The following code shows some variations on what you want. It matches and captures a string containing one or more no white space characters(represented by \S). The various options show the difference between greedy and non-greedy matching as well as anchoring on word boundaries.

#! /usr/bin/perl -w
use strict;
use warnings;

my $data = "abcdefghijklmnopFqrstuvwxyz arrrrhhhhg!";
# Greedily match as much non-whitespace (\S+) as we can that
# starts with an a and ends with an f or g
print "Greedily matched $1\n" if $data =~ /(a\S+(f|g))/i;

# Add the ? modifier to make the expression match minimally
print "Minimally matched $1\n" if $data =~ /(a\S+?(f|g))/i;

# Anchor each end of the expression to a word boundary (\b)
# so that we only match words that start with an a and ends with an f 
+or g
print "Word matched $1\n" if $data =~ /(\ba\S+(f|g)\b)/;

# Apply the same regular expression to pick out all of the
# matches in the data
while ($data =~ /(a\S+?(f|g))/ig)
{
    print "Word: $1\n";
}
[download]