Holiday Survival Guide: How to talk about scientific studies around the dinner table

[Kristin] (0:00 - 0:10)
I went to ChatGPT to ask if 0.4% could be right, and it said that at that point, your blood would be flammable. I don't know if that's literally true, but I found it pretty funny.

[Regina] (0:11 - 0:24)
Flammable blood. Well, the site I was reading when I looked it up, it said if your blood alcohol level is at 0.4% right now, you are probably not reading this because there's a good chance you are unconscious.

[Kristin] (0:30 - 0:54)
Welcome to Normal Curves. This is a podcast for anyone who wants to learn about scientific studies and the statistics behind them. It's like a journal club, except we pick topics that are fun, relevant, and sometimes a little spicy.

We evaluate the evidence, and we also give you the tools that you need to evaluate scientific studies on your own. I'm Kristin Sainani. I'm a professor at Stanford University.

[Regina] (0:54 - 0:59)
And I'm Regina Nuzzo. I'm a professor at Gallaudet University and part-time lecturer at Stanford.

[Kristin] (1:00 - 1:05)
We are not medical doctors. We are PhDs, so nothing in this podcast should be construed as medical advice.

[Regina] (1:05 - 1:10)
Also, this podcast is separate from our day jobs at Stanford and Gallaudet University.

[Kristin] (1:11 - 1:22)
So this week, we picked a fun paper. It's about alcohol and language learning, specifically whether a drink or two actually helps you speak a foreign language better.

[Regina] (1:22 - 1:27)
Mm-hmm. After a margarita, suddenly your Spanish sounds amazing.

[Kristin] (1:28 - 1:32)
And there was actually a study on this published in 2018.

[Regina] (1:33 - 1:50)
Yep. That study just won a 2025 Ig Nobel Prize. The Ig Nobel Prize, for those who don't know, it's a real award given out for real science, but it's for studies that catch your attention because they are kind of odd or funny.

[Kristin] (1:50 - 2:09)
Mm-hmm. Exactly. They're a lot of fun, Regina.

And we're going to do something a little different in this episode. We do have a claim, which is that a little alcohol helps you speak a foreign language better. But this episode is not really about the specific claim.

It's about how to talk through a study like this.

[Regina] (2:09 - 2:31)
Right. Because we are coming up on the holiday season. And let's just say we know how it can go when you're sitting around the dinner table with, you know, all of your loved ones, some of whom live in other states.

They're talking about science, and someone is going to bring up that study that they saw on Facebook, and they're going to want to argue about it.

[Kristin] (2:32 - 2:56)
And, Regina, let's call this person, let's say, Uncle Joe.

[Regina]
Uncle Joe. Yes.

[Kristin]
I do not actually have an Uncle Joe, by the way. So instead of rolling your eyes or launching into a lecture to Uncle Joe, we are going to teach you how to walk him through a study. What to look for, what questions to ask, and how to sound calm and curious instead of combative.

[Regina] (2:57 - 3:32)
Definitely not combative. So you can kind of think of this episode as your holiday survival guide to talking about science with family. And Kristin, of course, we also did another fun twist for this episode.

And that's because when we were working on the episode, we were lucky enough to be at the Science Writers' Conference in Chicago, SciWri 2025, which is full of science journalists who are game to do anything for science. And should we say what we did, Kristin, or keep the suspense?

[Kristin] (3:33 - 3:51)
Well, build the suspense, of course. But Regina, I do think we've picked a topic that could come up at the Thanksgiving dinner table. And let's talk about why anyone would think this was true, that alcohol could make you better at a foreign language, because biologically, it doesn't really make a lot of sense.

[Regina] (3:52 - 4:16)
It doesn't. Now, we've already talked about alcohol. We've had a full episode about it.

That was about whether alcohol is good for your heart. And in that episode, we also talked about some of the other positive effects of alcohol, you know, relaxation, disinhibition, and your friend getting funnier. We did not, however, talk, Kristin, about whether it helps you speak a foreign language better.

[Kristin] (4:17 - 4:39)
We did not address that in that episode. And if anything, you'd think alcohol should hurt language performance, right? Because speaking another language relies on what psychologists call executive functioning, your brain's ability to plan, switch tasks, and prevent the wrong words from coming out.

So, logically, we would think that drinking would lead to worse performance.

[Regina] (4:40 - 5:10)
When I drink, the wrong words definitely come out, but maybe that's what makes it so much fun? I don't know. So, if you ask people who are learning another language, they'll often swear the opposite of what biology might predict.

They'll say that a drink or two actually makes them more fluent. And I got this advice, actually, for the competency exam for American Sign Language that faculty need to pass where I teach at Gallaudet University.

[Kristin] (5:11 - 5:31)
Interesting, Regina. And just for people who don't know, Gallaudet is the only university designed for deaf and hard-of-hearing students, and everything is taught in American Sign Language, ASL, no spoken language. And Regina, I still can't believe that you teach statistics in ASL.

That kind of blows my mind, and yes, maybe a glass of wine would help with that.

[Regina] (5:32 - 5:45)
Well, the statistics in ASL, a conversation for another day, but yes, some faculty advised me to have a glass of wine right before the test. I have not, however, followed that particular piece of advice.

[Kristin] (5:47 - 6:00)
You should try it, Regina, and report back, right? Take the test with the wine, and then try it again without the wine, and voila, we have a controlled experiment with N of 1, who is also a lightweight drinker.

[Regina] (6:01 - 6:02)
That might bias it, yes.

[Kristin] (6:03 - 6:07)
True. All right, let's get back now to how our conversation with Uncle Joe would go.

[Regina] (6:07 - 6:35)
Right, that leads us to the first question that you need to ask when you're looking at a paper, and that is, what specific things did the study actually set out to test? So, it's not enough just to pull the headline and say, well, the headline says this. It's not enough for the researchers just to say, we wanted to see if a couple of beers makes you better at a foreign language.

You need to have these specific measurable questions to answer.

[Kristin] (6:35 - 7:11)
Right, you need the research hypotheses. These are specific and detailed, and you can usually find them hanging out at the end of the introduction section in the paper. For this study, they actually specified four hypotheses.

We're going to focus just on the first two because these were the most important. So first hypothesis was that drinking alcohol would cause people to rate themselves higher on foreign language performance, but hypothesis two was that those same tipsy speakers would sound worse to outside raters because alcohol would mess with their cognitive function.

[Regina] (7:12 - 7:19)
So basically, does a drink make you actually speak better or just feel better and feel like you're speaking better?

[Kristin] (7:20 - 7:35)
It definitely makes you feel better. So yeah, it could just be that you think you're doing better. All right, so now we have identified the research hypotheses.

That's step one. The next question I would ask is, Regina, was this a randomized controlled trial?

[Regina] (7:35 - 8:04)
All right. We ask that because randomized controlled trials, they are the gold standard study design for proving causality. They're the strong studies.

And yes, this was a randomized controlled trial. The participants were randomly assigned to drink either vodka or water. And then after they drank that, they had to speak a foreign language for a couple of minutes.

