In a late Tuesday night tweet, President Donald Trump seized on an online survey’s apparent shock finding that a majority of Americans think John Brennan’s security clearance should be revoked. The poll results, however, are far more ambiguous than the White House would like, and do not necessarily support the president’s actions against the former CIA director.

The online survey was conducted by controversial pollster and pundit Mark Penn and his Harris Poll. The poll has been dogged by its own controversies for the way it words questions and how Penn interprets the findings, and by the designation of a C+ (up from a recent C-) and right-leaning rating from Five Thirty-Eight. The survey results on security clearances were released exclusively to The Hill, but are now publicly available.

A deep problem with the survey is that it did not ask the kind of questions that would address President’s Trump’s revocation of Brennan’s clearance. Imagine if the survey just asked the question whether all former intelligence officials should retain or lose their clearances after leaving office, and then asked whether former CIA director Brennan should retain or lose his clearance after leaving office. The flaw in such a survey instrument would be obvious. That hypothetical is not that far from the mark of what the survey actually did.

What one would really want to know is whether Americans support the president’s revoking Brennan’s clearances based on what the former CIA Director had said about Mr. Trump and his relations with Russia. Put another way, one can easily imagine starkly different poll results if the survey asked whether respondents supported the revocation of former national security officials for publicly criticizing the president.

What did the survey actually do?

The survey asked: “Should former National Security officials who become consultants and TV news contributors retain or give up their national security clearances?” In answer to that question, 60 percent of registered voters said the former officials should relinquish their security clearances. The following question had to do with Brennan specifically.

According to the survey, 57 percent of registered voters said Brennan should have lost his security clearance “after he started making public statements.” In answer to a similarly worded question, a similar number (59 percent) said that Brennan should have lost his security clearance.

Should former National Security officials who become consultants and TV news contributors retain or give up their national security clearances?

Should retain security clearance: 40%
Should give it up: 60%

Do you think former CIA director John Brennan should have retained or lost his national security clearance after he started making public statements?

Retained his clearance: 43%
Lost his clearance: 57%

Do you think former CIA director John Brennan should have retained or lost his national security clearance?

Retained his clearance: 41%
Lost his clearance: 59%

In sum, what the survey tells us is fairly simple: most Americans think all former national security officials should lose their security clearances when they become consultants and TV news contributors, and almost the exact same percentage say that the former CIA director should as well once he started making public statements (in news media). But that information does not touch the heart of the issue with which our republic has had to grapple: whether it is legitimate for the White House to selectively revoke security clearances based on the particular content of political speech such as criticisms of the president.

A less important concern but not a minor quibble is the ambiguity in how respondents might have understood the wording of the question whether former officials should “retain or give up” their clearances and the question whether Brennan should have “retained or lost” his. The first formulation sounds like a potentially voluntary action on the part of the individual, and neither formulation explicitly involves a presidential decision to revoke the clearance.

A final quandary is how much faith we should have that respondents knew about even the bare bone details to answer these questions in an informed manner. Did they, for example, know who Brennan is, what Brennan said, or the like? It is remarkable that none of these questions appear to have provided an option for “Don’t Know,” “Unsure,” or “No Opinion”—options which were included in several other questions in the same Harris poll. Indeed, the poll was conducted among 1,330 registered voters, and every single one of those 1,330 individuals apparently responded with one of the two choices given for the security clearances questions. What’s more, other questions in the poll suggest a significant lack of knowledge about political affairs. For example, the survey showed only 70% of respondents had “ever heard of” Jeff Sessions, 70% had “ever heard of” Robert Mueller, and 57% had “ever heard of” John Kelly. So how many respondents had a sufficient basis to answer the Brennan questions, which presumed a prior state of knowledge?

The survey has already entered the bloodstream of American political discourse, assisted by the President’s tweet and right-wing headlines and spin. The Harris poll may now have a self-fulfilling effect—causing those who oppose the president’s security revocation policy to become more silent, those who support it to claim vindication, and Americans in the middle to be swayed by a false narrative that more people support the president’s actions than do.