ADVERTISEMENTS:
The concept of causality is an extremely complex one and it is not possible to present a thorough analysis of this concept here. Indeed, we may not do better than bring out the basic points necessary for a workable conversance with the concept.
What is a’ cause’? The first point that we must be clear about is that in science the causes which are discovered are, secondary’ or ’caused causes.’ They are only ‘efficient’ causes not the ‘final’ causes. They do not provide an answer to the question, ‘ultimately why?’ Purpose does exist in human affairs, there may be cosmic purposes also; but in science a final cause does not exist.
Francis Bacon decreed that concern for final causes be better left to philosophy. Scientists hold that purpose is not a necessary concept in the research for scientific laws. In sciences, the word cause is used in the sense indicated by J.S. Mill, “a cause which is itself a phenomenon without reference to the ultimate cause of anything. “As Mill puts it, “causation is simply uniform antecedence.”
ADVERTISEMENTS:
But even after gaining a clear understanding that science does not concern itself with, a first cause or a final cause, great ambiguities still remain. Professor Bergeson has pointed out that even in scientific discourse, three different meanings of the term ’cause’ are frequently confused. A cause may act by impelling, by releasing or by unwinding.
The billiard ball that strikes another determines its movement by impelling, the spark that explodes the gunpowder acts by releasing and the gradual relaxing of the spring makes the gramophone turn or unwind the disc, acts by unwinding. Only in the first cause, would the cause seem to explain the effect.
In the other two causes the effect is more or less given in advance and the antecedent invoked is its occasion rather than its cause. In the first case, where the cause acts by impulsion, what is in effect is already in the cause.
In the second cause where cause acts by releasing, it is an indispensable condition; it pulls the trigger, apart from which the effect would not occur. But it does not explain more than the rate or duration of the effect.
ADVERTISEMENTS:
In regard to this immensely complex concept of causality we cannot afford to miss out on the Humean view of causality. A central point of the Humean view is that when someone states that X causes Y he only expresses some reflection in his mind of the material objective world and not the material world directly.
It is as though he is talking of a moving picture of a landscape rather than of the landscape itself. The moving picture may be very public and most of us might agree on what the picture is of. But this moving picture is man-made just as the association or prediction is a product of human mind for it requires an observer to notice the association or to interpret the association.
David Humean did not, of course, insist that there was no real world in which things happen but what Humean is saying is simply that when a scientist observes an association and abstracts from the real world to make some scientific statement, the statement is not the same thing as what he has been observing.
It is a product of his mind or a picture of the world filtered through his perception. This is true of a statement of causality as of every association.
Hume says, “All reasoning concerning matter of fact seem to be founded on relationship of cause and effect,” We judge that the appearance of a table indicates the factual presence of the table on the grounds that presence ’causes’ appearances and we judge that the table is there (if in fact, it is) as a result of the earlier causal chain such as growth of a tree and subsequent actions of a carpenter.
To know the matter of fact involves for Hume the necessity of knowing the causal relation which links them to our perceptions or which links one event to another.
But when we turn to look for this causal relation among the events which we perceive, we find no trace of it. These are mere events; the pattern of events has a certain regularity but we are never able to dictate a relation between events — certainly not a causal one.
We may observe that one event is linked with another through a series of intermediate events or that one event never seems to occur except just before or just after another. Nevertheless they are all events.
The most Hume would concede to, was by characterizing causal relation as possessing three elements, viz., contiguity, succession and constant conjunction — these relations themselves being defined by means of pairs of events both of which must be observed if the relation is to be taken to obtain.
ADVERTISEMENTS:
But this kind of relation is clearly useless in establishing truths about the matters of fact, since one would have to have the matter of fact as well as the perceptions of them in order to appreciate that the former caused the latter.
Unfortunately, we can never get directly at the matters of fact but only at the perceptions of them and hence all fallible empirical knowledge is fillable resting as it does on totally un-provable surmises about the cause of what appears.
The same argument would apply in a slightly different way to the attempts at making predictions about the future on the basis of past observations. The analysis of causality into contiguity, succession and constant conjunction (Hume) has been a centerpiece of controversy.
Many philosophers have felt that the internal necessity which compels one state of affairs to give way to another is clearly open to rational if not to empirical scrutiny. They have thus dismissed Hume’s skeptical conclusion as an unworthy loss of faith in philosophy. But the alternative analyses have brought out that Hume’s intentions were not properly understood.