And no, they did not get to choose whether they were in the vodka group or the water group.

[Kristin] (8:04 - 8:27)
Which is great, because if this was an observational study, meaning you just observe people who chose to drink or chose not to drink, then you would need to be even more skeptical. Because of course, drinkers and non-drinkers differ in lots of important ways. And as we talked about back in the alcohol episode, it's difficult to untangle drinking from all the other behaviors that go along with drinking, right?

[Regina] (8:27 - 9:06)
Oh, yeah. I can imagine that people who enjoy hanging out drinking in pubs, you know, for a hobby, they might have different reactions to speaking a foreign language after a couple of drinks. And non-drinkers, on the other hand, they might just throw up, which might be what I might do.

So, yeah, the study is not randomized. Then you need to be more skeptical and ask a lot more questions. But for today's episode, we're going to focus on a randomized experiment.

And we'll talk Uncle Joe through an observational study in a future episode. Oh, good. Yeah, let's do that, Regina.

[Kristin] (9:06 - 9:21)
All right. Randomized trials, as we've said, they're the gold standard, but it's the randomization that makes them so powerful. So we also want to probe things like how good was the randomization?

And Regina, did this paper give any details about their randomization scheme?

[Regina] (9:22 - 9:24)
Sadly, no details at all.

[Kristin] (9:25 - 9:46)
You know, Regina, that's a common problem, lack of details, that I see in randomized trials. And this matters because, for example, if they just said, hey, we're going to give alcohol to the first 25 people who walk in the room and then water to the next 25, that is not truly random. Or like, hey, we're going to give alcohol to the prettiest girls.

Also not truly random.

[Regina] (9:47 - 10:00)
Not at all random. Yeah. So in this case, it might be that the randomization was actually done well, but it's that we would like to see the details just to reassure us of that.

[Kristin] (10:00 - 10:20)
Exactly. So not a fatal flaw, but a bit of a question mark. Okay, now we know that the study design is a randomized controlled trial.

Next, we would want to walk Uncle Joe through a set of specific questions that we like to ask about study design. And these questions are so common, in fact, that people have come up with a mnemonic for remembering these. It's called PICOT.

[Regina] (10:21 - 10:23)
Which is a dumb mnemonic.

[Kristin] (10:24 - 10:25)
Yeah, totally.

[Regina] (10:25 - 10:25)
Like, what is PICOT?

[Kristin] (10:25 - 10:42)
It's not really memorable. I only remember it because I've seen it too many times. Now, the good thing is, is it does work for both randomized trials and observational studies.

So if we were talking through an observational study, this framework would still apply. Regina, do you want to walk us through what PICOT stands for?

[Regina] (10:43 - 10:53)
Oh, I would love to. P. P is for penis.

No, wait, wait. It's for population.

[Kristin] (10:53 - 10:55)
Okay. Good try. Good try, Regina.

[Regina] (10:56 - 11:00)
Oh, come on. You were expecting that, Kristin. You know you were.

[Kristin] (11:00 - 11:14)
I was absolutely not, although I guess I should have been by now. Okay. P is for population.

Don't confuse people, Regina. Population is referring to the who, like what group of people, who are you choosing for your study?

[Regina] (11:14 - 11:31)
P, population. Yes. P is for participants.

I is for intervention. What intervention, what treatment did they give the participants? What variable did the researchers manipulate?

C is for comparator. What did you compare that intervention group to? What's the control group?

[Kristin] (11:32 - 11:39)
C, comparator or control, and control groups are so important, as everyone who's taken fifth grade science should know. Even Uncle Joe.

[Regina] (11:40 - 11:57)
Should know. Even Uncle Joe. Exactly.

Okay. P, I, C, O. O is for outcome.

What outcome variable did they measure and compare between the two groups? And T is for time frame. Over what time period did they look?

[Kristin] (11:57 - 12:10)
So it's like the journalists, who, what, where, when, and how, but for science, Regina. All right, Regina, let's start with P for population. And the questions I'm going to ask here are who was in the study and how big was the sample size?

[Regina] (12:10 - 12:51)
Right. So the WHO was undergraduates at a university in the Netherlands. Now Dutch was actually their foreign language.

They were native German speakers enrolled in a Dutch language psychology program. That is a very specific group. It is a very specific group, but it's also a strength of the study in a way, because the researchers wanted to make sure that the participants were all pretty similar in their language skills and that they were all fluent enough to even do this.

So they didn't get like Americans in Amsterdam just trying to order a beer in a bong. That wouldn't work.

[Kristin] (12:52 - 12:59)
That probably wouldn't have worked so well. And Regina, how did they actually recruit the students for the study? Like did they find them at the local bar or what?

[Regina] (13:00 - 13:02)
Yeah, they advertised on campus with posters.

[Kristin] (13:03 - 13:17)
Ah, okay. That's a typical way to recruit on college campuses. But notice, Regina, we have a pretty select group here.

They have to have passed this very specific test, be willing to drink, and be willing to spend their afternoon participating in a study.

[Regina] (13:19 - 13:36)
Which is important when you think about generalizability, right? Who can we generalize these results to? They apply to German-speaking psychology students at a Dutch university.

Not necessarily to you or your Uncle Joe after two beers on Thanksgiving.

[Kristin] (13:37 - 13:40)
Uncle Joe is not going to start spontaneously speaking Dutch after a few beers.

[Regina] (13:41 - 13:43)
Well, I don't know. Maybe.

[Kristin] (13:45 - 13:50)
All right. Regina, the other important question about population is what was the sample size?

[Regina] (13:51 - 13:55)
Yep. Just 50 people total, Kristin. Only 25 in each group.

[Kristin] (13:56 - 14:05)
That's not super big. And when a randomized trial has small samples like this, you can end up with imbalances between the groups just by chance.

[Regina] (14:06 - 14:19)
And when you have small samples, it's more likely that any signal you do find is just a fluke, right? Or looks bigger than it actually is. This is called winner's curse.

We talked about this back in the red dress episode.

[Kristin] (14:20 - 14:43)
That's right. You know, and Regina, that study had only like 11 or 12 per group. So we are actually doing quite a bit better than that.

Still, 50 is not huge, so we have to keep that in mind. Okay, Regina. So that's P again, population, not penises.

What about the I part? What intervention did they actually give people in the experimental group? The alcohol group got alcohol.

[Regina] (14:43 - 15:24)
Not surprising. Surprise, surprise. Surprise.

Specifically, something called Smirnoff Red, which I believe is cherry flavored. Ew. And sounds kind of gross.

Yeah. It's 37.5% alcohol by volume. And I looked that up.

Pretty standard for distilled liquor. Now the researchers mixed it with bitter lemon, which is apparently like tonic water, but with a lemon peel and lemon pith added. So it sounds kind of gross overall.

I don't know. Yeah. Cherry and lemon.

Yeah. And they asked the participants to drink it within 10 minutes.

