, id. at 264:1-14. Titus directed their attention to NFPA 921. Id. at 259:25-260:7. NFPA 921 circumscribes the use of canines; they are meant simply to be tools to help investigators narrow the search area for ignitable liquids. NFPA 921 § 14.5.3.5 ("Canine ignitable liquid detection should be used in conjunction with, and not in place of, the other fire investigation . . . methods. . . ."); id. ("The proper objective of the use of canine/handler teams is to assist with the selection of samples that have a higher probability of laboratory confirmation than samples selected without the canine's assistance."). Titus gave the Spinales an article from the Journal of Forensic Sciences that highlighted the varying levels of reliability in accelerant detection depending on the substance in question and the canine handler. Evid. Hr'g. Tr. 259:1-7.; Evid. Hr'g Ex. 12 (Michael E. Kurtz et al., Effect of Background Interference on Accelerant Detection by Canines, 41 J. Forensic Sci. 868 (1996)). Titus told them that concerns arose from the fact that the handler has to interpret the dog's actions. Evid. Hr'g Tr. 262:5-266:23. And he especially underscored the prejudice of allowing the testimony to be described as "accelerant detecting." What investigators refer to as "accelerants" actually represent a wide range of common and frequently benign materials. In addition, such chemicals can be created by the breaking down of materials during a fire, such as decomposing carpet and other adhesives. "Unlike explosive- or drug-detecting dogs, these canines are trained to detect substances that are common to our everyday environment. . . . [M]erely detecting such quantities is of limited evidential value." NFPA 921 § 14.5.3.5. For these reasons, NFPA 921 requires not just laboratory corroboration, but also comparison samples.Id. And it emphasized concerns about the canine's reliability — false positives and false negatives. Id. at § 14.5.35; see also, Evid. Hr'g. Exh. 12.

NFPA 921 is promulgated by the Technical Committee of the National Fire Protection Association ("NFPA"), the largest fire protection organization in the world and is widely accepted as the standard guide in the field of fire investigation. (NFPA 921: Guide for Fire and Explosion Investigations (2001).) See Evid. Hr'g Tr. 371:6-12, 443:10-16; Fireman's Fund Ins. Co. v. Canon U.S.A., Inc., 394 F.3d 1054, 1058-59 (8th Cir. 2005); Hartford Ins. Co. v. Gen. Elec. Co., 526 F. Supp. 2d 250, 257 (D.R.I. 2007); Royal Ins. Co. of Am. v. Joseph Daniel Constr., Inc., 208 F. Supp. 2d 423, 426 (S.D.N.Y. 2002); Travelers Prop. Cas. Corp. v. Gen. Elec. Co, 150 F. Supp. 2d 360, 366 (D. Conn. 2001); Technical Working Corp. on Fire/Arson Scene Investigation, Dep't of Justice, Fire and Arson Scene Evidence: A Guide for Public Safety Personnel (2000) (claiming that NFPA 921 is "a benchmark for the training and expertise of everyone who purports to be an expert in the origin and cause determination of fire investigation." See Evid. Hr'g. Tr. 371:6-12). Investigators are directed that "no specific hypothesis can be reasonably formed" until all relevant evidence is collected and all hypotheses must be based on data that is "capable of being verified." NFPA 921 §§ 2.3.1-2.3.7.
There was some debate during the evidentiary hearing as to whether the Spinales' reviewed the 2001 or the 2004 NFPA. The 2001 version was introduced at trial; however, counsel do not dispute that for the purpose of the sections cited here, there is no material difference between the two versions.

The government's experts repeatedly referred to Billy as an "accelerant canine" or an "accelerant-detection dog."

1. DOMINGOS: "Trooper Lynch . . . is trained in accelerant canines for the Fire Marshal's office." Trial Tr. Vol. 3, 42:2-5.

2. LYNCH: There was an opening for an accelerant detection dog. . . . I ended up getting a Labrador Retriever as my accelerant detection dog. . . . Id. at 141:11-16.

3. GOVERNMENT: [D]escribe what kind of training goes through for an accelerant detection dog. Id. at 144:25-145:3; see also id. at 148:15 ("accelerant detection dog"); id. at 150:11 ("accelerant dog"); id. at 157:13-14 ("accelerant detection canine").

Lentini reiterated many of the same concerns. He stated that the Forensic Science Committee of the International Association of Arson Investigators were so bothered by the use of canine alerts at trial that it issued a white paper. The white paper stated that canines were "[a]n excellent tool for the selection of samples but not appropriate for presentation to juries just because of the potential prejudice. . . . People identify with the dogs." Evid. Hr'g Tr. 54:14-25. Lentini continued that the dog does not add any probative force to the laboratory findings, and is in fact often prejudicial. Id. at 56:2-5. "The laboratory is really the final arbiter of what is in that sample." Id.
Professor Moriarty also underscored reliability problems with canine evidence in the literature and the case. Evid. Hr'g Tr. 343:13-25.

Titus testified in the evidentiary hearing — and I credit the testimony — that his testimony had been prematurely cut off by government counsel and not rehabilitated by the defense on redirect. He had answered that he was prepared to accept "Billy is 100 percent accurate" but was not allowed to explain the concerns about proficiency testing and false positives. Evid. Hr'g Tr. 262:5-13, 263:9-16, 264:15-25, 265:3-23.

(2) Laboratory Analysis

According to petitioner's experts, the lack of a comparison sample did not just call into question the investigator's conclusions; it fundamentally undermined its validity, a "critical error," according to Lentini. Evid. Hr'g Tr. 64:22-24. He pointed out that the convenience store had items that would also have tested as "light petroleum distillate." He listed building materials and glue as an example, but noted particular concern that a can of Zippo or Ronson lighter fluid might have been knocked over by the high powered hoses and spread throughout the store with the water. Id. at 48:7-16. Lentini also pointed out that NFPA 921 emphasized the need for comparison samples in such a case as the one at bar. NFPA 921 § 14.5.3.4 provides:

The collection of comparison samples is especially important in the collection of materials that are believed to contain liquid or solid accelerants. For example, the comparison sample for physical evidence consisting of a piece of carpeting believed to contain a liquid accelerant would be a piece of the same carpeting that does not contain any of the liquid accelerant. Comparison samples allow the laboratory to evaluate the possible contributions of volatile pyrolysis products to the analysis and also to estimate the flammability properties of the normal fuel present.

Lentini noted that there was carpet and flooring away from the alleged area of origin that could have been collected. Evid. Hr'g Tr. 45:19-21. In fact, he also pointed out that the carpet in the alleged area of origin was "in very good shape for a carpet that supposedly had ignitable liquid put on it." Id. at 40:13-16. Titus testified that he recounted these concerns about the validity of the laboratory test to the Spinales. In effect, without a comparison sample, the test told investigators very little.

"[T]hese chemicals were part of the environment. They were part of a newsstand, cigarette lighters, any one of a thousand or any several of a thousand several consumer items that contain hydrocarbons." Evid. Hr'g Tr. 262:21-24.

Moreover, Titus testified that the government's failure to further identify the chemical in the sample was problematic. "[T]he term `light petroleum distillate' is so generic and reflects so many different kinds of chemicals that either are parts of other products or parts of decomposition products in a fire that it has no meaning with regard to incendiary intent. . . ." Id. at 267:8-13.

