Publication Bias

Publication bias is the tendency for journals to publish positive results ("It works!") while rejecting negative results ("It didn't work.").

This creates a distorted view of reality. It is like watching a highlight reel of a football game and thinking every play is a touchdown.

The Symptom: The 100% Success Rate

In 2008, researchers looked at all the studies ever done on 12 antidepressants.

The Published View: Among the studies published in journals, 94% were positive. The drugs looked like miracles.
The Reality: When they dug into the FDA archives to find the unpublished studies, only 51% were positive.

The drugs were barely better than placebo. But doctors only saw the 94% success rate, so they prescribed them to millions.

The Mechanism: The Filter

Science has a filter that blocks "boring" news.

The Researcher (Self-Censorship): "This experiment failed. If I publish it, people will think I'm incompetent. I'll just put it in the file drawer and start a new one."
The Journal (Editorial Bias): "This study found that eating kale doesn't prevent cancer. That's not a headline. Rejected."
The Sponsor (Commercial Bias): "This study shows our drug doesn't work. If we publish it, our stock will drop. Let's just hide it."

The File Drawer Problem

Imagine a file drawer full of studies that say "We tried X and nothing happened." If you don't see the drawer, you think X works every time.

The Funnel Plot

Meta-analysts use a graph called a Funnel Plot to detect this.

It plots the size of the study against the result.
It should look like a symmetrical upside-down funnel.
If the bottom-left corner (small, negative studies) is missing, it means publication bias is eating the data.

The Prescription: Pre-Registration

The solution is Pre-registration.

Before starting a clinical trial, researchers must register it in a public database (like ClinicalTrials.gov). This creates a permanent record that the study exists.

How to check for Publication Bias:

Find the Registration Number in the paper (e.g., NCT01234567).
Go to ClinicalTrials.gov.
Check the Status.
- Did the study end 5 years ago?
- Are there no results posted?
- Red Flag. They are likely hiding a negative result.

At PaperScores, we automatically check if a study was pre-registered. If not, we lower its Transparency Score. Hidden data is not science; it is marketing.