Decentralized selection, defined as selection in the target environment, ha
s been used to emphasize favorable interactions when significant genotype b
y environment interactions exist. However, crop breeding based on decentral
ized selection can still miss its objectives if it does not utilize the far
mers' knowledge of the crops and the environment, and it may fail to fit cr
ops to the specific needs and uses of farmers' communities unless it become
s participatory. One cycle of decentralized participatory selection was con
ducted in eleven locations in Syria. 208 barley entries (fixed lines and se
gregating populations) were planted unreplicated in two research stations a
nd in the fields of nine Syrian (host) farmers, where they were managed (ex
cept planting) by the farmers themselves. Visual selection was conducted by
a breeder at all locations and by the host farmers on their own fields and
on both stations. In five farm locations, there was also a one-time group
selection by neighboring farmers. Host farmers were able to handle the larg
e number of entries making observations during the cropping season using di
fferent scoring methods. They did not use the performance of entries on sta
tion for their final selection and used a higher selection pressure than th
e breeder. In their own fields, they selected about one tenth the number of
entries selected by the breeder, while on station the farmers selected, on
average, about half the number of lines selected by the breeder. For some
broad attributes, such as modern germplasm versus landraces, selection was
mostly driven by environmental effects. Selection for other attributes was
partly environmentally driven and partly based on individual farmers prefer
ences. Selection preferences were similar for fixed or segregating populati
ons. There was wider diversity among farmers' selections in their own field
s than among farmers' selections on research stations and among breeder's s
elections, irrespective of where the selection was conducted. Larger kernel
s, higher grain yield and biomass, and taller plants (particularly in envir
onmentally stressed locations) were the characteristics most frequently use
d as selection criteria by both breeder and farmers. Entries selected by th
e farmers yielded as much, and in one case significantly more, than those s
elected by the breeder. Decentralized-participatory selection was significa
ntly more efficient in identifying the highest yielding entries in farmers'
fields than any other type of selection. There was also evidence suggestin
g that the breeder was more efficient in selecting higher yielding entries
in the research station in a high rainfall area, while the farmers were mor
e efficient in selecting under stress conditions. The results suggest that
farmers can handle selection choices among a large number of lines, and bec
ause farmers' selections are at least as high yielding as breeder's selecti
ons, it is possible to transfer the responsibility of selection to farmers
in their fields.