Methods of Research

Industrial/organizational psychology borrowed much from many other areas
of psychology during its growth and has retained the strong research orientation
common to them, along with many of the research methods each
has developed and many of the findings that each has generated. Bringing
psychological methods to work settings where experts from many other disciplines
are studying some of the same problems results in conflicts, but it
also produces a richness of information beyond the scope of any one of the
disciplines.

In most cases, the most feasible approach to data collection for I/O psychologists
is field research, an approach in which evidence is gathered in a
“natural” setting, such as the workplace. (By contrast, laboratory research
involves an artificial, contrived setting.) Systematic observation of ongoing
work can often give a psychologist needed information without greatly disturbing
the workers involved. Generally, they will be told that data are being
gathered, but when the known presence of an observer likely would change
what is being studied, unobtrusive methods might be used. Information
from hidden cameras, or observations from researchers pretending to be
workers and actually engaging in whatever must be done, can be used when
justified.
Again studying within the actual work setting, I/O psychologists may
sometimes take advantage of natural experiments, situations in which a
change not deliberately introduced may be studied for its effect on some important
outcome. If, for example, very extreme, unseasonable temperatures
resulted in uncontrollably high, or low, temperatures in an office setting, a
psychologist could assess the effects on employee discomfort, absenteeism,
or productivity.

Still studying within the actual work setting, an I/O psychologist may arrange
a quasi-experiment, a situation in which the researcher changes some
factor to assess its effect while having only partial control over other factors
that might influence that change. For example, the psychologist might
study the effects of different work schedules by assigning one schedule to
one department of a company, a second schedule to a second department,
and a third schedule to a third department. The departments, the people,
and the differences in the work itself would prevent the strategy from being
a true experiment, but it still could produce some useful data.
An experiment, as psychology and other sciences define it, is difficult to
arrange within work settings, but it may be worth the effort to evaluate information
gathered by other methods. In the simplest formof experiment, the
researcher randomly assigns the people studied into two groups and, while
holding constant all other factors that might influence the experiment’s
outcome, presents some condition (known as an independent variable) to
one group of subjects (the experimental group) and withholds it from another
(the control group). Finally, the researcher measures the outcome
(the dependent variable) for both groups.
Carrying out a true experiment almost always requires taking the people
involved away from their typical activities into a setting obviously designed
for study (usually called the laboratory, even though it may bear little resemblance
to a laboratory of, say, a chemist). The need to establish a new, artificial
setting and the need to pull workers away from their work to gather information
are both troublesome, as is the risk that what is learned in the
laboratory setting may not hold true back in the natural work setting.
Correlational methods, borrowed from psychometrics, complement the
observational and experimental techniques just described. Correlation is a mathematical technique for comparing the similarity of two sets of data (literally,
to determine their co-relation). An important example of the I/O
psychologist’s seeking information on relationships is found in the process
of hiring-test validation, answering the question of the extent to which test
scores and eventual work performance are correlated. To establish validity, a
researcher must demonstrate a substantial relationship between scores and
performance, evidence that the test is measuring what is intended.