i think i got fooled by your test Phillip. i got the question with the 3 in the first part, when i was looking at my choices quickly, i wanted to choose the 2nd option but i chose the third and submitted before i thought about it..(following the choose quickly thing). now i know what might have been the reason for that.. the bolded 3.. interesting experiment

Sammy Summers is a goddamn porn star name you fools, Fred is the better one. Anyway, to me it seems like the unseeded chose 1 & 2 because it was the 1st choices they saw, and disregarded the 3rd as "god, not ANOTHER choice!". While the seeded were consciously(or subconsciously) counting for 3 choices, seeing 1 as the start and then LOOKING for the end to end with 3, therefore paid most attention to the start and end, disregarding the 2nd as just "more irrelevant road to travel to get the destination of 3!"

Bob – trust the wisdom of the crowd. The question didn't specify a genre.. :D

Interesting experiment, Philipp!

I remember not noticing any kind of "A question with ...". Granted, I read your site from RSS and had to read the post a second time before clicking anything. I thought a few times about my choice to give the most honest answer, rather than click instinctively.

Disclaimer: I've done a few years of nlp and suggestion.
The user already knows there are 3 possible answers. 3 is also a very commonly used amount of possible answers for a question. I don't think the way you did it could have influenced the results.

Likely with stronger emphasis on the 3, you could achieve what you wanted. Here are a few tips:
- You bolded an entire sentence, the number goes unnoticed. Try bolding only the 3.

- You wrote something the user already knows (there are 3 results). Try writing something that isn't right (eg. 2 instead of 3)

- The notice wasn't mandatory read. Most people click buttons and sign papers without reading. Try making the 3 part of the questions themselves (not exactly fit for your test, but you see where I'm going with that).

- The experiment was not disparate enough. Try picking names from a larger pool at random, and giving, lets say, 5 possible answers.

Adys

> Sammy Summers was in the lead with 42%

You don't say. So what else is new?

http://google.com/search?q=%22The%20answer%20to%20everything%20is%2042%22

That out of the way, you, Philipp, concurred that the differences between seeded and unseeded returns were statistically insignificant. Even if they were, there's still one hitherto unpointed "anomaly" within them.
http://blogoscoped.com/little-experiment/results-rounded.png
While the "percentage gap" between the options in list item 1 and item 3 were the same, exactly one percent ("%"), the same in the middle item 2 jumped up by a factor of +200% to 2.1 "discrepancy gap percents."

Since both #1 and #3 were in effect "edge items," surrounded either above or below by sizable white space –which certainly abetted in their standing out on the page– #2 effectively was "boxed in" by its upper- and below "siblings."

This not-quite-yet-hypothesis of middle items' discovery disadvantage vs. that of those at either end could have been confirmed or rejected (if still with statistically insignificant values) by rerunning the test with the items arranged horizontally, or diagonally [think Mayakovsky, dammit!].

Sammy Summers made me think Sammy Davis Jr. and Susanne Summers.

> The experiment was not disparate enough. Try
> picking names from a larger pool at random,
> and giving, lets say, 5 possible answers.

Perhaps 5 choices, and an intro sentence including a lot of "4"s, as in "before", "forever", "4 seconds to choose", and so on. But then again it should probably also not be too obvious (or it would start testing your conscious reaction, not your subconscious one). Ah, I wish this was repeatable here :)

If you wanted the 3 to take real effect you should have numbered the choices. Now the brain has to make the translation "number 3" to "position three in the list" which takes time and interferes with what you intended to be a 'directed' subconscious choice.
E.g. like this

1. (O) Fred
2. (O) Sam
3. (O) Quayle

Bl*st! You did! Hmm that didn't do the trick did it...

interesting experiment.

i agree with Bob's comment further up though. Definitely had the same thought when I read the name Sammy Summers (though it wouldn't even be a very original pr0n star name)

I think it would be intersting if you had measured if the user took the test at all.
I imagine that it was more likely for the introduction text mentioning that there are only 3 choices (so the user knows what to expect and that it isn
t much work).

Sorry to pop your little bubble Bob but no actor that respects him self would call him self Fred!
Quayle Torton is the only one!
No but actually I even IMDB'ed the names and at the end I choosed Quayle Torton because it remineded me of Tarentino...

> If you wanted the 3 to take real effect you should have
> numbered the choices.

I did number them, for the reason you mention. As you can still see in the test, this is what shows:

1: ( ) Quayle Torton
2: ( ) Fred Ferentine
3: ( ) Sammy Summers

> I think it would be intersting if you had measured if the user
> took the test at all.

Good suggestion.

You made me in too much of a hurry reading it by saying I should answer them quickly, so I was expecting 3 questions rather than 3 choices :).

Question: What is the option that you think is going to be most voted?

A – Brad Pitt
B – Harrison Ford
C – Robert de Niro