NONPRECEDENTIAL DISPOSITION
                          To be cited only in accordance with
                                   FED . R. APP. P. 32.1



            United States Court of Appeals
                                 For the Seventh Circuit
                                 Chicago, Illinois 60604

                                  Argued March 2, 2010
                                  Decided April 13, 2010

                                          Before

                            DIANE P. WOOD, Circuit Judge

                            ANN CLAIRE WILLIAMS, Circuit Judge

                            DAVID F. HAMILTON, Circuit Judge

No. 09-2595

JENNIFER RICHARDS,                                 Appeal from the United States District
     Plaintiff-Appellant,                          Court for the Northern District of
                                                   Illinois, Eastern Division.
       v.
                                                   No. 08 C 972
MICHAEL J. ASTRUE,
Commissioner of Social Security                    Charles R. Norgle, Sr.,
    Defendant-Appellee.                            Judge.

                                        ORDER

        Jennifer Richards applied for disability insurance benefits after suffering a retinal
tear in her right eye, and she later amended her application to include claims of
disabling depression and anxiety. An administrative law judge concluded that Richards
could perform her past relevant work as a cashier and customer-service representative
and denied the application. The district court upheld the ALJ’s decision, but we
conclude that the ALJ erred in assessing Richards’s credibility, the limiting effects of her
No. 09-2595                                                                               2

mental impairments, and her residual functional capacity. Accordingly, we vacate and
remand for further proceedings.

                                    I. BACKGROUND

       Richards is a 51-year-old mother of four with a high-school education. In
September 2004 she underwent laser treatment to seal off a retinal tear in her right eye.
The surgery was successful, but Richards nevertheless applied for disability insurance
benefits the following month, claiming that she was disabled by an eye impairment. A
state-agency physician reviewed Richards’s medical records and noted that she had
some visual limitations and should avoid concentrated exposure to hazards such as
machinery and heights, but the agency concluded that she was not disabled and denied
her application.

        Richards requested a hearing before an ALJ and, in the interim, amended her
application to include claims of disabling depression and anxiety. In support of these
new claims, she submitted records dating back to the late 1980s, when she was placed
under the supervision of state child-welfare authorities following reports of abuse.
Since at least 1995 Richards has been diagnosed with depression and has regularly
attended therapy with a licensed clinical social worker. She submitted detailed records
from therapy sessions until 2003, which were sometimes as frequent as once a week and
focused primarily on stress caused by parenting and employment (or lack thereof).
Richards switched therapists in 2004, and the record includes only general “treatment
plans” that her new therapist, Marsha Smith, drew up at six-month intervals. These
plans contemplated that Smith would see Richards for hour-long individual therapy
sessions twice a month focused on managing her depression and stress, increasing her
social support, and developing strategies for finding employment. In a letter to the SSA
in July 2006, Smith opined that Richards “has symptoms of chronic depression and
anxiety which interfere with her ability to pursue or maintain employment.”

       Richards also submitted treatment records from a psychiatrist whom she
consulted three times between September 2005 and July 2006. Richards’s chief
complaints during these visits were lack of energy, difficulty sleeping, nightmares,
crying spells, irritability, and trouble controlling her anger, which she attributed to
memories of sexual abuse she suffered as a child. At the first visit, the psychiatrist
diagnosed Richards with depression and assigned her a Global Assessment of
Functioning (“GAF”) score of 60-70, which indicates a patient with “some mild
symptoms (e.g., depressed mood and mild insomnia) or some difficulty in social,
No. 09-2595                                                                                  3

