Ballot counters wait Friday at the Broward County Supervisor of Elections headquarters for the start of the manual recount of Florida’s U.S. Senate race in Lauderhill, Fla. (Photo by Jahi Chikwendiu/The Washington Post)

In Florida, as Yogi Berra might say, it’s deja vu all over again.

Eighteen years after the infamous recount to decide Florida’s vote in the 2000 presidential election, Florida Secretary of State Ken Detzner announced on Thursday that the state will conduct a manual recount of ballots in this year’s election for U.S. Senate — potentially overturning the slim lead currently held by Republican Gov. Rick Scott over Democratic incumbent Bill Nelson.

So how will Florida go about recounting these ballots? And can voters expect a fair count, or will it be tainted by partisanship?

Let’s look at what Florida law says about recounting ballots — and what political science research says about how that process actually might work.

What Florida law says

Florida’s Election Code requires a manual recount for any election in which the initial margin of victory is 0.25 percent or less. Essentially, in a manual recount election officials will reexamine any paper ballots that were initially set aside as overvotes (votes for multiple candidates in a given race) or undervotes (votes for none of the candidates in a given race), in hopes of discerning the voter’s intent. Fortunately, there are no dimpled chads this time around. But there are bound to be stray marks, scribbles and other anomalies that would prevent voting machines from recognizing these potential votes.

So how does an election official determine voter intent? Many states provide guidelines for doing so in the case of a manual recount. Some states, such as Minnesota, have very specific rules for determining intent. Others, such as North Carolina, offer less guidance. Florida gives ballot counters a few examples to help guide the recount. For each example, the question the ballot counter is trying to answer boils down to this: Regardless of whether voters followed the provided instructions, did they mark their ballot in a consistent manner that expresses “a definite choice”? If so, count the vote. If not, don’t count it.

Simple enough, right? Our research suggests the answer is no.

It matters who counts the ballots — and how

In 2010, we conducted an experiment to test how partisanship and other factors might influence a ballot recount. Our “ballot counters” consisted of 340 undergraduate students from Ohio State University who participated in exchange for extra credit in political science courses. Of these, 20 percent of participants identified as people of color, 45 percent were women, and their ages ranged from 18 to 45, with a median age of 20 years old.

We asked these participants to decide whether to count or reject ballots in a disputed election, using actual ballots from Minnesota’s contested 2008 U.S. Senate election between Democrat Al Franken and Republican Norm Coleman. But, using photo editing software, we manipulated the information that appeared on each ballot. And we randomly varied the ballot counting instructions provided to each of six groups.

Two groups of participants saw ballots with no party labels whatsoever; one of those two groups saw ballots with specific counting rules, and the other saw ambiguous ballots. The other four groups received party-labeled ballots with: (1) fictional candidates and specific counting rules, (2) fictional candidates and ambiguous counting rules, (3) real candidates and specific rules or (4) real candidates and ambiguous counting rules.

Clean ballot, no errors

Overvote — or retraction?

Overvote — or pencil rest mark?

Overvote — or cross-out?

Our results show that ballots can be recounted fairly, at least in some circumstances. Democrats and Republicans basically agreed on how to count disputed ballots when they were given specific rules for determining voter intent and when the ballots included fictional candidates from the two major parties.

But in other cases — including many similar to those that will be recounted in Florida — the voter’s intent was truly in the eye of the beholder. That is, Democrats and Republicans disagreed quite a bit on how to count ballots when they were provided with ambiguous rules for determining voter intent and when the ballots included actual candidates from a recent election, with their party labels.

And it is probably no coincidence that the vote counters tended to rule in favor of their party’s candidate. In fact, there was more than a one in three chance that someone who “strongly” identified as a Democrat or Republican would accept a ballot challenge favoring their party. But if it favored the other party, the chances of accepting a ballot challenge were only one in five.

In practical terms, for every 100 ballot challenges favoring one party, “strong” members of that party would accept 36. “Strong” members of the other party would only accept 20.

In a disputed election likely to be determined by a small number of votes, this difference between what ballot judges will accept based on their partisan preferences could be large enough to shape the outcome. But here’s a key point: This partisan difference is reduced by more than half when ballot judges are given specific, rather than ambiguous, instructions about how to determine voter intent. Our experimental results suggest that specific rules for judging disputed ballots can limit the effects of partisanship and thus lead to a fairer counting of votes.

In Florida, rules provide guidance for a variety of hypothetical scenarios, but there are bound to be ambiguously marked ballots. And, certainly, the stakes are high. Partisanship may play a big role in resolving ambiguously marked ballots that are not clearly addressed by ballot counting rules.

Can recounts be fair?

In our experiment, we found that this partisanship can bias the vote-counting process, especially when the political stakes are high. Clear recounting guidelines can mitigate these biases, but they can also lead to ballots being rejected because of accidental stray pen marks even when voter intent is otherwise clear.

We found that merely instructing ballot judges to determine intent is not explicit enough to override deep-seated partisan attachment. And in situations where every vote could tip the election results, these partisan attachments matter.

As voting by mail becomes more widespread, state election officials may find it even more difficult to craft rules that define which ballots “count.” That’s because election officials need to match the signatures on mail-in ballots to a voter’s signature at the time of registration. And such an interpretation is also subject to partisan and other biases. Is that “t” crossed in the same way — or the “i” dotted similarly? Is that signature slanted differently than this one? And who’s she voting for, anyway?

Here again, what should be an objective judgment could be open to interpretation. But as we show, clear and explicit guidelines can help to diminish the power of partisanship in our ever more partisan nation.

Kyle C. Kopko (@KyleKopko) is associate dean of institutional effectiveness, research and planning and an associate professor of political science at Elizabethtown College. Sarah M. Bryner is research director for the Center for Responsive Politics; her views here are her own. Jeffrey Budziak is an associate professor of political science at Western Kentucky University. Christopher J. Devine (@ProfDevine) is an assistant professor of political science at the University of Dayton. Steven P. Nawara is an assistant professor of political science at Lewis University.