Daniel Douglas Martin appeals his sentence on revocation of supervised release. Because we hold that Martin's due process right to confrontation was violated, we reverse and remand.

* Martin pleaded guilty to possession with intent to distribute heroin and was sentenced to 24 months' imprisonment and four years of supervised release. He was released from prison on January 23, 1992. On March 18, 1992, the district court revoked Martin's supervised release, finding four separate violations of the conditions of his release: 1) submission of two urine specimens that tested positive for cocaine and methadone; 2) failure to report for drug testing and counseling on two occasions; 3) failure to report termination of his employment; and 4) leaving the District of Hawaii without permission.

Martin admitted the last three violations and contested the first. As to the first alleged violation, he admitted using methadone as part of a treatment program but denied using cocaine.1 Martin concedes that the three admitted violations support revocation of supervised release. However, the contested violation led to an enhanced sentence. The district court found that Martin had possessed a controlled substance and applied 18 U.S.C. § 3583(g), which requires imprisonment for at least one-third of the term of supervised release; the court therefore sentenced Martin to 16 months' imprisonment.2 In the absence of section 3583(g), the recommended sentencing range under the Sentencing Guidelines would have been four to ten months. U.S.S.G. § 7B1.4(a).

The only evidence supporting the finding of possession of a controlled substance consisted of two laboratory urinalysis reports showing the presence of methadone and cocaine metabolites. Although actual testing of the samples was conducted by PharmChem Laboratories in Menlo Park, California, Gary Noland, a counselor at Drug Addiction Services of Hawaii (DASH) who collected the samples from Martin, testified to the test results. The government presented no other witnesses. Noland described collection and shipping procedures at DASH. On cross-examination, Noland was unable to testify to the particular tests employed on Martin's samples or to PharmChem's general testing and handling procedures. Noland indicated, however, that the samples should still be available for retesting. Upon learning of the samples' likely availability, Martin's counsel, who had been appointed the day of the final revocation hearing, requested that the court allow counsel to arrange for an independent retesting of the samples. Without explanation, the district court denied the request.

In Morrissey v. Brewer, 408 U.S. 471, 92 S. Ct. 2593, 33 L. Ed. 2d 484 (1972), the Supreme Court defined certain minimal due process requirements for parole revocation. The Court quickly extended these protections to probation revocation. See Gagnon v. Scarpelli, 411 U.S. 778, 93 S. Ct. 1756, 36 L. Ed. 2d 656 (1973). Fed.R.Crim.P. 32.1, which applies to supervised release revocation, incorporates these same minimal due process requisites. We must therefore decide if Martin's Morrissey due process rights were violated by the district court's admission of the laboratory results and its refusal to allow independent retesting of the urine specimens.

Among the Morrissey mandates is "the right to confront and cross-examine adverse witnesses (unless the hearing officer specifically finds good cause for not allowing confrontation)." 408 U.S. at 489, 92 S.Ct. at 2604; see also Fed.R.Crim.P. 32.1(a)(2)(D) ("opportunity to question adverse witnesses"). Although it is possible to evaluate Martin's claims under separate Morrissey requirements, we conclude that these alleged errors may be addressed most logically and meaningfully, together, under the right to confrontation.3 We construe that right as requiring that a supervised releasee receive a fair and meaningful opportunity to refute or impeach the evidence against him in order "to assure that the finding of a [supervised release] violation will be based on verified facts." See Morrissey, 408 U.S. at 484, 92 S.Ct. at 2601.

In cases involving the Morrissey right to confrontation, we employ "a process of balancing the [releasee's] right to confrontation against the Government's good cause for denying it." United States v. Simmons, 812 F.2d 561, 564 (9th Cir.1987). See also United States v. Kindred, 918 F.2d 485, 486 (5th Cir.1990); United States v. Bell, 785 F.2d 640, 642 (8th Cir.1986); United States v. Penn, 721 F.2d 762, 764 (11th Cir.1983). We address each side of this balance in turn.

* Simmons and the other cases offer minimal instruction on weighing the releasee's right to confrontation. The cases have tended to focus nearly exclusively on the "good cause" side of the balance. The Fifth Circuit has defined the "right" side of the balance as measuring the defendant's "interest in confronting a particular witness," Kindred, 918 F.2d at 486, but that definition provides little in the way of guidance. In general, the Morrissey right to confrontation has itself been but murkily defined. A releasee's right to confrontation is not equivalent to that afforded a criminal defendant at trial. Morrissey, 408 U.S. at 489, 92 S. Ct. at 2604; Simmons, 812 F.2d at 564. But aside from the "upper limit" imposed by this conclusion, precedent provides sparse indication of the more specific parameters of a releasee's right to confrontation.

This dearth of guidance does not mean that we can simply ignore one side of the Simmons right-cause balance. Doing so would suggest a static right, but Morrissey emphasizes the flexible nature of due process. "Its flexibility is in its scope once it has been determined that some process is due; it is a recognition that not all situations calling for procedural safeguards call for the same kind of procedure." 408 U.S. at 481, 92 S.Ct. at 2600. Morrissey invoked this precept in arriving at the minimal due process requisites of a revocation hearing, but the principle of flexibility applies equally to our further interpretation of the skeletal framework of rights put forward by the Court.

