Fifty-seven isolates of fungi of the Gaeumannomyces-Phialophora complex that had been maintained in stock culture were tested by host-infection (in vivo) and agar plate, filterpaper and flask culture (in vitro) methods for production of perithecia. Most isolates of each of the three varieties of G. graminis produced perithecia in one test or another ( avenae (Gga) , 100%; tritici (Ggt) , 75%; graminis (Ggg) , 71%), but stock cultures of the two varieties of Phialophora radicicola did not form perithecia by any method. For G. graminis , in vivo methods (89, 56–75, 50% of isolates producing perithecia) were generally rather more effective than in vitro methods ( 33 , 69 and 43%). No one in vitro method was the best for all varieties of G. graminis : full-strength potato-dextrose agar was the least effective medium for perithecia, but at quarter-strength and supplemented with yeast extract it ranked with malt extract broth agar and a filter paper method as the most effective in vitro methods for Ggt , but not Gga or Ggg . In host-infection tests using stock isolates those host-parasite combinations that produced most disease or most colonized roots tended also to be those in which the fungus produced perithecia. Perithecia that formed on roots excised from wheat seedlings used to bioassay field soil with little infestation by Ggt were mostly on roots without take-all lesions. In soils with more infestation there was a positive association between lesioned roots and perithecia. G. cylin- drosporus perithecia also occurred on these excised roots, but not on the same plants as Ggt and concurrence of Ggt and P. radicicola var. radicicola was ≤£ 30% of isolates. These fungi seemed to have different seasonal occurrences. Some factors that influence the production of perithecia in vivo are discussed.