William E. Butts appeals from Judge Kahn's order remanding Butts' claim for disability benefits to the Social Security Commissioner. The remand directs further proceedings pursuant to "sentence four" of 42 U.S.C. § 405(g) to determine whether adequate jobs exist in the national economy that Butts could perform. Butts asks us to remand only for a calculation of benefits because the Commissioner failed to sustain her burden of showing through the testimony of a vocational expert that there are jobs in the national economy that Butts could perform. The Commissioner agrees that a vocational expert was needed but argues that remand for an evidentiary hearing was within the district court's discretion. We hold that a district court's "sentence-four-remand order" is subject to review for abuse of discretion, that the district court did not abuse its discretion, but that any further proceedings must be completed within a time limit.

Butts is a high-school graduate who worked as an ironworker for a steel fabrication company from 1971 until September 30, 1991. At that time he suffered an injury to his back while lifting a heavy piece of steel on the job. Butts has since been treated for back and back-related problems by a host of doctors and has been diagnosed with a number of back conditions, including degenerative disk disease, congenital spinal stenosis, mild displacement of nerve root from disk bulging, moderate thecal sac compression, mild-to-moderate neural foraminal stenosis, mild intermittent radiculitis, and chronic lumbar syndrome and radiculopathy. Butts' problems began to subside in 1993, and a scheduled surgery was cancelled. However, Butts re-injured himself in 1995, and his condition again worsened. Butts' most recent physician, Dr. Kamlesh S. Desai, diagnosed Butts with chronic lumbar syndrome, intermittent left L-4 radiculopathy, and disc degeneration at L5-S1 and L4-5 with facet arthropathy.

Butts' job as an ironworker involved lifting up to 100 pounds on a regular basis. His doctors now agree that he can lift no more than half that weight and only from time to time. Based on his diagnoses in 1997 and 1998, Dr. Desai also placed Butts on permanent restrictions, directing him to "[a]void heavier bending, lifting, and twisting activities, prolonged sitting and prolonged driving, as well as prolonged standing in one position."

Butts filed his first application for Social Security disability benefits in May, 1996. The application was denied both upon initial determination and upon reconsideration. Butts filed a request for an administrative hearing in connection with the denial of his claim but later withdrew that request. On January 1, 1998, Butts filed an amended application for disability insurance benefits. This application was also denied both upon initial determination and upon reconsideration. Butts again requested a hearing before an administrative law judge ("ALJ"). At the hearing on January 6, 1999, Butts, then 49 years old, testified that he has trouble standing and sitting for long periods of time and experiences sharp pains while bending and lifting objects. The ALJ denied Butts' claim.

In his decision, the ALJ undertook the mandatory five-step sequential evaluation process for evaluating disability determinations. See 20 C.F.R. § 416.920. The ALJ first found that Butts had not engaged in substantial gainful employment since the onset of his impairment. Second, after reviewing the medical evidence, the ALJ concluded that Butts'"back condition has resulted in a significant restriction of his ability to perform basic work activities" therefore constituting a "`severe' impairment." Third, the ALJ determined that Butts' impairment did not qualify under the Social Security Act's "Listing of Impairments" in Appendix 1 to Subpart P of Section 404 of the Act. Had the ALJ found otherwise, Butts would have been deemed disabled without reference to vocational factors such as age, education, or work experience.

However, because Butts' impairment did not qualify as a listed impairment under Appendix 1, the ALJ moved to step four of the sequential evaluation to determine Butts' residual functional capacity for his work as an ironworker. Residual functional capacity involves the range of work activities Butts could still perform despite his impairment. Social Security regulations direct the ALJ to consider both objective medical evidence and any other evidence, including statements and reports from Butts and his physicians, relevant to how his impairments and related symptoms affect his ability to work. See 20 C.F.R. § 404.1529. The ALJ reviewed Butts' testimony, the diagnoses and treatment recommendations of his physicians, and the opinion of a state agency review physician. The state agency physician, while not an examining physician, evaluated the evidence and opined that Butts could frequently lift ten pounds, could sit and/or stand for a total of six hours in an eight hour work day, could occasionally climb and balance, but could never stoop, kneel, crouch or crawl.

