Right, my brain put a 'just' in front of bank teller in response A, then I started thinking about whether the facts in the paragraph would seem to have anything to do with whether she was a feminist, or had anything to do with banking. A re-parsing and I thought maybe it was about whether we associated feminists with protesters, since his previous paragraphs mentioned prejudices we have, etc.

I ended up not choosing an answer and reading on, as it seemed like a 'trick' type question, and indeed, I was thinking about what it had alluded to, not what the actual question was.

Various alternative explanations like that have been explored, but it seems the effect is not due to wording: http://lesswrong.com/lw/jj/conjunction_controversy_or_how_th...

Actually, I would take it to mean that the wording has a huge effect on how people respond, as per his previous post: http://lesswrong.com/lw/ji/conjunction_fallacy/

Notice that in the jazz playing accountant example, 92% of people answered A, E, C.

Notice now that in the gambling situation, only 65% of subjects answered 2.

92% for the "story" question, and 65% for the "hard data" question. I still look at the jazz question and read it through the cultural interpretation of A > E > C, despite it being incorrect in a strictly logical sense. That does not make my interpretation a conjunction fallacy, regardless of what the question creator intended.

This only goes to show that the conjunction fallacy has no hope of being measured accurately unless the question is completely unambiguous in subtext (meaning that it doesn't tread upon cultural aspects of language, where people fill in the blanks and read between the lines).

And even then, the dice rolling question, though devoid of cultural baggage, still might not be the best way to measure the conjunction fallacy. I know a great number of people who couldn't figure out probability to save themselves. This could have skewed the results higher or even lower, depending on what proportion of the test subjects could properly interpret probabilities. Now if this question were asked to statisticians, the results would be compelling.

That's not the reason. The question seems clear to me. Moreover, your assertion could only be true if people believed that more than half of female bank tellers are involved in the feminist movement, which is implausible.

The question most certainly did not seem clear to me. And the assertion does not require more than half of female bank tellers to be involved in the feminist movement.

Rather, bank teller becomes an irrelevant side-fact, and the focus of discussion becomes whether Linda is more likely to be part of the feminist movement given her being deeply concerned with issues of discrimination and social justice.

Human language is imprecise, and comes with assumptions based on context and culture. The people who came up with this question succumbed to the false assumption that the majority of people would interpret the wording of the question the same way they did.

Because of the flawed nature of the question an the incorrect assumptions of the test givers, this question cannot be relied upon to give any measure of the "conjunction fallacy".