Why We Sleep

Figure 14: Example of a Memory Association Network
Using  a  standard  computer  test,  Stickgold  measured  how  these  associative
networks  of  information  operated  following  NREM-sleep  and  REM-sleep
awakenings, and during standard performance during the waking day. When you
wake the brain from NREM or measure performance during the day, the operating
principles  of  the  brain  are  closely  and  logically  connected,  just  as  pictured  in
figure  14.  However,  wake  the  brain  up  from  REM  sleep  and  the  operating
algorithm  was  completely  different.  Gone  is  the  hierarchy  of  logical  associative
connection.  The  REM-sleep  dreaming  brain  was  utterly  uninterested  in  bland,
commonsense  links—the  one-step-to-the-next  associations.  Instead,  the  REM-
sleep brain was shortcutting the obvious links and favoring very distantly related
concepts.  The  logic  guards  had  left  the  REM-sleep  dreaming  brain.  Wonderfully
eclectic  lunatics  were  now  running  the  associative  memory  asylum.  From  the

---

REM-sleep  dreaming  state,  almost  anything  goes—and  the  more  bizarre  the
better, the results suggested.
The two experiments of anagram solving and semantic priming revealed how
radically different the operating principles of the dreaming brain were, relative to
those  of  NREM  sleep  and  wakefulness.  As  we  enter  REM  sleep  and  dreaming
takes hold, an inspired form of memory mixology begins to occur. No longer are
we constrained to see the most typical and plainly obvious connections between
memory units. On the contrary, the brain becomes actively biased toward seeking
out the most distant, nonobvious links between sets of information.
This widening of our memory aperture is akin to peering through a telescope
from  the  opposing  end.  When  we  are  awake  we  are  looking  through  the  wrong
end of the telescope if transformational creativity is our goal. We take a myopic,
hyperfocused, and narrow view that cannot capture the full informational cosmos
on  offer  in  the  cerebrum.  When  awake,  we  see  only  a  narrow  set  of  all  possible
memory  interrelationships.  The  opposite  is  true,  however,  when  we  enter  the
dream  state  and  start  looking  through  the  other  (correct)  end  of  the  memory-
surveying telescope. Using that wide-angle dream lens, we can apprehend the full
constellation of stored information and their diverse combinatorial possibilities,
all in creative servitude.
MEMORY MELDING IN THE FURNACE OF DREAMS
Overlay  these  two  experimental  findings  onto  the  dream-inspired-problem-
solving  claims,  such  as  those  of  Dmitri  Mendeleev,  and  two  clear,  scientifically
testable hypotheses emerge.
First,  if  we  feed  a  waking  brain  with  the  individual  ingredients  of  a  problem,
novel connections and problem solutions should preferentially—if not exclusively
—emerge after time spent in the REM dreaming state, relative to an equivalent
amount of deliberative time spent awake. Second, the content of people’s dreams,
above  and  beyond  simply  having  REM  sleep,  should  determine  the  success  of
those hyper-associative problem-solving benefits. As with the effects of REM sleep
on  our  emotional  and  mental  well-being  explored  in  the  previous  chapter,  the
latter would prove that REM sleep is necessary but not sufficient. It is both the act
of dreaming and the associated content of those dreams that determine creative
success.
That  is  precisely  what  we  and  others  have  found  time  and  again.  As  an
example, let’s say that I teach you a simple relationship between two objects, A
and B, such that A should be chosen over object B (A>B). Then I teach you another