III. ANALYSIS

To succeed on his ineffective assistance of counsel claim, Hebshie must show that his counsel's performance was deficient and that he suffered prejudice as a result. Strickland, 466 U.S. at 698.

A. Defense Counsel's Performance Was Constitutionally Deficient

To satisfy the "deficient performance" prong, Hebshie must show that his trial counsel's representation "fell below an objective standard of reasonableness." Id. at 687-88. The reasonableness of counsel's actions is considered in light of "prevailing professional norms." Id. at 688. There is a "strong presumption" that counsel's performance was reasonable. Indeed, courts are to apply "a heavy measure of deference to counsel's judgments." Id. at 691.

The case law describes two lines of cases. In one line, the record may show counsel's entire performance fell below the constitutional minimum. See, e.g., Williams v. Taylor, 529 U.S. 362 (2000) (finding counsel's performance ineffective where he failed to present substantial mitigation evidence to sentencing jury). In the other, the record may indicate that counsel, for the most part, provided adequate performance, yet he or she committed a single, critical error that renders the representation ineffective. Kimmelman v. Morrison, 477 U.S. 365, 386 (1986) (holding that counsel's "total failure to conduct pre-trial discovery" constituted ineffective assistance); see also Murray v. Carrier, 477 U.S. 478, 496 (1986) ("[T]he right to effective assistance of counsel . . . may in a particular case be violated by even an isolated error of counsel if that error is sufficiently egregious and prejudicial."); Prou v. United States, 199 F.3d 37, 47, 48 (1st Cir. 1999) (finding failure to object to untimely filing that resulted in unlawful mandatory minimum sentence was ineffective assistance); Mason v. Hanks, 97 F.3d 887, 902 (7th Cir. 1996) (finding that failure to raise issue of inadmissible hearsay constituted deficient performance).

This case fits more closely in the latter category. There was not a complete failure of defense (although candidly, counsel's performance came close). Defense counsel consulted with and presented an expert and cross-examined witnesses about some of the inconsistencies in their case. Cf. Dugas v. Coplan, 428 F.3d 317, 342 (1st Cir. 2005) (remanding on ineffective assistance of counsel where counsel did not present an arson expert on defense.) What counsel did not do is to move for a Daubert hearing prior to trial on any expert issue. They did not seek exclusion of any of the proposed expert testimony which was the core of the arson case, or move for its limitation. They did not argue that the expert testimony failed to meet the minimal threshold for reliability of scientific evidence in NFPA 921 and should not have been admitted at all. They did not alert the Court to the ways in which the government's investigation undermined their very ability to present a defense.

And their trial performance did not remotely make up for the inadequacies in their pretrial preparation, especially with respect to Lynch and Drugan, the arson evidence witnesses. They did not object to Lynch's testimony that went far beyond the proper use of canine evidence. They did not object to the government's improper argument about Billy's "failure" to alert anywhere else in the store, except for where Domingos concluded the fire started. They did not object when the government used Billy's failure to alert as substantive evidence of Hebshie's guilt.

I address the Daubert issues generally with respect to all of the expert evidence. I then turn to the unique problems with the arson evidence — the canine and the laboratory analyses. I then address Domingos' and Miller's testimony.

1. Counsel Erred in Failing to Move for a Daubert Hearing on any Expert Testimony

Scientific evidence, as recognized by Daubert, 509 U.S. 579, and its progeny, requires a different kind of trial preparation for both the court and counsel. Scientific evidence must be "not only relevant, but reliable," that is, "derived by the scientific method" and "supported by appropriate validation." Id. at 589-90. The court must conduct "a preliminary assessment of whether the reasoning or methodology underlying the testimony is scientifically valid and of whether that reasoning or methodology properly can be applied to the facts in issue." Id. at 592-93.

Daubert's extra vigilance is essential for two reasons: First, while opinion testimony is generally excluded, expert witnesses are permitted to opine on the basis of evidence not given to the jury. See Fed.R.Evid. 703. Second, as I have previously written, "a certain patina attaches to an expert's testimony unlike any other witness; this is `science,' a professional's judgment, the jury may think, and give more credence to the testimony than it may deserve." United States v. Hines, 55 F. Supp. 2d 62, 64 (D. Mass. 1999); see also Michigan Millers Mut. Ins. Corp. v. Benfield, 140 F.3d 915, 920 (11th Cir. 1998) ("The use of `science' to explain how something occurred has the potential to carry great weight with a jury, explaining both why counsel might seek to couch an expert witness's testimony in terms of science, as well as why the trial judge plays an important role as the gate-keeper in monitoring the evidentiary reliability of such testimony.").

When scientific evidence does not meet the minimal requirements, advocacy tools that may be effective as to other evidence are not effective here. Just because the testimony has been admitted sends the jury the message that the opinions are entitled to some weight. See N.J. Schweitzer Michael J. Saks,The Gatekeeper Effect, 15 Psychol. Pub. Pol'y L. 1, 12 (2009) ("[J]urors assume that judges review scientific evidence before it is presented to them, and that any evidence used in a trial must be above some threshold of quality. Because of these assumptions, jurors seem to be less critical of scientific evidence used in trials and are more persuaded by it."). Cross-examination suffices only when experts have reached different conclusions, but the underlying approach is sound. Where it is not, exclusion, or in some situations, limitation, is the only option. See United States v. Green, 405 F. Supp. 2d 104, 109 (D. Mass. 2005) (limiting the ballistics expert to describing the similarities between the shell casings, but not the ultimate conclusion that they came from a specific pistol "to the exclusion of every other firearm in the world").

Despite ample reasons for defense counsel to be on notice of serious problems with the government's expert evidence — from Titus, from Muse, arguably from their own investigation — they did not request a Daubert hearing as to anything. They knew that there were problems in the Domingos cause-and-origin investigation and the Drugan laboratory analysis that undermined their validity; they knew that the failure to take a control sample in this case was inconsistent with the scientific method and NFPA 921 § 14.5.3.4 (governing collection of comparison samples); they knew that the investigation of the basement was inadequate, or at least, not fully documented; and they knew or should have known that the canine evidence was supposed to be admitted for only a limited purpose, namely, assisting in the selection of samples that have a "higher probability of laboratory confirmation than samples selected without the canine's assistance," and that testimony beyond those purposes was potentially prejudicial, Evid. Hr'g Tr. 255:1-266:5.

Further, if Jay Spinale was as familiar with NFPA 921 as he testified he was at the evidentiary hearing, Evid. Hr'g Tr. 177:13-16, he should have recognized serious deficiencies in the documentation.

Contesting the reliability of this evidence in a Daubert hearing was plainly within the "prevailing professional norms." By 2006, when Hebshie was tried, the public and professional literature reflected increasing scrutiny of arson evidence by experts in both the scientific and legal fields as well as by the public at large. While counsel need not read every article pertaining to the subject at hand, he must "keep[] abreast of current legal literature and developments." Model Code of Prof'l Responsibility EC 6-2 (1980); see also Strickland, 466 U.S. at 688 ("Prevailing norms of practice as reflected in American Bar Association standards . . . [are] "guides to determining what is reasonable.")

