WALKER, Circuit Judge:
The government appeals from a judgment of conviction entered in the United States District Court for the Southern District of New York (Charles S. Haight,
District Judge).
On appeal, the government challenges the district court’s authority to depart downward from Williams’s guideline sentencing range so that he could enter a drug treatment program to which he had been admitted. Although we believe that the district court had authority to depart downward
in the circumstances of this case, we vacate the sentence so that the district court can impose a special condition of supervised release to ensure that Williams serves at least his guideline minimum sentence if he does not successfully complete the drug treatment program.
BACKGROUND
In the winter of 1993, Michael Williams participated in two hand-to-hand sales of approximately 25 and 26 grams of crack cocaine to an undercover agent of the Bureau of Alcohol, Tobacco, and Firearms. Williams was arrested on February 10,1993. On July 16, 1993, he pled guilty to a superseding information pursuant to a written plea agreement. Williams has been in federal custody since his arrest.
The information charged Williams with two counts of distributing and possessing with intent to distribute five grams and more of mixtures and substances containing detectable amounts of cocaine base in a form commonly known as “crack.” Under the plea agreement, the government agreed not to oppose a four-level reduction in Williams’s base offense level for minimal participation and a three-level reduction for acceptance of responsibility. The parties stipulated that Williams’s total offense level was 27, his criminal history category was VI, and his sentencing range was 130 to 162 months in prison. Both sides agreed not to seek any departures. The Probation Department filed a Presentenee Report stating that the 130-162 month sentencing range was appropriate and that no known circumstances warranted a departure from the guideline range.
Prior to the first sentencing hearing, defense counsel submitted a letter urging Judge Haight to consider that Williams’s criminal record was largely a product of drug addiction and that Williams had volunteered for a drug education program while detained at the Federal Correctional Institute at Otis-ville (“Otisville”). Defense counsel attached to the letter a copy of a report from a drug treatment specialist at the Bureau of Prisons, who commented on Williams’s flawless participation in the drug education program, and a letter from Williams expressing remorse and a desire to enter a drug treatment program at the federal prison in Butner, North Carolina.
At the initial hearing, Judge Haight notified the parties of his intention to depart downward
sua sponte,
stating that Williams’s brother appeared to be the “architect” of the drug distribution scheme and that Williams “was going along primarily for the purpose of indulging his own addiction.” Judge Haight observed further that Williams’s participation in the Otisville drug education program, his desire to attend the Butner treatment program, his expression of remorse, and his demeanor convinced him that Williams had a genuine desire for rehabilitation. He determined that the appropriate approach would be to depart downward to the mandatory minimum of five years imprisonment, but to extend the supervised release term to ten years so that “if even once he goes back to the drug life he led before ... [Williams] will go back to prison for a period of time comparable to that required by the guidelines.” Judge Haight then sentenced Williams to two concurrent terms of five years’ imprisonment followed by two concurrent terms of ten years of supervised release.
The government appealed from that sentence. In
United States v. Williams,
37 F.3d 82 (2d Cir.1994)
(Williams
I), we vacated the judgment and remanded for resentencing. While we recognized that “in an appropriate case, a defendant’s rehabilitative efforts in ending his drug dependence may be a permissible ground for a downward departure,”
id.
at 86 (citing
United States v. Maier,
975 F.2d 944, 948 (2d Cir.1992)), we could not conclude that “the district court’s departure based on Williams’s attendance in a drug education program and genuine desire to enroll in a drug treatment program was proper.”
Id.
Pointing out that “Williams has not yet even entered a drug treatment program and can point to few objective indications of his progress towards overcoming his drug dependence,” we held that Williams’s “genuine desire to seek rehabilitative treatment in the future” fell short of the “extraordinary” efforts at rehabilitation that justified a departure in
Maier. Id.
Soon thereafter, Williams completed his drag education program at Otisville. Although the district court had recommended that Williams be incarcerated at a prison where drag treatment would be available, the Bureau of Prisons transferred him to Ray Brook federal prison, a facility which did not have a program for drag therapy or rehabilitation. Finally, in September of 1994, Williams was transferred to the federal prison in Butner, North Carolina.
