I had planned to write a post about songbirds at some point in the future—how they’re used as a model for studying motor coordination, speech learning, and all kinds of cool stuff. But with recent news of the untimely passing of Allison Doupe, a longtime neuroscience faculty member at UCSF (where I got my PhD) and a pioneer in the songbird field, I think now is the right time to draw attention to the field and to her work in particular.
The first time I ever heard of songbirds as a neuroscience model, actually, was when Allison gave a guest lecture for one of my undergraduate classes back in 2005. At the time it was one of the coolest neuroscience talks I’d ever heard. I was not only impressed by her work, but also by her obvious scientific passion and her generosity in coming all the way down to Stanford to talk to a bunch of clueless undergrads. She even stayed after the talk to cheerfully answer all my questions, as painfully naive as I’m sure they were. It’s moments like this that are crucial for inspiring young scientists.
Avian love songs
So what’s so cool about songbirds? Let’s start with the basics. Male birds sing a song in order to woo females. Much like humans, female birds swoon over hot male singers who know their stuff. But male birds don’t write their own material. In fact, all males of the same species sing basically the same song, composed of syllables which have a characteristic pitch and duration.1 That means that not just the capacity for singing, but also the actual song itself is genetically hardwired into their DNA! Think about how crazy that is: it would be like if all of us humans were born knowing how to play Beethoven’s fifth symphony.
However, male birds start out as pretty crappy singers. Just like us, they need to practice to get better. Even if the basic template for the song is innate, producing a typical adult song requires learning.
Song learning occurs in multiple phases. First, the young bird hears a more experienced bird sing and forms a memory of what the song is supposed to sound like. Then the bird begins singing himself, starting initially with noises that bear only a passing resemblance to the real song. With practice, he eventually learns to sing a proper song capable of winning over the ladies.
Now you’re probably starting to get why songbirds are so interesting. Producing song requires generating a complex, coordinated motor sequence on very precise timescales; it’s not well understood how the brain does this kind of thing. Moreover, learning how to produce these precise sequences poses an even greater challenge for the brain.
Birdsong learning bears many similarities to speech learning in humans: both processes require first listening to others vocalize, then producing your own crude sounds (“babbling”), and eventually refining those sounds by trial and error. Both birdsong and speech learning require auditory feedback (you need to hear yourself) and must be learned during a specific time period early in life. So the hope is that studying birdsong will provide insight into how the human brain learns to generate complex motor patterns such as speech.
How the brain sings
There are numerous brain regions and neural mechanisms involved in birdsong, so I’m not going to review the entire literature here.2 Instead, I’m going to tell you about one important brain region whose role Allison Doupe’s work helped bring to light over the last couple decades (along with her husband Michael Brainard, also at UCSF). This region is called LMAN (you don’t want to know what it stands for),3 and I think it’s the most interesting part of the song circuit.
LMAN is part of a brain circuit called the basal ganglia. The basal ganglia circuit is known for its role in motor control and habit formation, and for being impaired in motor diseases such as Parkinson’s. In birds, LMAN represents the output of the basal ganglia, so it receives information from the rest of this circuit. LMAN then transmits signals to influence the song motor pathway, which directly controls the production and timing of song syllables.
When you inactivate LMAN in an adult songbird, you see (drumroll please): nothing. The bird can still sing perfectly well. However, young birds lacking LMAN never learn to sing normally.4
The reason these young birds can’t learn the song is because they stop tweaking it during practice.5 Their songs “crystallize”, or stop changing, so they’re stuck with whatever they’ve got. It turns out that LMAN is essentially a variability generator: it sends signals to the motor pathway that trigger more variation in the song.6
The importance of being variable
These results are really fascinating because we usually think of variability as a bad thing. Consistency is supposed to be good; variability is bad. Like if you’re preparing for a piano competition, you want to make sure you can hit all the right notes at all the right times, every time. That’s consistency. Variability is what makes you accidentally hit the wrong note in front of the judges, who frown disapprovingly into their scoresheets.
But variability is really important in certain contexts. And learning is one of them: when you’re learning, you need to explore the space of possible options to determine which option leads to the best outcome. For birds learning song or babies learning to speak, this means varying your vocalizations and listening to the results, gradually figuring out which motor commands produce the right sounds. LMAN generates the variability that enables this critical step in song learning.