occupational, or school functioning” but who is “generally functioning pretty well.”1
See AM . PSYCHIATRIC ASS’N , DIAGNOSTIC AND STATISTICAL MANUAL OF MENTAL
DISORDERS 32 (4th ed. 1994) (“DSM-IV”). The psychiatrist prescribed Lexapro, a drug
used to treat major depressive disorder and generalized anxiety disorder, see
PHYSICIAN ’S DESK REFERENCE 1160-61 (64th ed. 2010), and Ativan, another drug
prescribed for generalized anxiety disorder, see PDR FAMILY GUIDE TO PRESCRIPTION
DRUGS 66 (9th ed. 2002). The doctor renewed these prescriptions in May 2006, and at
Richards’s next appointment two months later, the psychiatrist noted that Richards was
“feeling better, but still depressed” and was suffering from “psychomotor retardation.”
The doctor increased her Lexapro dosage and added a prescription for Wellbutrin,
another antidepressant. See 

        At her hearing in July 2006, Richards testified that stress and nightmares kept her
up at night, that she had problems with concentration and memory, and that she got
upset and mad easily. She reported that her “whole body ache[d]” and that, as a result,
she seldom left the house. She attributed her depression primarily to earlier episodes of
sexual abuse and domestic violence, which she began thinking about with increased
frequency once her children had grown and left her home. Richards testified that
although she had worked at the Northwestern University bookstore for roughly five
hours a week during the back-to-school rush, she was collecting unemployment benefits
at the time of the hearing.

       After hearing Richards’s testimony, the ALJ asked a vocational expert if an
individual with Richards’s age, education, and work experience could perform her past
work as a cashier and customer-service representative if she (1) could sit, stand, and/or
walk for at least 6 hours out of an 8-hour workday, (2) could lift and carry up to 10
pounds frequently and up to 20 pounds occasionally, (3) needed to avoid concentrated
exposure to unprotected heights and hazardous machinery, and (4) could not perform a
job requiring good binocular vision or depth perception. The VE opined that such an
individual could perform Richards’s past jobs. When asked if the person could perform
those jobs if she “should have no regular general public contact, and is limited to jobs
that are low to moderate stress,” the VE opined that she could not but added that she
could perform other jobs including office helper, laundry worker, and production
worker.


       1
         A GAF score of 51-60 indicates a patient with moderate symptoms, while a score of 61-
70 indicates a patient with mild symptoms. DSM-IV at 32. It is unclear whether, by assigning
Richards a score of 60-70, the psychiatrist believed that her symptoms might fall in the
moderate range or whether, more likely, he simply made an oversight.
No. 09-2595                                                                                  4

       The ALJ evaluated Richards’s claim under the familiar sequential analysis. See 20
C.F.R. § 404.1520. At step one, the ALJ found that Richards had not performed
substantial gainful activity since the alleged onset of her disability. At step two, the ALJ
found that Richards had severe impairments of “slight decreased visual acuity,” “mild
depression/anxiety,” and obesity, but at step three she found that none of these met or
equaled a listed impairment. Moving on to step four, the ALJ concluded that Richards
had the residual functional capacity (“RFC”) to lift 20 pounds occasionally and 10
pounds frequently, and to stand or walk at least 6 hours and sit at least 8 hours in an 8-
hour workday. The ALJ also found that Richards could not perform “complex job
tasks” or jobs requiring good binocular vision or good depth perception and must avoid
exposure to unprotected heights and dangerous moving machinery. In reaching these
conclusions, the ALJ found Richards not credible and thus discounted her testimony
about the limiting effects of her depression and anxiety. Finally, the ALJ concluded that
Richards was capable of performing her past relevant work as a cashier and customer-
service representative and therefore was not disabled.

                                      II. ANALYSIS

       We review the district court’s decision de novo, reviewing the ALJ’s decision
directly. Moss v. Astrue, 

, 560 (7th Cir. 2009). We will uphold the ALJ’s
decision if it is supported by substantial evidence, meaning “such relevant evidence as a
reasonable mind might accept as adequate to support a conclusion.” Terry v. Astrue, 

, 475 (7th Cir. 2009) (internal quotation marks and citations omitted). The ALJ
need not address every piece of evidence in the record but must “build an accurate and
logical bridge from the evidence to his conclusion.” Clifford v. Apfel, 

, 872
(7th Cir. 2000).

