WHAT MAKES BIOLOGY
"SCIENTIFIC"?

Science as a Study of How the World Works

Explanations can be cultural, religion-based, or scientific, and there are no clear lines separating these - each is strongly influenced by aspects of the other.  The job of a book like this is to try to make clear what a "scientific" approach is, but no human researcher in the world is totally unaffected by the culture they were raised in or later exposed to, or the religions they follow (or reject).

One aspect of the philosophy of Post-Modernism is the idea that every choice we make is based on a personal world-view that affects what we are willing to perceive in the world.  The statement, "I wouldn't have believed it if I hadn't seen it," has a flip side, stated by Ashleigh Brilliant as,  "I wouldn't have seen it if I hadn't believed it."  Scientists like to believe that they are above such earthly influences, but the best scientists like to step back and look at their own biases whenever they make a conclusion, to try to actively detach themselves from their own worldview.  Do you think that is that even possible?

Having acknowledged the overlap of influences, it is possible to detail what is accepted as a scientific approach to explaining the world:  science is an approach that insists upon a fairly rigid structure, used to answer simple questions with controlled procedures designed to weed out answers that also appear but may not answer the question.  Confused yet?  Science is hard to reduce to a short definition;  the best way to explain it is to show you how it is supposed to work.

SCIENTIFIC METHOD AS A WAY TO ANSWER IMPORTANT QUESTIONS

SCIENTIFIC METHOD - OBSERVATION

As mentioned earlier, though, what we observe is based at least somewhat on what we expect:  one person's floating log is another's lake monster.  There are assumptions embedded in our observations that can greatly affect both what we sense and the sense we make of it.

SCIENTIFIC METHOD - HYPOTHESIS

First, a hypothesis should be predictive:  its should be clear enough that someone should be able to say,  "Well, if that explanation is right, then this should happen under these particular conditions."  Say you decide that plants are green because they all share some green chemical critical to their survival.  That would mean that removing the green chemical should kill a plant.  It would also mean that you shouldn't be able to find plants in the world that are not green (of course, to do this "being green" couldn't be used to define what a plant is!).

Second, a hypothesis should be testable:  you ought to be able to check to see if things work the way the predictions expect them to.

In real-world science, some popular ideas can fit the first criterion only, although it helps to be able to imagine some tests.  Many of Albert Einstein's theories about the nature of light, space, and gravity were such that tests could only be imagined, but they were still widely accepted long before technological developments allowed them to really be tested.

Scientific Method is based upon Logic.  This sounds obvious and a fairly easy rule to follow, but a hypothesis that seems logical to one person may not to another, and an "obvious" prediction may only seem obvious to the predictor.  This is especially true in biology, where "If A exists and B exists, then C should happen" are often based upon an incomplete understanding of A, B, and C.  Biology is full of results that amount to, "We checked for C, and we don't think we've found it - in fact, we're not sure what we've found."   Of course, that won't stop the concept of Application of Logic from showing up throughout the rest of the Method.

Hypothesis revisited.  When the testing phase has been done, a researcher looks at the results and decides whether the predictions were supported, but even this decision is a hypothesis.  People like to speak of science as being able to "prove" things, but all it can really do is collect evidence to support hypotheses.  For a very long time, test after test accumulated evidence to support the idea that gravity was an attractive force, and then Einstein came along and described it as a property that bent space, which is now the accepted explanation.  It is the strength of science that no idea is ever absolutely confirmed.  New ideas overturn and replace old ones all the time, or adjust how old ones are viewed.  In the media and fiction, a scientist's unwillingness to speak in absolute certainty is sometimes shown as a weakness, but it shouldn't be.  It's an easy trap - even folks who should know better will often use the "p-word."  (That's prove, folks.)

SCIENTIFIC METHOD - TESTING A HYPOTHESIS WITH CONTINUED OBSERVATIONS

In many instances where controlled experiments are not possible, or at least not practical, predictions are tested by follow-up observations, also called field tests.   The prediction that one should not be able to find non-green plants is really only testable this way - you go out looking extensively in the environment.

