[House Hearing, 110 Congress]
[From the U.S. Government Publishing Office]


 
                   GAO BRIEFING ON THE STATUS OF THE
      INVESTIGATION INTO THE FL-13 CONGRESSIONAL DISTRICT ELECTION

=======================================================================

                                MEETING

                               before the

                   COMMITTEE ON HOUSE ADMINISTRATION

    TASK FORCE FOR THE CONTESTED ELECTION IN THE 13TH CONGRESSIONAL 
                          DISTRICT OF FLORIDA

                        HOUSE OF REPRESENTATIVES

                       ONE HUNDRED TENTH CONGRESS

                             FIRST SESSION

                               __________

            MEETING HELD IN WASHINGTON, DC, OCTOBER 2, 2007

                               __________

      Printed for the use of the Committee on House Administration




                       Available on the Internet:
   http://www.gpoaccess.gov/congress/house/administration/index.html



                    U.S. GOVERNMENT PRINTING OFFICE
38-706                      WASHINGTON : 2007
_____________________________________________________________________________
For Sale by the Superintendent of Documents, U.S. Government Printing Office
Internet: bookstore.gpo.gov  Phone: toll free (866) 512-1800; (202) 512ï¿½091800  
Fax: (202) 512ï¿½092104 Mail: Stop IDCC, Washington, DC 20402ï¿½090001

                   COMMITTEE ON HOUSE ADMINISTRATION

                ROBERT A. BRADY, Pennsylvania, Chairman
ZOE LOFGREN, California,             VERNON J. EHLERS, Michigan,
  Vice-Chairwoman                      Ranking Minority Member
MICHAEL E. CAPUANO, Massachusetts    DANIEL E. LUNGREN, California
CHARLES A. GONZALEZ, Texas           KEVIN McCARTHY, California
SUSAN DAVIS, California
ARTUR DAVIS, Alabama
                 S. Elizabeth Birnbaum, Staff Director
                William Plaster, Minority Staff Director

    Task Force for the Contested Election in the 13th Congressional 
                          District of Florida

                  CHARLES A. GONZALEZ, Texas, Chairman
ZOE LOFGREN, California              KEVIN McCARTHY, California


