A sparrow on a railing is a useful observation. It tells us that someone identified a sparrow at a particular place and time. A checklist from ten minutes of watching that railing and the trees beside it can tell us something else: what the observer looked for, how long they looked, and which identifiable birds they did not report. That difference turns a record of a bird into a record of a birding trip.
The word complete can sound grander than it is. In eBird, it does not mean every bird present was found, or that the observer knew every call. It means birding was the main purpose and the observer reported every species they could identify by sight or sound, to the best of their ability. eBird's own definition allows a bird you could not identify to remain uncertain. It does not allow a familiar pigeon to vanish from the list because a more exciting bird took your attention.
The missing bird becomes a kind of information
Imagine twenty short watches of the same park corner. A swallow appears on two lists. If those are only two separate swallow sightings, we know it was seen twice, but not how many times anyone looked and saw no swallow. If all twenty watches were recorded as complete checklists, the eighteen lists without it become non-detections for those particular watches. The fraction is still not a count of all swallows in the park. It is a description of what those observers reported during those twenty searches.
This is why the checklist matters to analysis. In the Cornell Lab's eBird data guide, an observation record and a checklist record are treated as different pieces of data. The observation table contains the birds reported; the sampling-event table contains the searches. When researchers join them, a complete checklist with no record of a species can supply an inferred zero for that species on that outing. An incomplete list cannot do the same: the bird might have been missed, or it might have been noticed and deliberately left off.
The quietest line on a complete checklist can therefore be a useful one, even though it is never typed. It says, within the bounds of this search, I did not detect and identify that bird. It does not say, there were none here. Birds hide, fall silent, pass overhead between visits, or live in a part of the site that a route did not reach. The distinction between non-detection and true absence is central to occupancy research using eBird: estimating where a species occurs requires some account of the chance of finding it when it is there.
The search has a shape
Two complete lists need not represent comparable searches. Ten minutes at a bench and two hours walking three kilometres are different opportunities to find birds. So are one observer and four, dawn and midday, an open lawn and a wooded path. eBird asks for the date, location, start time, duration, distance and number of observers where the protocol calls for them. Its protocol guide distinguishes a stationary watch from a traveling count, and an incidental sighting from a trip made chiefly to bird.
Those details are not clerical decoration. They let a researcher limit a comparison to similar outings or include differences in effort in a model. In a 2021 analysis of eBird data, researchers tested ways of estimating bird distributions using more or less of that context. In their examples, complete checklists and variables describing effort and detectability produced some of the largest improvements in model performance. The size of the gain differed by species, data density and method; the study did not show that any one checklist is a flawless survey.
The person holding the binoculars matters as well. A bird that is obvious to a listener who knows its short call may pass unrecognized by someone learning the same park. In a study of eBird observers and 200 species, reporting rates increased with a measure of observer expertise for 197 species, and including that measure improved predictions. That is a reason to model differences among observers, not a reason for a beginner to stay home. A beginner's honestly complete ten-minute count still says what they could identify during that count.
Complete does not mean polished
There is no need to turn a walk into an exam. eBird says a complete list can include birds identified only to a broader group, such as a hawk that could not be named to species. It also says the observer need not have counted every individual to mark the list complete, although counts are preferred. Its reporting rules ask birders to include the birds they identify by ear as well as eye, and ordinary introduced birds that are free-flying. A list of highlights, however lovingly described, is incomplete if identifiable species were intentionally omitted.
An incidental observation is still worth recording. If a hawk crosses the street while you are carrying groceries, you did not conduct a search for all the birds around you. eBird's incidental protocol reflects that narrower event and automatically marks such a list incomplete. If you stop afterward and deliberately watch and listen for five minutes, that is a new, bounded birding effort; the start and end of the count should reflect those five minutes, rather than a retroactive claim about the whole errand. The protocol guidance makes this distinction because effort is part of what the checklist describes.
There are limits beyond the observer, too. People return to easy paths and famous hotspots; they do not visit every patch of habitat with equal frequency. A long traveling list may span several habitats while carrying only one mapped location. The eBird analysis guide recommends filtering by effort and considering spatial precision for the question at hand. In the 2021 modeling study, researchers also reduced the weight of places that had been sampled heavily. A 2023 eBird occupancy study found that sampled locations were not, on average, representative of nearby habitat; in its simulations, adding single-visit records could worsen bias when detection was low. Completeness gives the data a denominator, but it cannot by itself make birding locations random or every species equally detectable.
A useful small habit
For a routine walk, choose a stretch or a fixed spot you can describe honestly. Start the count when you begin looking and listening for birds; stop when you stop. Enter the protocol that matches what you actually did, and include every species you could identify, even if it was only a pigeon beside the sought-after warbler. Use a best count where you can, and keep uncertain identifications uncertain. If you later realize you left out birds you had recognized, add them if you can reconstruct the record accurately. Otherwise, mark the list incomplete rather than stretching the meaning of complete.
The reward is subtle. One sighting keeps the moment when the bird appeared. A complete checklist keeps the edges of the search around it. Those edges are what let the sighting join other visits without pretending that every silence means an empty sky.
A thought worth keeping?