At Butner, Williams was accepted to its Pilot Drag Abuse Program (“Butner program”). According to literature provided by Butner, the program is an outgrowth of a 1988 conference hosted by the Bureau of Prisons which gathered “many nationally recognized experts in the areas of corrections, drag abuse treatment, and research.” The conference recommended that the Bureau of Prisons establish an “intensive pilot drag abuse treatment program with a comprehensive research evaluation methodology at a federal correctional institution.” Butner was selected as the facility to develop that program. The program targets inmates with significant drag histories who volunteer for treatment. It integrates many different -kinds of services designed to help inmates rid themselves of drag dependency while in prison and avoid relapse thereafter. The Butner program operates with a 12:1 inmate to staff ratio and consists of five phases involving “1000 hours of intensive institutional treatment” over 12 to 15 months, in contemplation of a significant post-release program.
Designers of the Butner program limited enrollment to 100 inmates and established eight criteria for admission. The inmate, who could be of any security level designation, had to (1) be male; (2) be 18 to 36 months away from a confirmed release date; (3) have a history of moderate to severe substance abuse; (4) volunteer for program participation; (5) sign a written agreement to observe program requirements; (6) be free of violence-related disciplinary problems during incarceration; (7) write and speak English; and (8) have no serious medical, psychiatric, or psychological problems aside from drag abuse that would interfere with full program participation.
At resentencing on November 18,1994, the district court noted that it was bound by
Williams I
to the extent that the facts had not changed. It felt, however, that Williams’s admission to the selective drag treatment program based on objective factors and his subjective willingness to commit to the program regimen was a significant changed circumstance; it meant that Williams was no longer in the category of someone who simply manifested a desire to rehabilitate himself and had not even entered a treatment program. Moreover, Williams’s counsel represented to the court that in his efforts to secure placement for his client he had found an “extraordinary lack of drag programs w[ith]in the federal system” given that newly authorized programs were not slated to commence until 1996 or 1997. The district court noted that 18 U.S.C. § 3553(a)(2)(D) mandates a sentencing court to take account of the defendant’s need for “medical caret ] or other correctional treatment in the most effective manner.” It also pointed out that the vagaries of federal funding and the uncertain effect of prison life on Williams’s attitude might later thwart the opportunity for rehabilitation that now presented itself. The district court concluded that in setting the 130-162 month sentencing range applicable to a defendant with Williams’s criminal history who had committed the same offenses, the Sentencing Commission could not have considered the particular circumstances of this case: namely,
an inmate who fits within a rather narrow profile for a particular and unique and small and selectively available pilot program, which can do him some good now, one hopes, which in the absence of a downward departure would not be available to him for a significant number of years; if at the end of that significant number of years it was available to him at all, as to which there can be no guarantee.
Accordingly, the district court departed from the guideline range and imposed the same sentence that had been vacated. The government appeals, contending that the departure is impermissible under
Williams I.
DISCUSSION
We summarized the standard of review for challenges to a district court’s departure from the applicable sentencing range in
Williams I:
We review
de novo
the legal issue of whether a particular factor is a permissible ground for departure, and we review for clear error the district court’s factual findings as to the existence of that factor. If there is no clear error in the district court’s factual findings and no error in its conclusion that the factor is a permissible ground for departure, the court has discretion in determining what aggravating or mitigating circumstances to consider, and we review the resulting sentence to determine whether or not it is reasonable.
37 F.3d at 85 (citations and quotations omitted).
I. Permissibility of the Departure
In assessing the district court’s authority under the statutory scheme of the Sentencing Reform Act of 1984, Pub.L. No. 98-473, 98 Stat. 1987 (1984) (codified as amended at 18 U.S.C. §§ 3551 et seq. & 28 U.S.C. §§ 991-98) (“SRA”), which established the framework for the Sentencing Guidelines, we start with the recognition that “[t]he keystone section in the creation of the new sentencing system is 18 U.S.C. § 3553.”
United States v. Merritt,
988 F.2d 1298, 1306 (2d Cir.1993),
cert. denied,
— U.S. -, 113 S.Ct. 2933, 124 L.Ed.2d 683 (1994). In particular, “[s]ubparagraph (a) instructs the sentencing court as to the ‘[fjactors to be considered in imposing sentence.’ ”
Id.
