214
serine-9 or inactivation via other signaling
pathways. In contrast, growth-inhibitory
signals act in the opposite direction to
dephosphorylate the serine-9 residue.
All of these results suggest a potential
strategy for promoting axonal regeneration
after neural injury using pharmacological
inhibitors of GSK-3β. Indeed, we have
recently demonstrated that inactivation of
GSK-3β is necessary for axon regeneration
from mature neurons of the peripheral nervous
system after nerve transection (20).
However, it must be emphasized that efficient
axon extension probably requires localized
inactivation of GSK-3β at the distal end
of the growing axon, as well as activity of
GSK-3 toward “unprimed” substrates.
Although low concentrations of GSK-3β
inhibitors lead to elaboration of multiple
axons by hippocampal neurons, more complete
global inhibition of GSK-3β almost
certainly impedes axon growth (5, 14).Thus,
manipulation of GSK-3β may be a promising
therapeutic strategy to promote functional
recovery after injury to the central
nervous system. But an important challenge
remains how to achieve appropriate local
regulation of GSK-3β using pharmacological
inhibitors or gene therapeutic constructs.
GSK-3β may be a “master” kinase that
mediates convergent signals from multiple
extracellular cues affecting axon morphogenesis.
It achieves this by coordinating
regulation of all three aspects of microtubule
assembly: polymerization, stabilization,
and dynamics (see the figure).
However, regarding the functions of GSK-
3β in regulating neuronal morphology, its
control of microtubule assembly may only
be the “tip of the iceberg.” GSK-3β almost
certainly regulates actin filaments, the other
important component of the axonal
cytoskeleton (17). Furthermore, GSK-3β
presumably regulates metabolism required
for axon growth both locally in the axon and
at the level of gene transcription by controlling
the NFAT (nuclear factor of activated T
cells) family of transcription factors and
possibly other transcriptional regulators.
Surprisingly, in view of its importance, the
nervous system of mice that lack GSK-3β
has not been extensively analyzed, and no
animals lacking the related isoform, GSK-
3α, have yet been reported.
Developing inhibitors of GSK-3β is a
major target of the pharmaceutical industry
because of the potential of such drugs for
treating type 2 diabetes and Alzheimer’s disease.The
success of lithium in treating mood
disorders may reflect its ability to inhibit
GSK-3β, among other effects (6). Given the
evidence that GSK-3β is important for regulating
nervous system development, the use
of GSK-3β inhibitors during pregnancy will
need to be very carefully evaluated.
References
1. E.W. Dent, F. B. Gertler, Neuron 40, 209 (2003).
2. A.Akhmanova, C. C. Hoogenraad, Curr. Opin. Cell Biol.
17, 47 (2005).
3. H. Jiang,W. Guo, X. Liang,Y. Rao, Cell 120, 123 (2005).
4. T.Yoshimura et al., Cell 120, 137 (2005).
5. F. Q. Zhou, J. Zhou, S. Dedhar, Y. H. Wu, W. D. Snider,
Neuron 42, 897 (2004).
6. L. Meijer, M. Flajolet, P. Greengard, Trends Pharmacol.
Sci. 25, 471 (2004).
7. B.W.Doble,J.R.Woodgett,J. Cell Sci. 116,1175 (2003).
8. Y. Fukata et al., Nat. Cell Biol. 4, 583 (2002).
9. M. Bienz, Nat. Rev. Mol. Cell Biol. 3, 328 (2002).
10. J. Zumbrunn, K. Kinoshita, A. A. Hyman, I. S. Nathke,
Curr. Biol. 11, 44 (2001).
11. N.Trivedi, P. Marsh, R. G. Goold,A.Wood-Kaczmar, P. R.
Gordon-Weeks, J. Cell Sci. 118, 993 (2005).
12. R. G. Goold, P. R. Gordon-Weeks, Mol. Cell. Neurosci.
28, 524 (2005).
13. A. I. Lyuksyutova et al., Science 302, 1984 (2003).
14. O. Krylova et al., Neuron 35, 1043 (2002).
15. J.A. Del Rio et al., Curr. Biol. 14, 840 (2004).
16. R. Salie, V. Niederkofler, S. Arber, Neuron 45, 189
(2005).