       A.     Special Technique

       Richards first argues that the ALJ failed to follow the procedure for evaluating
mental limitations described in 20 C.F.R. § 404.1520a. See generally Craft v. Astrue, 

, 674-75 (7th Cir. 2008). Under this so-called “special technique,” the ALJ must,
in determining whether the claimant has a severe impairment (step two of the five-step
analysis), rate the degree of the functional limitation resulting from the claimant’s
impairment with respect to four broad functional areas: activities of daily living; social
functioning; concentration, persistence, or pace; and episodes of decompensation. 20
C.F.R. § 404.1520a(c)(3). The ALJ must rate the claimant’s limitation in the first three
categories as none, mild, moderate, marked, or extreme, and number the claimant’s
episodes of decompensation. 

If there are no episodes of
decompensation and the rating in each of the first three categories is none or mild, the
No. 09-2595                                                                                         5

impairment generally is not considered severe and the claimant thus is not disabled. 

Otherwise, the impairment is classified as severe, and the ALJ
continues on to steps three through five of the standard five-step analysis. 

ALJs formerly were required to enter this information on a standard
document known as a Psychiatric Review Technique Form (“PRTF”) and append it to
their decision, see Kasarsky v. Barnhart, 

, 543 (7th Cir. 2003), but now they
need only incorporate into their decision the pertinent findings and conclusions based
on the technique, see 20 C.F.R. § 404.1520a(e)(2).

        Although the government insists that the ALJ followed the special technique, it is
clear that she did not, at least not to the letter. The ALJ nowhere mentioned that she
was applying the technique, and although the ALJ happened to assign Richards a rating
in each of the four functional categories, she did this at step three of her five-step
analysis (not at step two, as the technique requires) and did not explain how she had
reached her conclusions. See 20 C.F.R. § 404.1520a(e)(2). The ALJ did recount some of
Richards’s mental-health history in the RFC analysis, but we have cautioned that “the
RFC analysis is not a substitute for the special technique, even though some of the
evidence considered may overlap.” See 

.

        An ALJ’s failure to explicitly use the special technique may be harmless error,

; see also Rabbers v. Comm’r SSA, 

, 654-57 (6th Cir. 2009),
but here, however, the ALJ’s misstep is compounded by other errors in her analysis,
and the combined effect of these errors requires a remand. Most significantly, we are
troubled that the ALJ rated Richards’s mental functional limitations without the benefit
of any medical professional’s assessment of her mental RFC. Typically, when an
applicant claims a mental impairment, the agency’s medical or psychological consultant
will complete a PRTF and assess the severity of the impairment before the case reaches
an ALJ. 20 C.F.R. § 404.1520a(e)(1); see, e.g., Villano v. Astrue, 

, 561 (7th Cir.
2009); Young v. Barnhart, 

, 999 (7th Cir. 2004). But Richards’s case is unusual
because her initial application claimed only an eye impairment, and thus the state-
agency physicians who reviewed her file evaluated only the effect of her visual
limitations on her ability to work. Richards’s therapist, however, opined that she “has
symptoms of chronic depression and anxiety which interfere with her ability to pursue
or maintain employment.”2



       2
        Although evidence from a social worker may not be used to establish whether a
claimant has a medically determinable impairment, it may be used to show the severity of the
claimant’s impairment(s) and how it affects her ability to work. 20 C.F.R. § 404.1513(a), (d)(1).
No. 09-2595                                                                                  6

        Although an applicant for disability benefits bears the burden of proving that she
is disabled, an ALJ may not draw conclusions based on an undeveloped record and
“has a duty to solicit additional information to flesh out an opinion for which the
medical support is not readily discernable.” Barnett v. Barnhart, 