Congress directs a sentencing court to consider, among other factors, the need for the sentence imposed—
(A) to reflect the seriousness of the offense, to promote respect for the law, and to provide just punishment for the offense;
(B) to afford adequate deterrence to criminal conduct;
(C) to protect the public from further crimes of the defendant; and
(D) to provide the defendant with needed educational or vocational training, medical care, or other correctional treatment in the most effective manner.
18 U.S.C. § 3553(a)(2).
In
United States v. Maier,
975 F.2d 944, 947 (2d Cir.1992), we concluded that § 3553(a)(2)(D) indicated that Congress did not abandon rehabilitation as a permissible goal of sentencing when it passed the SRA. We pointed out that Congress’s concern in enacting the SRA was not with the rehabilitative goal per se, but with “the vices of the indeterminate sentence regime of prior law, under which the effective time of incarceration was determined, not by the sentencing judge, but by the Parole Commission endeavoring to decide when the prisoner had been ‘rehabilitated.’”
Id.
at 947 n. 1. Because Congress continued to endorse rehabilitation as a goal of a sentencing in § 3553(a)(2)(D), and because there was no evidence that the Sentencing Commission had given adequate consideration to a defendant’s efforts at drug rehabilitation in formulating the Guidelines,
id.
at 948;
cf.
18 U.S.C. § 3553(b), we concluded that such endeavors could serve as a basis for a downward departure.
Maier,
975 F.2d at 948. Indeed, in light of § 3553(a)(2)(D), we questioned whether the Commission even had the authority to bar sentencing judges from considering a defendant’s efforts at drug rehabilitation.
Id.
However, “[although the Guidelines afford the district court flexibility in sentencing, the power to depart is to be used sparingly and is reserved for unusual cases.”
Williams I,
37 F.3d at 85. A tentative step towards rehabilitation is not usually enough to warrant a downward departure. In
Mai-er,
and again in
Williams I,
we emphasized that a drug rehabilitation program, which is “ ‘easily entered but difficult to sustain, cannot be permitted to become an automatic ground for obtaining a downward departure.’ ”
Id.
at 86 (quoting
Maier,
975 F.2d at 948). Accordingly, if departure is to be based on the defendant’s rehabilitative efforts, those efforts must be “extraordinary.”
Id.
In the government’s view, our holding in
Williams I
settles the matter. It argues that the only change in circumstances be
tween the first and the second sentencing is that Williams was granted admission to the Butner program, which
Williams I
and
Mai-er
indicate is not enough to warrant a departure.
To say that admission to a drug treatment program is not “an automatic ground for obtaining a downward departure,” however, is not to say that it can never be the basis for such a departure, provided that there exist other compelling circumstances not adequately considered by the Commission. Indeed, the Commission itself has disavowed the idea that sentencing can be reduced to simple rules mechanically applied without careful attention to the circumstances of a case. The Commission has “stresse[d] the difficulty of prescribing ‘a single set of guidelines that eneompass[es] the vast range of human conduct potentially relevant to a sentencing decision.’ ”
Merritt,
988 F.2d at 1309 (quoting U.S.S.C. Guidelines Manual [hereinafter “Manual”], Ch. 1, Pt. A, intro, cmt. 4(b) and altered to match original). Accordingly,
in the discussion of Departures in Part 4(b) of the Introduction [to the Manual], the Commission asserts that courts should treat each guideline as “carving out a ‘heartland,’ a set of typical cases” and that, “[w]hen a court finds an atypical ’case,” that “significantly differs from the norm, the court may consider whether a departure is warranted.” This discussion goes on to assert that (apart from a few enumerated factors that may not serve as a ground for departure), “the Commission does not intend to limit the kinds of factors, whether or not mentioned anywhere else in the Guidelines, that could constitute grounds for departure in an unusual ease.” ... [T]he overall treatment makes clear that departure in the appropriate case is essential to the satisfactory functioning of the sentencing system.
Id.
((quoting Manual, Ch. 1, Pt. A, intro, cmt. 4(b)) (citations omitted).
On remand, the district court did not depart from the guidelines sentencing range of 130 to 162 months simply because Williams had entered a drug treatment program. It departed because, on the facts of this case, there was effectively no other sentence that would accord with the requirements of 18 U.S.C. § 3553(a)(2)(D). The district court determined that Williams was an excellent candidate for rehabilitation given his prior history, demeanor, post-arrest resolve, and acceptance into a “special and selective” treatment program based on criteria devised by experts in the field. However, the only program available to Williams would not take him unless he were within 18 to 36 months of release. To sentence him to even the minimum term of 130 months, the district court reasoned, would require Williams to wait some six or seven years to begin treatment. If in the interim the Butner program were terminated for budgetary or other reasons or if Williams’s resolve weakened under the pressures of prison life, the chance of curing him of his addiction and perhaps his criminal ways would vanish.
