Bias

What bias actually is

In everyday speech, bias means an unfair preference. In research, it means something more specific. Bias is a systematic error in how a study is designed, run, or analyzed that pushes the results away from the truth in a consistent direction. It’s not random noise. It’s a tilt baked into the process.

That tilt is what makes bias dangerous. Random error scatters results in all directions and tends to average out across a large enough study. Bias doesn’t average out. It nudges every measurement the same way, so a biased study can look clean and confident while still being wrong.

How it works

Bias can sneak in at almost any stage. If the people in a treatment group differ from the comparison group at the start, that’s a problem. If patients who know they’re getting the real treatment report feeling better partly because they expect to, that’s a problem. If studies with exciting results get published while disappointing ones quietly disappear, that’s a problem too.

Good study design is mostly an effort to block these paths. Randomly assigning people to groups, keeping patients and researchers unaware of who got what, and registering studies in advance all exist to keep bias from creeping in. The methods can look fussy, but each one closes a specific door.

What it isn’t

Bias isn’t the same as a researcher being dishonest. Most bias is unintentional and happens despite everyone’s good intentions, which is exactly why careful methods matter more than good intentions do.

It also isn’t the same as random chance. A p-value addresses chance, not bias. A study can be very unlikely to be a fluke and still be badly biased. No amount of statistical significance fixes a flaw in how the data was collected.

Related terms you’ll see next

A randomized controlled trial uses random assignment to reduce bias. A double-blind design keeps both patients and researchers unaware of who got the treatment, which blocks another source of bias. A placebo gives a fair comparison so expectation doesn’t masquerade as effect. A systematic review tries hard to find and account for bias across many studies.

Why it matters when you read about mental health

When you read about a study, the loudest number is usually the result. The quieter, more important question is how the study was run. Bias is the reason a striking finding can still be unreliable, and it’s the reason the strongest evidence comes from designs built to keep it out. When a claim rests on a single small or poorly controlled study, bias is often the explanation for why it doesn’t hold up later.

<blockquote>
  <p><strong>Bias:</strong> In research, bias is a systematic error that pushes a study's results in a particular direction. It distorts findings in ways that random chance doesn't.</p>
  <p>&mdash; <a href="https://shrinktionary.com/terms/bias/">Shrinktionary</a>, a plain-English mental health dictionary, medically reviewed by a board certified psychiatrist.</p>
</blockquote>