Thanks very much for these answers. Note that I was already pretty sure the second and third results were correct because the function dgig is the density function of the generalized inverse Gaussian and I had done a Monte Carlo estimate of the integral (being a probability) using a sample of 10,000,000 and got the answer to at least 4 decimal places.

Your explanations did clarify for me what Maple was doing. That is important to me because I am trying to determine some reference values for the probability distribution. The message I got is that I need to use evalf, not just int to ensure I get a proper numerical calculation. Also, increasing the number of digits may be needed.

David Scott