[Kristin] (15:24 - 15:43)
Oh, that's pretty fast. I mean, not chugging it, but still. And they were trying to get their blood alcohol to a modest level, so I guess they didn't want them sitting around, you know, sipping for too long.

And Regina, they actually used an online calculator to set the alcohol dosage for each person individually according to their sex and body weight to get the right amount of vodka.

[Regina] (15:43 - 16:04)
Right. Very scientific. Now, Kristin, funny thing, when I first read the paper, I thought it said they were aiming for a blood alcohol level of 0.4%.

[Kristin]
You know, Regina, I thought the same thing because the numerals given for the blood alcohol were 0.4, and I was just paying attention to those numbers.

[Regina] (16:05 - 16:35)
Right. But it was followed by this funny symbol that almost looks like a percent sign but isn't. And so at first I thought it was just some fancy European percent sign.

[Kristin]
Me too.

[Regina]
But then I looked up 0.4% blood alcohol, and that got a bit of a shock. So I put my sex and weight into an online calculator to find out what it would take me to get to that level, and it said I would need to have seven drinks in 15 minutes.

[Kristin] (16:36 - 17:00)
Regina, if you did that, you would quite literally be dead of alcohol poisoning. I think, actually, if you had four drinks, you would probably be dead of alcohol poisoning, knowing you at least passed out. Regina, you know, I went to ChatGPT to ask if 0.4% could be right, and it said that at that point your blood would be flammable. I don't know if that's literally true, but I found it pretty funny.

[Regina] (17:01 - 17:22)
Flammable blood. Well, the site I was reading when I looked it up, it said if your blood alcohol level is at 0.4% right now, you are probably not reading this because there's a good chance you are unconscious. That this indicates alcohol poisoning, and there's a very serious chance of long-term damage or death.

[Kristin] (17:24 - 17:33)
Yeah, so at first we were pretty convinced, Regina, that it was a typo, but then we realized that that fancy symbol is actually not a percent sign.

[Regina] (17:33 - 17:55)
Not a percent sign. It was a per-mille sign. I don't even know how you pronounce it.

You can see that it has an extra zero, right? So the percent sign basically has a slash and two zeros in it, so it's per-cent, per-hundred. But the per-mille sign has a slash and three zeros, so it's per-thousand.

[Kristin] (17:55 - 18:00)
Right. I actually do know some French, and mille, in French, means 1,000.

[Regina] (18:01 - 18:37)
Ah, mille, per-mille. Maybe that's how it's pronounced?

[Kristin]
Yeah.

My French is a little rusty, but I think so.

[Regina]
Have you had drinks to try to boost it up before this?

[Kristin]
I have not.

[Regina]
You have not. Okay, so it's 0.4 per-mille, which is 0.4 out of 1,000, not 100. Big difference there.

To translate it into normal percentage, we need to move the decimal place one to the left, so it's 0.04%.

[Kristin] (18:37 - 18:54)
Which is actually much more reasonable, because 0.08%, that is the cutoff for the legal limit for driving. So above 0.08% is drunk driving. And Regina, it's kind of cool. I learned something new here.

I had never seen a per-mille sign before, and honestly, I don't love it. It's kind of confusing because we're all used to thinking in percents, but it's still kind of cool that it exists, and now I know about it.

[Regina] (18:54 - 19:01)
Yeah, I can't wait to trot this one out on like a bad first date or something at a bar. It's kind of boring.

[Kristin] (19:01 - 19:06)
If the conversation gets really boring, then we go to per-mille signs.

[Regina] (19:07 - 19:10)
Per-mille signs. Whew. You can tell I'm a lot of fun to hang out with.

[Kristin] (19:12 - 19:20)
Well, you know, that might be a great screen for somebody who is going to be good for you to date, Regina. If they enjoy the per-mille story, I'd say they're a keeper.

[Regina] (19:21 - 19:39)
Ooh, I like it. It's like a litmus test for hot nerds. Okay, so while I'm at the bar with this guy talking about per-mille and thinking about .04% blood alcohol level, I wasn't sure what it actually feels like. Do you know, Kristin? I looked it up.

[Kristin] (19:39 - 19:46)
Well, I'm kind of guessing since it's under the legal limit for driving, you're feeling a little buzzed, but you're definitely not flat-out drunk.

[Regina] (19:46 - 20:08)
Yep, exactly. One website said .04 to .07 level, you're feeling relaxed, talkative, mildly euphoric, your inhibitions are lowered, your reasoning and memory are impaired slightly, your coordination might be mildly impaired. So they're actually selling it.

Like this sounds like fun, right?

[Kristin] (20:08 - 20:20)
Hmm, that's a very nice description, actually. And, you know, Regina, one thing I like about this study is they didn't just ask people, like, do you feel like you're at .04% blood alcohol level? They actually gave everybody a breath test to check if they got the levels right.

[Regina] (20:20 - 20:41)
Mm-hmm. They said they managed to get the participants to between .02% and .06%, with an average of .03%. Now, they were aiming for .04, so it's not perfect. Some of these participants were drunker than they were supposed to be, and some were more sober, but it's not bad.

Not bad.

[Kristin] (20:41 - 20:51)
Yeah, I mean, it's not bad. It would have been nice if they had topped off the more sober people, but, you know, I can imagine that's quite logistically tricky, and at least they checked everyone, and they're all in the right ballpark.

[Regina] (20:52 - 21:01)
Okay, now that we have cleared up that the intervention did not actually make their blood flammable, I think we're ready to move on to C for comparator.

[Kristin] (21:02 - 21:41)
All right, that sounds good, Regina, but let's take a short break first.

Welcome back to Normal Curves. Today we're helping you discuss research studies with Uncle Joe around the Thanksgiving dinner table.

And we are walking our way through the PICOT framework, and we were at C. Regina, do you want to explain C?

[Regina] (21:41 - 21:49)
Right. The C in PICOT stands for comparator. Basically, what are we comparing that intervention to?

What's the control?

[Kristin] (21:49 - 21:56)
In this case, the control group just got plain water. Same volume of liquid, same 10 minutes to drink it, but just water.

[Regina] (21:56 - 22:15)
Mmm, plain ice water. They definitely did not try to do a placebo here. They did not spike the water with a dash of lemon, right, to match the flavor or turn it pink like the Smirnoff Red.

So that means everyone knew what they were getting or what they were not getting. The participants were not blinded.

[Kristin] (22:16 - 22:37)
And I'm a little surprised that they didn't even at least attempt to fool people, because this could definitely create a bias because people might perform differently or rate themselves differently if they know that they got alcohol. Of course, it's definitely hard to blind alcohol studies because people can tell if they have had a drink, but I actually have seen studies where they were able to fool some people.

[Regina] (22:38 - 23:09)
So the researchers in the paper said that this was the intention because research shows that alcohol placebos can influence how you feel and impair your cognitive functioning, just like real alcohol does. So it's like the placebo effect of thinking that you're getting drunk can actually be as strong as getting drunk, because I looked at the studies and they were actually pretty interesting. So maybe we'll do an episode on the alcohol placebo effect.

