Twenty-five obese and 23 reference women were compared with respect to their peripheral insulin concentrations in response to the sight and smell of food. An additional 21 obese women were examined for different control purposes. The women were examined after fasting for approximately 16 hr. Venous blood samples for determination of glucose and insulin were drawn 20, 10, and 1 min prior to the demonstration of food for 5 min. After the food had been presented to the subjects, samples were drawn at 1, 2, 3, 4, 5, 6, 10, 15, and 20 min. The response was calculated in two different ways: method I--the difference between meal basal insulin values and mean insulin values during and/or after stimulation, and method II--the "insulin area" over the mean basal concentration was calculated for 0-20 min after start of food presentation. Both methods resulted in significantly higher insulin responses in obese as compared to reference subjects. However, when performing duplicate experiments in the same subjects only method II resulted in reproducible results and even with this method the error was as high as 60%-90%. The high error of the method was partly expected since the insulin elevation is most likely not only a function of controlled external cues but also dependent on unknown sensorimotor and cognitive-affective alterations. No insulin response was observed when obese women were exposed to an external cue that was not food related. Atropine completely blocked the insulin elevation in response to food related external stimuli indicating that this insulin response is mediated via vagus.