BLAB: Do fruit flies dream of electric bananas?

[Colin Allen]

Do fruit flies dream of electric bananas?

New Scientist vol 181 issue 2434 - 14February2004, page 32


It may be no bigger than a poppy seed, but some neuroscientists think the
fruit fly's brain contains the rudiments of consciousness. Douglas Fox
peers through this unlikely window on the human mind


A FRUIT fly hovers in mid-air. Its bulbous eyes capture a panoramic view
of the world, but it ignores most of what it sees. Instead, it is
captivated by one small thing: a bright green stripe that just zipped by.
It's worth a closer look, worth landing on. The fly chases after it.

Or so it thinks. The fly is actually in a laboratory and suspended on a
wire inside a miniature flight simulator. Surrounding the fly is a
cylindrical LED screen, and travelling around that screen is a vertical
stripe of green light. Each time the stripe comes into view, the fly tries
to steer towards it. Sensors count the fly's wing beats, measure its
orientation, and control the motion of the stripe accordingly. If the fly
steers hard enough, it can halt the stripe and hold it steady in its view.

It's a neat gizmo, but watching it work I discover there's more to it than
that. Inserted in the fly's brain is an electrode, and buzzing through
that electrode, scrolling across a computer screen, are some curious
brainwaves. Whenever the fly steers towards the stripe, these saw-tooth
waves grow taller, and when the fix is lost, they shrink.

Ralph Greenspan, the researcher showing me this experiment, thinks these
little waves are a very big deal. "The fly thinks that something important
is happening to it," he says. "That stripe is what the fly considers to be
a significant stimulus from its environment." In other words, the fly is
paying attention.

That might not sound too impressive. Even a fruit fly needs to be able to
focus on important stimuli - how else would it find food or avoid danger?
But there's more to it than that. The brainwaves that Greenspan has found
look uncannily like the ones you see in a human brain when it is paying
attention.

This is a tantalising discovery. To neuroscientists, attention is a
profoundly interesting and important phenomenon. We are constantly
bombarded by information - smells, sounds, sights - yet we attend to only
the slenderest sliver of the whole; the rest we tune out, just as you tune
out the rumble of passing traffic as you read. Exactly how the brain
achieves this feat is one of neuroscience's biggest questions, and for
good reason: attention is intimately associated with consciousness. What
you pay attention to defines how you experience the world from moment to
moment.

Many neuroscientists believe that if they can discover how the brain
decides what to pay attention to, they will have taken the first step
towards teasing out the neuronal basis of consciousness. And that's why
finding human-like attention in a fly is so promising. Flies are much
easier to work with than people - could studying their brains open a new
window on the human mind?

I was eager to see how one studies a brain the size of a poppy seed, and
what, if anything, the results say about the human mind. So I decided to
visit Greenspan and his colleague Bruno van Swinderen at the Neurosciences
Institute in San Diego, California.

It's little wonder that insects are generally considered to be little more
than hard-wired, dim-witted automatons. A fruit fly brain contains just
250,000 neurons, compared with a human's 100 billion. Even the honeybee,
considered the brainiest of insects, has only a million neurons. And on
the surface, a fruit fly's brain looks nowhere near as complex as a
mammal's (see Diagram).

Yet recent behavioural studies have caught insects doing seemingly
un-insect-like things that hint at more complex brain functions. Sleep,
for example, was thought to be the sole preserve of vertebrates, but in
1999 Greenspan's team found that fruit flies sleep every night. They also
learn the way that Pavlov's dogs did: immerse a fly in peach odour and
shock its foot, and it will avoid peaches. And just as humans possess
separate short, medium and long-term memory, so do fruit flies: mutations
have been discovered that selectively block each type, proving they exist
in the first place. Flies also react to general anaesthetics in the same
way as humans, losing equivalent brain functions at equivalent doses.

Honeybees can learn the abstract concepts "same" and "different" and apply
them to novel situations. Researchers demonstrated this when training bees
to navigate mazes, by teaching them to always choose passages marked with
the same colour or smell they encountered at the maze entrance, regardless
of the particular colour or odour used. They also did the reverse,
training bees to choose passages marked differently from those at the maze
opening. "We are not saying that insects are Einstein," says Martin
Giurfa, a neuro-ethologist at the University of Toulouse (III), France,
who was involved in the honeybee study. "In contexts relevant for their
natural life they could be Einstein, but in other contexts they are
absolutely stupid."