[Kristin] (23:10 - 23:28)
Oh, I love that idea, Regina.

But this is interesting, Regina, because you're saying that the researchers didn't even try to correct for the placebo effect, but don't we usually want to try to correct for the placebo effect? Right. I mean, the hypothesis here was about actually drinking alcohol, not about thinking that you drink alcohol.

So it seems like they should have used a placebo.

[Regina] (23:29 - 23:45)
Yeah, it does, doesn't it? Maybe they were lazy. I can't tell.

I think they should have had three arms, right, one with vodka, one with water and then one with a believable mocktail. Like I've seen how they do this where they wipe the rim of the glass with alcohol to add that smell to it.

[Kristin] (23:45 - 23:52)
Oh, that is clever. Or, Regina, they could put sugar around the rim of the glass. That's how I used to like my alcohol.

[Regina] (23:52 - 23:54)
Mmm, I want salt on mine, though.

[Kristin] (23:55 - 24:00)
Ooh, or they could add an umbrella and a cherry. Then you really think you've got alcohol.

[Regina] (24:01 - 24:06)
They totally should have put the umbrella in that ice water glass.

[Kristin] (24:07 - 24:15)
Well, that is what I put in the green juice that I serve to guests now that I'm a teetotaler, just to try to make it seem more fun and less disgusting.

[Regina] (24:16 - 24:24)
I miss your green juice mocktails, Kristin. Yes, they were disgusting, along with the dirt juice, but they were more fancy with the little umbrella.

[Kristin] (24:24 - 24:45)
The umbrella makes all the difference. All right, so no placebo, no attempt at a placebo here. That is something to note for Uncle Joe.

Another thing to check with the control group is whether it's really a fair comparison. I mean, randomization should make the groups similar, but with a small sample, that doesn't always happen. And also, sometimes randomization just goes awry.

[Regina] (24:46 - 25:05)
We have seen many studies like this, Kristin, and if the groups end up being imbalanced because the randomization went off the rails, then if we do see differences in the outcome, that might just be due to imbalances in the groups rather than the actual intervention. That's the problem.

[Kristin] (25:05 - 25:10)
Exactly. And so, Regina, did they actually end up with balanced groups in this study?

[Regina] (25:11 - 25:31)
They said that both groups were 22 years old on average and had about 70 percent women and that there was no difference on their own self-assessed Dutch language skills that they asked before the experiment. So we don't have a ton of information, but based on what they gave us, they look pretty similar.

[Kristin] (25:31 - 25:43)
OK, so that's a good check, Regina. And randomized trials, really important that they report those baseline comparisons between the groups. All right, let's move on to O now, Regina.

[Regina] (25:47 - 25:50)
The big O. I'm always happy to move on to the big O.

[Kristin] (25:52 - 26:08)
OK, the O, let's remind everybody, the O in PICOT is for outcome. The next thing we would want to explain to Uncle Joe is what did they actually measure to decide whether the drink helped or hurt their Dutch speaking?

[Regina] (26:09 - 26:24)
Right. So 15 minutes after they consumed their drink, the participant needed to give a mini speech in Dutch for two minutes. They had to argue for or against animal testing, and they recorded them right there on the spot.

[Kristin] (26:24 - 26:44)
Oh, wow. Animal testing. That is a sophisticated topic.

I'm thinking their Dutch must have been pretty advanced, right? Just a little teaser here, Regina, in the fun experiment that we did at the Science Writers Conference recently, we did not make them discuss animal testing. They were more novice speakers, so we went with some more basic vocabulary.

[Regina] (26:45 - 26:50)
Yeah, basically, we went with whatever they wanted to talk about after we bought them drinks at the hotel bar.

[Kristin] (26:52 - 27:02)
But we'll get back to that at the end of this episode. All right. So they gave these speeches, monologues.

Who judged their Dutch skills, and what were they graded on?

[Regina] (27:02 - 27:51)
Right. There were two main measures, actually. So first, the participants rated themselves on how well they thought they did, and it was things like, was my vocabulary good enough to keep it all going?

Was my Dutch clear and understandable? And they scored each of these items on a scale from 0 to 100. And then the researchers averaged all of these scores together to get one overall self-rating of their language skills on this thing that they just talked about on a 0 to 100 scale.

So it scaled like an exam grade. Now, that was the self-rating, people rating themselves. And then there were also these two native Dutch speakers who listened to the recordings later and graded them on the same scales.

[Kristin] (27:51 - 28:04)
Sort of like teachers grading an exam. And Regina, one nice thing about this study, these outside graders did not know who was in the alcohol group and who was in the water group when they listened back to the speeches.

[Regina] (28:04 - 28:28)
Right. They were blinded, which is great, because if they'd known who drank alcohol, they might have subconsciously rated them higher or lower, depending on their expectations of what drunk Germans do when they try to speak Dutch. So a little spoiler here, the independent raters, Kristin, were actually nicer than the participants were to themselves.

[Kristin] (28:28 - 28:32)
Yeah, I found that really interesting. I guess we tend to self-grade more harshly.

[Regina] (28:33 - 28:33)
Yeah.

[Kristin] (28:33 - 28:52)
Now, Regina, I want to point out something. The paper did well. They made it clear that these two measurements, the self-rating and the independent rating, were the primary outcomes of the study.

And that's something you want to check when reading a study. Did the authors clearly identify the primary outcomes or were they cagey about it?

[Regina] (28:53 - 29:20)
Cagey, exactly. Because sometimes researchers try to, oh, let's just say hedge their bet, and they measure a whole bunch of outcomes and then just cherry pick anything that happens to show a signal. And when they do this, the results are not robust, because if you look at enough things, some things are going to come out statistically significant just by random chance.

[Kristin] (29:21 - 29:28)
This is the problem of multiple testing. We talked about that back in the alcohol episode with the CASCADE Wine Study. Do you remember, Regina?

[Regina] (29:29 - 29:32)
I do. It all comes back to alcohol, doesn't it?

[Kristin] (29:33 - 29:35)
Alcohol and penises. Two themes for this podcast.

[Regina] (29:37 - 29:41)
Which sometimes can be a great combination. Sometimes not.

[Kristin] (29:42 - 30:03)
In the CASCADE Wine Trial, they measured like everything under the sun, blood pressure, cholesterol, cholesterol ratios, LDL, HDL, insulin sensitivity. And then they cherry picked the few results that were statistically significant.

So those findings, as we talked about in that episode, were not robust.

[Regina] (30:03 - 30:27)
Right. We also talked about this problem of multiple testing in our Stats Reunion Review episode, and that was where we personified this whole thing with the character of multiple testing dude who's playing the field, but he's telling every hot date in every bar that, hey, you're my primary variable of interest, baby. It's just you.

[Kristin] (30:27 - 30:39)
Yep. So they get a gold star that they did not do that here. They were only dating two variables.

They had two clear primary outcomes, the self-rating and the independent outside rating.