    GAO BRIEFING ON THE STATUS OF THE INVESTIGATION INTO THE FL-13 
               CONGRESSIONAL DISTRICT CONTESTED ELECTION

                              ----------                              


                        TUESDAY, OCTOBER 2, 2007

                  House of Representatives,
                          Task Force on Florida-13,
                         Committee on House Administration,
                                                    Washington, DC.
    The committee met, pursuant to call, at 4:05 p.m., in Room 
1310, Longworth House Office Building, Hon. Charles A. Gonzalez 
[chairman of the task force] presiding.
    Present: Representatives Gonzalez, Lofgren, McCarthy, 
Ehlers, and Lungren.
    Staff Present: Thomas Hicks, Senior Election Counsel; 
Janelle Hu, Election Counsel; Jennifer Daehn, Election Counsel; 
Matt Pinkus, Professional Staff Member/Parliamentarian; Kristin 
McCowan, Chief Legislative Clerk; Daniel Favarulo, Staff 
Assistant; Matthew DeFreitas, Staff Assistant; Kyle Anderson, 
Press Director; Gineen Beach, Minority Counsel; and Bryan 
Dorsey, Minority Professional Staff Member.
    Mr. Gonzalez. I am going to call the meeting of the Task 
Force on Florida-13--this is a public meeting, of course, and 
it is going to be called to order.
    The record will reflect that I am Charles Gonzalez, and I 
am the chairman of the task force. And I am joined today by 
another member of the task force, Kevin McCarthy, a Congressman 
from California. And we have also Dan Lungren, who is not 
officially a member of the task force but has attended 
different meetings and hearings that we have had, and 
briefings. Congresswoman Zoe Lofgren, the third member of the 
task force, is delayed presently but will be joining us.
    Recognizing that we are probably going to have votes in 
about 30 minutes, we are going to try to get as much done as 
possible today. The witnesses that we have here--and I believe 
we may be hearing from all three if necessary, but I will start 
with the easier names. And that is going to be Jan Montgomery 
and Gloria Jarmon, and then we have Dr. Naba Barkakati, which 
is pretty good after all this time. Because I believe it is 
going to be Dr. Barkakati that is going to actually be making 
the report.
    By way of background, we had a briefing last week. A draft 
report was being prepared. It had to be vetted, basically, if 
you want to call it that, or presented to ES&S, the 
manufacturer of the voting machines in question, as well as 
Florida election officials, for their input, their comments and 
so on. My understanding is that has been completed, and today 
GAO will be making a report to us that should answer the 
question about whether there is a need for further testing.
    And with that, I will turn it over to the GAO 
representatives.
    Mr. Barkakati. Thank you. Mr. Chairman, Members----
    Mr. Gonzalez. If we could hold on----
    Mr. Barkakati. Sorry.
    Mr. Gonzalez. And I am sorry.
    I apologize to Congressman McCarthy. If any Member wishes 
to make an opening statement, we can do so at the present time.
    Mr. McCarthy. I thank Chairman Gonzalez.
    Today's public hearing will analyze the GAO's report that 
summarizes the investigation of the Florida-13 congressional 
election that took place almost a year ago between Congressman 
Buchanan and Ms. Jennings.
    I thank our witnesses for returning to publicly report 
their findings thus far and for their continued cooperation to 
keep the task force informed and to brief members last week.
    I am pleased that all parties are working well together. In 
what could be a contentious struggle to get information from 
the courts, the state, the vendors and experts, the GAO has 
received cooperation from all parties as it has gone over 
studies that have already been done and as it has formulated 
plans of further studies, checked protocols and rechecked 
protocols.
    The results of this investigation thus far seems to clearly 
point that there is no smoking gun, no evidence that the voting 
system would have caused the undervote, a conclusion similar to 
what Sarasota County, the State of Florida and the independent 
teams of experts have already derived.
    Right now, based upon GAO's findings, we seem to be nearing 
the conclusion of this contested election so that GAO can 
corroborate its findings with reasonable certainty on what it 
has found thus far, and that there is no evidence that voting 
systems caused undervotes and that the undervotes could instead 
have occurred because voters intentionally undervoted or did 
not properly cast their ballots, perhaps due to the ballot 
design.
    The GAO is recommending that it conduct a few additional 
tests. I credit the GAO with its professional undertaking of 
this very difficult responsibility in response to the task 
force's request. I agree with the GAO recommendations to move 
forward with those additional tests so it can make its eventual 
recommendation with the reasonable certainty and finally put to 
rest for the people of the 13th District of Florida the 
challenge against Congressman Vern Buchanan.
    And I yield back the balance of my time.
    [The statement of Mr. McCarthy follows:]

    [GRAPHIC] [TIFF OMITTED] T8706A.001
    
    [GRAPHIC] [TIFF OMITTED] T8706A.002
    
    [GRAPHIC] [TIFF OMITTED] T8706A.003
    
    Mr. Gonzalez. Thank you very much, Congressman McCarthy.
    And at this time, we will turn it over to the witnesses.

      STATEMENT OF MR. NABAJYOTI BARKAKATI, SENIOR-LEVEL 
  TECHNOLOGIST, APPLIED RESEARCH AND METHODS, U.S. GOVERNMENT 
                     ACCOUNTABILITY OFFICE