Indeed, in the early 2000s, there were two cases in this Circuit that raised the question of defense counsel's effectiveness in an arson case. See Dugas v. Coplan, 428 F.3d 317, 342 (1st Cir. 2005) (remanding for a hearing on the question of whether counsel was ineffective in not obtaining an arson expert); United States v. Correia, 77 F. App'x. 12 (1st Cir. 2003), aff'g No. 00-10246-RWZ, 2002 WL 31052766 (D. Mass. Sept. 13, 2002) (affirming the trial court's decision to grant a new trial based on counsel's ineffectiveness in not obtaining a defense arson expert, and failing to pursue other defense avenues).

Popular legal casebooks, such as David L. Faigman et al.,Science in the Law: Forensic Science Issues (2002), highlighted problems in many cause-and-origin techniques. In addition, in the early 2000s, several reports and articles were published that impugned arson investigatory techniques as well as forensic evidence more generally. Notably, in 2005 Congress passed legislation calling for a review of forensic evidence. See Science, State, Justice, Commerce, and Related Agencies Appropriations Act of 2006, Pub.L. No. 109-108, 119 Stat. 2290 (2005). Even the popular press reported on the concerns raised by forensic evidence. The Chicago Tribune ran a series of such articles in 2004, "Forensics Under the Microscope." This series included an exposé on arson evidence. Def.'s Mem. Supp. 2255 Pet. Mot. New Trial Ex. D (Maurice Possley, Arson Myths Fuel Errors, Chi. Trib., Oct. 18, 2004) (document #137-5).

This legislation resulted in the 2009 National Academy of Sciences Report, "Strengthening Forensic Science in the United States: A Path Forward," that questioned the reliability of forensic evidence used in courts across the country.

Just a few months before Hebshie's trial began, the Innocence Project Arson Review Committee released its report excoriating the suspect "science" used to convict Cameron Todd Willingham and Ernest Ray Willis of arson and murder. Arson Review Committee, Innocence Proj Report on the Peer Review of the Expert Testimony in the Cases of State of Texas v. Cameron Todd Willingham and State of Texas v. Ernest Ray Willis (2006) (Mem. in Support Ex. A (document #137-1)).

Finally, case law raised concerns about arson expert testimony similar to the concerns raised in this investigation. See, e.g.,Fireman's Fund Ins. Co. v. Canon U.S.A., Inc., 394 F.3d 1054, 1058 (8th Cir. 2005) (holding district court's exclusion of expert arson evidence proper where experts failed to compare hypothesis to evidence from scene in violation of NFPA 921);Mich. Millers, 140 F.3d at 920 (finding no abuse of discretion where trial court excluded arson expert's testimony because his methodology did not support his conclusion); Ind. Ins. Co. v. Gen. Elec. Co., 326 F. Supp. 2d 844, 850-51 (N.D. Ohio 2004) (holding that cause-and-origin expert's failure to properly collect evidence, in violation of NFPA 921, Section 9-5.1, "Documenting the Collection of Physical Evidence," made his investigation unreliable.); Am. Family Ins. Grp. v. JVC Am. Corp., No. 00-27 DSD/JMM, 2001 WL 1618454, at *3-4 (D. Minn. Apr. 30, 2001) (excluding expert testimony where expert did not apply methodology recommended by NFPA).

While most of the above cases are civil, it cannot be that science is different in criminal cases than in civil ones. Bad science is bad science; unreliable methodologies are unreliable methodologies, no matter the side of the docket. See Paul C. Giannelli, The Supreme Court's "Criminal" Daubert Cases, 33 Seton Hall L. Rev. 1071, 1111 (2003) ("Daubert has evolved into a stringent standard in civil cases; D. Michael Risinger,Navigating Expert Reliability: Are Criminal Standards of Certainty Being Left on the Dock?, 64 Alb. L. Rev. 99, 149 (2000) (comparing "the heightened standards of dependability imposed on expertise proffered in civil cases" with "expertise proffered . . . in criminal cases [which] has been largely insulated from any change in pre-Daubert standards or approach."). Paradoxically, and perhaps shamefully, this standard has not been consistently imposed in criminal cases.").

There was nothing remotely "strategic" about counsel's failures to request a Daubert hearing on Domingos' testimony. John T. Spinale said so. Evid. Hr'g Tr. 168:8-15. If he believed he could have, he would have tried to exclude the expert testimony. Id. His explanations for not doing so could not be more disingenuous: He claimed that he had not spoken to prior counsel about aDaubert hearing; he just "assumed" that counsel had waived it or that it was too late. I do not credit that testimony. Predecessor counsel Muse was "certain" that he had discussed challenging the government's experts with the Spinales. Evid. Hr'g Tr. 218:24-219:4.

More significant, Spinale could not have reasonably believed that a Daubert hearing was waived when the Court asked him over and over again if he wanted such a hearing — even mid-trial. Indeed, the government's claim that Spinale did not move for a hearing because he feared the Court would be aggravated is absurd. This Court explicitly invited the challenge so that critical scientific issues could be hashed out outside of the presence of the jury.

2. Failure to Challenge Canine Evidence Was Error

As described below, Lynch's testimony about the canine evidence played a role in this case that far outstripped the limited purposes for which canine testimony may be admitted. There were errors of two sorts: First, counsel should have asked for aDaubert/Kumho Tire hearing at the outset. Second, during the trial, counsel plainly should have objected on both Daubert and Rule 403, Federal Rules of Evidence grounds to Lynch's truly extravagant testimony about Billy the dog, her powers, her accuracy, her role in accelerant-detection in general and in this case, and to the government's closing.

a. Failure to Move for Daubert Hearing

The government argues that there was no basis for error in counsel's failure to move for a Daubert hearing in connection with the canine testimony, on three grounds: First, this claim was not included in the defendant's habeas petition; second, canine testimony is not the appropriate subject of a Daubert hearing; and third, the failure to call for one was fully justified by the case law.

There is no question that Hebshie challenged counsel's failure to move for a Daubert hearing on the canine testimony. His habeas petition includes as grounds for his ineffective assistance claim that "[t]rial counsel waived Daubert hearing with respect to . . . use of accelerant-detecting canine" and that "[t]rial counsel failed to object to prejudicial and irrelevant testimony concerning use of canine to locate petroleum derivative." Mot. to Vacate 5 (document #136). Hebshie more than adequately addressed the Daubert issue in his memorandum supporting his habeas petition, Def.'s Mem. Supp. 2255 Pet. and Mot. New Trial 4, 22 (document #137); the evidentiary hearing, Evid. Hr'g Tr. passim; and the supplemental habeas memorandum; Def.'s Post-Hr'g Mem. Supp. 2255 Pet. 8, 10, 13-15, 25 (document #177).

