United States Court of Appeals
                   For the First Circuit
                                         

No. 98-2231                      

                        MYRTLE THOMAS,

                    Plaintiff, Appellant,

                              v.

                    EASTMAN KODAK COMPANY,

                     Defendant, Appellee.

                                         

         APPEAL FROM THE UNITED STATES DISTRICT COURT

              FOR THE DISTRICT OF MASSACHUSETTS

      [Hon. W. Arthur Garrity, Jr., U.S. District Judge]

                                         

                            Before

                    Lynch, Circuit Judge,
               Bownes, Senior Circuit Judge,
                 and Lipez, Circuit Judge.
                                         

   Marisa A. Campagna, with whom the Law Offices of Marisa A.
Campagna were on brief, for appellant.
   Michael A. Fitzhugh, with whom Jon M. Nelson and Fitzhugh &
Associates were on brief, for appellee.

                                         

                       July 15, 1999
                                             LYNCH, Circuit Judge.  In 1993, Myrtle Thomas, the only
black Customer Service Representative in Eastman Kodak's Wellesley,
Massachusetts office, was laid off.  Thomas responded with a race
discrimination suit against Kodak under Title VII, 42 U.S.C.
2000e to e-17, arguing that Kodak's layoff decision was
discriminatory because it resulted from a ranking process that
relied on racially biased performance appraisals prepared in 1990,
1991, and 1992.  Kodak made two arguments in its motion for summary
judgment:  first, that Thomas's claim was time-barred because the
performance appraisals were conducted outside of Title VII's
statutory limitation period, and second, that Thomas failed in any
event to present enough evidence of racial animus to support a
disparate treatment claim.  The district court disagreed with the
first point but agreed with Kodak's second argument and granted
summary judgment.  Both issues are before us on appeal.
We decide both in favor of Thomas.  We find Thomas's
claim to be timely because the discriminatory appraisals that she
is challenging first caused her concrete harm when they led to her
layoff in 1993.  Because we also find that she has presented enough
evidence to support her claim that the performance appraisals were
racially biased, we reverse the district court's grant of summary
judgment and remand for further proceedings.
After thirty-five years of litigation under Title VII,
cases can still present new wrinkles.  This is one such case. 
Because it raises a number of important issues -- some new and some
familiar but difficult -- we preview the key holdings.
First, when an employer utilizes scores from past
performance appraisals in an objective formula to determine who
will be laid off, and the laid-off employee suffered no earlier
concrete harms from those appraisals, the accrual date for the
limitations period is the date of the notice of layoff, not the
date of the performance appraisals.  
Second, once there is sufficient evidence to create a
material issue of fact that the employer's articulated reason for
an adverse employment action is a pretext, there is no requirement
that a plaintiff always produce direct evidence to demonstrate that
the real reason was discriminatory.
Third, Title VII's prohibition against "disparate
treatment because of race" extends both to employer acts based on
conscious racial animus and to employer decisions that are based on
stereotyped thinking or other forms of less conscious bias.
Fourth, under the McDonnell Douglas/Burdine framework, a
court may not enter summary judgment for an employer based upon a
non-discriminatory reason not articulated by the employer but
identified sua sponte by the district court.
                             I
In reviewing a grant of summary judgment, we consider the
facts in the light most favorable to the nonmoving party, drawing
all reasonable inferences in that party's favor.  See Aponte Matos
v. Toledo Davila, 

