Science Concept 1: Scientific Methods

Matthew R. Bennett

The so-called scientific method underpins most scientific research and as such is in
theory relevant to all sub-disciplines of science. But what is it really and what does it
mean for me?

For many sciences it means the systematic observation, measurement, modelling, and
experimentation of ideas leading to the formulation, testing and modification of
hypotheses. Data or observations are made by yourself or others, from which you build
a tentative explanation (or explanations) which you formulate into statements
(hypotheses) which can be compared and evaluated often with additional data.

A hypothesis is a tentative explanation of a series of facts which emerge from data

and can in turn be tested by data. As students you are repeatedly exposed to the idea
of a single hypothesis in statistics so-called null-hypothesis. That is a statement that
summarises the no-difference, no-relationship, no-pattern, and no-story explanation.
It encourages people to think in terms of single hypothesis which has a set of inherent
problems, but more of that later.

Figure 1: The empirical cycle.

Figure 1 summarises the basic scientific method in its generic form and is often
referred to as the empirical cycle. It starts with observation, leading to inferences or
deductions that in turn lead to hypotheses that can be tested and if proven used to
develop general models, theories, or even scientific laws. We can apply this to any
scientific discipline and in Figure 2 it is applied to forensic science and in Figure 3 to
geology/archaeology. You might like to think of how you could modify this diagram for
your own discipline be it healthcare or psychology. Each of these steps requires a
different set of skills and I have listed these on top of the forensic version (Fig. 4).
Note just how often the word communicate features in this diagram. The ability to
communicate scientific ideas is a key element of almost every stage, whether that
 communication be verbal, written, numerical or visual.

Figure 2: Application of the empirical cycle to forensic science.

I have used the term SMART [FAB] tests in Figure 2 which is modified from the
language of performance management where a good line manger should set smart
goals. In our context we have SMART tests:
S – Specific and significant.
M – measurable and meaningful
A – Achievable and action-orientated.
R – Realistic, replicable and results orientated.
T – Time limited and tractable.

To these we can add FAB Tests:

F – falsification versus verification.
A – Accurate and accredited.
B – Best practice that allows precision.
The idea is that when you dream up a test for a specific hypothesis it should meet
these criteria. It should be specific to the hypothesis where possible, you should be
able to measure/observe the subject in a realistic time scales with the resources and
equipment available and it should be repeatable. The last issue is of growing
importance in science. The idea of falsification comes from the work of the scientific
philosopher Karl Popper who said it any amount of data can be used to verify a
hypothesis, but it only takes one observation to prove it wrong. Basically, we should
try to disprove an idea rather than prove it. These are good principles to try and follow
in our work.
 Figure 3: Application of the empirical cycle to areas of archaeology and geology.

Figure 4: Key scientific skills involved in the empirical cycle, note the importance of communication
skills.
Now in discussions of scientific method you often get lots of terms thrown around like
deductive and inductive reasoning. Some will firmly argue that one is better than
another but resist this with all your might – it is always about ‘horses for courses’. So,
let us examine some of these terms.
Deductive Reasoning. Here we move from the general to the particular producing a
conclusion based on the foundation on which it is made. If that foundation is
valid then the conclusion will be true. We are working from Theory to hypothesis,
to data and usually some form of statistical test.
Mythical animals do not exist [ Yes]
Unicorns are mythical animals [Yes]
Therefore, unicorns do not exist [Yes, but do not tell my daughter!]
Inductive Reasoning. Here we move from a specific set of observations to a general
theory. Observations lead to rules and if seen often enough led to a scientific
 law. We are working from Data via analysis and pattern spotting to explanation.
I see an apple falling from a tree [Yes]
I see it happen again and again [Yes]
I develop an explanation which with refinement is elevated to theory [Yes]
For many people inductive reasoning is often seen as a negative thing and to be
avoided -wasteful of resource and potentially wrong. The cartoon in Figure 5
summaries one problem with inductivism, but in truth inductive thinking underpins a
large part of explorative science. Before you can generate a hypothesis to test through
deductive processes you need to gather data and explore a problem, that is use
inductivism. The deductive bit follows when you have a tentative explanation and
hypothesis to explain something. When you are sick you visit your doctor, and you
present a series of symptoms. From this inductive basis they form an idea (hopefully)
of what is wrong and confirm this by further tests such as an ultrasound or blood test.
This is sometimes referred to as Abductive Reasoning.
Abductive Reasoning. Infers the best explanation for a set of observations or facts
even if that explanation includes unobservable facts. It is used by diagnosticians
and detectives.
If it rains, the grass becomes wet [yes]
The grass is wet [yes]
Therefore, it is most likely that it has rained [yes]
The conclusion is only probable since in this case someone may have watered the
lawn. A key part of this approach is something referred to as Occam’s Razor (Fig. 6),
basically you should favour the simplest explanation when given a choice of
explanations.

Figure 5: One problem with inductive reasoning, the unexpected!

 Figure 6: Occam’s Razor and a tongue in cheek version.
No one method is better than another, despite what you will hear from some
researchers. It is also discipline specific psychologists and ecologists like
experimental designs that favour a deductive approach. Take the example in Figure
7A as a typical question in applied science: does the crop supplement increase wheat
yield? You could equally substitute a medicine here. So, the research design involves
setting up trial plots in which the supplement is applied to some plots but not others.
You measure the yield and perform a statistical test to show whether the yield has
increased or not. The same type of experimental design would work with a
pharmaceutical or clinical practice trial.

