President Obama’s State of the Union address tonight is expected to draw about 30 million TV viewers. Last year, about 33 million tuned in—a record-low reflecting either changes in how Americans get their news or, if you believe conservatives, Obama’s declining relevance.

But sheer numbers don’t tell the whole story; not every viewer included in the official count is really watching. Even public officials in attendance sometimes can’t keep their eyes open—Senator John McCain apparently nodded off during the 2007 address, as did Supreme Court Justice Ruth Bader Ginsburg in 2013.

How can Obama ensure that his listeners stay awake? According to a new paper by Dutch scholars published in the Journal of Language and Social Psychology, he should be more negative than positive—and speak in lists, if at all possible.

Luuk Lagerwerf, an associate professor of language and communication, and his colleagues at VU University Amsterdam designed two experiments to explore what techniques and rhetorical devices make audiences pay attention to and remember political speeches, from moment to moment and as a whole.

For the first experiment, Lagerwerf recruited volunteers at a university, a tennis club and a newspaper office, winding up with a sample of 120 people—62 percent female and with an average age of 29.6. The participants had to listen to a set of political speeches on the radio, given by a representative of a made-up political party called “Ons Belang” (“Our Interest”). They were given a keyboard and told to press the “+” key whenever they noticed their attention increasing.

The researchers manipulated the “valence framing” of the speech: sometimes issues were presented in a positive frame, and sometimes in a negative one. They also varied whether the speeches contained rhetorical devices like lists of three, contrasts (for instance, “We’ve done a lot, but we’ve got a lot more to do” as Obama told the Iowa caucus in 2012) and negative naming (phrases like “It’s the economy, stupid,” a slogan Bill Clinton used against Bush in 1992).

They found that negative framing increases attention—both at short intervals (“immediate attention”) and throughout the speech (“global attention”). The presence of rhetorical devices increased attention in positively framed speeches, particularly in the two-second interval immediately following the use of the device, but had no effect on negatively framed speeches.

In the second experiment, Lagerwef assessed listeners’ attentiveness by measuring their reaction times at a secondary task: The more absorbed people were in the speech, the slower their reaction times would be. He recruited 80 participants—evenly divided by gender, and with an average age of 30—at public libraries and universities, and had them listen to another set of political speeches, on the topic of either education or development and cooperation, and instructed them to press the space bar on a computer as soon as the screen flashed blue.

The results of this experiment confirmed the results of the first: Reaction times were slowest when participants were listening to negatively framed speeches with rhetorical schemes, followed by negatively framed speeches without rhetorical schemes; positively framed speeches without rhetorical schemes translated to the quickest reaction times. In short, the study says: “People pay more immediate attention to speeches if they hear negative messages, or if they are exposed to rhetorical schemes. Positively framed speeches without rhetorical schemes attract the least immediate attention.”