When did life first appear on land? The answer to this question — one of the most fundamental in science — rather depends on the values you choose for ‘life’ and ‘land’.

There is certainly evidence for freshwater life — pond life, essentially — a billion years ago or so (P. K. Strother et al. Nature 473, 505–509; 2011). Apart from that, the evidence is indirect and inferred from signs of weathering of non-marine rocks and the presence of apparent palaeosols — sediments indicative of fossilized soil that, by definition, were exposed to the air. Actual fossils that might be signs of land life in the Precambrian eon (before 542 million years ago) are exceedingly rare — or, some say, mythical.

It is a highly controversial subject, and one scientist who didn’t shy away from controversy was the late palaeobotanist Jane Gray. Starting in the 1950s, Gray argued, often vociferously, for the presence of life on land at an early date. Being female, ferocious and an advocate of an unpopular view, she didn’t get many grants. But as biologist William Shear wrote in her obituary, she “was as astute at playing the stock market as she was at interpreting fossil spores, and used her independent wealth to fund her own research” (see Nature 405, 34; 2000).

Gregory Retallack of the University of Oregon in Eugene, like Jane Gray, is unafraid to plough a contrarian furrow in this much-debated area. For many years now, he has worked on palaeosols from the Precambrian. The problem with fossil soils is that they are conventionally recognized by traces of the organisms that lived in them, especially plant roots. There lies a conundrum — how do you recognize a palaeosol in sediment that lacks plant roots? The answer comes through careful geological work, to show that the palaeosol is associated with rock formed under non-marine conditions, together with work on geochemistry and data from stable isotopes. But there might also be direct evidence in the fossil soil, in the form of carbonate nodules, sand crystals and cracks caused by desiccation or the presence of ice.

Evidence of this sort has now led Retallack to infer the presence of palaeosols among rocks from the Ediacaran period (635 million to 542 million years ago) of South Australia. One might think that soils from the latest Precambrian are hardly controversial, even by the standards of a highly charged field. But these rocks contain fairly abundant evidence for the first macroscopic life — which, until now, was widely assumed to have been marine.

Ediacaran rocks feature a wide range of large, distinctive and yet enigmatic structures, usually thought to be fossils of living creatures. Originally found in South Australia, Ediacaran fossils have since been discovered in localities as far-flung as Newfoundland in Canada, Arctic Russia and the English Midlands.

The Ediacarans, however, were a rum lot. Although clearly highly organized, their precise nature has been elusive. If they were animals, they bore little or no resemblance to any other creatures, either fossil or extant. This has led to suggestions that they were giant protists, fungi, algae, lichens or even a kind of life entirely different from anything else known and now wholly extinct.

The stately progression of life from water to land was not necessarily a single, simple narrative.

Perhaps the only point of agreement is that, whatever else they were, the Ediacarans lived on the sandy beds of shallow, sunlit seas. This is where Retallack parts company with just about everyone else, because some of his Ediacaran palaeosols are associated with Ediacaran fossils. This would mean that at least some Ediacarans lived on land, under the sky, perhaps in the manner of lichens, or microbial colonies that form soil crusts. The Ediacarans, then, would be the now-not-so-rare (and not at all mythical) creatures that first colonized the land — not just in puddles, but in soils indicative of a dry, cold desert. This is as far away as imaginable from the oceanic idyll that many have assumed for Ediacaran organisms, and have reconstructed as such in a million coffee-table books.

These conclusions are published in a paper on Nature’s website this week (G. J. RetallackNaturehttp://dx.doi.org/10.1038/nature11777;2012) and will cause sharp intakes of breath in the palaeontological community; so much so that we have commissioned a News and Views Forum to air the arguments (S.Xiao&L.P.KnauthNaturehttp://dx.doi.org/10.1038/nature11765;2012).

More work — and more science — will be the only way to validate this challenging and exciting work. But the lesson from the past is clear. Jane Gray’s advocacy of land life in the relatively recent Ordovician period (485 million to 443 million years ago), once seen as off-the-wall, is now orthodoxy. There is nothing outré in principle in the supposition that life of some sort, however humble, lived on land from a very early date: the stately progression of life from water to land was not necessarily the single, simple narrative that seems so cut and dried in retrospect. The fossil record has this irritating habit: just when everyone thinks that the narrative has been sorted out, something comes along to force the story in a new, unexpected and breathtaking direction.