Across epithelia the water is secreted or absorbed through the cell membrane (transcellular pathway) or the tight junction (paracellular pathway), each route involving specific molecules such as aquaporin or claudin. The paracellular route also allows the passage of some substrates from the circulation depending upon the composition of serum. This idea has fascinated many scientists since long ago. The salivary sugar was recognized as an indicator of blood sugar before the discovery of insulin, but had no clinical application. Modern techniques should be developed for blood-free clinical tests in the near future. The filter size of the paracellular route was estimated by perfusion with various sizes of radioactive dextran in the isolated perfused submandibular gland and analysis reveals that a high proportion of the water crosses by the paracellular route (Murakami et al. 2001). Then discrimination of transcellular and paracellular fluid had been difficult until Segawa noticed the difference in Lucifer Yellow (LY) fluorescence of intercellular canaliculi (IC) between the whole perfused gland and the acini isolated enzymatically. In the latter case, when the enterotoxin (contaminated in the collagenase preparation) bound to claudin and ‘paralyzed’ the tight junction (TJ). Application of carbachol decreased the fluorescence in IC, indicating that the LY was diluted by fluid from the cytosol. We estimated transcellular fluid movement into the lumen due to dilution of LY in the IC of isolated acini (Segawa et al. 2002). By subtraction of this estimate from the whole fluid secretion, we could assess the time course of development of paracellular fluid transport by muscarinic stimulation. The paracellular component started to increase slowly after 30 s from start of transcellular secretion and reached the plateau levels around 1 min after start of transcellular output. The plateau level was higher than that of the transcellular fluid secretion level. How is the paracellular route modulated? During parasympathetic stimulation, addition of sympathetic stimulation increased the passage of sulphhate (Martin & Burgen, 1962), suggesting that the paracellular passage is activated by addition of b-adrenergic stimulation. This finding was reproduced by using LY during perfusion of the rat submandibular gland. To study the underlying mechanism for opening the paracellular route, we observed the TJ structure using freeze-fracture (FF) replicas on rapidly frozen tissues with liquid helium (Hashimoto et al. 2003). During combined stimulation with carbachol and isoproterenol (isoprenaline), microvilli disappeared, and the secretory granule fused with the plasmalemma. The meshwork of TJ strands were rearranged more irregularly and became more straggled. Furthermore, the alignment of TJ membrane particles was interrupted at several points and free ends and terminal loops appeared at the basal side. These changes in the TJ may be related to the subcellular structural modulation of the actin filament network and associated TJ proteins. Changes in fluid transport rates created by hyperosmolar sucrose during the perfusion of isolated rat submandibular glands (SMG) in vitro reduce secretion rates much more than predicted by the osmotic theory of fluid production. However, these are in accord with a theory involving AQP5 feedback control of paracellular fluid transfer (Hill & Shachar-Hill, 2006). The changes in transport rate can be predicted with parameters determined earlier for this gland (Murakami et al. 2001) and a model of the SMG system is presented. Experiments were performed with SMG from genetically selected rats that have very low levels of AQP5 as determined by Western blotting (Murdiastuti et al. 2002). The fluid secretion rates after osmolarity changes were those expected for the osmotic theory. We suggest that control of paracellular flow has been lost in these low AQP5 rats, which have reverted to osmotic fluid production. Retrograde injection of Hg ions into the duct of normal rats partially inhibited AQP5, leading to a concentration-dependent reduction in flow rates. However, reduction of fluid secretion after osmolarity changes was still close to that of normal rats. The results suggest the involvement of a feedback loop including AQP5 and paracellular fluid transport. The meshwork structure of TJ is simple in the salivary gland; to link the data from single cells to the whole organ, we must include the function of the paracellular route in the salivary gland.