    Mr. Barkakati. Thank you, Mr. Chairman.
    Mr. Chairman, Members of the task force, I am here today to 
present our findings regarding the Florida-13 review.
    Thanks in large part to the task force's continued support, 
we have had good cooperation from all parties involved, and we 
have been able to access all the information we needed to study 
whether the voting systems contributed to the undervote in the 
Florida-13 race in Sarasota County's 2006 general election.
    In that election, Sarasota County used ES&S voting systems, 
specifically the Unity Election Management System and 1,499 
iVotronic direct recording electronic, DRE, touchscreen 
machines, during the early voting and Election Day voting days.
    Our independent analysis of the 2006 general election data 
from the county, Sarasota County, confirmed the unusually large 
number of undervotes there in the Florida-13 race, and we found 
that the undervotes were generally distributed across all 
precincts and all machines.
    We found that the prior reviews and testing provided 
reasonable assurance that the Unity Election Management System 
did not contribute to the undervote, and the votes captured by 
the iVotronic DREs within the precincts matched the voter count 
from the precinct registers within an acceptable range of 
errors.
    However, these tests do not provide enough information to 
determine whether the iVotronics contributed to the undervote 
or not. The firmware comparison and parallel tests, which were 
done as part of Florida State's audit of the Sarasota County 
elections, provided useful information, but the results could 
not be applied to all iVotronics because the number machines 
that were tested was small. Additionally, the machines were not 
tested for all the different ways that the voter could have 
cast votes using the machines--a feature of an iVotronic 
machine. We also did not find prior testing of what happens 
when a touchscreen is deliberately miscalibrated.
    To address these issues, we are proposing three tests: a 
firmware verification test; a ballot test; and a calibration 
test, which should be conducted to try to obtain further 
assurance that the iVotronic DREs used in the Sarasota County 
elections did not cause the undervote.
    The proposed firmware verification test is similar to the 
one that was conducted by Florida State on six machines, but, 
in this case, we are conducting on a larger population. It is 
going to test a representative sample of iVotronic DREs and 
compare its installed firmware with the certified version.
    The ballot test is going to exercise the different ways of 
casting ballots, selecting candidates and casting a vote on 10 
iVotronic DREs.
    The calibration test would deliberately miscalibrate an 
iVotronic DRE and then test it to verify whether it works 
properly.
    We estimate that all three tests would take 2 weeks using a 
staff of about six to eight people, once we have made all the 
arrangements necessary for the tests.
    Should the task force ask GAO to conduct these tests, 
several matters need to be addressed before the testing could 
begin. We would need to obtain access to the machines that are 
currently sequestered under a court order. We have to arrange 
for a test site, obtain some tools that are needed for firmware 
verification tests, develop the test protocols and test 
procedures, and arrange for a video recording of the test.
    Sarasota County election officials have told us that, 
working around their current election schedule, they can help 
GAO access the machines and provide a test site between 
November 26 and December 7, 2007.
    Before I conclude, I should mention that we recognize that 
human interaction with the ballot layout could have been a 
potential cause of the undervote. And although we have not 
explored this issue in our review, we note that there is an 
ongoing academic study that is exploring this issue using 
machines from ES&S, the same manufacturer. We believe that such 
experiments could be useful and could provide insight into the 
ballot layout issue.
    We also noted that there are several suggestions which have 
been offered as possible ways to establish that the voters are 
intentionally undervoting and the machines are not causing the 
undervote.
    First, a voter-verified paper trail could provide an 
independent confirmation that the touchscreen machines did not 
malfunction in recording and counting the votes.
    Second, providing explicit feedback of the undervote and 
requiring positive verification before casting the ballot with 
an undervote might just prevent many voters from 
unintentionally undervoting.
    And third and finally, offering a ``none of the above'' 
option in a race could provide the voters with an option that 
enables them to indicate that they are intentionally 
undervoting.
    However, we emphasize that any decision about these or 
other suggestions about ballot layout or things related to the 
voting system functions should be informed by human-factors 
studies that assess their effectiveness in accurately recording 
the voters' intentions and making the voting systems easier to 
use and preventing unintentional undervotes.
    In conclusion, the prior tests and reviews of the Sarasota 
County 2006 election have provided valuable information about 
their voting systems. Our review has found that, in many cases, 
we could rely on those results--on that information to 
eliminate areas of concern. This has allowed us to identify the 
areas where increased assurances are needed to answer the 
questions being raised. Accordingly, the primary focus of the 
tests we are proposing is to obtain reasonable assurance that 
the results of prior reviews, as well as our proposed tests, 
could be applied to all the iVotronic DREs that were used in 
the 2006 election.
    However, even after completing all the proposed tests, we 
would not be able to provide absolute assurance that the 
iVotronics did not play any role in the large undervote because 
we are unable to recreate the exact conditions of the election 
in 2006.
    By successfully conducting the tests, however, we are going 
to be able to reduce the possibility that the machines were the 
reason of the undervote and shift attention to the other 
possibilities that the voters either intentionally undervoted 
or did not properly cast their ballot on the iVotronic, 
potentially because of issues related to the interaction 
between the voter and the ballot.
    Mr. Chairman, this concludes the summary of my written 
statement. I would be happy to respond to questions at this 
time from you and other Members of the task force. Thank you.
    [The statement of Mr. Barkakati follows:]