---

relationship,  which  is  that  object  B  should  be  chosen  over  object  C  (B>C).  Two
separate,  isolated  premises.  If  I  then  show  you  A  and  C  together,  and  ask  you
which you would choose, you would very likely pick A over C because your brain
made an inferential leap. You took two preexisting memories (A>B and B>C) and,
by flexibly interrelating them (A>B>C), came up with a completely novel answer
to a previously unasked question (A>C). This is the power of relational memory
processing, and it is one that receives an accelerated boost from REM sleep.
In  a  study  conducted  with  my  Harvard  colleague  Dr.  Jeffrey  Ellenbogen,  we
taught participants lots of these individual premises that were nested in a large
chain  of  interconnectedness.  Then  we  gave  them  tests  that  assessed  not  just
their  knowledge  of  these  individual  pairs,  but  also  assessed  whether  they  knew
how these items connected together in the associative chain. Only those who had
slept and obtained late-morning REM sleep, rich in dreaming, showed evidence of
linking  the  memory  elements  together  (A>B>C>D>E>F,  etc.),  making  them
capable  of  the  most  distant  associative  leaps  (e.g.,  B>E).  The  very  same  benefit
was found after daytime naps of sixty to ninety minutes that also included REM
sleep.
It  is  sleep  that  builds  connections  between  distantly  related  informational
elements  that  are  not  obvious  in  the  light  of  the  waking  day.  Our  participants
went  to  bed  with  disparate  pieces  of  the  jigsaw  and  woke  up  with  the  puzzle
complete.  It  is  the  difference  between  knowledge  (retention  of  individual  facts)
and wisdom (knowing what they all mean when you fit them together). Or, said
more  simply,  learning  versus  comprehension.  REM  sleep  allows  your  brain  to
move beyond the former and truly grasp the latter.
Some may consider this informational daisy-chaining to be trivial, but it is one
of the key operations differentiating your brain from your computer. Computers
can store thousands of individual files with precision. But standard computers do
not  intelligently  interlink  those  files  in  numerous  and  creative  combinations.
Instead, computer files sit like isolated islands. Our human memories are, on the
other  hand,  richly  interconnected  in  webs  of  associations  that  lead  to  flexible,
predictive  powers.  We  have  REM  sleep,  and  the  act  of  dreaming,  to  thank  for
much of that inventive hard work.
CODE CRACKING AND PROBLEM SOLVING
More  than  simply  melding  information  together  in  creative  ways,  REM-sleep
dreaming can take things a step further. REM sleep is capable of creating abstract
overarching  knowledge  and  super-ordinate  concepts  out  of  sets  of  information.

---

Think  of  an  experienced  physician  who  is  able  to  seemingly  intuit  a  diagnosis
from the many tens of varied, subtle symptoms she observes in a patient. While
this kind of abstractive skill can come after years of hard-earned experience, it is
also  the  very  same  accurate  gist  extraction  that  we  have  observed  REM  sleep
accomplishing within just one night.
A delightful example is observed in infants abstracting complex grammatical
rules in a language they must learn. Even eighteen-month-old babies have been
shown  to  deduce  high-level  grammatical  structure  from  novel  languages  they
hear,  but  only  after  they  have  slept  following  the  initial  exposure.  As  you  will
recall,  REM  sleep  is  especially  dominant  during  this  early-life  window,  and  it  is
that  REM  sleep  that  plays  a  critical  role  in  the  development  of  language,  we
believe. But that benefit extends beyond infancy—very similar results have been
reported  in  adults  who  are  required  to  learn  new  language  and  grammar
structures.
Perhaps  the  most  striking  proof  of  sleep-inspired  insight,  and  one  I  most
frequently  describe  when  giving  talks  to  start-up,  tech,  or  innovative  business
companies to help them prioritize employee sleep, comes from a study conducted
by Dr. Ullrich Wagner at the University of Lübeck, Germany. Trust me when I say
you’d  really  rather  not  be  a  participant  in  these  experiments.  Not  because  you
have to suffer extreme sleep deprivation for days, but because you have to work
through  hundreds  of  miserably  laborious  number-string  problems,  almost  like
having  to  do  long  division  for  an  hour  or  more.  Actually  “laborious”  is  far  too
generous a description. It’s possible some people have lost the will to live while
trying to sit and solve hundreds of these number problems! I know, I’ve taken the
test myself.
You will be told that you can work through these problems using specific rules
that are provided at the start of the experiment. Sneakily, what the researchers do
not tell you about is the existence of a hidden rule, or shortcut, common across all
the  problems.  If  you  figure  out  this  embedded  cheat,  you  can  solve  many  more
problems in a far shorter time. I’ll return to this shortcut in just a minute. After
having had participants perform hundreds of these problems, they were to return
twelve  hours  later  and  once  again  work  through  hundreds  more  of  these  mind-
numbing  problems.  However,  at  the  end  of  this  second  test  session,  the
researchers asked whether the subjects had cottoned on to the hidden rule. Some
of the participants spent that twelve-hour time delay awake across the day, while
for others, that time window included a full eight-hour night of sleep.

