Trust in a Bottle

flosofl writes "The BBC has a report on oxytocin and its ability to skew our trust levels. 'The participants in the study played a game, in which they were split into "investors" and "trustees." The investors were then given credits and told they could chose whether to hand over zero, four, eight or 12 credits to their assigned trustee.' Some of the investors were given oxytocin via nasal spray. The results were surprising: 'Of 29 investors who were given oxytocin, 13 (45%) displayed "maximal trust" by choosing to invest highly, compared to six (21%) of the 29 investors who were given the dummy spray.' When the trustee was a computer, there was no difference between the two test groups."

Comments

Too Small of a Test

[rhino_badlands]

For this to really be worked out you need to do multiple test on a much larger scale. The people in the one group could have just been suckers.

Re:Too Small of a Test

[NoData]

What mean? Mean of what?
29 people is plenty of subjects for a reliable statistical test. The t distribution is about equivalent to the normal distribution at about 30 samples. 30 samples is about the usual rule of thumb for adequate power for a cell in a behavioral experiment. But, you know, it really depends on the effect size of whatever you're studying.

Anyway, the right test to do here, just from the tiny snippet of info we're given about the study is a chi square test. According to TFA, a subject could invest 0, 4, 8, or 12 credits. If we assume that we would expect a uniform distribution of investment across these levels (and I don't know if that's a fair distrubtion to assume, perhaps normal is better--you'd expect more people to invest middle amounts than extremes, perhaps), then we expect 7.25 people to fall in each of the 4 cells. For just the oxytocin condition, they report 13 people invested 12 credits. Let's assume that the remaining 16 subjects were evenly distributed among the 0, 4. and 8 investment levels. That means 5 1/3 people in each of those cells. With those data, the chi square test gives you a p value of .11 (Chi sqaure score of 6.08), which means a less than 11% chance of getting these results just by chance. That's not exactly meeting the 5% standard alpha level for significance, but then again, I've made some horrible simplifying assumptions that stack the deck against significance. Besides my made up data, I'm sure there's a repeated measures component to this study...I doubt each subject had to only make ONE investment decision in the whole experiment. The repeated measures would lend a lot more power.

Re:If memory serves me correctly...

[ultramk]

It's released a lot of times... usually during intimacy.

it's released:
- during sex
- when a mother holds her baby
- when nursing
- when two people are holding each other

I had a professor who called it "the Cuddle Drug". It's been thought to play a major part of the "bonding" process, parent/child as well as romantic relationships.

And no, it's not Oxycontin, which is a completely different thing.

m-

Horny geeks, take note.

[Jerk City Troll]

This hormone is produced when female nipples are stimulated. Maybe you guys can get her to trust you.

Re:Number of participants

[tktk]

I used to be in a graduate psychology program.

If it was a social psych. experiment, 29 participants would have been considered enough. I have no idea why.

I was in developmental psych. and was expected to get a minimum of 200 participants for my own study. Likewise, people in the cogntive psych. program had to get 100 or more participants.

Re:Number of participants

[Skippy_kangaroo]

Not necessarily.

The standard deviation of a binomial distribution is sqrt(n.p.(1-p)) where n is the number of subjects and p is the probability of maximal trust.

Thus, out of a sample of 29 people and with p=0.21 the standard deviation is 2.2.

Thus, 13 is 3.2 standard deviations away from 6. There is only a 0.07% chance that these are from the same distribution.

Thus, they can conclusively conclude that this spray had a statistically significant effect on trust.

You can make it more complicated if you wish but the basic fact remains that you can get statistically significant results from small samples. In this case there is only a 0.07% chance that they are wrong.

Re:Number of participants

[yali]

The experimenters reported p=.029 (one-tailed) with their results. If you accept the one-tailed test, it is "statistically significant" by conventional standards.

And N=58 (29 people per group) is pretty typical for single studies in the behavioral sciences. Ultimately, the grandparent is right -- this needs to be replicated. But that's true of single studies in any scientific field, no matter the sample size or p-value. This is an exciting enough discovery that you can bet lots of scientists are going to try to replicate it.

Re:It's a BS experiment.

[XorNand]

Wow... I guess you've never been in the position of having to sell a product or service. It's a very core principle of professional salesmanship: if the prospect doesn't trust you, he will never open his wallet. 90% of the sales process is working to build rapport and trust with that person. People don't decide to buy or not to buy based on reading the stacks of whitepapers and literature thrown at them. They don't call your previous customers to ask how your track record looks. It all boils down to how they personally feel about the guy trying to influence them, aka "trust".

Re:Corporate uses

[BlueFashoo]

You're thinking of oxycontin, a.k.a., hillbilly heroin, and Rush's drug of choice. What they use here is oxitocin, a neurotransmitter associated with bonding, orgasms, and the milk let down response.