The Fe solubility test is a commonly used, easy and relatively cheap in vitro tool for predicting Fe bioavailability in food matrices. However, the outcome of a recent field trial comparing the effect on Fe status of Tanzanian infants of processed V. unprocessed complementary foods (CF), with otherwise the same composition, challenged the validity of this test for predicting Fe bioavailability. In the solubility test, significant more soluble Fe was observed in processed compared with unprocessed foods (mean 18·8 (sem 0·21) V. 4·8 (sem 0·23) %; V<0·001). However, in the field trial, no significant difference in Fe status was seen between processed and unprocessed CF groups after 6 months' follow-up. Therefore, twenty-four samples of these CF (twelve processed and twelve unprocessed batches) were analysed in triplicate for Fe availability using an in vitro digestion–Caco-2 cell culture method and results were compared with solubility results. Significantly more soluble Fe was presented to Caco-2 cells in the processed compared with unprocessed samples (mean 11·5 (sem 1·16) V. 8·5 (sem 2·54) % p=0·028), but proportionally less Fe was taken up by the cells (mean 3·0 (sem 0·40)p. 11·7 (sem 2·22) %; p=0·007). As a net result, absolute Fe uptake was lower (not significantly) in processed compared with unprocessed CF (mean 1·3 (sem 0·16) V. 3·4 (sem 0·83) nmol/mg cell protein; p=0·052). These data clearly demonstrate that the Fe solubility test was not a good indicator of Fe bioavailability in these particular food matrices. In contrast, the results of an in vitro Caco-2 model supported the effects observed in vitro.