First, the authors independently generated a large
pool of items for each of the constructs included in
the study. Care was taken to tap the domain of each
construct as closely as possible. For example, multiple
items were generated to correspond to each of the
three components of market orientation From this pool
of items, a subset was selected using the criteria of
uniqueness and the ability to convey "different shades
of meaning" to informants (see Churchill 1979). Several
items were reverse-scored in order to minimize
response set bias.