AH, THE joys of a good night鈥檚 sleep. Eight hours of blissful, uninterrupted
shut-eye. Except it never happens.
None of us just hits the pillow and lies comatose till morning. Even a
normal, healthy sleeper will wake up between 15 and 35 times each night. That鈥檚
not usually a problem. Healthy sleepers hardly notice they鈥檝e woken up before
their bodies pull them back to sleep. But, for some people, trying to sleep is a
waking nightmare. As many as 10 per cent of the adult population suffer from
more extreme sleep-wake disturbances, which leave them constantly tired.
Compared with other branches of medicine, our understanding of sleep is
pretty basic, and treating sleep problems is a slow process. Night after night,
at sleep clinics like the one run by Thomas Penzel and J枚rg-Hermann Peter
at the University of Marburg, researchers record and analyse the sleep habits
and the vital signs鈥攕uch as blood pressure, heart rate, breathing rates
and so on鈥攐f people with sleep problems, hoping to see something that will
help them diagnose a particular ailment.
Advertisement
Occasionally, though, they have a happier task: logging the vital signs of
people who sleep soundly. Penzel and Peter used to work out the proportion of
the night these healthy sleepers would spend asleep and awake, and use that as a
control against which to measure the problems of their patients. But that simple
view came to an end a couple of years ago, when Penzel met Plamen Ivanov, a
Boston University physicist.
Ivanov works in a research group that has an almost unhealthy obsession with
details. Its members are experts in signal analysis, looking for hidden patterns
in just about every place they can think of: the heart, the Earth鈥檚 crust,
financial markets. You name it, and they鈥檒l find the pattern in its
behaviour.
Between them, Ivanov and Penzel began to think there might be undiscovered
patterns in the way we sleep at night, and that they might provide a new way of
diagnosing and treating sleeping disorders. They and their colleagues have
encapsulated these ideas in a paper published in Europhysics Letters
this month.
To Ivanov, sleep is made up of events whose frequency you can plot
mathematically, just as you can plot the severity of earthquakes. You can take
all the earthquakes on the planet for one year, for example, and plot the
frequency for each magnitude of earthquake against the magnitude itself. The
result is a curve that shows the exact relationship between the size and how
often it occurs. With earthquakes, the kind of curve you get is called a power
law: the frequency of the quakes falls off in inverse proportion to their
magnitude raised to some power. Movements on the stock market behave the same
way. But with other phenomena, such as the intervals between signals in human
neurons, the frequency decays exponentially instead.
It sounds like an obscure distinction, but it is actually profound.
Exponential curves have an inbuilt scale factor: the size jump over which the
frequency falls by half is the same no matter where you take your starting
point. For power laws this doesn鈥檛 apply.
This kind of analysis could be just the thing to sort out the problem of a
bad night鈥檚 sleep. Graduate student Chung Chuan Lo got the job of analysing
Penzel鈥檚 sound sleepers. He looked at the length and number of waking periods
through the night, and plotted their frequency against how long they lasted.
Then he did the same for periods of sleep.
Lo discovered that the waking periods followed a power-law relationship,
while the sleep times are exponential. To his colleagues, that was absolutely
amazing. In all their years of analysis, the Boston researchers had never before
come across a case in which both patterns arise in one system. Normally it鈥檚
always one or the other. Having discovered this intriguing behaviour, they felt
compelled to find out what was at the root of it, and what it might reveal about
how sleep works. So they began to build a model.
Scans of brainwave patterns indicate that there are certain distinct sleep
states: REM sleep, which usually coincides with dreaming, two other light
sleeping states, and two deep states. So they divided the slumber spectrum into
a few sleep 鈥渕icrostates鈥�. It also seems reasonable that there are many
different degrees of waking, from barely conscious to fully alert, so they put a
whole set of different waking microstates into the model.
The next stage was based on the latest biological thinking about sleep. A
collaboration between Swiss and French neurobiologists published a couple of
years ago found that sleep is essentially a battle in the brain between
sleep-promoting and wake-inducing neurons (Nature, vol 404, p 6781).
