On top of all the other challenges that pollsters have faced in the past two presidential elections, this year has an additional, potentially significant, complication: a well-known third-party candidate.

Measuring support for third-party candidates has long been a particular challenge for pollsters. But it has been decades since the country has seen a third-party candidate as prominent as Robert F. Kennedy Jr., who has an average of about 10 percent of the vote in national polls.

Historically, polls overstate support for third-party candidates. When it comes to Mr. Kennedy, the biggest question may be by how much.

Consider this: In a two-part experiment conducted by The New York Times and the research firm Ipsos, a seemingly subtle difference across two versions yielded significantly different results for Mr. Kennedy.

What’s more, a candidate who is not on the ballot anywhere — a Times editor picked for inclusion thanks to his generic-sounding name — received a non-negligible share of support, highlighting just how much support for third-party candidates can come down to frustration with major-party candidates and a yearning for more options.

What does this all mean? It’s tempting to interpret these results as ungenerous to both voters and to polling. But the results say something real about how preferences work, and the central conundrum when it comes to third-party candidates. (This experiment is separate from Times/Siena College polling, though it was conducted with the same standards and rigor that we apply to all of our polling.)

In short: Much of what influences third-party candidate support isn’t just a straightforward desire to see that person become president. This poses a challenge for pollsters no matter what they do: Simply by listing third-party candidates, a poll might overstate their support. If a poll doesn’t list them, however, it can’t capture their support at all.

This year, to combat that concern, many reputable pollsters ask both versions of the question: one that poses a simple head-to-head contest between major-party candidates, and one that includes third-party candidates who may be on the ballot.

And which question gets asked first is where the difference comes in.

Question order matters

Here is the longer question asked by Times/Ipsos that includes the full field:

It contains a total of six options: the major-party candidates, the three established third-party candidates who have achieved ballot access in at least one swing state, and our wild card, William Davis, at No. 6.

And this is the shorter question that includes just President Biden vs. Donald J. Trump:

Our experiment worked like this: All respondents were shown both the long and short questions, but half were shown the full list first, and the other half were first shown the two-way race.

Among those who saw the long list first, Mr. Kennedy garnered 7 percent of the vote.

But among those respondents who encountered the head-to-head contest before seeing the full list, Mr. Kennedy’s support shot up six percentage points to 13 percent.

Why the increase, if the questions are the same? There are many factors that can explain this, but it is at least partly related to a phenomenon that pollsters call expressive responding. This is when people might use a survey response to show their frustration or express a particular feeling that’s not exactly what is being asked.

In this case, many respondents seem to be using the second question to convey frustration with the choices for president in the first question, whether or not their answers reflect their full views. When respondents have already been given a chance to express their support for one of the two major-party candidates, they seem to be more likely to register a protest of that first choice with their response to the fuller ballot. Some of the respondents given the longer list first are also probably expressing their frustration with the major-party candidates, but our results help demonstrate that effect is magnified when the longest list of candidates is asked second.

[You can find the full results of the poll, including the exact questions that were asked and how the poll was conducted, here.]

That might also explain why Mr. Davis, the Times editor who has no aspirations for higher office, won the support of about 1.5 percent of respondents, putting him on par with an actual Libertarian Party candidate. His support was only slightly lower among respondents who saw the third-party candidates first — evidence that voter frustration, though less pronounced under that scenario, still exists.

What’s more, Mr. Davis gets 4 percent among voters who feel unfavorably toward Mr. Biden and Mr. Trump.

The effects of this phenomenon show up when looking across many high-quality polls. Among 11 recent national polls, those that listed third-party candidates as the second question generally saw higher support for those candidates when compared with the polls that showed third-party candidates as the first question. (In the latest Times/Siena battleground polls released Monday, Mr. Kennedy was listed in the first question and received 10 percent support across the six states.)

An experiment like this can help us get a rough sense of how much support for Mr. Kennedy, and other third-party candidates, might come from voters expressing their frustration. But it also puts into perspective just how much his support can vary across polls and how hard it is to judge his real support.

It also illustrates some of the limitations surveys face. Pollsters can rely only on what voters tell us, and even voters themselves might not have fully thought through some of these questions.

Are voters consciously telling us they plan to vote for Mr. Kennedy, knowing that in the end they might support one of the two major-party candidates? Probably not. But they might be considering their options at a time when these decisions feel fairly abstract.

Is Kennedy a unique case?

History shows that third-party candidates often poll best in the spring and summer before an election — when everything feels fairly hypothetical — but lose steam as the election nears. Looking back at some of the strongest modern third-party candidates like Ross Perot or John Anderson, they often follow a similar path: strong support early in the race that slowly recedes by Election Day.

There’s at least one reason to believe Mr. Kennedy’s support may last longer: his name. In the Times/Siena battleground polls, we asked his supporters why they planned to vote for him. While most listed distaste for the alternatives as their motivation, for a handful — about 7 percent — his family is exactly why they are supporting him, even as many of his relatives have disavowed his candidacy. As one respondent put it: “Because he is a Kennedy.”

But current conditions are also ripe to have inflated third-party support. The two major-party candidates are deeply unpopular, providing an outlet for the type of expressive responding that pollsters worry about. And in our latest Times/Siena swing state polls, support for Mr. Kennedy appeared weak. Only about 30 percent of his supporters said they definitely planned to vote for him, compared with nearly 80 percent of Mr. Trump and Mr. Biden’s supporters who said they definitely planned to vote for their candidate.

So perhaps Mr. Davis, the Times editor, should not consider giving up his job and hitting the campaign trail anytime soon. But poll consumers should consider that even the best polls are imperfect, and it’s important to understand potential sources of error.

This year, Mr. Kennedy’s support is likely a big one.