Opinion by Judge O’SCANNLAIN; Partial Concurrence and Partial Dissent by Judge GRABER.
O’SCANNLAIN, Circuit Judge:We must decide whether the district court properly affirmed the Commissioner’s denial of Social Security Disability Insurance benefits in this case.
I
Carmickle alleges disability and seeks benefits under the Social Security Act, 42 U.S.C. § 423, due primarily to a back injury caused by a July 2001 on-the-job accident. In his fifties, Carmickle has an *1158llth-grade education with some college classes, and Ms past work experience includes construction worker, auto salesperson, construction supervisor, and counter salesperson. A few weeks before his accident, Carmickle was examined by Dr. Gan-janan Nilaver and reported a history of neck and back injuries and chronic back pain. Dr. Nilaver concluded that Carmick-le would benefit from physical therapy, but that he could perform sedentary activities.
After his work injury, Carmickle went to the emergency room complaining of lower back pain. He did not mention his work-related accident. Carmickle was diagnosed with left sciatica and mild lumbar degenerative joint disease, but lumbar disc disease was ruled out. A few months later, he was examined by Dr. Kevin Kane, an osteopath. Carmickle told Dr. Kane that he believed his onset of pain was caused by the work-related accident. Dr. Kane diagnosed moderate to severe degenerative disc disease in the lumbar region, but he did not believe it was related to the on-the-job injury. Dr. Kane noted preexisting degenerative lumbar spondylosis and disc disease, and he recommended physical therapy and provided a work excuse for two weeks. Dr. Kane noted Car-mickle “does have significant low back disability,” and “it is clear he did not tolerate return to sedentary or light work.”
A few weeks after Dr. Kane’s examination, Carmickle was examined by Dr. Victoria Carvalho. Carmickle again reported the work-related injury and complained of lower-back pain despite chiropractic treatment and physical therapy. Although he stated that he was unable to sit or stand for more than five to ten minutes, Dr. Carvalho noted that Carmickle sat in her office for at least half an hour. She diagnosed lumbosacral sprain secondary to the reported work injury and she prescribed Relafen and a moist heating pad. She also recommended neck and back exercises and gave Carmickle a two-week work excuse.
Carmickle filed a workers’ compensation claim with the State of Oregon based on the July 2001 accident. In November 2001, two physicians performed a joint independent medical evaluation in connection with this claim. Both doctors diagnosed multilevel lumbar degenerative disease and concluded that the lumbar strain was medically stationary and created “no permanent impairment.” The physicians also noted that, despite his claims of pain, Carmickle “s[at] comfortably during the interview portion of the examination” and “d[id] not appear to have any difficulty sitting or with any change of position.”
In December 2001, Dr. Michael Horowitz, an osteopath, recommended that Car-mickle use a “reclinable desk chair while working.”
Dr. Mark Patton, also an osteopath, treated Carmickle between December 2001 and November 2003. In December 2001, Dr. Patton diagnosed an L4-5 disc bulge with the possibility of spinal stenosis. He recommended walking or stretching every 45-60 minutes and “no prolonged sitting.” Shortly thereafter, Dr. Patton noted Car-mickle’s decreased sensation in the left foot and diagnosed peripheral neuropathy. Dr. Patton opined that Carmickle “would be better off in[a] job where he is up & moving rather than a sit down job.”
In January 2002, Carmickle had an MRI which showed no spinal stenosis or significant nerve root impairment. Dr. Patton summarized Carmickle’s condition, stating that the “only major functional limitation currently is the inability ... to sit for prolonged periods without developing numbness of his foot.” Several months later, he observed localized tenderness and swelling consistent with chronic muscle strain and recommended that Carmickle *1159have a Relafen injection. Carmickle declined.
