Benchmark:
timing 10 iterations of
 , regexem, scanem, splitem
...

   regexem:  2 wallclock secs ( 1.86 usr +  0.00 sys =  1.86 CPU) @  5.37/s (n=10),

    scanem:  1 wallclock secs ( 0.80 usr +  0.00 sys =  0.80 CPU) @ 12.48/s (n=10),

   splitem:  1 wallclock secs ( 1.20 usr +  0.00 sys =  1.20 CPU) @  8.32/s (n=10),

          Rate regexem splitem  scanem
regexem 5.37/s      --    -35%    -57%
splitem 8.32/s     55%      --    -33%
scanem  12.5/s    133%     50%      --
2000, 2000, 2000

c:\test>

##</code><code>##

#! perl -slw
use strict;
use vars qw/$test @splits @regexs @scans/;
use Benchmark qw/cmpthese/;

local $, = ', ';

sub gendata {
	my ($min, $max) = @_;
	my $count = 0;
	my $data = '';
	$data .= $count++ . ('X' x ($min + rand( $max-$min))) . ' ' while length $data <2000;
	return $data;
}

sub splitem (\$$) {
	my ($scalar, $fieldno) = @_;

	return (split ' ', $$scalar, $fieldno+1)[-2];
}

sub regexem (\$$) {
	my ($scalar, $fieldno) = @_;

	my ($result) = $$scalar =~ /(?:([^ ]*) ){$fieldno}/;
	return $result;
}

sub scanem (\$$) {
	my ($scalar, $fieldno) = @_;
	my $p = 0;
	$p = index( $$scalar, ' ', ++$p ) while --$fieldno >0;
	$p++ if $p;
	return substr $$scalar, $p, index($$scalar, ' ', $p)-$p;
}

$test = gendata 4, 10;

cmpthese( 10, {
	splitem	=>	'push @splits, splitem $test, $_ for 1 .. 200',
	regexem	=>	'push @regexs, regexem $test, $_ for 1 .. 200',
	scanem	=>	 'push @scans,  scanem  $test, $_ for 1 .. 200',
});

print ~~@splits, ~~@regexs, ~~@scans;
print 'splits differ' if "@splits" ne "@regexs" and "@regexs" eq "@scans";
print 'regexs differ' if "@splits" ne "@regexs" and "@splits" eq "@scans";
print 'scans  differ' if "@scans"  ne "@regexs" and "@splits" eq "@regexs";