These behavioural studies have gone a long way towards dispelling the
notion of insects as robotic dullards, but the brain circuits underlying
these behaviours were considered off-limits because insect brains are so
small. Not any more. Now that Greenspan and van Swinderen have found a way
to prise open the fly's brain and eavesdrop on its internal dialogue,
there are all sorts of possibilities.

It began during lunch one day as Greenspan and van Swinderen chatted with
colleague Douglas Nitz, a mouse neuroscientist. For a lark, they decided
to try recording an EEG from a fly's brain. The first run was as
garden-shed science as it gets: Nitz, who is blessed with bomb-squad
hands, held the electrode steady in the fly's brain with his fingers.
Unbelievably, it worked, and soon they had recorded the first EEG of fruit
fly sleep: its brain quietened down and produced fewer spikes, like a
mammal brainstem during non-REM sleep. After sleep, it was only natural
for the team to look for attention in the fly. After all, they are
opposite ends of the same spectrum.

In a closet-sized lab, I wheel about on a stool, careful not to bump van
Swinderen, who is bent over a microscope, wiring up today's fly. To me it
looks crude - the electrode seems almost the size of the fly's head - but
it's actually pretty exact. Adjusting the knobs of a micromanipulator, van
Swinderen places the electrode exactly where he wants it - between the
brain's mushroom bodies. If you are going to see the signature of
attention anywhere in a fly's brain, this is where you'd expect to find
it. The mushroom bodies receive multiple sensory inputs, including smell
and vision, and also recall memories. And sensory information plus memory
are the building blocks of attention: comparing a stimulus with past
experience to determine whether it's worth paying attention to, on account
of it being good, bad or new.

Van Swinderen hits a key and the show begins. A jagged Alps of brainwaves
dances across the computer screen, and each time the stripe whizzes past
the fly's field of vision, those waves change slightly. After a few runs
van Swinderen feeds the waves through a computer analysis and the picture
becomes clear. He is capturing a bundle of superimposed frequencies, but
one set - between 20 and 30 hertz - grow stronger when the stripe passes.
Greenspan and van Swinderen reckons these are an unmistakable signal of
what they call "salience" - the fly equivalent of attention. They prefer
"salience" to "attention" because they don't want to suggest, even
obliquely, that flies are conscious.

There are several reasons for believing they are right. In the flight
simulator, they see the 20 to 30 hertz signal whenever the fly is steering
towards a stripe. The first time a stripe comes into view, the signal
increases, but with each new pass the increase shrinks slightly as the fly
gradually loses interest - until a new stripe appears. What's more, the
signal increases when you enhance the importance of that stripe for the
fly - by simultaneously puffing the fly with banana odour, which it likes,
or heating it with a lamp, which it doesn't. And if you show the fly the
stripe alone, having previously shown it at the same time as heating the
fly, the signal is still elevated: the fly associates that stripe with
something bad, so watches it even more intently. It all suggests that the
20 to 30 hertz signal encompasses not just vision, but other senses too -
that it reflects some overall assessment of whether something is worth
keeping an eye on or can be safely ignored. What van Swinderen finds most
telling is that when the fly watches the stripe, it ignores everything
else. "Suppression is the hallmark of attention," he says.

There is another telltale sign. The electrode records from three different
regions of the brain. Normally, the electrical chatterings of these
regions are as different as the languages spoken in the Tower of Babel.
But show the fly a stripe and they suddenly fall into sync at 20 to 30
hertz, rising and falling in unison like a crowd doing a Mexican wave.
This is called synchrony and it is exactly what you see in a mouse or
human brain when it pays attention to something. Synchrony is attention
defined: all eyes focused on one stimulus, one stripe, one wave - and
everything else is ignored.

Synchrony is also interesting because neurologically it resembles what
consciousness probably is. Unlike, say, olfaction or face recognition,
which occur in specific brain regions, no one expects to find a localised
"consciousness centre". Instead, neuroscientists expect consciousness to
be about how brain regions are interconnected: the brain's various
regions, even distant ones, will be intimately wired together, equipped
for the continuous 10,000-way conference call that is consciousness.

