Xyloglucan endotransglycosylases (XETs) cleave and then re-join xyloglucan chains and may thus contribute to both wall-assembly and wall-loosening. The present experiments demonstrate the simultaneous occurrence in vivo of two types of interpolymeric transglycosylation: ‘integrational’ (in which a newly secreted xyloglucan reacts with a previously wall-bound one) and ‘restructuring’ (in which one previously wall-bound xyloglucan reacts with another). Xyloglucans synthesised by cultured rose (Rosa sp.) cells in ‘heavy’ or ‘light’ media (with [13C,2H]glucose or [12C,1H]glucose, respectively) had buoyant densities of 1.643 and 1.585 g ml−1, respectively, estimated by isopycnic centrifugation in caesium trifluoroacetate. To detect transglycosylation, we shifted heavy rose cells into light medium, then supplied a 2-h pulse of l-[1−3H]arabinose. Light [3H]xyloglucans were thus secreted into heavy, non-radioactive walls and chased by light, non-radioactive xyloglucans. At 2 h after the start of radiolabelling, the (neutral) [3H]xyloglucans were on average 29% heavy, indicating molecular grafting during integrational transglycosylation. The [3H]xyloglucans then gradually increased in density until, by 11 h, they were 38% heavy. This density increase suggests that restructuring transglycosylation reactions occurred between the now wall-bound [3H]xyloglucan and other (mainly older, i.e. heavy) wall-bound non-radioactive xyloglucans. Brefeldin A (BFA), which blocked xyloglucan secretion, did not prevent the increase in density of wall-bound [3H]xyloglucan (2−11 h). This confirms that restructuring transglycosylation occurred between pairs of previously wall-bound xyloglucans. After 7 d in BFA, the 3H was in hybrid xyloglucans in which on average 55% of the molecule was heavy. Exogenous xyloglucan oligosaccharides (competing acceptor substrates for XETs) did not affect integrational transglycosylation whereas they inhibited restructuring transglycosylation. Possible reasons for this difference are discussed. This is the first experimental evidence for restructuring transglycosylation in vivo. We argue that both integrational and restructuring transglycosylation can contribute to both wall-assembly and -loosening.