, 186 (1st Cir. 1998).  Given the
subtlety of the questions before us, we outline Thomas's
experiences at Kodak in some detail.
Thomas was a long-term Kodak employee.  She first began
working for the company in 1974.  In 1980, after working for six
years in clerical and administrative positions in Kodak's
Rochester, New York facility, she was promoted to Customer Support
Representative ("CSR") within the Office Imaging Division and
transferred to Kodak's office in Wellesley, Massachusetts. 
Along with five other CSRs working out of the Wellesley
office, Thomas supported customers in an assigned territory who
owned Kodak copiers and other Kodak equipment.  She helped
salespeople perform installations, trained customers in the use and
maintenance of Kodak equipment, facilitated communication between
customers and sales and service personnel, and provided other forms
of marketing support.  
Thomas generally performed her job well.  Kodak managers
who supervised Thomas during her first ten years as a CSR in the
Wellesley office reported variously that they were never
dissatisfied with her performance, that they were "delighted" with
Thomas, that her work was "excellent" and "far superior" to that of
some of the other CSRs, that she was "very much on top of things,"
and that she was "the perfect support person."  
Co-workers and customers expressed similar sentiments. 
A sales representative who worked with Thomas sent a memorandum to
Thomas's supervisor praising her "continuous professionalism,"
"very high level of commitment," and "total dedication."  The sales
representative later noted that he was particularly impressed with
the way a certain customer "really went out of his way" to
emphasize his satisfaction with Thomas's support.  Another customer
who contacted Kodak after Thomas's layoff described Thomas as "an
irreplaceable part of the Kodak team" and explained that Thomas was
the primary reason for his selection of Kodak copiers over copiers
from other companies.  The customer concluded:  "In my many
contacts with company representatives, I have not met anyone of the
class and caliber of Myrtle Thomas."
Because of her high level of performance, Thomas received
awards and bonuses from Kodak.  In 1989, the company also changed
Thomas's grade from K4 to K6, which resulted in a salary increase.
According to Kodak's job description, the K6 grade was limited to
CSRs  who were "making an outstanding contribution to the support
activities," defined as "servicing the largest and/or most
sensitive accounts, developing and giving individualized
presentations and/or demonstrations, and training new CSRs." 
Thomas received at least eight other salary increases during her
years as a CSR.  At the time of the 1993 layoff, Thomas was the
fourth most senior CSR in the Wellesley office and earned the third
highest salary. 
The Kodak compensation plan stated that "[t]he company's
goal for [its] pay program is to reward each individual's job
performance appropriately" and to ensure that "[p]eople with higher
performance will, over time, be paid more than average performers."
The company made use of annual performance appraisals in order to
reach this goal.  According to the compensation plan, performance
appraisals were also used for a number of other purposes,
including:
  A. Evaluating and documenting the performance of each
individual in comparison with performance expectations
for the job.
  B.  Providing individuals with constructive feedback.
  C. Identifying the guidance and training that can help
individuals be as successful as their ability permits.
  [and]
  D.  Determining who should be promoted, transferred,
demoted, terminated, laid off, and re-employed.
To conduct an appraisal, the supervisor who directed an
employee's day-to-day activities filled out an appraisal form.  The
form contained a section pertaining to "basic performance
measures," which included categories for quality of results,
quantity of results, job skills, and teamwork.  Another section
pertained to "additional performance measures," including
dependability, versatility, communications, and leadership.  The
form required the supervisor to give the employee a rating from 1
to 7 for each applicable category, as well as an overall rating.
A separate section of the appraisal form contained space for the
supervisor's comments on the evaluated categories, a description of
any developmental opportunities, a description of matters discussed
during the post-appraisal interview with the employee, and both
supervisor and employee signatures.  Each appraisal was also
reviewed and signed by the appraiser's supervisor.  The appraiser
and the appraiser's supervisor were together responsible for the
appraisal's accuracy, consistency, and conformance to company
policy, including a policy favoring fair and objective evaluations,
conducted "without regard to non-job-related criteria such as[]
race." 
Kodak intended the appraisal scores to be on a curve,
company-wide.  The compensation plan suggested that "performance
appraisal ratings for large groups of people (approximately 100 or
more) [should] average around the middle of the rating scale," but
acknowledged that factors such as the amount of turnover and the
percentage of long-term employees could influence the distribution
of ratings in any given group, particularly groups with a small
number of employees.  The compensation plan suggested that
appraisers attempt to validate their distribution of appraisal
ratings by rank-ordering employees in the same grade and job
category.  After the rank ordering was complete, supervisors would
finalize the preliminary ratings to agree with the rank-order
results.  However, these rank orderings were not discussed with
employees.  Supervisors were told that "[a]ppraisals verified
through this process [should be] communicated by referencing
performance relative to job expectations with no reference to the
rank-order process." 
Thomas's appraisals for 1988 and 1989 show that she was
performing at a high level.  In 1988, when she was responsible for
more than 400 machines, Thomas received seven 5 ratings and one 6
rating, for an overall rating of 5, which was a rating "appropriate
for individuals who not only achieve results and meet all
expectations on a regular basis, but who go beyond these
requirements from time to time."  Her supervisor's comments were
uniformly positive.  In 1989, when she was responsible for more
than 500 machines, Thomas received even higher ratings -- five 6s
and three 5s, for an overall rating of 6, which meant that her
"[p]erformance consistently exceed[ed] the requirements of the
position and [was] characterized by unusual initiative,
resourcefulness, and creativity."  Once again, the supervisor's
comments were uniformly positive.
 In 1989, Kodak created the position of Customer Support
Manager throughout its Office Imaging organization.  Thomas asked
to be considered for the Wellesley position, but was told that she
was not qualified.  Instead, the position went to Claire Flannery,
a former CSR who had been working as a division secretary.  As
Customer Support Manager, Flannery was responsible for supervising
the six Wellesley CSRs.  She remained in this position until 1993,
when both she and Thomas were laid off as part of a company-wide
reduction in force.
 The appointment of Flannery as Customer Support Manager
marked a significant downturn in Thomas's fortunes at Kodak. 
Although Thomas states that she and Flannery "were on a
professional basis," and Flannery denies having any problems with
Thomas's job performance, it appears that their working
relationship was strained.  Thomas alleges that Flannery treated
her differently from the other five CSRs, all of whom were white
(as were, in fact, all of the other CSRs during Thomas's thirteen
years in the Wellesley office).  
She describes a number of occasions on which she claims
Flannery unnecessarily damaged her professional standing with
customers.  For example, on the only occasion on which Flannery
accompanied Thomas to a customer training session in order to
observe Thomas's work, Flannery instead took over the session and
conducted the entire training herself.  On another occasion,
Flannery told Thomas the wrong time for a training session Flannery
had scheduled on site, and then refused to write to the customer,
who was upset, to explain why Thomas had been several hours late. 
(After Thomas proved to Flannery what had occurred by playing back
Flannery's voice mail message to her, Flannery did agree to call --
but not to write -- the customer.)  On a third occasion, Flannery
became quite angry and attempted physically to block Thomas from
leaving a CSR meeting which had been scheduled at the same time as
an important training session for one of Thomas's customers.  
 Thomas also provides evidence from which the inference
can be drawn that Flannery did not evaluate her skills fairly or
give her appropriate opportunities for growth and success.  For
example, Flannery did not travel with Thomas as she did with the
other CSRs in order to observe Thomas's interactions with
customers.  Flannery did not give Thomas the same type of
developmental opportunities available to other CSRs.  She
criticized Thomas for lack of computer skills, but then failed to
train Thomas when computer equipment became available.  After
Thomas was asked to prepare a presentation for a Kodak meeting,
Flannery told her that there would not be time for her
presentation, although time was found for a presentation by a white
employee on a less pressing topic.  Thomas was denied the
opportunity to apply for sales jobs, despite her good track record
in sales support activities.  Flannery expressly discouraged her
from applying for a management position in another division,
telling her she was not qualified because she did not yet have a
master's degree (Thomas was studying for one at the time), even
though none of the other managers in that position had a master's
degree or even a bachelor's degree.  Finally, Flannery went to
great lengths to prevent Thomas from meeting with Bill Cassidy, the
Regional Vice President for Office Imaging, to discuss advancement
opportunities, and became angry when Thomas nevertheless managed to
schedule an appointment.  
 Thomas's most significant and concrete allegation is
that Flannery gave her inaccurately low scores on her annual
performance appraisals.  For example, after receiving only 5s and
6s in 1988 and 1989, Thomas received a 2, four 3s, and a 4 from
Flannery in 1990, for an overall score of 3.  This was a below-
average rating, appropriate for employees who had "a need for
further improvement to achieve a middle rating [of 4]" or for
employees "whose overall performance has slipped from a higher
level."  Thomas's performance appraisal scores in 1991 and 1992,
while higher than her 1990 scores, were also inappropriately low,
in Thomas's estimation -- especially when compared to the higher
scores that Flannery gave to other CSRs.  Thomas presents specific
comparisons, discussed further below.
 Thomas was distressed by the 1990 performance appraisal
and refused to sign it.  She also refused to sign the 1992
appraisal, and signed her 1991 appraisal only "'out of  a joke,'
because it was a joke."  It is clear that Thomas disagreed with
Flannery's evaluations; more than that -- she found them
"insulting" and "shameful."  To the extent that it was possible for
her to compare salary and raises with other CSRs, however, she
would not have noted any obvious effects of the negative
appraisals, since neither her salary nor the raises she was given
during Flannery's tenure as Customer Support Manager differed
significantly from those of other employees.  She did complain
about the appraisals, both to Flannery herself and to others within
Kodak, including Bill Cassidy, the Regional Vice President for
Office Imaging, and Patti Weissinger, a Human Resources
Representative.  However, fearing retaliation from her new boss,
she did not file a formal charge against Flannery with the Human
Resources Department, and Kodak did not take any action in response
to her informal complaints. 
In January 1993, Kodak decided to reduce the number of
employees in its Office Imaging Division.  It selected employees
for layoff using a "Performance Appraisal Ranking Process ("PAR
process"), which produced a numerical score for each employee by
adding together the employee's overall performance appraisal score
for each of the three preceding years, after weighting the most
recent score by a factor of 25 and the second most recent score by
a factor of 5.  Thomas's PAR ranking, which was derived from the
three appraisals conducted by Flannery, was the second lowest of
the Wellesley CSRs.  Since Kodak had decided to cut two Wellesley
CSR positions, as well as the Customer Support Manager position,
Flannery, Thomas, and the lowest ranked CSR (Eileen Lavallee) were
laid off in March 1993.
                             II
In July 1993, four months after her layoff, Thomas filed
a charge of race discrimination with the Equal Employment
Opportunity Commission.  The EEOC issued a "right to sue" letter in
February 1996, and Thomas brought suit in May 1996. 
Although her complaint asserts a variety of race
discrimination claims, Thomas has focused primarily on Kodak's 1993
decision to terminate her, arguing that this decision was
discriminatory because it was based on discriminatory performance
appraisals conducted by Flannery from 1990 through 1992.   It is
the layoff which is the subject of this appeal.  Kodak moved for
summary judgment, contending that Thomas's claim was time-barred,
and in the alternative, that she had failed to present enough
evidence of racial animus to reach a jury under the First Circuit's
standard for showing disparate treatment once pretext has been
shown.  See Udo v. Tomes, 

, 12-13 (1st Cir. 1995)
(describing standard).  The district court found Thomas's claim
timely, but granted summary judgment to Kodak on the merits
argument.  See Thomas v. Kodak, 

, 133-38 (D.
Mass. 1998).  On appeal, Thomas argues that the district court
applied the standard incorrectly.  Kodak defends the district
court's application of the standard, while also continuing to argue
that Thomas's claim is time-barred.
                            III
We review the district court's grant of summary judgment
de novo, see Lennon v. Rubin, 

, 8 (1999), and will find
summary judgment appropriate if "the pleadings, depositions,
answers to interrogatories, and admissions on file, together with
the affidavits, if any, show that there is no genuine issue as to
any material fact and that the moving party is entitled to a
judgment as a matter of law,"  Fed. R. Civ. P. 56(c).    
                             A
We first address Kodak's argument that the district court
erred in not dismissing the action as untimely.  Title VII requires
aggrieved individuals to file a charge with the EEOC "within one
hundred and eighty days after the alleged unlawful employment
practice occurred."  42 U.S.C.  2000e-5(e)(1).  In a "deferral
jurisdiction," such as Massachusetts, this period is extended to
three hundred days.  See 42 U.S.C.  2000e-5(e)(1); Mohasco Corp.
v. Silver, 

, 814 n.16 (1980); Mack v. Great Atl. & Pac.
Tea Co., 

, 181-82 (1st Cir. 1989).  This relatively
short limitations period serves important interests.  "The
limitations period[], while guaranteeing the protection of the
civil rights laws to those who promptly assert their rights, also
protect[s] employers from the burden of defending claims arising
from employment decisions that are long past."  Delaware State
College v. Ricks, 

, 256-57 (1980).  In the Title VII
context, as in others, "the period allowed for instituting suit
inevitably reflects a value judgment concerning the point at which
the interests in favor of protecting valid claims are outweighed by
the interests in prohibiting the prosecution of stale ones."  

