IF YOU’VE EVER posted a photo on the massive community science project called iNaturalist and wondered how such contributions get used in research… well, today’s guest is here to tell us about one especially stunning example. It involves 1.6 million such crowd-sourced observations, and the timing of the migration of hummingbirds in Eastern North America.

You’ve probably heard it said that hummingbirds love red flowers, and scientists in the Hopkins Lab at the Arnold Arboretum of Harvard University took a much closer look at that connection thanks to all that community data, and the use of artificial intelligence to sort through it all.

One of the Harvard scientists, a post-doctoral researcher in the plant evolution lab named Patrick McKenzie, part of the team at Dr. Robin Hopkins’s lab, is here to explain what they learned and how.

Patrick has written that, “Quiet hours in the sun, meditating with the bugs, plants, and birds, are my inspiration as an evolutionary biologist.”

He is always on the lookout for patterns—and then asking himself why each pattern unfolds—like the why of red flowers and hummingbirds, for example. Besides his extensive training in plant evolution, Patrick is a keen birder, and I was glad for the chance to chat with him.

Read along as you listen to the June 1, 2026 edition of my public-radio show and podcast using the player below. You can subscribe to all future editions on Apple Podcasts (iTunes) or Spotify (and browse my archive of podcasts here).

hummingbirds and red flowers with patrick mckenzie

Download file | Play in new window | Duration: 00:27:00

Subscribe: Apple Podcasts | Spotify

Margaret Roach: Yes. So on a non-hummingbird topic—well, it might be hummingbird topic, but a non-red flower topic—I’ve read that your favorite wildflower species is Monarda fistulosa, the wild bergamot or bee balm, and that you spent many hours watching bumblebees forage from its lavender-colored flowers. So that’s one of your favorite observations that you like to go watch?

Patrick McKenzie: Yeah, definitely. Yeah. Monarda fistulosa is one that I grew up around in Arkansas and it’s kind of followed me my whole life. So now there’s a giant patch of it next to the building where I work. And so in the summertime, actually starting in just a few weeks, the Monarda fistulosa will be blooming here. And yeah, watching the pollinator interactions with them is really, really amazing because they’re visited by bumblebees very obviously, but if you stand around and watch long enough, you’ll see all sorts of different insect visitors; occasionally hummingbirds, too, come by to visit those flowers.

Margaret: Yeah. So how did the question about the commonly repeated notion that hummingbirds favor red and I think tubular and nectar-rich flowers, how did that sort of question get on your radar as something to look deeper into at the lab? And then that led to the publication, I believe it was last May in the journal “Current Biology,” of your research. So were you watching hummingbirds or red flowers or what happened? [Laughter.]

Patrick: Well, yeah, there are a few things. So the concept of what makes a “hummingbird flower” is something that scientists have talked about back and forth in the literature for a long time. And whether it’s truly a red flower preference for hummingbirds; whether that pattern that red flowers are more often hummingbird flowers, whether that’s true to begin with.

And then also there are these other physical characteristics that are common in hummingbird flowers, we think—like having a tubular shape, like trumpet-shaped flowers, and being rich in nectar—that scientists have been trying to study at large scales for a long time.

I think for us, there’s this observation that if you track the flowering time or the phenology, the seasonality, of when flowers are appearing in nature up the Eastern U.S., I think a simple observation that was motivating for us is that flowers are appearing in the Eastern U.S. before hummingbird migration has occurred. So there’s this couple-of-week-long period I think in late March or early April, right before hummingbirds have moved up to say Virginia, New York, where I am in Massachusetts, where flowers are blooming, but the birds aren’t here yet. And so I have this idea that if this is true, probably these hummingbird flowers are going to be underrepresented among the flowers that we see. And maybe we won’t be seeing many of these red or orange tubular, nectar-rich flowers.

Margaret: O.K. So that got you wondering, but it wasn’t like you could get nine million researchers [laughter] to go man the coast all the way from the overwintering grounds to the breeding grounds for a month or something or a few weeks or even a week. I mean, how do you possibly … So you had this idea of how you could possibly accomplish this.

Patrick: Right. Yeah. So there’s this history of these phenology studies being done, these studies of seasonality at local scales. So you can find within say a state park or something, a botanist would’ve recorded the flowering phenology, the flowering time, of these different plant species and published that in the literature. And so there was some of that highly localized information.