---

After  time  spent  awake  across  the  day,  despite  the  chance  to  consciously
deliberate on the problem as much as they desired, a rather paltry 20 percent of
participants  were  able  to  extract  the  embedded  shortcut.  Things  were  very
different  for  those  participants  who  had  obtained  a  full  night  of  sleep—one
dressed  with  late-morning,  REM-rich  slumber.  Almost  60  percent  returned  and
had  the  “ah-ha!”  moment  of  spotting  the  hidden  cheat—which  is  a  threefold
difference in creative solution insight afforded by sleep!
Little  wonder,  then,  that  you  have  never  been  told  to  “stay  awake  on  a
problem.” Instead, you are instructed to “sleep on it.” Interestingly, this phrase, or
something  close  to  it,  exists  in  most  languages  (from  the  French  dormir  sur  un
problem, to the Swahili kulala juu ya tatizo),  indicating  that  the  problem-solving
benefit of dream sleep is universal, common across the globe.
FUNCTION FOLLOWS FORM—DREAM CONTENT MATTERS
The author John Steinbeck wrote, “A problem difficult at night is resolved in the
morning after the committee of sleep has worked on it.” Should he have prefaced
“committee” with the word “dream”? It appears so. The content of one’s dreams,
more than simply dreaming per se, or even sleeping, determines problem-solving
success.  Though  such  a  claim  has  long  been  made,  it  took  the  advent  of  virtual
reality  for  us  to  prove  as  much—and  in  the  process,  shore  up  the  claims  of
Mendeleev, Loewi, and many other nocturnal troubleshooters.
Enter my collaborator Robert Stickgold, who designed a clever experiment in
which participants explored a computerized virtual reality maze. During an initial
learning session, he would start participants off from different random locations
within  the  virtual  maze  and  ask  them  to  navigate  their  way  out  through
exploratory trial and error. To aid their learning, Stickgold placed unique objects,
such  as  a  Christmas  tree,  to  act  as  orientation  or  anchor  points  at  specific
locations within the virtual maze.
Almost  a  hundred  research  participants  explored  the  maze  during  the  first
learning  session.  Thereafter,  half  of  them  took  a  ninety-minute  nap,  while  the
other  half  remained  awake  and  watched  a  video,  all  monitored  with  electrodes
placed  on  the  head  and  face.  Throughout  the  ninety-minute  epoch,  Stickgold
would occasionally wake the napping individuals and ask them about the content
of any dreams they were having, or for the group that remained awake, ask them
to report any particular thoughts that were going through their minds at the time.
Following  the  ninety-minute  period,  and  after  another  hour  or  so  to  overcome
sleep  inertia  in  those  who  napped,  everyone  was  dropped  back  into  the  virtual