A few months later, Dr. Patton again recommended an injection after noticing Carmickle’s peripheral extremities showed signs of chronic peripheral vascular disease and that Carmickle was still experiencing tenderness. Again, Carmickle declined. Dr. Patton reported that although Relafen “typically has not been covered” by Carmickle’s insurance, it “has been the only thing that has provided significant relief without addiction potential or intolerable side effects.” Dr. Patton further opined that Carmickle likely is incapable of returning to construction work due to his health problems, but a “retraining program is certainly a viable alternative.” Dr. Patton agreed with Dr. Horowitz’s recommendation that Carmickle use a reclina-ble chair for work.
In addition to his back impairments, Carmickle also alleges that he suffers from mental impairments. In September 2003, he complained of worsening memory loss, lethargy, and possible depression. Dr. Patton reported that Carmickle’s Zung profile was normal, but that he displayed a flat affect. At this same time, Dr. Patton diagnosed tendonitis after an x-ray of Car-mickle’s elbows showed bilateral bone spurs.
Carmickle filed his application for Disability Insurance benefits in May 2002. In November 2003, Carmickle had a hearing before an administrative law judge (“ALJ”). Carmickle was represented by counsel and he testified on his own behalf. Medical Expert (“ME”) Dr. William L. DeBolt, and Vocational Expert (“VE”) Patricia Ayerza also testified at the hearing. Lay witness Tom Tucker submitted a statement on Carmickle’s behalf. After the hearing, the ALJ concluded that Car-mickle does have severe impairments, but that he is not entitled to disability benefits because he retains the residual functional capacity (“RFC”) to perform his past relevant work as a construction supervisor and counter salesperson. The Appeals Council denied review and adopted the ALJ’s decision as the final decision of the Commissioner on August 26, 2004. Carmickle sought review in the district court pursuant to 42 U.S.C. § 405(g), and the district court affirmed the ALJ. This appeal followed.
II
“The district court properly affirms the Commissioner’s decision denying benefits if it is supported by substantial evidence and based on the application of correct legal standards.” Sandgathe v. Chater, 108 F.3d 978, 980 (9th Cir.1997) (per curiam). “Substantial evidence means more than a mere scintilla but less than a preponderance; it is such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.” Andrews v. Shalala, 53 F.3d 1035, 1039 (9th Cir.1995).
A claimant is entitled to Social Security disability benefits only if he suffers from a “medically determinable physical or mental impairment” that prevents him from performing his prior work activities and “any other substantial gainful employment that exists in the national economy.” Tackett v. Apfel, 180 F.3d 1094, 1098 (9th Cir.1999) (citing 42 U.S.C. § 423(d)(2)(A)). The Social Security Administration employs a five-step sequential analysis to determine disability. Id. Here, Carmickle contends the agency erred in finding him not disabled because the ALJ erroneously (a) rejected his testimony and the lay witness testimony as not fully credible, (b) rejected some of the medical evidence, and (c) classified his prior relevant work as a construction supervisor and counter salesperson. We address each issue in turn.
*1160A
The ALJ found Carmickle’s testimony “not entirely credible in light of the treatment record and his daily activities.” Carmickle had testified that he is in constant pain due to his back condition and that he cannot sit or stand for more than 15 minutes at a time. He stated that when sitting, he has to “change positions constantly.” He also stated that his lifting ability is significantly limited and that he can lift only between 10 and 20 pounds occasionally. The ALJ concluded Carmickle’s allegations regarding his condition are inconsistent with (1) his full-time college attendance, (2) his receipt of unemployment benefits, (3) his minimal pain treatment regime, and (4) Dr. Patton’s opinion that Carmickle can lift 10 pounds frequently and 20 pounds occasionally. The district court, addressing only the first and fourth inconsistencies, affirmed the ALJ’s adverse credibility assessment. In doing so, however, the district court concluded the “clear and convincing reasons” standard we established in Lester v. Chafer, 81 F.3d 821, 834 (9th Cir.1996), did not apply, but rather, that the ALJ need only “ ‘identify what testimony is not credible and what evidence undermines [Carmickle]’s complaints.’ ”
1