This type of test, and experiments as well, have some basic requirements within the Method.  First, it is useful if not essential to try not to test too many things simultaneously - if your hypothesis about plants was that they share a green chemical that is necessary to their life because it allows them to process sunlight but wouldn't work without a water source, that's a valid hypothesis but difficult to test all at once.  It would be easier to make and test predictions based on parts of the hypothesis, fitting the pieces together as the testing phase progresses.  Second, any questionable terms should be defined - if greenness isn't going to define plants in your study, then how will you decide what is a plant?  

A part of good science is that tests should be reproducible by other testers - they need to know how you did the test, and the first part of understanding that is knowing how you defined your terms.  If you are unclear on how you decided a plant is a plant in your study, someone else could do the test with their own assumed definition and include fungi (they used to be considered as types of plants, so that's not very far-fetched) while yours did not, and they will be running a very different test than yours.   Part of making a test reproducible is designing it clearly enough that if someone else runs it, they will get comparable results.   The clearer your methods and descriptions, the better - what are you looking for, and how are you going to look?  If measurements are involved, exactly what will you be measuring and how?  When will you be done looking, and then what are you planning to do with the data you've collected?  "Let's look for some stuff, and then do some stuff with it" is not a good plan.

Most scientists prefer measurements to consist of quantitative data rather than qualitative data.  Quantitative data involves quantities, or discrete numbers;  qualitative data involve quality, based more on feelings or opinions.  In a test of headache medication, it is more useful to have subjects rank their pain, say, on a scale of 1 to 10 (Scale definition is still important - which end has more pain?  Is 10 the worst headache the subject has ever had or the worst headache they could imagine?).  Pain is qualitative, it doesn't have numbered notches, but it will be easier to track the effects of the drugs if you have before-and-after numbers to compare.  In a case like this, the numbers are not really comparable - one person's "3" might be nothing like another's - but if we're looking for the pain numbers to go down as an indication of the drug working,  we have a legitimate measure of effectiveness.

SCIENTIFIC METHOD - TESTING A HYPOTHESIS WITH A CONTROLLED EXPERIMENT

Like a system of follow-up observations, and experiment should test a prediction based upon a hypothesis, and again, the more limited and specific the prediction, the clearer the results.  It will be necessary as well to set up tests that someone else should be able to reproduce, so clearly stating the terms and methods, understanding ahead of time how measurements will be made and data recorded, and how the data might be analyzed are all aspects of good science.  Often medical experiments reported on in the media seem to be contradictory (and sometimes they are), but often the studies were answering similar questions in different ways.  For instance, two studies on the effectiveness of long-term treatment of heart attack patients seemed to have conflicting results, but close inspection would reveal that one study looked at recurrence of heart attacks, which dropped, and the other looked at mortality, which remained the same.  From the two tests together, the treatments seem to lessen a person's chances of having another heart attack without changing their likelihood of dying.

So a test is designed.  By classic Scientific Method, what is hoped for is an experiment that allows for a control test:  a duplicate of the experimental test, with allowances made confounding factors, aspects that are not what you are testing but which might be changing your results.  At its simplest, all confounding factors can be accounted for by isolating what exactly is being tested as a variable.  In a test for headache medication, the medication in the treatment would be the variable:  the experimental test would involve giving the medication to people with headaches (how would you measure the effectiveness of such a drug?), while the control test would involve giving equivalent treatments but with the medication removed.  If the only difference between the two tests is exactly what is being tested, in this case the medication, any differences in the results of the tests can be linked to just the medication.  Any other differences confuses the issue:  if you gave your experimental subjects pills but your controls nothing, how could you tell if just the act of being given a pill might affect a headache?  Humans do react to just the act of being treated - this reaction is called the placebo effect  (a placebo is a false treatment, and is used in conteol tests).

Usually in biological research, a group of organisms or some other complex system has something done to it, and a specific type of response is looked for.  For instance, a test could be set up to extract the green chemical from a group of plants and then check their growth (Terms! - What would you measure about a plant to determine growth?  How many different ways could "growth" be measured?) for a period of time.  Aha!  They all not only stopped growing, but by the end of your study they were all dead!  That green chemical must be critical for their growth and survival.

