Information about http://www.stanford.edu/dept/soc/people/mgranovetter/documents/granignoranceknowledge_000.pdf
…
Tags: convection, density changes, drifter, drifts, fresh water, heat flux, holbrook, insta, intrusions, j phys, kilometers, kinetic energy, molecular diffusion, motions, ocean currents, panel c, salt fingers, salty water, time of arrival, water panel,Pages: 2
Language: english
Created: Wed Aug 6 14:46:25 2003
Display cached document
Page 1
Page 2
PERSPECTIVES
eter (panel E, green regions) and flow ver- ed by Holbrook et al. (3). The equatorial having a single explanation. Detailed data
tically, driven by molecular diffusion of intrusions are almost certainly affected by from the techniques reported by Holbrook
heat. Similarly, where warm, salty water salt fingers. Indeed, many characteristics et al. (3) should help to narrow down the
lies underneath cool, fresh water (panel D, of the intrusions, such as the density per- mechanisms underlying ocean mixing.
blue areas), diffusive convection causes en- turbations, suggest that fingers are driving
hanced upward heat flux (panel F). the layers. Yet, a new theory by Richards References
The fluxes from these phenomena cause and Edwards (4) shows that inertial insta- 1. A SOFAR (sound fixing and ranging) float drifts at a
depth of 1500 m while emitting coded sounds that
density changes in the intrusive layers bility could convert the kinetic energy of travel for thousands of kilometers. The time of arrival
(panel C) that can both create intrusions ocean currents into intrusive motions-- of the sounds can be used to locate the drifter.
and drive them horizontally. In this way, without the need for salt fingers. And May 2. L. Armi et al., J. Phys. Oceanogr. 19, 354 (1989).
3. W. S. Holbrook, P. Páramo, S. Pearse, R. W. Schmitt
layers as thick as the height of a building and Kelley (5) have found that the potential Science 301, 821 (2003).
are moved hundreds of kilometers by struc- energy due to sloping density layers can al- 4. K. J. Richards, N. R. Edwards, in preparation.
tures that are as thin as an upraised arm. so drive intrusions. 5. B. D. May, D. E. Kelley, J. Phys. Oceanogr. 27, 1997
However, the story is more complicated As the oceanographer Rob Pinkel once (1997).
6. K. J. Richards, H. T. Banks, J. Geophys. Res. 107, 3231
than this, underscoring the need for the de- said, having multiple explanations for a (2002); published online 26 December 2002
tailed visualization techniques demonstrat- phenomenon is not an improvement over (10.1029/2001JC000971).
SOCIAL SCIENCE
day life, keep our network knowledge
small and nonrandom. Even prolific and
Ignorance, Knowledge, and determined "networkers" cannot hope for
more.
Outcomes in a Small World Yet, enduring fascination with the
"small world" phenomenon rests on the
perception that we are sharply impacted by
Mark Granovetter the length of network paths. Short paths
may bring their endpoints
hirty-six years ago, Stanley Milgram new and better jobs (3) or
T concluded from his "small world" ex-
periment that any individual is con-
nected to any other through a short chain of
political favors, but also may
lead to infection with a dread
disease or the unwanted ex-
social ties, the average chain length being posure of illicit or shameful
six people (1). On page 827 of this issue, acts. The "small world" task
Dodds, Muhamad, and Watts (2) provide of launching a message to-
the first large-scale replication of ward a target through an ac-
Milgram's experiment. Their analysis of quaintance is a clever way
60,000 e-mail users--who were asked to to estimate actual chain
reach 1 of 18 target persons in 13 countries lengths. But it offers limited
by forwarding messages to acquaintanc- evidence of how much peo-
es--suggests that Milgram's surprising re- ple actually know about their
sults are robust. This excellent new study networks, and how they nor-
raises but cannot resolve the important mally use them.