So for many reasons attention and synchrony in a fly are very interesting,
but what can you do with them? Greenspan has some ideas. "The salience
response in flies looks like a very simplified version of the
attentiveness response seen in humans," he says. And that's exciting,
because it suggests that maybe this would be a model system for dissecting
human attention.

With this model system, researchers could finally combine the ability to
record complex brain activity - traditionally only done in mammals - with
the sophisticated genetic manipulation that is possible only in fruit
flies. They could screen hundreds of mutations to find genes involved in
attention, and then knock out the same genes in cells in the fly brain -
even in single neurons. In this way, they could eventually map the entire
brain circuits that control attention in flies. That would be a step
towards doing the same in mammals. And once you have identified mammalian
attention circuits, you would - perhaps - learn something about human
consciousness.

For now, though, it's what's going on in fly brains that is getting all
the attention. When the researchers presented their results at the
International Congress of Genetics in Melbourne, Australia, last July,
newspapers trumpeted the earth-shattering news: "Fruit flies are
conscious". Those stories were overblown, yet there was something to them.

"With consciousness there's the whole human baggage which is not worth
talking about in flies," says van Swinderen. "Philosophers waste a lot of
time thinking about such things. But a phenomenon like attention can be
completely understood - how an animal assigns salience to objects based on
experience, and how its internal representations of the world match the
external world."

To him, it is significant that if you are showing a fly a rotating stripe,
and then add a second stripe, the salience signal suddenly switches to the
new stripe. The fly's attention shifts to something new. This suggests
that the fly has something akin to a stream of consciousness. I put this
to van Swinderen and he sets me straight: not a stream of consciousness, a
stream of attention.

What's the difference? "Attention builds consciousness," he explains. "If
you couple attention to a load of memory, then maybe you can be conscious
after a lot of learning through a long life." A fruit fly, whose entire
being is shoehorned into 250,000 neurons and 30 days of life, never gets
there and never could. But it dabbles in the process, and van Swinderen
hopes to study it.

He envisages building a virtual maze for the flies to explore. First he
would present the fly with two visual stimuli to see which evokes a larger
salience response. Then, depending which it chooses, he would present it
with two more stimuli, and so on. The fly could navigate this maze for
hours.

"That's what I'm really excited about: having the fly tell me what's
important to it," he says. "If you let it choose the images, you can learn
a lot about attention and learning."

Van Swinderen thinks he could use this experimental design to identify how
the fly decides if something is salient. Show the fly two visual cues and
see which it pays more attention to. Then show it two variations of the
one it has chosen. Repeat this over many rounds and you end up optimising
the salience of the cue. Using this system you would discover to what
degree those salience criteria are hard-wired in all flies, as opposed to
being shaped by each fly's experiences.

Or if the fly learned to avoid something unpleasant, van Swinderen could
determine how far back the association went. For example, if a square
leads to a triangle leads to a circle leads to heat, could flies learn to
avoid the square, or would they recognise the threat only after seeing the
circle? Such experiments would be useful because they look at the fly's
use of memory to judge salience, and this could be a first step towards
asking whether they merely access their memories, or can somehow pay
attention to them.

The salience signal could advance dozens of other research areas, too.
Fruit flies are great for identifying genes involved in brain functions
such as learning and memory. But researchers must often rely on vague
behavioural measurements, such as courtship, to assess what those genes
do. Such assessments are notoriously ambiguous. A mutation that garbles
maze navigation, for example, might do so in dozens of different ways - by
disrupting memory, eyesight or even flight muscles. Maze navigation is the
sum of all these things so a behavioural change might not reveal the
underlying cause. But the salience signal allows researchers to look
directly at the brain, without having to guess.

"It's a tremendous tool," says Howard Nash, who studies fruit fly
anaesthesia at the National Institutes of Health in Bethesda, Maryland.
"It will allow us to look at things which a year ago I would have said
we'd never learn from a fly."

For me it explains a lot, too, not least why flies are so adept at
disappearing as soon as you arrive with a fly swatter. So next time you
pit your wits against an insect and lose, maybe you shouldn't feel so bad.
Its mind may be elementary and diminutive, yet it is not so different from
your own.




Douglas Fox
Douglas Fox is a freelance science writer living in northern California