of the balance between water inputs (precipitation and/or irrigation) and losses through evapotranspiration (ET), and has been used to examine the effects of climatic variability on terrestrial systems (Kahmen et al., 2011; Sternberg, 2009). Although extensively used in living plants (e.g., tree rings), the analysis of cellulose is limited by its typically short life after decomposition of plant biomass. Under particular conditions, cellulose extracted from fossil plants have been used to produce climate records that go back millions of years (Jahren and Sternberg, 2003), but such well-preserved samples are rare. In contrast to cellulose, plantderived lipids can persist in soils and sediments due to their low biodegradability and stabilisation into hydrophobic domains of organic matter (Matsumoto et al., 2007; Assis et al., 2011), providing a potential long-term record of terrestrial water balance. Plant lipids may be broadly defined as hydrophobic or amphiphilic molecules originating entirely or in part from two types of biochemical subunits, ketoacyl and isoprene groups, which are building blocks for molecules of various structures and polarities, including some with no oxygen at all (e.g., alkanes in leaf surface wax) and others that carry various amount of oxygen (Fahy et al., 2009). The value of lipid hydrogen isotope analysis (δ 2 H) has already been recognised for specific compounds recovered from soils and sediments (Sachse et al., 2012). The study of lipid δ 18 O signals has the potential to improve δ 2 H records by, for example, allowing water loss to be partitioned into evaporation and transpiration (Voelker et al., 2014). The first step towards the use of lipid δ 18 O records is to demonstrate that a signal related to changes in water balance is recorded in such compounds. The second step is then to demonstrate that this signal is preserved in soils and sediments. The present study is concerned with the first step. However, without losing sight of the second step, we imposed the restriction that lipid extracts should contain only the most apolar compounds present in the plant biomass, under the assumption that these are also the most likely to persist after deposition (Matsumoto et al., 2007). To advance the interpretation of lipid δ 18 O signals, we compared measurements of cellulose and hexane-extractable compounds of C 3 and C 4 species, grown under contrasting water regimes in replicated field experiments. In addition, we analysed the carbon isotope composition (δ 13 C) of cellulose and lipids as a way to assess plant water-use efficiency (Farquhar and Von Caemmerer, 1982). Alongside these isotopic measurements, we used spectroscopic analyses to qualitatively describe the molecular composition of lipid extracts and all results are integrated into a consideration of potential applications in ancient and contemporary settings. Methods Field Experiment. The field experiment was conducted at the University of California Five Points Experimental Station (36°20'10.28"N, 120°6'38.40"W). This site experie...