We believe that the district court had the authority to depart downward in order to facilitate Williams’s rehabilitation given the atypical facts of this case, which place it outside the “heartland” of usual cases involving defendants who may benefit from drug treatment. As we stated in
Maier,
“awareness of one’s [drug dependency] and the demonstrated willingness to act to achieve rehabilitation, thereby benefiting the individual and society,” may justify a downward departure. 975 F.2d at 948. We clarified in
Williams I
that “demonstrated willingness” to rehabilitate one’s self must be manifested by objective indicia of extraordinary efforts to that end. 37 F.3d at 86. But when a defendant who has been in federal custody since his arrest has had no opportunity to pursue any rehabilitation, when he has been admitted to a selective and intensive inmate drug treatment program, and when a sentence within the guideline range would effectively deprive him of his only opportunity to rehabilitate himself while incarcerated, we think a departure is within the district court’s discretion. If the Sentencing Commission did not give adequate consideration to the mitigating circumstance of drug rehabilitation generally,
Maier,
975 F.2d at 948, it certainly did not consider the unique constellation of mitigating circumstances in this case.
The government’s position, in essence, would limit the authority to depart on the basis of drug rehabilitation to the extraordinary facts of
Maier.
In that case, the defendant, while on bail, was admitted to a residential drug treatment program before sentencing. At the defendant’s request, the district court postponed sentencing for three months to allow her to attend that program, and then for another year to pursue additional rehabilitation programs. 975 F.2d at 945. Since 18 U.S.C. § 3553(a)(2)(D) requires the district court to consider the need for
future
medical or correctional treatment in devising a sentence, we see no reason to limit the power to grant departures for rehabilitative purposes to defendants who, by grace of a district court’s delay in sentencing, have had an opportunity to develop a record of extraordinary
past
progress.
Moreover, if the district court had no power to adjust the term of incarceration to facilitate rehabilitation, it is difficult to see how the defendant’s need for medical or correctional treatment could otherwise be taken into account in sentencing. A sentencing court has no authority to order that a convicted defendant be confined in a particular facility, much less placed in a particular treatment program; those decisions are within the sole discretion of the Bureau of Prisons.
See
18 U.S.C. § 3621(b);
United States v. Restrepo,
999 F.2d 640, 644 — 45 (2d Cir.),
cert. denied,
— U.S.-, 114 S.Ct. 405, 126 L.Ed.2d 352 (1993);
United States v. Voda,
994 F.2d 149, 151-52 (5th Cir.1993).
In upholding the district court’s authority, we by no means imply that district courts should automatically grant downward departures to defendants in circumstances similar to that of Williams. We intend only to acknowledge that the district court has the discretion that remains a vital component of individualized sentencing under the Guidelines regime. As we said when commenting on the Guidelines generally:
The legislative history reflects that it was not Congress’ aim to straitjaeket a sentencing court, compelling it to impose sentences like a robot inside a Guidelines’ glass bubble, and preventing it from exercising discretion, flexibility or independent judgment. [As the sponsoring committee wrote,]
The Committee does not intend that the guidelines be imposed in a mechanistic fashion. It believes that the sentencing judge has an obligation to consider all the relevant factors in a case and to impose a sentence outside the guidelines in an appropriate case. The purpose of the sentencing guidelines is to provide a structure for evaluating the fairness and appropriateness of the sentence for an individual offender, not to eliminate the thoughtful imposition of individualized sentences.
United States v. Lara,
905 F.2d 599, 604 (2d Cir.1990) (quoting S.Rep. No. 225, 98th Cong., 2d Sess. 52 (1984),
reprinted in
1984 U.S.C.C.A.N. 3182, 3235). Here, the district court was impressed by Williams’s demeanor and resolve to change his ways,
cf. id.
at 603 (deferring to district court as being in the “best position to observe the defendant’s demeanor and characteristics”), convinced that Williams’s criminality was largely a product of his addiction and his brother’s malignant influence, and persuaded that there was only a limited window of opportunity for rehabilitation. Consequently, while the district court was not required to grant a downward departure, it acted within its power in doing so.
