Data Tampering / Sabotage / Fabrication

Published by the National Science Foundation, Office of Inspector General on 1997-10-20.

Below is a raw (and likely hideous) rendition of the original report. (PDF)

                                        CLOSEOUT FOR M95060027

a   -   -
         officer in the

                 This case came to OIG on June 26, 1995, when ~             r        . NSF's program
                                                           rograrn in the Division of 1_-__

                                                                that research supported by his program
         had been falsified. OIG later learned that the subject of the allegation was Mr. ,-
         formerly a computer programmer at                         Attached is the OIG investigation
         report, including its appendices. Also attached is the notice of misconduct in science
         determination and proposed debarment that NSF sent to Mr.               which explains NSF's
         adjudicative decision. These documents explain the actions subsequently taken by OIG and
         NSF in this case.

        cc: AIG-0, IG
                            NATIONAL SCIENCE FOUNDATION
                               4201 WILSON BOULEVARD
                              ARLINGTON, VIRGINIA 22230

                                              October 15, 1997


    Mr.   1             -

    Re:    Notice of Misconduct in Science Determination
           and Proposed Debarment
    Dear Mr. B:
    This letter and the attached investigative report serve as formal
    notice that the National Science Foundation (NSFI proposes to
    debar you from directly or indirectly obtaining the benefits of
    Federal research grants for a period of three years. A person
    who is debarred will be excluded during the period of debarment
    from Federal financial and non-financial assistance and benefits
    under non-procurement Federal programs and activities. See 45
    CFR P620.110, S620.200. In addition, you will also be prohibited
    from receiving any Federal contracts or approved subcontracts
    under the Federal Acquisition Regulations (FAR) at 48 CFR Subpart
    9.4 for the period of this debarment. 45 CFR §620.110(c).
    Debarment of an individual is effective throughout the executive
-   branch of the Federal Government.

    NSFts decision to propose debarment is based upon a referral from
    our Office of Inspector General (OIG). The ~oundation's
    administrative record indicates that you were formerly a computer
    programmer at .-                 While at -,     you falsified
    data by wilfully designing computer programs for an NSF-funded
    scientific research project to alter the experimental research
    results and make the results confirm hypotheses that researchers
    on the project sought to test.
    Under NSF1s misconduct in science and engineering regulations,
     ami is conduct' is defined to include "fabrication, falsification,
    plagiarism, or other serious deviation from accepted practices in
    proposing, carrying out or reporting results from activities
    funded by NSF . . . l t 45 CFR §689.l(a). Your falsification of
    data constitutes falsification and is a serious deviation from
    accepted practices within the scientific community. We therefore
    conclude that you committed misconduct in science.
     Peaulat o w Basis for Debarment
    In deciding what response is appropriate when misconduct is
    found, NSF must consider the seriousness of the misconduct;
    whether it was deliberate or careless; whether it was an isolated
    event or part of a pattern; and whether the misconduct affects
    only certain funding requests or has implications for any
    application for funding involving the subject of the misconduct
    finding. 45 CFR §689.2(b). Severe misconduct is a cause for
    debarment because (i) it affects the integrity of NSF research or
    education programe, see 45 CFR §620.305(b); 45 CFR 0689.1(e), and
    (ii) it is a 'cause of ao serious or compelling a nature that it
    affects the present responsibility of a personN, 45 CFR
    4620,305 (d).
    Debarment must be for a period commensurate with the seriousness
    of the cause. 45 CFR 9620.320(a). The burden of proof is on the
    government to establish facts which justify debarment by a
    preponderance of the evidence. 45 CFR §689.2(d), §620.314(c).
    According to the Investigative Report, you deliberately engaged
    in falsification. You intentionally designed computer programs
    that falsified research data in order to confirm an untested
    research hypothesis. You purposefully wrote the computer
    programs in a manner that would avoid detection by your research
-   colleagues. In addition, when your colleagues became suspicious
    about the errors in your programs, you lied to them about the
    cause of the errors and continued to write programs that
    falsified the data.
    Falsification of data is an extremely serious offense because it
    distorts the scientific record. The scientific record is the
    foundation for all future research. Both the Federal Government
    and the scientific community have a vital interest in protecting
    the integrity of the research process. To undo the damage you
    caused to the research, your       colleagues had to expend
    considerable amounts of time and expense rewriting the data
    analysis and collection programs and replicating the'ir research
    findings from two years of research.
    The purpose of debarment is to protect the Federal Government
    from dealing with individuals who lack present responsiblity. In