, 669 (7th
Cir. 2004); see also Nelms v. Astrue, 

, 1098 (7th Cir. 2009); Smith v. Apfel, 

, 437-38 (7th Cir. 2000); 20 C.F.R. § 404.1545(a)(3) (“[B]efore we make a
determination that you are not disabled, we are responsible for developing your
complete medical history. . . .”). The government is correct that there is no absolute
requirement that an ALJ remand a case simply because a PRTF was not completed at
the initial or reconsideration level. See 20 C.F.R. § 404.1520a(e)(3) (providing that ALJ
may consult medical expert or remand to state agency if unable to apply special
technique on her own). But one of the stated goals of the special technique is to help the
agency “[i]dentify the need for additional evidence to determine impairment severity,”

and the need for additional evidence about the limiting effects of
Richards’s depression and anxiety was apparent.

       Richards’s psychiatrist noted that she suffered psychomotor retardation, lack of
energy, difficulty sleeping, nightmares, crying spells, irritability, and trouble controlling
her anger—all symptoms which would certainly bear on her ability to work. The
psychiatrist also diagnosed Richards with depression and noted on Axis IV of her DSM
assessment that “problems with social environment” affect her ability to function, but
he never opined on how these impairments affect Richards’s functional capacity for
employment. Yet, without any medical professional having rated Richards’s limitations
in the areas of daily living, social functioning, and concentration, persistence, and pace,
the ALJ assigned a rating of “mild” in each category. In the absence of any expert
foundation for these ratings, we cannot discern the necessary logical bridge from the
evidence to the ALJ’s conclusions.

       B.     Credibility

        We are also troubled by the ALJ’s credibility assessment. A credibility finding is
entitled to considerable deference, but only if the ALJ justifies her conclusions with
reasons that are supported by the record. 

. Here, the ALJ failed to
substantiate her conclusions with accurate references to the record. For example, the
ALJ stated that Richards had “described daily activities that are not limited to the extent
one would expect, given the complaints of disabling symptoms and limitations,” but the
ALJ did not specify what activities she meant. In fact, Richards testified that her daily
activities were often limited to sleeping and looking out the window, and she reported
that on most days she was too depressed to leave the house, even for groceries, or to
No. 09-2595                                                                                 7

socialize with family or friends. Although she testified that she went to doctor’s
appointments and sometimes cooked and cleaned, she explained that her daughter
frequently helped her with these chores. The ALJ did not adequately explain how the
very minimal activities Richards described contradict a claim of a disabling mental
disorder. See Bauer v. Astrue, 

, 608 (7th Cir. 2008); Zurawski v. Halter, 

, 887 (7th Cir. 2001).

       The ALJ also supported her adverse credibility finding by noting that Richards
had not gone to therapy on a regular basis, but this mischaracterizes the record.
Although Richards did not submit detailed records from her sessions with her most
recent therapist, the therapist’s treatment plans indicate that Richards was attending
hour-long individual therapy sessions twice a month. Nor was there support for the
ALJ’s conclusion that Richards’s medication had been “relatively effective” in
controlling her symptoms. At their last consultation, Richards’s psychiatrist noted new
symptoms of psychomotor retardation and tearfulness, doubled Richards’s Lexapro
dosage, and added a prescription for Wellbutrin, a drug used to treat major depression.

       Also flawed is the ALJ’s finding that, although Richards “alleges she is too
stressed to go out and work, she works at a stressful cashier job and only at the busiest
times.” This is a significant overstatement. Richards testified that she was a standby
employee at the Northwestern University bookstore and worked roughly five hours a
week during the back-to-school rush. But she also testified that she was late for work
every day because nightmares kept her up at night and that her boss allowed her to
take breaks when she was not feeling well or was having a panic attack. That Richards
was able to maintain temporary employment for a few hours a week with
accommodations from a generous supervisor does not contradict her claim of disability.
See Henderson v. Barnhart, 

, 435-36 (7th Cir. 2003). Further, the ALJ failed to
consider the difference between being able to work a few hours a week and being able
to work eight hours a day five days a week. See SSR 96-8p, 

, at *1;
Carradine v. Barnhart, 

, 755 (7th Cir. 2004).

