Why We Sleep

participant, you have to remember not only the individual items you have been
shown, but also their spatial location on the screen. You will be shown a hundred
of these items. After sleep, picture objects will again appear on the screen, now in
the center, some of which you have seen before, some you have not. You have to
decide if you remember the picture object or not, and if you do, you must move
that  picture  object  to  the  spatial  location  on  the  screen  where  it  originally
appeared, using a mouse. In this way, we can assess whether you remember the
object, and also how accurately you can remember its location.
But  here  is  the  intriguing  twist.  As  you  were  originally  learning  the  images
before sleep, each time an object was presented on the screen, a corresponding
sound was played. For example, you would hear “meow” when the cat picture was
shown, or “ding-a-ling” when the bell was shown. All picture objects are paired, or
“auditory-tagged,” with a semantically matching sound. When you are asleep, and
in  NREM  sleep  specifically,  an  experimenter  will  replay  half  of  the  previously
tagged  sounds  (fifty  of  the  total  hundred)  to  your  sleeping  brain  at  low  volume
using speakers on either side of the bed. As if helping guide the brain in a targeted
search-and-retrieve  effort,  we  can  trigger  the  selective  reactivation  of
corresponding  individual  memories,  prioritizing  them  for  sleep-strengthening,
relative to those that were not reactivated during NREM sleep.
When you are tested the following morning, you will have a quite remarkable
bias in your recollection, remembering far more of the items that we reactivated
during sleep using the sound cues than those not reactivated. Note that all one
hundred  of  the  original  memory  items  passed  through  sleep.  However,  using
sound  cuing,  we  avoid  indiscriminate  enhancement  of  all  that  you  learned.
Analogous  to  looping  your  favorite  songs  in  a  repeating  playlist  at  night,  we
cherry-pick  specific  slices  of  your  autobiographical  past,  and  preferentially
strengthen them by using the individualized sound cues during sleep.
VIII
I’m sure you can imagine innumerable uses for such a method. That said, you
may  also  feel  ethically  uncomfortable  about  the  prospect,  considering  that  you
would have the power to write and rewrite your own remembered life narrative
or, more concerning, that of someone else. This moral dilemma is somewhat far
in the future, but should such methods continue to be refined, it is one we may
face.

---

SLEEP TO FORGET?
Up to this point, we have discussed the power of sleep after learning to enhance
remembering and avoid forgetting. However, the capacity to forget can, in certain
contexts,  be  as  important  as  the  need  for  remembering,  both  in  day-to-day  life
(e.g., forgetting  last week’s  parking  spot in  preference  for today’s)  and  clinically
(e.g.,  in  excising  painful,  disabling  memories,  or  in  extinguishing  craving  in
addiction  disorders).  Moreover,  forgetting  is  not  just  beneficial  to  delete  stored
information  we  no  longer  need.  It  also  lowers  the  brain  resources  required  for
retrieving  those  memories  we  want  to  retain,  similar  to  the  ease  of  finding
important documents on a neatly organized, clutter-free desk. In this way, sleep
helps you retain everything you need and nothing that you don’t, improving the
ease of memory recollection. Said another way, forgetting is the price we pay for
remembering.
In 1983, the Nobel Laureate Francis Crick, who discovered the helical structure
of  DNA,  decided  to  turn  his  theoretical  mind  toward  the  topic  of  sleep.  He
suggested that the function of REM-sleep dreaming was to remove unwanted or
overlapping  copies  of  information  in  the  brain:  what  he  termed  “parasitic
memories.”  It  was  a  fascinating  idea,  but  it  remained  just  that—an  idea—for
almost thirty years, receiving no  formal  examination.  In  2009,  a  young  graduate
student and I put the hypothesis to the test. The results brought more than a few
surprises.
We  designed  an  experiment  that  again  used  daytime  naps.  At  midday,  our
research  subjects  studied  a  long  list  of  words  presented  one  at  a  time  on  a
computer screen. After each word had been presented on the screen, however, a
large green “R” or a large red “F” was displayed, indicating to the participant that
they  should  remember  the  prior  word  (R)  or  forget  the  prior  word  (F).  It  is  not
dissimilar  to  being  in  a  class  and,  after  having  been  told  a  fact,  the  teacher
impresses upon you that it is especially important to remember that information
for the exam, or instead that they made an error and the fact was incorrect, or the
fact  will  not  be  tested  on  the  exam,  so  you  don’t  need  to  worry  about
remembering  it  for  the  test.  We  were  effectively  doing  the  same  thing  for  each
word right after learning, tagging it with the label “to be remembered” or “to be
forgotten.”
Half  of  the  participants  were  then  allowed  a  ninety-minute  afternoon  nap,
while the other half remained awake. At six p.m. we tested everyone’s memory for
all  of  the  words.  We  told  participants  that  regardless  of  the  tag  previously
associated with a word—to be remembered or to be forgotten—they should try to

