This study investigated the effects of a structured procedure for collaborative writing (Paired Writing) on the quality of creative writing and attitudes to writing in 8-year-old children. Each of two classes featured an experimental collaborative writing group and a control group, which produced writing individually. In one class the experimental group collaborated in different-ability pairs with fixed roles as helper and writer; in the other class, in same-ability pairs reciprocating roles. The 8 week project involved training for the participating children. Analyses were made of quality of individual writing before and after the project and of collaborative writing during the project. Pre-post gains in individual writing were statistically significant for the cross-ability experimentals but not for their controls. Pre and post scores for same-ability experimentals were not significantly different, nor for their controls. However, the same-ability experimentals improved while their controls deteriorated, and the difference in gain between experimentals and controls was significant. Also, the collaborative writing of same-ability pairs scored significantly higher than their pre-project individual writing, although this was not the case for cross-ability pairs. Both types of Paired Writers reported finding the method easy to use and a majority showed positive attitudes to it. It was concluded that both formats of Paired Writing could be effective. However, ensuring short term gains for the more able helpers in cross-ability pairs could be problematic. Further research is needed, including trials of cross-ability reciprocal role pairing.