THE NATURE OF SCIENTIFIC PROOF

http://www.holycross.edu/departments/biology/kprestwi/behavior/e&be_notes/E&BE_04_Sci_Meth&Philo.pdf

THE NATURE OF SCIENTIFIC PROOF
This handout has some general comments about experiments and "proof" in science. It is for
review purposes -- you should be familiar with these ideas from Introductory Biology.
I. THE NATURE OF SCIENTIFIC PROOF
A. Is there proof in science?
1. In the sense that the word proof is used in mathematics and philosophy, nothing
is ever proven in science. There is always some uncertainty about the actual value of results
obtained from some experiment or their interpretation.
2. The more times an observation is repeated and the greater number of different
observations and theories that it ties into and agrees with, the more confident we are about how well
we actually understand something.
3. However, in the strictest sense, we never arrive at "proof"; we simply arrive at a
very high degree of probability that we understand something. Thus, it is important that you shift
your frame of reference from one of proof and certainty of knowledge and interpretation of facts to
one that is PROBABILISTIC in nature, where our confidence in whether or not we understand
something properly is not and never can be absolute. Thus, you are well advised to remove the word
"proof" from your vocabulary as far as science is concerned.
This should be no big surprise -- truth and proof in our own lives are generally probabilistic in
nature. In fact, it is only in philosophy and mathematics where the criteria are rules of logic where
the idea of proof, in its purest sense, ever has absolute meaning.
B. APPROACHES TO FINDING TRUTH IN BIOLOGY
1. Many define science as a METHODOLOGY -- not a body facts about certain
things. Thus, what makes a scientist are the methods that she or he uses to produce the body of
knowledge that is associated with a particular branch of science. Anyone who carefully uses these
methods is therefore a scientist.
2. The GENERAL METHODOLOGY USED IN ANSWERING QUESTIONS
IN BIOLOGY. It is largely derived from the work of the German-English philosopher of science,
Sir Karl Popper.
a. The principle element and starting point is usually OBSERVATION.
1. Observation is always guided, either consciously or
unconsciously, by some sort of theoretical view of the world that is the result of any observers'
experience.
2. Thus, to some degree or another, it is vital to realize that all
observation is TAINTED BY BIAS.
i. It is well known that people with different ideas about the
world and different experiences will tend to see events in either slightly or very different ways.
ii. Can biologists expect to be any different? The answer is
no, although, we try to minimize any biases we have, in part by becoming aware of them.
iii. Nevertheless, it is vital to realize that no observation, not
even that taken by a machine can be regarded as non-biased.
In the second class of the semester we will discuss techniques behaviorists use to obtain data and to
try to avoid bias.
3. Thus, it should not surprise anyone that occasionally large bodies
of observations are re-interpreted and the way that future scientists view certain "facts" changes.
Two examples are (i) the changes in explanations for the appearance of biological species that
resulted from acceptance of Darwin's view of life and (ii) the acceptance of the particulate nature of
Science, Philosophy, Methodology and the Nature of "Proof" 2
genes and the probabilistic nature of inheritance as the result of the work of Mendel and others. We
will see other examples of such "REVOLUTIONS" or PARADIGM SHIFTS in this course.
b. As a result of observation, or sometimes simply as a result of pre-existing ideas,
scientists speculate about the way things are.
1. For instance, in asking proximate questions, scientists become very
interested in chains of causality, that is, cause and effect types of relationships.
2. They speculate on the nature of these cause and effect relationships and
the result of such speculation is some sort of HYPOTHESIS. This represents a set of ideas about
how something works or how it happened based on observation and theory.
c. So far nothing has occurred that distinguishes science from any other type of
human knowledge.
1. However, now a crucial step occurs. This is the generation of a series of
SPECIFIC PREDICTIONS from the hypotheses that can be tested by some sort of observation.
2. Karl Popper stated that the key attribute that these predictions (and
therefore the hypotheses from which they sprung) must possess is that they must be falsifiable
BY OBSERVATION OR BY EXPERIMENT, that is, it must be possible to show that they are
wrong.
i. If it is not possible to show such a thing, then the hypothesis that
generated them is unscientific and the nature of the work being done is also unscientific.
ii. Thus, explanations (hypotheses) relating to the existence of a
supernatural cause of an event are unscientific since they are not potentially falsifiable in the sense
that they can be tested through observation: supernatural events are those that do not have a natural
cause.
iii. As we will see below, the best hypotheses will be those that can
generate a number of quantitative predictions, each of which can be tested independently and with a
minimum of subjective judgment on the part of the investigator.
d. According to Popper, THE ONLY WAY THAT SCIENCE CAN SEEK TO
VALIDATE OR SUPPORT A HYPOTHESIS IS TO TRY TO SHOW THAT IT IS FALSE.
1. As mentioned above, no idea is ever proven correct.
2. FALSE IDEAS ARE DISCARDED, BUT THIS DOES NOT PROVE
THE ALTERNATIVES THAT REMAIN. THEY SIMPLY HAVE NOT BEEN DISPROVED
YET.
a. However, the more often they withstand the test of falsification, the
more confident that we feel that they are likely to be correct.
b. Note again that scientific truth is probabilistic, not absolute. There
is always room to overturn the ideas we have held previously about how the universe works. This is
much of the excitement of science.
e. When hypotheses have repeatedly withstood the tests of falsification in many
different ways, we tend to elevate them to the level of THEORY.
1. As mentioned in the previous class, when used in science, theory comes
close of the idea of fact as is used in everyday life.
