Infinite scroll has been called autopagerize, unpaginate, endless pages. But essentially it is pre-fetching content from a subsequent page and adding it directly to the user’s current page.
User is browsing paged content.
- Retaining the user is important and clicking “Next Page” is a usability barrier.
- The full content available is too large to show on initial load.
- The content is available in paged chunks: search results, blog posts, product listings portfolio features.
- Users are retained on the site far better.
- Users are less likely to continue on to the next “page” if they have to click something versus it being delivered automatically to them. 
- Requires no adjustment in a user’s typical reading habits.
- The added functionality needs no affordances or instruction.
- The “footer” of the page will be typically impossible to reach.
- Currently there is no way to cancel or opt-out of the behavior.
- There is no permalink to a given state of the page.
- Dynamically adding more content to the page increases the memory footprint of the browser. Depending on the browser, this could account for around 50megs of RAM.
- Analytics will not immediately capture the event, so custom configuration is required.
- FOR GOD’S SAKE, DON’T BREAK THE BACK BUTTON.
- Depending on site latency, new content can be fetched aggressively (when a user has 500px of the page left to scroll) versus at the last moment (with ~100px left).
- User should be notified when there is no more content available.
- Consider a faux pagination widget to allow for navigation within the new content. Travis Isaacs has ideas on a running pagination.
Examples in the wild:
- Humanized Reader
- Live.com Image Search
- Soup tumblelogs
- Google Reader, Yahoo Mail use the same technique but within a container
- Dazed Digital