Next, the government argues — improbably, to be sure — that canine testimony is not the appropriate subject of a Daubert hearing because "neither the canine nor its handler is offering expert opinion testimony," just "whether and where the canine" alerted. Gov't Post-Hr'g Mem. Supp. Opp'n Def.'s 2255 Mot. 10 (document #179). The argument is simply extraordinary given the testimony in this case. Lynch opined about the dog's superior olfactory capabilities, his unique ability to read the dog, the protocol for canine investigations, how he trains the dog, and even Billy's error rate. Rule 702 of the Federal Rules of Evidence encompasses all "scientific," "technical," and "other specialized" knowledge. Lynch had specialized training and knowledge to assist in his handling of the dog and his interpretation of the dog's reactions. His testimony rested "upon an experience confessedly foreign in kind to the jury's own."Kumho Tire, 526 U.S. at 149. Even if Lynch's testimony were not, strictly speaking, scientific testimony as Daubert suggests, it was surely expert, experiential testimony, which Kumho Tire addresses; either way a hearing was required.

In fact, one court, applying both the federal standard and an analogous state standard, excluded testimony of a canine handler on precisely the grounds argued here. In Farm Bureau Mut. Ins. Co. v. Foote, 14 S.W.3d 512, 518 (Ark. 2000), for example, the Supreme Court of Arkansas affirmed the trial court's exclusion of a canine handler who sought to testify about "the alleged superior ability of his canine partner, Benjamin, to detect the presence of accelerants after a fire," that he could discriminate between different types of chemicals," and that he had an accuracy rate of "100%." The court held:

[W]e conclude that the proffered testimony concerning the dog's alleged superior ability to detect the presence of accelerants does not pass muster using either the Daubert or Prater [v. State, 820 S.W.2d 429 (Ark. 1991)] analysis. Farm Bureau simply did not make any showing regarding the scientific validity of the evidence. For instance, [the canine handler] did not produce the study allegedly conducted by [the director of the Florida State Crime Laboratory, cited by the canine handler], so there was no way of ascertaining the techniques used or the potential rate of error.

Id. at 520.

Likewise, another court, responding to just the kind of motion that ought to have been filed in the case at bar, excluded the testimony of a dog handler. It did so because the dog's alert was not confirmed by laboratory analysis. And without such confirmation, the dog's alert was in effect being used as substantive evidence of arson. The court noted:

[T]he use of the dog alert as substantive evidence is beyond the accepted scope and application of the technique as described in the NFPA guide. See § 16.5.4.7 (describing the role of canine investigation as "assisting in the location and collection of samples for laboratory analysis"). Use of expert testimony beyond its proper application and scope is neither relevant nor helpful to the trier of fact.

United States v. Myers, No. 3:10-00039, 2010 WL 2723196, at *3 (S.D.W.Va. July 8, 2010). To be sure, this case presents different facts. Billy's "alert" was confirmed by the laboratory, albeit with the deficiencies described below (a single sample, a generic result, from a restricted area on the fire scene). But what makes Myers analogous was the way in which the government in this case used Billy's failure to alert anywhere else as substantive evidence of guilt.

Several courts have held it in the district court's discretion to admit canine evidence but that fact hardly ends the inquiry.United States v. Berrelleza, 90 F. App'x. 361, 365 (10th Cir. 2004) (finding that failure to subject canine evidence to Daubert hearing was not abuse of discretion); United States v. Paccione, Nos. 99-1211(L), 99-1212, 2000 WL 34251719, at *3 (2d Cir. Feb. 3, 2000) (holding the same); United States v. Marji, 158 F.3d 60, 63 (2d Cir. 1998) (finding no abuse of discretion where the district court admitted canine evidence corroborated by "substantial additional evidence"). Just because an issue involves the exercise of discretion does not mean it is automatically admitted. Discretion, after all, only means that the Court can admit, exclude, or limit, based on the facts of the case at bar.

Indeed, the purpose of a pretrial hearing is to help the Court determine just how to exercise its discretion. The hearing is not about whether a particular methodology is valid in general, but whether it was properly applied in this case, and appropriate to the task at hand. See Kumho Tire, 526 U.S. at 153-54. Without a Daubert/Kumho Tire hearing to test the metes and bounds of the testimony, the Court may have no idea of the context for the expert testimony — no idea what the "task at hand" is.

See D. Michael Risinger, Defining the "Task at Hand": Non-Science Forensic Science after Kumho Tire v. Carmichael, 57 Wash. Lee L. Rev. 767, 773 (2000).

Justice Breyer, in Kumho Tire discusses Daubert's general holding:

[T]he Federal Rules of Evidence "assign to the trial judge the task of ensuring that an expert's testimony both rests on a reliable foundation and is relevant to the task at hand["] . . . [and its discussion of] certain more specific factors, such as testing, peer review, error rates, and "acceptability" in the relevant scientific community, some or all of which might prove helpful in determining the reliability of a particular scientific "theory or technique."

526 U.S. at 141 (internal citations omitted).

The Court's concerns in Daubert and Kumho Tire apply with special force here: First, the appropriate use of canine evidence is merely to assist investigators in the selection of samples, not to replace any other forms of investigation. NFPA 921 § 14.5.3.5 ("Canine ignitable liquid detection should be used in conjunction with, and not in place of, the other fire investigation . . . methods. . . .").

Second, NFPA 921 contains a lengthy warning about the proper use of canine evidence and the significant possibility of canine error — both false positives and false negatives. "Research has shown that canines have responded or have been alerted to pyrolysis products that are not produced by an ignitable liquid and have not always responded when an ignitable liquid accelerant was known to be present." Id. at § 14.5.3.5 (italics supplied);see, also, Evid. Hr'g Ex. 12, Michael E. Kurz et al., Effect of Background Interference on Accelerant Detection by Canines, 41 J. Forensic Sci. 868 (1996). Notwithstanding that caution, the government argued that the jury should consider the dog's failure to alert to other areas as substantive evidence that the positive alert was inculpatory. And Lynch opined without objection that Billy was correct 97% of the time and when she was not, it was because of her handler's error. Counsel not only failed to challenge any of this, he could not recall even asking the government for reports on the testing of the dog at issue. Evid. Hr'g Tr. 141:21-25.

Third, NFPA 921 warns that canines are trained to alert to common benign substances. § 14.5.3.5. ("Unlike explosive- or drug-detecting dogs, these canines are trained to detect substances that are common to our everyday environment. . . . [M]erely detecting such quantities is of limited evidential value.").

Fourth, the firefighting operations had severely damaged the building and moved around material within the store. Even if Billy alerted on some materials, the material could be from somewhere else in the room, or the building. Moreover, Billy was deployed only in Hebshie's store and only in the cleared area on its left side, where Domingos believed the fire had started.

Fifth, Titus expressly warned about the use of the phrase "accelerant-detection" to describe what Billy did. Billy alerted to a broad class of common chemicals, some of which were not "accelerants" in the sense of substances brought into a scene for the purposes of arson. As NFPA § 14.5.3.5 notes:

Unlike explosive- or drug-detecting dogs, these canines are trained to detect substances that are common to our everyday environment. The techniques exist today for forensic laboratories to detect submicroliter quantities of ignitable liquids, but because these substances are intrinsic to our mechanized world, merely detecting such quantities is of limited evidential value.

Sixth, these concerns made Lynch's decision to take only one sample, no comparison or control samples, all the more problematic. True, NFPA 921 § 14.5.3.4 describes instances in which comparison samples may not be necessary for the "valid identification of ignitable liquid residue," but that was not the case here. The ambient environment of the convenience store included a range of common products, which could well have tested positive. The fire itself could have produced substances that would have tested positive. And, in any event, comparison samples were essential precisely to test the observations of the officers on the scene. "The determination of whether comparison samples are necessary is made by the laboratory analyst, but because it is usually impossible for an investigator to return to a scene to collect comparison samples, they should be collected at the time of the initial investigation." Id. (emphasis added).