---

maze  and  tested  once  more  to  see  if  their  performance  was  any  better  than
during initial learning.
It should come as no surprise by now that those participants who took a nap
showed superior memory performance on the maze task. They could locate the
navigation clues with ease, finding their way around and out of the maze faster
than those who had not slept. The novel result, however, was the difference that
dreaming made. Participants who slept and reported dreaming of elements of the
maze,  and  themes  around  experiences  clearly  related  to  it,  showed  almost  ten
times more improvement in their task performance upon awakening than those
who  slept  just  as  much,  and  also  dreamed,  but  did  not  dream  of  maze-related
experiences.
As  in  his  earlier  studies,  Stickgold  found  that  the  dreams  of  these  super-
navigators  were  not  a  precise  replay  of  the  initial  learning  experience  while
awake. For example, one participant’s dream report stated: “I was thinking about
the maze and kinda having people as checkpoints, I guess, and then that led me to
think about when I went on this trip a few years ago and we went to see these bat
caves,  and  they’re  kind  of  like,  maze-like.”  There  were  no  bats  in  Stickgold’s
virtual  maze,  nor  were  there  any  other  people  or  checkpoints.  Clearly,  the
dreaming  brain  was  not  simply  recapitulating  or  re-creating  exactly  what
happened to them in the maze. Rather, the dream algorithm was cherry-picking
salient fragments of the prior learning experience, and then attempting to place
those new experiences within the back catalog of preexisting knowledge.
Like  an  insightful  interviewer,  dreaming  takes  the  approach  of  interrogating
our  recent  autobiographical  experience  and  skillfully  positioning  it  within  the
context  of  past  experiences  and  accomplishments,  building  a  rich  tapestry  of
meaning. “How can I understand and connect that which I have recently learned
with  that  I  already  know,  and  in  doing  so,  discover  insightful  new  links  and
revelations?”  Moreover,  “What  have  I  done  in  the  past  that  might  be  useful  in
potentially solving this newly experienced problem in the future?” Different from
solidifying  memories,  which  we  now  realize  to  be  the  job  of  NREM  sleep,  REM
sleep,  and  the  act  of  dreaming,  takes  that  which  we  have  learned  in  one
experience setting and seeks to apply it to others stored in memory.
When  I  have  discussed  these  scientific  discoveries  in  public  lectures,  some
individuals  will  question  their  validity  on  the  grounds  of  historical  legends  who
were  acclaimed  short-sleepers,  yet  still  demonstrated  remarkable  creative
prowess. One common name that I frequently encounter in such rebuttals is the
inventor  Thomas  Edison.  We  will  never  truly  know  if  Edison  was  the  short-

---

sleeper that some, including himself, claim. What we do know, however, is that
Edison  was  a  habitual  daytime  napper.  He  understood  the  creative  brilliance  of
dreaming, and used it ruthlessly as a tool, describing it as “the genius gap.”
Edison would allegedly position a chair with armrests at the side of his study
desk, on top of which he would place a pad of paper and a pen. Then he would take
a  metal  saucepan  and  turn  it  upside  down,  carefully  positioning  it  on  the  floor
directly below the right-side armrest of the chair. If that were not strange enough,
he would pick up two or three steel ball bearings in his right hand. Finally, Edison
would  settle  himself  down  into  the  chair,  right  hand  supported  by  the  armrest,
grasping the ball bearings. Only then would Edison ease back and allow sleep to
consume him whole. At the moment he began to dream, his muscle tone would
relax  and  he  would  release  the  ball  bearings,  which  would  crash  on  the  metal
saucepan  below,  waking  him  up.  He  would  then  write  down  all  of  the  creative
ideas that were flooding his dreaming mind. Genius, wouldn’t you agree?
CONTROLLING YOUR DREAMS—LUCIDITY
No  chapter  on  dreaming  can  go  unfinished  without  mention  of  lucidity.  Lucid
dreaming occurs at the moment when an individual becomes aware that he or she
is  dreaming.  However,  the  term  is  more  colloquially  used  to  describe  gaining
volitional control of what an individual is dreaming, and the ability to manipulate
that experience, such as deciding to fly, or perhaps even the functions of it, such
as problem solving.
The concept of lucid dreaming was once considered a sham. Scientists debated
its  very  existence.  You  can  understand  the  skepticism.  First,  the  assertion  of
conscious control over a normally non-volitional process injects a heavy dose of
ludicrous into the already preposterous experience we call dreaming. Second, how
can you objectively prove a subjective claim, especially when the individual is fast
asleep during the act?
Four  years  ago,  an  ingenious  experiment  removed  all  such  doubt.  Scientists
placed  lucid  dreamers  inside  an  MRI  scanner.  While  awake,  these  participants
first  clenched  their  left  and  then  right  hand,  over  and  over.  Researchers  took
snapshots  of  brain  activity,  allowing  them  to  define  the  precise  brain  areas
controlling each hand of each individual.
The participants were allowed to fall asleep in the MRI scanner, entering REM
sleep where they could dream. During REM sleep, however, all voluntary muscles
are  paralyzed,  preventing  the  dreamer  from  acting  out  ongoing  mental
experience. Yet, the muscles that control the eyes are spared from this paralysis,

