OPINION OF THE COURT
Maria Mendes appeals from the district court’s entry of judgment against her on her claim for Social Security disability benefits. The only question is whether Mendes proved that she had a severe impairment before December 31, 1990, which was her last insured date. Mendes offered laboratory evidence in the form of a 1992 MRI report that she had two ruptured disks and she offered other evidence that she suffered low back pain before the insurance cut-off date. We conclude that Mendes adduced evidence that her impairment began before December 31, 1990 and that the administrative law judge (hereafter referred to as the ALJ) erred in finding she had not done so. We will reverse with directions to remand to the Commissioner of Social Security for further proceedings.
Mendes applied for Title II Disability Insurance benefits on November 20, 1997, alleging that she became disabled on May 10, 1985. She quit work in 1985 when the factory where she worked closed. She testified that at the time she quit working, “I was not feeling very well already.... I had the problem with my back and my legs were getting [inaudible] I couldn’t move my legs and my arms.” Mendes’s treating physician, Dr. Rodolfo Colaco, had records of visits with complaints of skeletal pain beginning in 1987. Dr. Colaco’s first note about Mendes’s back is dated September 13, 1989 and states: “Low back pain with radiating left foot.” Dr. Colaco recorded no clinical findings, signs, or diagnosis at that time. The first objective evidence of Mendes’s ruptured disks came from an MRI in June 1992, which showed two herniated disks in her lower back.
After an initial hearing, the ALJ concluded that there was insufficient evidence of any restrictions to Mendes’s residual functional capacity before 1990, so she scheduled a supplemental hearing to adduce evidence on that issue. Dr. Albert G. Mylod submitted an interrogatory on that subject, but before the date of the hearing, *349Dr. Mylod became unavailable for medical reasons.1 Therefore, the ALJ appointed a different medical expert, Dr. Harlan Melik.
After hearing Dr. Melik’s testimony, the ALJ concluded that there was no medical evidence of an objective nature to establish that Mendes had a severe impairment before her last insured date. The ALJ therefore denied benefits, and the Appeals Council denied review, making the denial the Commissioner’s final decision.
Mendes filed this suit in the district court seeking review of the Commissioner’s decision, and the district court entered judgment for the Commissioner. The district court had jurisdiction under 42 U.S.C. § 405(g) (2000) to review the Commissioner’s determination to deny benefits, and our jurisdiction arises under 28 U.S.C. § 1291 (2000). We review the district court’s judgment de novo. Newell v. Com’r of Soc. Sec., 347 F.3d 541, 545 (3d Cir.2003). Section 405(g) provides that the Commissioner’s findings of fact are conclusive if supported by substantial evidence. Substantial evidence is such evidence as a reasonable mind would accept as adequate to support a conclusion. Newell, 347 F.3d at 545. We exercise plenary review over questions of law. Id.
Under 20 C.F.R. § 404.1520 (2003) (as revised by 68 F.R. 51153 (Aug. 26, 2003)), evaluation of disability proceeds in a five-step sequence. At the first step, the Commissioner asks whether the claimant is still working, § 404.1520(a)(4)(i); if so, the claimant is not disabled. Second, the Commissioner determines whether the claimant has a severe impairment — in other words, an impairment which significantly limits his or her physical or mental ability to do basic work activities. § 404.1520(a)(4)(ii). If not, the claimant is not disabled. If so, the Commissioner proceeds to the third step, determining whether the claimant’s condition is the same or equal to the listing of impairments in Appendix 1 to Subpart P to Part 404. § 404.1520(a)(4)(iii). If so, the claimant is disabled. If the claimant’s condition does not meet or equal the listings, the Commissioner asks whether the claimant’s residual functional capacity allows the claimant to do the kind of work the claimant has done in the past. § 404.1520(a)(4)(iv). If so, the claimant is not disabled. If not, the Commissioner will then consider whether the claimant’s residual functional capacity and other characteristics would allow the claimant to do other work. § 404.1520(a)(4)(v). If so, the claimant is not disabled; if not, the claimant is disabled. The burden of proof is on the claimant through the first four steps, but shifts to the Commissioner if the process reaches the fifth step. Newell, 347 F.3d at 546.
The ALJ found that Mendes would meet the listings based on her current condition, but that she failed to prove the existence of a severe impairment before December 31, 1990. A “severe impairment” is a threshold test and it is not a demanding one. See generally Newell, 347 F.3d at 545-46. “[Bjecause step two is to be rarely utilized as basis for the denial of benefits, its invocation is certain to raise a judicial eyebrow.” McCrea v. Com’r of Soc. Sec., 370 F.3d 357, 361 (3d Cir.2004) (citation omitted). “An impairment or combination of impairments is not severe if *350it does not significantly limit your physical or mental ability to do basic work activities.” 20 C.F.R. § 404.1521(a). Examples of basic work activities include “[p]hysical functions such as walking, standing, sitting, lifting, pushing, pulling, reaching, carrying, or handling.” § 404.1521(b)(1).