    [GRAPHIC] [TIFF OMITTED] T8706A.004
    
    [GRAPHIC] [TIFF OMITTED] T8706A.005
    
    [GRAPHIC] [TIFF OMITTED] T8706A.006
    
    [GRAPHIC] [TIFF OMITTED] T8706A.007
    
    [GRAPHIC] [TIFF OMITTED] T8706A.008
    
    [GRAPHIC] [TIFF OMITTED] T8706A.009
    
    [GRAPHIC] [TIFF OMITTED] T8706A.010
    
    [GRAPHIC] [TIFF OMITTED] T8706A.011
    
    [GRAPHIC] [TIFF OMITTED] T8706A.012
    
    [GRAPHIC] [TIFF OMITTED] T8706A.013
    
    [GRAPHIC] [TIFF OMITTED] T8706A.014
    
    [GRAPHIC] [TIFF OMITTED] T8706A.015
    
    [GRAPHIC] [TIFF OMITTED] T8706A.016
    
    [GRAPHIC] [TIFF OMITTED] T8706A.017
    
    [GRAPHIC] [TIFF OMITTED] T8706A.018
    
    [GRAPHIC] [TIFF OMITTED] T8706A.019
    
    [GRAPHIC] [TIFF OMITTED] T8706A.020
    
    [GRAPHIC] [TIFF OMITTED] T8706A.021
    
    [GRAPHIC] [TIFF OMITTED] T8706A.022
    
    [GRAPHIC] [TIFF OMITTED] T8706A.023
    
    [GRAPHIC] [TIFF OMITTED] T8706A.024
    
    [GRAPHIC] [TIFF OMITTED] T8706A.025
    
    [GRAPHIC] [TIFF OMITTED] T8706A.026
    
    [GRAPHIC] [TIFF OMITTED] T8706A.027
    
    [GRAPHIC] [TIFF OMITTED] T8706A.028
    
    [GRAPHIC] [TIFF OMITTED] T8706A.029
    
    [GRAPHIC] [TIFF OMITTED] T8706A.030
    
    Mr. Gonzalez. Thank you very much, Doctor.
    I have been informed that Congresswoman Lofgren may not be 
able to make it back before the recorded votes. It would be my 
intention, then, that we would be recessing once we are called 
for votes if we still have business going on and then 
reconvening. I apologize. But Zoe's--obviously, there are other 
pieces of legislation. She is meeting with leadership, and she 
is actually with the Speaker, as we speak.
    But a couple things, Doctor, let me ask you on some of the 
things that are kind of, you know, reasonably understandable by 
the lay person.
    You are going to be conducting tests on a greater number of 
machines than were used in the previous testing. Is that 
correct?
    Mr. Barkakati. Yes. I didn't mention specifically, but the 
number always depends on the level of confidence they want and 
the error you are going to tolerate.
    And if you want to conclude at a 99 percent confidence 
level that, at the most, 4 percent of the machines may have 
different software, then you have to test 115, according to our 
statisticians, you have to test 115 machines, which would be 
selected out of the ones that are sequestered as well as the 
ones that have been used since the 2006 election. And that is 
what we are proposing, 115 machines to test.
    Mr. Gonzalez. As opposed to how many that were tested 
previously?
    Mr. Barkakati. There were six machines that were tested in 
Florida State's audit. And that was the point that 
statisticians made, that it was an insufficient number to 
conclude that the results could apply to all the machines.
    Mr. Gonzalez. The other thing you pointed out was also the 
different options or variations that could actually occur in 
voting in the Florida-13. I believe that there are over 100 
variations; is that correct?
    Mr. Barkakati. Yes. The reason for looking through all the 
variations is that we are trying to use the machine's features 
in determining how many ways you could go back and change your 
vote and, you know, maybe confirm what you have done and then 
vote. Because of the ways you can do so, by paging back, 
because there are multiple pages of ballots, and you can go 
from the review screen where you see the summary and can touch 
and go back--because of those combinations, it comes out to be 
112 for a simple, like, 2-person race, where you are having to 
consider many different ways of going back and forth. And that 
is the reason we have 112 ways we determined that we should be 
testing.
    Mr. Gonzalez. And you are going to use how many variations?
    Mr. Barkakati. We are going to use all those 112 ways of 
selecting a candidate and casting a ballot.
    Mr. Gonzalez. And how many were used in the testing?
    Mr. Barkakati. In the previous tests, we had 10 ways of 
combinations that were used in the Florida State audit. And 
then the normal testing of candidates verifies three ways, 
which is like each candidate and an undervote. So those three 
ways were covered. So 13 ways were covered by previous tests.
    But I should emphasize that we are not putting as much--
that is not the biggest item. More important to ask was the 
confirmation of the firmware, that every machine is running the 
same software. That was the important part. And then this is 
another additional assurance to confirm the proper working of 
the machine.
    Mr. Gonzalez. Okay.
    And I wish to acknowledge that we have been joined by 
Congressman Ehlers, who is the ranking member of the full 
committee.
    Welcome, Congressman.
    The other thing that was of concern and I think we have 
touched on in the past--of course, time is always of the 