---

and give this stage of sleep its frenetic name. Lucid dreamers were able to take
advantage  of  this  ocular  freedom,  communicating  with  the  researchers  through
eye  movements.  Pre-defined  eye  movements  would  therefore  inform  the
researchers  of  the  nature  of  the  lucid  dream  (e.g.,  the  participant  made  three
deliberate  leftward  eye  movements  when  they  gained  lucid  dream  control,  two
rightward  eye  movements  before  clenching  their  right  hand,  etc.).  Non-lucid
dreamers  find  it  difficult  to  believe  that  such  deliberate  eye  movements  are
possible  while  someone  is  asleep,  but  watch  a  lucid  dreamer  do  it  a  number  of
times, and it is impossible to deny.
When  participants  signaled  the  beginning  of  the  lucid  dream  state,  the
scientists  began  taking  MRI  pictures  of  brain  activity.  Soon  after,  the  sleeping
participants  signaled  their  intent  to  dream  about  moving  their  left  hand,  then
their  right  hand,  alternating  over  and  over  again,  just  as  they  did  when  awake.
Their hands were not physically moving—they could not, due to the REM-sleep
paralysis. But they were moving in the dream.
At least, that was the subjective claim from the participants upon awakening.
The  results  of  the  MRI  scans  objectively  proved  they  were  not  lying.  The  same
regions of the brain that were active during physical right and left voluntary hand
movements  observed  while  the  individuals  were  awake  similarly  lit  up  in
corresponding ways during times when the lucid participants signaled that they
were clenching their hands while dreaming!
There could be no question. Scientists had gained objective, brain-based proof
that  lucid  dreamers  can  control  when  and  what  they  dream  while  they  are
dreaming.  Other  studies  using  similar  eye  movement  communication  designs
have  further  shown  that  individuals  can  deliberately  bring  themselves  to  timed
orgasm  during  lucid  dreaming,  an  outcome  that,  especially  in  males,  can  be
objectively verified using physiological measures by (brave) scientists.
It remains unclear whether lucid dreaming is beneficial or detrimental, since
well  over  80  percent  of  the  general  populace  are  not  natural  lucid  dreamers.  If
gaining voluntary dream control were so useful, surely Mother Nature would have
imbued the masses with such a skill.
However,  this  argument  makes  the  erroneous  assumption  that  we  have
stopped evolving. It is possible that lucid dreamers represent the next iteration in
Homo  sapiens’  evolution.  Will  these  individuals  be  preferentially  selected  for  in
the  future,  in  part  on  the  basis  of  this  unusual  dreaming  ability—one  that  may
allow  them  to  turn  the  creative  problem-solving  spotlight  of  dreaming  on  the

---

waking  challenges  faced  by  themselves  or  the  human  race,  and  advantageously
harness its power more deliberately?
I
. One example is language learning, and the extraction of new grammatical rules. Children exemplify this.
They  will  start  using  the  laws  of  grammar  (e.g.,  conjunctions,  tenses,  pronouns,  etc.)  long  before  they
understand what these things are. It is during sleep that their brains implicitly extract these rules, based on
waking experience, despite the child lacking explicit awareness of the rules.
II
.  Quoted  by  B.  M.  Kedrov  in  his  text,  “On  the  question  of  the  psychology  of  scientific  creativity  (on  the
occassion of the discovery by D. I. Mendeleev of the periodic law).” Soviet Psychology, 1957, 3:91–113.
III
. This ode to the creative juices of dream sleep is sometimes also attributed to the French Symbolist poet
Paul-Pierre Roux.

---