To characterize the relationship between replication and external validity, they show that sign-congruent external validity and harmonization of studies are required to guarantee target-congruence between studies. You can read the paper here: www.cambridge.org/core/journal...

We also consider recently popular extrapolation approaches that impute from an estimated treatment effect in one setting to its analogue in another setting. These methods also invoke uniting principles which are typically very strong.

Part 2 explores applications.

We show that *meta-analysis* and *replication* both invoke distinct uniting principles. By making explicit the assumptions each approach invokes, we can see what is learned---or not---from each method.

We advocate for measurement perspectivism, which emphasizes that substantive phenomena are distinct from the tools we use to observe and measure them. Researchers make choices about what concepts to measure, how to observe those measures, and how to quantify those observations.

Takeaways from Part I:
When studying a mechanism empirically one must consider how its influence is measured

Assessing general phenomenon engages with some form of external validity

Accumulating evidence requires uniting principles characterizing relationships between studies

Uniting principles are the set of arguments that relate studies both qualitatively and quantitatively.

We explore two:

1. Common Concepts articulates the common mechanism across studies

2. Quantitative Connection gives the quantitative relationship uniting different findings

Evidence accumulation is about whether empirical findings, collected in one particular place and time, speak to a broader social phenomenon. We develop a new concept for evidence accumulation, uniting principles, which provide the foundation of any effort to accumulate evidence.

Part 1 develops key theoretical concepts for evidence accumulation. Part 2 applies those concepts to meta-analysis, replication, and extrapolation