---

recall as many words as possible. Our question was this: Does sleep improve the
retention of all words equally, or does sleep obey the waking command only to
remember  some  items  while  forgetting  others,  based  on  the  tags  we  had
connected to each?
The  results  were  clear.  Sleep  powerfully,  yet  very  selectively,  boosted  the
retention  of  those  words  previously  tagged  for  “remembering,”  yet  actively
avoided the strengthening of those memories tagged for “forgetting.” Participants
who  did  not  sleep  showed  no  such  impressive  parsing  and  differential  saving  of
the memories.
IX
We had learned a subtle, but important, lesson: sleep was far more intelligent
than we had once imagined. Counter to earlier assumptions in the twentieth and
twenty-first  centuries,  sleep  does  not  offer  a  general,  nonspecific  (and  hence
verbose)  preservation  of  all  the  information  you  learn  during  the  day.  Instead,
sleep  is  able  to  offer  a  far  more  discerning  hand  in  memory  improvement:  one
that preferentially picks and chooses what information is, and is not, ultimately
strengthened.  Sleep  accomplishes  this  by  using  meaningful  tags  that  have  been
hung onto those memories during initial learning, or potentially identified during
sleep  itself.  Numerous  studies  have  shown  a  similarly  intelligent  form  of  sleep-
dependent memory selection across both daytime naps and a full night of sleep.
When  we  analyzed  the  sleep  records  of  those  individuals  who  napped,  we
gained  another  insight.  Contrary  to  Francis  Crick’s  prediction,  it  was  not  REM
sleep  that  was  sifting  through  the  list  of  prior  words,  separating  out  those  that
should be retained and those that should be removed. Rather, it was NREM sleep,
and especially the very quickest of the sleep spindles that helped bend apart the
curves of remembering and forgetting. The more of those spindles a participant
had during a nap, the greater the efficiency with which they strengthened items
tagged for remembering and actively eliminated those designated for forgetting.
Exactly  how  sleep  spindles  accomplish  this  clever  memory  trick  remains
unclear.  What  we  have  at  least  discovered  is  a  rather  telling  pattern  of  looping
activity in the brain that coincides with these speedy sleep spindles. The activity
circles  between  the  memory  storage  site  (the  hippocampus)  and  those  regions
that program the decision of intentionality (in the frontal lobe), such as “This is
important”  or  “This  is  irrelevant.”  The  recursive  cycle  of  activity  between  these
two  areas  (memory  and  intentionality),  which  happens  ten  to  fifteen  times  per
second  during  the  spindles,  may  help  explain  NREM  sleep’s  discerning  memory
influence.  Much  like  selecting  intentional  filters  on  an  Internet  search  or  a
shopping app, spindles offer a refining benefit to memory by allowing the storage

---

site of your hippocampus to check in with the intentional filters carried in your
astute frontal lobes, allowing selection only of that which you need to save, while
discarding that which you do not.
We are now exploring ways of harnessing this remarkably intelligent service of
selective remembering and forgetting with painful or problematic memories. The
idea may invoke the premise of the Oscar-winning movie Eternal Sunshine of the
Spotless  Mind,  in  which  individuals  can  have  unwanted  memories  deleted  by  a
special  brain-scanning  machine.  In  contrast,  my  real-world  hope  is  to  develop
accurate methods for selectively weakening or erasing certain memories from an
individual’s  memory  library  when  there  is  a  confirmed  clinical  need,  such  as  in
trauma, drug addiction, or substance abuse.
SLEEP FOR OTHER TYPES OF MEMORY
All of the studies I have described so far deal with one type of memory—that for
facts, which we associate with textbooks or remembering someone’s name. There
are,  however,  many  other  types  of  memory  within  the  brain,  including  skill
memory. Take riding a bike, for example. As a child, your parents did not give you
a textbook called How to Ride a Bike, ask you to study it, and then expect you to
immediately begin riding your bike with skilled aplomb. Nobody can tell you how
to ride a bike. Well, they can try, but it will do them—and more importantly you—
no  good.  You  can  only  learn  how  to  ride  a  bike  by  doing  rather  than  reading.
Which is to say by practicing. The same is true for all motor skills, whether you
are learning a musical instrument, an athletic sport, a surgical procedure, or how
to fly a plane.
The term “muscle memory” is a misnomer. Muscles themselves have no such
memory:  a  muscle  that  is  not  connected  to  a  brain  cannot  perform  any  skilled
actions, nor does a muscle store skilled routines. Muscle memory is, in fact, brain
memory.  Training  and  strengthening  muscles  can  help  you  better  execute  a
skilled  memory  routine.  But  the  routine  itself—the  memory  program—resides
firmly and exclusively within the brain.
Years  before  I  explored  the  effects  of  sleep  on  fact-based,  textbook-like
learning, I examined motor skill memory. Two experiences shaped my decision to
perform these studies. The first was given to me as a young student at the Queen’s
Medical  Center—a  large  teaching  hospital  in  Nottingham,  England.  Here,  I
performed research on the topic of movement disorders, specifically spinal-cord
injury. I was trying to discover ways of reconnecting spinal cords that had been
severed,  with  the  ultimate  goal  of  reuniting  the  brain  with  the  body.  Sadly,  my