That’s why researchers hadn’t noticed any effect when they inactivated LMAN in adult birds: because adults have already learned a nice consistent song that they don’t need to vary. Or so people thought. Actually, even adult birds are continuously trying to optimize their song. They may belt out a consistently flawless melody when they’re trying to impress a female, but when they’re alone they sing more variable renditions. It’s like when you’re playing Smash Brothers by yourself and you test out all kinds of crazy button combos hoping to find a new trick, but when your friends come over to play you stick to what you know works.
By modulating song variability, LMAN controls the switch between practice and performance. LMAN neurons are more active during practice singing sessions than during female-directed singing and show different activity patterns in each context.7,8 Moreover, inactivating LMAN in adults does have a subtle effect after all: these adults are unable to vary their song when they’re practicing, behaving as if there’s always a female around.6,9 So LMAN’s job is to promote song variability in both young and adult birds, enabling them to not only learn the song but also retain the flexibility to modify and optimize it throughout their lifetime.10
A scientific legacy
One of the legacies of Allison Doupe’s work is the discovery of a variability generator inside our brains, along with the idea that variability is crucial for learning and adaptation. Traditionally, neuroscientists never paid much attention to the variability they saw in the brain. We assume that neural variability is just meaningless noise, an accidental byproduct of the real signals we care about. We take for granted that the brain strives to be as precise as possible, and any residual variability is just due to its intrinsic limitations.
But thanks to work like Allison’s, we now know that some of that variability is intentional, generated by specific brain regions such as LMAN. It’s not just accidental noise, but rather a strategy that the brain uses to accomplish a particular goal. Promoting variability may in fact represent a core function of the basal ganglia that underlies its important role in motor learning.
Allison Doupe was one of the early pioneers in advancing the birdsong model from a niche field to an important model system in neuroscience. With the contributions from her lab as well as many others who are continuing this work, the ultimate goal of truly understanding the birdsong circuit—with its many implications for motor learning in general—seems remarkably within reach.
1. Or more accurately, males of the same species each sing their own cover version of the same song: the song can vary by region as well as from bird to bird.
2. If you’d like to learn more, this review provides an excellent summary of the field:
Brainard MS, Doupe AJ. Translating birdsong: songbirds as a model for basic and applied medical research. Annu Rev Neurosci 36:489-517 (2013).
3. LMAN stands for lateral magnocellular nucleus of the anterior nidopallium. See, I told you that you didn’t want to know.
4. Bottjer SW, Miesner EA, Arnold AP. Forebrain lesions disrupt development but not maintenance of song in passerine birds. Science 224:901-3 (1984).
5. Scharff C, Nottebohm F. A comparative study of the behavioral deficits following lesions of various parts of the zebra finch song system: implications for vocal learning. J Neurosci 11:2896-913 (1991).
6. Kao MH, Doupe AJ, Brainard MS. Contributions of an avian basal ganglia-forebrain circuit to real-time modulation of song. Nature 433:638-43 (2005).
7. Hessler NA, Doupe AJ. Social context modulates singing-related neural activity in the songbird forebrain. Nat Neurosci 2:209-11 (1999).
8. Kao MH, Wright BD, Doupe AJ. Neurons in a forebrain nucleus required for vocal plasticity rapidly switch between precise firing and variable bursting depending on social context. J Neurosci 28:13232-47 (2008).
9. Kao MH, Brainard MS. Lesions of an avian basal ganglia circuit prevent context-dependent changes to song variability. J Neurophysiol 96:1441-55 (2006).
10. I should mention that further studies have demonstrated a more complicated role for LMAN in motor learning. In addition to promoting variability, LMAN receives feedback about which motor commands lead to the best outcomes, and it can use that feedback to bias the variation that it generates in specific ways. For example, you can train an adult bird to either increase or decrease the pitch of his song, and LMAN plays an instructive role in controlling the direction of the change. See the following papers for more details:
Andalman AS, Fee MS. A basal ganglia-forebrain circuit in the songbird biases motor output to avoid vocal errors. Proc Natl Acad Sci U S A 106:12518-23 (2009).
Warren TL, Tumer EC, Charlesworth JD, Brainard MS. Mechanisms and time course of vocal learning and consolidation in the adult songbird. J Neurophysiol 106:1806-21 (2011).