(internal quotation marks omitted) (quoting Johnson v.
Railway Express Agency, Inc., 

, 463-64 (1975)). 
Congress has made that value judgment.
However, Title VII's statute of limitations is not self-
executing.  The three hundred day (or one hundred and eighty day)
rule cannot be mechanically applied, because it is not possible to
pinpoint when the limitations period begins -- i.e., when the
plaintiff's claim "accrues" -- without first determining the date,
or the temporal boundary, of the "alleged unlawful employment
practice"  referred to in  2000e-5(e)(1).
The crux of Thomas's claim is that the 1990-1992
performance appraisals were tainted by Flannery's alleged racial
bias.  Thomas does not argue that Kodak's decision to utilize the
PAR process to determine who would be laid off was racially biased. 
See 

("[P]laintiff does not contend
that Kodak's . . . PAR layoff process[] [was itself]
discriminatory.").  As we understand her argument, she does not
allege that Flannery and other Kodak employees participated in a
conspiracy to oust her from her job because of her race, using the
performance appraisals and the PAR process as a thin cover for this
express discriminatory purpose.  Rather, she argues that the PAR
process was illegitimate under Title VII in a derivative way,
because its calculations were based on discriminatory appraisal
scores.  If racially biased scores were plugged in to the PAR
formula, she argues, then the ranking that came out must also be
biased.
According to Kodak, this argument can only mean that the
performance appraisals themselves constitute the "unlawful
employment practice" at issue under  2000e-5(e)(1), and thus that
any claim regarding the allegedly biased performance appraisals
must have accrued at the time the appraisals were conducted.  Since
even the most recent appraisal (presented to Thomas in May 1992)
was conducted more than three hundred days before Thomas filed her
complaint with the EEOC in July 1993, Kodak asserts that Thomas is
precluded from questioning the racial neutrality of any of the
three performance appraisals.  And without the ability to challenge
the appraisals upon which the PAR process was based,  Kodak notes,
Thomas's discriminatory termination claim collapses.
According to Thomas, the fact that the performance
appraisals were conducted more than three hundred days before she
filed her EEOC charge is irrelevant, because the adverse employment
decision that she is challenging -- the "unlawful employment
practice" under  2000e-5(e)(1) -- is Kodak's decision to lay her
off, and this decision indisputably fell within the statutory
period. 
Kodak's general policy stance makes sense:  given that
the central purpose of the statute of limitations is to protect
employers from stale claims, it cannot be true that plaintiffs have
an unfettered ability to reach back and litigate biased
evaluations.  Yet Thomas's position is also reasonable, since it
seems unlikely that Title VII permits employers to rely on
discriminatory evaluations as long as those evaluations were
conducted more than three hundred days before action is taken on
them.
Counsel did not call the court's attention to any First
Circuit caselaw on this point.  The district court, led to believe
that the First Circuit had not addressed this question, "look[ed]
for guidance elsewhere."  

.  The court
focused on the decision of the Third Circuit in Colgan v. Fisher
Scientific Co., 

(3d Cir.), cert. denied, 

(1991).  The facts in Colgan do closely resemble the facts here. 
Like Thomas, the Colgan plaintiff was fired after receiving a
negative performance review.  See 

 He filed a
charge of age discrimination with the EEOC within three hundred
days of his termination, but more than three hundred days after the
negative performance review.  See 

 The Third Circuit
determined that Colgan's claim was not time-barred because it
accrued at Colgan's termination, not when he received the negative
review.  See 

 The court concluded that "an alleged
unlawful employment practice, here the performance evaluation, must
have inflicted harm which was or should have been noticed, or it
will not have triggered the limitations period."  

Colgan's review stated that he failed to meet job requirements and
expressly warned that action would be taken unless his performance
improved.  See 

 Nonetheless, the court found that the
review did not produce the requisite notice of harm to trigger the
running of the statute of limitations, because "[t]he performance
evaluation had no immediate consequence on Colgan's employment,
such as a loss of seniority, nor did it alert Colgan to the
possibility that consequences would flow from it without an
opportunity for him to improve his performance."  

The Third Circuit based its Colgan analysis on a trilogy
of Supreme Court cases, which we summarize briefly here.  In United
Air Lines v. Evans, 

(1977), the Supreme Court held
that a flight attendant who was wrongfully forced to resign and
then later rehired could not challenge the airline's refusal to
grant her seniority as if there had been no break in her
employment, since the relevant wrongful act -- Evans's being forced
to resign -- occurred outside of the limitations period.  See 

 In Ricks, the Court held that the college's denial of
tenure, not a later discharge pursuant to that denial, was the
event that triggered Title VII's statute of limitations.  See

.  Finally, in Lorance v. AT&T
Technologies, 

(1989), the Court held that the Title
VII limitations period begins to run when an employer adopts a
seniority system, not when an employee is demoted pursuant to that
system.  See 

 Read together, this trilogy defines
a notice rule:  an employer action only triggers the running of the
statute of limitations if that action has concrete, negative
consequences for an employee, and the employee is aware or should
have been aware of those consequences.  See 

; see also 2 B. Lindemann & P. Grossman, Employment
Discrimination Law 1349 (3d ed. 1996) (observing that Supreme Court
cases "seem to establish a relatively simple 'notice' rule as to
when discrimination 'occurs'").      
 We think many aspects of Colgan's reasoning are correct. 
However, we find more direct guidance in a First Circuit case
decided several years before Colgan.  In Johnson v. General
Electric, 

(1st Cir. 1988), a black employee claimed
that he had been denied a promotion based on a special review
process that was intentionally designed to prevent him from
qualifying for promotion.  See 

 The employer contended
that the plaintiff's claim accrued when the review process was
first put into place, while the plaintiff insisted that the clock
did not start running on his claim until he was informed that he
had failed the review process and would not be promoted.  See 

court, relying on Ricks, held that the establishment
of the review process was the "unlawful employment practice" of
which the plaintiff complained, and therefore started the running
of the statute from that point.  See 

court
affirmed the dismissal on other grounds, it disagreed with the
district court's analysis, noting that the "question [is] . . .
whether a claim accrues when an employee is made subject to an . .
. evaluation . . . , or when that . . . evaluation is applied to
deny the plaintiff particular benefits or positions."  

chose the latter rule, holding that the notice standard is
met and the statute of limitations is triggered only if "the
implications [of the evaluation] have crystallized" and "some
tangible effects of the discrimination were apparent to the
plaintiff," i.e., if "the plaintiff is aware that he will in fact
be injured by the challenged practice."  

The decision was motivated, in part, by concerns about
ripeness.  The court noted that "it is far from clear whether
claims under Title VII and analogous statutes would be ripe for
adjudication until discriminatory systems [or evaluations] were
actually applied to plaintiffs in particular employment decisions," 
and observed that "[i]t is unwise to encourage lawsuits before the
injuries resulting from the violations are delineated, or before it
is even certain that injuries will occur at all."  

Johnson's notice standard serves multiple functions.  By
starting the limitations clock as soon as harm is noticed, or
should be noticed, the standard protects employers from stale
claims while also lessening the risk that employees will bring
unripe claims. The notice standard protects employees as well,
because it ensures that claims will not be foreclosed before
employees have had a fair chance to bring them.   
The notice standard in Johnson also resolves the apparent
contradiction between the two reasonable policy stances pressed by
the litigants here.  Thomas is correct that Title VII extends to a
neutral employer decision-making process that relies on
discriminatory evaluations.  But Kodak is also right that employees
do not have an unfettered right to reach back to challenge previous
evaluations.  The key is whether those evaluations had tangible,
concrete effects at the time they were conducted.  If the
evaluation did cause tangible, concrete harm, the notice standard
requires the injured employee to promptly bring suit to recover for
those harms.  Failure to do so will render any later claim
regarding those particular harms time-barred.  If the evaluations
did not have tangible, concrete effects at the time, no claim
regarding the discriminatory evaluations accrues.  Under the notice
standard, no claim will accrue until and unless the evaluations
result in a tangible injury.
This means that an employer could be exposed to Title VII
liability for harms stemming from discriminatory evaluations some
years after the evaluations were conducted, if the evaluations
first cause tangible harm to the employee at that later point. 
Kodak argues that this puts employers in an untenable position.  We
disagree, for three reasons.  
First, as a matter of practicality, employers are
unlikely to rely on "stale" evaluations.  The older the evaluation,
the less likely an employer would be to use that evaluation as the
basis of an employment decision. Employers' preference for more
recent evaluations is demonstrated dramatically by Kodak's own PAR
formula, which used only the latest three evaluations and then
weighted the most recent by a factor of 25.  
Second, the passage of time would affect employees as
well as employers.  An employee who sought to challenge an old
evaluation would bear the burden of proving that the evaluation was
discriminatory -- and the older the evaluation, the more difficult
that task would be. 
Third, the standard gives concomitant advantages to
employers.   It avoids forcing employees to "run to the EEOC" each
time they disagree with a performance evaluation.  As the district
court aptly observed:
  If we apply the time bar to plaintiff's [performance
appraisals], then we require a given plaintiff to file
EEOC charges successively for each performance
evaluation, informal feedback from a supervisor, or
office rumor, so long as these events -- even if
nonharmful in themselves -- might be informed by racial
animus and could someday contribute to a later, harmful
result.  This requirement would surely disrupt the
American workplace . . . .