For their model, the Boston researchers assumed that the result of this complex
chemical battle is a kind of random walk: one step towards awake, three steps
back towards sleep, another couple of steps back towards wakefulness, and so on
in a random pattern.
And that nightly tug of war must be one-sided鈥攁 stronger force pulls
healthy sleepers back to the land of nod than pulls them awake. So it is common
to hit the deepest limit of sleep, but rare that the random walk will fully wake
you. 鈥淓ach time the brain enters the wake state, it is pulled back to sleep
quickly,鈥� Lo says.
The researchers incorporated all this in an equation and looked to see if it
modelled the way people drift in and out of sleep. As they had hoped, it
produced the same sleep patterns as the doctors saw in real
people鈥攊ncluding the all-important mixture of power law and exponential
distribution. To the Boston researchers, this is a sign that their model
reflects the processes behind sleep.
Night dreaming
But it鈥檚 not proof. Peretz Lavie, a sleep researcher at Technion in Israel,
says that the Boston model is not the only approach that could explain the
strange dynamics. He thinks that alternatives might concentrate on the
transition from and to REM sleep, which occurs roughly every 90 minutes and is
where most of our brief awakenings are clustered.
The Boston team admit that their model is crude and that it will be hard to
test the assumptions it鈥檚 based on, but they still believe it is going to be
extremely useful. 鈥淚t鈥檚 a toy model,鈥� says Ivanov. 鈥淏ut it is the first one to
give a rough idea and explanation for a very complex and intriguing phenomenon.鈥�
And it might have some strange implications.
If sleep really is a random walk, there can be no link between the length of
one sleeping period and the length of the waking period that follows. There
would also be no correlation between the length of one wake period and the next,
or one sleep period and the next. It could explain why you sometimes wake up for
no apparent reason, and why people鈥檚 sleep behaviour is so varied. Light
sleepers presumably have too weak a team of sleep neurons. Those who quickly
drift off again even after being shaken awake might be blessed with a beefier
bunch.
The Boston model is already providing a new set of tools for sleep
researchers. The team has looked at sleep apnoea, in which patients have
breathing problems that cause them to wake up frequently. Not surprisingly, they
saw more short waking periods than normal, but there were also fewer long waking
periods. This kind of shift may provide a new way to characterise sleep apnoea,
and may also help researchers find out more about how the body responds to such
problems. The pattern might indicate, for example, that the sleep neurons tug
harder to compensate for the frequent waking.
Back at the Marburg clinic, psychiatrist and neuroscientist Martin Huber is
interested in using the analysis to diagnose disorders such as depression that
affect moods and feelings. 鈥淎 disorder of sleep structure is very common in
affective disorders,鈥� he says. Huber hopes eventually to match different sleep
patterns with particular diseases. That might even allow doctors to diagnose
diseases before other symptoms show up.
Ivanov is collaborating with Huber to look at how drugs affect sleep
patterns. They believe some drugs may limit the number of available wake
microstates. It鈥檚 possible that you may hit your peak wakefulness during a
night鈥檚 sleep, shifting your wake state away from power-law behaviour. Looking
for such a shift might provide a way to monitor the side-effects of
medications.
All these efforts will, in turn, help refine the Boston researchers鈥� model of
sleep. Once a disease or a drug can be linked to a particular sleep pattern,
whatever is known about that disease or drug鈥檚 effects on our bodies should
suggest ways to alter the model and faithfully reproduce the altered sleep
patterns. As the model improves, it might be possible to see how other, subtle
physiological changes affect our sleep.
The researchers are confident they have found an important new path into the
relatively unexplored field of sleep. They have certainly made one thing clear:
sleep is more than a welcome switch-off at the end of each day. It is a game of
chance, a means of diagnosing illness, and a strange new phenomenon in physics.
Really, it鈥檚 enough to keep you awake at night.