       Finally, the ALJ also erred when she questioned the reliability of Richards’s
claims on the ground that she was receiving unemployment compensation. Although
we have noted that a claimant’s representations in seeking unemployment benefits may
be relevant in assessing the credibility of her representations to the SSA, Schmidt v.
Barnhart, 

, 746 (7th Cir. 2005), Richards testified that she sought
unemployment benefits only because she had no other source of income. A desperate
person might force herself to work—or in this case, certify that she is able to work—but
that does not necessarily mean she is not disabled. See Gentle v. Barnhart, 

,
No. 09-2595                                                                                   8

867 (7th Cir. 2005); Hawkins v. First Union Corp. Long-Term Disability Plan, 

,
918 (7th Cir. 2003).

       C.     RFC Analysis

        The ALJ’s errors above also led to a flawed analysis of Richards’s residual
functional capacity. The only mental limitation the ALJ noted in the RFC was Richards
inability “to perform complex job tasks.” But the ALJ did not explain which of
Richards’s mental limitations she meant to capture with this finding. See Young v.
Barnhart, 

, 1002-03 (7th Cir. 2004). Although the ALJ apparently credited
notations in Richards’s psychiatric records that she had depression, anxiety, irritability,
difficulty controlling her anger, and problems with her social environment, the ALJ did
not address how these limitations might impact Richards’s ability to respond
appropriately to supervision, co-workers, and work pressures, or reduce her ability to
do her past work or other work. See 20 C.F.R. § 404.1545(c); 

(explaining that ALJ must evaluate all limitations that arise from medically
determinable impairments, even those that are not severe). The ALJ apparently
contemplated a more limited mental RFC during the hearing—the second hypothetical
she posed to the VE included limitations of “no regular general public contact” and
“jobs that are low to moderate stress”—but abandoned these limitations without
explanation in her final decision.

       D.     Obesity and Visual Impairments

        Richards’s remaining arguments are less convincing. She argues that the ALJ
erred in failing to consider the effect of her obesity on her other impairments, but we
have repeatedly characterized an ALJ’s failure to explicitly discuss a claimant’s obesity
as harmless error when the ALJ factors obesity “indirectly” into her decision. See, e.g.,
Prochaska v. Barnhart, 

, 736-37 (7th Cir. 2006); Skarbek v. Barnhart, 

, 504 (7th Cir. 2004). Here, the ALJ explicitly noted that Richards was morbidly
obese but implicitly concluded that her obesity had no effect on her RFC. Richards
recites from Social Security Ruling 02-1p, 

, some of the possible ways
obesity may affect other impairments, but she does not contend that any of those
examples apply to her or identify any actual effect of her obesity, and nothing in her
medical records suggests that it aggravated her depression or anxiety.

       Richards also faults the ALJ for not fully incorporating her visual impairments
into the RFC: Although the state-agency physicians who reviewed her records checked
off boxes on a standard form to reflect that she was limited in her near and far acuity,
accommodation, and field of vision, the ALJ made a general finding that Richards could
No. 09-2595                                                                                  9

not perform jobs “requiring good binocular vision or good depth perception.” But any
error in not reciting the state-agency physicians’ precise findings is harmless because it
is clear that the ALJ simply collapsed them into her broader finding. Indeed, when
asked at the hearing about her visual limitations, Richards testified only that she had
last seen her eye doctor about a year earlier when he prescribed new glasses, which she
needs only for reading and driving.

                                   III. CONCLUSION

      We VACATE the judgment of the district court and REMAND for further
proceedings. On remand the agency should reevaluate Richards’s mental limitations
and residual functional capacity with the benefit of an expert opinion and reassess
Richards’s credibility in light of the entire record.