In addition to being unfaithful to Morrissey, a static right to confrontation would undercut the balancing process adopted in Simmons. The test would make little sense if one side of the balance was a constant. If the right were unchanging, we could simply define a requisite level of government good cause to overcome it, and measure each case against this single-variable barometer. Cf. Penn, 721 F.2d at 765 ("[o]bviously, since this is a balancing test, there can be no fixed rules on what would constitute good cause in every case"). We adopted a balancing approach because sufficient good cause in one set of circumstances may be insufficient in another. We must therefore weigh Martin's right to confrontation under the specific circumstances presented.

Recognizing that a releasee's Morrissey rights at a revocation hearing do not rise to the level of similar rights at a criminal trial, we nonetheless have no difficulty in concluding that Martin's Morrissey right to confrontation was substantial under these circumstances. The factors that lead us to this conclusion include the importance of the evidence to the court's ultimate finding, the virtually complete denial of any opportunity to refute the evidence, and the consequences of the court's finding.4

Our first consideration--the importance of the evidence to the court's ultimate finding--follows from the Supreme Court's emphasis on assuring that findings be based on "verified facts." Morrissey, 408 U.S. at 484, 92 S.Ct. at 2602. The more significant particular evidence is to a finding, the more important it is that the releasee be given an opportunity to demonstrate that the proffered evidence does not reflect "verified fact." Cases from this and other circuits focusing upon "reliance" by the trier of fact further support this principle. See, e.g., United States v. Garcia, 771 F.2d 1369, 1371 n. 2 (9th Cir.1985) (appellate court's first inquiry should be into "evidence actually relied upon by the trier of fact"); United States v. Taylor, 931 F.2d 842, 847 (11th Cir.1991) (probationer must show "that the court explicitly relied on the information"), cert. denied, --- U.S. ----, 112 S. Ct. 1191, 117 L. Ed. 2d 433 (1992).

Here, the laboratory results were uniquely important to the court's finding that Martin possessed a controlled substance. Martin denied using cocaine, and the government presented no other evidence of cocaine use, much less possession.5 Thus, Martin had a very strong interest in refuting the laboratory results, upon which the district court exclusively relied in finding possession.

Despite this strong interest, Martin had virtually no opportunity to refute the test results. This nearly complete denial of any confrontation is the second factor contributing to our conclusion that Martin's right weighs heavily in the Simmons balance. That this factor is an appropriate consideration follows from the Morrissey flexibility principle, as well as our own cases emphasizing "prejudice." See, e.g., Standlee v. Rhay, 557 F.2d 1303, 1307-08 (9th Cir.1977) (parolee must show absence of procedural safeguard was prejudicial). Had Martin been given some opportunity--even if not the traditional opportunity to cross-examine a live witness--to refute this important evidence, then that opportunity could be considered in partial mitigation of his right to confrontation. No such partial opportunity was here afforded.

Although Martin did cross-examine Noland, Noland's only knowledge related to collection and shipping procedures for the urine samples. Noland was unable to shed any light on chain of custody or testing procedures once the samples left DASH and presumably arrived at PharmChem. Moreover, the court denied Martin the opportunity to retest the samples independently, which would have allowed him to impeach more directly the positive laboratory results. In effect, Martin was denied all but the most cursory opportunity to impeach or refute the evidence supporting the court's finding of possession.

Ordinarily, the importance of the test results and the denial of any meaningful opportunity to impeach the results would suffice to conclude that Martin's right to confrontation was substantial. Here, however, this conclusion finds further support in our third factor--the consequences of the possession finding. As noted above, the court's finding that Martin possessed a controlled substance triggered application of the mandatory minimum specified in 18 U.S.C. § 3583(g). Application of the statute divested the district court of sentencing discretion and required a term of imprisonment sixty percent higher than the maximum in the range suggested by the Sentencing Guidelines policy statements, and four times higher than the minimum in that range. Thus, the evidence for which Martin alleges he was denied the right to confrontation resulted in a significant enhancement of his prison term.

Morrissey suggests that these consequences are appropriate considerations in the Simmons right-cause balance. In describing the need for informal revocation hearings, the Court envisioned a two-step decision-making process: an initial "factual" phase to determine whether the defendant violated conditions, and a subsequent "predictive and discretionary" phase to determine what appropriate steps should be taken in light of any violations. 408 U.S. at 480, 92 S.Ct. at 2600. Section 3583(g) severely restricts the discretion involved in the second phase.6 Because the Supreme Court premised, in part, its prescription of informality and flexibility upon the existence of such discretion, constriction of that discretion requires even greater protection of the defendant's right to confrontation in section 3583(g) cases. Thus, in applying the right-cause balance, we properly take account of the mandatory nature of the sentence imposed by section 3583(g).

Because the evidence Martin sought to refute was singularly important to the district court's finding, because Martin was denied virtually any opportunity to refute that evidence, and because the court's finding had such severe consequences, Martin's Morrissey right to confrontation weighs heavily in the Simmons balance. As a result, the government's "good cause" for denying confrontation must be particularly significant if it is to outweigh Martin's right.

