The Long Ashton breeding programme

Four cultivars, in four crosses. Dabinett and Michelin supplied the bittersweet side; James Grieve and Worcester Pearmain supplied acid and, in James Grieve’s case, a Scottish hardiness the West Country fruit did not have. The seedling numbers were unequal: about 500 from Dabinett × James Grieve, 650 from Michelin × James Grieve, 200 from Dabinett × Worcester Pearmain and 150 from Michelin × Worcester Pearmain.

The choice of parents says what the programme was for. Dabinett and Michelin were, and remain, the two most widely planted English cider apples, valued for reliable tannin and manageable trees rather than for character. Crossing them with sharp dessert fruit was an attempt to keep the tannin and the tree habit while adding acid — which is exactly the correction a West Country blend needs and which the traditional repertoire supplies by blending rather than by breeding.

The station closed in 2003. Its cider work had in fact ended earlier: the Pomology and Plant Breeding Division and the Food and Beverage Division were both closed in 1981, and arable crops replaced pomology as the site’s main research. The National Association of Cider Makers took over responsibility for cider pomology research in 1990, and its archive is where much of the programme’s surviving documentation now sits.

The parentage of the individual selections went missing. That sounds careless and is ordinary: a programme raising fifteen hundred seedlings across four crosses keeps its records in the working documents of the people running it, and those people retire, the division closes, and the trees continue to exist without them.

In 2019 a University of Bristol team genotyped 245 apple cultivars at 1,500 single-nucleotide polymorphism markers, drawing samples from the surviving Long Ashton-derived trees, from the John Innes Centre, from a commercial cider maker’s heritage orchards and from transplanted trees. The explicit objective was to recover the parentage of The Girls, and it did.

The result splits them into three groups rather than one. Fourteen selections group with Dabinett and James Grieve; seven with Michelin and James Grieve; two with Dabinett and Worcester Pearmain. Four more — Amelia, Connie, Eleni and Nicky — could not be resolved, because only two samples of each were available and the two disagreed, and the study says so rather than picking one.

Two findings in the paper have nothing to do with the breeding programme and are more useful than most of what the site can say about cultivar identity.

The first is a rate. Of 76 cultivars sampled in replicate from different orchards, 59 clustered with themselves and 17 did not — meaning that in nearly a quarter of cases, two trees carrying the same label were not the same fruit. That is the identity problem quantified from inside a collection, by people who went looking, rather than asserted as a general caution.

The second is a limit. The method separated every cultivar sampled except sports: Cox’s Orange Pippin and Queens Cox differed by a single marker, and Tom Putt and Red Tom Putt were 99.6 per cent identical. A sport is a mutation in a single branch propagated onward, so genetically it is the parent cultivar with a small local change, and a marker panel designed to tell cultivars apart cannot tell those apart. Genotyping settles a great deal about cider apple identity and it does not settle everything.

The third is a warning about the alternative. Two Dorset trees found in a search for old cider apples, judged the same variety because they had the same growth habit and the same juicing characteristics, turned out to share only half their markers. Identifying fruit by how it behaves is a method that fails in exactly the way it cannot detect.

Much less is known about them as fruit than as genomes, which is an unusual thing to be able to say about an apple. No juice analysis of any of the selections has been published anywhere CiderHQ has reached.

What exists is a one-page assessment from a 2005 rootstock trial at Kinnersley in Herefordshire, where fifteen selections were grown on an M9 interstock over three rootstocks and each given a line of description and a taste classification. Three Counties was assessed as a full bittersweet, Angela as a full bittersharp, Jean as a high-acid bittersweet. Gilly was poor and ugly on one rootstock and very good on the other two; Prince William was weak on all three.

That trial is worth reading as a demonstration of something separate. The same selection on three rootstocks gave three different trees, and the difference between them was larger than the difference between several of the selections. A cultivar’s habit is not a property of the cultivar alone.

The table below is where CiderHQ holds these twelve. Until Mega Sprint 7 each had a page of its own, and that was the wrong shape: with no juice analysis published for any of them, a page per selection could carry a name, a cross and one assessment word, and every other sentence on it was a sentence about the programme. Twelve pages were one page twelve times. Sorted by assessment rather than by cross they become something twelve separate pages could never be — a list a reader can compare, in which the Dabinett crosses spread across the entire bittersweet-to-bittersharp range while the Michelin crosses cluster in the middle of it. The records themselves stay in the graph, so the programme’s output remains countable and the parentage edges from Dabinett, Michelin and James Grieve are intact.

The programme is the clearest case in English cider of research changing what is planted. The selections were bred to a specification — the tannin and habit of a modern bittersweet with acid added — and several of them are grown commercially, which means the composition of English orchards today reflects a decision taken in a laboratory rather than only the accumulation of found trees.

It is also a case study in how easily institutional knowledge is lost. The trees survived, the names survived, and the single piece of information that made them scientifically legible — which cross each came from — did not. Recovering it took a genotyping study run more than fifty years after the crosses were made, and four of the selections were not recovered at all.