Earth exploration still has the power to astound. Mid-ocean ridges, also called spreading centres, produce new ocean crust at different rates and are classified accordingly. Dick and colleagues (page 405 of this issue1) show that ridges exhibit fascinating behaviour when the spreading rate is 'ultraslow', less than about 1.2 cm yr−1. The ridges they have investigated, the Gakkel and southwest Indian ridges, have long, quiescent stretches and oblique, non-volcanic, spreading segments that further differentiate them from their faster-spreading relatives.

Mid-ocean ridges were first intensively studied in the late 1950s, when it became evident that the Mid-Atlantic Ridge has the same form as the coastlines of South America, North America and Africa — almost as if all of the sea floor between these continents had been created by volcanism along a narrow ridge (which indeed it was). By 1972, it was believed that mid-ocean ridges are the sites of 90% of Earth's volcanic activity, and spreading rates were found to range typically between about 2 and 18 cm yr−1. A further observation was that of a remarkable transition in ridge morphology at the intermediate spreading rate of around 6 cm yr−1: the 'median valley' (2–3 km deep, 20–30 km broad) typical of slower rates changes to a structure some 200 m high and 2 km broad. The ocean crust created by volcanism at all well-studied ridges was found to be a fairly uniform 6.5 km thick2,3, except near volcanic hotspots such as Iceland, where it could be twice that value.

Also noted (but still acknowledged to be less well understood) was that ridges typically form straight, volcano-rich segments where basalt rocks are erupted and intruded between the spreading plates. These segments are oriented perpendicular to the direction of plate spreading, and are offset by features known as transform faults, where the two plates slip sideways past each other (the San Andreas Fault is the most familiar example of a transform fault, albeit in an atypical continental context). In the ocean basins, transform faults typically lie within a rugged 'fracture-zone valley' that is strikingly similar in width and relief to the median valley seen at slower-spreading ridges.

This overall picture was refined in the 1980s and 1990s, leading to better quantification of mid-ocean-ridge dynamics. But many of these studies focused on details of ridge segmentation, the magmatic plumbing of ridge volcanic centres, and the hydrothermal activity and associated organisms that inhabit these active deep-sea sites. Little attention was paid to the slowest-spreading ridges, largely because they are especially tough to study. For instance, the Gakkel ridge lies beneath the Arctic Ocean, investigations in which often require the assistance of ice-breakers. And the southwest Indian ridge underlies the southern oceans, halfway between southern Africa and Antarctica, which experience some of the most extreme weather in the world.

These logistical issues also mean a higher cost for field research, which makes it more difficult to justify research proposals within the modern peer-review funding process. Geological 'fishing' expeditions went out of fashion in the late 1970s in favour of hypothesis-driven science in more accessible regions. So even though there was evidence in the early 1980s that the ocean volcanic crust appeared to be much thinner than normal at an ultraslow-spreading ridge4, this intriguing observation was largely neglected. As a recent burst of publications5,6,7 attests, however, ultraslow ridges are at last being subjected to modern marine field research. The paper by Dick et al.1 is the latest in line, and documents striking contrasts between such ridges and their faster-moving cousins.

There are two especially notable differences. One is the apparent absence of volcanic activity on long stretches of ridge (for example, there is no apparent volcanism along an 80-km section of the Gakkel ridge). The other is the frequent occurrence of oblique, 'amagmatic' segments instead of transform faults. The existence of these segments, which are nonetheless producing ocean floor and spreading, is a challenge to current theories of how rock melts in the Earth's mantle and migrates beneath ridges to create observed patterns of volcanism.

If melting has indeed ceased beneath these segments of the ridge, there are three possible explanations. First, that melting is taking place at depths of less than about 25 km, which can be cooled by heat conduction to the surface at ultraslow upwelling rates, instead of the 25–60-km melting layer inferred from the chemistry of mid-ocean-ridge basalts. Second, that the mantle temperature beneath amagmatic segments is anomalously cool, and so fails to produce melts (an idea proposed to explain the occasional absence of melting at so-called 'non-volcanic rifted margins'8). Third, that the temperature at which this particular mantle material would melt is anomalously high. All of these possibilities seem strange. But they are testable.

The last two would also imply that there is correlation between lack of melting and ultraslow spreading, and that ridges that don't melt have a greater viscous resistance to spreading apart — an idea in apparent conflict with the well-accepted concept that ridges are a passive response to the separation of tectonic plates. Alternatively, the absence of volcanism could be a by-product of the lateral migration of melts in the mantle beneath the ridge, for at least 40 km. If this is so, then the mantle rocks collected along these segments should show evidence that they underwent the same amount of melting as mantle rocks dredged from faster-spreading ridges. Erupted basalts, where found, should also look similar to basalts erupted at faster-spreading ridges — and they do not, as Dick et al.1 discuss.

What about the presence of oblique, amagmatic spreading segments at ultraslow ridges? The contrast with transform faults could be a telling observation. It may finally allow us to figure out why transform faults are so common at ridges, and why they are almost invariably found within a rugged fracture-zone valley. The best partial answer I've heard to the question 'Why fracture zones?' is a suggestion9 put forward almost 30 years ago — that such zones form in response to contraction, parallel to the ridge, induced by the cooling of the new plate as it ages and moves away from a ridge. At ultraslow, oblique spreading segments, perhaps sea water can penetrate along faults to a much greater depth than at other, faster-spreading ridges? The consequent reaction between water and mantle rocks would transform high-density mantle into serpentine, a rock type of much lower density, that could 'fill' the component of plate contraction along the ridge axis.

Answers to the geological puzzles raised by ultraslow ridges are still unclear. But it's a sure bet that such ridges will be getting a lot more attention over the next few years.