Science
Home

References

Take a quiz

The basic goal of science is to understand and explain the behaviour of the objects under study. This goal is for both basic and applied research; in basic research it is the final goal but in applied research it is a prerequisite for application.

It is generally agreed that there are 4 levels of science.

• Level 1: Observation and Facts
• Level 2: Laws
• Level 3: Theories
• Level 4: Models
• Integration of the Levels

Involves the collection of data by measurements which use a standardised procedure. This standardised procedure allows for the measurements to be reliable and replicable.

Facts are related to observations in that they are statements about the results of reproducible procedures for measurement.

Variables are dimensions which need to be quantified in some way in order to properly manipulate and observe the effects of the manipulations in an experiment.

Continuous variables are variables which can have any real number value. These can be divided into ordinal, interval and ratio values.

1. Ordinal variables have values that are ordered but the difference between these values need not necessarily be equal.
2. Interval variables has the same ordered structure but require that the differences between the values are equal.
3. Ratio variables have the characteristics of an ordered scale and equal differences between values as that of interval Variables and has an absolute zero.

Discrete variables have discrete values, such as "yes" or ''no". In cases such as these where there are only two options, they are called dichotomous variables or dichotomies. When the discrete variable value simply names some characteristic of an object e.g. male or female, they are called nominal variables. There is no order in the variables as in continuous variables, the values are simply different. Continuous variables can be made into discrete variables e.g. height in centimetres can be changed into categories of "tall" and "short", but the reverse of changing discrete variables into continuous variables does not hold.

Scientists are interested in the relationship between variables how one variable changes as another changes. To do the above requires that scientists take measurements of an object or behaviour on two variables simultaneously e.g. Pavlovian conditioning has no. pairing and unconditioned response.

Scientists try to state the relationship either verbally or mathematically. Being able to make such statements implies that there is some orderly way in which the behaviour of one variable relates to the behaviour of the other.

Relationships can be causal in which the change in one variable causes a change in a related variable. Usually the directions of the change is critical and the temporal order of causal relationships is also important (i.e. one event must happen before the other can happen). The relationships are not necessarily absolute, but are usually stated as a function of its probability.

Some variables may appear to be related causally (e.g. IQ and shoe size) when in fact these are linked to a third variable (age).

They reflect the two ways in which variables are used.

In causal relationships, a change in the independent variable will lead to a change in a dependent variable. In noncausal relationships, the independent variable is usually the one that is predicted from while the dependent variable is the one that is predicted. These can usually be reversed because the relationship is noncausal but this cannot be done in causal relationships.

3 ways to determine the relationship between variables.

1. Experiments: a set of operations designed to determine the extent and nature of the causal relation between two variables. The independent variable is varied and its effect on the dependent variable is observed. The aim of experiments is to design them in such a way that the effect of the dependent variable can be attributed only to the manipulations of the independent variable - these are controls. Controls will help to eliminate alternative explanations of the results. Controls can be done by randomisation or systematic control of extraneous variables.
2. Natural Experiments: differ from "true" experiments in that the experimenter does not actively manipulate the independent variable. This is necessary when the experimenter cannot control the independent variable e.g. weather. Because of the lack of active manipulations, the conclusions from these experiments are less firm than those from "true" experiments.
3. Pseudo Experiments: the experimenter does not actively vary the independent variable e.g. the independent variable may be a dichotomy based on male and female. These can appear similar to natural experiments; natural experiments appear to differ in that the independent variable is allowed to vary naturally.

Problems in experiments include:

1. Measurement disrupts the behaviour
2. Technological errors related to measuring the independent and dependent variables
3. Intra-individual variability - subjects variation in behaviour over the experiment
4. Individual differences - related to different samples of subjects
5. Rosenthal effects - biases due to experimenter expectations
6. Hawthorne effects - the behaviours of people may differ from whether they know they are being observed or not.

Replication is to copy the experiment exactly how it was done but with a different groups of subjects. Replications are rare.

Laws are general statements that describe or summarise findings in terms that are more general than were the original findings. They describe a stable dependency between an independent variable and a dependent variable.