[Regina] (30:39 - 31:03)
Right. And they did measure some other additional things, but it was nice because they made it clear that these were the secondary outcomes. And just for example, in addition to that overall score that I just talked about, the independent raters also gave separate scores on a scale of 1 to 10 on pronunciation, grammar, vocabulary, and argumentation.

[Kristin] (31:03 - 31:12)
Right. But they made it clear that these were not the main outcomes. All right, Regina, finally, we are getting to the tea and PICOT, the time frame.

[Regina] (31:12 - 31:19)
Well, here it's pretty short. They downed a drink in 10 minutes, they waited 15 minutes, and then they spoke Dutch for two minutes.

[Kristin] (31:20 - 31:26)
Right. So we're looking at an immediate effect, not whether alcohol helps you learn language over a long time.

[Regina] (31:27 - 31:31)
Right. Right. Quick buzz, not a whole semester of bar hopping abroad.

[Kristin] (31:32 - 31:49)
Yes. And in fact, I'm thinking that if you're drinking a lot over the long run when you're trying to learn a language, that might be counterproductive, like if you're out partying every night rather than studying. But it's important to understand the time frame of studies because some studies do in fact play out over many years.

[Regina] (31:49 - 32:28)
Yep. Okay, right. That's PICOT.

Now going through PICOT can help us understand the study design, the study methods. And Kristin, I just want to recap some of the highlights, the pluses and minuses on PICOT for this. It was randomized.

It was a small study, but not the worst we've ever seen. Pretty select group. It was a well-planned intervention that did not in fact involve a lethal amount of alcohol, despite our initial assumptions and confusion.

The observers were blinded, but not the participants themselves, and they did have clearly defined primary outcomes.

[Kristin] (32:29 - 32:56)
Right, so reasonably strong study with a few weaknesses or caveats. It did win an Ig Nobel Prize, after all. And even though the winners of that prize are funny, they do care about quality, the selection committee.

Regina, so that's the methods, PICOT, but you know, I don't know of a similar mnemonic for talking through the results. So I decided that I would come up with my own.

[Regina] (32:56 - 32:57)
Hmm. Did it involve sex? Kristin, please say yes.

[Kristin] (32:57 - 33:15)
No, I tried. I thought about that, but I'm very sorry to disappoint you, Regina. I did not come up with one about sex, but I think it's cute and memorable anyway and matches the theme of this episode.

So here we go. It's called SMART, S-M-A-R-T, because you want to sound smart at the dinner table.

[Regina] (33:16 - 33:20)
Oh, I actually like it. I mean, almost as good as the sexy angle.

[Kristin] (33:20 - 34:04)
I leave the sexy angle for you, Regina, but SMART, let's go with it. All right, here it is. S is for signal.

Does the signal exceed the noise, which we usually evaluate with statistical significance? M is for magnitude. How big was the effect?

A is for alternative explanations for the results. Like, could the results be explained by biases or confounders? R is for reality check.

This is the gut check. Do the results pass the smell test? Do they make you feel good?

Does it make sense? And are there any red flags or errors that undermine the study? And finally, T is for trustworthiness or transparency.

Basically, were they honest and transparent, which, of course, increases our trust in the work?

[Regina] (34:04 - 34:54)
Oh, Kristin, I know I complained, but I actually like this a lot. You did a very nice job here. So let's test this out on this study with Uncle Joe.

But, Kristin, let's take a short break first.

Welcome back to Normal Curves. Today, we are walking through a study about alcohol and foreign languages.

Ooh, la la. At the dinner table with Uncle Joe. And, Kristin, you were about to walk us through your smart mnemonic.

[Kristin] (34:54 - 35:05)
Right. Let's start with S for signal. Were the results statistically significant or, Regina, what you call statistically discernible?

And here we need to look at the P values.

[Regina] (35:06 - 35:24)
I just want to jump in here and confess, Kristin, that I don't love that we're starting with P values instead of magnitude, because I always start with how big rather than, you know, how significant. But I do realize that M-SART does not have the same ring to it. So just wanted to get that out there.

Start with P values.

[Kristin] (35:24 - 36:06)
You know, maybe if I'd come up with a French mnemonic, but this was kind of the best I could do in English. So we're going to start with P values. And I'll remind everyone that we talked about P values back in our P value episode.

And we talked about statistical significance in our stats reunion episode. And, Regina, you'll notice how we are referring back to lots of previous episodes. So I want to tell our listeners, if you really want to be prepared for Thanksgiving dinner, we highly recommend binging on our previous episodes, if you haven't listened to them already.

Lots of good lessons there for talking with Uncle Joe and sounding smart at the dinner table, like you can whip out Mendelian randomization and inverse probability of treatment weighting.

[Regina] (36:07 - 36:45)
Yeah. And, Kristin, after a couple of drinks, you could probably explain those better or at least maybe think that you're explaining that better. Probably just think that you're explaining them better, Regina.

OK, but back to S, poor signal. We're talking about P values here. And here are the self ratings.

That was where the participants rated their own performance. The difference between the two groups was not significantly different. The P value there was 0.64, which is definitely bigger than 0.05. But the independent ratings, those outside raters, the difference between the two groups was significantly different. P value there was 0.02.

[Kristin] (36:46 - 37:02)
Regina, this is fascinating because they found the opposite of what they were expecting, right? They did not find that the participants in the alcohol group rated themselves better than those in the water group. But they did find that the independent raters rated the alcohol group more highly.

[Regina] (37:03 - 37:41)
Yeah, it's the exact opposite what they hypothesized. But to their credit, the authors were open about that. They were open minded, too.

The data went against their hypothesis and they talked about it. They confessed to it. But interestingly, the authors also did some exploratory analyses and they found the characteristic where those independent raters found a significant difference.

It was not on grammar, not on vocabulary or argumentation. It was on pronunciation. Does that make sense, Kristin?

The whole effect might be driven by an actual fluency in producing the sounds, you know, the accent, the flow.

[Kristin] (37:42 - 38:15)
That's interesting. So the overall effect was significant and this one characteristic was significant. It's at least consistent.

And Regina, I just want to point out, we'd be more skeptical if the results turned out to be that the overall grade did not differ and only the pronunciation got better. That would feel a little cherry picked, right? Because of an exploratory or secondary outcome is significant, but the primary one is not.

You don't want to make too much of a story out of that. That could just be a lucky chance finding. But that wasn't the case here.

The results were consistent. The primary outcome and one of the secondary outcomes.

[Regina] (38:15 - 38:30)
Right, right. So we do like that. OK, we've got S.

Now we are ready to move on to M in your smart mnemonic, magnitude. So forget about P values and whether it was statistically discernible. Was it big enough?

Was the effect big enough to care about?

[Kristin] (38:31 - 39:12)
So the objective graders rated the alcohol drinkers about five points higher than the water drinkers on this zero to 100 scale. It was an average score of about 57 in the water group versus 62 in the alcohol group. So not huge, but not nothing.

It's kind of like going up half a grade on a test. Right. I mean, I guess from a F to a D minus.

