ORDER
Charles Perkins claims that he is disabled by congestive heart failure and hypertension. He applied for disability insurance benefits and social security income, but the Social Security Administration denied his application at each stage of review and a district court upheld the denial. On appeal he argues that the ALJ’s residual functional capacity and credibility determinations are flawed. We agree with Perkins that the ALJ did not adequately explain these parts of the decision and remand this case to the agency for further proceedings.
Perkins, 56 years old, applied for disability insurance benefits and supplemental security income, alleging that he became disabled in 2005. For much of 2005 he had been experiencing shortness of breath, and he was brought to the emergency room in September after fainting. He was diagnosed with congestive heart failure and hypertension, and was prescribed medications to manage his ailments. A few weeks later he returned to the emergency room, suffering from shortness of breath and swelling in his legs. He again was diagnosed with congestive heart failure and hypertension (described this time as “severe”). Perkins’s physician noted that he had not been taking his medications consistently and tested positive for cocaine.
Since then, Perkins has continued to experience shortness of breath and fatigue, *643though the severity of his symptoms has varied. At examinations in the first half of 2006, his medications were controlling his shortness of breath, but his heart failure was nonetheless markedly limiting his physical activity. In 2007 he complained of fatigue and shortness of breath when he was active, though not when at rest. Later that year, however, he reported shortness of breath when sitting, walking short distances, and eating. As for his daily activities around that time, Perkins said that he found dressing, tying his shoes, and showering “exhausting,” that he needed to rest after a few minutes of mild activity, and that he took six or more rest periods during the day.
Two state-agency physicians reviewed Perkins’s medical records and suggested modest work limitations. They determined that he could stand for 6 hours of an 8-hour workday, sit for 6 hours of an 8-hour workday, occasionally lift 20 pounds, and frequently lift 10 pounds.
At his hearing in the fall of 2008, Perkins testified about the effects of his heart failure and hypertension. An Air Force veteran, he had worked as a building inspector, warehouse clerk, and sports official. He tried and was unable to perform a job in 2008 packing boxes in a warehouse because extended periods of standing made him short of breath and caused swelling in his legs. In response to questions about his history of drug abuse, Perkins replied that he had not abused drugs since 2005 — testimony that, it turns out, conflicted with some of his medical records. A vocational expert also testified, opining that if Perkins were able to stand or walk for six hours each day, he could not perform his past work but he could perform light work. If he could stand for only 30 minutes at a time, the expert testified, then he would be limited to sedentary work, none of which he was qualified to do.
The ALJ denied Perkins’s claim for benefits after applying the five-step evaluation process. See 20 C.F.R. § 404.1520(a)(4). The ALJ concluded that Perkins hadn’t worked since his alleged onset date (Step 1); his congestive heart failure and hypertension were severe impairments (Step 2); his impairments did not meet or medically equal a listed impairment (Step 3); his residual functional capacity allowed him to perform light work, which precluded his past work (Step 4); and he could work as a cashier, fastfood worker, or host, all classified as light work (Step 5). The ALJ also found Perkins’s testimony about the severity of his symptoms incredible.
On appeal, Perkins first challenges the AL J’s conclusion that he had the residual functional capacity to perform light work. Perkins could perform light work, the ALJ determined, because his medications had been “relatively effective in controlling his symptoms, when taken regularly as prescribed.” Perkins argues that the ALJ selectively cited the medical records, failing to consider evidence showing the severity of his symptoms even when he was taking his medications.
An ALJ may not ignore entire lines of contrary evidence or selectively consider medical reports, Myles v. Astrue, 582 F.3d 672, 678 (7th Cir.2009); Terry v. Astrue, 580 F.3d 471, 477 (7th Cir.2009), and here, the ALJ overlooked several pieces of contrary evidence. Most significantly, the ALJ failed to acknowledge that a cardiologist characterized Perkins’s heart failure as Class III under the New York Heart Association classification system. This is important because he had been taking his medications at the time, and Class III heart failure means that a person becomes short of breath and fatigued with less-than-ordinary activity. Classification of-functional capacity and objective assessment, American Heart Association, http:// my.americanheart.org/professional/ *644StatementsGuidelines/ ByPublieation-Date/PreviousYe ars/Classification-of-Functional-Capacity-and-Objective-Assessment — UCM—423811—Article.jsp (last updated Mar. 18, 2011). The ALJ also cited a physician’s evaluation notes from spring 2007 to show that Perkins’s medications were controlling his symptoms, but disregarded notes in the same evaluation that Perkins’s hypertension was uncontrolled and that Perkins reported both shortness of breath on exertion and “significant fatigue.” And the ALJ improperly discounted records from late 2007 that noted how poorly Perkins’s heart was pumping blood; in the ALJ’s view, this problem could be explained by Perkins’s failure to take his medications, but — significantly— the ALJ overlooked a physician’s note on the same page that Perkins was “Compliant with Medication.” Though the government has offered its own explanations for the residual functional capacity decision, we may look only to the rationales offered by the ALJ. See Spiva v. Astrue, 628 F.3d 346, 353 (7th Cir.2010); Parker v. Astrue, 597 F.3d 920, 922 (7th Cir.2010).
Perkins next argues that the ALJ did not adequately explain the adverse credibility determination, offering little more than boilerplate language. Substantial evidence does not support the ALJ’s credibility determination, even with the special deference we accord it. First, the ALJ trotted out what we have criticized as “meaningless boilerplate” when he concluded that Perkins’s “statements concerning the intensity, persistence and limiting effects of [his] symptoms are not credible to the extent they are inconsistent with the above residual functional capacity assessment.” See, e.g., Bjornson v. Astrue, 671 F.3d 640, 645 (7th Cir.2012). Second, although the ALJ discussed Perkins’s failure to take his medications consistently, he never explored the reasons for Perkins’s failure to follow a treatment plan and thus should not have drawn negative inferences from this failure. See Myles, 582 F.3d at 677; Moss v. Astrue, 555 F.3d 556, 562 (7th Cir.2009); Craft v. Astrue, 539 F.3d 668, 679 (7th Cir.2008). This is especially important here because Perkins testified that he was sometimes unable to receive his medications due to delays and mix-ups at the VA pharmacy rather than any fault of his own. Third, the ALJ only briefly mentioned Perkins’s dishonesty about his drug abuse, without explaining whether or how much this undermined Perkins’s credibility. See S.S.R. 96-7p,1996 SSR LEXIS 4, at *3-4; Craft, 539 F.3d at 678; Zurawski v. Halter, 245 F.3d 881, 887 (7th Cir.2001). Last, the ALJ noted Perkins’s comment to a nurse that he walks a mile daily, but the ALJ again did not explain how his ability to walk one mile is inconsistent with the limitations that he claimed, or consistent with standing or walking for 6 hours of an 8-hour workday. See, e.g., Carradine v. Barnhart, 360 F.3d 751, 755-56 (7th Cir.2004) (explaining that ability to exercise does not necessarily mean person can work). And here, again, although the government has given its own explanations for the ALJ’s determination, we review only the ALJ’s reasoning. See Spiva, 628 F.3d at 353; Parker, 597 F.3d at 922.
Perkins lastly argues that the ALJ inappropriately ignored the vocational expert’s testimony that, if he were limited to sedentary work, there would be no jobs available for him. But this argument essentially reconsiders the validity of the ALJ’s residual functional capacity determination, and so collapses into his other arguments. See Arnett v. Astrue, 676 F.3d 586, 591 (7th Cir.2012).
Accordingly, we VACATE the district court’s judgment and REMAND this case to the agency for further proceedings.