HOW do we ever survive a day? We slip, we blunder, we goof, we dribble, we
say the wrong things at completely inappropriate times—and given half a
chance we’ll try and do the whole lot together. One thing’s certain: we’d have
many more problems if it weren’t for our fix-it-quick brains watching over us,
alerting us to impending doom and jumping in to rescue us. Dozens of times a
day, at the first sign of trouble, the mental equivalent of an alarm bell goes
off inside our head, screaming out a warning that something isn’t quite
right.
Researchers are now tuning in to these mental alarms with the latest in brain
imaging techniques, trying to get a handle on exactly where the signals come
from and how they keep us on the straight and narrow. Ever since they were
discovered, the alarms have been thought of as the cry of a dedicated blunder
detector—the brain’s attempt to flag our mistakes the instant they occur,
telling us how we’ve messed up and perhaps warning us to slow down, listen up,
or try again.
But one group of researchers is questioning this view, because they think
they have caught the alarms going off at the wrong time. Is our blunder detector
itself prone to errors, or might these mental alarms actually be a step ahead of
the game, screaming at us to “watch out” when things get confusing, when extra
attention might yet avert a mistake?
Advertisement
One of the first people to spot these alarm signals was Michael Coles, a
cognitive neuroscientist from the University of Illinois in Urbana-Champaign.
Scientists have long known that the brain somehow monitors everything we do,
keeping track of our every move, recording and predicting. So all our
experiences are reflected in our brain activity—even when things don’t go
quite as planned. In 1990, Coles and psychologist William Gehring, now at the
University of Michigan in Ann Arbor, were busy recording this mass of activity
in the brains of volunteers wearing a cap of scalp electrodes. The volunteers
were doing a simple test of reaction time, but Coles and Gehring found that the
most intriguing brain activity appeared not when they got the test
right—but when they blundered.
Coles asked his volunteers to look for the letter H or S in the middle of a
computer screen. He told them to squeeze a button with their left hand when H
popped up; and another button with their right hand when an S appeared. This was
easy enough until the screen started to fill with extraneous letters. Some
combinations—in particular, a central H or S surrounded by the opposite
letter—invariably cause even focused folk to respond incorrectly most of
the time.
Every time these people made an error, the brain produced a telltale streak
of nerve impulses as if it had instantly caught the mistake. Researchers know
the brain somehow spots errors—after all, people learn by making, and then
correcting, mistakes. These impulses must be error messages, says Coles—a
neural “gotcha” or a critical “uh-oh!” that cues a person to correct their
mistake, or proceed more carefully next time around. He has traced the alarm to
a busy curve of grey matter just under the frontal lobes, known as the anterior
cingulate cortex or ACC.
Coles and his colleagues think that the brain takes stock of any task at
hand, noting the correct answer—for example, H, left button; S, right
button. If you respond impulsively, before your brain has a chance to process
the task properly, you’re likely to err. And when that happens, Coles thinks the
lightning-fast ACC compares the correct answer with your response, and if they
don’t match, it rockets off a neural SOS to frontal “planning” areas of the
brain. In experiment after experiment in Coles’s lab, when people estimate time
incorrectly, choose the wrong word, or move the wrong hand or foot, he sees the
same characteristic ACC signal. Seconds later, the misguided person often tries
the correct answer, and responds more cautiously on the next run.
All this leads Coles to conclude that he has captured the essence of error
detection. He believes this is the brain’s generic system for detecting
mistakes, and possibly the first step towards recovering from typing errors,
social howlers and every gaffe in between. Whether you’ve mispronounced a word
or got caught in the lift door, your ACC shrieks and you rethink your behaviour,
says Coles.
But not everyone agrees that Coles has found a generic blunder-spotter that
somehow “knows” the right answer. Some researchers argue that the ACC is not so
smart. They say that Coles has actually caught something far more simple: a
neural cry for attention that is made whenever a task grows complex, when there
is more scope for making mistakes.
