Dienekes, Razib and others have already reported on this. The paper just came out this morning.
The skeletons were found in Germany, Hungary, Poland and Lithuania. They were dated to 7,000 to 8,000 years ago (these are the seven neolithic ones).
All seven neolithic samples were homozygous for the non lactase persistent allele 13910- C.
They provide a list of the individuals who worked in the lab and I was surprised by how high the frequency of this non-lactase-persistent allele was- so maybe not so easy to rule out contamination. The authors make the interesting point that contamination of ancient nuclear DNA is not as common as ancient mtDNA contamination. There are some other issues with ancient DNA (allelic dropout), but they seem to have controlled for them pretty thouroughly (I'm not the person to ask though).
Aren't the highest frequencies of 13910-T found in northwestern Europe (see possible clue here)? This is maybe where selection for this allele began (according to haplotype decay analysis - about 10,000 years ago -- with a big confidence interval-- reference in paper), and then spread to other areas in Europe? The dates of these skeletons are pretty close to the middle of the range of when selection might have began.
I just noticed this paper was received for review back in September. What's taking so long?
Absence of the lactase-persistence-associated allele in early Neolithic Europeans
J. Burger , M. Kirchner , B. Bramanti , W. Haak , and M. G. Thomas
PNAS Published online before print February 28, 2007
Abstract: Lactase persistence (LP), the dominant Mendelian trait conferring the ability to digest the milk sugar lactose in adults, has risen to high frequency in central and northern Europeans in the last 20,000 years. This trait is likely to have conferred a selective advantage in individuals who consume appreciable amounts of unfermented milk. Some have argued for the "culture-historical hypothesis," whereby LP alleles were rare until the advent of dairying early in the Neolithic but then rose rapidly in frequency under natural selection. Others favor the "reverse cause hypothesis," whereby dairying was adopted in populations with preadaptive high LP allele frequencies. Analysis based on the conservation of lactase gene haplotypes indicates a recent origin and high selection coefficients for LP, although it has not been possible to say whether early Neolithic European populations were lactase persistent at appreciable frequencies. We developed a stepwise strategy for obtaining reliable nuclear ancient DNA from ancient skeletons, based on (i) the selection of skeletons from archaeological sites that showed excellent biomolecular preservation, (ii) obtaining highly reproducible human mitochondrial DNA sequences, and (iii) reliable short tandem repeat (STR) genotypes from the same specimens. By applying this experimental strategy, we have obtained high-confidence LP-associated genotypes from eight Neolithic and one Mesolithic human remains, using a range of strict criteria for ancient DNA work. We did not observe the allele most commonly associated with LP in Europeans, thus providing evidence for the culture-historical hypothesis, and indicating that LP was rare in early European farmers.