The ALJ concluded that although Butts could no longer perform his prior heavy lifting work—which includes "exertional" activities—he retained the residual functional capacity for the full range of light work—which includes "nonexertional" activities such as bending, kneeling, crouching, and crawling—as well as for sedentary work.1 In reaching this conclusion the ALJ relied principally upon the state agency review physician's opinion which it found to be supported by the opinions of Butts' treating physicians. Based on Butts' residual functional capacity to perform at most light work, the ALJ determined at step four of the sequential evaluation that Butts was incapable of returning to his former occupation as an ironworker.

At step five, the burden shifts to the Commissioner to show that there were a significant number of jobs in the national economy that Butts could perform based on his residual functional capacity, age, education, and prior vocational experience. See 20 C.F.R. § 404.1560. However, the Commissioner is not represented at an administrative hearing; rather, the ALJ marshals the evidence and makes the relevant determination. See Seavey v. Barnhart, 276 F.3d 1, 8 (1st Cir.2001) (citing Sims v. Apfel, 530 U.S. 103, 110, 120 S. Ct. 2080, 147 L. Ed. 2d 80 (2000)). In Butts' case, the ALJ used criteria from the Social Security Act's table of medical-vocational guidelines, referred to as "grid rules" or "grids," to conclude that Butts was capable of performing other jobs existing in significant numbers in the national economy and therefore did not meet the requirements for disability status. The ALJ did not call a vocational expert to testify as to Butts' ability to perform other jobs but relied exclusively on the grid rules. Butts filed a request for review of the ALJ's decision. On December 8, 2000, the Appeals Council denied Butts' request for review, thus rendering the ALJ's decision the final decision of the Commissioner of Social Security in Butts' case.

On January 24, 2001, Butts commenced the present action. Magistrate Judge Homer issued a report and recommendation on December 2, 2002, finding that according to Butts' physicians, he was capable of returning to work, but not to his prior heavy work. Specifically, the magistrate judge found that Butts was limited to jobs where he did not have to bend, twist, stoop, kneel, crouch, crawl or lift heavy objects, and where he could alternate between sitting or standing and walking. On this basis, the magistrate judge held that substantial evidence supported the Commissioner's determination that Butts was able to perform the full range of light and sedentary work. However, because Butts has non-exertional impairments that limit his exertional capabilities, the magistrate judge found that the ALJ erred by relying on the grids and concluded that a vocational expert should have been utilized to determine whether jobs exist in the national economy that Butts is capable of performing. The magistrate judge then recommended a remand for further consideration.

The district court adopted the magistrate judge's recommendation. The district court found that while Butts could perform sedentary work, the medical evidence of his limitations such as his inability to bend, twist, stoop, kneel, etc., and the need for him to alternate between sitting and standing, did not support a finding that he could perform the full range of light work.2 The district court took particular issue with the ALJ's selective reliance on physicians' reports from only the period before Butts re-injured himself, with the ALJ's incomplete and cursory consideration of more recent treating physicians' reports, and with the ALJ's reliance on the state agency reviewing physician's opinion, which the district court found to be internally inconsistent. The district court agreed with the magistrate judge that the ALJ's application of the grids was inappropriate, and that "given the uncertain state of the record with regard to the availability of jobs in the national economy specific to Butts' limitations ... remand and not reversal is warranted."3 Accordingly, the district court vacated the Commissioner's decision and remanded the action to the Social Security Commission pursuant to sentence four of 42 U.S.C. § 405(g), which provides that "[t]he court shall have power to enter, upon the pleadings and transcript of the record, a judgment affirming, modifying, or reversing the decision of the Commissioner of Social Security, with or without remanding the cause for a rehearing."

This appeal followed.

