DETECTION OFDECEPTION:
EVENT-RELATEDPOTENTIALS
P300 is a brain wave derived from the electroen-
cephalogram (EEG), which has recently been used as a
novel information channel in the detection of decep-
tion. The traditional channels are recorded from the
autonomic nervous system and include physiological
activity such as respiration pattern, blood pressure, and
skin conductance. In contrast, the EEG is a record of
sequential, spontaneously changing voltages as a func-
tion of time, recorded from the scalp surface in humans.
It reflects the spontaneous activity from the underlying
cerebral cortex. If as these changing voltages occur, a
discrete stimulus event (such as a light flash) occurs,
the EEG breaks into a series of somewhat larger peaks
and troughs, called components. This series of waves is
called an event-related potential (ERP).
These early peaks and troughs represent sensory
activity (exogenous ERP components), and the later
(endogenous) components may represent the psycho-
logical reaction to the sensory events. P300 is the
name of one heavily researched ERP. It is elicited by
stimulus events that are rare and meaningful to sub-
jects. For example, if a stimulus series consists of a set
of randomly occurring first names, each presented
singly on a display screen about every 3 s, and the
subject’s own first name is one of the stimuli pre-
sented about 15% of the time, with the remaining 85%
of the presentations being of other, unfamiliar names,
the P300 will be elicited by the rare, meaningful (sub-
ject’s own) name. P300 is named in respect of its pos-
itive (P) polarity and its occurrence at about 300 to
800 ms after the stimulus onset. Simple stimuli such
as brief sounds elicit early P300 peaks (300–400 ms),
whereas more complex stimuli such as words elicit
later peaks (500–800 ms).
It occurred to Dr. J. Peter Rosenfeld and colleagues
in the early 1980s that P300 might be used in decep-
tion detection situations to index recognition of the
presentation of crime scene details known only to per-
petrators (and the authorities) and not to innocent sus-
pects. The protocol would involve presentation (usually
on a display screen) of items of information, such as
possible murder weapons (e.g., pistol, rifle, knife,
axe). The guilty party, but not the innocent subject,
would react with a P300 to the actual murder weapon
(e.g., the pistol), called the probestimulus. Neither
guilty nor innocent subjects would react to the other,irrelevantitems from the weapons category, which
were not actually used in the crime, as the guilty party
would know. Thus, the difference in P300 amplitude
between the probe-evoked ERP and the irrelevant-
evoked ERP indicates guilt. This protocol was closely
related to the Guilty Knowledge Test (GKT) invented
by David Lykken in 1959, which used autonomic ner-
vous system responses to stimuli. One difference was
that in the P300 protocol, there was usually a third
stimulus type used, also rarely presented, called the
target.This was typically one other irrelevant item but
one to which the subject is told to respond by pressing
a unique button. In one version of the protocol, the
subject is told to press a “No” button (for “No, I don’t
recognize this”) in response to both probes and irrele-
vant items and “Yes” (“I do recognize this”) in
response to targets. Of course, in saying “No” to the
probe, the guilty subject lies, but it is hoped that his
P300 ERP reveals his guilty recognition all the same.
The target stimulus is used to force attention onto the
display screen, since the three stimulus types are pre-
sented unpredictably in random sequence, and if the
subject neglects to respond to the target stimulus as
instructed, the operator knows that the subject is not
paying attention and will report this to the authorities.
But if the subject is always paying attention, he or she
cannot avoid seeing the probe stimuli, which evoke
P300s in guilty subjects.
Early reports in the 1990s (by J. Peter Rosenfeld and
colleagues and by John J. B. Allen and Emanuel
Donchin and their respective coworkers) with this pro-
tocol reported high overall accuracy (80–95% correct
classification of guilty and innocent subjects), and they
were received with considerable enthusiasm; it was
naively believed that because the P300s occurred with
such short reaction times (fractions of a second post-
stimulus) relative to the slow autonomic reaction times
in the GKT, the P300-based protocols would resist
countermeasures(CMs), intentional covert responses
subjects can learn to make that can defeat the GKT.
Unfortunately, J. Peter Rosenfeld and colleagues
showed in 2004 that P300-based GKTs were also vul-
nerable to CMs. The guilty subjects were simply
trained to covertly respond (e.g., with secret toe wig-
gles) systematically to irrelevant stimuli, thus turning
them into P300-evoking targets. It then became impos-
sible to distinguish between probe and irrelevant P300s,
whose typical difference without CMs indexed guilt.
Reports from John J. B. Allen’s lab showed similar
results.Detection of Deception: Event-Related Potentials——— 197D-Cutler (Encyc)-45463.qxd 11/18/2007 12:41 PM Page 197