Given everything that the Spinales knew about the canine alert in this case, it was constitutional error for them not to ask for a Daubert hearing. There was no "imaginable strategic or tactical reason for the omission." Prou, 199 F.3d at 48. While they indicated that they did not challenge the evidence or ask for a hearing because they thought the evidence was admissible, they were simply wrong.See Evid. Hr'g Tr. 136:1-22, 178:17-25. The testimony may be admissible — but only if certain conditions are met. They were not.

The subsequent testimony of Lynch — which went so far afield — made the failure to air these issues in advance in a Daubert hearing all the more troubling.

b. Failure to Object to Lynch's Testimony

The testimony of canine handler Lynch was a lengthy, almost mystical, account of the dog's powers. He went on about the dogs he had trained and his attachment to "his" dogs, including an unrelated dog, Amore, a narcotics-sniffing dog who died six years before the fire at issue. He touted the unique olfactory qualities of the dogs with whom he had worked. He spoke emotionally about his particular relationship with Billy, who had died before the trial, and his uncanny ability to communicate with her. (Lynch's testimony extended over 56 pages of direct examination.) His testimony was frequently entirely subjective — how strong the alert was, how he understood Billy's expressions, interpreted the sounds she made even in the dark. And he repeatedly vouched for the animal — that she was 97% accurate, that if Billy were wrong it was the handler's error not the dog's.

Billy was identified as an "accelerant-detection" dog, as if the substance she identified — "light petroleum distillate" — was necessarily associated with arson fires; it was not. Lynch was permitted to testify that the dog did not alert to anything else on the premises, Trial Tr. vol. 3, 177:9-15; Trial Tr. vol. 4, 17:22-25, 19:25-20:4, under circumstances when the scientific literature (described above) casts doubt on the significance of the dog's failure to alert, and the testimony suggested that Billy had been directed only to the left side of Hebshie's store. Although counsel cross examined Lynch, challenging the fact that only one sample was taken, and that Billy was limited to certain areas, and although he conceded that Lynch's testimony was "over the top," Evid. Hr'g. Tr. 148:8, he never objected. Lynch's testimony went far beyond any plausible relevance of the use of a dog to help select the locations for the collection of a sample to be tested.

Indeed, Lynch's testimony was not the first time that the jury heard about Billy's prowess. Even before Lynch testified, Domingos recounted what he had seen when the dog was brought into Hebshie's store, the kind of alert she gave, and what it meant, without objection.

And the theme was recited yet again in the government's closing. The government gave credence to the dog's evidence, far beyond what it deserved, arguing, "Billy knew what she was smelling for. She indicated an alert on one place and one place alone. It was a strong alert. If it was the carpet that was setting her off, then she would have triggered everywhere else around her." Trial Tr. vol. 6, 171:5-9 (italics supplied). The Spinales remained silent.

On the contrary. In John Spinale's closing, he noted how "impressive" Lynch and his dog were. Trial Tr. Vol. 6, 189:10.

In addition to the grounds described above (which would have provided a basis for a Daubert/Kumho Tire hearing) counsel should have objected to the testimony based on the following:

On the 97% accuracy: There are no peer reviewed standardized methods of training detective dogs; their reliability is in fact highly variable. See NFPA § 14.5.3.5 (concerning false positives and false negatives). Indeed, as of the time the case went to trial, the Journal of Forensic Science had reported the results of studies indicating a substantial error rate with canine alerts, far higher than Lynch was describing. Evid. Hr'g Tr. 343:7-12, see also, Evid Hr'g Ex. 12, Michael E. Kurz et al., Effect of Background Interference on Accelerant by Canines, 41 J. Forensic Sci. 868 (1996). There is no indication that the Spinales had any idea what the 97% was based upon. Lynch apparently just announced the figure ipse dixit.

See Illinois v. Caballes, 543 U.S. 405, 411-12 (Souter, J., dissenting) ("The infallible dog, however, is a creature of legal fiction. . . . [T]heir supposed infallibility is belied by judicial opinions describing well-trained animals sniffing and alerting with less than perfect accuracy, whether owing to errors by their handlers, the limitations of the dogs themselves, or even the pervasive contamination of currency by cocaine." (discussing drug-sniffing dogs)).

On Lynch's references to drug-detecting and explosive-detecting dogs that he had trained: As the above reference suggests, drug-detecting and explosive-detecting dogs are different from accelerant-detecting dogs, although Lynch's testimony put all such canine evidence in the same category. Drug-detecting and explosive-detecting dogs typically do not alert to substances common in the every day environment.

On the subjectivity of the testimony: Lynch went on and on about what he understood about Billy, as if his relationship with Billy somehow enhanced the reliability and probative value of the results — that she was unique, that he could "read her face," that he was with her 365 days a year, that he knew her personality, "the way her eyes shifted," the ways her ear shifted, etc.

On taking comparison samples: According to Lynch, his special relationship with his dog justified his failure to take comparison samples. His decision was based on the "type of alert" he had from the dog as well as whether Domingos requested that he take a sample. The need for a comparison sample, however, does not relate to the accuracy of the dog, but to the significance of the laboratory analysis to the investigation — the chemistry of pyrolysis, the nature of the environment.

By the end of the testimony, Lynch had gone far beyond the limited purpose for which canine evidence could lawfully be presented. Spinale should have objected under Rule 403 or Rule 702 of the Federal Rules of Evidence to all statements that suggested that the alert had any validity beyond helping the fire investigators to choose samples. And the Spinales' failure to object to Lynch's testimony was exacerbated by their failure to object to the government's closing and their own remarks to the jury.

Billy's alert was not dispositive of either where the fire began or that it was arson. She had not been permitted to go "everywhere" in the fire scene. She had been directed only to an area that was safe for her, and only to the left wall in Hebshie's store. And her olfactory skills and the "strength" of the alert, as understood by Lynch, hardly entitled the jury to conclude that the substance alerted to had a role in an arson, rather than a relatively common light petroleum distillate, especially without comparison samples. In effect, everything that the literature warned about in dealing with canine evidence happened here. Spinale's failure to object was not remotely "sound strategy." Kimmelman, 477 U.S. at 381.

3. Failure to Request a Daubert Hearing on the Laboratory Test and Drugan's Testimony Was Error

A reasonably competent lawyer surely should have moved for aDaubert hearing on the laboratory analysis. The laboratory test proving the presence of an accelerant was the most powerful evidence of intent to commit arson. Even if the jury credited Domingos' cause-and-origin testimony — that the fire began in Hebshie's store — they did not have to conclude that the fire was arson, particularly given the substances for sale in the store. Not only did John Spinale fail to request a Daubert hearing on the laboratory accelerant test, he stipulated to it. Trial Tr. vol. 3, 181:14-16.

