Re: Event Related Potentials

From: HARNAD Stevan (
Date: Wed May 08 1996 - 15:51:36 BST

> From: "Whitehouse Chantal" <>
> Date: Wed, 8 May 1996 10:45:03 GMT
> I understand what event related potentials are but I'm unsure about
> their use. I read that they are used to "study the time course of
> higher-level processes in the human brain, such as perception and
> attention" but I don't really understand how this is concluded from
> the fluctuation which is produced on the EEG. I also can't see the
> use of this information. What do we learn about the mind or the brain
> from knowing how long it takes to perceive a light flash or a word
> being spoken?

First, you have to ask how the timing of processes could be done
without Event-Related Potentials (ERPs): by varying (1) the order of
tasks, or by varying (2) the complexity of tasks, and looking at (3)
how well people accomplish them, and (4) how quickly (reaction time,
RT), using (5) Donders's subtractive method (see the Posner paper for
an explanation of what that is). Together with (6) a theory of what the
unobservable internal processes might be, (1) - (5) can be used, with
difficulty, to test it.

You might infer that to be able to do X (say, understand a word) you
must first do A (detect that there's something there), (b) recognise
that it's a word, (c) identify which word it is, and (d) retrieve its
meaning. So your theory might say that there are at least 4 processes
that must take place: Are there really four? If so, do they happen in
that order, one after the other? Or do some or all of them occur at the
same time, in parallel? These are some of the kinds of questions you
might want to try to answer with behaviour alone. Posner describes the
Sternberg reaction time experiments; these are representative examples.

Now what more can you do if you also have the help of ERPs? First, ERPs
are not just EEG (brain waves): They are specific bits of EEG that are
time-locked to an EVENT (that's why they're called Event Related
Potentials). You get ERPs by averaging the EEG that comes just before
and after an event, say, an input stimulus like a spoken word or a
picture of an apple. From analysing the time-locked EEG, averaging it
over and over, you can eventually derive the smooth averaged wave
called the ERP. Then you can say more things: You can say where on the
scalp the EEG was most prominent (over the hearing areas, or the visual
areas, or the motor areas, over the right or the left, etc.) and you
can analyse the components of the ERP (the early parts, which tend to
be sensory, and the later parts, which tend to be cognitive, sensitive
to expectancy, attention and memory).

Now let's go back to where you were without ERPs: You had a theory about
what some of the processes might be that go on in your head when you do
something. With behaviour alone, you could only test this theory by
varying the task and measuring the reaction time and the accuracy of
the subject performing the task. With ERPs, you could, conceivably
identify an ERP or part of an ERP that corresponds to your hypothetical
process, and test whether it appears when your theory predicts it should
happen: Does the "decoding" process happen after the "detection"
process? If you are lucky (or clever) in your theorising, you may even
be able to say where in the brain these processes happen, not just when.
Posner gives good examples of this.

Note that the new brain imaging techniques Posner describes (PET, MRI)
do not have the accuracy to pick out a brief process in time: They show
activity deep in the brain, which is good, because the ERP only
measures activity at the top of the head, but PETs and MRIs are
"smeared" across time: they measure an overall state, spread across
seconds or minutes, not the brief process, possibly only a few milliseconds
long, that an ERP can detect and time.

That having been said, note that neither the timing nor the localising
capacity can reveal what the process really is: That depends on the mind
of the theorist: How well it can guess or deduce what's going on in
there. The timing and localising tools are no better than the hypotheses
they are used to test. Where are those hypotheses to come from?


Here is an example of the kinds of things you can find out from ERPs.
This is a study of the "unexpected" in the appreciation of music. Music
theorists have long suggested that some of the enjoyment we get from
music has to do with expectations and violations of expectations.
For more information, email to:
Or see her web page:

      1995, Vol.21, No.6, pp.1278-1296
      Musicians and nonmusicians listened to musical phrases that were
      either selected from the classical repertoire or composed for the
      experiments. The phrases ended either congruously or with a
      nondiatonic, diatonic, or rhythmic violation. Percentage of correct
      responses was analyzed in Experiment 1, and event-related potentials
      (ERPs) were recorded in Experiments 2 and 3. Musicians performed
      better than nonmusicians in recognizing familiar musical phrases and
      classifying terminal notes. The differences found as a function of
      expertise were larger for unfamiliar than for familiar melodies. The
      expertise were larger for unfamiliar than for familiar melodies. The
      ERPs to the end notes differed both in terms of amplitude and latency
      between musicians and nonmusicians, and as a function of
      participants' familiarity with the melodies and type of violation.
      Results show that expertise influences the decisional rather than the
      purely perceptual aspects of music processing and that ERPs can
      provide important insight into the study of music perception.

      NEUROSCIENCE LETTERS, 1994, Vol.168, No.1-2, pp.101-105
      Musicians and non-musicians were presented with short musical phrases
      that were either selected from the classical musical repertoire or
      composed for the experiment. The phrases terminated either in a
      congruous or a 'harmonically', 'melodically', or 'rhythmically'
      incongruous note. The brain waves produced by the end-notes differed
      greatly between musicians and non-musicians, and as a function of the
      subject's familiarity with the melodies and the type of incongruity.
      The timing of these brain waves revealed that musicians are faster
      than non-musicians in detecting incongruities. This study provides
      further neurophysiological evidence concerning the mechanisms
      further neurophysiological evidence concerning the mechanisms
      underlying music perception and the differences between musical and
      linguistic processing.

This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:39 GMT