Technical CommentsPaleoanthropology

Response to Comment on “U-Th dating of carbonate crusts reveals Neandertal origin of Iberian cave art”

See allHide authors and affiliations

Science  12 Oct 2018:
Vol. 362, Issue 6411, eaau1736
DOI: 10.1126/science.aau1736


Slimak et al. challenge the reliability of our oldest (>65,000 years) U-Th dates on carbonates associated with cave paintings in Spain. They cite a supposed lack of parietal art for the 25,000 years following this date, along with potential methodological issues relating to open-system behavior and corrections to detrital or source water 230Th. We show that their criticisms are unfounded.

Slimak et al.’s (1) supposed ~25,000-year (25-ka) hiatus in the production of parietal art comes from a misunderstanding of the logic of working with minimum ages. Our results (2) cannot be taken to imply the existence of such a hiatus. The minimum age of 45.9 ka for Ardales ARD16 and the minimum-maximum pair of 32.1 and 63.7 ka for ARD08, 09, and 06 bound painting episodes that could fall within Slimak et al.’s “hiatus,” as could the El Castillo red disk dated to before 40.8 ka ago (3). Indeed, if dates older than 65 ka are excluded, the hundreds of minimum ages we have obtained are all consistent with dates in the 40- to 65-ka interval for the stratigraphically associated paintings. The origin of the red pigment at Ardales is also questioned, but its anthropogenic nature is backed by more than a century of research (46), and with careful inspection it is even possible to recognize technical characters linked to the execution processes used.

Slimak et al.’s methodological objections relate to (i) open-system behavior, (ii) nonradiogenic 230Th in source water, and (iii) detrital contamination corrections. These topics have formed the focus of discussion in previous publications (7, 8) and are thoroughly assessed in (2).

As Slimak et al. acknowledge, we use a sequential sampling methodology to test for open-system behavior. When dates for subsamples are in correct stratigraphic order (i.e., from younger to older systematically from the “outside” of a crust inward toward the pigment), we can be confident that the carbonate has remained a closed system. In an open system, preservation of the chronological order of subsamples is highly unlikely. We have published multiple sequences of three or more subsamples with ages in the expected stratigraphic order, including examples from all three caves under consideration; open-system behavior is not an issue here.

Concerning nonradiogenic 230Th entering the carbonates from the source water, we have dated samples from all three sites to the very recent past, i.e., ~1 ka (e.g., PAS35a and -c) (2). This is entirely inconsistent with the hypothesis of high 230Th drip water; dates as young as ~1 ka cannot be obtained by U-Th if the drip water has a high 230Th content.

When considering detrital contamination corrections, it is true that La Pasiega PAS34c has large uncertainties due to the detrital Th correction; this was discussed at length in the supplementary materials of (2). There, we demonstrated that the chosen correction factor is appropriate by looking at the 234U/238U activity ratio, (234U/238U)A, which is also affected by the detrital correction. A higher detrital (238U/232Th)A yields an initial (234U/238U)A for PAS34c inconsistent with all other samples from this cave. Slimak et al. propose using such an elevated detrital value, yet make no attempt to explain the effect this would have on the (234U/238U)A. Furthermore, even if PAS34c is disregarded, PAS34a and PAS34b provide a minimum age of 53.0 ka, which still implies pre–Upper Paleolithic painting activity.

Slimak et al. argue for a younger minimum age for PAS34 based on an isochron derived from our results. However, deriving an isochron from three data points is not scientifically sound; a minimum of five would be needed. Furthermore, the assumption that these types of crust form within a short time is unsupported by previous results (2, 7). A hypothetical example for flowstone dated sequentially by U-Th shows how misleading Slimak et al.’s pseudo-isochron is (Fig. 1A). The suite of dating results shows a 6-ka-long hiatus in growth at 40 mm, and the sample just below the hiatus is more contaminated than the two above. If we follow the same approach as Slimak et al. and use only three data points, one just below the hiatus and two above, to derive an “isochron” (i.e., assuming that all are of similar age and that the difference in detritus is the reason for the age difference), then we obtain an age of 36 ± 3 ka and a detrital (238U/232Th)A of 5.7 ± 0.5 (Fig. 1B). This age is clearly wrong for the sample below the hiatus, and the very high detrital correction is largely a result of the faulty assumption that the samples are coeval. The pseudo-isochron is biased by the pair of younger samples, which coincidentally are less contaminated, exactly as is the case for PAS34. Unless Slimak et al. can demonstrate that PAS34a, -b, and -c are contemporary, their approach is inappropriate.