Our cases provide greater guidance on weighing government good cause in the Simmons balance. Previously, we have looked to both the "difficulty and expense of procuring witnesses," Gagnon, 411 U.S. at 783 n. 5, 93 S. Ct. at 1760 n. 5, and the "traditional indicia of reliability" borne by the evidence, Simmons, 812 F.2d at 564, in evaluating good cause. After considering these two factors, we conclude that the government's good cause fails to outweigh Martin's substantial Morrissey right to confrontation.

We dismiss the first factor, difficulty and expense, as carrying little weight in this case because the government provided absolutely no substitute for live testimony. Assuming that producing a live witness from PharmChem in California would entail significant expense, the government nonetheless provided none of the "conventional substitutes for live testimony" identified by the Supreme Court, "including affidavits, depositions, and documentary evidence." Gagnon, 411 U.S. at 783 n. 5, 93 S. Ct. at 1760 n. 5. Neither did the government provide, nor the district court allow, another possible alternative to live testimony, germane to this particular class of cases--independent retesting of preserved specimens. Each of these presumably less expensive alternatives would have provided Martin with some opportunity to impeach the test results.7 Because the government eschewed even these alternatives, the expense or difficulty of procuring witnesses carries no weight in our consideration of good cause.

The "reliability" factor presents a more difficult issue. Three other circuits have previously upheld the admission of urinalysis reports in revocation hearings on the basis of their indicia of reliability. See United States v. Kindred, 918 F.2d 485, 487 (5th Cir.1990); United States v. Burton, 866 F.2d 1057, 1059 (8th Cir.), cert. denied, 490 U.S. 1110, 109 S. Ct. 3167, 104 L. Ed. 2d 1029 (1989); United States v. Bell, 785 F.2d 640, 643 (8th Cir.1986); United States v. Penn, 721 F.2d 762, 765-66 (11th Cir.1983).

Each of these cases is distinguishable. For example, in Kindred and Penn, the defendants "did not contest the allegations of drug use or the accuracy of the urinalysis test." Kindred, 918 F.2d at 487. Here, Martin denies cocaine use and has a correspondingly greater interest in the opportunity to refute the laboratory reports. Penn, in addition, rested on a specific finding of reliability, which the court of appeals held was not "clearly erroneous." 721 F.2d at 766. Here, the district court made no such specific finding. In Burton, the government supported the urinalysis results with an affidavit from the director of the laboratory. 866 F.2d at 1058. As we note above, here the government introduced nothing regarding the laboratory procedures. In Bell, there was separate evidence corroborating the defendant's use of marijuana, including an arrest for possession of marijuana and narcotic paraphernalia. See 785 F.2d at 642. In this case, the urinalyses constitute the only evidence of cocaine use or possession.

Despite these and other distinctions,8 the government argues that we should affirm on the basis of Penn, Bell, and Burton. In essence, the government urges us to hold that urinalysis reports are so inherently reliable that they may be introduced in any revocation hearing. While it is possible to read some of the other cases, in particular Bell, as announcing such a blanket rule, we decline to adopt such a doctrine in this circuit. First, neither the government, nor these other cases, demonstrate that custody problems and testing errors happen with such rarity at testing laboratories that their reports are always inherently reliable. Second, such a blanket rule would be tantamount to abandonment of the Simmons balancing test; we would effectively hold that the weight of the defendant's right to confrontation is irrelevant in revocations involving urinalyses, which is likely a significant class of cases.

We see no reason to adopt such a departure from the test announced in Simmons. The right-cause balance is flexible, as mandated by Morrissey, without abandoning the respect for individual rights that is just as significant a part of Morrissey due process as flexibility.9 The balancing test is a workable means "to assure that the finding of a [supervised release] violation will be based on verified facts and that the exercise of discretion will be informed by an accurate knowledge of the [releasee's] behavior." Morrissey, 408 U.S. at 484, 92 S.Ct. at 2601. District courts should apply the balancing test to every alleged violation of the Morrissey right to confrontation; urinalysis evidence provides no exception.

Having declined to adopt a blanket rule, we nonetheless accord a certain amount of weight to the test results' indicia of reliability. Like the Eighth Circuit, we recognize that these "are the regular reports of a company whose business it is to conduct such tests, and which expects its clients to act on the basis of its reports." Bell, 785 F.2d at 643. However, because the government presented negligible information about PharmChem's experience or qualifications in this area,10 we accord little weight to this conclusion. Moreover, even if we were to assume that PharmChem has extensive experience in this area,11 and that, as a consequence, its reports carry greater indications of reliability,12 this level of good cause still would not outweigh Martin's substantial right to confrontation.

We conclude that, under the circumstances presented, the government's good cause for denying confrontation fails to outweigh Martin's right to confrontation. Accordingly, the district court, by admitting the laboratory results solely upon the testimony of counselor Noland and refusing to allow independent retesting, violated Martin's right to due process. We reverse and remand for further proceedings consistent with this opinion.

REVERSED and REMANDED.