The visible record has already been filtered

Cross-Domain EchoesShared pattern · Selection Bias

A literature review may encounter the studies with the most striking results while many ordinary or null results remain unavailable. An evolutionary analysis restricted to old surviving lineages may encounter unusually favorable early growth because that growth helped those lineages survive. In both, the record was filtered before the analyst began. A pattern in the selected sample can therefore reflect the gate as well as the underlying process. Publication involves human reporting decisions; the evolutionary example follows survival conditioning in a stochastic diversification model. Neither comparison says the visible evidence is false. It asks what became visible, and why.

Written comparison

The population before selection

Scientific evidence

Conducted studies and measured outcomes

Evolutionary history

Possible diversification histories

The target of inference is broader than the visible or surviving record.

The inclusion gate

Scientific evidence

Results influence dissemination

Evolutionary history

Early history influences survival

Inclusion is associated with the very pattern being studied, so the gate is not a neutral sampling step.

The selected record

Scientific evidence

A literature enriched for striking findings

Evolutionary history

Surviving clades enriched for strong early growth

The observed collection can accurately describe its members while misrepresenting the broader process.

The misleading appearance

Scientific evidence

An effect seems stronger and more consistent

Evolutionary history

Later ordinary growth looks like a decline

The bias is a change in what the analyst sees, not necessarily a change in the underlying phenomenon.

What carries across

Explain how cases entered the record before interpreting their pattern. A selected history can make exceptional outcomes look like the ordinary process.

Where the comparison stops

Publication and survival are different selection mechanisms. Deliberate or institutional reporting decisions do not have a literal counterpart in lineage survival.

• The evolutionary comparison is conditional on a diversification model; an apparent slowdown can have other causes. The comparison does not dismiss every slowdown or every published effect as selection bias.

Conditions for this comparison

• Entry into the observed record is related to the feature being interpreted.
• The intended inference concerns more than just the selected set itself.

Source entries

Shared pattern

Selection Bias

Prime

Core Idea

Selection bias is the fundamental principle that a study's inference can be distorted when the process by which units enter, remain in, or contribute data is associated with both the exposure and the outcome — whether through self-selection into recruitment, differential retention, survivorship patterns, or the structural conditioning on common effects (colliders). This distortion can make observed exposure-outcome associations unrepresentative of the target population or arise entirely from the selection mechanism itself, independent of true causal relationships. The concept spans both experimental-design/statistics origins (Berkson's 1946 hospital-bias recognition, Neyman's sampling-selection discussion) and econometrics origins (Heckman's 1979 formal treatment and 2000 Nobel-winning sample-selection model), making it a dual-origin principle essential to causal inference, observational research, and meta-analysis

Scientific evidence

Publication Bias

Domain-specific abstraction

Core Idea

When the probability of crossing that gate depends on direction, magnitude, statistical significance, novelty, or sponsor favorability, the visible record no longer represents the research that was conducted

Examples

Twenty teams test a weak effect. Sampling noise gives two teams large positive estimates that cross a significance threshold; those studies are submitted and published. The other eighteen studies remain unavailable. A later reviewer sees two concordant positive reports rather than twenty noisy attempts and overestimates both effect size and evidential consistency.

Evolutionary history

Push of the past

Domain-specific abstraction

Core Idea

Push of the past is the statistical enrichment of surviving old clades for favorable early diversification histories.