    your response to the OIG, you state that you have a dissociative
    disorder which causes you to have selective memory and differing
    motivations in differing mental states. In your opinion, this
    disorder caused you to engage in the data falsification. YOU
    acknowledge that you will need additional professional treatment
    before you can ensure that the misconduct that occurred at
    will not happen again. My duty is to ensure that the Government
    is fully protected from a reoccurrence of the misconduct.
    Accordingly, I am proposing that you be debarred for three years.
     -       Procedures Governina-Proposed~ebarment/ScientificMisconduct
             Under our regulations, you have 30 days after receipt of this
             notice to submit -- in person, in writing, or through a
             representative -- information and argument in opposition to the
             proposed debarment. 45 CFR 9620.313 (a) During this 30-day
             period you may also review the attached ~nvestigativeReport and
             submit comments or rebuttal. 45 CFR 9689.8 (c)(11, 9689.1 (e)
             Comments submitted within the 30-day period will receive full
             consideration and may lead to revision or withdrawal of the
             Investigation Report or of the recommended disposition.
             Any response should be addressed to Lawrence Rudolph, General
             Counsel, National Science Foundation, 4201 Wilson Boulevard, "~oom
             1265, Arlington, VA 22230. If you have any question~i please
             contact Mr. Rudolph at (703) 306-1060. We are attaching a copy
             of the Foundation's regulations on Nan-Procurement Debarment and
             Misconduct in Science and Engineering.


                                                Joseph Bordogna.
                                                Acting Deputy Director

--           Attachments (4)
             Investigative Report
             Nonprocurernent Debarment Regulations
 ,       .   FAR Regulations
             Misconduct in Science Regulations
                             NATIONAL SCIENCE FOUNDATION
                                 4201 WILSON BOULEVARD
                                ARLINGTON, VIRGINIA 22230

                                          December 11, 1997

            omcE OF THE



     Re:    Debarment
     Dear Mr. -:
     On October 15, 1997, the National Science Foundation (NSF) sent
     you a Notice of Misconduct in Science Determination and Proposed
     Debarment in which NSF proposed to debar you from directly or
     indirectly obtaining the benefits of Federal grants and contracts
     for a period of three years from the date of the letter. NSF's
     debarment action is based upon your severe misconduct in science
     discussed in detail in the Notice of Proposed Debarment. In that
     Notice, NSF informed you that you had a period of 30 days within
     which to respond to the proposed, debarment.
    NSF did not receive a response from you. Accordingly, you are
    debarred until October 15, 2000. You will be excluded during the
-   period of debarment from Federal non-procurement Federal programs
    and activities with the Executive branch of the Federal
    Government, as either a'participant or principal. See 45 CFR
    Sections 620.110 and 620.200. Nonprocurement transactions
    include grants, cooperative agreements, scholarships,
    fellowships, contracts of assistance, loans, loan guarantees,
    subsidies, insurance, payments for specified use, and donation
    agreements. In addition, you will also be prohibited from
    receiving any Federal contracts or approved subcontracts under
    the Federal Acquisition Regulations (FAR) at 48 CFR Subpart 9.4
    for the period of this debarment. 45 CFR Section 620.110(c).
     Under NSF's regulations, you have 30 days to submit an appeal of
    'this decision, in writing, to the Director of the Foundation.
     45 CFR Section 689.8 (c)(1)(iii). Any appeal should be addressed
     to the Director, National Science Foundation, 4201Wilson
     Boulevard, Arlington, Virginia, 22230. (A copy of the applicable
     regulations was enclosed with the previous notice of Proposed
     Debarment. )
If you have any questions, please feel free to contact our
General Counsel, Lawrence Rudolph, at (703) 306-1060.