Because we find that the district court had authority to depart from the guideline range, and because we find none of its factual findings clearly erroneous, we ton now to the reasonableness of the sentence imposed.
Williams I,
37 F.3d at 85;
Maier,
975 F.2d at 949.
II. Reasonableness of the Sentence
Under 18 U.S.C. § 3742(e)(3), if the sentence imposed is outside the applicable guideline range, we must determine whether that sentence is “unreasonable.”
Maier,
975 F.2d at 949;
United States v. Palta,
880 F.2d 636, 639 (2d Cir.1989). In assessing reasonableness, we must “ ‘hav[e] regard for — (A) the factors to be considered in imposing a sentence, as set forth in chapter 227 of this title [18 U.S.C. §§ 3551-3586 (factors listed in section 3553(a)) ]; and (B) the reasons for
the imposition of the particular sentence, as stated by the district court pursuant to the provisions of section 3553(e).’”
Maier,
975 F.2d at 949 (quoting 18 U.S.C. § 3742(e)(3)).
The district court departed from the range of 130-162 months imprisonment prescribed by the guidelines and imposed the mandatory minimum prison term of 60 months, with ten years of supervised release. Departure to the mandatory minimum does not strike us as unreasonable in light of the district court’s assessment of Williams and the Butner program’s requirement that inmates have 36 months or less remaining of their prison term. Moreover, the sentence reasonably accounts for Williams’s rehabilitative needs as described by 18 U.S.C. § 3553(a)(2)(D) and, if Williams is cured of his addiction, may ultimately serve to protect the public from future criminal acts that Williams might otherwise commit, in furtherance of § 3553(a)(2)(C).
Moreover, the district court imposed a lengthy supervised-release term of ten years to allow it to sentence Williams to a prison term within his guideline range should he fail to meet the requirements of supervised release. Pursuant to § 3553(a)(4)(A), the district court adequately considered the guideline range in devising the sentence.
We are less confident, however, that the sentence imposed reasonably accounts for one factor that Congress directed the district court to consider: “the need to avoid unwarranted sentence disparities among defendants with similar records who have been found guilty of similar conduct,”
id.
§ 3553(a)(6). The government argues convincingly that, as the sentence now stands, Williams could simply withdraw from the Butner program at any time. Having outfoxed the district court (and this court), Williams would go free at the end of five years while similar defendants who committed similar crimes would serve another six to nine years, rendering the disparity “unwarranted.”
We agree with the government that taking such a risk is impermissible. Indeed, the ease with which one can enter and then abandon drug treatment programs is one of the reasons for the narrowness of our holdings in
Maier
and
Williams I.
That risk, however, can be mitigated at resentencing if the district court were to impose special conditions of supervised release. The statute authorizing supervised release permits a court to impose “any ... condition it considers to be appropriate” to the extent that such condition
(1) is reasonably related to the factors set forth in section 3553(a)(1), (a)(2)(B), (a)(2)(C), and (a)(2)(D);
(2) involves no greater deprivation of liberty than is reasonably necessary for the purposes set forth in section 3553(a)(2)(B), (a)(2)(C), and (a)(2)(D); and
(3) is consistent with any pertinent policy statements issued by the Sentencing Commission pursuant to 28 U.S.C. 994(a).
18 U.S.C. § 3583(d).
Although the district court in this case set a supervised release term of ten years, it did not impose any special conditions. Thus, Williams would be returned to prison after having served the mandatory minimum only if he committed another crime or possessed illegal controlled substances,
id.,
or if he violated the standard conditions of supervised release. The standard conditions do include prohibitions on purchase, possession, use, distribution, or administration of controlled substances and on the frequenting of places where controlled substances are illegally sold, used, distributed, or administered.
See
U.S.S.G. § 5B1.4(a)(7) & (8). But each of these conditions demands detection. An unrehabilitated Williams, who in the past has violated parole, would be free to resume his criminal ways after five years as long as he were able to avoid getting caught. If Williams fails to rehabilitate himself, nothing warrants his being sentenced to a lesser term than that prescribed by the Commission for similar crimes committed by defendants with similar criminal histories.
