Starch and soluble sugars are the major photosynthetic products, and their carbon isotope signatures reflect external versus internal limitations of CO2 fixation. There has been recent renewed interest in the isotope composition of carbohydrates, mainly for use in CO2 flux partitioning studies at the ecosystem level. The major obstacle to the use of carbohydrates in such studies has been the lack of an acknowledged method to isolate starch and soluble sugars for isotopic measurements. We here report on the comparison and evaluation of existing methods (acid and enzymatic hydrolysis for starch; ion-exchange purification and compound-specific analysis for sugars). The selectivity and reproducibility of the methods were tested using three approaches: (i) an artificial leaf composed of a mixture of isotopically defined compounds, (ii) a C4 leaf spiked with C3 starch, and (iii) two natural plant samples (root, leaf). Starch preparation methods based on enzymatic or acid hydrolysis did not yield similar results and exhibited contaminations by non-starch compounds. The specificity of the acidic hydrolysis method was especially low, and we therefore suggest terming these preparations as HCl-hydrolysable carbon, rather than starch. Despite being more specific, enzyme-based methods to isolate starch also need to be further optimized to increase specificity. The analysis of sugars by ion-exchange methods (bulk preparations) was fast but produced more variable isotope compositions than compound-specific methods. Compound-specific approaches did not in all cases correctly reproduce the target values, mainly due to unsatisfactory separation of sugars and background contamination. Our study demonstrates that, despite their wide application, methods for the preparation of starch and soluble sugars for the analysis of carbon isotope composition are not (yet) reliable enough to be routinely applied and further research is urgently needed to resolve the identified problems. Copyright © 2009 John Wiley & Sons, Ltd.