                         Joseph Bordogna
                            Acting Deputy Director
                       SCIENCE AND ENGINEERING


        The Ofice of Inspector General (OIG) has determined that Mr.
subject), formerly a computer programmer at -(the
by willfully designing the computer programs he wrote for an ~ S sponsored
                                                                        F        research project
to alter experimental research results and make them confiirm hypotheses that researchers on
the project sought to test. This conclusion is based on our analysis of evidence collected by the
University. OIG recommends that NSF find that the subject committed misconduct and take
the following actions as a final disposition in this case. The subject should be told that NSF
has made a finding of misconduct and should receive a letter of reprimand from the NSF
Deputy Director. NSF should enter into a voluntary exclusion agreement with the subject for a
period of three years or, as a .alternative, debar the subject from participation in federally
funded projects for a suitable period of time. The subjixt has indicated that he is agreeable to

a voluntary exclusion agreement such as that described in this report.

                              THE UNIVERSITY'S INVESTIGATION

        This case came to OIG on ~ u h 26,e 1995, w         h     e     n      , NSF's program
officer in the - ' P r o g r a m                             of the Division ofB-1
                          informed us of an allegation that research supported by his program
                                  - 4
had been falsified. Two days later,                           the University dean responsible for
misconduct inquiries and investigations (the Dean), informed us that, at the conclusion of a
University misconduct inquiry, the subject had admitted to falsifying data, whereupon the
University had made a finding of misconduct and termhated the subject's employment.' The
Dean told us that the University planned to investigate further to verify that the subject had
confessed to the full extent of his data falsification, and she promised to send OIG a report of
the University's investigation. On July 1, 1996, the Dean sent OIG a letter stating the
University's conclusions in this case. Attached to the letter were nine appendices, some
divided into alphabetically labeled parts, containing the evidence on which those conclusions
were based.

       The Dean's letter and the relevant portions of the appendices to it comprise Appendix 1
to OIG's report; except for Appendix 2, which is the subject's reply to OIG's draft report,
appendix numbers and letters below refer to the appendices to the Dean's letter. For
convenience, we have attached to our report only those appendices to the Dean's letter to
which we refer in the text. The remaining appendices areavailable on request.

 The University's misconduct procedures (point #5) permit a finding of misconduct at the close of an inquiry
under these circumstances. See Appendix IX.
 --     The subject worked as a computer programmer on research funded by NSF awards
               (Award #la),
 These awards supported studies of how humans
                                                 (Award #lb), and

 experimental results" indicating that there were sm
         s        e inaccuracies fit an-pattern:
                                                                                 w ward #2).2
                                                                     In 1992, the PIS published
                                                              research participants alkteiy

 human subjects, some studim focused on people suffering f r 6 m d i s e a s e .
                                                                              .   .

                                                 ordinarily in the C computer language, to
                                                                  and to analyze the data
                                                                  the subject in the research
 group as follows ("Amplification of My Views, " p. 1, Appendix 1.A):

         [The subject] has developed the real-time data collection programs that have been
         standardly used in our laboratory for several years. He has also developed the more
         difficult data analysis programs. . . . He has provided programs to many researchers
         working on o                                        [student #l] ,                  [the
                                              [Student # 21,                [the co-PJJ,
                       [Student #3], me, and several undergraduate honors students
                                  [the undergraduate], and .                He is an excellent
         programmer, with a thorough knowledge of many computer languages and many
         computer architectures.

   Dt. 0of the                            Department (the PI) was the principal investigator for Award #la, which
 was awarded to the University. Dr. -(the                     co-PI)was the principal investigator for Award #lb,
 which was awarded to the University o-f                where the CQ-PIis a member of the faculty. These were
 identical proposals and were proposed to NSF as a collaborative effort, with funds administered by both
                                        " -
 universities. The projects were entitled                                           These awards funded the PIS'
 research until January, 1994. Award #2, entitled                         was an award to the University under the
 direction of the PI and the co-PI. This award became effective on January 1, 1995.
 'The PIS also conducted contemporaneous, related research on
 supported by National Institutes of Mental Health (NLMH) award  '!'mm-          in animals. This research was
                                                                                        . The subject served as a
 programmer for this project. The University concluded that the subject did not commit misconduct in connection
 with this research, and the Department of Health and Human Services Office of Research Integrity, which is
 responsible for handling misconduct cases involving NIMH awards, accepted this conclusion.
.4 +                                                   (1992).   9
 degree and i;ow affiliated with the Department of             at
                                                                     the laboratow. She has since wmoleted her
                                                                           She was also married to the subject, a
 fact that she did not reveal to any other members of the research group until August 30, 1995, when she informed
 the PI. The University concluded that there is no evidence that she was involved in the data falsification, and OIG
 agrees with this conclusion. See below.
       scientific meetings, such as the Society for
       participant in our weekly-eetings
        He is a creative designer of transducers and interfaces. . . .' The isometric pressure
        transducer, and associated interface and software that he has developed, produces much
        more reliable data than a standard switch. . . .