17. B.J.Eickholt,F.S.Walsh,P.Doherty,J. Cell Biol. 157,211
(2002).
18. J. K.Atwal, K. K. Singh, M.Tessier-Lavigne, F. D. Miller, D.
R. Kaplan, J. Neurosci. 23, 7602 (2003).
19. Y. Uchida et al., Genes Cells 10, 165 (2005).
20. F. Q. Zhou et al., in preparation.
10.1126/science.1110301
A
lthough primates and other animals
seem to have some understanding of
mind (that is, the behavior of others),
the concept of belief seems to be a specifically
human ability. Comprehending false
belief is the clearest sign of understanding a
critical aspect of the mind: its subjectivity
and its susceptibility to manipulation by
information. It is thought that children
develop an understanding of false belief
around 4 years of age. However, on page 255
of this issue, Onishi and Baillargeon (1)
report that infants as young as 15 months
have insight into whether a person acts on
the basis of a mistaken view (false belief)
about the world. This discrepancy touches
on important issues. An understanding of
false belief at 4 years of age suggests that
this ability may be constructed in a cultural
process tied to language acquisition. In contrast,
competence at 15 months suggests that
this ability is part of our purely biological
inheritance. What could account for the discrepant
findings?
Children’s understanding of false belief
has hitherto been assessed using a verbal
false-belief task in which the experimenter
enacts stories.An example of such a story is
as follows: A protagonist (let’s call him
Max) puts a toy or doll (object) in one location
and then doesn’t see it moved to a second
location (2). When asked by the experimenter,
most 3-year-olds wrongly claim
that Max will look for the object in the second
location (where they know it is). This
finding with 3-year-olds has been confirmed
despite many attempts to improve
the potential shortcomings of the verbal
false-belief task [see meta-analysis by
Wellman et al. (3)]. These results contrast
with those from Onishi and Baillargeon’s
study in which 15-month-old infants were
tested with a nonverbal false-belief test. In
this test, infants were familiarized with an
adult actor hiding and then retrieving a toy
(a plastic slice of water melon) in either a
yellow or a green box (see the figure). The
looking times of the infant subjects were
then computed in a series of trials that
tested whether the actor held a true or false
belief about the location of the toy. Onishi
and Baillargeon found that the infants
“expected” the actor to search for the toy
based on the actor’s belief about its location,
regardless of whether the location was
actually correct. So, why would 3-year-olds
fail to provide the correct answer in a verbal
false-belief test, when 15-month-old infants
can correctly anticipate erroneous actions in
the nonverbal false-belief test?
Part of the explanation might come from
previous studies that used eye gaze as a
measure of understanding in 3-year-olds.
Three-year-olds look to the correct (initial)
location when anticipating Max’s return
there, even when they explicitly make the
incorrect claim that Max will go to the second
location.This early indication of understanding
Max’s mistake has been dubbed
implicit, because many of these children
show no awareness of the knowledge
implicitly conveyed in their correct eye
gaze (4). Nonetheless, children at the age of
21⁄2 years show absolutely no sign of this
earlier, implicit understanding (5).
P S Y C H O L O G Y
Infants’ Insight into the Mind:
How Deep?
Josef Perner and Ted Ruffman
J. Perner is in the Department of Psychology,
University of Salzburg, A-5020 Salzburg, Austria. T.
Ruffman is in the Department of Psychology,
University of Otago, Dunedin, New Zealand. E-mail:
josef.perner@ sbg.ac.at; tedr@psy.otago.ac.nz
8 APRIL 2005 VOL 308 SCIENCE www.sciencemag.org
P E R S P E C T I V E S
Published by AAAS
215
Converging evidence comes from children’s
word learning, which also shows sensitivity
to false belief around 3 years and not before
(6). In sum, the evidence of an earlier,
implicit understanding does not solve but
rather exacerbates the puzzle about Onishi
and Baillargeon’s finding with infants:
Where would the implicit understanding be
hiding between 15 months and 3 years?