But in parallel to all of this, I’ve been getting more and more fascinated with iNaturalist, this community science or—I call it community science, other people call it citizen science or participatory science. It’s all referring to these platforms by which anyone can walk around in nature with a cellphone and not have a science background but still be contributing useful data. And iNaturalist is I think one of, if not the most popular platform by which people are often engaging with community science. And so I’m starting to get really excited just in all aspects of the science I’m doing with the opportunity to use this data to study natural history at these really giant scales.

So like you said, yeah, it would be really, really difficult to send thousands of researchers up and down the Eastern U.S. to track when things are flowering and where hummingbirds are. But because we have people walking around with their cellphones taking pictures of plants and birds and uploading them to iNaturalist, we can do it from our living room couch. Or in my case, in the building where I work, I can look at observations from all over the country and all over the world.

So yes, I think in North America, there are millions of observations of flowering plants and that provided a really amazing dataset for us to dive into for this project.

Margaret: And now you didn’t really sit on your living room couch going through the pieces of data. [Laughter.] You actually did something even more forward-thinking and modern, which is I think you used artificial intelligence to help do the analysis. Is that correct?

Patrick: Right, that is correct. And actually, this idea is something that I originally had probably… I think I started trying to work on this with some friends during my PhD five or six years ago. And at the time, artificial intelligence as an opportunity for data processing wasn’t really on our radar at all. And at the time I took a bunch of people and a giant spreadsheet and tried to subdivide the work of manually labeling flower colors across the spreadsheet among eight or 10 people, something like that. And I was like, if we all spend 10 hours clicking rows, maybe we can come out with a data set of like 10,000 label observations and maybe that would be enough to study this phenomenon.

And then fast forward to I think late 2023 is when I was able to revisit this project and ChatGPT had just matured in kind of a new way. They’d just introduced this computer-vision model, and computer vision is something that I think is famously hard for doing complex tasks but really excels at very simple tasks.

And so the kind of question that I had been asking up to this point was looking at iNaturalist observations and asking what flower color I’m looking at in a given image. And this is the exact kind of question that something like the ChatGPT vision model is really, really good at performing, answering very simple questions and then doing that many times. And so I was able to come back to this dataset and rather than trying to rope in a bunch of people to just assemble a small dataset, suddenly I could turn the computer loose on this dataset of what ended up being a dataset of 1.6 million observations.

And for every observation, we can ask this very simple question of there’s a flower in this image out of this list of colors, yellow, blue, green, red, what color is the flower that you’re seeing? And scaled up massively, that becomes a very rich source of information that humans have a very difficult time assembling.

Margaret: And so we know also from other observations, we know where the hummingbirds are, from the migration reports and so forth.

Patrick: Yeah.

Margaret: Because I guess maybe we should say that hummingbirds, they’re not just taking nectar, they also act as pollinators to some of these plants too. Is that correct? There’s a two-way relationship.

Patrick: That’s exactly right. Yeah. The plants provide a service to the hummingbirds and the hummingbirds provide a service to the plants. And it’s thought that this specialization of these physical traits of the flowers of being red and of being tubular and of being nectar rich are often more to the flowers’ benefit. I think specifically the specialization on red is thought to be the plants specializing to the hummingbirds.

There’s been some previous work showing that hummingbirds are really great learners, and if they are in an environment where there aren’t red and orange flowers that they can learn to find nectar in flowers of other colors, and that this red color that the flowers take help them stand out to the hummingbirds to the exclusion of things like bumblebees. So they can specialize really well on hummingbird visitors for pollination. And then the hummingbirds help plants cross-pollinate with one another, which helps maintain diversity for the plants.

But yes, getting back to your original question, hummingbirds are also very frequent in the iNaturalist dataset. They’re also tracked separately. There’s a separate community-science platform called eBird that has loads of information about hummingbird migration as well. For simplicity for this paper, we just used the iNaturalist dataset, but we did check informally to see that there’s really good concordance in both datasets for tracking this northward advancement of hummingbird migration.

Margaret: So here you have all this data, it’s getting analyzed by AI and it’s all … And what did you find out? Because again, I’ve heard for years, “Oh, they like red flowers,” and there may be some correlation to their migration patterns or whatever. People say things (but sometimes it sounds like a little bit of conventional wisdom or an old wive’s tale) like, “Oh, they come at the time of such and such flower.” But not necessarily backed up by research. So what did you find out? [Above, Monarda didyma, scarlet bee balm, with red tubular flowers.]