       He . . . deeply cares about the substance of the          search. He spends vacations at
                                                                                He is a major
                                                   and a collaborator in many of our research
       projects. Because we work closely as a group, it is difficult to identify the source of
       many ideas, but it is my impression that his ideas have been at least as important as
       those of any other participant in our group. If asked to order the quality of intellectual
       contributions to our research in the last few years, I would probably list [the subject],
       [the recent graduate], and me (in that order). His technical contributions to our
       research are clearly enormous.

The subject was listed in the proposal for award #2 as a co-author of two published articles and
a third article that had been submitted for publication. This evidence indicates that the subject
played an important scientific role in the project and had a clear responsibility to understand
thoroughly the ethical requirements of that role.

        In a February 26, 1995, memorandum, "Facts about which we are certain" (Appendix
I.A), Student #2, Student #3, and the co-PI described numerous errors in computer programs
and experimental apparatuses developed by the subject. An undated second document,
"Additional Findings" (Appendix LA), reports information and conclusions "discovered after
February 27, 1995" @. 1). These documents were submitted to the PI, who concluded that the
subject had made certain technical errors and that at least one such error might be interpreted
as intentional and, hence, as misconduct in science. Along with Student #2, Student #3, and
the co-PI, the PI took steps to gather evidence of the subject's actions and to assess the extent
of his errors. According to the PI'S March 20, 1995, memorandum to the Dean ("Report of
Concern," p. 1, Appendix I.A), the PI then discussed "the general nature of the problem" with
the subject and "asked him many specific technical questions." Although the PI concluded that
"there is no evidence of scientific misconduct," he reported this matter to the Dean so that she
might appoint a disinterested committee to deterhine whether there was substantial evidence
indicating that the errors resulted from misconduct in science by the subject. To head the
committee, the Dean, at the PI'S suggestion, named Dr.                     - a colleague in the
PI'S department who, aceording to the PI, possessed app-                     lolowledge,
familiarity with the people involved, and no prior knowledge of the incidents in question.

        The committee reported its conclusions to the Dean on May 10, 1995 (Appendix LC,
"Professor flll(ll(bLaboratory - Report of Findings"). After interviewing relevant members
of the research group and examining computer codes and other pertinent documents, the
committee concluded @p.2-3):

       The strong argument against [the subject] is the sheer number of errors. While it is
       difficult for us to accept that such numerous and pervasive errors could go undetected,
         nevertheless, we are unable to determine that any of his actions were deliberate,
         intentional, or with motivation to deceive. Therefore, we cannot conclude that fraud,
         dishonest behavior or scientific misconduct has occurred. We base this conclusion on
         the following:

                  1) [The subjectl's explanations and answers to all of our questions were
                  reasonable and plausible. . . .

                  2) It is extremely difficult to determine a motive for [the subject] to fabricate
                  data or to act dishonestly. As a technical assistant he did not have the pressures
                  experienced by graduate students and faculty to achieve certain results and/or to
                  develop data in support of a specific hypothesis. . . .

                 3) me subjectl's health problems [the report makes reference earlier to
                 C r o b l e m s that had caused the subject to be hospitalized] have been
                 documented and could indeed produce a state of depression that would lead to
                 negligence of the type that has occurred.

                 4) During all of our questioning [the subject] made no attempt to cover up
                 errors or blame someone else or be uncooperative in any way.

        Therefore we conclude that the errors represent a severe case of incompetent
        performance. . . .