By adopting particular assumptions
about how infants encode events and behavior,
we propose two explanations for the
apparent early competence of infants that
imply an evolutionary, innate bias for understanding
the mind. Infants encode events and
behavior the way they do because this encoding
captures something useful about how
people tend to act only because people are
endowed with minds. Yet there is no need to
assume an understanding on the infant’s part
that a mind mediates a particular behavior.
Our first account of Onishi and
Baillargeon’s data is based on neuronal
activation as babies process the
events of the nonverbal false-belief
task (see the figure). Our suggestion is
that babies create three-way actorobject-location
associations. During
the familiarization phase of the test, an
adult actor watched by the infants last
observes the object (water melon
slice) in the yellow box under two conditions
(right column) and in the green
box under another two conditions (left
column). Neurons remember this
information both in an active manner
(through sustained firing in the prefrontal
cortex) and in a latent manner
(through altered firing thresholds in
nonfrontal regions) (7). If an association
of elements “actor-object-yellow
box” is still sustained in the frontal
cortex when babies are exposed to the
test stimuli, a consistent test combination
will need less processing and,
consequently, a shorter “looking” time
than a new combination of elements
(e.g., actor-object-green box). In the
latter case, babies might need longer
looking times because, when they
examine the new combination, they must
form a new association.
A similar increase in looking times may
also stem from changes in latent activation
in nonfrontal regions, where neurons code
for the recency of exposure and increase
their firing when a nonrecent stimulus is
presented. Even rats are able to represent
new arrangements of three familiar stimuli,
resulting in increased neuronal activation in
the postrhinal cortex and hippocampus and
less activation in the dentate gyrus and
subiculum (8). New arrangements of “actorobject-location
(yellow box versus green
box)” could result in longer observation
times because of the differential activation
of neurons that code for the recency of stimuli.
Both of these explanations have a clear
testable prediction that differs from explanations
based on understanding belief: The
actor’s intentional search for the object in a
box in the test phase is not critical.The actor
could do something equally interesting (but
nonintentional) at either box, and this would
also result in the same pattern of differential
looking in infants.
For our second account of Onishi and
Baillargeon’s findings, we acknowledge
their suggestion that infants expect the
observed person to act in a particular way.
However, we propose that this can be based
on behavior rules. Infants may have noticed
(or are innately predisposed to assume) that
people look for an object where they last saw
it and not necessarily where the object actually
is. Again, such a rule captures something
implicit about the mind, because the
rule only applies as a result of the mind
mediating between seeing and acting.
Nonetheless, infants can simply know the
rule without any conception that the mind is
the mediator. For instance, O’Neill (9) found
that when requesting an object, 2-year-olds
gesture more to the object’s location if a parent
had not witnessed its placement on a
shelf. This finding is compatible with a 2year-old’s
understanding of the parent’s need
for knowledge. It is also equally compatible
with a 2-year-old’s understanding of the link
between the behavior of not having looked at
the object in its new location and the likely
action of looking in the wrong place (which
needs to be prevented by gesturing to the
right location).
Povinelli and Vonk (10) recently argued
that extant evidence for the social intelligence
of primates leaves open the question
of whether they merely know about behavior
or whether they also know about the mental
states that mediate behavior. No explanatory
power or theoretical parsimony is gained by
assuming that animals know about the mind.
This criticism also applies to the traditional
verbal false-belief task (featuring Max) that
children do not pass until 4 years of age. So,
it is important to realize that claims about 4year-olds’understanding
of belief cannot be
based solely on their positive response to
this particular test. The conclusions drawn
from the false belief task are warranted only
because understanding of false belief
around 4 years can be demonstrated in a
CREDIT:PRESTONHUEY/SCIENCE,ADAPTEDFROM(1)
TB-green + FB-green
Familiarization phase Belief-induction phase
TB-yellow + FB-yellow
Last association before test
of actor + object + location
New association in test trial
of actor + object + location
Similarity to last association greater in…
Conditions
Smaller in…
Yellow-box trial
than in
green-box trial
Green-box trial
than in
yellow-box trial
Yellow-box test trial Green-box test trial
Green-box test trial Yellow-box test trial
Predicted results:
Longer looking time in:
Now you see it…now you don’t. Reanalysis of Onishi and Baillargeon’s experimental conditions
in a nonverbal false-belief task (1).The left column shows the critical events for test trial
conditions in which infants looked longer in the yellow-box than in the green-box test trials.