First, a reasonably competent counsel would have challenged the test results of "light petroleum distillate" and their characterization as accelerants. Titus warned the Spinales as much, alerting them that "light petroleum distillate" is such a generic category and includes so many benign chemicals that it literally has no meaning in connection with a finding of "incendiary intent." Evid. Hr'g Tr. 267:8-13. Spinale's files contained laboratory analysis from K-Chem Laboratories, the experts Muse had retained, which disputed the significance of the finding.

Titus also discussed the need for a control sample with the Spinales. Evid. Hr'g Tr. 268:15-269:6.

In similar incidences finding petroleum products that are natural to the scene (sold there) is expected. These products heat up normally and the containers will either melt or burst allowing the liquid petroleum product to spill. Even if these products were stored in a different location, the action of the firefighters spraying hoses can move these petroleum products around allowing them to be found in a different location. This in turn can cause accelerant detecting K-9's to find false samples that will mislead investigators.

Evid. Hr'g Ex. 17, at 1-2.

Second, the laboratory results were especially problematic here without comparison samples. NFPA 921, which Jay Spinale claimed that he and his father had read and re-read, underscores the importance of collecting comparison samples when testing for accelerants. See NFPA 921 § 14.5.3.4 ("The collection of comparison samples is especially important in the collection of materials that are believed to contain liquid or solid accelerant."). As described above, the laboratory test was significant only if a substance were found in the fire scene that was not there before, not part of the ambient environment of Hebshie's convenience store, or not produced by the heat of the fire itself.

Spinale was on notice with regard to these problems; there was simply no downside to requesting a Daubert hearing. "Where, as here, an attorney fails to raise an important, obvious defense without any imaginable strategic or tactical reason for the omission, his performance falls below the standard of proficient representation that the Constitution demands." Prou, 199 F.3d at 48; see also Rompilla v. Beard, 545 U.S. 374, 394 (2005) (O'Connor, J., concurring) (finding that counsel's performance was objectively unreasonable where counsel failed to investigate an issue that was "at the very heart of the prosecution's case" and that "threatened to eviscerate one of the defense's primary . . . arguments").

4. Failure to Request Daubert Hearing on Cause-and-Origin Evidence Was Error

The government suggests that since there was no reason to doubt the reliability of the cause-and-origin testimony, the Spinales' failure to request a Daubert hearing could not have been deficient. See United States v. Davis, 406 F.3d 505, 509 (8th Cir. 2005) ("[C]ounsel cannot be said to be ineffective for failing to challenge the FBI's methodology on a basis that was not advanced by the scientific community at the time of trial."). The claim is extraordinary. There were surely concerns about which the Spinales were well aware — the failure to take control samples, to document the investigation of the basement, the interpretation of the thermal camera, inconsistent testimony about the building construction. And, as with the laboratory and canine evidence, there was no reason not to.

Here again, the Spinales' failure to ask for a Daubert hearing on Domingos' testimony, prior to trial, or even during the trial when the Court invited him to do so, was not remotely a "strategic choice[] made after thorough investigation of law and facts relevant to plausible options," Strickland, 466 U.S. at 690, as the government argues. It was not remotely the product of an "informed tactical decision," Rompilla 545 U.S. at 395 (O'Connor, J., concurring), but instead, of a lack of preparation, even thought. It was error, plain and simple.

One can envision how the failure to request a Daubert hearing could be a strategic choice. For example, counsel, fully prepared to cross-examine an expert, and less than certain that the challenges would lead to exclusion after a Daubert hearing, may prefer not to alert the government to the points he or she is likely to make on cross-examination. That kind of strategic calculus was not at all involved in the case at bar.

Nor is a lack of time a reasonable excuse. Counsel had a year and three months to prepare for trial.

B. Counsel's Deficient Performance Prejudiced Hebshie

To satisfy the "prejudice" prong, Hebshie must establish that but for his counsel's deficiency, there is a "reasonable probability" that the outcome would have been different. Peralta v. United States, 597 F.3d 74, 79 (1st Cir. 2010). He need not show "that counsel's deficient conduct more likely than not altered the outcome" of his trial, only "a probability sufficient to undermine confidence in [that] outcome." Strickland, 466 U.S. at 693-94.

With respect to the Rule 403 objections (to Lynch, the government's closing), Hebshie must show that the Court would have sustained the objections and that the exclusion of the testimony would likewise "undermine confidence" in the verdict. With respect to the Daubert issues, Hebshie must demonstrate a reasonable probability not only that the Court would have granted a hearing, but that the outcome would have been exclusion or limitation of the testimony, a result that would have "undermine[d] confidence" in the verdict.

The First Circuit permits courts to examine prejudice cumulatively. "`Strickland clearly allows the court to consider the cumulative effect of counsel's errors in determining whether a defendant was prejudiced.'" Dugas, 428 F.3d at 335 (quotingKubat v. Thieret, 867 F.2d 351, 370 (7th Cir. 1989)) ("We thus find that these three issues, taken together . . . are sufficient to raise a genuine dispute of material fact concerning prejudice."); see also Reiner v. United States, No. 09-118-P-H, 2010 WL 1490303, at *9 (D. Me. Apr. 13, 2010) (evaluating errors cumulatively).

1. Is There a Reasonable Probability the Court Would Have Granted a Daubert Hearing?

The answer is an unequivocally, yes, as to the cause-and-origin testimony, the laboratory accelerant sample, and the canine evidence.

2. Is There a Reasonable Probability That as a Result of the Daubert Hearing, the Court Would Have Excluded or Limited Evidence?

In order to evaluate the outcome of a potential Daubert hearing, the evidence needs to be divided into two categories — the cause-and-origin testimony, which purported to show where the fire began, on the one hand, and the laboratory and canine evidence, which purported to show that the fire was incendiary in nature. There is no question that had evidence in the latter categories been excluded or limited, even if the cause-and-origin testimony remained, confidence in the verdict would be undermined. I therefore begin with laboratory and canine evidence.

a. Laboratory Accelerant Sample Test

There is a "reasonable probability" the Court would have excluded the laboratory accelerant sample test, for all of the reasons already described: Investigators did not take any comparison samples in direct contravention of NFPA 921 guidelines. They took a sample from only one location — the area that they had already concluded was the point of origin. Control samples are essential in any experiment or test in order to discount false positives. The logic applies even more forcefully here. As even Domingos and Drugan acknowledged, many common materials will test positive for accelerants. Trial Tr. vol. 3, 98:14-99:15; Trial Tr. vol. 4, 48:4-49:6; Lentini Aff. Ex. 4, at 50 ("Forensic Science Committee Positions on Comparison Samples") (document #139-4); see also, Evid. Hr'g Tr. 48:7-16; Evid. Hr'g Ex. 17, (Higgins' Letter), at 1-2. Moreover, the laboratory result was generic and broad classification into which a host of benign materials may fall.

The collection of comparison samples is especially important in the collection of materials that are believed to contain liquid or solid accelerants. For example, the comparison sample for physical evidence consisting of a piece of carpeting believed to contain a liquid accelerant would be a piece of the same carpeting that does not contain any of the liquid accelerant. Comparison samples allow the laboratory to evaluate the possible contributions of volatile pyrolysis products to the analysis and also to estimate the flammability properties of the normal fuel present.

NFPA 921 § 14.5.3.4.