         On April 1, 1995, the research group reported some of its results in a talk at the
                        Association meetings in          m6Dr. of-                the Department
  university                             (the repli&tor) heard the talk and attempted to replicate

the reported results. She informed the PI on M a y 27, 1995, that she was unable to do so. A
discussion of the experimental methods in the two laboratories revealed no significant
differences. The replicator sent the PI data on one of her subjects, and Student #2 then
analyzed the data and found no evidence of b  the
                                                - f                    This led Student #2 to

reexamine the analyses from the PI'S laboratory and eventually to discover evidence of
falsification, which he reported to the PL7

        On June 23, 1995, the PI sent a memorandum to the Dean reporting that the subject had
confessed to falsifying data and explaining how a data collection program written by the
subject had falsified data in                             experiments. The memorandum (Appendix 11,
p. 1) states:

 [Student #2, the recent graduate, the subject, and the PI], "

                          t h m an August 31. 1995, memorandum fmm the PI to the Dean ('Repon," p.4.
Appendix VII).
       [The subject] gave [Student #2] a program to present    a     of varying a    n      d to
       record the location of a line to which subjects pointed with a cursor controlled by a
       mouse. . . . The program introduced an artifact in the data, a small systematic error in
       the reported location of the cursor that was functionally related to the '  o     f the
       stimulus. The code that produced that artifact apparently was put into one of the
       standard C libraries. . . . The code had no other purpose than to create this small
       systematic error.

       By writing this code and giving it to [Student #2], [the subject] created a data collection
       procedure that necessarily would produce falsified data. The nature of the small
       systematic error was to produce 0                     tasks that would have been of
       particular theoretical interest.

The memorandum goes-on to describe how the subject programmed the computer to
accomplish the falsification and what steps the PI proposed to take to ascertain the extent of the
falsification and to correct the scientific record.

        On June 24, 1995, the subject wrote a "voluntary" and "unsolicited" statement
outlining his method of falsifying data, surveying the extent of his falsifications, and explaining
the motive for his action (Appendix IV).

       With regard to method, the subject stated (p. 1):

       In all cases the mechanism of falsification was to alter the standard C libraries with
       special pattern matching code that looked for patterns in the data, and once these were
       found, modified the data by adding a small artifact to the real data. The source code
       for the original libraries was available in the package that the C compiler came in. The
       purpose of performing data falsification W way was to allow the actual source code to
       be distributed to members of the lab and also to allow me to write programs in a normal
       manner. This served the purpose of deceiving myself as well as others. . . .

Thus either the subject or another researcher on the project would write a source computer
program that was not corrupted. The source program would send input to the library and
direct the library to perform specified calculations. The library, which the subject had
corrupted, then performed calculations that were different fiom the calculations the project's
scientists expected. The library next returned the results of these corrupted calculations as
output to the source computer program. The source program, in turn, relied on these results in
its subsequent operations. Researchers on the project had copies of the source program and
could examine it for errors, but the library was part of the computer system software and was
therefore effectively invisible to the researchers. ,
               All versions of programs used to collect data for the   -
       With regard to the extent of the falsification, the subject stated (p. 1):

                                                                                  estimation task
       were altered. They contain various routines that were placed into various standard C
       library functions to add a small, nonlinear, categorical artifact to all data that were of
       nearly linear form with a few outliers. While there were many false or modified
       functions that were contained in the libraries that I modified, the one that had the
       primary effect was located in the "fprintf" function.

                 All data collected and report& in [the undergraduatel's Thesis and the

      m?              portion of [another studentl's Thesis contained no modifications. The data
       co ect subsequent to that also contained no modifications, except for the most recent
       program used by [Student #1] (This was unintentional, it was just that it was compiled
        with corrupted libraries). All analyses performed on data collected from "normal"
       subjects afier september 1993 weri falsified or altered in some way. AII (IY
       -analysis                   pro@ams that were written after September of 1993 are
       suspect. . . .

               All raw animal data is [sic] clean and unaffected by any tampering. . . . A
       subset of recent analysis programs was unintentionally compiled with the corrupted
       libraries and may not be trustworthy, though the exact effect of this is not yet
       known. . . .

The subject reported that all other data collection and analysis programs for both human and
animal data were "clean from any intentional defect" (p.2).

       With regard to the motive for his actions, the subject explained as follows (p.2):


              The primary motivation to modify these data was to insure that all results
       obtained in experiments on "normal" subjects that could be potentially used with
      -patients              in a short test period, such as an hour or less, would show the
        same pattern (or                       as that obtained by [the undergraduate] and myself
        in her honor's
                    yl1-ht                        important to me that the m
      t         a          s        k show this result because it could provide strong evidence
       that the nonlinearities i n w t a were in some way 'central" to the s t e m
       and not related to any peripheral n o i n e a r i t i e s . If all "normal"
       subjects showed this pattern then the differing pattern seen in the -tien@
       would have to be seen as resulting from the disease.

