1. Three passages, one observation
Suppose three passages say a device works after being submerged. One reports a demonstration; another quotes that report; a third paraphrases the quotation. Counting three passages as three confirmations inflates support: all three depend on one observation. Their repetition establishes that a claim circulated, not that the device passed three independent trials. These examples are illustrative, not reported test results. The assumption to reconsider is that more matching passages automatically mean stronger evidence.
2. Follow the researcher
The link between research and researcher offers a useful reminder: publications have origins. Ask who observed what, using which records or measurements. Different authors can still depend on the same dataset, press statement or witness. Conversely, similar wording can accompany independent observations. Overlap is therefore a reason to inspect dependence, not a verdict that evidence is worthless. A second analysis of shared data may expose an error without adding another observation.
3. Check the contribution
Choose one precise claim and compare the passages supporting it. For each, identify the underlying observation, its origin and any additional measurements or reasoning. Trace quotations as far as the supplied material allows. Group passages that rely on the same observation, then reassess the claim using those groups. If removing duplicate wording leaves the same observation and reasoning, do not count the repetition as extra empirical support. If origins remain unclear, mark independence as unresolved; neither assume corroboration nor accuse anyone of copying.
4. Hypothetical boundary: shared equipment
Imagine two teams independently submerge separate devices but use the same miscalibrated timer. Their reports describe distinct trials, yet a claim about survival duration inherits a shared measurement risk. Decision rule: count the trials as distinct observations, while withholding independent corroboration of timing accuracy until that shared risk is checked. Seek calibration records or measurements from another timer. This preserves useful evidence without allowing repeated conclusions to conceal a common weakness. Revise confidence only as far as the checks justify.