essence when we have these contests. I think we all agree on 
that. But we wanted to do this in a very orderly manner. It is 
somewhat disturbing that we would not be able to actually 
conclude some of this testing until late November, early 
December.
    My understanding is the reason that we would do that is 
that is the only time that the Florida officials would be 
available to assist you. To try to do it any other way actually 
may even prolong it. In other words, if we assume 
responsibility for the premises, the security of the premises, 
the sequestered machines, rather than the present situation 
where everybody in Florida is duly acknowledged and authorized 
to maintain that kind of control, to do it any other way would 
probably even extend it beyond November 26 through the first 
week of December.
    Is that correct?
    Mr. Barkakati. That is true. We had considered based on the 
task force's need for doing it faster, and we also concluded 
that it would be hard to get it done any other way because of 
all the reasons that you cited. And what they had told us is 
that if we cannot do it within that 2-week time period, then it 
could be as late as in February because of other primaries 
coming up for the presidential election.
    Mr. Gonzalez. Okay. So we are faced with the situation that 
November 26 may be the earliest. And that is a window of 
opportunity. And I will tell you now that we need to take 
advantage of that opportunity and in no way wait until 2008.
    At this time, I will recognize my colleague, Congressman 
McCarthy.
    Mr. McCarthy. Well, thank you, Mr. Chairman. I just ask 
that we could keep this ratio in the full committee as well. 
[Laughter.]
    Let me just touch on a few--and I appreciate the help. Now, 
you have looked at the reports that Sarasota, the reports that 
Florida has done, and the reports of the experts. Have you 
found anything in those reports in your testing that doesn't 
conclude what they have said?
    Mr. Barkakati. No, we didn't find anything--we didn't find 
anything that concluded anything different from what they have 
already concluded.
    Mr. McCarthy. Okay. And if I understand correctly, you have 
some assurance already, and to take it to 99 percent assurance 
would take 115 machines?
    Mr. Barkakati. Yes.
    Mr. McCarthy. And of those machines that you tested, were 
all those machines used in the election?
    Mr. Barkakati. Oh, no, I am sorry. We have not tested any 
machines yet. What we are proposing is that we would test 115 
machines.
    Mr. McCarthy. I thought you said six machines were tested.
    Mr. Barkakati. Oh, I am sorry. Yeah, you are right. Those 
six machines were tested by Florida State during their audit.
    Mr. McCarthy. And those were used in the election?
    Mr. Barkakati. They were. Those six machines were--I think 
they were used in the election maybe----
    Mr. McCarthy. And in the precinct?
    Mr. Barkakati. I don't know precisely. At least some of 
them were used.
    Mr. McCarthy. It is my understanding those six were used in 
this congressional election in those precincts. And you found 
nothing wrong with these?
    Mr. Barkakati. No. Those tests didn't find anything wrong.
    Mr. McCarthy. All right. So if the 115 machines give you 99 
percent assurance, where are you currently in your assurance of 
nothing going wrong?
    Mr. Barkakati. The statisticians, I mean, we didn't ask 
them to calculate it with the six machines tested. That is 
going to be pretty low, though, you know, in terms of 
statistical terms, it will be very low assurance that all the 
population of machines are running the same software.
    Mr. McCarthy. So you just want to have that reasonable 
assurance?
    Mr. Barkakati. Right. I mean, we wanted----
    Mr. McCarthy. You wanted to be at absolute?
    Mr. Barkakati. Yes. We could probably get back to you with 
the number. But that would be, honestly speaking, quite low, I 
mean, in the tens or less probably. But for 99 percent 
assurance that all the machines are running--see, we are kind 
of running into the situation that we don't know if all the 
machines are the same or not, even though there is no reason to 
say they are not the same.
    Mr. McCarthy. Of this that you have already tested, has 
there been any smoking gun or any signs that would drive you to 
look at something else?
    Mr. Barkakati. No.
    Mr. McCarthy. Okay. So you have some assurance already. The 
ballot design, does that move up on the list of maybe the 
outcome of what has caused this?
    Mr. Barkakati. Actually, we were trying to say it in this 
manner. If we did the tests of the 115 machines and there was 
nothing found in the ones we are proposing, then definitely the 
remaining next cause, only remaining cause would be--after all, 
voters can intentionally do it. That is a different issue. Then 
unintentionally missing it because of ballot layout is the 
