The axial and radial refilling with water of cut dry branches (up to 80 cm tall) of the resurrection plant Myrothamnus flabellifolia was studied in both acro- and basipetal directions by using 1H-NMR imaging. NMR measurements showed that the conducting elements were not filled simultaneously. Axial water ascent occurred initially only in a cluster of a very few conducting elements. Refilling of the other conducting elements and of the living cells was mainly achieved by radial extraction of water from these initial conducting elements. With time, xylem elements in a few further regions were apparently refilled axially. Radial water spread through the tissue occurred almost linearly with time, but much faster in the acropetal than in the basipetal direction. Application of hydrostatic pressure (up to 16 kPa) produced similar temporal and spatial radial refilling patterns, except that more conducting elements were refilled axially during the first phase of water rise. The addition of raffinose to the water considerably reduced axial and radial spreading rates. The polarity of water climbing was supported by measurements of the water rise in dry branches using the ‘light refraction’ (and, sometimes, the ‘leaf recurving’) method. Basipetal refilling of the xylem conduit exhibited biphasic kinetics; the final rise height did not exceed 20–30 cm. Three-cm-long branch pieces also showed a directionality of water climbing, ruling out the possibility that changes in the conducting area from the base to the apex of the branches were responsible for this effect. The polarity of water ascent was independent of gravity and also did not change when the ambient temperature was raised to c. 40 °C. At external pressures of 50–100 kPa the polarity disappeared, with basipetal and acropetal refill times of the xylem conduit of tall branches becoming comparable. Refilling of branches dried horizontally (with a clinostat) or inverted (in the direction of gravity) showed a pronounced reduction of the acropetal water rise to or below basipetal water climbing level (which was unaffected by this treatment). Unlike water, benzene and acetone climbing showed no polarity. In the case of benzene, the rise kinetics (including the final heights) were comparable with those measured acropetally for water, whereas with acetone the rise height was less. Transmission electron microscopy of dry branches demonstrated that the inner surfaces of the conducting tracheids and vessels were lined with a continuous osmiophilic (lipid) layer, as postulated by the kinetic analysis and light microscopy studies. The thickness of the layer varied between 20 and 80 nm. The parenchymal and intervessel pits as well as numerous tracheid corners contained opaque inclusions, presumably also consisting of lipids. Electron microscopy of rehydrated plants showed that the lipid layer was either thinned or had disintegrated and that numerous vesicle-like structures and lipid bodies were formed (together with various intermediate structural elements). These, many other data and the physical–chemical literature imply that several (radial) driving forces (such as capillary condensation, Marangoni forces, capillary, osmotic and turgor pressure forces) operate when a few conducting elements become axially refilled with water. These forces apparently lead to an avalanche-like radial refilling of most of the conducting elements and living cells, and thus to the removal of the ‘internal cuticle’ and of the hydrophobic inclusions in the pits. The polarity of water movement presumably results from high resistances in the basipetal direction, which are created by local gradients in the thickness of the lipid film as a result of draining under gravity in response to drought. There are striking similarities in morphology and function between the xylem-lining lipid film and the lung surfactant film lining the pulmonary air spaces of mammals.