Skip to the page
Sniff, Purr, Preen & Shoal

You are the habitat.

Sniff Test: How Studies, Numbers and Pictures Are Checked

Revised

A sniff test is a quick check before something is accepted. The checks here are slower and written down: how studies, figures, formulas and pictures are handled, and what each label on a page means.

Which source decides

  • Details of one study, such as its sample size, method or data, come from that study's own paper. If a secondary account disagrees, the paper wins.
  • A broader conclusion, such as whether cats form attachments to people, weighs study quality, how directly each study addressed the question, whether other teams repeated the result, and the direction of the evidence as a whole. A systematic review is consulted first where one exists; a newer study that disagrees is set beside it with the difference explained, not used to overturn it, and the most surprising result is not picked as the answer.
  • University, clinic and organization pages lead to research but are not cited for it. News stories count only as evidence that a claim circulates.
  • Names. Fish names follow Eschmeyer's Catalog of Fishes. An amphibian name follows a taxonomic database whose terms allow this use, or else the paper or original report the page cites, which the page names. Fish and amphibian names are checked against the Integrated Taxonomic Information System (ITIS). Conservation status is given as the IUCN Red List category name, together with the published source the page read it in, such as Reiß's 2022 history of axolotl research.
  • General encyclopedias that anyone can edit are never cited.

How much of each source was read

Each item in a page's source list carries one of these labels, and the page's wording follows it.

LabelWhat the page does with it
Full text, open licenseShort quotations; figures may go into the page's data table
Full text, read onlyFacts and figures restated in the site's own words with the DOI; no sentences, tables or figures reproduced
Abstract onlyThe page says the study was read in abstract and gives no number the abstract lacks; other Doorplate fields say not in the abstract read
Bibliographic record onlyThe study's existence and title, nothing about its content
Public-domain originalQuoted at length, with lecture or page number

Papers under a non-commercial license are not read or cited. A claim found only in secondary media is not used, nor is a DOI that does not resolve or whose author and year do not match the record.

Where the animals came from

Every research story marks where a study's animals came from, which the site calls the couch sample. A study's first figure on a page carries one of four markers: lab colony, shelter, pet homes or mixed. They are categories, not grades, and a pet-homes sample can still be one breed or whoever happened to sign up. The Doorplate's Limits field adds breed, age, recruitment, whether testing took place in a laboratory or a home, and the task. If a paper does not say where its animals came from, the page says no marker can be assigned.

Counts and percentages

  • A percentage is recalculated from the study's own counts, with the number of animals actually analyzed as the denominator; exclusions go in a separate column.
  • Two groups from one paper are never merged. In Bol and colleagues' 2017 catnip study, 32 of 99 cats did not respond to catnip, and a separate subset of 95 cats tested with all four plants is reported as its own group.
  • If a recalculated figure does not match the paper, the paper's figure stands and the mismatch is noted.
  • Each row of a data table stands on its own: claim, source and reading level share the row, and qualifiers, approximations and ranges have their own columns.

Formulas and the two calculators

Dog years. The calculator uses human age = 16 ln(dog age) + 31 from Wang and colleagues (2020), read as the authors' manuscript. Their 104 dogs were mostly Labrador retrievers, aged 0.1 to 16 years when sampled, with 9 removed from the analysis. The calculator accepts ages from about 7.5 weeks to 16 years: the oldest dog in the study was 16, and below about 0.144 years the formula returns zero or a negative age. The formula crosses the times-seven rule at about 0.15 and 9.6 years. The calculator states that it compares DNA changes and is not a health measure, and the paper notes that other breeds could follow a different curve.

Tank volume. NIST Handbook 44, Appendix C, defines the US gallon as exactly 231 cubic inches and the inch as exactly 2.54 centimeters, so one gallon is 3.785411784 liters. From a tank's outside dimensions the calculator removes the glass on four sides and the bottom, the gap between waterline and rim, and the substrate layer. Decorations and water held between grains of gravel are not counted, a bow front is approximated with a circular arc and labeled approximate, and examples built on assumed sizes say so. The result is water volume only, never a number of fish.

With scripts turned off, each calculator page still shows a diagram, a table of worked values, the formula and the data file, and the table rows come from the same formula the calculator uses.

Models and diagrams

  • Models are labeled model. The two-part model of a falling cat shows how a body can turn over with nothing to push against; its proportions are illustrative, and it is simpler than the mechanics in the research.
  • Schematics are labeled schematic, and drawings not to scale say so.
  • Measurements made in different ways are labeled not a like-for-like measurement when shown together, as with dog cone peaks from behavioral tests on three dogs beside human peaks measured in single cone cells.
  • Result charts are redrawn from a paper's numbers, never traced from its figure. Redrawing a protected diagram, anatomy plate or commercial simulation is still copying, so without a license it is described in words.
  • Every chart and step diagram comes with an HTML table or expandable text carrying the same information.

Counting words in an old book

The Pavlov's dogs page counts how often each sound stimulus is named in G. V. Anrep's 1927 English translation of Conditioned Reflexes. The rules were fixed before counting began:

  • Only the text of Lectures I to XXIII is counted; the index is counted separately and not added. Tables and captions could not be separated from the text and are included.
  • Singular, plural and possessive forms count, verb forms such as buzzing are listed apart, matching ignores capital letters, and a word split by a line break counts only when the two halves form a counted word.
  • Words in the scan's text layer within two letters of a counted term were compared with the page by eye; three were misreadings of a counted term and were added.
  • Two counting methods were compared and differed once, where a table split a term across two cells; the page image settled it.
  • On five pages, counts were checked against the page images and 13 of 13 matched. Those pages were chosen because they sit near the rare word bell, so the check leans toward rare terms.

The counts describe this translation only.

Photographs, paintings and prints

  • Each picture's license, version, creator and requested credit are checked on its own file page; a collection's reputation is not proof.
  • A public-domain claim names its basis, such as the creator's year of death and the jurisdiction, or an official statement. A date of publication alone is not treated as a basis that holds in every country.
  • Scans with extra conditions, such as a non-commercial restriction, are not used, and scans made by Google are not used at all. Where a holding institution's own record could not be consulted, the credit says the picture was checked at its file page.
  • A picture under CC BY-SA can illustrate the body of a page but is never used as a card picture.
  • Photographs showing children are not used, and at launch none shows an identifiable face.
  • A painting's subject has to match the page: a dead hound or game is never used for a page about sleep or scent.
  • A species or breed name appears in a caption or alt text only when the source labels it; no animal is identified by looking at a picture.

Rechecks and limits

Research stories, species names, IUCN categories and the calculators' sources are rechecked every 12 months, and the Doorplate's Replicated field changes when a result is repeated or overturned. Recalculating a figure confirms the arithmetic and the drawing; whether the input was right depends on the chain of sources, which is why every figure keeps its source and reading level. None of this turns a page into veterinary, training or care advice. Front door lists what the site does not do, and Mending basket records every substantive change.