Actually, your regex, while good, has a slight problem. In a character class (something like [a-zA-Z]), the target text is filtered against each individual character in the class. No bar '|' for alternation is necessary. Your class actually is testing for a dash, a bar, and a space, rather than alternating between dash and space. Here's this code slightly cleaned up:

#!/usr/local/bin/perl -w

use strict;

while (<DATA>) {
    s /[\n\r]//g;

    print "Testing with $_, result is ";

    m/(?:1[- ]?)?    # Testing for an leading 1 and dash or space behi
+nd it
        \(?          # Optional parentheses around the area code
          (\d{3})    # Area code
        \)?          # Close optional parentheses around area code
        [- ]?        # Optional dash or space after area code
        (\d{3})      # Exchange
        [- ]?        # Optional dash or space after exchange
        (\d{4})      # Line
    /x;              # /x modifier allows whitespace in a regex

    my $areacode = $1;
    my $exchange = $2;
    my $line     = $3;

    print "($areacode) $exchange-$line\n";
}

__DATA__
212-555-1212
(212)555-1213
1-(212)-555-1214
1-212-555-1215
212 555 1216
(212) 555 1217
1 (212) 555 1218
1 212 555 1219
12125551210
[download]

A couple of notes on the code: when single character alternates are allowed (that's not quite the same thing as single byte alternates), a character class is much faster than alternation. In other words, we're using [- ] rather than (-| ).

I also switched the first set of parentheses from (1[- ]?) to (?:1[- ]?). The (?:someregex) construct allows for grouping without backreferencing to a $somenumber variable. This is more efficient than straight parentheses for grouping and should be used, when possible.