These two conditions have in common that the last appearance of the adult actor was during
the familiarization phase, where the actor was grasping the object in the green box (“actorobject-green
box” association).This makes it possible that the looking time by the infant subjects
was shorter in the green-box test trials—as reported (1)—because the actor-object-location
combination in this test trial is more similar to the last seen association of these three elements
than in the yellow-box test trial.The right column shows the same results for the other
two conditions, in which looking time for test trials was reversed.This can be explained by the
fact that the last appearance of the adult actor was during the belief-induction phase, where
the actor observed the object disappear into the yellow box (“actor-object-yellow box” association).
Hence, under these conditions, infants found the yellow-box test trials more similar to
the last actor-object-location combination than the green box trials, which explains the reversal
in looking times. (TB, true belief; FB; false belief)
www.sciencemag.org SCIENCE VOL 308 8 APRIL 2005
P E R S P E C T I V E S
Published by AAAS
216
variety of belief-inducing situations [in
which behavior rules would lead to contradictory
predictions of actions (11)]. Only the
assumption that children acquire an understanding
of belief at this point can explain
why they start to make correct predictions of
actions in these different situations at the
same time. Demonstration of such flexible
use of belief understanding is missing from
studies of both primates and infants (and
from studies of the implicit understanding of
3-year-olds).
Assuming that primates have a genetic
predisposition to acquiring behavior rules,
we can concoct a plausible story about
human development. Inheriting from our
evolutionary ancestors this predisposition,
infants start with a “core theory” (12)—that
is, knowledge that stays close to the perceptible.
Then, children develop a deeper mental
understanding of behavior through enculturation
into a language community. This contention
is supported by increasing evidence
that the explicit understanding of belief
around 4 years of age strongly relates to language
development. Most notably, deaf children
raised by hearing parents suffer from a
language delay of several years that is also
reflected in their late understanding of false
beliefs.Thus, we can conclude that the acquisition
of our adult theory of mind has a strong
evolutionary basis and is deepened by universal
aspects of culture, and by linguistic
communication in particular.
References and Notes
1. K. H. Onishi, R. Baillargeon, Science 308, 255 (2005).
2. H.Wimmer, J. Perner, Cognition 13, 103 (1983).
3. H. M.Wellman, D. Cross, J.Watson, Child Dev. 72, 655
(2001).
4. T. Ruffman, W. Garnham, A. Import, D. J. Connolly, J.
Exp. Child Psychol. 80, 201 (2001).
5. W.A. Clements, J. Perner, Cognit. Dev. 9, 377 (1994).
6. M. Carpenter, J. Call, M. Tomasello, Br. J. Dev. Psychol.
20, 393 (2002).
7. J. B. Morton, Y. Munakata, Dev. Psychobiol. 40, 255
(2002).
8. H. Wan, J. P. Aggleton, M. W. Brown, J. Neurosci. 19,
1142 (1999).
9. D. K. O’Neill, Child Dev. 67, 659 (1996).
10. D. J. Povinelli, J.Vonk, Mind Lang. 19, 1 (2004).
11. J. Perner, S. R. Leekam, H. Wimmer, Br. J. Dev. Psychol.
5, 125 (1987).
12. S. Carey, E. Spelke, Philos. Sci. 63, 515 (1996).
13. We thank M.Taumoepeau for helpful comments.
10.1126/science.1111656
I
n a metathesis reaction, two molecules
exchange atoms or groups of atoms. For
example, two alkynes—hydrocarbons
that contain a carbon-carbon triple bond—
may swap their substituents (see the first
figure, top panel). Such alkyne metathesis
reactions, discovered by Mortreux in 1974
(1), are very useful for synthesizing complex
natural products (2), semiconducting
polymers, and macrocyclic compounds (3,
4). On page 234 of this issue, Bino et al. (5)
shed light on how some of these reactions
are catalyzed.