But this isn't exactly like an exam grade. And, you know, Regina, I gave my undergraduates a contest recently and the prize for the winner was five bonus points on the midterm exam and more than a third of the class bothered to enter. So clearly five extra points was very meaningful to them.

[Regina] (39:13 - 39:55)
Oh, a contest. Oh, that's exciting.

[Kristin]
Yeah, we made beautiful graphics in R. It was fun.

[Regina]
Oh, my favorite.

OK, back to our participants. That was the independent raters. But when the participants rated themselves, they gave an average of 54 in the water group and only 56 in the alcohol group.

And it was interesting because the standard deviation was quite high. It was 15. And that means there's a lot of variability.

That standard deviation of 15 tells us that the typical deviation from the mean is about 15. So the individual scores probably varied quite wildly. Maybe a lot of people around 70 or even as high as 85 or, you know, a lot of people as low as 40 or as low as 25.

[Kristin] (39:56 - 40:21)
You know, this is why I hate self-grading, Regina. It's just too hard for people. Nobody knows how to grade themselves.

You don't want to seem braggy, but you don't want to be too self-deprecating. So it's kind of no wonder to me that those are all over the map. And, Regina, I'll point out that the standard deviation among the outside raters was much, much smaller.

There was a lot more consistency because they were probably trying to grade in a very consistent manner.

[Regina] (40:22 - 40:31)
Yep. Good point.
OK, Kristin, what is next in SMART? It's A, A for alternative explanations. Right.

Which I love. What else could have explained the results?

[Kristin] (40:32 - 40:46)
Right. This is where we look at possible biases or the role of random chance. And in this case, we have to look at things like the participants weren't blinded and self-rating is hard, as we've said.

So that could have affected those self-rating results.

[Regina] (40:46 - 41:11)
Right. Or even affected their actual performance, right? If they knew they were getting alcohol, maybe they expected to feel more relaxed, which actually made them more relaxed.

Yeah, could be. Also, maybe the participants rated themselves lower because they knew they had the alcohol and they were expecting to do worse. I don't know.

But this would not have affected, Kristin, the independent raters because those were blinded.

[Kristin] (41:11 - 41:31)
Right. And that is a strength of the study, the blinding of the independent raters. Regina, we do also have to worry here about random chance.

They did pre-specify two primary outcomes, as we've said. That's good. But sometimes there are false positives, especially with a small sample size.

Maybe a few really good speakers ended up in the alcohol group just by chance.

[Regina] (41:31 - 41:34)
Exactly. Which would mean it has nothing to do with the alcohol.

[Kristin] (41:35 - 41:35)
Right.

[Regina] (41:35 - 41:41)
OK, so that was A, alternative explanations. Now, what about R, reality check?

[Kristin] (41:41 - 42:34)
All right. This one's my favorite. This is where we do a gut check.

Like, are there things in the study that just don't make sense at face value, numbers that just couldn't be possible? Or is the math obviously wrong? Like, remember back in the vitamin D 1 episode, we had a situation where it was reported that 76 percent of 40 moms were vitamin D deficient, which is impossible because that would mean we would have a fraction of a mom.

Or are there graphs that are obviously wrong? Like back in the chat GPT episode, there were bar graphs that showed 63 people when there were only 55 in the study. And there were graphs where 100 percent of people had no response, but then there were responses.Didn't make any sense.

Regina, I didn't find anything obvious here, like no glaring red flags jumping out at me. Did you see anything?

[Regina] (42:35 - 42:46)
Well, I found a couple of things that weren't consistent. Nothing huge, but the numbers didn't add up for the part where they asked people to rate their Dutch skills in general before the experiment.

[Kristin] (42:47 - 42:48)
Oh, those baseline ratings. Yes.

[Regina] (42:49 - 43:04)
Right. They asked everyone to rate themselves from one to five. Well, one is poor, five is excellent.

And it was on four dimensions. They said they had a sample size of 50, and then they reported the means and standard deviations for the full group of participants.

[Kristin] (43:05 - 43:48)
And I quickly subjected these statistics to the GRIM test, which is a fun and easy statistical sleuthing test that we talked about in the Scary Bridge Halloween episode. So easy that even Uncle Joe could do it. We're assuming here that that one to five scale is what we call a Likert scale.

That is, you can only pick integers. And when you have these Likert type whole number ratings and you know the sample size, it turns out that some means and standard deviations are actually mathematically impossible. So if you detect means and standard deviations that are mathematically impossible, that tells you that something went wrong.

And did you find mathematically impossible results, Regina, what we call GRIM Reaper math?

[Regina] (43:49 - 44:23)
GRIM Reaper math, I did, which does not fit well with the Thanksgiving theme. But that's OK. That's OK.

Yeah. OK. I did.

So they had, like I said, the means and standard deviations for four different measures. The results for one measure were fine, but not for the other three. So I went back and checked even further.

So two of those three mathematically impossible results, two of them do work if you have a sample size of 49, not 50. But one did not work for any sample size I tried.

[Kristin] (44:24 - 44:35)
Right. So, you know, it could be the case that one person didn't respond on a few of the questions, so they didn't have all the data and the researchers didn't tell us that the sample size was incomplete for those measures. That does happen sometimes.

[Regina] (44:35 - 44:48)
Yeah. You know, it's not it's not huge. It's not like we found tables and tables of mathematically impossible results.

Or if we found something where it looks like the sample size was consistently, like, very different from what was reported.

[Kristin] (44:49 - 44:56)
Right. It's also not on the outcome variables. It's on something they measured at baseline.

But, you know, still, it means the paper's not perfect in the math.

[Regina] (44:57 - 45:34)
Right. Not the end of the world. I just checked it because it is easy and fun to do.

But reality check, a different type of reality check, Kristin. I can't help but wonder, is one drink and a 0.04% blood level really enough to bring out the effects of alcohol in this group? Right.

Picturing I'm a young German college student going to school in the Netherlands. I probably have several beers for breakfast every morning. So a little vodka tonic in the lab in the afternoon.

Is that really enough to get my inner language superstar flowing? I don't know.

[Kristin] (45:34 - 45:38)
That's a good point. Yeah. 0.03% average, not a huge amount.

[Regina] (45:38 - 45:57)
I also wonder about whether this task means anything in the real world. Kristin, you know how many times I'm just sitting in a bar happily drinking the night away and someone comes up to me and tells me to give a two minute argument about animal testing.

[Kristin] (45:58 - 46:33)
I'm guessing that hasn't happened a lot.

[Regina]
Right. Yeah. Yeah.

[Kristin]
This may not be a practical, right. We might have wanted to test like how well they did in a conversation, for example. Right.

Regina, we also like to ask here about biological plausibility. Is this even plausible? And I think the authors had some plausible explanations.

For example, they said it might have to do with the disinhibiting effects of alcohol. Right. People probably get pretty anxious when they have to speak a foreign language.

Right. So maybe they felt less anxious and thus spoke a little better with a modest amount of alcohol because they felt a little more uninhibited.