The government argues that the single sample would have passedDaubert scrutiny because NFPA 921 states that while multiple samples are desirable, "[i]t is recognized that comparison samples may be unavailable due to the condition of the fire scene. It is also recognized that comparison samples are frequently unnecessary for the valid identification of ignitable liquid residue." NFPA 921 § 14.5.3.4. That reasoning simply does not apply here.

One sample was demonstrably unreliable here. Without a control, the jury could not know if "light petroleum distillate" was found on the carpet because Hebshie used an ignitable liquid to start a fire or because of something in the ambient environment or a product of the fire. A test without a control is as good as no test at all.

Making matters worse, the investigator's failure to collect additional samples was compounded by the razing of the building. The defense had no way to contest the significance of the laboratory test. They were entirely dependent on the government's limited investigation. See NFPA 921 § 14.3 ("Every attempt should be made to protect and preserve the fire scene as intact and undisturbed as possible. . . . Generally, the cause of a fire or explosion is not known until near the end of the investigation . . . As a result, the entire fire scene should be considered physical evidence and should be protected and preserved.)."

As the Supreme Court pointed out in Kumho, when performing aDaubert inquiry, the relevant question is "not the reasonableness in general" of a particular methodology, but "the reasonableness of using such an approach . . . to draw a conclusion regarding the particular matter to which the expert testimony was directly relevant." Kumho Tire Co., 526 U.S. at 153; see also Faigman et al., supra, at § 7-1.1. (Daubert's "gatekeeping requirement would be defeated if multiple techniques were allowed to hide behind a global claim of expertise. In other words, Daubert appears to require that an expertise be unpacked, so that the court can permit those methods and principles it is persuaded are valid, and only those, to be offered to the factfinder. Kumho Tire makes that requirement explicit").

The government cannot just say that in some situations, or in general, a single sample will suffice. They have to show that in this case, on these facts, a single sample was reliable. And it was not.

b. Canine Evidence

There is also a "reasonable probability" that the Court would have excluded the canine testimony or severely limited it. If the laboratory accelerant test were excluded, the canine alert would also have been excluded. NFPA 921 § 14.5.3.5 warns: "Any canine alert not confirmed by laboratory analysis should not be considered validated." If the dog alerts on a specific spot, the investigators know there is a better chance that laboratory analysis will reveal the presence of an accelerant. But without laboratory corroboration, the canine alert is questionable. See Carr v. State, 482 S.E.2d 314, 318 (Ga. 1997), (overturning a murder conviction because laboratory sample tested negative for the presence of accelerants), ( overruled on other grounds by Clark v. State, 515 S.E.2d 155 (Ga. 1999).

In any event, apart from the laboratory test, if, at a Daubert hearing, the government had presented an offer of proof mirroring Lynch's testimony, there is no question that testimony would have been severely limited or excluded and if it had not been excluded or limited before the trial, it would have been excluded afterwards as Lynch began to wax poetic about his dog's unique prowess. As noted above, NFPA 921 and the scientific literature have accepted canine alerts only as tools to help investigators narrow the search area. Lynch's testimony went far beyond that.

c. Cause-and-Origin Testimony (Domingos and Miller)

Domingos (and Miller on rebuttal) testified both as to cause-and-origin and their conclusion that the fire was arson. As to the latter testimony — that the fire was arson — there is surely a "reasonable probability" of exclusion. Without a valid accelerant laboratory test, Domingos' and Millers' opinions that the fire was incendiary was simply not reliable.

Whether I would have allowed Domingos and Miller to conclude that the fire started in Hebshie's store is a closer question. Domingos' interpretations of the burn patterns, the heating grate, and other such evidence likely would have been admitted since they were his opinions based on his observation of the scene and experience in other fire scenes. While defense counsel had disagreed with these interpretations, they countered through cross-examination and rebuttal testimony, although, to be sure, examination that was not as effective as it could have been had there been a hearing.

I have greater concern about the investigator's failure to document any other potential cause — take photographs of the basement, take samples from other areas of the store for laboratory analysis. See Pekarek v. Sunbeam Prods. Inc., 672 F. Supp. 2d 1161, 1176-77 (D. Kan. 2008) (allowing expert testimony ruling out specific items as possible causes of fire but excluding opinion that cause of fire was electrical blanket where expert failed to follow reliable methodology). Petitioner argues that Domingos' failure to abide by essential steps in the scientific method means that his conclusions were not supported by "appropriate validation," Daubert, 509 U.S. at 590, or "accompanied by a sufficient factual foundation;" Boucher v. U.S. Suzuki Motor Corp., 73 F.3d 18, 22 (2d Cir. 1996) (quoting Gumbs v. Int'l Harvester, Inc., 718 F.2d 88, 98 (3d Cir. 1983)).

NFPA 921 repeatedly insists on the importance of documenting the investigation. NFPA 921 § 13.1 states: "Thorough and accurate documentation of the investigation is critical because it is from this compilation of factual data that investigative opinions and conclusions will be supported and verified." See also id., at § 13.2 ("A visual documentation of the fire scene is "the most efficient reminder[] of what the investigator saw while at the scene. Patterns and items may become evident that were overlooked at the time the photographs or videos were made."); id. at § 14.3 (Preservation of the Fire Scene and Physical Evidence). Section 13.2.5.6(a) directs investigators to take photographs of areas near the supposed point of origin, even if there is no visible damage. See id. at § 13.2.5.6(a) ("Rooms within the immediate area of the fire origin should be photographed, even if there is no damage."). The DOJ Manual, Fire and Arson Scene Evidence,supra, at 29-30, similarly emphasizes the necessity of proper documentation.

With the building demolished shortly after the fire, there was no way for investigators to test other hypotheses about how the fire started. While the government pointed to the sizable number of pictures taken — over a hundred — proper documentation is about quality, not just quantity. See Ind. Ins. Co. v. Gen. Elec. Co., 326 F. Supp. 2d 844, 850-51 (N.D. Ohio 2004) (holding that cause-and-origin expert's failure to properly collect evidence, in violation of the section of NFPA 921 entitled, "Documenting the Collection of Physical Evidence," made his investigation unreliable).

Documentation is necessary to test a hypothesis; in fact, reproducibility is the sine qua non of "science."

[A] key question to be answered in determining whether a theory or technique is scientific knowledge . . . [is] whether it can be (and has been) tested. `Scientific methodology today is based on generating hypotheses and testing them to see if they can be falsified; indeed, this methodology is what distinguishes science from other fields of human inquiry.'

Daubert, 509 U.S. at 593 (quoting Michael D. Green, Expert Witnesses and Sufficiency of Evidence in Toxic Substances Litigation: The Legacy of Agent Orange and Bendectin Litigation, 86 Nw. U. L. Rev. 643, 645 (1992)); see also id. (quoting Karl Popper, Conjectures and Refutations: The Growth of Scientific Knowledge 37 (5th ed. 1989) ("`[T]he criterion of the scientific status of a theory is its falsifiability, or refutability, or testability'"). The government experts had it both ways; their documentation had significant gaps and the scene was long gone.

