The real win in using pseudohashes is when creating and using objects. As an example, if I create an object based on a hash:

package hashobject;

sub new {
  my ($class) = @_;
  $class = ref($class) || $class;
  my $self = {};
  $self->{variable} = 'foo';
  return(bless($self, $class));
}
[download]

and then use the resulting object, I can still autovivify member variables, leading to a whole raft of typo problems:

my $obj = new hashobject;

$obj->{variablee} = 'bar';  # TYPO...
print $obj->{variable};  # Prints 'foo'
[download]

If we move to a pseudohash, the object looks like this:

package pseudohashobject;
use fields qw(variable);

sub new {
  my ($class) = @_;
  $class = ref($class) || $class;
  my $self = bless([\%FIELDS], $class]);
  $self->{variable} = 'foo';
  return($self);
}
[download]

And then (mis)use the object in the same way:

my $obj = new pseudohashobject;
$obj->{variablee} = 'bar';  # TYPO
print $obj->{variable};
[download]

the typo actually causes a runtime exception! This makes very hard to find bugs easy to find, and hence is a good thing (tm). I would suggest looking into Damien Conway's Object Oriented Perl (Manning) for more information. I would almost rate this book as more important than the Camel!