Or is it?  Experiments can be full of confounding factors.  One common type of confounding factor is called an artifact:  this is some aspect of the study process itself that produces results independent of what we are testing.  In the case of our plants, the treatment we used to extract the green might have killed them itself.  It might be impossible to ever determine whether the death of our plants wasn't poisoning rather than lack of green.

Many experiments cannot be designed with a clear-cut control;  our green chemical test could be controlled if we could magically remove the green and disturb no other aspect of the plant, but that can't be done.  The best that can be done is test the various parts of the extraction process and try to figure how much a plant is hurt by them.

Designing a good experiment is as much an art as a science.  It requires imagination, because many confounding factors are not immediately obvious, and trying to control for them may be challenging.  Often there is a level of guesswork and cost-benefit analysis - you may decide that one factor's impact would be too small to worry about, or another has to be accepted and worked with in the conclusion stage because actually dealing with it would be impractical (many experiments on humans would be more reliable if subjects could be locked away with every bit of their lives controlled, but researchers know they cannot do this).  It is almost a given that, after all is done, someone looking to criticize your work can come up with a valid confounding factor that you either didn't anticipate or that you had accepted but dealt with.

SOME EXAMPLES OF CONFOUNDING FACTORS

You may hear that aluminum pots and utensils can give people Alzheimer's disease.  This goes back to studies that found aluminum residue in the brain tissue of Alzheimer's patients, residue that was not present in controls' (people without Alzheimer's) tissue.  But what turned out to be happening is that something about the chemistry of Alzheimer's tissue - the disease definitely changes brain chemistry - pulled aluminum out of the preserving fluid, while normal brain tissue did not - the aluminum had not been there until the tissues had been removed from corpses and preserved!  Oddly enough, not everyone agrees that the aluminum was completely an artifact, and an aluminum-Alzheimer's connection is still being studied, with unclear results.

Studies involving groups of humans are notoriously difficult to control.  How do you match exactly two groups of people for everything except one variable?  For instance, a hypothesis linking Sudden Infant Death Syndrome (SIDS) to the cultural practice of sleeping with infants compared the U.S., where parents rarely sleep with infants, to an African nation where the practice is common, and found that the SIDS rate was much higher in the U.S.  One important confounding factor involves the data used, which were cause-of-death statistics:  it seems like SIDS is rarely considered as a cause-of-death by medical personnel in the African nation, even when no clear cause is present, so it's very unlikely to be recorded as such.  The SIDS differences could be just a reflection of this records-keeping artifact.

Human expectations can often be a confounding factor, and sometimes expectations are not even controlled for because everyone accepts them.  It is widely reported, and accepted in several scientific areas, that crack cocaine has profound effects on the behavior of infants, so-called "crack babies."  This has become an acceptable premise, an assumption when looking for effects on their later development, but no one had really tested the idea that crack babies really are different from regular ones.  Finally, someone tested the premise:  nurses and caregivers in a hospital nursery unit were given the task of sorting the crack babies from the regular ones without knowing ahead of time which were born to addicts.  The hospital personnel were sure they would be able to do this, were certain that the behavioral differences would make it easy, but were no better able to choose which were which than a control flipping a coin (often, controls involving choice are not a classic control test, but a way of choosing randomly;  a real choice should not duplicate a random one).

The placebo effect was recognized fairly quickly when research into medical treatments became integrated with the Scientific Method, and it was quickly decided that patients should have no knowledge of whether they were in a study group or a control group - they are informed that they are in a case-control test, but are not told anything beyond that.  These were blind studies, sometimes called single-blind tests.  It was much later that the hypothesis arose that the researchers administering the treatments, knowing who was in which group, might treat the patients in each group differently.  Would your attitude be different giving an experimental treatment to one patient and a false treatment to another?  Even when the researchers tried to control themselves, follow-up observations suggested that subtle signs crept through that might signal to patients which groups they were in.  Modern medical tests often are conducted as double-blind tests, in which the researchers in direct contact with patients do not themselves know which sort of treatment they are giving out;  treatments are assigned randomly by researchers one level removed from the procedures and tracked by code numbers.  As far as the treating personnel are concerned, there is only the one group.

