What it is
The rel=canonical pointer exists to tell search engines that one page is the child, duplicate, or near-duplicate of another. Examples might be category pages where order is not enforced like www.yoursite.com/tshirts/green and www.yoursite.com/green/t-shirts or minor variations on a page like www,yoursite.com/tshirts?sort_by_price.
All SEO value accrues to the nominated parent page. If search crawlers follow the canonical indicator (which is taken as a guideline rather than a hard and fast rule) the child pages will not be indexed as separate entities.
Why it matters
Simply removing one copy from indexation would mean that any external links leading to the child page would not deliver any SEO value.
What to do
After a full crawl of the site has been made, duplicate and near duplicate pages should be identified and decisions made as to which should be considered the canonical parent and which the child/children. Canonical tags can then be applied.