Use brackets to memorize an chunk, and then \1, \2, etc to refer to the memorized chunk inside the regex ($1, $2 outside it)

For example, to search for words which have the same letter twice;

my $string = <>;
$string =~ /(\w*(\w)\w*\2\w*)/;
# The beginning of the word, then a notable letter, 
# maybe some middle letters, then that same notable letter as before, 
# then the rest of the word.
print "Found word '$1', with duplicated letter '$2'\n";
[download]

This generates:

>image imagination dream
Found word 'imagination', with duplicated letter 'n'
>
[download]

If you want to find the duplicated letter which appears first rather than last, change the first \w* to the non-greedy form \w*?. It should be easy to extend this principle to whole sentences/strings.