Patrick: Yeah, so what we found was something that to me is really delightfully simple, which is that if you just forget about the species identity of all of these different flowers, if you just take these 1.7 million observations and you break them down into colors—red, blue, green, yellow—we see this sort of perfect looking concordance in flowering time for everything that’s not red and orange flowers.

So basically early in the year, all of these different colors start flowering in the Southern United States in the Eastern half of the country. And then throughout the spring, this flowering advances northwards. So we have early in the year, everything’s blooming in the South. Later in the spring, by May and by June, flowers are peaking in the Northeast. And then red and orange really stand out as lagging behind in this northward advancement. And so in March, in April when the hummingbirds haven’t yet arrived, there are basically no red and orange flowers blooming in the Eastern U.S.

And then as soon as the hummingbirds show up, we see this really perfect correlation in the northward advancement of the Eastern U.S. of hummingbird migration alongside where these red- and orange-colored flowers are appearing.

And so it’s the first time that this lag driven by flower color alone has been documented in the scientific literature. And as an evolutionary biologist, this is really exciting because it suggests something about a generality. Generalities can be hard to come by in evolution and in ecology, because biology is just so messy. But it suggests there’s this generality of a link between red and orange flowers regardless of the evolutionary histories of these different species, and the time of the season at which they’re appearing. And that’s we think driven by the appearance of hummingbirds seasonally.

Margaret: So because again, relationship? Because over all these however many millennia, etc., and beyond— relationship, is that why? I mean, the two organisms—they’re not two, but you know what I mean, the flowers of these colors and the hummingbirds. Do we know that or—

Patrick: Yeah, I think there’s still more testing to be done. So what we have right now, it’s a good correlation between the appearance of the red and orange flowers and the appearance of the hummingbirds, but I think it’s really highly suggestive of a co-evolutionary relationship between these things. So there is some dependence, right, on hummingbird presence of success for these red and orange flowers and getting successfully pollinated.

And yeah, I think for millions of years, they will have been building this relationship, and it’s contingent on the fact that hummingbirds are migratory. So this aspect of flowering time is contingent on this completely separate taxon. Which if you look across on the other side of the country at the West Coast like California, Anna’s hummingbirds are there year round, and we don’t see this same pattern. And so it’s this process of migration in the ruby-throated hummingbirds in the Eastern U.S. that I think is driving this.

Margaret: I kind of love that it’s ruby-throated and that it’s the red and orange flower [laughter]. I just kind of love that. It’s all just very styled. It’s very fashionable. So some of the flowers, just as examples that in this palette [laughter] going up the country, are there some that people would recognize that as “hummingbird plants” kind of?

Patrick: Yeah, definitely. Yeah, some popular garden plants for native flower gardens: I think Lobelia cardinals, cardinal flower [photo, top of page], is one of the most frequent in the dataset, as well as Aquilegia canadensis, wild columbine. If you’re more in the Southeast, Campsis radicans, the trumpet creeper, is a more common orange flower. Later in the season we see jewelweed appear as well.

The dataset’s really interesting because yeah, it’s dominated very much by some of these very, very, I think, common and showy flowers. I think something that sort of innately appeals to me about these questions of hummingbird pollination is that I think the flowers that are very attractive to hummingbirds are also the flowers that are very attractive to me personally. So I think of these as being the showy, charismatic flowers that are often associated with hummingbird pollination. [Above, wild columbine; photo from Wikimedia by Ragesoss.]

Margaret: Right. And it is interesting, of course, now thinking about it, I’m thinking, well, you’re completely right. I never thought of it consciously. I mean, I always thought about the part of early spring is pastel-y, but I never thought about the other part, dot, dot, dot, but the red and orange stuff comes later. I didn’t think of it exactly that way. I thought of what was present, not what was missing.

Patrick: Right, right. Well, for the early spring stuff, and actually just I think looking across all of the other ways that a flower can look, there are other types of pollination syndromes. So pollination syndrome is a word that’s referring to whether something is a hummingbird flower or a bumblebee flower or a fly flower, these groups of traits that are associated with one type of floral visitor versus another.