.  Instead, the Johnson notice
standard allows employees to give employers the benefit of the
doubt, where employees suspect that an evaluation which has yet to
cause tangible harm might be tainted by bias, and encourages them
to try to solve the problem by proving their actual worth to the
employer.  And while the standard places an ongoing obligation on
employers to monitor their evaluation processes to ensure that they
are free from illegal bias, employers also benefit from the
opportunity to base personnel decisions on accurate, bias-free
evaluations.  In any event, this sort of obligation is foreseen by
Title VII, since the statute encourages the elimination of both
obvious and subtle forms of discrimination.  See McDonnell Douglas
Corp. v. Green, 

, 801 (1973).
In arguing against the application of the Colgan rule
(which, as explained above, resembles the First Circuit standard
announced in Johnson), Kodak has presented a two-stranded argument. 
The first strand is based on notions of fair notice to the
employer.  The second asserts that Thomas's timeliness argument
relies on an impermissible combination of the equitable tolling and
continuing violation doctrines.
Kodak would substitute for the accrual rule in Johnson a
rule that focuses on notice to the employer.  According to Kodak,
the Colgan accrual rule makes "bad law because it allows an
employee to introduce a time-barred incident where the employee has
never previously contended, even in an informal process afforded by
the employer, that the employee was subjected to unlawful
discrimination."  Appellee's Brief at 17.  According to Kodak, this
violates "the policy underlying Title VII [entitling] an employer
. . . to some kind of prior notice."  

not an irrational position, but Kodak cites no
authority for it.  The argument calls to mind the exhaustion rule
imposed by courts in the ERISA context, see Employee Retirement
Income Security Act, 29 U.S.C.  1001 et seq., which requires
employees to take full advantage of employer-internal appeals
processes before bringing suit to recover denied benefits, see
McMahon v. Digital Equipment Corp., 

, 40 (1st Cir. 1998)
(noting court-imposed exhaustion rule for ERISA benefit claims). 
But the analogy is inappropriate.  ERISA requires employers to
offer employees a speedy appeals procedure for denied benefits. 
See 29 U.S.C.  1133 (stating that employers must "afford a
reasonable opportunity to any participant whose claim for benefits
has been denied for a full and fair review by the appropriate named
fiduciary of the decision denying the claim"); 29 C.F.R.
2560.503-1 (imposing time limits).  Only after this process is
complete does the ERISA statute of limitations begin to run.  See,
e.g., Godfrey v. BellSouth Telecomms., Inc., 

, 759-60
(11th Cir. 1996); Martin v. Construction Laborer's Pension Trust,

, 1385 (9th Cir. 1991).  
In Title VII, by contrast, Congress chose not to impose
a particular employer-internal appeals procedure.  Furthermore, the
statute of limitations for a Title VII claim is not tolled while an
employee exhausts any internal remedy the employer has made
available.  See 

("[T]he pendency of a
grievance, or some other method of collateral review of an
employment decision, does not toll the running of the limitations
periods."); International Union of Elec. Workers v. Robbins &
Myers, Inc., 

, 236-37 (1976) (holding that the Title
VII statute of limitations is not tolled by a collective-bargaining
grievance procedure).  Thus, a court-imposed exhaustion requirement
would not work well in the Title VII context, because it would
place employees in the position of having to exhaust an internal
appeals process of uncertain length while also bringing suit within
three hundred (or one hundred and eighty) days of the employer's
allegedly wrongful act.        
Kodak's position bears an even closer resemblance to one
aspect of the approach first explicated in Stoller v. Marsh, 

(D.C. Cir. 1982), and followed by a number of other
courts, see, e.g., Hale v. Marsh, 

, 620 (7th Cir.
1986); Brown v. City of New York, 

, 169 (S.D.N.Y.
1994); Woolery v. Brady, 

, 669-70 (E.D. Mich.
1990).  The Stoller court held that a plaintiff could challenge an
employer's wrongful reliance on discriminatory evaluations, even
where the plaintiff has acknowledged that the employer's reliance
itself was not intentionally discriminatory.  See 

("Otherwise an organization could separate illegal motive
from decisionmaking responsibility, contrary to the principle that
Title VII applies to the employer as an organization.").  This is
essentially the same result that Johnson permits.  See 

, 137 (citing this portion of Stoller).  
But the Stoller court added a second part to this rule: 
  [A]fter preparation of employee evaluations, the
employing organization may protect itself from Title VII
liability by establishing procedures to allow employees
to screen their personnel files and to remove damaging,
discriminatory information. . . . If established
procedures have given an employee a reasonable
opportunity to inspect the supervisory evaluations in his
or her file, to challenge allegedly inaccurate materials,
and to have such materials corrected or removed, and if
the organization gives its employees adequate notice that
these rights may be exercised, then it may rely in good
faith on such evaluations in making subsequent employment
decisions without violating Title VII.

(footnotes omitted).  Kodak's "notice to
the employer" argument could refer to this second half of the
Stoller rule:  Kodak apparently wants to be able to rely on
Thomas's performance appraisals, because in its view Thomas had a
reasonable opportunity to inspect the appraisals and to challenge
inaccuracies through the "Open Door" appeals process. 
We can imagine advantages of the Stoller rule for both
employers and employees.  Employers, of course, would be able to
limit their liability.  Employees could conceivably benefit as
well, if the rule increased the likelihood that employers would
implement effective review procedures.  But we decline to adopt
this rule -- which was, in any event, implicitly rejected in
Johnson -- because we think any possible advantages would be
outweighed by two serious risks.  One is the obvious risk that
employers would have carte blanche to rely on discriminatory
appraisals as long as they offered their employees an "Open Door." 
Avoiding the first risk would entail a second:  courts would need
to interpret strictly the concept of "[adequate] procedures to
allow employees to screen their personnel files and remove
damaging, discriminatory information."  This would in turn entail
a degree of judicial scrutiny and interference with day-to-day
employer operations that employers would be unlikely to welcome and
which courts interpreting Title VII have tried to avoid.  See 
Hidalgo v. Overseas Condado Ins. Agencies, Inc., 

, 337
(1st Cir. 1997) (noting courts' reluctance to "'sit as super
personnel departments'" (quoting Mesnick v. General Elec. Co., 

, 825 (1st Cir. 1991)).
The second strand of Kodak's timeliness argument is the
claim that both Thomas and the district court have relied on an
illegitimate "hybrid" of the equitable tolling and continuing
violation doctrines.  Thomas disavows any such reliance.  We make
several observations in response to Kodak's argument.  
First, the First Circuit doctrine of equitable tolling is
simply inapplicable to these facts.  Some courts permit tolling of
the statute of limitations if the plaintiff knew of a harm but not
of its discriminatory basis.  See 2 Lindemann & 

.  But our approach to equitable tolling is narrower; First
Circuit law permits equitable tolling only where the employer has
actively misled the employee.  See 

(noting
that the First Circuit's "narrow view" of equitable tolling reaches
only "active deception" (internal quotation marks and citations
omitted)); Jensen v. Frank, 

, 521 (1st Cir. 1990).
There is no allegation here that Kodak actively attempted to
mislead Thomas about her performance appraisals.
Second, there is no need to apply the continuing
violation doctrine.  Commentators have labeled this doctrine "the
most muddled area in all of employment discrimination law," 2
Lindemann & 

, and some courts have gone so
far as to conclude that the entire doctrine is misguided and
unnecessary, see, e.g., Moskowitz v. Trustees of Purdue Univ., 

, 282 (7th Cir. 1993).  We need not enter into that debate
here.  We describe the First Circuit's continuing violation
doctrine with only as much detail as is necessary to show that it
is inapplicable to Thomas's claim.  
The First Circuit has recognized two different types of
continuing violations: systemic violations, which "ha[ve] [their]
roots in a discriminatory policy or practice . . . [that] itself
continues into the limitation period," DeNovellis v. Shalala, 

, 307 (1st Cir. 1997) (quoting 

),
and serial violations, which are "composed of a number of
discriminatory acts emanating from the same discriminatory animus,
[with] each act constituting a separate wrong actionable under
Title VII," id.; see also 

.  The systemic
violation doctrine is clearly inapposite because Thomas has not
alleged that Kodak has "an overarching policy or practice" of
conducting discriminatory evaluations.  

