In the U.S., Spanish – unquestionably. Like it or not, there are millions of Spanish-speaking U.S. citizens, and the percentage is growing. Likewise, for a Canadian, French – and there, it’s required in school in most places.
In general, though: you could make a case for a number of languages. Mandarin is probably the language with the most native speakers – but how often will you use it, outside of being in China or doing work that involves translating to and from Chinese in one fashion or another.
Perhaps the best way to a non IMHO answer migth be to graph it out two-dimensionally. One axis would be number of native speakers, with a factor relative to how many nations it’s dominant in. Russian and Portuguese are big, but outside Russia and Brazil, there are few native speakers. On the other hand, Arabic and Spanish are major languages in a large number of countries – with dialectal variations, to be sure, but still widespread.
The Y axis would be number of people factoring in number of locations where it is used as a second language, a lingua franca. Very few Danes and Dutch speak English in the hume, but English is so widespread in Denmark and the Netherlands as a second language that one almost need not know Danish or Dutch. There are some odd results here – Fulani (Peul, Fulfulde), a rather small language, is spoken across a large swath of West Africa. Tamil pops up in the strangest places, from Guyana to Malaysia – of all the languages of India, it has the widest use beyond the Republic.
But the results should be informative – the languages that are not only the first, home language for the most people, but also the most widely spoken beyond their homelands. If 50 million Italians but almost nobody else speak Italian, compared to something spoken by 25 million natives and 25 million others as a second language, I’d give the nod to the second language as between the two – it’s more widely spoken, even though the numbers are comparable.
I’d be very intrigued to see what the results of such an analysis come out to.