It seems that here and elsewhere, some people are unable to grasp the concept of normal distribution as it applies to stomach emptying. I will now attempt to help by explaining how the known research evidence relates to the known evidence from the Kercher case.
Firstly, this is a normal distribution curve:
http://en.wikipedia.org/wiki/File:PR_and_NCE.gif
The normal curve follows a "bell-shaped" pattern, centred upon the median/mean value (the median and mean are the same in a normal distribution). The height of the curve at any given point reflects the relative frequency of the value under inspection, at the corresponding point on the x-axis. In the case of stomach T(lag), the x-axis represents time in minutes, and the y-axis represents frequency of experimental subjects whose T(lag) equals the corresponding point (dropping a certical line) to the x-axis.
As you can see, the curve drops off either side of the mean/median. The rate of decrease in the curve can either be shown in relation to standard deviations either side of the mean (which might be too statistical), or in terms of percentiles either side of the mean. If using percentiles (which are probably more intuitive to people less familiar with statistical analysis), the mean/median lies at the 50% mark - indicating that 50% of subjects lie below the mean/median (and, of course, 50% lie above the mean/median). The percentile scale is shown on the linked normal curve which I've given above.
Now, how does this relate to T(lag) values. Well, experimental data suggest that the mean/median time for T(lag) is 82 minutes. So the 50% mark is at 82 minutes.
In addition, experimental data show that the 75% mark lies at 102 minutes. In other words, 75% of the experimental subjects had a T(lag) of less than 102 minutes - or, put the other way, 25% of subjects had a T(lag) of more than 102 minutes.
What we can do now is plot the median t value and the 75% t value on the normal curve, to give a properly scaled x-axis. Going back to the link above, we make t=82 at the 50% mark, and t=102 at the 75% mark. We can now extrapolate the x-axis to estimate what value of t corresponds to the 95% mark, the 99% mark, the 99.9% mark, and so on.
If we do this, we find that the 95% mark lies at around t=130. In other words, only 5% of people have a T(lag) of greater than 130 minutes (2 hours and 10 minutes). And we find that the 99.9% mark lies at around 170 minutes (2 hours and 50 minutes). So only 0.1% of people have a T(lag) of over 170 minutes. And, although the link I gave doesn't go beyond the 99.9% mark, at t=240 minutes (4 hours), the corresponding percentile is beyond 99.995% - meaning that fewer than one person per 20,000 has a T(lag) time greater than 4 hours.
Having calibrated the normal curve using experimental data, we can now match it to the Kercher case. The clock starts when Meredith started her final pizza meal. For the sake of argument, let's say that she started that meal at 18.30. Of course, if Meredith had been right on the mean/median mark of the normal curve, her T(lag) would have been 82 minutes. 18.30 + 82 minutes = 19.52. But we know categorically that she was still alive at ten minutes to 8, since she was watching a DVD with her friends. In fact, we categorically know that she was alive until 21.00, which is 150 minutes after the start of her meal.
Now we go back to the values we extrapolated earlier on the normal curve. They show that only 5% of people have a T(lag) of more than 130 minutes, and only 0.1% have a T(lag) of more than 170 minutes.
So, since we know that Meredith definitely passed the 150-minute mark, we must accept that she is already very unusual in her T(lag) physiology. In fact, 150 minutes corresponds to around the 98% mark, implying that only 2% of people have a T(lag) of 150 minutes or more. But only 0.1% of people have a T(lag) of 170 minutes or more, and less than 0.005% have a T(lag) of over 240 minutes.
So, since we know that Meredith's T(lag) has to be greater than 150 minutes, we can estimate the probability that it was between 150 minutes and 170 minutes, or between 170-240 minutes, or over 240 minutes:
Suppose we took 100,000 people and analysed their T(lag). The experimental data suggest that 2% of these people would have T(lag) greater than 150 minutes, equating to 2,000 people. Meredith must be one of these people.
Now, we know that 0.1% of people have a T(lag) of over 170 minutes, which equates to 100 people from our 100,000 group.
So, of the 2,000 people with T(lag) over 150 minutes, 100 of them have a T(lag) greater than 170 minutes. This also of course means that 1,900 of them have a T(lag) of 150-170 minutes.
And we also know that 0.005% have a T(lag) greater than 240 minutes, equating to 5 of our 100,000 sample group. So of the 100 with T(lag) greater than 170 minutes, 5 have T(lag) greater than 240 minutes, and 95 have T(lag) between 170-240 minutes.
Now we can put everything together. Of the 2,000 people with T(lag) over 150 minutes, 1,900 had T(lag) of 150-170 minutes, 95 had T(lag) of 170-240 minutes, and 5 had T(lag) over 240 minutes.
Plugging this back into Meredith's timeline, it means that her probability of dying between 9.00pm and 9.20pm (t=150-170 minutes) was 1900/2000 = 95%. Her probability of dying between 9.20pm and 10.30pm (t=170-240 minutes) was 95/2000 = 4.75%. And her probability of dying after 10.30pm (t>240 minutes) was 5/2000 = 0.25%.