Generally, an impairment must be demonstrable by medically acceptable clinical and laboratory diagnostic techniques. 42 U.S.C. § 423(d)(5)(A) (2000); 20 C.F.R. §§ 404.1527(a), 404.1529(b). In this case, there is no dispute that Mendes has a current impairment which was established by medically acceptable clinical and laboratory evidence, including the MRI that showed the two ruptured disks.
Once an impairment has been established by suitable medical evidence, the onset date of the impairment may be established by evidence other than clinical and laboratory evidence. See Newell, 347 F.3d at 548 (In determining onset of established impairment, “[l]ay evidence need not be corroborated by contemporaneous medical evidence to be credible.”); Grebenick v. Chater, 121 F.3d 1193, 1199 (8th Cir.1997) (“Once the diagnosis is established, but the severity of the degenerative condition during the relevant period is unanswered, the claimant may fill the evidentiary gap with lay testimony. The ALJ must consider this evidence, even if it is uncorroborated by objective medical evidence.” (internal citations omitted)). The standards governing proof of the onset of impairments are set out in Social Security Ruling 83-20, “Titles II and XVI: Onset of Disability.”2 This policy statement provides that in disabilities of nontraumatic origin, “the determination of onset involves consideration of the applicant’s allegations, work history, if any, and the medical and other evidence concerning impairment severity.” The starting point in determining the onset date is the claimant’s statement.
With slowly progressive impairments, it is sometimes impossible to obtain medical evidence establishing the precise date an impairment became disabling. Determining the proper onset date is particularly difficult, when, for example, the alleged onset and the date last worked are far in the past and adequate medical records are not available. In such cases, it will be necessary to infer the onset date from the medical and other evidence that describe the history and symptomatology of the disease process.
In determining the date of onset of disability, the date alleged by the individual should be used if it is consistent with all the evidence available....
In some cases, it may be possible, based on the medical evidence to reasonably infer that the onset of a disabling impairment(s) occurred some time prior to the date of the first recorded medical examination, e.g., the date the claimant stopped working. How long the disease may be determined to have existed at a disabling level of severity depends on an informed judgment of the facts in the particular case. This judgment, however, must have a legitimate medical basis —
If reasonable inferences about the progression of the impairment cannot be made on the basis of the evidence in the file and additional relevant medical evidence is not available, it may be necessary to explore other sources of documentation. Information may be obtained from family members, friends, and former employers to ascertain why *351medical evidence is not available for the pertinent period and to furnish additional evidence regarding the course of the individual’s condition.... The impact of lay evidence on the decision of onset will be limited to the degree it is not contrary to the medical evidence of record.
S.S.R. 83-20.
Here, of course, Mendes’s own testimony is that her impairment began in 1985 when she quit working and when she began to have trouble using her legs. Mendes’s treating physician, Dr. Colaco, submitted a letter stating: “Mrs. Maria Mendes has been having excruciating back pain since the late eighties. She was seen by me for the same on 3/16/87 and was placed on Feldene and advised to be followed by Orthopedic Surgeon [or] a Chiroprator [sic]. On 4/13/89 she was seen with radiculopathy to the left foot.” Mendes’s attorney introduced correspondence from a chiropractor in 1999 that stated:
[A]s we have informed your office we do not have the records on the patient, since she has been inactive for more than ten years. We were able to find a new patient logbook where it showed that Mrs. Mendes was first examined on June 13, 1987 for a low back condition. She was under treatment for approximately three months.
The crucial evidence on which the ALJ based her finding was from the medical expert, Dr. Melik. First the ALJ asked him if he had an opinion as to whether Mendes’s impairments met or equalled the listings as of the end of 1990. He responded:
I cannot tell, your Honor. There’s no objective data even close to that date in the file. And the earliest data that I can find, I believe, is the MRI of June 22nd, 1992, which shows central and right herniated nucleus propulses L3, L4 and central and right herniated nucleus pro-pulses L5, SI.
The ALJ then asked: “Based on SSI 83, 20, [sic] I’m required to ask whether, to a reasonable degree of medical certainty, you can relate that back to the date of alleged onset.” Dr. Melik replied:
Well, I, I think the answer to that is yes. Just based upon the client’s — claimant’s testimony, although, that would be the, the most compelling reason. I have no reason to disbelieve the history as reported by the patient. But I would not know — it’s compatible with the, the testimony. I would not know when this occurred, though.
(emphasis added). After reviewing Dr. Colaco’s notes, Dr. Melik returned to the onset question: “[I]n answer to the — to the very first question you posed, I think it’s reasonable to say that the herniated disk [sic] were, were present there, but one can’t say for certain, but she certainly had back pain and some radiation at that time [apparently referring to Dr. Colaco’s notes of September 13, 1989].” The ALJ then asked Dr. Melik to render an opinion on Mendes’s limitations “at that time.” He said:
Based upon what I read there and giving you the worst case scenario type of, of focus, she would be able to sit six to eight hours, now she may have to shift around; stand and or walk two to three hours in total, that’s in a six to eight hour day; lift and or carry 10 pounds frequently or occasionally; would have problems with repetitive bending, stooping, climbing, use of foot pedals. At that time, she’d have no problem with fine or gross manipulation.