[Regina] (46:34 - 46:42)
Yeah. So they're less self-conscious and they could just go with it and let their brain do its thing instead of getting all caught up in their head.

[Kristin] (46:42 - 47:19)
Exactly. OK. That was R for reality check.

Now T is for trustworthiness or transparency. And this is where we ask things like, did the authors pre-register the study and did they provide their raw data and code? And just to remind people, study pre-registration is where researchers publicly post their protocol ahead of time, including all their data analysis plans.

And this is to rein them in and keep them from cherry picking and cheating on their data analysis later. Regina, I didn't find, however, that this study was pre-registered and I couldn't find their data anywhere. Did you have any luck?

[Regina] (47:19 - 47:23)
No, I couldn't either. So that's not great.

[Kristin] (47:23 - 47:32)
Right. I will say that most studies do not provide pre-registration or raw data, but we really like to see that because that definitely increases our trust in the study.

[Regina] (47:32 - 47:58)
Yeah, I was hoping for this. But on the plus side, I think there were a few other things they did that increased my trust in them. Like you said, they were transparent about what were their primary analyses and what were their exploratory analyses.

And that's nice. And another thing that I liked, they were honest that they didn't find evidence to support their hypothesis. But I think what they found was even more interesting, but they were very transparent about that.

[Kristin] (47:58 - 48:53)
Right. They didn't try to make the data fit into what they thought was going to happen. Right.

OK, Regina, to recap SMART, they found the opposite of what they were expecting. The participants in the alcohol group did not rate themselves better, but the independent raters did. It was about a half a grade worth, so not huge, but not nothing.

We found a few potential arithmetic errors, but nothing that would undermine this study. We don't have their raw data or their protocol to check things. So I would say, you know, on balance, the results are interesting, but maybe not a total slam dunk.

And let's also do a quick recap of how we talked through this paper. We first asked, what was the study hypothesis or hypotheses? We then asked, was it a randomized trial or observational study?

And then we applied the PICOT and SMART frameworks to talk through the study. SMART being an invention of our own.

[Regina] (48:54 - 48:59)
Let's do the pop quiz to make sure we remember PICOT and SMART. P is for?

[Kristin] (48:59 - 49:00)
Population.

[Regina] (49:01 - 49:24)
I had to do it. Intervention, comparator, outcome, time. SMART, S is signal.

M is magnitude. A is alternative explanations. R is reality check.

And T is transparency and trust. Oh, Kristin, I really like SMART. I like it so much more than PICOT, I've got to say.

[Kristin] (49:24 - 49:26)
Well, it's easier to remember.

[Regina] (49:26 - 49:29)
Yes, it is. All right. I think we are ready to wrap up now, Kristin.

[Kristin] (49:29 - 50:17)
Everybody's ready for Thanksgiving dinner. But first, Regina, as promised, we have to play our experiment from Science Writers 2025. We, of course, pre-taped this.

So I'm going to play it now.

So Regina and I are here at Science Writers 2025, and we decided to recreate this little experiment. So we have some guinea pigs, Christie Aschwanden and Helene Engler, and they're both going to try out our little experiment here with French before and after drinking alcohol.

[Christie]
I'm Christie Aschwanden, author of Good to Go, What the Athlete and All of Us Can Learn from the Strange Science of Recovery, which seems like it's a sports book, but it's actually a book about statistics and study design. Unfortunately, my editor made me take out the section on p-values, but I can assure you that the first draft had it in there.

[Kristin]
All right, Helene.

[Helene] (50:17 - 50:40)
I'm Helene Engler. I am an assistant professor at the University of Texas in Austin. I used to be a butterfly biologist, and then I reinvented myself as a science writer for K-12 and then a journalist.

And then I went back, and now I'm teaching at the University of Texas. So, yeah, and I am actually writing a book about butterflies and a science memoir, and I'm at the conference. And meeting wonderful people.

[Christie] (50:41 - 51:01)
I'm learning French because I want to try and be better able to talk to my husband's family in Switzerland. He has some cousins that do not speak English. They speak French.

And we were actually wine touring with them. They're in the wine business. And so I did notice that after some alcohol, we were communicating better.

True story.

[Kristin]
Well, I think that proves the hypothesis for today.

[Helene] (51:02 - 51:18)
I was married to a Frenchman for 10 years, and so I learned a little bit of French. It's been 20 years since I took a French class. So, but my kids are half French and none of us can really speak the language.

And I'm really interested in learning it again. So this has inspired me to start up again.

[Kristin] (51:19 - 51:39)
Excellent. Just a little alcohol on the side and you'll be fluent. All right.

So they are going to try speaking a little French before alcohol and then after we get them a drink. So who wants to go first? Helene?

[Helene]
Bonjour. Bonsoir. Je voudrais une verre vingt des rouges.

[Kristin]
Excellent.

And now Christie.

[Christie] (51:43 � 52:14)
Bonjour.
J'ai voudrais une vingt.

J'ai a sept chats et un ch�teau.

[Kristin]
Excellent. And let's because in the experiment, they actually asked the participants how they thought they did.

So how do you think you did on your French?

[Christie]
I think I was terrible. I'm really struggling to learn French.

I know some Italian and German, and it's much easier. You can't look at a French word and know how it's pronounced.

[Kristin]
A little alcohol may help.

OK, Helene, how do you feel like you did?

[Helene] (52:14 - 52:27)
I also think I did pretty, pretty terrible. Mal. Pas bien.

I learned French 20 years ago and I was married to a Frenchman. I think I've blocked it from my brain. So maybe the alcohol will help me remember.

[Kristin] (52:28 - 52:47)
All right. So now we're going to go get the alcohol and then we'll come back and see after the bar. So Christie and Helene have now had vodka.

One drink each. That should get them to about the 0.04 percent that we're looking for. Clearly, this is highly scientific, but we've waited the appropriate 15 minutes and now they're ready to try their French again.

Helene, you want to go first?

[Helene] (52:48 - 52:57)
Vive la France. Je voudrais beber une bi�re. C'est la vie.

C'est bon. Au revoir.

[Kristin] (52:58 - 53:01)
Excellent. All right. And then Christie.

[Regina] (53:01 - 53:13)
Je voudrais une bi�re et vignt. J'ai un chat.

Chateau. Bienvenue. Bonjour. Bonsoir.

[Kristin] (53:14 - 53:24)
Excellent.

All right. So now we're going to let them self-judge. Do you think you did better this time?

[Helene] (53:24 - 53:34)
I did. I felt more comfortable. I felt like phrases were coming to my brain.

French phrases. It was out of nowhere. So, yeah, I feel very good about it.

[Christie] (53:35 - 53:47)
Oh, I did so much better.

You know, beer goggles. Although I did have a vodka tonic. Just it's all confidence.

[Kristin]
That was the hypothesis of the study. That it was confidence. But I don't think that's what they found in the end.

[Regina] (53:48 - 53:50)
They also asked about self-esteem.

[Kristin] (53:50 - 53:52)
How's your self-esteem right now?

