Bread owes its appeal to its aerated structure which directly relies on the bubbles entrained into the dough during mixing. If the bubble size distribution (BSD) in the dough can be determined at the end of mixing, then the resulting loaf quality could be predicted before bread is fully manufactured. However, non-invasively monitoring the structure of a fragile opaque soft solid such as dough is challenging. This thesis addressed the challenge by determining dough’s BSD and its evolution using ultrasound and X-ray microtomography. Using a resonant scattering model and the frequency dependence of the ultrasonic parameters measured in the dough, the change in the BSD in dough (made without yeast) with time as a result of disproportionation was determined. At 30 min after mixing, the median radius (R0) of the lognormal BSD was 6.5 microns. Converting the BSD to the radius dependence of bubble volume fraction (BVF(R)), R0V (the median radius of BVF(R)) was 66.4 microns and increased 18 % in the succeeding 90 min. In order to validate the bubble sizes determined ultrasonically, X-rays from a synchrotron source were utilized to examine dough’s microstructure. Large numbers of very small bubbles were discovered and it was apparent that lognormality did not describe the BSDs. Nevertheless, lognormal characterization of the BVF(R) was appropriate. At 30 min after mixing R0V of the BVF(R) was 32.5 microns and it increased by 20 % in the succeeding 90 min, supporting the ultrasonic quantification of bubble volume changes due to disproportionation. Changes in the mode, median and mean of the BVF(R) with time after mixing had the same trend for ultrasound and for X-ray microtomography. The time evolution of the mode of the BVF(R) obtained by ultrasound and X-ray microtomography matched very well; both increasing linearly as a function of time. Ultrasonic assessments of bubble sizes and their changes with time are very encouraging, but the ultrasonic model should use distribution functions that precisely define the empirical data, perhaps not making ‘pre-assumptions’ of lognormality for the BSD data. / February 2015