children could easily have been concerned with what the adult experimenters — who stayed in the room as the children decided — thought of them as cooperators.

A notable result was a change to other-regarding behaviour with age: younger chil-dren (3–4 years old) were mostly self-centred, whereas the older children (7–8 years old) were concerned with equality — within their group, that is. Interestingly, this developmental pattern is different from that of another form of altruism, namely ‘instrumental helping’. Human infants as young as 14–18 months readily help others do such things as fetch or stack objects, or open cabinets7. For this kind of helping there is direct evidence that infants are not influenced by external rewards, and indeed our nearest primate relatives sometimes help others achieve their goals as well7,8. So, in the case of instrumental helping, the develop-mental trajectory is of early altruism and other-regarding preferences, and then, presumably, growing concern with reciprocity and reputa-tion as children learn to be selective altruists so as to avoid being exploited by others.

One difference between these studies on instrumental helping and Fehr and colleagues’ research is that, in the former case, children were faced with a real individual, not pictures, and this might influence their altruism. That is, in more natural settings, where resources are being distributed among individuals who are physically present, young children might very well start out being altruistic as well. But it is also possible that altruism is not a general trait, and that there are actually two different evolutionary scenarios involved here: one for helping behaviourally and one for sharing resources. Young infants may thus be helpful behaviourally because it involves only energetic and opportunity costs. But being generous with food is another story, which might be governed especially strongly by such things as expecta-tions of reciprocity, adult encouragement and social norms.

Fehr et al.3 thus contribute to a grow-ing collection of direct investigations of the tendency of human beings to care about and act on behalf of others. Their paper shows that, in the case of sharing resources equitably, the tendency emerges in middle childhood. It is also significant that this tendency, even in childhood, is directed mostly at others within the group, for much evidence now suggests that human cooperation is at root parochial9. In the end, research of this kind should help us to determine the degree to which human social life, morality and culture derive from a unique other-regarding psychology that emerges early in life. ■

Michael Tomasello and Felix Warneken are in the Department of Developmental and Comparative Psychology, Max Planck Institute for Evolutionary Anthropology, Deutscher Platz 6, D-04103 Leipzig, Germany. e-mails: tomasello@eva.mpg.de; warneken@eva.mpg.de

EARTH SCIENCE

A sheet-metal geodynamoUlrich R. Christensen

A decade of modelling Earth’s core on computers has led to the belief that we understand what produces Earth’s magnetic field. More realistic simulations are now shaking that complacency.

Earth’s magnetic field is often depicted as though a gigantic bar magnet resides inside our planet. In fact, the magnetic field is created by strong electrical currents gener-ated by a dynamo process resulting from the flow of molten iron in Earth’s outer core. An important feature of this motion is thought to be spiralling flows in the form of long columns running parallel to Earth’s axis. Despite numerous simplifying assumptions, previ-ous computer simulations of Earth’s core have reproduced both this kind of flow and the properties of the magnetic field remarkably well. On page 1106 of this issue, Kageyama et al.1 present a simulation that should more closely model the conditions in Earth’s core. Instead of columns, the model shows thin sheets, which undermines our present thoughts on exactly how the geodynamo works.

In a simple dynamo, as might power the lights on a bicycle, the motion of a wire through a mag-netic field generates electrical current in that wire. In more sophisticated, technical dynamos found in cars or power stations, the magnetic field is generated by the dynamo’s own current rather than being externally applied. These dynamos work because the current is guided by wires arranged in coils. But Earth’s core is an unstructured sphere of iron, which is liquid in its outer part. A simple rotational motion of the core would be insufficient to produce a magnetic field. Rather, the liquid metal must follow complex paths reminiscent of the coils in a technical dynamo.

