Categories and Checklists for Observational Research
 |
Researchers
using structured observational methods specify in advance the behaviors or
events to be observed and use record-keeping forms that yield numeric
information. Observers using structured observation are still required to make
some inferences and exercise judgment, but they are restrained with regard to
the kinds of phenomena that will be watched and recorded.
The creativity of
structured observation lies not in the observation itself but rather in the
formulation of a system for accurately categorizing, recording, and encoding
the observations. Because structured techniques depend on plans developed
before the actual observation, they are not appropriate when researchers have
limited knowledge about the phenomena under investigation.
Categories
and Checklists
The
most common approach to making structured observations consists of constructing
a category system for classifying observed phenomena. A category system
represents an attempt to designate in a systematic or quantitative fashion the
qualitative behaviors and events transpiring in the observational setting.
Considerations in Using Category Systems A critical requirement for a good
category system is the careful and explicit definition of behaviors and
characteristics to be observed.
Each category must be explained in detail with
an operational definition so that observers have relatively clearcut criteria
for determining the occurrence of a specified phenomenon. Virtually all
category systems do, however, require observers to make some inferences,
although there is considerable variability on this dimension.
In
this system, assuming that observers were properly trained, relatively little
inference would be required to allocate the infant’s state to the proper
category. Other category systems, however, require considerable inference.
In
scales like the AIMS, even when categories are defined in detail, a heavy
inferential burden is placed on observers. The decision concerning how much
observer inference is appropriate depends on a number of factors, including the
research purposes and the observers’ skills. Beginning researchers are advised
to construct or use category systems that require only a moderate degree of
inference.
Another consideration in structured category systems concerns the
exhaustiveness of what is to be observed. Some category systems are constructed
to classify all observed behaviors of a certain type (eg, all body movements)
into mutually exclusive categories. A contrasting technique is to develop a non
exhaustive system in which only particular types of behavior are categorized.
For example, if we were observing children’s aggressive behavior, we might develop
such categories as “strikes another child,” “calls other children names,”
“throws objects around the room,” and so forth. In this category system, many
behaviors (all those that are non aggressive) would not be classified. Such
non-exhaustive systems are adequate for many research purposes, but they run
the risk of providing data that are difficult to interpret.
When a large number
of observed behaviors are unclassified, investigators may have difficulty
placing categorized behaviors into a proper context. When observers use an
exhaustive system that is, when all behaviors of a certain type, such as
verbal interaction, are observed and recorded researchers must be careful to
define categories so that observers know when one behavior ends and a new one
begins.
Another essential feature is that referent behaviors should be mutually
exclusive. If overlapping categories are not eliminated, observers will have
difficulty deciding how to classify a particular observation. The underlying
assumption in the use of such a category system is that behaviors, events, or
attributes that are allocated to a particular category are equivalent to every
other behavior, event, or attribute in that same category.
Checklists
for Exhaustive Systems
A
category system is the basis for constructing a checklist, which is the
instrument observers use to record observed phenomena. The checklist is usually
formatted with the list of behaviors or events from the category system on the
left and space for tallying the frequency or duration of occurrence of
behaviors on the right.
In complex social situations with multiple actors, the
right-hand portion may be divided into panels according to characteristics of
the actors (eg, nurse/physician; male patients/ female patients) or by individual
subjects’ names or identification numbers. The observers’ task with an
exhaustive checklist is to place all behaviors in only one category for each
element. By element, we refer to either a unit of behavior, such as a sentence
in a conversation, or to a time interval.
To illustrate, suppose we were
studying the problem-solving behavior of a group of public health workers
discussing a new intervention for the homeless. Our category system involves
eight categories:
(1) seeks information
(2) gives information
(3) describes
problem
(4) offers suggestion
(5) opposes suggestion
(6) supports
suggestion
(7) summarizes
8) Miscellaneous
Observers would be required
to classify every group member’s contribution using, for example, each sentence
as the element in terms of one of these eight categories. Another approach with
exhaustive systems is to categorize relevant behaviors at regular time
intervals.
For example, in a category system for infants’ motor activities, the
researcher might use 15-second time intervals as the element; observers would
categorize infant movements within 15-second periods. Checklists based on
exhaustive category systems are demanding because the recording task is
continuous.
Checklists
for Non Exhaustive Systems
The
second approach, which is sometimes referred to as a sign system, begins with a
list of behaviors (or symptoms) that subjects may or may not manifest. The
observer’s task is to watch for instances of the behaviors on the list. When a
behavior occurs, observers either place a check mark beside the behavior
category to designate its occurrence or make a cumulative tally of the number
of times the behavior occurred.
The resulting product is a kind of demography
of events transpiring in the observational period. With this type of checklist,
the observer does not classify all behaviors or characteristics of those being
observed, but rather identifies the occurrence of particular ones.
Rating
Scales
The
major alternative to a checklist for recording structured observations is a
rating scale that requires observers to rate a phenomenon along a descriptive
continuum that is typically bipolar. The ratings are quantified for subsequent
statistical analysis. Observers may be required to rate behaviors or events at
specified intervals throughout the observational period (eg, every 15 minutes),
in much the same way that a checklist would be used.
Alternatively, observers
may rate entire events or transactions after observations are completed.
Postobservation ratings require observers to integrate a number of activities
and to judge which point on a scale most closely fits their interpretation of
the overall situation.
For example, suppose we were comparing the behaviors of
nurses working in intensive care units with those of nurses in other units.
After each 15-minute observation session, observers are asked to rate the
perceived degree of tension of nurses in each unit. The rating scale might take
the form of a graphic rating scale:
Global
observational rating scales are often included at the end of structured
interviews. For example, in a study of the health problems of nearly 4000
welfare mothers, interviewers were asked to observe and rate the safety of the
home environment with regard to structural or potential health hazards to the
children on five-point scales, from completely safe to extremely unsafe (Polit,
London, & Martinez, 2001).
Rating scales can also be used as an extension
of checklists, in which observers not only record the occurrence of a behavior
but also rate some qualitative aspect of it, such as its magnitude or
intensity. A particularly good example is Weiss’s (1992) Tactile Interaction
Index (TII) for observing patterns of interpersonal touch.
The TII comprises
four dimensions: location (part of body touched, such as arm, abdomen); action
(the specific gesture or movement used, such as grabbing, hitting, patting);
duration (temporal length of the touch); and intensity. Observers using the
index must both classify the nature and duration of the touch and rate the intensity
on a four-point scale: light, moderate, strong, and deep.
When rating scales
are coupled with a category scheme in this fashion, considerable information
about a phenomenon can be obtained, but it places an immense burden on
observers, particularly if there is extensive activity.