And so we’ve been talking about this red and orange, tubular, nectar-rich flower for hummingbirds,. But there are also associated sweets of traits with bumblebees, often having spotting on the petals [as above, with Monarda]; they often have like a landing pad for the bumblebees to hold onto. Just the sort of generalities of the appearance of these flowers. And in contrast, we have for fly-pollinated flowers, we often have really strong aromas. They’re often white or sometimes like maroon or brown.

And they’re also famous hawk moth pollination syndromes, bat pollination, all of which are generalities because these syndromes can arise sort of across the tree of life and might look slightly different in agave versus morning glory or something, but still you might see this trend toward, oh, if it’s hawk moth pollinated, it might be white.

I think what stands out here is that there is this lag, the seasonal lag, in the appearance of ruby-throated hummingbirds that we don’t see in bumblebees versus flies. I think maybe we would also see a similar difference in flowering time of our bumblebee-pollinated things and our fly-pollinated things, except I think both of those groups of pollinators get, they become active around the same time of year, and so we don’t see the same lag there.

Margaret: It made me want to ask about, does this experience with this research done this way using this kind of data, do you now have a list of all the other things you want to explore this way [laughter]? Uh-oh. I think laughing. I guess he does.

Patrick: It can really spiral, can’t it? I think yes, absolutely. I think iNaturalist and community science in general is something that if we were to awaken a botanist from 200 years ago and show that person iNaturalist and what it can provide as a resource for asking these kinds of natural history questions at large scale, I really think it’d be absolutely mind-blowing for that person.

It is a treasure trove that I think provides endless opportunities for questions about the natural world. And so yeah, I’ve started digging in more to specific organisms. So we brought up Monarda fistulosa earlier and we have a recent paper out about exploring flower color variation in Monarda fistulosa specifically. So rather than asking are we looking at red versus yellow versus blue, we’re asking, O.K., all of these flowers are generally lavender-colored, but is there variation in that color across space? And in that project, we find that all of the Western Monarda fistulosa seem to be this richer, deeper lavender than those in the East. And so I’ve explored that a bit more with more of an organismal target.

There are also new discoveries every day with iNaturalist of new species, or it can be used to track the spread of invasive plants, too. So I’m thinking about these things as well.

I think maybe in the short term, something I’ve been really excited about exploring in the future is to actually turn to eBird, this other community-science platform, and browse through all of the images of hummingbirds and try to assemble a list or a map or something of if there is a flower being visited by the hummingbird in that picture, what type of flower is that and can this community-science dataset also just generally support the idea that this physical association of this red tubular flower really does hold up. Again, I think this is something that we generally know to be true, but every different facet we can take to support that I think makes that generality more convincing.

Margaret: And selfishly, I want to say as a gardener, I always say to people, plant some “hummingbird flowers” near your favorite windows [laughter] because there’s such little birds and they’re fast and so forth, but it’s so incredible to watch them up close, to have the privilege to watch them up close. Do you know what I mean?

Patrick: It really is.

Margaret: I have some honeysuckle, Lonicera sempervirens, around a porch post right outside my kitchen windows and it’s just like, I mean, just watching is just selfish but wonderful. So hopefully they’re as happy as I am about it.

Patrick: Well, they’re so entertaining, right?

Margaret: Yeah, exactly.

Patrick: I’ve lived near Monarda didyma before, the scarlet bee balm [above].

Margaret: Yes.

Patrick: And yeah, really, really similar experience. It can provide a whole lot of entertainment, especially when you get more than one hummingbird. It can be a lot of fun.

Margaret: Well, Patrick McKenzie, I’m so glad to talk to you and I was so glad that you shared with me in correspondence and so forth recently this research and alerted me to this because this is just really fascinating, both the process and what you learned. So thank you. And I hope I’ll talk to you again soon, and happy birding meantime, right?

Patrick: Yeah. Thank you so much. It’s been a pleasure.

(Photos by Patrick McKenzie except as noted.)

prefer the podcast version of the show?

MY WEEKLY public-radio show, rated a “top-5 garden podcast” by “The Guardian” newspaper in the UK, began its 17th year in March 2026. It’s produced at Robin Hood Radio, the smallest NPR station in the nation. Listen locally in the Hudson Valley (NY)-Berkshires (MA)-Litchfield Hills (CT) Mondays at 8:30 AM Eastern, rerun at 8:30 Saturdays. Or play the June1, 2026 show using the player near the top of this transcript. You can subscribe to all future editions on iTunes/Apple Podcasts or Spotify (and browse my archive of podcasts here).