       Circumstances surround in^ Motive:

                  I have a dissociative disorder that causes me to have selective memories and
          differing motivations in differing mental states. I have been told by several
                                       @at this could be the result of the
     taht-                                a switch from one state to another can be "triggered" by
          external events. In September o                                               s JO who acted
          v ~ r ymuch like -who

       -!-               isease not too 1
          was e trigger. At the time
          look like the original normal data collected by [the undergraduate] and myself and thus
         causing all differences between normal subjects' data and 4 u b j e c t s ' data to
          look as though the resulted from the disease then our results could lead to an early
         detection test for -disease                    and perhaps some idea of the mechanism of the
         disease and possibly a treatment or a cure. Upon reflection and much therapy, I realize
         that what I was trying to do was to esadicue'r                       in the hopes of (I
                                                                                               ( II 1
               I realize that this motive does not make any real causal sense, but it is the truth.
       I seemed to think that I was doing some sort of good by modifying the programs and
       data the way I did. I am fairly certain that I believed that the pattern that [the
       undergraduate] saw in the                                            was the "true" pattern
       and I don't think I trusted                                       riment "properly" to
       find that pattern.

        In closing his statement @.3), the subject emphasized that he acted alone in falsifying
data, that no one else was aware of his falsifications or tried to conceal them, that he was
"very deeply sorry" about what he had done, and that he would accept "any reasonable
sanctions" that university authorities imposed.

        On June 28, 1995, the Dean sent a letter to the subject (see Appendix V) informing him
that the University President had found that the subject committed misconduct and had
accepted the Dean's recommendation to terminate the subject's employment on June 30, 1995,
on grounds of 'gross misconduct."

       On August 3, 1995, the Dean appointed D-r.                     of the University's
Department of                 (the monitor) to monitor the efforts of the PIS and their research
group to assess the extent of the damage done by the subject's falsification and to determine
whether scientific findings of studies in which the subject participated could be replicated. The
Dean's charge to the monitor is in Appendix V1.A. The monitor's September 5, 1995, report
    (Appendix V1.C.) explains that the research group chose to discard all data collection and
.   analysis programs designed by the subject.' The monitor explained @p. 2-3) that correcting the
    programs would not have been feasible:

             At least some of [the subjectl's programs clearly had mechanisms designed to hide their
             data fabrication procedures from attempts at systematic testing. Discovering these
             mechanisms in other, as yet untested, programs may not be straightforward, and
             certainly would be very time consuming. . . . [The subject] explained that his
             falsifications were done by modifying the source code for the standard C libraries that
             came with the C compiler; the source code for original data analysis programs was thus
             not obviously corrupted and could be distributed freely to lab personnel. Programs
             compiled by [the subject] with the corrupted libraries, however, produced fabricated
             data. [The subject] himself seems to be unsure which programs may have been
             compiled with corrupted libraries. Apparently all of these libraries and compilers were
             on [the subjectl's personal computer, and he claimed that all were lost or erased.
             Despite [the subjectl's claims, [Student #2] and [Student #3] stated to me that they
             suspect [the subject] may not have actually used this mechanism for falsifying his
             programs, although they have not been able to document specifically an alternative
             method. . . . mt was quite difficult to evaluate and document the nature [ofi the
             programs' behaviors.

    The monitor, summarizing his conclusions concerning preliminary results of studies the
    research group had undertaken to replicate findings that the subject had helped to generate,
    said that "the results of these replications so far provide no evidence for the fabrication or
    falsification of previous studies on       but "definitive evidence for the falsification of a l l
    previous human                          esults" (p.6). This summary accords with the subject's
    descriptione -ofh t                       falsification, although, as the monitor repeatedly
    cautioned, the subject's precise statements about the falsification are not reliable. In particular,
    the monitor expressed doubt about the date the subject gave for the onset of his miscond~ct.~
    The PI'S August 31, 1995, report on the replication efforts (Section II.F., p.6; the report is in
    Appendix VII) reaches conclusions similar to those of the monitor. Additional reports written
    in November, 1995, by the two PIS separately and in collaboration provide further evidence
    supporting these conclusions (see Appendix V1II.B)

             On August 30, 1995, the recent graduate revealed to the PI that she was married to the
    subject. There is no evidence that before this date anyone else in the research group was
    aware that the two were married. The PI informed the Dean of the marriage in a September 2,
    1995, memorandum ("Addendum to Report of August 31 ," Appendix VII). He noted that this
    information did not alter his view that the subject was solely responsible for the misconduct.