We have consistently held that where the record includes objective medical evidence establishing that the claimant suffers from an impairment that could reasonably produce the symptoms of which he complains, an adverse credibility finding must be based on “ ‘clear and convincing reasons.’ ” Lingenfelter v. Astrue, 504 F.3d 1028, 1036 (9th Cir.2007) (quoting Smolen v. Chater, 80 F.3d 1273, 1281 (9th Cir.1996)). The only time this standard does not apply is when there is affirmative evidence that the claimant is malingering. Greger v. Barnhart, 464 F.3d 968, 972 (9th Cir.2006); Morgan v. Comm’r of Soc. Sec. Admin., 169 F.3d 595, 599 (9th Cir.1999).1
Here, the ALJ did not find that Car-mickle is malingering and we see no affirmative evidence of such. Rather, the district court, relying on Swenson v. Sullivan, 876 F.2d 683, 687 (9th Cir.1989), concluded that Carmickle’s alleged symptoms are not “medically related” because he is not alleging that his symptoms “are somehow more disabling to him than to others.” As just stated, the clear-and-convincing standard applies only where objective medical evidence establishes that the claimant suffers from an impairment that could reasonably be expected to cause the symptoms of which he complains. Smolen, 80 F.3d at 1281. Once such an impairment is established, however, the claimant’s reasonably expected symptoms are deemed “medically related.” In Smolen, we explained:
The claimant need not produce objective medical evidence of the [symptom] itself, or the severity thereof. Nor must the claimant produce objective medical evidence of the causal relationship between the medically determinable impairment and the symptom. By requiring that the medical impairment “could reasonably be expected to produce” pain or another symptom, the Cotton [v. Bowen, 799 F.2d 1403 (9th Cir.1986)] test *1161requires only that the causal relationship be a reasonable inference, not a medically proven phenomenon.
Id. at 1282 (citations omitted).
In support of its decision that specific testimony from Carmickle was required before the clear-and-convincing standard would apply, the district court relied on our statement in Swenson that “[i]f a claimant submits objective medical findings of an impairment that would normally produce a given symptom, but testifies that he experiences the symptom to a greater degree than would normally be expected, the Secretary may disbelieve that but must make specific findings justifying his decision.” 876 F.2d at 687. This principle remains good law, but it does not apply here. Swenson addresses only how the ALJ must treat a claimant’s excess symptom testimony. Id. It does not establish a standard for assessing the claimant’s testimony generally. And it does not establish the standard for determining whether a claimant’s alleged symptoms are “medically related.” On that point, our Smolen line of cases controls.
Here, Carmickle has shown that he suffers from a medically-established back impairment that could reasonably be expected to produce back pain and reduced mobility. Thus, his subjective testimony about his back pain can be rejected only for clear and convincing reasons.
2
Accordingly, our next task is to determine whether the ALJ’s adverse credibility finding of Carmickle’s testimony is supported by substantial evidence under the clear-and-convincing standard. The ALJ rejected Carmickle’s assertion that he has to “change positions constantly” when sitting, finding it inconsistent with Carmick-le’s full-time college attendance. When discussing his school attendance, Carmick-le stated that he has trouble sitting through “the classes that are a little bit longer.” He also indicated that propping his feet up or leaning forward helps relieve his discomfort. He did not state, however, that he constantly adjusts his position while sitting in class. Rather, when pressed by counsel, he admitted that he can sit for “[ajbout 15 minutes in one particular position.” The ALJ included this 15-minute limitation in his RFC assessment, and on this record, we conclude this interpretation of the evidence is reasonable and we will not second-guess it. Rollins v. Massanari, 261 F.3d 853, 857 (9th Cir.2001).2
The ALJ also rejected Carmickle’s testimony that he can lift only 10 pounds occasionally in favor of Dr. Patton’s contradictory opinion that he can lift up to 10 pounds frequently. Contradiction with the medical record is a sufficient basis for rejecting the claimant’s subjective testimony. Johnson v. Shalala, 60 F.3d 1428, 1434 (9th Cir.1995). Thus, we conclude that both of these reasons relied on by the ALJ are supported by substantial evidence in the record.