---

research was a failure. But during that time, I learned about patients with varied
forms  of  motor  disorders,  including  stroke.  What  struck  me  about  so  many  of
these  patients  was  an  iterative,  step-by-step  recovery  of  their  motor  function
after  the  stroke,  be  it  legs,  arms,  fingers,  or  speech.  Rarely  was  the  recovery
complete, but day by day, month by month, they all showed some improvement.
The second transformative experience happened some years later while I was
obtaining my PhD. It was 2000, and the scientific community had proclaimed that
the next ten years would be “The Decade of the Brain,” forecasting (accurately, as
it turned out) what would be remarkable progress within the neurosciences. I had
been asked to give a public lecture on the topic of sleep at a celebratory event. At
the  time,  we  still  knew  relatively  little  about  the  effects  of  sleep  on  memory,
though I made brief mention of the embryonic findings that were available.
After  my  lecture,  a  distinguished-looking  gentleman  with  a  kindly  affect,
dressed  in  a  tweed  suit  jacket  with  a  subtle  yellow-green  hue  that  I  still  vividly
recall to this day, approached me. It was a brief conversation, but one of the most
scientifically important of my life. He thanked me for the presentation, and told
me that he was a pianist. He said he was intrigued by my description of sleep as an
active brain state, one in which we may review and even strengthen those things
we have previously learned. Then came a comment that would leave me reeling,
and trigger a major focus of my research for years to come. “As a pianist,” he said,
“I have an experience that seems far too frequent to be chance. I will be practicing
a  particular  piece,  even  late  into  the  evening,  and  I  cannot  seem  to  master  it.
Often, I make the same mistake at the same place in a particular movement. I go
to bed frustrated. But when I wake up the next morning and sit back down at the
piano, I can just play, perfectly.”
“I can just play.” The words reverberated in my mind as I tried to compose a
response. I told the gentleman that it was a fascinating idea, and it was certainly
possible that sleep assisted musicianship and led to error-free performance, but
that  I  knew  of  no  scientific  evidence  to  support  the  claim.  He  smiled,  seeming
unfazed by the absence of empirical affirmation, thanked me again for my lecture,
and  walked  toward  the  reception  hall.  I,  on  the  other  hand,  remained  in  the
auditorium,  realizing  that  this  gentleman  had  just  told  me  something  that
violated the most repeated and entrusted teaching edict: practice makes perfect.
Not so, it seemed. Perhaps it was practice, with sleep, that makes perfect?
After  three  years  of  subsequent  research,  I  published  a  paper  with  a  similar
title, and in the studies that followed gathered evidence that ultimately confirmed
all of the pianist’s wonderful intuitions about sleep. The findings also shed light

---

on  how  the  brain,  after  injury  or  damage  by  a  stroke,  gradually  regains  some
ability to guide skill movements day by day—or should I say, night by night.
By  that  time,  I  had  taken  a  position  at  Harvard  Medical  School,  and  with
Robert  Stickgold,  a  mentor  and  now  a  longtime  collaborator  and  friend,  we  set
about trying to determine if and how the brain continues to learn in the absence
of  any  further  practice.  Time  was  clearly  doing  something.  But  it  seemed  that
there were, in fact, three distinct possibilities to discriminate among. Was it (1)
time, (2) time awake, or (3) time asleep that incubated skilled memory perfection?
I took a large group of right-handed individuals and had them learn to type a
number sequence on a keyboard with their left hand, such as 4-1-3-2-4, as quickly
and as accurately as possible. Like learning a piano scale, subjects practiced the
motor  skill  sequence  over  and  over  again,  for  a  total  of  twelve  minutes,  taking
short  breaks  throughout.  Unsurprisingly,  the  participants  improved  in  their
performance across the training session; practice, after all, is supposed to make
perfect.  We  then  tested  the  participants  twelve  hours  later.  Half  of  the