81 patients with chronic low back pain (average duration 10 years) were randomised to two treatment groups. 40 received an empirically devised regimen of forceful spinal manipulation and injections of a dextrose-glycerine-phenol ("proliferant") solution into soft-tissue structures, as part of a programme to decrease pain and disability. The other 41 patients received parallel treatment in which the main differences were less extensive initial local anaesthesia and manipulation, and substitution of saline for proliferant. Neither patients nor assessors knew which treatment had been given. When assessed by disability scores the experimental group had greater improvement than the control group at one (p less than 0.001), three (p less than 0.004), and six (p less than 0.001) months from the end of treatment; at six months an improvement of more than 50% was recorded in 35 of the experimental group versus 16 of the control group and the numbers free from disability were 15 and 4, respectively (p less than 0.003). Visual analogue pain scores and pain diagrams likewise showed significant advantages for the experimental regimen.