A diamond composition ATR probe was used in situ to obtain IR spectra on replicate Escherichia coli fermentations involving a complex medium. The probe showed excellent stability over a 6-month operating period and was unaffected by either agitation or aeration. The formation of an unknown was observed from IR spectra obtained during the sterilization; subsequent experiments proved this to be a reaction product between yeast extract and the phosphates used as buffer salts. Partial-least-squares-based calibration/prediction models were developed for both glucose and acetate using in-process samples. The resulting models had prediction errors of ±0.26 and ±0.75 g/L for glucose and acetic acid, respectively, errors which were statistically equivalent to the estimated experimental errors in the reference measurements. Relative concentration profiles for the unknown formed during sterilization could be generated either by tracking peak height at an independent wavelength or by self-modeling curve resolution of the spectral region overlapping that of glucose. These profiles indicated that this compound was metabolized simultaneously with glucose; upon depletion of the glucose, when the microorganism switched to consumption of acetic acid, utilization continued but at a lower rate. The data presented provide an extensive characterization of the performance characteristics of this in situ analysis and clearly demonstrate its utility not just in the quantitative measurement of multiple known species but in the qualitative evaluation of unknown species.