.  
Kodak apparently believes that Thomas is inappropriately
attempting to shoe-horn her claim into the serial violation
category.  This is not so.  Whether a serial violation exists does
depend in part on a type of notice standard similar to the one in
Johnson.  See Sabree v. United Bhd. of Carpenters, Local No. 33,

, 402 (1st Cir. 1990) (stating that if an earlier event
had "'the degree of permanence which should trigger an employee's
awareness and duty to assert his or her rights,'" then it is not
substantially related to the later event, and therefore cannot form
part of a continuing violation (quoting Berry v. Board of
Supervisors of L.S.U., 

, 981 (5th Cir. 1983))).  This
similarity may be the source of Kodak's argument.  But the serial
violation doctrine, like the systemic violation doctrine, stands as
an exception to the accrual rule recognized in Johnson.  The
purpose of this exception is to permit suit on later wrongs where
a wrongdoer would otherwise be able to repeat a wrongful act
indefinitely merely because the first instance of wrongdoing was
not timely challenged.  See D. Laycock, Continuing Violations,
Disparate Impact in Compensation, and Other Title VII Issues, 49
Law & Contemp. Probs. 53, 55 (1986) (comparing the application of
the continuing violation doctrine in employment discrimination
cases to its application in an antitrust case, Hanover Shoe, Inc.
v. United Shoe Machinery Corp., 

(1968), and observing
that "[o]bviously, United should not be able to continue its
illegal conduct forever because no one challenged it during World
War I"); J. MacAyeal, The Discovery Rule and the Continuing
Violation Doctrine as Exceptions to the Statute of Limitations for
Civil Environmental Penalty Claims, 15 Va. Envtl. L.J. 589, 615-22
(1996) (describing the history and purpose of the continuing
violation doctrine); see also 

(discussing
proper remedy).
In the Title VII context, the continuing violation
doctrine applies where "a number of discriminatory acts emanat[e]
from the same discriminatory animus, [with] each act constituting
a separate wrong actionable under Title VII."  

.  Sexual harassment, failure to promote, and pay inequity cases
often fall into this category.  Plaintiffs in these cases
experience harm from each employer act (e.g., a harassing comment,
a denied promotion, or a smaller paycheck), for which they could
recover under Title VII.  Thus, under the Johnson standard, each
act could trigger the running of the statute of limitations.  The
continuing violation doctrine ensures that these plaintiffs' claims
are not foreclosed merely because the plaintiffs needed to see a
pattern of repeated acts before they realized that the individual
acts were discriminatory.      
Thomas's claim presents a different question:  not
whether she is permitted to bring suit after many repeated harms,
but rather whether harm cognizable under Title VII existed at the
time she first received her performance appraisals.  This question
is more appropriately addressed under the Johnson notice standard.
Applying that standard, we consider when Thomas's
performance appraisals first caused her tangible, "crystallized"
harm.  According to Kodak's compensation plan, appraisal scores
were intended to affect salary levels and determine "who should be
promoted, transferred, demoted, terminated, laid off, and re-
employed."  It is not clear whether Kodak informed employees of
this intention.  But even if employees were familiar with the
intended possible uses of the scores, it appears that the effects
listed in the compensation plan remained abstract -- mere
possibilities, not certainties.
  This was particularly true in Thomas's case.  Flannery's
appraisals evaluated Thomas's performance as only average or below
average.  But the appraisals did not specify that Thomas was to
suffer an immediate consequence for her alleged performance
failures, much less that they would mechanically lead to her being
laid off.   Cf. 

, 1419-20 (finding notice
lacking even where an evaluation explicitly warned that adverse
actions would be taken if the employee's performance did not
improve).  She was not told that layoffs were impending and that
her scores placed her at high risk for layoff.  Even if she did
have this information, it would fall short of the "crystallized"
implications required under 

, since being
at risk for layoff (due to a low ranking relative to other
employees) and actually being selected for layoff are two quite
different things.
Kodak emphasizes the fact that Thomas refused to sign two
of the three performance appraisals.  According to Kodak, her
refusal to sign "demonstrates that she was dissatisfied with [the
appraisals] for some reason" and makes it "evident that she then
believed she had been treated unfairly."  Appellee's Brief at 18
n.6.  A trier of fact could, but need not necessarily, conclude
that Thomas had notice of the appraisals' possible racial bias as
soon as they were presented to her.  However, this fact does not go
to the relevant question under Johnson's notice rule:  whether
Thomas had notice of immediate, tangible consequences of her poor
scores.  Kodak's argument pertains to notice of bias rather than
the notice of harm required under Johnson.  But notice of bias
alone, absent harm, is clearly not sufficient under Johnson.
We hold that the performance appraisals Thomas received
in 1990, 1991, and 1992 did not trigger the statute of limitations
in  2000e-5(e)(1) at the time they were presented to Thomas,
because they did not initially have any crystallized implications
or apparent tangible effects.  Thomas's low appraisal scores first
resulted in a concrete injury in 1993, when they led to her layoff.
Thus, under Johnson, her layoff marked the point of accrual.  Since
Thomas filed a charge with the EEOC within one hundred and eighty
days of her layoff (the shortest possible period imposed under
2000e-5(e)(1)), her claim is timely. 
                             B
The second question on appeal is whether a reasonable
jury could find, based on Thomas's evidence, that Flannery
discriminated against Thomas because of her race when she assigned
Thomas's performance appraisal scores.  See 42 U.S.C.  2000e-
2(a)(1).  To reach an answer, we follow the three stages of the
"familiar burden-shifting framework," Mulero-Rodriguez v. Ponte,
Inc., 

, 673 (1st Cir. 1996), that was first outlined in
McDonnell Douglas Corp. v. Green, 

(1973), and further
explained in Texas Department of Community Affairs v. Burdine, 

(1981).  The three stages can be summarized as follows:
  First, the plaintiff[] must establish a prima facie case
that [plaintiff] (1) was within a protected class;  (2)
met [the employer's] legitimate performance expectations;
(3) was adversely affected;  and (4) was replaced by
another with similar skills and qualifications.  Once
[plaintiff] do[es] so, the burden shifts to [the
employer] to produce a valid and nondiscriminatory reason
for the dismissal.  In the final stage, the burden shifts
back to the plaintiff[] to show that [the employer's]
stated reason for [plaintiff's] dismissal was false and
but a pretext for discrimination. 