The flow of material in the outer, liquid, core is produced by convection, driven mainly by a flux of heat and buoyant elements (such as silicon and sulphur) that emanate from the inner, solid, core. The flow is strongly influ-enced by forces due to Earth’s rotation — such as the Coriolis force, which causes low- pressure weather systems to rotate anticlockwise in the Northern Hemisphere, and clockwise in the Southern Hemisphere. As long as convection is weak, the flow takes the form of columns, aligned north–south, through which fluid particles move with a corkscrew-like trajectory

(Fig. 1). Theoretical work has shown that this situation could generate a magnetic field with a dipolar (bar-magnet-like) structure2. Previ-ous computer simulations of the geodynamo that solved the equations of fluid flow and magnetic induction in a rotating sphere sup-port the existence of such columns and their role in the generation of magnetic fields3,4. They show magnetic field lines emerging at the top of Earth’s core mainly in so-called flux lobes, which appear in pairs aligned on the same meridian line at around 60° latitude in the Northern and Southern Hemispheres. The locations of the lobes are thought to be places where individual convection columns touch the outer boundary of the core5.

All computer models of the geodynamo suffer from being unable to run with true core condi-tions, because this would require much more powerful computers than are currently available. For example, the viscosity of the liquid core must be made unrealistically large to suppress small-scale eddies in the flow that cannot be resolved on present computers. A useful measure of vis-cosity here is the Ekman number, which is the ratio of viscous friction to the Coriolis force. The Ekman number in the core is tiny, around 10−15, but until now geodynamo simulations have only been able to operate down to Ekman numbers of 10−6, with larger values being used in most cases. Kageyama et al.1 used 4,096 processors at the Earth simulator in Japan to run a vigor-ously convecting dynamo model with an Ekman number approaching 10−7.

This may seem a small step, given the discrepancy remaining — eight orders of magnitude — but such efforts are important to discover whether simulations with high Ekman numbers produce qualitatively similar results to those with more realistic conditions, or whether lowering the Ekman number results in marked alterations. It was expected that, at lower Ekman numbers, the vortex columns would become narrower and more numerous. Instead, Kageyama et al.1 find that the flows become flat-tened into thin sheets radiating out from Earth’s axis, reminiscent of the gills of a mushroom

1. Silk, J. B. et al. Nature 437, 1357–1359 (2005).2. Jensen, K., Hare, B., Call, J. & Tomasello, M. Proc. R. Soc.

Lond. B 273, 1013–1021 (2006).3. Fehr, E., Bernhard, H. & Rockenbach, B. Nature 454,

1079–1083 (2008). 4. Hook, J. G. & Cook, T. D. Psychol. Bull. 86, 429–445 (1979).5. Boehm, C. Hierarchy in the Forest (Harvard Univ. Press,

1999).

6. Haley, K. J. & Fessler, D. M. T. Evol. Hum. Behav. 26, 245–256 (2005).

7. Warneken, F. & Tomasello, M. Science 311, 1301–1303 (2006).

8. Warneken, F., Hare, B., Melis, A., Hanus, D. & Tomasello, M. PLoS Biol. 5, e184 (2007).

9. Bernhard, H., Fischbacher, U. & Fehr, E. Nature 442, 912–915 (2006).

1058

NATURE|Vol 454|28 August 2008NEWS & VIEWS

ba

radiating away from its stem. Similar patterns have been observed in laboratory convection experiments6. If the flow in Earth’s core has the form of elongated sheets rather than columns, the interpretation of magnetic flux lobes at the top of the core must be reconsidered.

Kageyama and colleagues’ simulations1 show that a laminated flow pattern still acts as an efficient dynamo. The conventional way of visualizing dynamo simulations is to map the resulting magnetic field lines. In complex dynamos the result often looks like a bowl of spaghetti and is hard to interpret. Kageyama

et al. instead map lines of electrical current to produce a more coherent image, showing a large number of structures similar to the coils used in electrical engineering. However, unlike the Earth, this model produces a magnetic field that is not predominantly dipolar.

It is disturbing that dynamo models with grossly simplified parameters reproduce Earth’s field well, whereas this agreement degrades when conditions are made more realistic. Some proposed explanations for this discrep-ancy highlight the problems encountered when performing simulations at the edge of what is

currently possible. Even though the simulation took many months to run, it covered only 2,000 years of modelled time. If the flow in Earth’s core were to stop suddenly, the dipole field would take 20,000 years to decay and so a similar time scale might be needed to create the field from scratch, requiring computer runs of years rather than months. In addition, the Ekman number is not the only parameter that has had to be set at unrealistic levels in geo dynamo models. For example, the ratio between viscosity and elec-trical resistivity (magnetic Prandtl number) is set to 1 in simulations, despite being about 10−6 in the core.