Figure 7: Two contrasting methodological approaches which fall crudely along the deductive versus
inductive spectrum.
Contrast this with Figure 7B where a farmer has asked you to explain why a particular
patch of his field always gives a poor crop yield. Here you are investigating a specific
problem and an inductive design is required in which gathers data and looks for an
explanation. In this approach preconceived hypotheses can in fact be dangerous
because you may bias an investigation or overlook something important.

Application 1.1: Science of Detection and Beyond

The fictional character Sherlock Holmes has a set of catchphrases of which
‘elementary my Dear Watson, it is simply a matter of deduction’ is perhaps the most
famous. Author Conan Doyle does write a lot about deduction, but there is not a line
anywhere that reads ‘elementary my Dear Watson’. You have to thank Basil
Rathbone’s Hollywood portrayal of Holmes for that (Fig. 8).

Figure 8: Basal Rathbone as Sherlock Holmes he played the great detective between 1939 and 1946.
Copyright: http://www.basilrathbone.net/gallery/sherlockholmes/
A typical statement from Holmes reads like this, in this case from the Study of Scarlet:
From a drop of water a logician could infer the possibility of an Atlantic or a Niagara
without having seen or heard of one or the other. Essentially the form is from a
particular set of observations, what can we infer about the more general circumstances
that relating to them? This is not deduction at all but induction! The blogger Kincaird
(2015) puts it nicely.
Deductive reasoning seeks to prove another statement based on
previously “known” information. An example might be: Sherlock walks
into a dank bedroom with the premise that there have been a rash of
murders in this area, so the woman on the floor is likely another victim of
that same predator, and then he gathers information — blood, nicotine
stains, ash in a corner, a fragment of glass — and narrows the scope of
that available information until his premise is the only logical conclusion
remaining. This is the kind of logic that, in fact, many of the police in the
Sherlock Holmes stories (and our modern world) use; they have a
suspect in mind based on previously available information that is not
necessarily related to the case, and then they seek out evidence to prove
the guilt of that suspect. It’s been proven over time to be a useful tool in
crime-solving and law enforcement, but it sure doesn’t seem like the
Sherlock method.
Inductive reasoning, however, allows Sherlock to extrapolate from the
information observed in order to arrive at conclusions about events that
have not been observed. Here, we see the intrepid detective walking into
the scene a blank slate; he has no pre-supposed ideas about what may
 have taken place. Maybe the woman on the ground died of natural
causes, maybe she was murdered, maybe she fell through the ceiling
from an airplane flying thousands of miles above. He simply doesn’t
know. So, he looks. There’s the ash from her cigarette, so she couldn’t
have fallen from a great height. There’s the glass from a broken window,
but this might have already been broken when she arrived, or it may have
been broken in a struggle. Does she have lacerations? Is her dress damp
from the rain outside or dry? And he goes on to gather information until
he arrives at a conclusion — that still may be incorrect. And yet more
information will come and come until he is quite sure he’s arrived at the
right conclusion.
You might well ask what relevance this has to me as a science student – its all about
a fictional character! Absolutely but it helps us understand the concepts is my
response. So, you have to do a research project as part of your Foundation Degree
and if you proceed further with your studies you will likely do another much large one
before you graduate. In formulating your research design, you will need to think about
the logical design you are using. In Figure 9 I have translated the different logical
reasoning methods into a series of possible project types.

Figure 9: Typical logical design around student projects. In this diagram ‘x’ stands for any subject,
location, or skill.

Application 1.2: Detection versus crime analysis

There is a lot of crime and in general it is on the rise. For quite a while it has no longer
been practical to investigate each and every crime. The alternative is crime analysis
which is also increasingly called intelligence policing. The crime analyst is different
from a detective. They have no badge, no power of arrest, no car, and no truncheon,
but what they do have is data. Data allows them to seek out patterns that allow crime
to be prevented before it happens or accelerate its investigation.

Henderson and Armitage (2018) discuss the cost-benefit of introducing automated

footwear capture in custody suites in London. If you get arrested as you are processed
through a custody suite, they will photograph you, take your fingerprints and in theory
they should take an impression of the soles of your shoes. Traditionally these shoe
impressions were captured by using colourless vegetable dye and paper. The suspect
steps on a sponge of dye and then on to a piece of paper leaving a black and white
impression of their shoeprint. These are then coded – different types of impression
have different descriptors – and the National Footwear Database is searched for
similar shoe impression recovered from local crime scenes. Now this takes time, and
the suspect is often bailed or released before any searches are done. In the
Metropolitan Police area, they now have custody scanners on which the suspect walks
and this captures instantly a shoe impression. This impression is automatically
compared, using machine learning, to stored patterns and potential matches are
returned within a few minutes. If there is a match it does not prove that the suspect
was at a particular crime scene, but it raises the possibility that they were. As such it
is simple enough for additional questions to be asked during a formal interview. You
are arrested for one crime but can be asked about others. So much of policing is
about asking the right question at the right time. The digital custody capture enables
this process to operate sufficiently quickly to be of investigative value. This is the power
of inductive research, in this case done by a machine learning algorithm rather than a
person. This is intelligence lead policing at its best. Why not give the paper by
Henderson and Armitage (2018) a read?

Further Reading
Henderson, J. and Armitage, R., 2018. If the shoe fits: proposing a randomised control
trial on the effect of a digitised in-custody footwear technology compared to a paper-
based footwear method. Crime Security and Society, 1, 2.
[ https://core.ac.uk/download/pdf/151171479.pdf]