question of how much people know about Dodds et al. suggest that
their own social networks and why this in the "small world" task "all
matters. targets may in fact be reach-
Milgram named the "small world prob- able from random initial
lem" after the obligatory cocktail party re- senders in only a few steps,"
sponse of strangers who unexpectedly dis- but that small differences in
cover that they share an acquaintance. incentives and confidence
Thus, the very name of the phenomenon may produce large differ-
alludes to surprise at and ignorance of ences in completion rates.
one's social network. Indeed, Milgram Given the likely high level of
liked to recount that before his results be- aration" has made the naiveté that amused network ignorance, few can know the most
came well known, he would ask generally Milgram rare. But it does not follow that efficient way to complete a chain. Thus,
sophisticated audiences to guess how many people now know more about their social observed chain lengths are only an upper
personal acquaintance links were required networks. Just 500 acquaintances require bound on actual minima. We can expect re-
to connect randomly chosen endpoints in more than 100,000 bits of information to spondents to fashion strategies from the
the United States. Many guessed in the track who knows whom, and if each ac- very limited information offered to them
hundreds; hardly any imagined the half- quaintance knows 500 others, there can be about targets. If occupation and location
dozen or so of his and later experiments. as many as a quarter of a million people at are the most salient part of this informa-
Recent attention to "six degrees of sep- one remove. Overlap in ties reduces these tion, it is not surprising that, as Table 2 of
CREDIT: JOE SUTLIFF
numbers but introduces structural com- the Dodds et al. paper shows, respondents
The author is in the Department of Sociology,
plexity that is equally if differently daunt- channel the message through contacts who
Stanford University, Stanford, CA 943052047, USA. ing. Limitations of cognition and time, not match the target on these characteristics.
E-mail: mgranovetter@stanford.edu to mention competing obligations of every- This produces the appearance of a network
www.sciencemag.org SCIENCE VOL 301 8 AUGUST 2003 773
PERSPECTIVES
where reachability depends on people as- cause such individuals are considerably while chain length is important, informa-
signing nodes to "identity categories" more likely than others to be chosen in tion reaches prospects who did not seek it
[compare (4)], but is not good evidence network relations (6). Thus, the data in about one out of three cases--more for
that these cues are the most salient in may seriously understate the number of better jobs (3). Though people contract
everyday searches, where different, more hubs, making it difficult to assess the diseases when network distance to the al-
complex and subtle part they play in social ready-infected is short, few searched their
information may be searches. More gener- networks to achieve this outcome. Chain
available. ally, close analysis of length and knowledge are critical, but the
We should also in- "Small-world studies the characteristics of search complicated, when chain end-
terpret with much cau- the specific nodes that points have opposing goals, as when
tion the conclusion of
offer tantalizing leads ended failed chains, searches are actively resisted by crimi-
Dodds et al. that hubs about connectivity in comparison to spe- nals, mob informers, missing deposed
(nodes with many cific nodes that reached dictators or terrorist icons. Small-world
links to others) are and processes in intended targets, might studies offer tantalizing leads about con-
less relevant to social tell us something be- nectivity and processes in natural net-
searches than has been natural networks..." yond what can be works, but then need confirmation and
suggested by Barabasi learned by treating all enrichment by studies that emerge from
(5) and others. This nodes in all chains the laboratory to track actual networks.
conclusion relies on equally. As in other branches of science, progress
identifying hubs by respondents' reports The broadest issue this research raises in understanding requires that tightly con-
of whether they chose a particular ac- is what chain-length estimation experi- trolled experiment and real-world com-
quaintance to send on the message be- ments tell us about natural social process- plexity regularly and systematically in-
cause that person had many friends. But es. They suggest the need to extend our form one another.
when giving a single response justifying study to a wide range of situations where
one's choice of contact, location and occu- network chain length actually matters. This References
pation may seem more salient than num- requires far more information than we now 1. S. Milgram, Psych. Today 2, 60 (May 1967).
