126 • Alternatives to Animal Use in Research, Testing, and Education 
way in 1985 involved research in cardiology, phys- 
iology, endocrinology, toxicology, and neurology. 
Specific simulations included: 
• regulation of sodium , potas sium , and calcium 
in heart muscle; 
• electrolyte diffusion in heart muscle; 
• propagation of activity in heart muscle; 
• heart volume potentials; 
• mathematical modeling of blood coagulation; 
• regional dose responses in the human and ani- 
mal lung; 
• ciliary motility; 
• cochlear function in the inner ear; and 
• a molecular model of ion transport in nerves 
and muscles. 
Limitations on the utility of computer simula- 
tions stem from the lack of knowledge of all possi- 
ble parameters that may play a role, however slight, 
in the melange of feedback mechanisms that con- 
stitute living systems. Basic biomedical research 
at all levels, some of it involving live animals, will 
continue to provide the new knowledge required 
to improve existing simulations and develop models 
where no satisfactory one exists . The development 
of increasingly powerful computer programs de- 
pends on the use of animals in biomedical research. 
CONTINUED, BUT MODIFIED, USE OF ANIMALS 
IN BEHAVIORAL RESEARCH 
As in biomedical research, the continued, but 
modified, use of animals in behavioral research 
encompasses reducing the number of animals used 
through changes in experimental design and sta- 
tistical analyses, substituting cold- for warm- 
blooded vertebrates, and lessening the degree of 
pain or experimental insult in general, and in pain 
research in particular. Compared with biomedi- 
cal research, behavioral research offers markedly 
fewer opportunities to substitute cold- for warm- 
blooded vertebrates and to use in vitro cultures, 
and it holds little chance of using nonliving systems. 
Reduction in the Number of 
Animals Used 
Improved Experimental Design and 
Statistical Analyses 
Individual animals vary in their behavior both 
between subjects and, in the case of one subject, 
over time. The goal of a behavioral experiment 
is to identify patterns that remain when these two 
sources of variability have been eliminated or taken 
into account. An investigator attempts to conclude 
that observed effects are due to the conditions be- 
ing manipulated in the experiment and not to ex- 
traneous factors . This decision usually rests on the 
outcome of statistical tests. Ensuring the validity 
of such tests or improving their design can mean 
that fewer experiments are needed. Enhanced sta- 
tistical rigor, however, may lead to increases or 
decreases in the number of animals required in 
a particular protocol. 
Statistical Power.— A statistical test’s sensitiv- 
ity in detecting experimental effects is termed its 
"power.” The most widely recognized method of 
increasing power and, hence, the sensitivity of an 
experiment is to use a large sample of subjects. 
Typically, the more variable the results, the more 
power is needed to detect an effect and, therefore, 
the greater the need for large samples. Although 
the magnitude of variability cannot be determined 
prior to an experiment, the amount of variability 
likely to be encountered can be estimated by con- 
ducting small, pilot studies or by examining previ- 
ous research in the same or related areas. Given 
an estimate of variability, statistical tables can be 
used to determine the sample size needed to at- 
tain certain levels of power (221). 
In certain instances, the methods of increasing 
power may reduce, not increase, the number of 
animals needed: 
• Choosing a lower level of statistical signifi- 
cance (i.e., the likelihood that the results were 
due to chance) increases power and reduces 