    * "It is impossible or impractical to determine the precise start time or scope of [the subjectl's misconduct,
    although it dates at the latest to June, 1993." ("Report of research observations in [the Pu's laboratory," p.6,
    Appendix V1.C) The subject had suggested September, 1993, as the date he began falsifying the data (See the
    subject's statement in Appendix IV).
 The PI's memorandum went on to discuss preliminary replication data that tended to confirm
 results that the recent graduate reported in her dissertation and in a published article.

                                        OIG'S ANALYSIS

        For NSF to make a finding of misconduct, a preponderance of the evidence must show
that the subject committed culpable acts with a culpable state of mind. The pattern of errors in
the subject's computer programs and the subject's confession that these errors were part of a
deliberate effort to create data sets and data analyses falsely supporting certain scientific
hypotheses provide incontrovertible evidence that the subject falsified data.

         Above all, the scientific community values truth. Deliberate distortion of the scientific
record to create evidence that falsely tends to confirm a hypothesis is a direct affront to the
community's defining value commitment. The subject's actions pervert the main purpose for
which NSF funds research-to advance scientific knowledge. NSF's Regulation on Misconduct
in Science and Engineering defmes misconduct in part as a "serious deviation from accepted
practices in . . . carrying out or reporting results from" research (45 C.F.R. $689.1(a)(l)).
The regulation specifically mentions only three examples of misconduct, and one of these is
falsification. There is no question that the subject's actions seriously deviate from accepted

         OIG believes that the subject acted willfully. His falsifications were animated by a
coherent purpose, albeit one that did not initially make sense to colleagues and investigators
unacquainted with the subject's claimed psychological disorder. He chose to modify the
standard C libraries so that he could distribute programs to other researchers that did not show
traces of his falsifications, which were hidden within subroutines not ordinarily accessible to
the project's researchers. When colleagues raised suspicions about numerous, uncharacteristic
errors in the subject's programs, the subject lied convincingly to investigators and continued to
write programs that falsified data. The subject's years of research experience and deep
involvement with the substance of the research should have impressed upon him how
profoundly wrong it was to distort scientific data. His actions indicate that he engaged in
falsification knowing that it was wrong to do so.

       OIG agrees with the University that there is no evidence to indicate that anyone in
addition to the subject was responsible for falsifying data. In all instances, the subject had both
the ability and the opportunity to execute the falsification without assistance. In particular, we
conclude that there is no evidence to implicate the subject's wife in his misconduct. However
odd it may appear that the subject's wife did not reveal her personal relationship with the
subject earlier, the pattern of falsification tends to c o n f m the subject's account of his motive
and is inconsistent with the idea that either he or his wife falsified data to advance the wife's
scientific career. The falsified data did not involve work distinctively associated with his wife,
nor did the falsification appear to benefit her in any special way. Independent replication has
tended to confirm the research findings with which the subject's wife was most closely
associated. The wife's (and the PI's) initial reluctance to believe that the subject was guilty of
falsification strikes us as no more than a natural reluctance to believe that a close colleague
could commit misconduct. It is not evidence of complicity in the misconduct, any more than
the University's initial conclusion that the subject was guilty only of "a severe case of
incompetent performance" suggests that it was complicit. In addition, there is no evidence that
the subject's wife acted in any way to hinder or obstruct the University's efforts to investigate
this case.

OIG9sConclusion Regarding Misconduct in Science

       OIG concludes that a preponderance of the evidence supports the conclusion that the
subject committed misconduct in science by willfully falsifying data, and we recommend that
NSF make a finding to that effect.

                               OIG'S RECOMMENDED DISPOSITION

        Under 5 689.2@)of NSF's misconduct in science and engineering regulation, upon
making a finding of misconduct, NSF, in determining what actions it should take, must
consider the seriousness of the misconduct. This includes considering the state of mind with
which the subject committed misconduct and whether the misconduct "was an isolated event or
part of a pattern." We have explained why the subject's actions are a serious deviation from
accepted practices and hence misconduct; this section explains OIG's recommended actions in
light of our assessment of the seriousness of the subject's misconduct, i.e., our assessment of
how serious this instance of misconduct is in relation to other instances.