3
The ALJ also gave less weight to Car-mickle’s testimony because he received unemployment benefits during the time he alleges disability and because he took only Ibuprofen to treat his pain. First, while receipt of unemployment benefits can un*1162dermine a claimant’s alleged inability to work fulltime, see Copeland v. Bowen, 861 F.2d 536, 542 (9th Cir.1988); accord Schmidt v. Barnhart, 395 F.3d 737, 745-46 (7th Cir.2005) (recognizing receipt of unemployment benefits could impact a claimant’s disability claim), the record here does not establish whether Carmickle held himself out as available for full-time or part-time work. Only the former is inconsistent with his disability allegations. Thus, such basis for the ALJ’s credibility finding is not supported by substantial evidence.
Second, although a conservative course of treatment can undermine allegations of debilitating pain, such fact is not a proper basis for rejecting the claimant’s credibility where the claimant has a good reason for not seeking more aggressive treatment. Orn v. Astrue, 495 F.3d 625, 638 (9th Cir.2007). Carmickle testified that he does not take other pain medication because of adverse side effects. In 2003, he also indicated that he would prefer to take Relafen, which was prescribed by Dr. Patton, but his insurance does not cover this medication. Both of these assertions are supported by Dr. Patton’s treatment notes, which state: “Prior auth [sic] is filled out for his Relafen which typically has not been covered, but, unfortunately, has been the only thing that has provided significant relief without addiction potential or intolerable side effects.” On this record, Car-mickle’s minimal treatment regime is not a proper basis for finding him non-credible.
4
Because we conclude that two of the ALJ’s reasons supporting his adverse credibility finding are invalid, we must determine whether the ALJ’s reliance on such reasons was harmless error. See Batson v. Comm’r of Soc. Sec. Admin., 359 F.3d 1190, 1195-97 (9th Cir.2004) (applying harmless error standard where one of the ALJ’s several reasons supporting an adverse credibility finding was held invalid). Our decision in Batson makes clear that reviewing the ALJ’s credibility determination where the ALJ provides specific reasons supporting such is a substantive analysis. So long as there remains “substantial evidence supporting the ALJ’s conclusions on ... credibility” and the error “does not negate the validity of the ALJ’s ultimate [credibility] conclusion,” such is deemed harmless and does not warrant reversal. Id. at 1197; see also Stout, 454 F.3d at 1055 (defining harmless error as such error that is “inconsequential to the ultimate nondisability determination”).
Contrary to the dissent’s assertion, the relevant inquiry in this context is not whether the ALJ would have made a different decision absent any error, see Dissent at 1168, it is whether the ALJ’s decision remains legally valid, despite such error. In Batson, we concluded that the ALJ erred in relying on one of several reasons in support of an adverse credibility determination, but that such error did not affect the ALJ’s decision, and therefore was harmless, because the ALJ’s remaining reasoning and ultimate credibility determination were adequately supported by substantial evidence in the record. 359 F.3d at 1197. We never considered what the ALJ would do if directed to reassess credibility on remand — we focused on whether the error impacted the validity of the ALJ’s decision. Id. Likewise, in Stout, after surveying our precedent applying harmless error in social security cases, we concluded that “in each case, the ALJ’s error ... was inconsequential to the ultimate non-disability determination.” 454 F.3d at 1055 (emphasis added).