Hume did not deny that our idea of causality is derived from regularity in experience nor did he doubt that men have a tendency to expect such regularity in future experiences; he only denied that we can have any knowledge other than the experience of regularity itself to base these expectations which are philosophically groundless.
Some other philosophers seem to have felt that logical connections in the realms of thought and of language were so clear that they indicate real connections in the world of precaution and the natural world. Hume admitted the plausibility to this argument to the extent of defining causality as the tendency of our mind to produce the idea of what is called the ‘effect’ when the idea of what is called the ’cause’ is presented to it.
But the fundamental epistemological problem is exactly to discover the grounds on which we may presume the connections and tendencies in the world. Indeed, no theory of causality has succeeded in doing this.
From Aristotle’s Four Causes:
Material, efficient, formal and final to Mill’s inductive method of determining which element of some antecedent situation is to be matched with which element of consequent situation has cause the effect, this parallel between the real and ideal relation has been assumed.
Some theorists like Hegel have tried to identify them but even this is of no help since it leaves us with the question whether our understanding of amalgam of what is actual with what is thought, affords us an accurate representation of it.
Mill’s methods are no doubt an elegant recipe for detecting constant conjunction that Hume speaks of. Mill’s methods of residues and the method of concomitant variation. The first three deal with sets of antecedents and consequents.
If we are looking for the cause of some consequence C and suppose a number of sets of antecedent (A) after each of which C is observed, the method of agreement directs us to look for the cause of C among those antecedents which are members of all the sets.
Alternatively, suppose the consequence C follows after only one of these sets of antecedents, the method of differences directs us to look for the causes among those members of that set which it does not share with any of the other sets which failed to produce C.
The method of residues prompts us to discard from the set of antecedents any elements whose effects are known to be different from the consequence in question and to look for its cause among those which are left out after the operation is competed.
Lastly, the method of concomitant variation directs our search towards the cause of any event or process whose intensity varies with time among other phenomena whose contemporaneous or little prior intensity varies in some simple way with respect to the intensity of the first.
But all these methods in the light of the rule of constant conjunction (Hume) are obvious; they barely prove helpful in solving the Humean problem.
Hume’s answer to his critics was that as an agent he would be quite willing to concede their points but as a philosopher with some share of curiosity he would want to know the foundation of this inference. There may or may not be any internal necessity of linking events in the world and we cannot know whether there is such a linkage or not, but it is only reasonable to behave as if there was.
But then what would an answer to the question of causality come to? The law of the uniformity of nature phrased in causal language says that similar causes are always followed by similar effects and enables us to use the relation past-present as an analogy for present-future.
But suppose it was suddenly revealed to us that this law was about break down and that from tomorrow similar cause may not lead to similar effects.
Now, unless we are told in advance what the differences were going to be, we will have to wait for the change in order to be able to base new kinds of predictions or new kinds of observation. But this activity itself would presuppose the same regularity of causal connection to which the change was offered as a counter-example.
A failure of the principle genuinely would involve complete chaos but one would have no way of knowing about it by the fact that this chaos would extend to our perception and thought.
If all that is argued for is the occasional fallibility of the causal principle then this argument does not hold and we are once again driven to a sceptical impasse. Hence, the solution would seem to lie not in trying to establish the truth of the principle but rather in asserting it.
It is to be noted that in any particular test, the cause and effect to be abstracted from a complex setting or background. Thus, a better formulation of the principle would be that ‘similar causes lead to similar effects if the backgrounds are similar.’
In other words, if other things are equal (ceteris paribus), we may resolve to proceed on the assumption that the causal principle holds but at the same time we may treat with courteous skepticism any claim to have decisively established it.
We may envisage the present state of universe as the effect of its previous state and the cause of what will follow. By causal relationship is meant an effectively productive relationship between antecedent conditions and subsequent results. Hume could discover no such relationship, one merely saw the antecedent conditions and then the subsequent results.
The conclusion of the pristine Humean view is that there is no difference between statement of cause and effect and all other statements of association. But this view is not very satisfactory because social scientists talk and behave as if some associations belong to a different class from other associations.
Many an attempt has been made to propose an inclusive and realistic definition of causality. M. Bunge and Blalock have defined causality mostly by offering synonyms for it.