Cf. Merritt,
988 F.2d at 1307 (disparities in sentencing are permissible only if warranted by the circumstances of the case).
The risk of an unwarranted sentencing disparity would be allayed if the district court were to add two special conditions: (1) at the end of Williams’s term of incarceration and at the beginning of his term of supervised release, Williams must present to his probation officer certification from a drug treatment program at his place of incarceration that he has entered an available program at the first opportunity and remained in this program until the earlier of his release from confinement or the completion of the program, and that he is currently drug-free, and (2) Williams must submit to drug testing throughout the period of his supervised release and, if the United States Probation Office directs, he must continue to participate in an approved drug treatment program,
see
U.S.S.G. § 5B1.4(23). Although the first condition may appear unorthodox since it makes a condition of supervised release dependent on acts the defendant performs while in prison, we know of no law prohibiting a condition that ultimately bears upon his proclivity to use drugs after release. In our view, these special conditions are permissible under § 3583(d)(1) since they are “reasonably related” to the rehabilitation goal of § 3553(a)(2)(D) and the public-protection goal of § 3553(a)(2)(C). They are also consistent with the Commission’s directive that special conditions of supervised release be related to,
inter alia,
the offender’s need for effective medical or other correctional treatment. U.S.S.G. § 5D1.3(b). If such conditions are added, we would deem the sentence reasonable under 18 U.S.C. § 3742(e)(3).
As a final matter, in order to say that the sentence imposed by the district court would not be unreasonable if special conditions of supervised release were added, we must address the permissibility of the ten-year term of supervised release. Under 18 U.S.C. § 3583(b)(1) (and U.S.S.G. § 5D1.2(1)), the maximum authorized term of supervised release for Class A or B felonies is five years, except as provided otherwise by statute. A Class B felony is one in which the maximum term of imprisonment is twenty-five years or more (but not life). 18 U.S.C. § 3559(a)(2). The maximum term of imprisonment for violations of 21 U.S.C. § 841(a) involving 5 or more, but less than 50, grams of a substance with cocaine base (Williams pled guilty to two counts involving 25 and 26 grams respectively) is forty years if death or serious bodily injury does not result from the use of such substance. 21 U.S.C. § 841(b)(l)(B)(iii). Since Williams pled guilty to two counts of Class B felonies, the maximum term of supervised release for each count would be five years unless another statute provides otherwise.
We hold that § 3583(b)(1) does not apply to an offense under 21 U.S.C. § 841(b)(1)(B). The latter statute sets a minimum term of supervised release of “at least 4 years” for the offense of which Williams was convicted, if there has been no prior narcotics felony conviction.
In
United States v. Eng,
14 F.3d 165 (2d Cir.),
cert. denied,
— U.S. -, 115 S.Ct. 54, 130 L.Ed.2d 13 (1994), we held that a provision of § 841(b)(1)(A), which set a minimum term of supervised release of “at least 5-years” for a first-time offender, was not subject to the five year maximum of § 3583(b)(1) since such a construction would rob the term “at least” of any meaning.
Id.
at 172-73. Similarly, it would be strange indeed for Congress to have limited a district court’s discretion by requiring it to impose a supervised-release term of between four and five years for an offense under 21 U.S.C. § 841(b)(1)(B). A more natural reading of that statute, which was part of the act that amended § 3583(b) by adding the introductory phrase “Except as otherwise provided,”
Eng,
14 F.3d at 172-73, is that “Congress intended to enhance the penalties available to combat drug offenses” and thus overrode the generally applicable supervised-release máxi-mums of § 3583(b),
id.
at 173. A term of supervised release of ten years is therefore not impermissible in this case.
In summary, we wish to emphasize that the Sentencing Guidelines do not dis
place the “traditional role of a district judge in bringing compassion and common sense to the sentencing process.”
United States v. Rogers,
972 F.2d 489, 492 (2d Cir.1992). In areas where the Sentencing Commission has not spoken — areas outside the “heartland” cases considered in the Guidelines — district courts should not hesitate to use their discretion in devising sentences that provide individualized justice, provided they give reasons for their departure that allow us to ensure that the departure was permissible under the law.
While we agree with the district court that it had authority to depart downward and to impose a supervised-release term of ten years, for the reasons set forth above, we vacate the sentence and remand for resen-tencing in light of this opinion.