Our specific holding in Stout does require the court to consider whether the ALJ would have made a different decision, but significantly, in that case the ALJ *1163failed to provide any reasons for rejecting the evidence at issue. See id. at 1054-55; see also Robbins, 466 F.3d at 885 (citing Stout and concluding ALJ’s failure to consider lay testimony was not harmless error). There was simply nothing in the record for the court to review to determine whether the ALJ’s decision was adequately supported. Here, however, as in Bat-son, the ALJ considered Carmickle’s testimony and provided specific reasons for finding him less than fully credible. Thus, there is a basis for the court to review the ALJ’s decision, and the analysis set forth in Batson, rather than Stout, controls.3 And as just discussed, in Batson we focused on the validity of the ALJ’s underlying decision, and not necessarily on whether the ALJ would come out differently if the case were remanded after error was identified by the court.4
Here, the ALJ’s decision finding Car-mickle less than fully credible is valid, despite the errors identified above. The ALJ did not wholly reject Carmickle’s allegations; indeed, the RFC assessment is largely consistent with his testimony. And, as discussed above, to the extent the ALJ rejected Carmickle’s allegations that he needs to change positions constantly when sitting and that he can lift only 10 pounds occasionally, such findings were based on substantial evidence in the record. These findings also support the ALJ’s conclusion that Carmickle’s daily activities are inconsistent with his allegations of disability. On this record, the ALJ’s error in relying on Carmickle’s receipt of unemployment benefits and on his relatively conservative pain treatment regime does not “negate the validity” of the ALJ’s adverse credibility finding. Batson, 359 F.3d at 1197. Contrary to the dissent’s assertion, the remaining valid reasons supporting the ALJ’s determination are not “relatively minor.” See Dissent at 1170. They are specific findings related to Carmickle’s ability to perform vocational functions, and they clearly demonstrate that to the extent the ALJ found Carmickle’s testimony incredible, the ALJ did not do so arbitrarily. Rollins, 261 F.3d at 856-57.
5
Carmickle also argues that the ALJ erred in finding lay witness Tucker’s *1164testimony “not entirely credible.” The ALJ must consider competent lay testimony but in rejecting such evidence, he need only provide reasons for doing so that are “germane to [the] witness.” Greger v. Barnhart, 464 F.3d 968, 972 (9th Cir.2006) (internal quotation marks omitted). Tucker, one of Carmickle’s classmates, testified that Carmickle often appears uncomfortable in class and has to lean back in his chair with his feet propped up on his wheeled book carrier. Tucker also stated Carmickle has trouble understanding at times and sometimes appears confused. The ALJ rejected this evidence finding it inconsistent with Carmickle’s successful completion of continuous full-time coursework. This reason is germane to Tucker. Therefore, we conclude that the ALJ had a proper basis on which to reject Tucker’s testimony.
B
The ALJ is responsible for resolving conflicts in the medical record. Benton v. Barnhart, 331 F.3d 1030, 1040 (9th Cir.2003). Those physicians with the most significant clinical relationship with the claimant are generally entitled to more weight than those physicians with lesser relationships. Lester, 81 F.3d at 830; 20 C.F.R. §§ 404.1527(d), 416.927(d). As such, the ALJ may only reject a treating or examining physician’s uncon-tradicted medical opinion based on “clear and convincing reasons.” Lester, 81 F.3d at 830-31. Where such an opinion is contradicted, however, it may be rejected for “specific and legitimate reasons that are supported by substantial evidence in the record.” Id. Here, Carmickle asserts the ALJ erred in rejecting medical evidence from Dr. Patton, Dr. Kane, Dr. Horowitz, and Dr. Nilaver.
1
Carmickle argues the ALJ erred by failing to include all of the limitations identified by Dr. Patton, a treating physician, in assessing Carmickle’s RFC. Dr. Patton submitted three RFC reports. Though largely consistent, there are some variations among the reports. The record also includes contradictory opinions from reviewing physicians. In resolving these conflicts, the ALJ gave Carmickle “the benefit of the doubt” and assigned great weight to Dr. Patton’s second report, dated November 2002, because the ME who testified at the hearing stated that this report included “generous” limitations. The ALJ also rejected some of the new limitations identified in Dr. Patton’s third report, dated November 2003, because they were based on a recent tendonitis diagnosis that “is not expected to result in any significant work-related functional limitation for any 12-month period.”
The ALJ is required to consider all of the limitations imposed by the claimant’s impairments, even those that are not severe. Social Security Ruling (“SSR”) 96-8p (1996). Even though a non-severe “impairment ] standing alone may not significantly limit an individual’s ability to do basic work activities, it may — when considered with limitations or restrictions due to other impairments — be critical to the outcome of a claim.” Id. Dr. Patton opined that Carmickle’s tendonitis significantly limits his ability to perform rotary movement. The ALJ erred in not including this limitation in his assessment of Carmickle’s RFC.
