From the account of your process: 1st dilution tested is 0.1g/ml of which you use 0.1ml in each of your replicate tests, so the initial quantity tested is 0.01g; subsequent dilutions will give you 0.0001, 0.0001, etc.   However, since you are testing only 3 replicates/dilution level the precision of the test is very low.  A preliminary assessment  of the microbial load would enable you to increase the number of replicates tested to (say) 10/dilution and to reduce the number of dilutions tested. We published a revised approach to calculation of MPNs (see Jarvis, Wilrich & Wilrich, 2010. The paper is available for download from Research Gate) .  The spread sheet software is available from http://www.wiwiss.fu-berlin.de/fachbereich/vwl/iso/ehemalige/wilrich/index.htmland from the ISO website,

In addition, if you want further info on the background to MPNs see Jarvis (2008) Statistical Aspects of the Microbiological Examination of Foods, 2nd Edn, Chapter 8)

Patrick, I'm doing everything in triplicate. Hani, thank you for the link. I've read the link before doing my assay, but find that it falls short of sample preparation. However, I think I have figured it out. Thank you both!

MPN (Most Probable Number) is a statistical method of counting E. coli and other in food/other sample, usually using 3 sets of dilution of 1/10 each with N 5, that is 15 tubes for a single sample. Your dilution depends on original loads of bacteria. It Is better if you have an idea before hand, otherwise you better try to 2 different sets for getting close counts. Good luck.

The following links may help you:-

1. http://www.fda.gov/Food/FoodScienceResearch/LaboratoryMethods/ucm109656.htm

2. http://www.fda.gov/Food/FoodScienceResearch/LaboratoryMethods/ucm064948.htm

3. http://www.fda.gov/Food/FoodScienceResearch/LaboratoryMethods/ucm071429.htm

we use this method for Ecoli as fecal contaminations work well as mazumder said

From the account of your process: 1st dilution tested is 0.1g/ml of which you use 0.1ml in each of your replicate tests, so the initial quantity tested is 0.01g; subsequent dilutions will give you 0.0001, 0.0001, etc.   However, since you are testing only 3 replicates/dilution level the precision of the test is very low.  A preliminary assessment  of the microbial load would enable you to increase the number of replicates tested to (say) 10/dilution and to reduce the number of dilutions tested. We published a revised approach to calculation of MPNs (see Jarvis, Wilrich & Wilrich, 2010. The paper is available for download from Research Gate) .  The spread sheet software is available from http://www.wiwiss.fu-berlin.de/fachbereich/vwl/iso/ehemalige/wilrich/index.htmland from the ISO website,

In addition, if you want further info on the background to MPNs see Jarvis (2008) Statistical Aspects of the Microbiological Examination of Foods, 2nd Edn, Chapter 8)

Regrettably, I've noted 2 typos in my answer:

Line 3 should read "subsequent dilutions will give 0.001, 0.0001, etc"

The web link should end at /index.html

Basil, thank you very much! 

it is a good methos to study fecal contamination  on food , vegetables and woter samples

Dear Sivia Giono,

You are right. I used to sue the MPN-method frequently in '80s, for counting E.coli and fecal coliform(if any) in water, shirmp, Icecream, and other fresh/raw foods.

David's question was very specific: does he use the actual amount of sample in each replicate for each series, or the concentration of GI content in each of these tubes.

According to the definition of MPN, and the formula, you should use the amount of grams that you actually are testing, not g/mL. For example, assuming that you will place you place 0.1 g in whatever broth you plan to use (MUG, lactose broth etc.), that would be your first series. Every dilution, as pointed out by Jarvis, will have 10 times less sample. Since the amount of sample is constant, the actual volume of the broth is likely irrelevant. The FDA-BAM MPN charts are based on the MPN equation, where you will need to know the amount of positive and negative tubes (equaling amount of positive tests as a function of amount of sample) and requires repetitions because, as other counting methods such as plate counts, assumes a Poisson distribution. 

I hope this helps.