The Cornell Lab of Ornithology has been an essential resource for birders for over a century, and Merlin is its terrifically nifty interactive app for learning about your local birds in real time in the field. You are probably already familiar with eBird, Cornell Ornithology’s enormously successful citizen science program that birders all over the world use to submit observations and to explore bird sightings in a highly localized way. They also run a whole range of other interactive programs, such as Project Feederwatch, Nestwatch, the Great Backyard Birdcount, camps and clubs for young birders, as well as informational resources like All About Birds, Macaulay Library and Birds of the World (not to mention birdcams, online courses, research studies, and partnerships with bird organizations worldwide). They are passionate about birds and nature, and create tools like these to help people learn and to become part of the larger birding community.
Merlin is a relatively new tool, first launched in 2014 in a limited form and which has now been developed into a very sophisticated but easy-to-use method of identifying birds in the field. Currently, the app (available FREE for both iPhone and Android) can identify more than 6,000 bird species across six continents visually and by sound. In 2017 it was upgraded, now using AI-powered image and sound recognition technology. The database is derived from their own participant-generated information: eBird’s 1.3 billion contributed bird observations from ~820,000 participants since it started in 2002, reporting on 10,511 of the world’s 10,585 bird species. 40 million images, 1.5 million audio recordings and ~200,000 video clips from checklists are stored in their Macaulay Library (which contains insect, fish, mammal and frog recordings as well). As for the quality of this data, eBird has a system for ensuring the accuracy of contributions, using regional reviewers and editors. The technology the app relies on can access this vast quantity of data and compare it to a Merlin user’s field experience in a fraction of a second to generate identifications.
Now, Merlin is not perfect, and you don’t want to assume the IDs are infallible. There’s a reason Merlin lists can’t be uploaded directly to eBird, and in fact the increasing trend of eBirders to report Merlin results is making the work of reviewers more difficult. But if you keep certain factors in mind, your results can be very very good. More on those later.
So, how does it work? After downloading the free Merlin app onto your phone, how can you use it?
The function most like a traditional field guide is the first on the Home Screen, “Start Bird ID”. It asks you a series of questions: location, date, approx size, main colors, behavior/habitat. Then it presents you with a list of possibilities (names, photos, thumbnail description). If you can’t get a photo or recording, this function might be helpful, although to be honest I don’t use it.
Second on the Home screen is “Photo ID”. This function can be very handy. It’s similar to iNaturalist but more efficient to run through the steps. It also provides key behavioral traits. For example, for this nonbreeding warbler I saw on Little Cayman in April, on successive screens, I (1) added the photo, (2) cropped it, (3) confirmed date and location, and Merlin came up with a few likely possibilities. In the series of screenshots below we can see that both the behavior (not flicking tail, not on ground) and beak detail point to Cape May warbler over Palm. Of note, when I asked the Dawn Chorus community about this bird in my Little Cayman diary, two commenters brought up behavior as an important confirming detail differentiating these two. I didn’t think to use Merlin for this until just recently!
However it’s the third function on the Merlin Home screen that I use every day and have found to be like magic: “Sound ID”. You simply start recording from wherever you are, and the app extracts identifications from the spectrogram generated by ambient sound. Invariably where I am there are multiple bird songs and calls so the spectrogram is a mix. It’s displayed on the top third of the phone screen so you can follow along if you want, but frankly I can’t possibly tease out individual bird sounds, even if I could hear the whole thing (I have high frequency hearing loss like most of us oldsters). But the Merlin technology can read the subtleties of the spectrogram mix, find matches in the Macaulay Library database, and display the birds on your phone screen.
Birds are listed in order they are ID’d and then highlighted in yellow when they sing/call again. Only 5 are displayed on the screen at a time while recording so you have to scroll up and down to see who’s being highlighted when you’re in the field. But this is a great way to learn songs: watch the screen for a highlighted bird as it repeats its song. I’ve noticed there’s a momentary delay between a song/call and the listing or highlighting….processing…. so keep that in mind as you correlate the song and highlighting.
The fourth function on the Home Screen, “Explore Birds”, is purely informational, but a terrific personalized reference resource where you can access seasonal bar charts and bring up species descriptions for birds most likely to be found in your locale. The screenshots below make it very clear why Merlin has become a real game changer for me. In winter, birds are more visible, either in leafless trees or more commonly, out on the water. Birdwatching is primarily a visual activity (although there is nothing cuter than listening to Buffleheads chattering amongst themselves).