Carmickle also argues that the ALJ erred by not classifying his carpal tunnel syndrome as a “severe” impairment at step two of the analysis. However, the only medical evidence addressing such impairment is a letter dated in 1996 (well before Carmickle’s alleged onset of disability) stating that the physician “identified a probably unrelated carpal tunnel syndrome and peripheral neuropathy” and *1165notes from a follow-up visit indicating that “over the past couple of weeks [Carmickle] has had further improvement in his ... arm symptoms such that he states that over the past several days he has few, if any, arm complaints.” Furthermore, the medical record does not establish any work-related limitations as a result of this impairment. See Burch v. Barnhart, 400 F.3d 676, 682 (9th Cir.2005) (holding that a medical impairment is deemed “ ‘severe’ ... when alone or in combination with other medically determinable physical or mental impairment(s), it significantly limits an individual’s physical or mental ability to do basic work activities.”) (internal quotation marks omitted). Thus, we conclude that the ALJ did not err in his assessment of Carmickle’s carpal tunnel syndrome.
2
Carmickle next argues the ALJ erred in rejecting treating physician Dr. Kane’s statement that Carmickle cannot tolerate sedentary or light work. A few months after Carmickle’s disability-triggering accident, Dr. Kane diagnosed “moderate to severe degenerative disc disease in the lumbar spine” and gave Carmickle a two-week excuse from work. At that time, he stated: “I think it is clear that [Car-mickle] did not tolerate return to sedentary or light work.” Two months later, however, Dr. Kane released Carmickle to return to full-time work. The ALJ gave Dr. Kane’s opinion “little weight in assessing the claimant’s long-term functioning.” We conclude the ALJ’s assessment is supported by substantial evidence.
3
Carmickle also contends the ALJ erred in rejecting Dr. Horowitz’s recommendation that Carmickle use a “reclinable desk chair while working.” Based on Dr. Patton’s reports, the ALJ concluded a reclining chair is not necessary if Carmickle has a sit/stand option. In his first RFC report, Dr. Patton indicated that Carmickle can sit for two hours in a reclining chair and 30-45 minutes in a standard chair, but “rotation between sitting!,] standing!,] walking works best.” In his later reports, Dr. Patton makes no distinction between the type of chair, simply indicating that Car-mickle can sit for only short amounts of time and that he needs the option to alternate positions frequently. The difficulty is that Dr. Patton also stated that he agreed with Dr. Horowitz’s recommendation. Dr. Horowitz’s proposal was offered as a recommendation, not an imperative. Thus, the ALJ’s decision primarily to rely on Dr. Patton’s specific statements regarding Carmickle’s limitations, rather than his summary agreement with Dr. Horowitz, was rational, and we will not disturb it. See Allen v. Heckler, 749 F.2d 577, 579 (9th Cir.1985) (“If the evidence admits of more than one rational interpretation, we must uphold the decision of the ALJ.”).
4
Carmickle lastly points to the medical evidence from Dr. Nilaver, an examining physician. The ALJ gave little weight to Dr. Nilaver’s opinion because it was provided before Carmickle’s alleged onset of disability at a time when Carmick-le was working two jobs that he never indicated having trouble performing before his on-the-job injury. Medical opinions that predate the alleged onset of disability are of limited relevance. See Fair v. Bowen, 885 F.2d 597, 600 (9th Cir.1989). This is especially true in cases such as this where disability is allegedly caused by a discrete event. See SSR 83-20 (1983). As such, we conclude the ALJ did not err in his treatment of Dr. Nilaver’s evidence.
C
The final issue on appeal is whether the ALJ erred at step four in finding *1166that Carmickle can perform his past relevant work as a construction supervisor and counter salesperson. True, because the ALJ erred in excluding some of Carmick-le’s limitations from the RFC assessment, see supra Part 4.B.1, and thus from the VE hypothetical, the VE’s testimony “has no evidentiary value.” Russell v. Sullivan, 930 F.2d 1443, 1445 (9th Cir.1991) (order), abrogated on other grounds in Sorenson v. Mink, 239 F.3d 1140, 1149 (9th Cir.2001). However, because Carmickle also asserts the VE misclassified his past relevant work and erroneously conflated the step-four and step-five analyses, we must address this issue.
