A Blackburnian warbler perches in a conifer tree. The species is featured in the author's discussion of Merlin's occasional bird-song identification mistakes. Credit: Bob Duchesne

Outdoors
The BDN outdoors section brings readers into the woods, waters and wild places of Maine. It features stories on hunting, fishing, wildlife, conservation and recreation, told by people who live these experiences. This section emphasizes hands-on knowledge, field reports, issues, trends and the traditions that define life outside in Maine. Read more Outdoors stories here. 

For 22 years, I’ve had the honor of kicking off the Downeast Spring Birding Festival with a talk on how to identify birds by sound. During this year’s Q&A, somebody asked: “How many people in the audience use Merlin?” Almost every hand shot up.

Merlin is the bird identification app from the Cornell Lab of Ornithology. Download it onto a smartphone, and you’ve got a free bird expert in your pocket. It can identify many birds by sight and sound. It can also misidentify them.

To be fair, Merlin doesn’t actually identify birds. It suggests potential identifications. Merlin provides best guesses by creating a visual picture of each bird song — a spectrogram — and then comparing those spectrograms to stored pictures in its vast database of bird recordings. Essentially, the spectrogram is somewhat like an audio fingerprint. It works amazingly well.

I played with Merlin this week, trying to figure out why some of its guesses miss the mark. Since the weather was ideal for recording bird songs, I ventured outdoors at daybreak for five straight mornings.

When a bird sings its typical song, Merlin’s suggestion is almost always accurate. Most songs are unique and distinct. But birds make other noises besides their typical songs — call notes, alarm notes and begging calls. These can be similar between species, leading Merlin to misguess.

Some species’ songs are very similar. Red-eyed and Philadelphia vireos sound nearly identical. That’s intentional, as the two species overlap in range and habitat, competing for the best nesting spots. My favorite spot to hear this conflict is Roaring Brook Campground in Baxter State Park.

The red-eyed vireo is slightly larger and can bully its cousin. The Philadelphia vireo compensates by migrating earlier and imitating the red-eyed vireo song, trying to convince its cousins to stay away — because this territory is already taken. If the vireos can’t tell each other apart, it’s a small wonder that Merlin struggles.

Some birds mimic other birds in their musical repertoires. Gray catbirds and brown thrashers steal other birds’ songs, but their vocal quality is different enough that Merlin usually guesses correctly. However, northern mockingbirds are such good imitators, their impressions are often better than the original bird. If you want to torture Merlin, point it at a talented mockingbird.

A red-eyed vireo’s song closely resembles that of the Philadelphia vireo, making identification challenging. Credit: Bob Duchesne

So there you have it. Merlin is frequently right and infrequently wrong. It’s a good tool for identifying bird songs, as long as you treat its results as suggestions, not gospel. It can tell you all the birds that are in your neighborhood, though occasionally incorrectly. Unfortunately, it can also make it harder to master the songs yourself.

First, if you don’t practice learning bird songs, you won’t learn them. No surprise. Since acquiring a wristwatch with a calculator, I’ve forgotten how to do long division.

Second, Merlin is blind and has only one ear. It can’t see the bird to verify its identity. It can’t tell you where the bird is, so you can verify it yourself. It’s a tease.

Third, Merlin creates a list of every song it detects during a recording session. However, every species is listed only once, no matter how long the session lasts or how many times a bird sings. It gives you information, but you have no idea how to use the information.

A Philadelphia vireo perches among spring leaves. Its song is often confused with that of the red-eyed vireo. Credit: Bob Duchesne

Fourth, the longer Merlin listens, the more misjudgments it might make, and the less reliable that list becomes.

Fifth, once a bird goes on the list, Merlin can’t remove it. Wednesday morning, as I walked under a singing Blackburnian warbler, Merlin suggested it was a bay-breasted warbler. The songs of both species are relatively weak. Moments later, Merlin recognized the Blackburnian song and added it, yet the erroneous bay-breasted warbler remained on the list. If I hadn’t known better, I might have been misled into thinking I had both warblers in my neighborhood.

Sixth, all the birds get thrown onto the list as fast as Merlin hears them. During the dawn chorus on a loud morning, Merlin’s bird suggestions pile up quickly. Tuesday morning, on my walk to get the newspaper, I had 24 species before I even reached the box. It’s like drinking through a fire hose — too much information too fast to be helpful.

Conclusion: there’s no substitute for learning bird songs yourself. Nonetheless, I highly recommend Merlin for inexperienced birders, if it’s used as a tool and you understand its limitations. Merlin opens up a whole new world of birds in your own backyard — birds you didn’t know were there.

Bob Duchesne serves as vice president of Maine Audubon’s Penobscot Valley Chapter. He developed the Maine Birding Trail, with information at mainebirdingtrail.com. He can be reached at duchesne@midmaine.com.

Leave a comment

Your email address will not be published. Required fields are marked *