In Chester Valley Coach Works v. Fisher-Price, Inc., No. Civ. A. 99 cv 4197, 2001 WL 1160012 (E.D.Pa. Aug. 29, 2001), the trial court excluded expert testimony under Daubert where the expert failed to follow the scientific method set out in NFPA 921. The expert "put[] the cart before the horse" by approaching his fire investigation with a hypothesis already in place, id. at *5, and never seriously considered other possible causes of the fire; id. at *7. Similarly, in Bryte ex rel. Bryte v. Am. Household, Inc., 429 F.3d 469 (4th Cir. 2005), the Fourth Circuit held that where the fire expert failed to consider and rule out other possible causes, his inspection was properly excluded for being "inconsistent with the NFPA standards, which require investigators to exclude `[a]ll other reasonable origins and causes.'" Id. at 478 (quoting NFPA 921 § 2-3.6 (1999 ed.)).

In Fireman's Fund Ins. Co. v. Canon U.S.A., Inc., 394 F.3d 1054, 1058 (8th Cir. 2005), the Eighth Circuit held that the district court appropriately excluded expert cause-and-origin testimony where the experts failed to "carefully examine[] th[eir] hypothesis of fire origin against empirical data obtained from fire scene analysis and appropriate testing, as required by NFPA 921." Id. at 1059; see also Pekarek v. Sunbeam Prods, Inc., 672 F. Supp. 2d 1161, 1176 (D. Kan. 2008) ("[T]roubling is [the expert]'s failure to investigate or consider the significance of the two tripped or off circuit breakers in the panel, and whether their condition provided any evidence of the source of the fire. Additionally, [his] report and deposition testimony reflect a failure to fully account for what was obviously a possible alternative cause of the fire. . . .").

Because Miller's conclusion vis-a-vis the cause of the fire is drawn entirely from Domingos' testimony and report, it presents the same problems. Miller never went to Hebshie's store. (Indeed, it was long gone. Trial Tr. vol. 6, 136:25-127:1. He testified that he based his opinions on the fire marshal's report and photographs, the ATF reports, a time sheet from the alarm company, and an additional interview he conducted with Domingos. Id. at 109:19-25, 113:18-114:1. He did no independent evaluation of the evidence. And while Miller gave his opinion with certainty, he acknowledged the incomplete nature of the investigation and that evidence regarding the basement would have been important. Id. at 138:10-23.

Q: And any opinions that you've [sic] drawing today or in the past are based in part on your conversations with Mr. Domingos; isn't that true? A: Some of the conclusions, yes. Trial Tr. vol. 6, 137:16-19.

He repeatedly limits his conclusions to "the information given to him." See id. at 112:1-2 ("It is the lowest burn that I have information about being at and near floor level."); Id. 112:2-3 (" I have no information that there was any fire in the basement."); id. at 113:25-114:1 ("[E]limination of other factors based on the information I received.").

Testimony similar to that of Miller was excluded in Weisgram v. Marley Co., 169 F.3d 514, 519-20 (8th Cir. 1999). In Weisgram, the Eighth Circuit held that the district court had abused its discretion in admitting the testimony of two experts on the cause-and-origin of a fire. It held that opinions given by Freeman, the fire captain who investigated the scene, about the cause of the fire should have been excluded because they were not supported by the record. It also excluded the testimony of an electrician, who based much of his conclusions on the faulty observations and opinions of Freeman. Id. at 519l; see also Kozar v. Sharp Elecs. Corp., No. 04-901, 2005 WL 2456227, at *3 (W.D. Pa. Sept. 30, 2005) (excluding expert report where expert failed to comply with NFPA 921 and his report was based primarily on opinion of another expert, who was precluded through Daubert).

Nevertheless, I need not reach that question of whether I would have excluded Domingos' and Millers' cause-and-origin testimony. As I describe below, the exclusion of the laboratory analysis and the limitation of the canine evidence would surely be enough to undermine confidence in this verdict. Without them, Domingos' and Millers' cause-and-origin testimony was not enough to prove a crime. Just because the fire may have started in Hebshie's store does not rule out accidental ignition.

3. Is There a Reasonable Probability That Exclusion or Limitation of the Expert Evidence Would Have Undermined Confidence in the Verdict?

The Court considers three factors in its prejudice inquiry: first, the strength of the state's case; second the effectiveness of the defense's case at trial; and third, the effect of the evidence's exclusion in undermining the government's case.Gonzalez-Soberal v. United States, 244 F.3d 273, 278 (1st Cir. 2001).

The government's case consisted of (1) the accelerant laboratory test and canine evidence; (2) the cause-and-origin testimony placing the fire's location in Hebshie's store; and (3) evidence about Hebshie's motive, including testimony that he owed $5,000 to the lottery, that a potential buyer had fallen through, that Hebshie had a $30,000 insurance policy, and that he had been uncooperative with the insurance claims investigator.

Without the laboratory test or the canine evidence, there was no other direct evidence of arson.

The motive testimony was not substantial and, to a degree, was undercut by other evidence. First, the fact that Hebshie left just minutes before the alarm went off cuts both ways. It could suggest that he set the fire and took off. At the same time, the fact that Hebshie set the alarm with his personal passcode makes arson more improbable. Similarly, the timing of the fire seems very risky for a relatively small award. It makes little sense to set a fire at the busiest time of day — midday on a Saturday — when both other stores in the building were open and there might be foot traffic on the street?

The government argues that counsel's errors did not prejudice Hebshie because Spinale cross-examined Domingos and Drugan about the lack of a control and the possibility that the positive could have come from other products. But as has already been shown, cross-examination was not remotely to deal with severe and fundamental problems with the laboratory test and the canine evidence. Indeed, to say that the exclusion of either the accelerant laboratory analysis or the canine evidence would have undermined this verdict and grievously prejudiced Hebshie, is an understatement. There would have been no case at all.

Hebshie also claims Strickland errors in the Spinales' failure to object to testimony that a firefighter suffered a heart attack. I agree. Indeed, counsel's error in not objecting to this testimony was compounded by this Court's error. The Court wrongly overruled counsel's objection to testimony concerning how much money the owners of the other stores lost. Finally, Hebshie claims a Strickland error in connection with counsel's failure to object to the Court's erroneous mail fraud instruction. See United States v. Hebshie, 549 F.3d 30, 31 (1st Cir. 2008). While I agree that these two errors represent deficient performance, standing alone, they did not prejudice Hebshie.

IV. CONCLUSION

As Larry A. Hammond, past president of the American Judicature Society, noted:

By routinely allowing into evidence expert testimony that we know should have been excluded, and by closing courthouse doors to claims for redress after conviction, the courts have contributed to the problems we face today.

Larry A. Hammond, The Failure of Forensic Science Reform in Arizona, 93 Judicature 227, 2 (2010). In fact, he suggested, one thing that each of the cases in which there have been wrongful convictions necessarily have in common is that each were presided over by a judge, an appellate court, and typically had post conviction habeas review. And then he concluded:

One would hope that with the announcement of every exoneration, the judges across whose desks these cases passed would pause to ask, `what can we do to make sure that this doesn't happen again?

This Court recognizes the importance of finality in criminal cases, particularly after time and resources have gone into the trial, and after a jury has pronounced guilt. But finality cannot trump fairness or justice. If Hebshie's defense of serious charges was fatally undermined by ineffective counsel, I am duty bound to say so. Therefore, Hebshie's habeas petition is GRANTED.

SO ORDERED.