Laws are the result of induction, or the process of reasoning from specific facts to general principles (generalising). Usually these specific facts come from several separate observations.

Laws must earn its right to be called a law by being able to efficiently encapsulate in a single statements a wide range of separate facts and to be able to predict new facts before they are found.

Laws do not provide an understanding of a behaviour, but they are an important step towards it.

Two different laws may be formulated to explain the same set of facts. The better law of the options is the one that is consistent with all known facts and is most parsimonious or most general in its application. Laws can also be tested by determining whether the predictions they make are true or not- This is because laws allow scientists to predict (the process of deductions).

Theories seek to explain or understand the reasons behind facts or laws. Theories aim to relate a group of laws together by providing a set of processes or mechanisms that seem to account for the relations among the independent variables and the dependent variables. This brings the level of analysis one step deeper than that of laws.

Theories contain the basic elements of

1. Hypothetical constructs: serve as the building blocks for the theory and represent imagined elements. It is defined usually only by a set of postulates.
2. Postulates are statements that relate hypothetical constructs together and thereby provide a definition for them. With constructs and postulates, deductive logic can be used to make specific statements.
3. Co-ordinating definitions: also called operational definitions provide a way for the construct to be defined in terms of actual measurement techniques. The operational definitions allow the postulates to be tested. If the postulates cannot be tested, then the theory they make up is useless.

Theory building: there must be a set of laws and the paradigms which generate them of which they are not under relative dispute. Theories then propose hypothetical underlying causes for these lawful relations. Theories are formed from hypothetical constructs, postulates and co-ordinating definitions which attempt to deduce the laws that exist in an area of study. This process is based on the rules of logic to deduce the specific statements that are the laws existing in a given area (i.e. make predictions). The theory must be able to predict all known laws in an area of study.

Theory testing: predictions are made from the theory that are not already known and so new observations are made to see if the predictions are correct. These predictions thus correspond to derived laws, and a called hypotheses.

Theories may be disputed over whether the operational definitions hold i.e. if the constructs can be measured in way the theory says.

If a theories predictions are confirmed by an experiment, the results are said to support the theory, it is not shown to be "proved" or "correct". All it shows is that there is another result that can be predicted by the theory.

But when a set of observations is not predicted by the theory, the theory is disproved. Thus the most powerful experiments are those that have results which are counter to the predictions from a theory. But a number of experiments are required to show these results because it may have been a result of error. Another problem may arise in that the way the constructs were manipulated in the experiments in a way that is inconsistent with the operational definitions of the theory. Null results occur when the theory predicts an effect on a variable but no effect is observed. Null results may not necessarily show that a theory is incorrect because the experiment may have not been sensitive enough to detect the effect.

It follows that a powerful way to test theories is to design an experiment such that if the theory is correct then there will be no change in the variable but if the theory is incorrect there will a change in the variable. Efficient experiments are able to test two theories at the same time by making them predict opposite results from the same experiment.

Models are analogies to theories which enable the theory to be visualised more easily.

Models have several important functions.

1. They enable new predictions to be made that would have otherwise gone unnoticed.
2. They are good teaching devices.

But models should not be used instead of theories because they do not contain all the relevant information of the theory.

Science systematically eliminates incorrect explanations and creates more correct ones. Other important factors include criticism of work, replication and referencing of submitted articles.

Science aims to acquire knowledge in a way that in unaffected by prejudices and biases. Conducting experiments is one way to achieve this. But because humans are humans, subjective biases can come in.

Empiricism is generally more concerned with collecting facts and observations for their own sake without regard for the theories they are based on. For scientists, these theories are the basis for collecting facts. Although the scientist uses a more efficient method of understanding the world, some argue that it forces scientists to think in a predetermined way and may result in missing some facts that are important for the theory being investigated.

The more applied an experiment is then it has more relevance for some practical problem. Applied research, however, does not have the degree of contribution to theory development as is the case with basic research. Basic research often has more generality than applied research.

Please read the Copyright information.
February, 1998

e-mail: ottmar@psy.uq.edu.au
Web design: David Neumann