Not so clever
The chief proponents of this rival theory are Cameron Carter, a psychiatrist
at the University of Pittsburgh in Pennsylvania, and Jonathan Cohen, a
neuroscientist at Pittsburgh and now also at Princeton University in New Jersey.
Last year, they published their evidence for an ACC attention alarm in the
journal Science(vol 280, p 747). According to their line of thinking,
the ACC is clueless about the correct response—all it knows is that
several stimuli are clashing in a confusing way. As confusion often goes hand in
hand with errors, they believe this could explain why Coles spots the mental
alarms going off whenever we blunder.
Carter, Cohen and their colleagues used magnetic resonance imaging to peer
into the brains of people making judgments about sequences of letters. First
they showed a letter A or B, then 10 seconds later an X or Y. They asked people
to push one button when they had seen the sequence AX. All other sequences
required a different response. In most cases, A and X appeared one after the
other, prompting people to expect that sequence. But when a different sequence
appeared the ACC immediately fired up, even when a person responded correctly,
and especially when the signal was partly correct—in other words AY or BX.
It looked as though the researchers had found Coles’s error signal without any
error.
In this experiment, Carter and Cohen say, the ACC fireworks were signalling
conflict or confusion. Anticipating the AX sequence, people were ready to push a
certain button. But faced with a different stimulus, they were forced to
override that urge in order to respond correctly. Carter and Cohen suggest that
the signal is not a neural “gotcha” message, but more of a call to “be careful”.
This kind of alarm may come into play whenever we are faced with competing
stimuli—perhaps talking on the phone and hearing someone else talk in the
other ear, or reading the word “blue” when the letters are coloured red (the
so-called Stroop effect).
Carter asserts that picturing the ACC as a complexity detector, rather than
an error detector, is a more “parsimonious” way of viewing the brain. To detect
complexity, your brain need only sense an increasing number of signals. But to
detect an error, your brain must somehow know both the right and wrong solutions
to a given task, and must compare them with your responses. “We’re arguing for a
slightly dumber ACC,” Carter says.
Great expectations
Coles calls this hypothesis “provocative”, but he’s not convinced. He
believes if subjects expect to see AX, they will prepare to answer as
appropriate to that sequence. Then if they see something different, the answer
they prepared could be thought of as an error. For another thing, he says, the
brain must have some way to spot errors. Where would we be, he asks, if we
failed to learn from our mistakes? We’d all flunk out of school or wreak havoc
whenever we got in a car. And, in turn, Coles has found alarms ringing at times
that wouldn’t be predicted by the rival theory—in response to an error
without a hint of conflict.
In 1997, Coles, Wolfgang Miltner of Friedrich-Schiller University in Jena,
Germany, and Christoph Braun of Eberhard-Karls University in Tübingen
reported a similar set of reaction time tests, but this time the correct answers
were harder to spot and the subjects couldn’t tell when they had erred. After
each response, the researchers gave their volunteers feedback, effectively
telling them whether they were right or wrong. And when people were told they’d
been wrong, the ACC shot off those familiar neural signals.
In this case, the signals occurred with error feedback alone. “So in some
situations, clearly conflict is not involved, and yet the ACC response still
occurs,” Coles says. But Carter and Cohen believe that the feedback experiment
is open to interpretation. Perhaps, they say, a person who learns of an earlier
mistake mentally replays the scene, experiencing the conflicting responses anew.
Coles dismisses this explanation as “hand-waving”.
His dismissal may be justified by an array of other studies which seem to
back the idea that the ACC is an error detector. Some say that hyperactive
signals from the ACC contribute to obsessive-compulsive disorder (OCD), in which
a person anxiously repeats menial tasks, like washing their hands or locking the door
(“Over and over and over”, 91av, 2 August 1997, p 26).
Last year, Gehring found that OCD sufferers display unusually large and
lingering ACC impulses when they make mistakes during classic tests of reaction
time. Gehring says these hyperactive error signals may incorrectly convince OCD
sufferers that their hands are still dirty or that the door is still
unlocked.