Fig. 1 A hypothetical example of erroneous isochron dating.

(A) U-Th dating results across a 5-cm section of a flowstone (where distance is measured from the base). The results clearly reveal a 6-ka-long growth arrest at 40 mm. Dating results before and after the arrest are supported by the full dataset. (B) The three gray data points in (A) are used to obtain an Osmond-type pseudo-isochron, as done for PAS34a, -b, and -c by Slimak et al. The isochron gives an age of 36 ± 3 ka, which is clearly too young. Error bars denote 2 SD.

All carbonate samples will be contaminated by detrital Th to some degree, and the threshold of reliability based on measured (232Th/238U)A or (232Th/234U)A that Slimak et al. suggest is entirely arbitrary. Of more importance is the sensitivity to the applied correction of the resulting corrected age. Figure 2 shows corrected ages and (232Th/234U)A for all the published Ardales and Maltravieso data (2) using our detrital (238U/232Th)A values and elevated ones. It is apparent that there is no clear positive correlation between age and (232Th/234U)A for either site, and the dates are relatively insensitive to the detrital correction. Critically, this means the shift in the two sets of corrected ages is not critical to our conclusion that some of the art is Neanderthal.

Fig. 2 Corrected ages and (232Th/234U)A for carbonate samples associated with art (i.e., as maximum or minimum ages).

(A) Ardales cave; (B) panel GS3b in Maltravieso cave (2). Error bars denote 2 SD.

For Ardales (Fig. 2A), even with an unrealistic (238U/232Th)A value of 5, ARD13b still gives a minimum age of 59.0 ka. A highly unrealistic detrital (238U/232Th)A value of ≥11 is required before the corrected age of this sample is on the order of the ~47 ka that Slimak et al. prefer. When applying detrital Th corrections to relatively clean samples such as those from Ardales, using the bulk-earth value of (238U/232Th)A with a conservative error is adequate, and our applied detrital corrections are robust.

The samples from Maltravieso are characterized by higher detrital Th; thus, extra effort was made to characterize the detrital component directly. Sediment from the cave was collected and analyzed as a proxy for the samples’ detrital fraction. A speleothem column was also sampled and a series of six growth layers dated to provide a control for this sediment-derived correction (2). In Fig. 2B, two detrital (238U/232Th)A values are used: 3.3 ± 0.2 (i.e., the sediment-derived correction) and an elevated value of 4 ± 2. Detrital (238U/232Th)A values of ≥4 are not possible, as beyond this limit the equivalent measured (230Th/232Th)A of one of the samples is exceeded. Shifting the detrital (238U/232Th)A to 4 ± 2 has very little effect on the corrected ages, giving, for example, a minimum age of 64.9 ka (instead of 66.7 ka) for MAL13a. This sample does contain a notable detrital component, but not enough to critically affect the corrected age. Finally, Slimak et al. cast further doubt by incorrectly claiming that the proposed Middle Paleolithic age of the Maltravieso hand stencil is based on a single sample. On the contrary, it is supported by a second sample, MAL17d (Embedded Image ka).

On the basis of present evidence, the most likely scenario is that in Europe, parietal art emerged prior to 65 ka ago and continued, perhaps episodically, throughout the remainder of the Paleolithic. Slimak et al.’s speculation that two technocomplexes dated to ~50 ka ago—the Bohunician and the Neronian—are possibly associated with modern humans sheds light on their willingness to accept a minimum age of 47 ka but not older. Their speculation is groundless. The earliest remains of modern humans in Europe, the Oase fossils from Romania, date to ~40 ka ago, and Neanderthal remains directly dated as recently as 40 to 50 ka are known across all of the then-inhabited Europe, east to west and north to south (9). There is no escaping the conclusion that these temporal patterns imply Neanderthal authorship of Europe’s earliest cave art.

References and Notes

Acknowledgments: Supported by Natural Environment Research Council (UK) grant NE/K015184/1, National Geographic Society grant EC0603-12, the Max Planck Society, a Royal Society Wolfson Research Merit Award (A.W.G.P.), and Research Group IT622-13 of the Basque government (M.G.-D.).

Stay Connected to Science

Navigate This Article