        We believe that the subject's action, measured by the scientific significance of what was
falsified, was extremely serious. His falsification did not merely alter a few data points or
strengthen the case for a hypothesis that was already well supported with genuine data.
Rather, his falsification was designed to c o n f i i a hitherto untested scientific hypothesis. It
prompted the PIS to draw conclusions that were sufficiently significant to include in their
progress report to NSF and present at a scientific conference.

         This is also a serious case of falsification if we measure seriousness by the labor
required to undo the damage to the PIS' research program and the human and financial cost
incurred. Several researchers on the PIS' project each spent several months rewriting data
analysis and collection programs and replicating the major fmdings of two years of research to
assess the effects of the subject's actions. A scientist at another institution wasted time and
effort trying to replicate falsified fmdings. Graduate students in the PIS' research group spent
countless hours that could have been devoted to productive research ferreting out errors that
the subject willfully incorporated into his computer program^.^

 We note that two of the three graduate students whose research was affected by the subject's falsification chose
not to return to the PI's laboratory and to interrupt or terminate graduate study. (Student #1 took a leave of
absence and Student #3 requested a separation from the University [PI's "Report," Appendix VII, p.2 and p.31.
        The subject's account of the motive for his actions suggests that NSF needs to take
special precautions to safeguard the government's interests. By the subject's own admission,
the impetus for his misconduct is outside his rational control and makes him resistant to
ordinary moral suasion. Moreover, colleagues unaware of the subject's mental condition
would be, as they were in this case, understandably slow to detect future misconduct because it
would likely not be instnunentally motivated (i.e., it would not likely flow from a "rationaln
motive). The subject's technical ability with computers enables him to hide misconduct from
colleagues without similar ability. All of these factors suggest that the subject's actions need to
be carefully monitored.

        We believe that the subject willfully engaged in extremely serious misconduct and
ought to be excluded from employment in projects funded by federal awards for a minimum of
three years. Our draft report recommended that NSF seek to enter into a three year voluntary
exclusion agreement with the subject to be followed by a two year monitoring period in which
the government enlists the subject's cooperation in protecting the government's interests (see
45 C.F.R. 5 620.105). As part of the agreement governing this two year period, we
recommended that the subject stipulate that, before accepting employment on a federally
sponsored project, he would inform the head of the project and the federal official responsible
for the award that NSF found he had committed misconduct in science; that, as a result, he
agreed to exclude himself from participation in federally sponsored projects for three years;
and that his misconduct was caused, at least in part, by a long standing psychiatric disorder.

        In his response to our draft report (Appendix 2), the subject stated that he "would most
certainly be willing to enter into the voluntary agreement that is described in the draft report,"
but would need more detail as to its exact provisions. He went on to describe the information
that he currently provides employers to enable them to protect themselves from possible

       In the light of the subject's response, we have reaffiied and finalized the findings and
conclusions of our draft report.

         NSF's Debarment and Suspension regulation (45 C.F.R. 5 620.320 (a) (I)), provides
that "debarment shall be for a period commensurate with the seriousness of the cause(s)* and
that, though debarments generally "should not exceed three years[,] where circumstances
warrant, a longer period of debarment may be imposed." The monitoring we propose is fully
consistent with these provisions. The mental condition that apparently caused the subject's
misconduct is, by his own account, deep seated and long standing. It will not be neutralized
because the subject has been chastised or has learned a lesson. It is an unusual cause for
misconduct, and it justifies the unusual remedy we propose.

Although no single factor determines student career decisions, it is quite possible that trauma created by the
subject's falsification played a role.
      If the subject proves unwilling to enter into an acceptable voluntary exclusion
agreement, we recommend that NSF debar the subject for a suitable period of time. The
debarment is a Group 111action (see 5 689.2(a) (3) (ii)).

       NSF should also send the subject a letter of reprimand, which is a Group I action (see
0 689.2 (a) (1) 0).
        We believe that the actions described above adequately protect the federal interest in the
integrity of work carried out under federal awards and are proportionate to the seriousness of
the subject's misconduct. The subject's reply to our report, along with his belated confession
of wrongdoing and expression of regret to the University, encourage us to hope that this matter
can be brought to an appropriate resolution.