On appeal, neither the Commissioner nor Butts challenge the district court's ruling that Butts cannot perform the full range of light work and may be "disabled" within the meaning of the Act depending on whether work is available in the national economy for one with his limitations. The parties further agree that whether Butts is disabled is unknown because the Commissioner failed to consult a vocational expert at step five of the hearing process. The parties also agree that the Commissioner bore the burden of showing that Butts was not disabled at step five.

The sole issue on appeal therefore is whether the proper remedy is a remand for an evidentiary hearing or for a benefit calculation. Butts does not claim that he is disabled as a matter of law; rather, he argues that because there was no testimony by a vocational expert, the Commissioner failed to prove that he was not disabled and, therefore, he is entitled to benefits as a matter of law. The Commissioner agrees that she failed to meet her burden of proving that Butts was disabled but argues that the district court's decision to remand for an evidentiary hearing was proper.

a) The Five Step Procedure

The Social Security Act defines "disability" in pertinent part as the "inability to engage in any substantial gainful activity by reason of any medically determinable physical or mental impairment which can be expected to result in death or which has lasted or can be expected to last for a continuous period of not less than 12 months." 42 U.S.C. § 423(d)(1)(A). Further, the impairment must be "of such severity that [the claimant] is not only unable to do his previous work but cannot, considering his age, education, and work experience, engage in any other kind of substantial gainful work which exists in the national economy." Id. § 423(d)(2)(A). The Social Security Administration has a five-step procedure for evaluating disability claims. See 20 C.F.R. § 404.1520. The claimant bears the burden of proving his or her case at steps one through four; the parties agree that Butts has satisfied this burden; the only remaining issue is the outcome of step five.

At step five the burden shifts to the Commissioner to "show there is other gainful work in the national economy [which] the claimant could perform." Balsamo v. Chater, 142 F.3d 75, 80 (2d Cir. 1998) (internal quotation marks and citation omitted). See also Curry v. Apfel, 209 F.3d 117, 123 (2d Cir.2000) ("Once a disability claimant proves that his severe impairment prevents him from performing his past work, the [Commissioner] then has the burden of proving that the claimant still retains a residual functional capacity to perform alternative substantial gainful work which exists in the national economy.") (internal quotation marks and citation omitted; alteration in original). "In the ordinary case, the Commissioner meets his burden at the fifth step by resorting to the applicable medical vocational guidelines (the grids)." Rosa v. Callahan, 168 F.3d 72, 78 (2d Cir.1999) (internal quotation marks and citations omitted). However, "exclusive reliance on the grids is inappropriate where the guidelines fail to describe the full extent of a claimant's physical limitations." Id.

In particular, sole reliance on the [g]rid[s] may be precluded where the claimant's exertional impairments are compounded by significant nonexertional impairments that limit the range of sedentary work that the claimant can perform. In these circumstances, the Commissioner must introduce the testimony of a vocational expert (or other similar evidence) that jobs exist in the economy which claimant can obtain and perform.

Id. (internal quotation marks and citations omitted; alterations in original). It is undisputed that Butts has both exertional and nonexertional limitations. However, the ALJ relied exclusively upon the grid rules to conclude that jobs exist in the national economy that Butts is capable of performing. It is clear under our cases that the Commissioner could have satisfied her burden at step five through a vocational expert's testimony that Butts could engage in other work existing in the national economy. Therefore, as the parties now agree, the ALJ should have called a vocational expert in step five but did not.

b) Standard of Review

In reviewing a final decision of the Commissioner, a district court must determine whether the correct legal standards were applied and whether substantial evidence supports the decision. Machadio v. Apfel, 276 F.3d 103, 108 (2d Cir.2002). Sentence four of 42 U.S.C. § 405(g) provides that, after reviewing the Commissioner's determination, a court may:

enter, upon the pleadings and transcript of the record, a judgment affirming, modifying, or reversing the decision of the Commissioner of Social Security, with or without remanding the cause for a rehearing.4