Mortreux found that treating a disubstituted
alkyne with a mixture of two compounds—molybdenum
hexacarbonyl
[Mo(CO)6] and a phenol derivative
called resorcinol—led to
the formation of an “in situ”
catalyst and yielded a statistical
mixture of three alkynes, with
their substituents scrambled
(1).Variants of this catalyst system
containing more acidic
phenols have been used successfully
in the synthesis of
polymers and natural products
(1–4, 6, 7). However, the chemical nature of
the in situ catalysts remains unknown.
In an effort to understand how alkyne
metathesis reactions are catalyzed, Schrock
and co-workers (8, 9) have produced tungsten
and molybdenum carbynes of the type
(RO)3M≡C-CMe3, where M isW or Mo and
R is Me3C or aryl (see the first figure, top
panel). These complexes, which are air sensitive
and are now commercially available,
are highly active, structurally defined catalysts
for alkyne metathesis. In the case of
tungsten carbynes, the reaction proceeds
via two cyclic intermediates, which each
fragment into a metal carbyne and an
alkyne (see the first figure, bottom panel)
(8, 9). If one of the product alkynes is
removed from the reaction
mixture—for example, by distilling
off the alkyne with the
lower boiling point—the
metathesis reaction will yield
only the desired metathesis
product.
Zhang and Moore have obtained variants
of the Schrock carbynes (10, 11) by reaction
of the molybdenum complex (R2N)3Mo≡CEt
(1, 12) with 4-nitrophenol. The resulting
carbyne, (O2N-C6H4-O)3Mo≡C-Et, is the
most active alkyne metathesis catalyst
reported so far and tolerates sulfur atoms
adjacent to the catalytic center: It catalyzes
the conversion of thiophene-containing
monomers into organic semiconductors.
This is an important achievement in the field
of alkyne metathesis; none of the earlier
alkyne metathesis catalysts are able to perform
this transformation.
Nevertheless, the simpler Mortreuxtype
in situ catalysts are powerful for synthetic
applications, provided the substrates
are sufficiently robust and do not contain
nitrogen or sulfur functionalities in close
proximity to the alkyne unit. But despite
their usefulness, there is no mechanistic
understanding of this reaction. It has been
suggested that it may proceed via an intermediate
resembling the carbyne complexes
discussed above, but it remains unclear how
these carbynes would form from molybdenum
hexacarbonyl and phenols. Bino et al.
(5) bridge this chasm in understanding and
shed some light on possible intermediates
of alkyne metathesis with in situ catalysts.
Reaction of molybdenum hexacarbonyl
C H E M I S T RY
How Are Alkynes Scrambled?
Uwe H. F. Bunz
R2
2 R1
R1R1 R2R2+
(RO)3M C-R1
or
Mo(CO)
Catalyst:
6/phenol
M OR
RO OR
R1
R1
R2
M
OR
OR
OR
R1R1
R 2
+
Reaction mechanism with Schrock carbynes
General alkyne metathesis reaction
M OR
RO OR
R1
R2
R2
R2
R1
FragmentationFragmentation
M OR
RO OR
R1R2
R2
R2
R1
M
RO
RO
RO R1
Alkyne metathesis and its mechanism. (Top) Two types of compounds—the original Mortreuxtype
in situ catalysts and the carbynes pioneered by Schrock and co-workers—are known to catalyze
alkyne metathesis. (Bottom) In the case of carbynes, insights have been gained into the catalytic
intermediates. R1 and R2 are organic residues, M is tungsten or molybdenum, and RO are bulky alkoxide
or aroxide ligands.
The author is in the School of Chemistry and
Biochemistry, Georgia Institute of Technology,
Atlanta, GA 30332, USA. E-mail: uwe.bunz@chem-
istry.gatech.edu
P E R S P E C T I V E S
8 APRIL 2005 VOL 308 SCIENCE www.sciencemag.org
Published by AAAS