(citations omitted). 
We focus particularly on the final stage of the McDonnell
Douglas/Burdine framework, because we agree with the district
court's careful analysis of the first two stages.  See 

.  Thomas has established a prima facie case of
discrimination by showing that she is a member of a protected class
who met Kodak's legitimate performance expectations and was laid
off, while Kodak retained persons outside the protected class.  See

a presumption that Kodak unlawfully discriminated
against her.  See St. Mary's Honor Ctr. v. Hicks, 

, 506
(1993) (citing 

).  To counter this
presumption, Kodak must "articulate some legitimate,
nondiscriminatory reason" for its action, McDonnell 

, that is, allege "reasons for its action which, if
believed by the trier of fact, would support a finding that
unlawful discrimination was not the cause of the employment
action,"  

(citing 

-
55).  At this second stage, the framework imposes on the defendant
only a burden of production.  The burden of persuasion remains at
all times with the plaintiff.  See 

(citing 

).  Kodak met its burden of production -- and eliminated
the presumption that it had discriminated -- by contending that its
layoff decision was based solely on racially neutral performance
appraisal scores.  See 

.
At the third stage of the McDonnell Douglas/Burdine
framework, the ultimate burden is on the plaintiff to persuade the
trier of fact that she has been treated differently because of her
race.  See 

.  This burden is often broken
into two separate tasks.  The plaintiff must present sufficient
evidence to show both that "the employer's articulated reason for
laying off the plaintiff is a pretext" and that "the true reason is
discriminatory."   Udo v. Tomes, 

, 13 (1st Cir. 1995)
(citing Smith v. Stratus Computer, Inc., 

, 16 (1st Cir.
1994)).  For expository convenience, this court has sometimes
labeled these two findings "pretext" and "plus" and has referred to
the First Circuit rule as a "pretext-plus" standard.  See, e.g.,
Mullin v. Raytheon Co., 

, 699 (1st Cir. 1999)
(describing the difference between the "federal 'pretext-plus'
standard and the Massachusetts 'pretext-only' standard"); Dichner
v. Liberty Travel, 

, 30 (1st Cir. 1998) (describing the
"'pretext plus' approach"). 
Applying this standard, the district court found that
Thomas had met the first part of her burden.  It determined that a
jury could find that the appraisal scores were objectively unfair
or in some sense skewed against Thomas.  Accordingly, the court
found that "[b]ased only on the evidence presented so far, a
reasonable fact finder could determine that Flannery's reasons for
lowering plaintiff's [performance appraisal] ratings were
pretextual."  

.  As explained below,
we agree.
But the court also found that Thomas did not meet the
second, "plus" part of her burden:  she failed to present evidence
to create a genuine issue of fact about whether the allegedly
unfair scores were due to her race and not some other factor.  See

 The court compared the evidence presented by
Thomas, which in the court's view amounted only to evidence of "an
unwelcoming office environment," "conspicuously unfair treatment"
by Flannery, and "a personality conflict between Flannery and
plaintiff," 

evidence presented by plaintiffs in other
First Circuit cases in which plaintiffs had survived summary
judgment:
  In all of these cases, the plaintiff alleged at least one
piece of evidence that explicitly referred to plaintiff's
membership in a protected class, and stated or implied
that this membership was or would soon adversely affect
plaintiff's employment prospects.  By contrast, in cases
where plaintiff fails to make any connection between
adverse employment actions and membership in a protected
class, summary judgment is granted to the employer, and
judgment is affirmed.

 The court granted summary judgment to Kodak because it
found that Thomas had failed to "make any connection" between
Flannery's actions and her own membership in a protected class. 

the court found that Thomas failed to present
any evidence that Flannery had expressly linked the low scores to
Thomas's race and that Thomas's evidence of pretext was not
"flagrant" enough to support a finding of discrimination on its
own.  

The district court's analysis mischaracterizes the
plaintiff's burden at the third stage of the McDonnell
Douglas/Burdine framework.  To clarify the actual nature of that
burden, we make a number of points.
First, the labels "pretext" and "plus" must be used with
great care.  Although it uses the label "plus," the First Circuit's
"pretext-plus" standard "does not necessarily require the
introduction of additional evidence" beyond that required to show
"pretext," i.e., evidence showing that the employer's articulated
reason is false.  

.  Plaintiffs may use the
same evidence to support both conclusions, "provided that the
evidence is adequate to enable a rational factfinder reasonably to
infer that unlawful discrimination was a determinative factor in
the adverse employment action."  Rodriguez-Cuervos v. Wal-Mart
Stores, Inc., No. 98-1732, 

, at *7 n.5 (1st Cir. June
11, 1999); see also 

.  
A corollary is that there can be no mechanical formula at
the third stage of the McDonnell Douglas/Burdine framework.  "The
strength of the prima facie case and the significance of the
disbelieved pretext will vary from case to case depending on the
circumstances.  In short, everything depends on the individual
facts."  Woods v. Friction Materials, Inc., 

, 260 n.3
(1st Cir. 1994).  Other circuits have highlighted the same point: 
"The sufficiency of the finding of pretext to support a finding of
discrimination depends on the circumstances of the case."  Fisher
v. Vassar College, 

, 1338 (2d Cir. 1997).  Moreover,
"it is difficult, if not impossible, to say in any concise or
generic way under what precise circumstances . . . an inference [of
discrimination from a showing of pretext] will be inappropriate." 
Aka v. Washington Hosp. Ctr., 

, 1294 (D.C. Cir. 1998). 
Because discrimination, and discrimination cases, come in many
different forms, a case-by-case analysis is always necessary. 
There can be no rigid requirement that plaintiffs introduce a
separate "plus" factor, such as a negative employer comment about
the plaintiff's protected class, in order to prove discrimination. 
Otherwise, the McDonnell Douglas/Burdine framework would no longer
serve the purpose for which it was designed:  allowing plaintiffs
to prove discrimination by circumstantial evidence.  See Smith v.
F.W. Morse & Co., 

, 420-21 (1st Cir. 1996). 
Our second point is related.  That "everything depends on
the individual facts" means we need to look carefully at the
particular type of claim that Thomas is bringing.  As we understand
Thomas's argument, she alleges a more subtle type of disparate
treatment than the type often used to exemplify the operation of
the McDonnell Douglas/Burdine framework.  She does not argue that
Kodak has articulated a false reason for her layoff (for example,
excessive tardiness) in order to disguise the actual, unrelated 
reason (her race) -- what one might describe as a "truth versus
lies" claim -- rather, she challenges the racial neutrality of the
proffered reason itself.  The latter type of challenge is also
cognizable as a form of disparate treatment:  if an employer
evaluates employees of one race less favorably than employees of
another race who have performed equivalently, and if race, rather
than some other factor, is the basis for the difference in
evaluations, then the disfavored employees have been subjected to
"discriminat[ion] . . . because of  . . . race."   2000e-2(a)(1). 
The "pretext-plus" label may seem inapposite for this
type of claim, since the word "pretext" could suggest a conscious
lie.  But issues of pretext should not be confused with the issue
of whether there has been discrimination "because of race."  As the
Supreme Court has used the term "pretext," the term provides "no
justification for assuming . . . that those employers whose
evidence is disbelieved are perjurers and liars."  

.  According to the Hicks Court, "[t]o say that the company
which in good faith introduces such testimony, or even the
testifying employee himself, becomes a liar and a perjurer when the
testimony is not believed, is nothing short of absurd."  

The ultimate question is whether the employee has been
treated disparately "because of race."   This is so regardless of
whether the employer consciously intended to base the evaluations
on race, or simply did so because of unthinking stereotypes or
bias.  See Robinson v. Polaroid Corp., 

, 1015 (1st
Cir. 1984) (noting that plaintiffs in a disparate treatment case
can challenge "subjective evaluations which could easily mask
covert or unconscious race discrimination on the part of
predominantly white managers"); Sweeney v. Board of Trustees of
Keene State College, 

, 179 (1st Cir.) (permitting a
challenge to a decision process in which "bias may often be
unconscious and unexpressed"), vacated on other grounds, 

, 24 (1978), aff'd after remand, 

, 114 (1st Cir.
1979) (noting again that sex discrimination includes "the practice,
whether conscious or unconscious, of subjecting women to higher
standards of evaluation than are applied to their male
counterparts").
The Supreme Court has long recognized that unlawful
discrimination can stem from stereotypes and other types of
cognitive biases, as well as from conscious animus.  Indeed,
discussing age discrimination in Hazen Paper Co. v. Biggins, 

(1993), the Court characterized employer decisions "'based
in large part on stereotypes unsupported by objective fact,'"  

(quoting EEOC v. Wyoming, 

, 231 (1983)), as
"the essence of what Congress sought to prohibit in the ADEA," 

(emphasis added).  And the prohibition against this form of
disparate treatment is not limited to the context of age
discrimination.  The Court has interpreted Title VII as
"'prohibit[ing] all practices in whatever form which create
inequality in employment opportunity due to discrimination on the
basis of race, religion, sex, or national origin,'" County of
Washington v. Gunther, 

, 180 (1981) (quoting Franks v.
Bowman Transp. Co., 

, 763 (1976)), and has noted that
this includes "the entire spectrum of disparate treatment of men
and women resulting from sex stereotypes," 