When reviewing the district court's determination as to the final decision of the Commissioner "[w]e review the administrative record de novo," using the same standard applied by the district court. See Machadio, 276 F.3d at 108; Shaw v. Chater, 221 F.3d 126, 131 (2d Cir.2000); Williams v. Apfel, 204 F.3d 48, 49 (2d Cir.1999). However, the proper standard of review regarding a district court's determination of whether to remand for further proceedings or for a calculation of benefits is an issue of first impression in this circuit.5 Three circuits have ruled that the appropriate standard is abuse of discretion, Nelson v. Apfel, 210 F.3d 799, 802 (7th Cir.2000); Harman v. Apfel, 211 F.3d 1172, 1173 (9th Cir.2000); Higgins v. Apfel, 222 F.3d 504, 505 (8th Cir.2000), while one has reviewed such decisions de novo, Seavey v. Barnhart, 276 F.3d 1, 9 (1st Cir.2001).

We agree with the Seventh, Eighth, and Ninth Circuits that the proper standard of review is abuse of discretion. We rely principally on the language of Section 405(g) in reaching our conclusion.6 See Cmty. for Creative Non-Violence v. Reid, 490 U.S. 730, 739, 109 S. Ct. 2166, 104 L. Ed. 2d 811 (1989) ("The starting point for our interpretation of a statute is always its language."). Sentence four of Section 405(g) provides district courts with the authority to affirm, reverse, or modify a decision of the Commissioner "with or without remanding the cause for a rehearing." 42 U.S.C. § 405(g). This language clearly augments the authority of the district court by granting it an additional power as to the order of remand. See Harman, 211 F.3d at 1178 ("Without this additional authority, a district court could not remand a case for immediate payment of benefits but could only remand the case for rehearing."); Nelson, 210 F.3d at 801-02 (Sentence four "empowers" district courts to reverse or modify without remanding). "Because Congress did not state that district courts `shall' exercise this additional power but simply gave district courts the authority to do so in an appropriate case, it reasonably may be inferred that the district court's exercise of such authority was intended to be discretionary and should be reviewed for abuse of discretion." Harman, 211 F.3d at 1178. We agree with this analysis and review the district court's order of remand for abuse of discretion.

c) The Remand

We now must determine whether the district court abused its discretion by remanding Butts' claim for further proceedings rather than for a calculation of benefits. In deciding whether a remand is the proper remedy, we have stated that where the administrative record contains gaps, remand to the Commissioner for further development of the evidence is appropriate. See Rosa, 168 F.3d at 82-83. That is, when "further findings would so plainly help to assure the proper disposition of [the] claim, we believe that remand is particularly appropriate." Id. at 83 (internal quotation marks omitted). On the other hand, "where this Court has had no apparent basis to conclude that a more complete record might support the Commissioner's decision, we have opted simply to remand for a calculation of benefits." Id.

For instance, in Curry the claimant was 44 years old with a high school education and had been working as a plumber before his injury. 209 F.3d at 119. In light of back and knee problems and severe pain, his doctors stated that he could no longer work as a plumber and that during an eight hour day he could sit for no more than two or three hours, stand a total of only one hour, and walk a total of 30 minutes. Id. at 121. After rejecting this evidence as not credible, the ALJ concluded that there was no evidence that the claimant could not engage in sedentary work and denied disability status. Id. The district court affirmed this decision, but we reversed and remanded for calculation of benefits. Id. at 124. We held that because the Commissioner had been unable to prove that Curry could work, "no purpose would be served by our remanding the case for rehearing" and "remand for the sole purpose of calculating an award of benefits [was] mandated." Id. (internal quotation marks and citation omitted); see also Rosa, 168 F.3d at 83 (remand for benefit calculation appropriate where there was no "basis to conclude that a more complete record might support the Commissioner's decision" that the claimant would be classified as able to work); Balsamo v. Chater, 142 F.3d 75, 82 (2d Cir. 1998); Carroll v. Sec'y of Health and Human Servs., 705 F.2d 638, 644 (2d Cir. 1983).

