What is real scientific integrity?

According to the Oxford English Dictionary Integrity = "The quality of being honest and having strong moral principles".

When people talk about scientific integrity and misconduct they mostly have a narrow definition which means "don't make up data."

However, I think we need to consider a broader definition of integrity that relates to all communications and messages.

Scientists talk about their research in a wide range of forums:

• private discussions
• articles in luxury journals
• articles in professional society journals (PRA, JCP etc)
• grant applications and job applications
• seminars at universities and conference presentations
• press releases and interviews
• public lectures and popular books

Yet it seems it has now become quite acceptable to have different messages (claims and conclusions) in different forums. This post was stimulated by a perceptive comment by Steve W on a previous post.

My finding is if you talk to the authors of luxury papers with controversial or sexy explanations, that they will be the first to admit their own skepticism regarding their explanations of the data. But somehow this skepticism is not conferred to the text, because the luxury journals like clear, concise, authoritative explanations. Most of the details get hashed out later in less prominent, but longer form journals, and these are only followed closely by those within the specific community. 

For a concrete example see a recent post by Peter Woit about the basic question, "Is string theory experimentally testable?" He highlights a significant inconsistency between the answers in a preprint, the published version in PRL, a press release, and a public talk by Amanda Peet.