The effects of numbers.  In biology especially, there is enough variety among individual organisms or even laboratory systems that any single individual or set-up might be unusual and produce unusual results.  The more individuals that can be used and "blended" into a statistical response, the less probable that blind chance will have a significant effect on outcomes.  You know that people you know have traits that aren't exactly universal human traits, and it would not be scientific to use just one of them to make pronouncements about humanity as a whole.  Single- or limited-case evidence is called anecdotal evidence and is a driving force of the "health supplement and herbs" industry, where a couple of success stories are used to imply that everyone can benefit from their product.  Good scientific studies involve large numbers of test subjects or many repetitions, so odd rare results have little impact on the overall data.

SCIENTIFIC METHOD - PERFORMING EXPERIMENTS

All data must be recorded in an organized fashion according to the experimental design.  A good experimenter also keeps extensive notes as they go along, relying on their training and powers of observation to notice things beyond what is being officially measured.  The best researchers record everything that they notice, without trying to pick out what might be important.  It may turn out later that a seemingly minor happening has major significance - using our green chemical example, it could happen that a small notation about how the extraction process is working might later supply a big clue to just exactly what the green chemical is.

In biology especially, often things happen during the experiment that were not anticipated in the experimental design, and decisions have to be made about how to adapt.  Good scientists have to be good reactors and good decision-makers.

SCIENTIFIC METHOD AND EXPERIMENTAL RESULTS

The control becomes significant at this stage as a comparison - if both your experimental results and your control results are roughly the same, then the variable you were testing had no real effect.  If your headache medication results were an average of a 3.2 drop on your 10-point pain scale but your placebo group had an average of a 3.1 drop, your drug isn't very effective over just any old pill.

Statistics are an extremely important aspect of modern science.  This type of math can allow comparisons of things or groups that seem incomparable - an average or ratio can be used to compare groups of very different sizes, for instance.  There are very many ways to mathematically manipulate data, and statistics are famous both for its ability to find patterns that are not obvious and, unfortunately, for the flexibility they afford someone desperate to wring some sort of support out of their data.  This is why statistical analysis should be part of the starting design - if you plan ahead of time the best way to get useful information out of your results, you can resist the temptation later to just keep trying different statistical approaches until the results look like they are "supposed" to.

Conclusions, as mentioned earlier, are actually a new set of hypotheses built upon the results of the tests.  When you say, "This happened and this is what it means," others might think that your results indicate something entirely different.  As you become trained, you will find published studies where your own interpretation may be so contrary to the researchers' that it won't seem to you like they really looked at their own results.

MODERN SCIENCE AND THE NULL HYPOTHESIS

However, a researcher new to a field could certainly benefit by being able to see where the dead-ends are, and soon there may be a Journal devoted to negative results, at least in the biomedical field.

MATH AND MISINTERPRETATION OR MISREPRESENTATION

SCIENCE AND PUBLISHING

Peer review can also happen at other steps of the process, such as:  when funds are needed for research, it can be the first step;  in large laboratories, there may be peer oversight at many stages;  ideas are presented at conferences and receive input.  The concept of review by others in the field is important.

Virtually every field of study, and most subfields, has a journal devoted to that brand of research.  Many are available, at least partially, online.

SOME BASIC SCIENCE TERMS AND BIOLOGY USAGE

In other science courses, you may have been given lists of terms and definitions;  the terms will show up in biology sources, but the definitions may not be the same.  The term hypothesis, for instance, has a particular meaning in the Scientific Method as a working explanation that can be tested, very much a preliminary idea, and you may have been taught that a theory is like a hypothesis but widely-accepted as the probable explanation for some phenomenon.  You may also have learned that a law is a rule, an explanation or feature that always works a certain way, without exception.  NONE of those terms can be reliably tracked in biology - sometimes they fit, as in Cell Theory, sometimes they don't, as in the widely-accepted Gaia Hypothesis.  And there are virtually no laws in the classic sense in biology, because somewhere some living thing is breaking whatever rule someone has tried to stretch across the breadth of every organism on the planet.

This can be very disturbing to some students who like the feeling that the world can be reduced down to regular rules.  The best that you can do is get used to the idea that it doesn't work that way with biology.  Or shift to physics, although quantum physics seems as crazy as biology sometimes.  There are a couple of reasons why much of modern biology in performed on the molecular level, and one of those reasons is that molecules behave better than organisms.