Angeles
Dept. of Water & Power v. Manhart, 

, 707 n.13 (1978)). 
Stereotypes or cognitive biases based on race are as incompatible
with Title VII's mandate as stereotypes based on age or sex; here
too, "the entire spectrum of disparate treatment" is prohibited.
The role of such stereotyping has been discussed most
thoroughly in that branch of disparate treatment law developed
apart from the McDonnell Douglas/Burdine framework and known as the
Price Waterhouse framework.  See Price Waterhouse v. Hopkins, 

, 239-58 (1989); 

(comparing the
two frameworks).  The district court in Price Waterhouse found that
the employer, an accounting firm, had discriminated against Ann
Hopkins by permitting stereotypical attitudes about women to play
a role in its decision not to invite her to become partner.  See
Hopkins v. Price Waterhouse, 

, 1118-19 (D.D.C.
1985).  On appeal, the employer contended that it could not be
liable for disparate treatment because "Hopkins did not prove
'intentional' discrimination on the part of the [decision-making]
Board, but only 'unconscious' sexual stereotyping by unidentified
partners who participated in the selection process."  Hopkins v.
Price Waterhouse, 

, 468 (D.C. Cir. 1987), aff'g in part
and rev'g in part 

.  The D.C.
Circuit squarely rejected this argument:
  In keeping with [Title VII's remedial] purpose, the
Supreme Court has never applied the concept of intent so
as to excuse an artificial, gender-based employment
barrier simply because the employer involved did not
harbor the requisite degree of ill-will towards the
person in question.  As the evidentiary framework
established in McDonnell Douglas makes clear, the
requirement[] of discriminatory motive in disparate
treatment cases does not function as a "state of mind"
element, but as a method of ensuring that only those
arbitrary or artificial employment barriers that are
related to an employee or applicant's race, sex,
religion, or national origin are eliminated.  

(footnotes omitted); see also Lynn v. Regents of the
Univ. of Cal., 

, 1343 n.5 (9th Cir. 1981) ("[W]hen
plaintiffs establish that decisions regarding . . . employment are
motivated by discriminatory attitudes relating to race or sex, or
are rooted in concepts which reflect such attitudes, however
subtly, courts are obligated to afford the relief provided by Title
VII.").
The court of appeals found it unsurprising that the
disparate treatment doctrine focuses on causality rather than
conscious motivations, since "unwitting or ingrained bias is no
less injurious or worthy of eradication than blatant or calculated
discrimination."  

.  If the plaintiff has
shown that she was treated less favorably because of her gender,
the court said, "the fact that some or all of the partners at Price
Waterhouse may have been unaware of that motivation, even within
themselves, neither alters the fact of its existence nor excuses
it."  

Court upheld this portion of the D.C.
Circuit's decision.  See Price 

(plurality opinion); 

(White, J., concurring in the
judgment); 

272, 277-78 (O'Connor, J., concurring in the
judgment); see also Hopkins v. Price Waterhouse, 

, 969
(D.C. Cir. 1990) (affirming district court decision after remand
from Supreme Court) (noting that this portion of 

, was upheld by the Supreme Court).  The Price Waterhouse
plurality explained:  "In the specific context of sex stereotyping,
an employer who acts on the basis of a belief that a woman cannot
be aggressive, or that she must not be, has acted on the basis of
gender."  Price 

.  The plurality added
that "[b]y focusing on [the plaintiff's] specific proof, . . . we
do not suggest a limitation on the possible ways of proving that
stereotyping played a motivating role in an employment decision." 

The concept of "stereotyping" includes not only simple
beliefs such as "women are not aggressive" but also a host of more
subtle cognitive phenomena which can skew perceptions and
judgments.  Price Waterhouse highlighted one such phenomenon:  the
tendency of "unique" employees (that is, single employees belonging
to a protected class, such as a single female or a single minority
in the pool of employees) to be evaluated more harshly in a
subjective evaluation process.  See 

see also
Villanueva v. Wellesley College, 

, 131 (1st Cir. 1991)
(discussing possible relevance of plaintiff's status as the sole
employee within a protected class).  Other types of biased thinking
are also widely recognized.  See generally  L. H. Krieger, The
Content of Our Categories:  A Cognitive Bias Approach to
Discrimination and Equal Employment Opportunity, 47 Stan. L. Rev.
1161, 1186-1217 (1995) (describing varieties of cognitive bias); C.
R. Lawrence III, The Id, the Ego, and Equal Protection:  Reckoning
with Unconscious Racism, 39 Stan. L. Rev. 317, 328-44 (1987)
(discussing forms of racism); see also D. Charny & G. M. Gulati,
Efficiency-Wages, Tournaments, and Discrimination: A Theory of
Employment Discrimination for "High-Level" Jobs, 33 Harv. C.R.-C.L.
L. Rev. 57, 77-78, 83 (1998) (noting the inability of the market
alone to correct the effects of cognitive biases and stereotypes).
We make one final general point about the third stage of
the McDonnell Douglas/Burdine framework:  in the context of a
motion for summary judgment, courts must be very cautious about sua
sponte finding non-discriminatory reasons for apparently disparate
treatment.  Although the presumption of discrimination has dropped
out of the case by the third stage of the McDonnell Douglas/Burdine
framework, its rationale remains relevant:  discrimination, rarely
explicit and thus rarely the subject of direct evidence, may be
proven through the elimination of other plausible non-
discriminatory reasons until the most plausible reason remaining is
discrimination.  See 

n.8 (noting that "the
allocation of burdens" in a Title VII case "is intended
progressively to sharpen the inquiry into the elusive factual
question of intentional discrimination").  In this case, for
instance, the plaintiff has argued that her appraisal scores were
unfairly low because of her race, while the defendant has argued
that the scores were objectively fair.  By finding that plaintiff
has presented enough evidence to suggest that her scores were
unfairly low, and then speculating that a personality conflict
could be the source of that unfairness, the district court
erroneously adopted a third view.  See 

-38.  No party in this case has argued that a personality
conflict explained any disparity in the scores assigned by
Flannery, and the district court should not, at summary judgment,
have replaced Kodak's articulated explanation sua sponte with its
own.  It may be that a jury (or judge, if no jury trial is claimed)
would find at trial that a personality conflict rather than racial
bias is the best explanation for any difference in scores.  But the
question at summary judgment is not which of the possible
explanations is most convincing; it is whether the plaintiff has
produced enough evidence to raise a genuine issue of fact regarding
her explanation.  
Having clarified the actual nature of the plaintiff's
burden under McDonnell Douglas/Burdine, we now consider whether
Thomas has met that burden.  Because the employer here has
articulated a non-discriminatory reason for its action, the initial
presumption of discrimination created under McDonnell
Douglas/Burdine has dropped from the case.  See 

; 

.  This leaves the plaintiff with the
ultimate burden of proving discrimination.  She must produce
evidence to create a genuine issue of fact with respect to two
points:  whether the employer's articulated reason for its adverse
action was a pretext and whether the real reason was race
discrimination.  See 

.  In other words, she must
produce evidence to permit a reasonable jury to conclude both that
disparate treatment occurred and that the difference in treatment
was because of race.  The same evidence may support both showings. 
See 

portion of Thomas's third-stage burden is to 
produce sufficient evidence to show that she was not evaluated on
the same terms as her colleagues, but rather was evaluated more
harshly than the other, non-minority CSRs.  The district court
found that the plaintiff met this portion of her burden:  she
produced enough evidence to allow a reasonable factfinder to
determine that Flannery's reasons for assigning her scores were
"pretextual."  

