[mjn]

Varies quite a bit by field afaict. In CS, I've never paid to publish in journals, and I think it would be seen as weird. However we do often publish papers at conferences (big conferences like SIGGRAPH, CHI, and IJCAI are possibly actually more prestigious than journals to publish in), and in that case it costs money to attend and register for the conference.

[omnisci]

I didn't know that:) In the Life sciences, we tend to pay. We also publish images which are very costly. We have one publication that is going to cost us around $3k to publish as it has a lot of microscopy images.

[MrGunn]

I don't work for them (in fact, they could be considered a competitor), but check out PeerJ. $99 per author for an article. PLOS ONE also has fee waivers if you really can't afford it.

[fludlight]

Is there a legitimate reason why scientific journals still produce dead-trees versions? Why not just publish PDFs? The reader can print if he really must.