prominent reason, probably, remaining at that point.
    Mr. McCarthy. Well, I keep seeing, with the testing that we 
have done, the county did the testing, the State did the 
testing, brought in the independents, and now you have done so 
much that you come back all with the same answer, the 
assurance. And then we are going to go one more.
    Just for the voters of the 13th, so far everything that was 
said on Election Day has come true. Would this be the end of 
this? Would we be able to tell the voters then--I guess right 
now we can tell them we have reasonable assurance that, yes, 
the outcome is true--that we would be able to say, yes, the 
outcome is totally true.
    Mr. Barkakati. I think, actually, after we finish the tests 
we proposed, then we can say that we have a reasonable 
assurance that the machines did not probably cause the 
undervote. At that point, we can make the statement, because 
then we have tested it. Right now, our problem is that we--
basically, the sticking point is that we don't know if all the 
machines are truly----
    Mr. McCarthy. We want to get to the highest percentage so 
we can say we are at 99 percent. We can never get to 100. But 
we just want to reaffirm what we already found out from our 
testing.
    Mr. Barkakati. Yes. To be able to apply the results we know 
so far, which have been showing that the machine hasn't caused 
anything.
    Mr. McCarthy. So we just want to jump through the next hoop 
to reassure what we already know.
    Mr. Barkakati. The problem we are running into--the 
statisticians tell us that, with the six machines tested, I 
could not say that we have reasonable assurance yet. We are 
getting that after we have finished the 115 machines. Then we 
can say that all the tests that were done so far tell us now 
that we are very much certain that they are----
    Mr. McCarthy. Well, I believe, Mr. Chairman, that I would 
support to do this further testing. But I would like to come to 
a conclusion. And I know we have an assurance so far. And I 
will go to the 99 percent, but I don't want to carry it on 2 
years until the next election.
    I yield back.
    Mr. Gonzalez. Thank you very much, Mr. McCarthy. I know 
there is only a certain degree of, again, certainty that we can 
arrive at. But I understand what you are saying.
    At this time, does anyone else wish to be recognized?
    Mr. Lungren.
    Mr. Lungren. Thank you very much, Mr. Chairman.
    You may have already answered this, but let me just get 
this correct. If you do the test that you are asking the task 
force to permit you to do, you can then come back with your 
results and say that you have reached a reasonable assurance 
that either the machines did malfunction or did not 
malfunction?
    Mr. Barkakati. Yes. We will be able to say it at that 
point.
    Mr. Lungren. Okay. If you say that, is there a possibility 
you would ask for further tests?
    Mr. Barkakati. No. I mean, we have----
    Mr. Lungren. Okay.
    Mr. Barkakati. We can never say that we considered all 
possible tests by the machine.
    Mr. Lungren. I understand that. But I just want to make 
sure that, from your standpoint, you are telling us if the task 
force okays the testing, when you finish that, you will be able 
to give us what you call a, quote/unquote, ``reasonable 
assurance'' level conclusions?
    Mr. Barkakati. Yes.
    Mr. Lungren. Okay. Now, here is the question I got. On page 
15 of the draft that we had received, you explain the 
justification for one of the three additional tests you are 
proposing, which expanded the number of ways a voter could 
select the candidate in the congressional race from the 13 
tests to the 112.
    Now, see if I have got this right. To get to the 112 ways a 
voter could choose a candidate, it appears the voters would 
have to cast and change their votes in this race four times. 
Now, just to an average person or even someone who has been 
involved in politics a long time, both as a candidate and as a 
voter, changing your vote in a single race four times would 
seem highly unusual, to say the least. But your report says 
that the 112 permutations would compile the expected forms of 
voter behavior.
    What I am trying to understand is, how is that within the 
universe of the expected forms of voter behavior? I would call 
that aberrate behavior, unusual behavior, something that, if I 
saw it, would either send up a red flag or I would say, ``That 
is so unusual, I wouldn't expect to see it again.''
    Mr. Barkakati. Actually, I should say that those 112 
include all the combinations. Of course, one person doing--I 
mean, there are some that are like--we have eliminated some 
that wouldn't make any sense.
    Mr. Lungren. I am trying to understand. Am I reading it 
correctly? Is that what the report said? You would have to 
change your vote in this race four times to come to that----