[Helene] (53:52 - 53:59)
I feel so much better about myself. And I'm not a drinker. I feel like I, yeah, I need to drink some more I think.

[Kristin] (54:00 - 54:00)
Christie?

[Christie] (54:01 - 54:06)
Well, I'm feeling really great about myself right now. Because I've got a little buzz. And people always tell me that I'm so fun when I'm drunk.

[Kristin] (54:09 - 54:18)
Yeah, we'll second that. And I don't know. Do we have to judge?

I think they did better after the wine. What do you think, Regina?

[Regina]
I don't know.

[Kristin]
It's not wine. Sorry.

[Regina] (54:22 - 54:26)
They seemed much more confident and fluent. And just kind of relaxed and a little smooth.

[Kristin] (54:27 - 54:34)
I think they seemed a little smoother afterwards. Yeah, that's our unbiased take on it. Yes.

Thank you both for joining us for this little experiment.

[Regina] (54:34 - 54:48)
Let's go have some more.

Kristin, that was so much fun. So much fun to do in the moment.

But also so much fun to listen to as well. Christie and Helene are just a hoot.

[Kristin] (54:48 - 54:56)
Yeah. I really want to give them a big thank you for being the first guests on the Normal Curves podcast. And also the first drunk guests.

Maybe not drunk.

[Regina] (54:57 - 55:15)
Not quite. 0.04%. That was excellent. I hope we have more experimentation in the future.

Kristin ran around the conference asking random people, hey, do you want a drink for science? I can't believe we only got two people.

[Kristin] (55:15 - 55:28)
I know. I thought that was such an easy sell. I'm feeling, though, as a teetotaler, in pretty good company.

Because I didn't get a lot of takers on that one. Maybe it was the foreign language, though, and not the alcohol that was inhibiting people.

[Regina] (55:29 - 55:42)
Good point. Okay. So we usually, in our episode, use our Smooch rating scale for rating the strength of evidence.

But I'm not sure you should use this one with Uncle Joe.

[Kristin] (55:43 - 55:53)
It doesn't seem appropriate for the Thanksgiving dinner table with family. So I was thinking, Regina, maybe this time we use a turkey scale or a martini scale.

[Regina] (55:55 - 56:00)
Oh, martini.

[Kristin]
Pumpkin pie scale.

[Regina]
No, Martini. Well we�ve got to use a martini scale for this one in particular.

[Kristin] (56:00 - 56:13)
Oh, that's true.

Yeah, alcohol, right.

So one to five martinis. How strong is the evidence for the claim we looked at today? And remember, the claim is that a little alcohol helps you speak a foreign language better.

So, Regina, are you drinking martinis tonight?

[Regina] (56:14 - 56:30)
I was thinking about changing this to J�germeister shots. I don't know. It just seems even more exciting.

Okay, martinis. We'll go classy with the martinis. I think I'm going to drink three martinis.

[Kristin] (56:31 - 56:37)
Oh, my God. You're going to be drunk, Regina.

You may not even be conscious. I've never seen you consume three alcoholic beverages in one sitting.

[Regina] (56:38 - 56:42)
No, there's a good reason for that. It might not be pretty.

[Kristin] (56:43 - 56:46)
You might all of a sudden be able to speak Dutch, though.

[Regina] (56:47 - 57:30)
I might levitate at that point. Yes, so three martinis because for some strange reason I found this paper to be oddly convincing. And I don't know if that's just because it makes kind of common sense, but I like the randomized trial part of this, and maybe I like the fact that it came out opposite what they expected.

Things seem to fit in with a story that we could tell about this, and maybe I'm just falling to that kind of bias. But pronunciation, if this is all driven by pronunciation, and pronunciation is a little bit of inhabiting the language, then I could see this working. So, three martinis.

What the heck?

[Kristin] (57:30 - 57:38)
But, Regina, here's the real question. Are you going to have the glass of wine before the ASL test? Yes or no?

That tells us what you really think.

[Regina] (57:38 - 57:41)
I feel like I need to try it for science. Right?

[Kristin] (57:42 - 57:50)
I think you should. Yeah. In your case, maybe a half a glass, though, because I think that would get you up to 0.04%. You're such a lightweight.

[Regina] (57:51 - 58:00)
Yes. When I calculated 0.04%, it was actually less than a full drink. So, I think about...

[Kristin]
I believe that for you.
[Regina]
What about you?

[Kristin] (58:01 - 58:31)
I'm going to go with 2.5 martinis, though, make my non-alcoholic please. I'm somewhat convinced. I mean, it was a randomized trial, and it was small, but not a teeny randomized trial.

I also really like the fact that the raters who were blinded are the ones that found a difference, so we can be more convinced by that. There really is some biological plausibility here, right? If you loosen up a little, your language might flow a little better just because you're not tense and afraid of saying the wrong thing.

So, I'm halfway convinced. 2.5 for me.

[Regina] (58:32 - 58:34)
2.5 martinis. Okay.

[Kristin] (58:34 - 58:49)
Regina, I do want to point out something we've been talking about as we've gone along here. I don't know how generalizable these results are, right? This was this very specific group of students who spoke pretty good Dutch to begin with, and it was also as we talked about, a pretty small amount of alcohol.

[Regina] (58:50 - 59:02)
This was just a modest amount of alcohol. They were not trying to get them sloppy drunk, because that would make things worse for sure, slurring words, passing out. The amount of alcohol is very important.

[Kristin] (59:02 - 59:13)
Right. If you're unconscious, you're not going to speak Dutch better. Right.

So, if you're going to try this at Thanksgiving dinner, stick to like one, one and a half drinks. Once you're up to two or three, this is not going to work.

[Regina] (59:14 - 59:20)
Probably not. Okay, what about methodological morals? Do you have one for this one, Kristin?

[Kristin] (59:21 - 59:29)
I do. Here it is. You don't need a PhD to read a study. Just remember, PICOT and SMART.

Oh, I like that.

[Regina] (59:30 - 59:31)
That is very good marketing too.

[Kristin] (59:32 - 59:32)
How about you, Regina?

[Regina] (59:33 - 59:44)
I'm going to focus on that per-mille thing, because I still find that fascinating, right?

A decimal point can mean the difference between life and death. Details matter.

[Kristin] (59:44 - 59:56)
That is so true, Regina, because a decimal point to the left or right is an order of magnitude, and, you know, that can screw up medication and kill somebody for sure. Or alcohol levels.

[Regina] (59:57 - 59:59)
Per-mille sign. Good to know.

[Kristin] (59:59 - 1:00:08)
We all learned something from this episode. All right, Regina, I think everybody is now ready for the Thanksgiving dinner table. So, happy Thanksgiving early, Regina.

[Regina] (1:00:08 - 1:00:14)
Happy Thanksgiving, Kristin. I hope you enjoy your green juice cocktail, mocktail.

[Kristin] (1:00:14 - 1:00:19)
It will certainly have an umbrella and a cherry.

[Regina]
Thank you, Kristin.

[Kristin]
Thanks, Regina, and thanks everyone for listening.