Page 61 - Media Effects Advances in Theory and Research

50 GERBNER ET AL.

potential “lessons” of the television world and when viewing data are
available for the respondents.
Television viewing is usually assessed by asking about the amount of
time respondents watch television on an “average day.” Multiple mea-
sures are used when available. Because these measures of amount of view-
ing are assumed to provide relative, not absolute, indicators, the determi-
nation of what constitutes “light,” “medium,” and “heavy” viewing is
made on a sample-by-sample basis, using as close to an even three-way
split of hours of daily television viewing as possible. What is important is
that there should be significant relative differences in viewing levels, not
the actual or specific amount of viewing. The heaviest viewers of any sam-
ple of respondents form the population on which cultivation can be
3
tested. The analysis of simple patterns across light, medium, and heavy
viewing groups (overall and in key subgroups) is useful to illuminate the
general nature of the cultivation relationship, but it is normally followed
up with more stringent multivariate analysis using continuous data.
The observable evidence of cultivation is likely to be modest in terms of
absolute size. Even “light” viewers may be watching several hours of tele-
vision a day and, of course, live in the same general culture as heavy view-
ers. Therefore, the discovery of a consistent pattern of even small but per-
vasive differences between light and heavy viewers may be of far-reaching
consequence. Extensive and systematic reexamination of hundreds of cul-
tivation studies carried out over more than two decades (using the statisti-
cal techniques of meta-analysis; Shanahan & Morgan, 1999) has shown
that cultivation relationships typically manifest a strength of about .10
using a common metric, the Pearson correlation coefficient.
What some critics belittle as “small effects” may have significant reper-
cussions. It takes but a few degrees shift in the average temperature to have
an ice age or global warming. The 2000 U.S. presidential elections showed
the havoc that could be wreaked by a miniscule percentage of votes. A
range of 5 to 15% margins (typical of our “cultivation differentials”) in a
large and otherwise stable field often signals a landslide, a market takeover,
or an epidemic, and it overwhelmingly tips the scale of any closely bal-
anced choice, vote, or other decision. A single percentage point ratings dif-
ference is worth many millions of dollars in advertising revenue—as the
media know only too well. Thus, a slight but pervasive (e.g., generational)
shift in the cultivation of common perspectives may alter the cultural cli-
mate and upset the balance of social and political decision making.

3 In all analyses we use a number of demographic variables as controls. These are applied
both separately and simultaneously. Included are gender, age, race, education, income, and
political self-designation (liberal, moderate, conservative). Where applicable, other controls,
such as urban-rural residence, newspaper reading, and party affiliation, are also used.