Jan 29

The frustrating puzzle of dream amnesia

All of us, I’m sure, have many times had the experience of waking from a dream with an intensely vivid and detailed memory of a series of fascinating events—but five minutes later the memory is virtually gone.  We can barely even remember the topic of the dream, and all we remember about the events is that they were exciting.  This “dream amnesia” is not an exceptional thing, rather the opposite:  it is nearly universal.  The only way to remember a dream is to write down an account of it immediately after waking, and even then it is not clear whether our later memory is of the dream itself or of thinking about it afterward.  We might remember snippets of old dreams, but we can never get back that original vividness.

Dreams thus seem to contradict the basic rules of memory.  Normally the more exciting an experience, the more easily it is remembered; and even boring events are usually remembered for at least a few hours.  Dreams are universally fascinating, yet their memory fades in minutes.  What is the cause of this very striking amnesia?

It is surprising how little attention has been paid to this question.  The effects of REM sleep on memory consolidation have been studied quite extensively, but only in the context of memories established prior to sleep.  There have been a number of studies of psychological factors associated with dream recall, but as far as I can tell there has been hardly any research directly addressed to the physiology of dream amnesia itself.

And yet to a neuroscientist familiar with the neurophysiology of memory, there is a rather natural hypothesis.  We know, or at least have very strong evidence, that the formation of memory relies on a process called “long-term potentiation”, abbreviated LTP.  LTP takes place when two neurons that are connected to each other are simultaneously active, and results in strengthening of the synaptic connection between them.  LTP was first observed in the hippocampus, but also takes place in the neocortex and other brain areas.

LTP has been one of the hottest topics in neuroscience for three decades, and we have learned an enormous amount about it—the rules that govern it and the physiological mechanisms that implement it. One of the most important facts is that although LTP can be produced in seconds, it creates a memory trace that can last for days, or weeks, or even longer.  But this long-lasting trace does not arise immediately:  it requires a series of “consolidation” processes.

In more detail, when two connected neurons are strongly activated, the synapse between them is strengthened very rapidly.  This strengthening is produced by activation of some intracellular chemicals called kinases—I’m going to gloss over the details.  However, the strengthening doesn’t necessarily last very long.  Unless those kinases remain activated, the synaptic strength drops back to baseline over the course of a few minutes.  In other words, the memory trace disappears.  There is also a later phase of consolidation that relies on protein synthesis, but it is not relevant to my point here so I am not going to say any more about it.

Thus following the initial stimulus, which only needs to last for a few seconds (or even less), LTP goes through a series of stages:

  1. The induction phase, dependent on activation of several kinases, lasting a few minutes.
  2. The early maintenance phase, dependent on continued activation of kinases, and particularly on activation of one called Protein Kinase Mζ, lasting a few hours.
  3. The late maintenance phase, dependent on protein synthesis, lasting for days or longer, perhaps in some cases indefinitely.

Knowing those facts, it seems natural to hypothesize that dream amnesia results from a failure of LTP to consolidate from the induction phase to the early maintenance phase.  This would give rise to a fading of the memory trace over the course of minutes, just as we actually experience following a dream.

An additional encouragement for that hypothesis comes from the fact that, even though the state of the brain during REM sleep is in most respects very similar to the brain state during wakefulness, one of the main differences is in the activity of several neurotransmitter systems.  The norepinephrine and serotonin systems, in particular, are very strongly suppressed during REM—and there is quite a bit of evidence that norepinephrine plays an important role in modulating LTP.  (Allan Hobson pointed out this connection as long ago as the 1970s, and has discussed it many times.)

But here’s the problem:  the experimental evidence, such as it is, just doesn’t support the LTP-maintenance-failure hypothesis.

To be sure, there isn’t a great deal of evidence.  As far as I can tell, only one study has directly examined the time course of LTP induced during REM sleep:  Bramham, C. R., Maho, C., & Laroche, S. (1994). Suppression of long-term potentiation induction during alert wakefulness but not during ‘enhanced’REM sleep after avoidance learning. Neuroscience, 59(3), 501-509.  Working with rats, Clive Bramham and his colleagues found it just as easy to induce LTP during REM sleep as during wakefulness, and didn’t see any shortening of its duration.

Does that mean the hypothesis is wrong?  Possibly.  It would be nice to see a replication, just for the sake of confidence, but even if the observations hold up, there may still be an escape route.  The Bramham study induced LTP by the easiest method, using a high-frequency train of strong stimuli to hippocampal inputs.  That’s a much stronger stimulus than occurs naturally in the brain.  It seems possible that LTP induced in a different way—by “theta burst” stimulation, for example—might result in a more labile memory trace.  Beyond this, there is evidence that the timing of hippocampal cell activity with respect to the theta rhythm changes during REM sleep (Poe et al, 2000), and there are known to be interactions between theta and LTP—the Bramham study wouldn’t say anything about that.

One way or another, the bottom line is that dream amnesia is an important part of human experience.  It would be nice to know why it happens, and we really ought to have the tools to figure it out.


  • Poe, G. R., Nitz, D. A., McNaughton, B. L., & Barnes, C. A. (2000). Experience-dependent phase-reversal of hippocampal neuron firing during REM sleep. Brain research, 855(1), 176-180.
  • Hobson, J. A., Pace-Schott, E. F., & Stickgold, R. (2000). Dreaming and the brain: toward a cognitive neuroscience of conscious states. Behavioral and brain sciences, 23(06), 793-842.


Older posts «