The ALJ’s discussion of Dr. Melik’s testimony stated in relevant part: “He opined that it was reasonable to conclude that the *352claimant had some back pain. He further noted that while the MRI in 1992 demonstrated herniated discs at L3-L4 and L5-Sl, he could not state when this occurred.” She said that Colaco’s records show that “Complaints of low back pain began in 1989,” but that Colaco did not record any evidence of an objective nature, such as range of motion testing, strength testing, etc. She dismissed the chiropractor’s evidence because it did not qualify as medical evidence. On this record, the ALJ found there was “no evidence to support a finding of a severe impairment prior to the date the claimant was last insured.” She stated, “I am constrained to find that the claimant was <not disabled’ prior to the expiration of her last insured status, December 31, 1990. This is most unfortunate as, the record clearly demonstrates that at the present time Ms. Mendes’s condition meets the requirements for disability.”
To the extent the ALJ’s decision was based on a legal determination that the onset date of an established impairment had to be proved by medical evidence of “an objective nature,” it is erroneous. S.S.R. 83-20 allows the use of lay testimony and medical evidence describing history and symptomatology to link an impairment to a date earlier than the first diagnosis documented by laboratory findings. See Newell, 347 F.3d at 547 (“Retrospective diagnosis of an impairment, even if uncorroborated by contemporaneous medical records, but corroborated by lay evidence relating back to the claimed period of disability, can support a finding of past impairment.”). In a comparable case, we reversed the Commissioner’s denial of benefits, stating (as one of several grounds) that the ALJ could not reject the opinion of a treating physician “solely on the basis that his opinion was based on information supplied by [the claimant].” Walton v. Halter, 243 F.3d 703, 710 (3d Cir.2001).
Our review of the record shows that there is evidence of the sort contemplated by S.S.R. 83-20 that the impairment documented in 1992 began before the end of 1990. There was not only Mendes’s testimony, but the records of her treating physician and her chiropractor that Mendes sought treatment for lower back pain before 1990. The ALJ did not reject Mendes’s or her doctors’ testimony on credibility grounds.
Mendes did not supply an explicit diagnosis linking her pain in the' eighties to the ruptured disks. Dr. Colaco only stated that Mendes had excruciating back pain since the late eighties; he did not explicitly diagnose the pain as resulting from the ruptured disks or say that it created any particular limitations. We have no need to decide whether this lack of a formal diagnosis relating the pain to the subsequent laboratory findings would result in a failure of proof, because the missing links were supplied by Dr. Melik, the medical expert called by the ALJ.
Dr. Melik said he had no reason to doubt Ms. Mendes’s account and he referred to Dr. Colaco’s notes as showing that there was back pain and radiation, apparently in 1989. When the ALJ asked him if he could relate the ruptured disks back to the alleged date of onset (which was 1985), he said: ‘Well, I, I think the answer to that is yes.” He later said it was “reasonable” to say the disks were herniated at the relevant time. Social Security Ruling 83-20 permits reasonable inferences. Finally, when asked about Mendes’s limitations “at that time,” which appears to refer to the alleged date of onset, Dr. Melik described a residual capacity which presumed certain limitations and explicitly stated others. For instance, he described her as having the ability to lift ten pounds, implicitly negating an ability to lift more, and he stated that she *353would have trouble with repetitive bending, stooping, climbing and use of foot pedals. An impairment that resulted in such restrictions would significantly limit Mendes’s work activities, thus satisfying step two of the sequential analysis. See 20 C.F.R § 404.1521. By way of comparison, we recently reversed a decision denying benefits at step two where the record showed that the claimant’s condition prevented her from sitting more than two hours a day, carrying objects weighing more than twenty pounds at any time, and carrying objects weighing more than ten pounds for more than two hours a day. See McCrea, 370 F.3d at 361.
The ALJ’s finding that there was no evidence of severe impairment before December 31, 1990 was not supported by substantial evidence. The Commissioner responds, “[E]ven if Mendes had established that she had herniated lumbar discs prior to the expiration of her insured status, that fact alone would not entitle her to disability insurance benefits. Rather, Mendes must show that her impairment resulted in functional limitations that precluded all substantial gainful activity.” This statement is correct as a matter of the ultimate award of benefits, but it ignores the fact that the ALJ found that Mendes failed to prove severe impairment, and therefore the analysis ceased at step two, without ever reaching the question of whether Mendes’s condition precluded performance of her past work or all substantial gainful activity. We will reverse and remand to the district court with instructions to remand to the Commissioner for further proceedings to ascertain whether the claimant was disabled at steps three, four or five.