2. Again, note that while we have great confidence in anything that we call a
theory (such as evolution by natural selection), we still are not absolutely certain about their
correctness.
f. To complete the circuit, theories (along with observations and hypothesis) will of
course influence how scientists come to view the world; it will determine the way their observations
are made and interpreted and the types of hypotheses they generate.
Science, Philosophy, Methodology and the Nature of "Proof" 3
g. Notice that throughout this discussion, we have left out the concept of
EXPERIMENT. It is crucial to the method and therefore it will be discussed separately at this
time.
1. It is in the type of experiment used that we see the main differences
between procedures used to answer proximate and ultimate questions of causation. In both cases
the role of an experiment is to find a situation where observation will unambiguously be able to
falsify a hypothesis.
2. When PROXIMAL QUESTIONS (questions relating to mechanism --
we will discuss these in class) are being asked, it is usually possible to perform a DIRECT
MANIPULATION AND OBSERVE THE RESULT and then evaluate competing hypotheses.
a. For example, we may believe that a certain hormone will increase
the amount of neurotransmitter released by a certain class of axons.
1. Two hypotheses (at least) could be generated, one that the
hormone did increase the neurotransmitter release and the other that it didn't.
2. A possible manipulation to test these hypotheses would be
to
(i) measure neurotransmitter release rates of all
subjects before and after the hormone was administered or
(ii) measure the neurotransmitter release rates in two
different groups of similar individuals, one that received the hormone (or naturally possessed it in
high levels) and one that received no hormone and instead got a substance known to have no effect
(a PLACEBO).
! In many types of behavioral experiments, even those where nothing is administered to the subject,
it is common to administer similar handling or even surgery to the control subjects. Such
procedures are the analog of placebos and are often referred to as SHAM PROCEDURES (for
instance -- a sham surgery to some part of the animal believed to have no effect on the variable of
interest but otherwise involving the administration of the same anesthetics, producing similar
discomfort, recovery times etc.
Why are sham procedures considered so important?
3. In both cases above, there are two groups: CONTROLS
who either haven't yet or never will receive the drug and groups that receive the EXPERIMENTAL
manipulation (the drug). Observation of the differences, if any, of response in control and
experimental groups guides the scientist as to which hypothesis to reject. As I am sure you are
aware and as will be discussed in class, statistical analysis is an important part of this process.
3. By contrast, in questions of ultimate causation (where we are interested in the
evolutionary aspects of causation) it if often not possible to perform a direct manipulation. The
events of interest may well have happened long ago. What to do?
a. One possibility is to use the NATURAL EXPERIMENT. In this case
the scientist looks for a series of natural events that present the conditions needed to test a
hypothesis and uses the results of this "already performed" experiment to accept or reject his or her
hypotheses.
b. This is the technique that Darwin used when testing his ideas about
evolution by natural selection after he returned from his famous voyage on the Beagle. Just below,
you will see how this technique is often used in Behavioral Ecology and Ethology.
Observation vs. natural experiment: You may ask yourself, "What is the difference between a
natural experiment and observation -- or for that matter observation and a "normal" experiment?"
The answer is -- not much and the terms may be used interchangeably. But if you really want to use
experiment correctly -- think of it as observation that is directed at testing a hypothesis. By
Science, Philosophy, Methodology and the Nature of "Proof" 4
contrast, observation could simply be the first part of science with the intention of producing an
initial induction or generalization -- not yet part of a rigorous test of a hypothesis.
Thus, observation is always part of experiment but not all observations are connected
with experiments!
II. A BIT MORE ABOUT UNCERTAINTY AND HYPOTHESIS TESTING -- THE ROLE OF
STATISTICS IN MODERN BIOLOGY
(Don't worry -- this will not be a quick course in statistics)
A. One of the fundamental problems in biology is that of variation and how to deal with it
1. In any experiment of any type we can be certain that variation will come from at
least three sources:
a. Genotypic differences between individuals
b. Different environmental experience, including different experimental
manipulations
c. Variation due to SAMPLING ERROR -- that is, BIAS in what we observe
due to the fact that we generally do not see all individuals within a population.
In the Biomedical Sciences that you are doubtlessly most familiar with, great pains are
usually taken to minimize the genetic differences between subjects (HOW WOULD THIS BE
DONE WITH ANIMAL AND WITH HUMAN SUBJECTS?) and often to minimize as many
environmental differences as possible, except for the actual experimental element (HOW ARE
SUCH THINGS DONE?)
By contrast, in NATURALISTIC SCIENCES such as ethology, behavioral ecology,
ecology, environmental physiology, population genetics and evolution it is very important that
the populations reflect natural variation induced by differences in genotype and
environment since the goal of the study is to understand how organism or system works
in its natural setting.
? What does this do to the variability of observation? To the confidence you have in the conclusions
(abstractions derived from scientific study)?
2. Statistics has two major roles:
a. to accurately DESCRIBE the variability of POPULATION by using a
SAMPLE.
--------------------------------------------------------------
IMPORTANT DEFINITIONS:
POPULATION: A group that has been defined by the experimenter. In many cases it will
correspond to some sort of naturally occurring group. In other cases it is used in a more abstract
manner to refer to a group that may differ from others only a single basis.
For instance, in a simple lab experiment, there would generally be at least two potential
populations – the control and experimental groups. As you will see below, it is the role of inferential
statistics to decide the likelihood that two populations were "created" by an experimental
manipulation.
SAMPLE: A subgroup drawn from a population. It is used as an economy so that not all
members of a population need to be measured. It is important in that the values of a properly drawn
sample should closely approximate population values, most of the time.
--------------------------------------------------------------
Science, Philosophy,