The project is based primarily on data that are "vouchered" by preserved specimens stored in museum collections (see Version 3.0, where we add non-vouchered data to the database). These data are the highest quality occurrence data available since they can be verified at any time by inspection of specimens that will persist indefinitely. Other data sources without vouchers are unverifiable. The ability to verify determinations is critical and our recent work towards correcting errors of determination in the Fishes of Texas Database illustrates the importance of museum specimens. Additionally, data backed by specimens are useful since those specimens can then be used for research in many fields of biological study.
...
The only way to verify anyone's determination of a species' identification is by examination of a specimen, and if one does not exist, questions will always remain. We thus chose to focus primarily on museum specimen-vouchered data as a means to reconstruct the historic historical record. Our work on this project demonstrates that fish identification errors are common. At the time of this writing, over half of our flagged (as geographic outliers) records had erroneous identifications, but we have found large error rates among species within their known ranges as well and believe all determinations should be seriously questioned.
In many cases specimens are identified incorrectly at the time of deposition in a museum, but in some instances errors are 'created' by the progress of science since species are often divided by taxonomists into two or more new species. The original species names remain in museum databases and labels, but now with incorrect determinations. Museum curated specimens allow researchers to determine the historic historical distribution of newly-split species, provided that determinations can be based on preserved morphology.
With the introduction of our Track 3 dataset (Version 3.0 of the Fishes of Texas Project) we introduced non-specimen vouchered data including, those from citizen science applications, agency databases, researcher databases, and others. These data are often based on an observation, but sometimes photos, tissues, or audio recordings are available. These data compliment the specimen record, but often cannot be verified.