In contrast, in cases where "the ALJ failed to develop the record sufficiently to make" appropriate disability determinations, a remand for "further findings [that] would so plainly help to assure the proper disposition of [the] claim ... is particularly appropriate." Rosa, 168 F.3d at 83 (internal quotation marks omitted). In Rosa, the ALJ failed to obtain adequate information from the claimant's treating physician and a number of other doctors and treatment facilities as to extent of the claimant's injuries and consequently reached unsupported conclusions. Id. We vacated the district court's affirmance of the ALJ's denial of benefits and directed a remand to the Commissioner for further proceedings. Id.

Moreover, while it is true that the Commissioner bears the burden at step five, "the ALJ, unlike a judge in a trial, must [her]self affirmatively develop the record in light of the essentially non-adversarial nature of a benefits proceeding." Pratts v. Chater, 94 F.3d 34, 37 (2d Cir. 1996) (internal quotation marks omitted; alteration in original). Because Social Security disability determinations are nonadversarial, "the Commissioner is not a litigant and has no representative at the agency level. Indeed, the model is investigatory, or inquisitorial, rather than adversarial." Seavey, 276 F.3d at 8. Thus, "[i]t is the ALJ's duty to investigate and develop the facts and develop the arguments both for and against the granting of benefits." Id. The ALJ therefore owes a duty to the Commissioner as well as to the claimant.

Unlike Curry, where "no purpose would be served by our remanding," 209 F.3d at 124 (internal quotation marks omitted), the evidence in the instant case is sufficient to support a finding that Butts could perform sedentary and perhaps some light work. However, the ALJ failed to develop the record as to whether Butts' nonexertional limitations preclude him from performing other work in the national economy. Like the ALJ in Rosa, the ALJ at Butts' hearing "failed to fulfill her duty," 168 F.3d at 83 (internal quotation marks omitted)—in this case, by neglecting to consult a vocational expert. Because the ALJ failed to call a vocational expert, the record is incomplete and "further findings" are appropriate "to assure the proper disposition of [the] claim." Id. (internal quotation marks omitted). The district court's remand for further proceedings was therefore within its discretion.

In so holding, we note two caveats. First, a remand is proper where the error is found in an ALJ's failure to apply correctly the distinction between cases where reliance on the grid suffices and those where the testimony of a vocational expert is essential to a denial of benefits. See Rosa, 168 F.3d at 78. However, were a claimant to show that Social Security ALJs reject this distinction as a matter of policy and rely solely on the grid in all or most similar proceedings, other remedies, including judgment for the claimant, should be considered. Cf. Schisler v. Heckler, 787 F.2d 76, 82, 84 (2d Cir.1986) (SSA had "adopted a formal policy of non-acquiescence in response to legal rulings of Courts of Appeals with which it disagree[d]"; thus, even though SSA claimed to follow a certain rule, injunction ordering SSA to publish its acquiescence in that rule appropriate).

Also, we are mindful of the "often painfully slow process by which disability determinations are made," Carroll, 705 F.2d at 644, and that "a remand for further evidentiary proceedings (and the possibility of further appeal) could result in substantial, additional delay." Curry, 209 F.3d at 124. Because, as the Commissioner argues, a remand is within the discretion of a district court, the principles calling for some evaluation of relative hardships that govern a discretionary selection of alternative remedies apply, and the hardship to a claimant of further delay should be considered. For this reason, in cases involving an ALJ's failure to call a vocational expert, district courts that select remand as a remedy should consider imposing a time limit on the subsequent proceedings. In this case, the past delay is of such magnitude—years— that a time limit is imperative. We therefore instruct the district court to direct that further proceedings before an ALJ be completed within 60 days of the issuance of the district court's order and, if that decision is a denial of benefits, a final decision of the Commissioner be rendered within 60 days of Butts' appeal from the ALJ's decision. The district court's order should provide that, if these deadlines are not observed, a calculation of benefits owed Butts must be made immediately.

The district court's order of remand is affirmed but with instructions to impose the time limits stated.