2. P. S. Dodds, R. Muhamad, D. J. Watts, Science 301, 827
ber of ties. Moreover, peoples' informa- have on what people know about their net- (2003).
tion about how many contacts one's works, how they use this knowledge during 3. M. Granovetter, Getting a Job: A Study of Contacts
friends have is generally poor. Yet, it is searches, and how chain length impacts in- and Careers (Univ. of Chicago Press, Chicago, 1995).
4. D. J. Watts, P. S. Dodds, M. E. J. Newman, Science 296,
likely that the choices respondents made dividuals even when the search is absent, as 1302 (2002).
were typically for individuals with a is common. Studies of how people find 5. A-.L. Barabasi, Linked (Perseus, Cambridge, MA, 2002).
greater than average number of ties, be- jobs through social networks show that 6. S. L. Feld, Am. J. Sociol. 96, 1464 (1991).
DEVELOPMENT
naling may control tissue sculpting through
selective cell death and survival.
Longing for Ligand: Hedgehog, To initiate the cell death program, pro-
apoptotic dependence receptors require
Patched, and Cell Death preliminary cleavage of their intracellular
domain (at the DXXD site) by caspase en-
zymes. Thibert et al. present several lines
Isabel Guerrero and Ariel Ruiz i Altaba of evidence indicating that Ptc1 is a
proapoptotic dependence receptor. They
ependence on developmental signals, Thibert et al. (1) provide evidence suggest- show that overexpression of Ptc1 in cul-
D like dependence on love, can have
catastrophic effects. Being unloved
or lost in a multicellular organism can lead to
ing that the Patched1 (Ptc1) receptor is a de-
pendence receptor that induces programmed
cell death during chick neural tube develop-
tured cells induces apoptosis, which is
blocked by addition of Shh. In the develop-
ing chick neural tube, removal of the ven-
self-inflicted death that people call suicide ment in the absence of its ligand, the signal- tral source of Shh causes massive cell
and cells call apoptosis. But how do lost ing molecule Sonic hedgehog (Shh). death, which is rescued by expression of a
cells recognize that they lack direction from During neural development in verte- dominant-negative form of Ptc1 that inter-
developmental signals? That job may belong brates, Shh is produced first by the noto- feres with the proposed function of wild-
to ligand dependence receptors that induce chord and later by the floor plate (see the fig- type Ptc1 in apoptosis. Cleavage of Ptc1 by
cells bearing them to undergo apoptosis if ure). A ventrodorsal gradient of Shh directs caspase-3 exposes a carboxyl-terminal
the receptor remains unoccupied by ligand. ventral patterning and cell differentiation apoptotic domain. Transfecting cultured
These receptors are thought to ensure the [reviewed in (2)]. The Thibert et al. results cells with the carboxyl-terminal region of
survival of cells that remain close to the provide insight into how the neural tube is Ptc1 is sufficient to induce cell death. In
source of the appropriate developmental sig- shaped during development. In multicellular this region, there is a conserved aspartic
nal (the ligand), and the death of those cells organisms, cells that are poorly positioned as acid residue in human, mouse, and chick-
that do not. On page 843 of this issue, a result of developmental errors can be elim- en; mutation of this site (D1392N) in
inated because of failure to receive cues in- mouse Ptc1 prevents apoptosis when this
I. Guerrero is in the Centro de Biología Molecular structing them that they are in the correct lo- receptor is unoccupied by ligand.
"Severo Ochoa," CSIC-UAM, Universidad Autónoma de cation [reviewed in (3)]. Given the wide- Transfection of cultured cells with Ptc1
Madrid, Madrid E-28049, Spain. E-mail: iguerrero@
cbm.uam.es A. Ruiz i Altaba is in the Skirball Institute,
spread importance of hedgehog (Hh) signal- truncated at the caspase-cleavage site in-
New York University School of Medicine, New York, NY ing during development, the results of duces apoptosis that cannot be rescued by
10016, USA. E-mail: ria@saturn.med.nyu.edu Thibert et al. suggest how Ptc1 and Shh sig- addition of Shh.
774 8 AUGUST 2003 VOL 301 SCIENCE www.sciencemag.org