    Mr. Barkakati. The combinations include everything from 
single, one selection to changing once to other combinations of 
changing, going forward, coming back and changing. So it is a 
combination of all of them coming out to be 112.
    To get to the 112, you do include some what might seem like 
odd behavior. But the counts come up because it includes 
everything from simple to complicated ones.
    And the assumption is, if you did not know anything about 
the machine, then you are trying to make it do all the stuff 
that it can do, to confirm that it works under all those 
circumstances. And that was the reason for picking that.
    I mean, in some ways, because you have to realize that 112 
ballots can be cast within about 1 day basically, so we thought 
that is a good test to exercise the machine.
    Mr. Lungren. Sure. You are making sure you take the whole 
universe in of possible voter behavior, it would seem to me, 
rather than just expected forms.
    Let me ask you this, and I didn't quite understand what you 
were saying about, you call it, ballot layout. I have heard the 
expression, ``ballot design.'' Nothing you are doing goes to 
the question of ballot design or ballot layout, is that 
correct, in your tests?
    Mr. Barkakati. I should qualify it in this way. The machine 
that we are testing is loaded with the ballot that was used in 
the 2006 election. So in that sense, it has the layout that is 
there. We are testing that.
    But the human interaction, the voters' reaction to the 
ballot, is not something we have tested or have proposed right 
now. And which is why we kind of ended up saying that, if we do 
this, the machine could be eliminated as the reason, but----
    Mr. Lungren. Machines would be eliminated.
    Mr. Barkakati. Yeah, but the voters----
    Mr. Lungren. Okay. I just wanted to make sure that is what 
you were talking about. Thank you very much.
    Thank you, Mr. Chairman.
    Mr. Gonzalez. You are welcome.
    Mr. Ehlers, do you have any questions?
    Mr. Ehlers. Thank you, Mr. Chairman. Just a few.
    As a scientist, I am a little worried about the terms used 
here: partial assurance, absolute assurance, increased 
assurance and so forth.
    When you have concluded the next set of tests, will you be 
able to express that in terms of a percentage likelihood, 
rather than the nondescriptive terms or nonquantitative terms 
you used?
    Mr. Barkakati. Yeah, I think we will be able to do the 
statistical sample of the machine in the primary comparison. 
That part we can definitely do in a quantifiable way.
    And as far as our statisticians, the next part depends on 
the logic, which says that machines that run the same software 
would behave the same way. If you accept that logic, then we 
can extend it to that level, that, yes, that will be 
quantifiable.
    And if it is not quantifiable, what we are saying is that 
our reasonable assurance is that machines do not cause--you 
know, it is not 100 percent, but with some very high degree of 
certainty, the machine did not cause the problem. That would be 
our----
    Mr. Ehlers. Well, you talk about your assurance now. What 
would you guess is your percentage assurance now that----
    Mr. Barkakati. I should really not probably guess because--
I should have asked our statisticians to calculate that. They 
would be able to tell with the sample of six what the 
percentage would be. Unfortunately I shouldn't say. I mean, I 
know it is low because they said, ``Oh, that is not good 
enough.'' But I wouldn't quantify it. At this point, I don't 
know statistically.
    Mr. Ehlers. What concerns me a little is we keep going by 
iteration, and we may end up with a 99 percent assurance. Are 
you going to come back and say, ``Well, we should do a few more 
tests and maybe we can get it at 99.5''? Where are you going to 
draw the line?
    Mr. Barkakati. Actually, we did pick the 99 percent 
confidence level based on the sample size we could test in a 
reasonable amount of time. And even though we are cautious in 
how we present our results and everything in a very careful 
way, we do expect that that will give us what we might call 
reasonable assurance.
    And like I said, at that point, we can say that no more 
than 4 percent of the machines could have had any problem, you 
know, essentially. That only gives you an error level of 4 
percent, unfortunately. That is how it goes.
    Mr. Ehlers. Okay. But our decision here is basically 
whether Mr. Buchanan remains seated as the Congressman or he 
does not. And you have to have--the evidence for saying that he 
is not has to be very, very strong. In other words, you may be 
talking about a 99 percent assurance that the machines worked 
right, but if you are trying to show the other--or anyone is 
trying to show that he should not be seated, that Ms. Jennings 
had won, there has to be a very high probability in her favor.