1
At step four of the sequential analysis, the claimant has the burden to prove that he cannot perform his prior relevant work “either as actually performed or as generally performed in the national economy.” Lewis v. Barnhart, 281 F.3d 1081, 1083 (9th Cir.2002). Carmickle labeled his construction supervisor position using various titles, including “job supervisor,” “construction supervisor (working),” “carpenter,” and “lead carpenter.” In his work history report, Carmickle described the job as follows: “Remodeled house — use saws, tape measures, drew plans, all skills required to remove & renew anything & everything in a residential home, supervised 3 employees, arranged equip, rentals ect [sic]. Constantly lifting or carrying something from a hammer to heavy beams or rolls of vinyl & carpet.” He frequently lifted 50 pounds or more, and the heaviest weight he lifted was 100 or more pounds. He stated that his supervisory duties comprised 20 percent of his time.
The VE conceded that as actually performed by Carmickle, this was a heavy-duty position. However, she classified the position as “Superintendent, Construction,” number 182.167-026 in the Dictionary of Occupational Titles (“DOT”), which is a purely supervisory position that requires no manual labor. And based on this classification, the VE concluded that as normally performed in the national economy, this is a light-duty position.
Carmickle argues such classification improperly focuses on the supervisory aspect of his work and that the proper classification is “Straw Boss,” which the DOT defines as someone who “takes the lead in a construction or laboring crew.” The exer-tional level of this occupation depends on the specific type of work the crew performs. Thus, because his crew did carpentry work, which is medium exertion, Carmickle argues his prior work as a construction supervisor is properly categorized as medium-duty.
The DOT is “the best source for how a job is generally performed.” Pinto v. Massanari, 249 F.3d 840, 845 (9th Cir.2001). In classifying prior work, the agency must keep in mind that every occupation involves various tasks that may require differing levels of physical exertion. It is error for the ALJ to classify an occupation “according to the least demanding function.” Valencia v. Heckler, 751 F.2d 1082, 1086 (9th Cir.1985). Here, the ALJ did just that. Only 20 percent of Carmickle’s duties as a construction supervisor involved supervision. The remainder of his time was spent performing manual labor. Yet the VE’s classification, which the ALJ accepted, was a purely supervisory position.
To the extent the VE, and thereby the ALJ, may have concluded that the supervisory skills Carmickle gained from this pri- or position are transferrable to a purely supervisory position, this was also error. The step-four analysis is limited to determining whether the claimant can perform his past relevant work. Id. at 1086-87; 20 C.F.R. §§ 404.1520, 416.920. Only if the ALJ finds that the claimant can no longer *1167perform his past work, as properly classified, does the analysis move to the fifth and final step of determining whether the claimant can perform any other work that exists in the national economy. Valencia, 751 F.2d at 1086-87. Here, the ALJ’s decision, by its own terms, was resolved at step four.
2
Carmickle also argues that the VE erred in classifying his prior work as a counter salesperson. He described the functions of this job as follows: “Used cash registers, saws, computer, advised customers as to product suggestions.... Occasional lifting of lumber to help customers and stock shelves, this was varying weights.... ” Carmickle frequently lifted less than 10 pounds, and the heaviest weight he lifted was 100 pounds or more. He had to “stand & walk all day long.”
Carmickle contends that as described, this position is properly classified as “Sales Attendant, Building Materials,” DOT number 299.677-014. The VE, however, failed to specify a DOT classification, instead generically referring to the position as a “counter sales position” or “customer service job.” The VE further concluded this job was a light-duty position.