Conversely, there is some preliminary evidence that muted ACC signals may
reduce a person’s ability to distinguish right from wrong. In a small experiment
last year, John Allen of the University of Arizona in Tucson found that people
who describe themselves as having withdrawn and poorly adjusted behaviour, and
other characteristics often ascribed to psychopaths, have particularly weak ACC
messages when they err on simple tasks, even when they receive a punishment
signal in response to an error (in this case, a high-pitched tone). Allen
believes that his research may eventually shed light on some undesirable
behaviours, such as stealing and running red lights. People with unusually weak
error signals may be more likely to steal because they don’t sense the
impropriety of their actions, Allen says.
Even mild variations in ACC signals seem to affect our ability to predict
errors. In a study about to be published in the journal Psychophysiology
(vol 36, p149), Coles and his colleagues persuaded six
people to stay awake all night performing visual search and memory tests while
their brain activity was monitored. The ability to detect errors and the size of
the ACC signals both fell as time passed. The later it got, the worse the
sleep-deprived participants were at finding the right answers. The experiment
knits performance and error detection more tightly together, Coles says.
But not everyone is happy with this interpretation. A growing body of
research shows the ACC is in a prime attention-grabbing position in the brain,
perfectly suited to working in the way Carter and Cohen propose. Studying rhesus
monkeys, Gary Aston-Jones, a psychiatrist at the University of Pennsylvania in
Philadelphia, and his colleagues recently recorded what went on in the locus
coeruleus —a nervous centre at the base of the brain, which receives
neural messages from the ACC.
Aston-Jones found evidence that the locus coeruleus can switch the brain
between two modes of attention—a focused state, in which the monkeys
concentrate solely on the job at hand, and a flexible, scanning state, in which
they pay just enough attention to get the job done. He speculates that when the
task grows harder or more mundane, the ACC could prompt the locus coeruleus to
change the mode of attention accordingly.
Last year, George Bush, a neuroscientist at Harvard University in Boston and
his colleagues described a task where this sort of attention switch might be
called on. They used a modified version of a test for the classic Stroop
effect—a kind of cognitive multi-tasking problem, which would call on the
brain to perk up and concentrate. They asked people to indicate the number of
words on a screen, regardless of what the words were. This is easy when the
words are neutral animal names, say. But if the words are distracting—for
example, the word “two” written three times—it’s much harder to count
correctly. In response to these brain teasers, neural sparks fired off in the
ACC—possibly a cry for more attention, Bush says.
Still, Coles notes that even this ACC activity might be cast as error
detection. “Generally, we assume that people err because they act too quickly,”
Coles says. “They decide what to do before they’ve had a chance to fully analyse
the stimulus. You might say they guess, and that could be the case here,
ٴǴ.”
But need the results described by coles and by Carter and Cohen be in
conflict? Last year, Michael Posner from the University of Oregon in Eugene
found that different parts of the ACC were active in slightly different
situations. He and his colleagues used scalp electrodes to record ACC activity
while people made errors and while they were receiving delayed feedback on their
errors. The researchers found that when errors occur, the ACC’s middle section
immediately lights up. Being told of an error after the fact sets an area
further back in the ACC ablaze.
Given the ACC’s many talents, Posner says that Coles and the Carter and Cohen
team might both be right. And Michael Falkenstein, a neuroscientist at the
University of Dortmund in Germany—who discovered ACC alarms about the same
time as Coles—also believes that the ACC might react whether we respond
correctly or incorrectly, in both conflicting and non-conflicting situations.
One part of the ACC may detect conflicts. Next door, another spot may pick up
errors. It’s possible.
Even though today’s brain imaging techniques are better than ever before,
scientists still struggle to pinpoint neurons that rest just centimetres apart.
Muddying the debate, the different research teams use different techniques to
record ACC fireworks. Coles can get very precise measurements of the timing of
the signals, but can only achieve slightly fuzzy localisation of that activity.
Carter and Cohen’s method, on the other hand, can pinpoint the position of the
neural impulses in the ACC more precisely, but to record the signals they must
average the activity over time. “We might all be touching the same elephant, but
feeling different parts,” Cohen acknowledges.