But in summer, the nearshore sea is empty of ducks and my attention shifts to woods and thickets. That’s where the summer migrants live and nest during their time here before heading back south in fall. Our warblers, flycatchers, swallows, waxwings, vireos, not to mention House wrens, Rufous hummers, Red Crossbills, Swainsons thrushes, and a lot of sparrows all show up in spring as everything leafs out. The woods are full of songs and chirping. Invisibly so for the most part. Since I got a smart phone for the first time last winter and downloaded Merlin, all of a sudden my nearby environment has become much richer to me. And I’ve been able to start learning songs, more able to identify birdsong now in the field even without the app. It’s also helped in visual ID: I listen for the highlighted song, triangulate on location, and search for that particular bird there. I saw my first Orange-crowned warbler this year that way.
Now for the caveats. Merlin is magical, but magic is how we describe things we can’t explain, and in this case Arthur C. Clarke’s Third Law applies: “Any sufficiently advanced technology is indistinguishable from magic.” When we don’t understand how the sausage is made we can too easily make assumptions. Cornell Ornithology itself is pretty clear about the fact the app is not infallible:
Is Merlin always right?
No. Merlin shows a list of possible birds based on the songs and calls you recorded, and matches that with the birds that are likely in your area. Explore the matches provided by Merlin and compare the provided audio recordings to see if it matches your bird. Merlin is a great birding companion that suggests which species are more likely, but it relies on you to confirm what you’re hearing or seeing. Sound ID Help Center
How can we maximize the accuracy of our Merlin results, given the mysterious technological wizardry going on under the hood?
First, let’s look at the issues that complicate ID. The app can only analyze the data it can perceive. With incomplete incoming information, no matter how good the reference database, the AI can’t get good matches. GIGO. If you are too far from a bird or there’s a lot of ambient noise like traffic, lawn mowers, leaf blowers, people talking, even wind, you may see birds mistakenly show up on your sound ID because the app gets misled. Or the sound is too faint for it to lock onto (even if I can hear enough to ID, like distant ravens). Or a call may be IDd that is very similar to a bird that IS there. For example, a Blue-grey Gnatcatcher popped up one time on my Merlin app and those birds have never been seen in my county, very few ever found in Washington state in general and those long ago. Maybe the app heard a Towhee? Sometimes an oddball name will be listed and then disappear in a few seconds, which suggests there’s some internal checking in real time.
The nondescript quiet generic little cheeps have to give Merlin fits!
Beware the mimics. Birds like Mockingbirds, Blue Jays, Catbirds, and Starlings can sound like other birds. Then there are unintentional mimics, like the woman who had a big sneeze that was IDd as a Great Blue Heron hahaha. I’ve heard examples of frogs, car alarms, dogs barking and other sounds that show up as birds on the app.
There’s also the issue that matchups are based on user-contributed images and recordings. Any low-quality recordings in the database compromise the accuracy. I have no doubt improvements are being made on that front. This is a pretty new app.
Also, the database is accessed as a whole (from what I’ve read) and there are regional song differences. For example, the Golden-crowned sparrows that winter in my neighborhood sound a bit different from the collection of song recordings at All About Birds. Same for my Marsh Wrens, even the western ones. Optimally Merlin would use local recordings but I suspect there may not be enough in their database for that yet, especially in less reported areas.
So, how can you filter questionable results for accuracy?
—> See if a red dot accompanies the listing. That means it’s rare for your time and location, and makes it less likely you actually recorded it. On the other hand I’ve had red-dot birds get listed that I’m looking right at, like a Caspian Tern the other day, so use your judgement.
—> Watch for a listed ID to get highlighted. If the name pops up but is never highlighted, that’s a red flag, especially if songs are being repeated where you are at the time. Keep the recording going for a while to be sure. To check, tell the app to play back the section of the spectrogram where the questionable bird sang and then listen some more in real time. Save the recording so you can pass it along to an expert if you want to pursue it.
—> For photo IDs, get close and take the best possible photo you can. If you upload a low-quality photo, the app will have a hard time parsing the bird’s features for a matchup.
—> For sound ID, confirm the app’s results by actually looking for the bird. Make a personal confirmation. If a named birdsong is coming from a particular spot over and over, seek it out where it’s singing. You can bring up what the bird looks like on the app if necessary.
With all this in mind, it’s definitely worth downloading and using Merlin! It will make you a better birder and will enrich your experience of the natural world by revealing the many birds sharing your neighborhood with you.
Apple: Merlin Bird ID runs on iPhones and iPads with iOS 15 or newer, and M1/M2-equipped Apple computers. Download app here.
Android: Merlin Bird ID runs on devices with Android 6 or newer. Download app here.
You download a “Bird Pack” for your region of the US, and any other parts of the world where you might be traveling. They are listed in alphabetical order, from Angola to Japan to Zimbabwe and everywhere in between. Merlin will automatically use the Bird Pack database for wherever you are, based on your phone GPS (that worked for me when I was in the Caribbean).
Who are you seeing and hearing in your birdy world lately?