The ALJ’s reliance on the VE’s generic classification was error. “[B]road generic occupational classifications ... are insufficient to test whether a claimant can perform past relevant work.” Vertigan v. Halter, 260 F.3d 1044, 1051 (9th Cir.2001); see also SSR 82-61 (1982) (“Finding that a claimant has the capacity to do past relevant work on the basis of a generic occupational classification of the work is likely to be fallacious and unsupportable.”). And not only did the ALJ rely on a generic classification, he failed to explain his reasons for doing so. At the hearing, the VE simply asserted, without explanation, that this is a light-duty position. Likewise, the ALJ baldly concluded that this position is “classified as light work and do[es] not require the performance of tasks precluded by [Carmickle’s] [RFC].” Though there may be cases where generic classifications are appropriate, or perhaps even necessary, the ALJ always “has a duty to make the requisite factual findings to support his conclusion” at step four. Pinto, 249 F.3d at 844. Otherwise, the court has no basis on which to review the agency’s decision. Id. at 847. Here, the ALJ failed sufficiently to support his conclusion.
Ill
For the foregoing reasons, the district court’s decision affirming the Commissioner’s decision is AFFIRMED IN PART and REVERSED IN PART with instructions to the district court to REMAND this case to the Commissioner for further proceedings. On remand, the ALJ is directed to reassess Carmickle’s RFC, including all of his relevant limitations. The ALJ is further directed to reassess the step-four conclusion in light of Carmickle’s complete RFC and this opinion, and if necessary, proceed to step five.
. As we noted in a recent unpublished decision, the statement in Robbins v. Social Security Administration, 466 F.3d 880, 883 (9th Cir.2006), suggesting that the ALJ must make a specific finding of malingering before the clear-and-convincing-reasons standard applies is an anomaly in this Circuit’s caselaw. Schow v. Astrue, 272 Fed.Appx. 647, 650-52 (9th Cir.2008). In fact, in cases both pre- and post-dating Robbins, we have held that the clear-and-convincing-reasons standard applies so long as there is “affirmative evidence suggesting ... malingering.” Smolen, 80 F.3d at 1283-84; see Lingenfelter, 504 F.3d at 1036 (same).
. The ALJ also rejected Carmickle's assertion that he has memory problems on the basis that he has successfully completed several full-time college terms. We do not address this finding because Carmickle failed to argue this issue with any specificity in his briefing. See Paladin Assocs., Inc. v. Mont. Power Co., 328 F.3d 1145, 1164 (9th Cir.2003) (noting that we “ordinarily will not consider matters on appeal that are not specifically and distinctly argued in an appellant's opening brief”).
. The dissent contends that we are erroneously adopting the dissenting view expressed in Stout and Robbins. See Dissent at 1168-69. Such allegation, however, mischaracterizes the disagreement in this case and misunderstands our caselaw. The disagreement here is in how to interpret our decision in Batson, which established the harmless error standard in cases where the ALJ provides specific reasons supporting its credibility determination. As explained, Stout and Robbins purportedly extended Batson and created a harmless error standard applicable where the ALJ fails to make a reasoned credibility determination. Such cases did not, however, overrule what we said in Batson, nor could they. And while it is true the dissenting view in Stout and Robbins disagreed with such extension, see Robbins, 466 F.3d at 889 (O’Scann-lain, J., dissenting) ("In Batson ... we simply asked whether there remained ‘substantial evidence supporting the ALJ's decision,’ or whether the error in any way 'negated[d] the validity of the ALJ’s ultimate conclusion.’ A similar analysis would have been proper in this case.”) (quoting Batson, 359 F.3d at 1197) (internal citation omitted), it is not this view that we adopt today, but rather an interpretation of Batson itself that is based solely on the language we used therein.
. The dissent also relies on the statement in Stout asserting that the error in Batson was harmless because it "did not materially impact [the AU’s] decision.” Dissent at 1168 (quoting Stout, 454 F.3d at 1055) (emphasis added). For the reasons previously discussed, we conclude that Batsons "materiality” analysis considers whether the ALJ's underlying decision remains supported, in spite of any error, and not whether the ALJ would necessarily reach the same result on remand. Thus, we also read this statement in Stout as referring to a material impact to the validity of the ALJ's decision.