By making a quick comparison among several language phonologies (from various language families), I could observe that some phones occur very frequently, such as [m], [p], [b], [h], [a] and [i]. Others seem to be much rarer, among which, [ʕ], [ɰ], [ⱱ] and [ɪ̈].
Do linguists know the frequency of each phone across languages? And, perhaps more importantly, why are some phones more frequent than others?
The most frequently-cited work on this topic is probably Maddieson (1984) (see also Ladefoged & Maddieson 1996), and the associated UPSID database, from which you can get frequency data for different types of contrastive sounds in languages. Taking a diachronic perspective, we can say that rare sounds are those which tend to get easily lost in sound change, because they are difficult to produce or perceive, and those for which there are few pathways for their innovation (even if they are not difficult to produce, they just require very specific conditions to arise).
An interesting discussion of aerodynamic factors influencing the prevalence of different types of stop consonants is found in Ohala (1983); a proposal for the rarity of bilabial trills is found in Maddieson (1989: 91--115).
Peter Ladefoged & Ian Maddieson (1996). The sounds of the world's languages. Oxford: Blackwell. Maddieson, I. 1984. Patterns of Sounds. Cambridge University Press, Cambridge
Phonetics isn't my forte but I might offer some help.
I am aware of at least two theories that try to explain why certain sounds are more wide-spread than others. The first one is called Quantal Theory, proposed by Kenneth Stevens (1972, 1989, 2002, inter alia). There are regions of stability in the articulatory to acoustics mapping (aka quantal regions), when changes in articulation have little impact on acoustics. Sounds that occupy those stable regions tend to be more common in languages. Have a look at this handout
The other one is called Adaptive Dispersion, first proposed by Liljencrants & Lindblom (1972) (the 1971 working paper version can be downloaded for free). Its basic tenet is that sounds in a language tend to be placed in such a way so that perceptual (auditory) contrast between them is at maximum. See this paper by Keith Johnson
Stevens, K.N. 1972. The quantal nature of speech: Evidence from articulatoryacoustic data. In Denes, P.B. and David Jr., E.E. (eds.), Human Communication, A Unified View, 51-66. New York, McGraw-Hill.
Stevens, K.N. 1989. On the quantal nature of speech. Journal of Phonetics 17, 3-46. [the same issue of JPhonet. contains multiple critical responses to Stevens' theory]