.  We agree.
One strong piece of evidence is the sharp drop in
Thomas's overall score (from 6 in 1989 to 3 in 1990) after Flannery
became her supervisor.  While the drop in score would not
demonstrate uneven treatment if Flannery were simply a "tough
grader," this does not appear to have been the case.  In 1990, when
Thomas received a 3, all of the other CSRs whom Flannery graded
received either a 5 or a 6.  Furthermore, over her three years as
supervisor Flannery gave CSRs other than Thomas a number of 7s in
individual categories, a score that earlier supervisors
characterized as extremely rare or even unheard of. 
The drop in Thomas's score from 1989 to 1990 would also
fail to demonstrate uneven treatment if it merely reflected a
higher standard imposed upon Thomas due to her grade change in
1989.  According to Kodak, "the higher the wage grade and salary,
the more Kodak expected from the employee."  Appellee's Brief at
34.  But Thomas presents deposition testimony from other
supervisors suggesting that there was no such rule, and the scores
that Flannery assigned to other Wellesley CSRs seem to bear this
out:  one CSR who received a grade change in 1991 received a 5 in
1992 after receiving a 6 in 1991 (a drop of only one point rather
than three after the grade change), and another CSR who received a
grade change in 1991 received a higher score the following year
(from a 4 in 1991 to a 5 in 1992).
Thomas makes other specific comparisons as well.  The
strongest of these concern her scores for "quantity of results,"
which was the most objective of the categories appraised, since it
was based directly on the number of machines and installations for
which a CSR was responsible.   Thomas's scores appear low when
compared to the scores for quantity given to other CSRs.  In 1990,
for instance, Thomas received only a 3 for quantity of results --
a below average score -- for managing 730 machines and 110
installations, while another CSR, also in the K6 grade, received a
5 for quantity in 1991 for managing far fewer machines (504) and
fewer installations (81).  Kodak argues that the comparison is
"unfair and analytically unsound" because the other employee had
been at the K6 grade for many years, and "[s]ustained performance
at a higher grade counts for more than a relatively shorter period
of performance at the higher grade."  Appellee's Brief at 35.  This
argument might persuade a jury to discount the value of this
comparison, but it does not render the comparison meaningless at
summary judgment, where Thomas need only produce evidence
sufficient to support her contention that her scores were lower
than those given to similarly performing non-minority CSRs.  See
Molloy v. Blanchard, 

, 91 (1st Cir. 1997) (noting that
"'[e]xact correlation is neither likely nor necessary'" and 
stating that "'[t]he test is whether a prudent person, looking
objectively at the incidents, would think them roughly
equivalent.'" (quoting Dartmouth Review v. Dartmouth College, 

, 19 (1st Cir. 1989))).  
In addition to the performance appraisals themselves,
Thomas presents other evidence to show that Flannery treated her
differently than she treated the other CSRs.  According to Thomas,
Flannery used flimsy grounds to prevent her from delivering an
important presentation at a Kodak meeting, refused to provide her
with computer training, failed to allow her appropriate
developmental opportunities, and failed to evaluate her accurately
on the basis of her interaction with customers (since she never
accompanied Thomas on site in order to observe that interaction, as
she did with the other CSRs). 
On the whole Thomas has presented enough evidence to
survive summary judgment on the question of disparate treatment. 
Some of Thomas's scores, and particularly the 1990 scores, appear
lower than one would expect given her apparent performance level. 
A reasonable jury, examining the evidence Thomas has submitted,
could conclude that Thomas was evaluated more harshly than other,
non-minority CSRs supervised by Flannery.  
Kodak argues that, even if true, this would not matter,
because Thomas would have been laid off anyway.  Even if Thomas had
received the highest possible score (7) in 1990, Kodak says, her
PAR ranking would not have changed:  because the PAR formula so
heavily weighted the two more recent scores, Thomas would still
have had the second lowest PAR score among the CSRs and would still
have been selected for layoff.   There are two flaws with this
argument.  First, it ignores the natural implications of a finding
that the 1990 scores were unfairly low.  To the extent that
Thomas's 1990 scores were inappropriately low, a factfinder could
infer that her later scores -- although higher than her 1990 scores
-- were inappropriately low as well (and indeed, they were
consistently lower than the scores given to the other CSRs, as
demonstrated by the performance appraisals that Kodak has produced
for 1991 and 1992).   Second, Kodak's argument ignores the fact
that performance appraisal scores were expressly on a curve. 
According to Kodak's compensation plan, scores were to be
"validated" by reference to an informal supervisor ranking of
employees in the group.  Thus, if Thomas's scores had gone up,
other CSRs' scores might have gone down.  Whether accurate scoring
of Thomas in relation to the other CSRs would have saved her from
layoff is something Thomas must prove at trial.  But she has
created a genuine issue with respect to this question.  Not only
has she produced evidence which suggests that Flannery treated her
differently and graded her more harshly than the other CSRs, she
has also produced evidence to support the claim that her earlier
evaluations were a more accurate appraisal of her performance: (1)
her promotion in 1989 to the K6 grade, which was reserved for
"outstanding" CSRs "servicing the largest and/or most sensitive
accounts," (2) her salary, which was higher than her seniority
would predict, presumably reflecting the fact that Kodak had, over
the years, found her performance level to be comparatively high,
and (3) the positive comments from customers and from other
supervisors, which suggest that she was in fact one of the better
CSRs.
 Our inquiry does not end with this finding, since even
the most blatant unfairness cannot, on its own, support a Title VII
claim.  "Title VII does not grant relief to a plaintiff who has
been discharged unfairly, even by the most irrational of managers,
unless facts and circumstances indicate that discriminatory animus
was the reason for the decision."  

; see also
Rodriguez-Cuervos, 

, at *5.  Instead, our focus
shifts to the question whether Thomas has produced sufficient
evidence to support a finding that the apparent disparity in
treatment was due to her race. 
As we emphasized above, Thomas can meet her burden with
respect to this question using the same evidence used to show
unequal treatment.  The district court's opinion can be read to say
that at this stage the plaintiff must produce what has been called
"direct" evidence,  such as racially explicit statements.  That is
not so.  There is no requirement that a plaintiff, having shown
differential treatment and pretext, must present direct, "smoking
gun" evidence of racially biased decisionmaking in order to
prevail.  Where the disparity in treatment is striking enough, a
jury may infer that race was the cause, especially if no
explanation is offered other than the reason rejected as
pretextual.  That is the case here, since Kodak argues that
Thomas's scores were objectively fair, not that they were unfair
due to a personality conflict, as the district court surmised.
In fact, Thomas does present additional evidence on this
point.  She describes incidents and situations which suggest that
Flannery had a general disregard for her professional abilities and
status.  It also appears, from the picture that Thomas paints (and
which Kodak does not dispute) that Flannery was at times
inappropriately upset or angry with Thomas, to the point of
behaving unprofessionally.  This, in turn, suggests that she did
not respond neutrally to Thomas.  A jury might reasonably infer
from Thomas's description of these incidents that Thomas's race was
an issue for Flannery and that Flannery's evaluations of Thomas
were affected by some form of conscious animus or less conscious
bias.
Our assessment of the evidence would be quite different
if Thomas had been one of several black employees supervised by
Flannery, some of whose scores were not low relative to those of
the non-minority CSRs.  But that was not the case.  Thomas was the
only black CSR, and one can infer from the evidence that she was
also the only CSR who was evaluated unfairly. Given this, it is
reasonable to infer that race played a determinative role in the
evaluation process -- especially since there is also other evidence
that Flannery treated Thomas poorly.  Indeed, the very fact that
Thomas was the only black CSR at the Wellesley office may have
increased the likelihood that she would be evaluated more harshly. 
See Price 

; 

.
In the end, Thomas is in a different position than a
plaintiff who merely "disagree[s] with [the employer's] assessment
of her relative performance."  Compton v. GTE Gov't Sys., Corp.,
No. 93-11383, 

, at *6 (D. Mass. Dec. 7, 1995), cited
in 

; see also Blick v. Pitney Bowes
Management Servs., Inc., No. 93-11573, 

, at *5 (D.
Mass. Dec. 25, 1995) (noting that a plaintiff's mere disagreement
with the employer's "assessment of his work attitude . . . is not
sufficient to create a triable federal discrimination claim"),
cited in 

.  Thomas has presented
"evidence from which the trier of fact reasonably could conclude
that [her] abilities and qualifications were equal or superior to
employees who were retained."  Goldman v. First Nat'l Bank of
Boston, 

, 1119 (1st Cir. 1993).  It is not an
"improbable inference[]" or "unsupported speculation," Medina-Munoz
v. R.J. Reynolds Tobacco Co., 

, 8 (1st Cir. 1990), to
further conclude, based on the evidence available at summary
judgment, that Kodak failed to recognize this fact because it
relied on an evaluation procedure that was tainted by racial bias.
                              IV
Thomas's claim is timely, and it is based on sufficient
evidence to permit a reasonable jury to find that Flannery
discriminated against Thomas because of her race when she assigned
the performance appraisal scores that led to Thomas's layoff. 
Accordingly, we reverse the district court's grant of summary
judgment and remand for further proceedings consistent with this
opinion.  
Costs are awarded to appellant.