It will remain impossible to model the geodynamo at true core conditions for the fore-seeable future. The recipe for getting models with an Earth-like field might therefore be to have the parameters ‘wrong in the right propor-tions’. The question is then what relationships between parameters must be kept fixed as we approach more realistic values. Kageyama et al.1 have penetrated territory previously inacces-sible to simulation. The next step will be even more demanding: scouting this new terrain to identify the true path through parameter space to understanding. ■

Ulrich R. Christensen is at the Max Planck Institute for Solar System Research, Max-Planck Strasse 2, 37191 Katlenburg-Lindau, Germany.e-mail: christensen@mps.mpg.de

1. Kageyama, A., Miyagoshi, T. & Sato, T. Nature 454, 1106–1109 (2008).

2. Busse, F. H. Geophys. J. R. Astron. Soc. 42, 437–459 (1975).3. Kageyama, A. & Sato, T. Phys. Rev. E 55, 4617–4626 (1997).4. Olson, P., Christensen, U. & Glatzmaier, G. A. J. Geophys.

Res. 104, 10383–10404 (1999).5. Gubbins, D. & Bloxham, J. Nature 325, 509–511 (1985).6. Sumita, I. & Olson, P. Phys. Earth Planet. Inter. 117,

153–170 (2000).

SYSTEMS BIOLOGY

Reverse engineering the cellNicholas T. Ingolia and Jonathan S. Weissman

Borrowing ideas that were originally developed to study electronic circuits,two reports decipher how yeast reacts to changes in its environment by analysing the organism’s responses to oscillating input signals.

Systems biology has certainly caught people’s attention. But when pressed to define precisely what it is, let alone how it will complement clas-sic reductionist approaches, concrete answers can be hard to find. One immediate, practi-cal contribution of systems biology is that it suggests new ways of interrogating cellular responses to stimuli, and provides a framework for understanding the results1. In this spirit, two papers2,3 now borrow ideas from the theory of system identification — a formal strategy for building mathematical models of dynamic sys-tems based on experimental data — to examine the response of yeast cells to changes in their

environment. In both cases, the authors use microfluidics to control precisely the external stimuli and to quantify the cells’ responses. The authors thus unravel some of the ‘wiring’ of the signal transduction networks that monitor these environmental changes.

Studies of gene regulatory networks typi-cally measure the way in which cells respond to strong perturbations, such as changes in tem-perature or abrupt variations in nutrient con-centration. Such marked changes are valuable for inducing signalling pathways and for iden-tifying the molecules involved, but are not likely to be representative of physiological conditions.

Figure 1 | Alternative geodynamos. The magnetic field is generated in Earth’s liquid outer core. a, In the conventional concept, the molten iron circulates along a spiralling path in columns aligned in the north–south direction, generating electrical currents that set up the dipolar magnetic field. The concentration of field lines into anticyclonic vortices (rotating in the same direction as air around a region of high pressure) has been thought to explain the intense magnetic lobes found in Earth’s field at the top of the core. b, The computer simulations of Kageyama et al.1 show that the flow pattern in the core may bear more resemblance to a set of thin sheets, which puts the conventional concept into question. The sheet-like flow is efficient at generating a magnetic field, but how it might lead to the dipole geometry of Earth’s field is not yet clear.

At worst, they might even give a distorted view of the normal function of the pathways. For-tunately, more nuanced approaches can help overcome these problems.

Engineers know that certain dynamic systems, such as electronic circuits, can be stud-ied by measuring their response to stimuli that vary with a sinusoidal pattern4. In this approach, the response of the system is measured at differ-ent frequencies of periodic signals; well-devel-oped mathematical methods are then used to convert this ‘output’ into a model of the inner workings of the system. The frequency depen-dences of two quantities are particularly infor-mative: the amplitude of the response and the lag (phase difference) between the input stimu-lus and the output behaviour (Fig. 1, overleaf). This approach has previously been used to study bacterial movement in response to different frequencies of pulses of a chemical attractant5. But broader biological applications of the strat-egy have had to wait for technical advances that are only now coming into play — such as microfluidics and fluorescent cell reporters.

Enter Mettetal et al.3, who report in Science

1059

NATURE|Vol 454|28 August 2008 NEWS & VIEWS