    What I am getting at is it appears, what you have so far, 
there is no probability that, given the evidence you have--or 
very low probability that she was the actual winner. And it is 
a much higher probability that he is the winner. Is that 
correct?
    Mr. Barkakati. I think after we finish our proposed 
testing, that will be certainly the indication, that the 
machines were--I mean, if your decision is based on the 
machines' performance, whether it is working or not working, it 
is most likely that everything succeeds, then we conclude the 
machines were not the problem. And then it will be up to the 
task force to decide how to use that knowledge, I think.
    Mr. Ehlers. Okay. Then we get into issues of ballot design, 
human behavior and so forth.
    Mr. Barkakati. Yes.
    Mr. Ehlers. Thank you very much.
    Thank you, Mr. Chairman.
    Mr. Gonzalez. Thank you, Mr. Ehlers.
    We have to make it real clear, the GAO are really evidence-
gatherers for us, and then we will make those determinations.
    But I do want to point out that, when we tasked GAO as to 
exactly what they were going to be doing, we were very clear 
that, in the final analysis, this first part of the work that 
you were doing, the last subpart was: Considering the tests 
that were conducted on the voting systems from Sarasota County 
after the general election, are additional tests needed to 
determine whether voting systems contributed to the undervote? 
That is what brings us here today.
    And it seems pretty clear, on page 11, it says, ``Prior 
tests and reviews provide some assurance, but do not provide 
reasonable assurance, that the iVotronic DREs did not 
contribute to the undervote. Prior tests and reviews of 
Sarasota County voting systems provide useful information but 
have some shortcomings.''
    So that is why it is necessary. You will be in a much 
better position to give us more probative and valuable 
information on which we may predicate a decision on the 
reliability of the results that were reported to us on the 
election in Florida-13.
    And we have just been joined by Congresswoman Lofgren.
    And thank you very much. I know you rushed over here.
    And by the way, I have been told that she was on ``The 
View'' this morning.
    Ms. Lofgren. No, not me. It was----
    Mr. Gonzalez. I thought it was you.
    Ms. Lofgren. The Speaker was on there.
    Mr. Gonzalez. Oh, the Speaker. I thought it was you. We 
were saying Lofgren.
    Oh, well, given my choice, it would have been 
Representative Lofgren.
    This is being recorded, and I am sure the Speaker is 
watching. [Laughter.]
    Ms. Lofgren. That would not be my choice.
    Mr. Gonzalez. All right. Representative Lofgren, at this 
point, if you wish to ask any questions--I know you were not 
here, but you attended last week's briefing, and I think you 
were provided a draft of the report.
    Ms. Lofgren. Right.
    Mr. Gonzalez. If you have any questions----
    Ms. Lofgren. No, I think this is proceeding properly. And 
at the appropriate time, I will have a motion to offer when 
others are through with their questions. If that is now, I will 
do it now.
    Mr. Gonzalez. All right.
    Ms. Lofgren. Everyone is ready.
    ``I move that the task force approve the proposed GAO 
testing plan and associated protocols as follows: A, firmware 
testing to verify that the firmware in the iVotronic DREs used 
in Sarasota County matches the certified version; B, ballot 
testing of iVotronic DREs to confirm correct operation; C, 
miscalibration of an iVotronic DRE to understand the effect on 
the undervote.
    I move further that the Chairman request that all 
individuals, offices and entities whose cooperation is 
necessary fully, promptly and voluntarily assist the GAO to 
enable it to conduct the testing described above.''
    That would be the motion.
    [The information follows:]

    [GRAPHIC] [TIFF OMITTED] T8706A.031
    
    Mr. Gonzalez. All right.
    All in favor of the motion, vote, ``Aye.''
    It is unanimous, and the record will reflect such.
    Also, I want to make sure that the record will include, 
being part of the record, and that is the report as submitted 
by GAO today.
    And we look forward to continuing working with you. Advise 
us immediately if you need some additional assistance in 
gaining access, any problems you have with logistics.
    Anything further?
    Mr. McCarthy. Mr. Chairman, I just want to congratulate you 
on this committee, because everything that we have done has 
been unanimous in our approach. And I think this is probably 
going to show, in the long run, how contested elections should 
be done in the future. So, thank you.
    Mr. Gonzalez. Well, I want to thank my colleagues for that.
    And we stand adjourned.
    [Whereupon, at 4:37 p.m., the task force was adjourned.]

                                  
