Q: "For dry food we have 40 + 24.4 = 64.4, which means no more than 35.4% for carbohydrates. But Bob says kibbled foods are 50% or more carbohydrates. Please explain the discrepancy." A: There are no discrepancies a lobotomy wouldn't cure. For me, that is. This is an excellent question and it looks like you have caught me in a boo boo, but not the one you assume. I have been writing about kibbles and extruded foods and carbohydrates and stuff for such a long time I simply make the assumption that everyone who care's enough to read it has also read the older stuff. So, to save precious FML space, I sometimes take shortcuts. The 50% figure is a mean, with a range from 40-60%, as I have reported in older posts. Using the 50% mean wasn't an attempt to infuse evil into kibble; I was just trying to save a line and make the post easier to read. However, I agree that saying "50% or more carbohydrates" can be misunderstood by people who have not been following this thread, and if so, I sincerely apologize. Interestingly, I just received a new book on comparative animal nutrition which has an extensive set of tables comparing manufacturing processes and food components required for specific pet foods. It was just published in the last month or so (I bought it for $160 nearly a year ago at a conference and forgot all about it), so the info is about as up to date as can be expected. It indicates extruded dry cat and dog foods contain 35-70% carbohydrates (+/-5%), with a 47% mean, with cat foods generally on the lower end and dog foods generally on the high end. In a footnote, it suggests the change from historic figures is attributed to increased consumer demand for high quality food. I will be posting this reference on the Geek's List shortly. My figures of 40-60%, with a 50% mean are not far off the latest numbers. Your original figures were 36% protein, 22% fat and 10% moisture. To figure out the carbohydrate portion, you do simple subtraction: 100 - 36 - 22 = 42% crude carbohydrate (and ash, trace minerals, etc.). You ignore the moisture content when making the computation because in the chemical analyses to determine fat and protein contents, moisture content isn't considered. Moisture only has computational value when compared to the complete food, NOT the individual components, because that is how it was derived. So, you cannot use dehydrated variables to compute the missing (and hydrated) unknown. The corrected figure of 42% is within the range I have reported; there is no discrepancy. One way to determine the moisture content of a food is to weigh it, drive off the moisture, then reweigh it. The difference in weight represents the moisture content of the complete food. The value does NOT represent the moisture content within individual components. You can see this by looking at a bowl of Alpo; obviously the "gravy" component has more moisture than the "meat" component. When you compare various brands of food, you correct for moisture by subtracting it from individual compounds. This type of analysis gives a rough estimate of dry weight percentages, allowing you to compare one type of food to another. It allows for rough comparisons, BUT it is by no means precise enough to do accurate computations. The reason is because the analysis method to determine moisture content does not consider if the moisture is coming from the protein, fat or carbohydrate components. (similarly, the tests to determine proteins and fats do not determine their points of origin either). It is well known that soluble fiber (a type of carbohydrate) absorbs a tremendous amount of water compared to proteins and fats, so the moisture content associated with proteins would be different than from carbohydrates. Also, proteins have more water bound to them than fats. We know this, but ignore it in the rough computation we use to compare one food to another. For purposes of crude comparison, it is useful, but in terms of determining actual percentages, like "40 + 24.4 = 64.4, ... 35.4% for carbohydrates," it has no mathematical precision. There is simply too much error for reliable or accurate computations. Finally, if you "dehydrate" the 35.6% figure in your question like the rest of the components, you get: 35.6/90 x 100 = 39.6%, which considering the amount of error introduced (as discussed above), is EXTREMELY close to the more correct number of 42% (it is only off 2.4%, well within the +/- 5% error). We could get into discussions of ash percents lowering the carbohydrate estimation, but it would only lower the percentage slightly, making it even closer to the "dehydrated" value and STILL within the predicted norms and degree of error. Like I said, there is no discrepancy. Bob C and 16 Mo' Polecat Pseudomathematicians [Posted in FML issue 3032]