Causality, say Blalock, is conceived as involving the notion of production, that is, causes produce effects. Production is obviously used as a synonym for causation. But giving synonyms can be useful when one is clarifying what a particular word means in a particular language. Obviously, synonymization does not help solve basic scientific problems of causal labelling.
Definition may be offered by naming some properties of the concept. This type of definition would aim to state, for instance, what causality is. This is an ontological definition in terms of certain material properties of our world.
Such a definition can help us to convey to others a general feeling of what one has in mind. For example, a horse is an animal with four feet used for riding, or culture is a grand pattern comprising habits, customs, thought and adaptive skills acquired by members of a society.
Such a definition of causality has been tried by philosophers for centuries now, without any success. Bridgeman criticizing the sub-definitions argued that defining words in terms of properties creates walls to understanding. Instead, he advocated that definitions should be formulated in terms of operations.
Hume demonstrated typical flaws in ontological definition of causality without offering a substitute definition of causality in terms of operations. Instead, he suggested that the term ‘causality’ was useless and should be dispensed with. This view was one of the most influential ones prevailing among the twentieth century philosophers including Bertrand Russell.
Terms can be defined denotatively, that is, with examples. But one needs more than denotations to clarify the scientific concept of causality.
When there is disagreement among scientists about the application of a term and when they are keen to increase the likelihood that the same terms will be applied to same empirical phenomena, they must turn to operational definitions, i.e., by making sense of a concept by reference to the operations involved.
An operational definition of causality may reasonably be proposed in terms of the following procedure:
(1) The stimulus is varied and variations (if any) in the response observed.
(2) A number of other stimuli are used to observe if same response occurs.
(3) If the above two steps yield appropriate results the relationship between stimulus and response may be called ‘causal.’
Defining causality in situations where structured experiments are not feasible is obviously fraught with hazards. However, a worthwhile operational definition of causality in non-experimental setting would mean that the definition results in many scientists reaching the same judgement. Secondly, the proposed operational definition fits closely the hypothetical concept of causality held by most scientists.
It makes sense to say that causal relationships are a sub-class of associations. In other words, all causal relationships are associations but all associations may not be causal relationships. A causal and effect statement may be understood as a type of scientific explanation but not all explanations are causal statements.
The question now is how to effect a distinction between those associations that are within the sub-class of statement and those that are not. Quite a few attempts have been made to find a method for deciding whether or not a particular association may be included in causal or non-causal associations.
Many writers have opined that association which can be verified experimentally deserves the title causal, no others. Although this has been a useful rule in much of science it cannot be said to be perfect rule. In any experiment some hidden third factor rather than changes in the hypothesized independent variable might be responsible for the changes in the dependent variable.
Besides, many situations do not allow for experimentation. Since a hidden third factor may turn out to be the real cause, a single experiment understandably fails to provide a comprehensive operational definition of causality.
It is necessary under the circumstances, to run related experiments varying different parameters of the situation. It is only after the important possibilities are exhausted in the course of the experiment-series that we can conclude or rightfully say that the experimental stimulus causes the response.
Where experimentation is possible, the operational definition of causality may be proposed as follows:
If the response follows the experimentation stimulus and if this experimental relationship persists even if other elements of the situation are subjected to variation, the observed relationship may be called causal.
Situations in which no experiment is possible and hence the test of experimental confirmation cannot operate as a criterion for defining casual statements which throw up a number of questions about causal statements. Such situations characterizes most social sciences.
Wold (1966) attempted to bring non-experimental situations within the reach of the experimental verification principle by asking whether or not a non-experimental situation is fictitious or the hypothetical experiment.
That is whether the natural situation has in it many of the elements of an actual experiment. But this conceptualization is not without shortcomings. Firstly, the essence of the experiments as an operational definition of causality is that it is the actual observed result of a real experiment that serves to determine whether or not the relationship is to be called causal.
Secondly, the very act of choosing to label relationship as causal is an operation that defines causality. But such an operational” definition lacks validity because it hardly goes far toward in resolving disagreements among people.
Logicians and philosophers have tried various combinations of conditional statements of the ‘if-then’ variety. They have attempted to find out some logical formulation that successfully distinguishes between causal and non-causal associations.
This quest has, however, failed to reach the goal. Yet another type of attempt has recently been made by H. Simon, Blalock and others abstracting from original work of P. Lazarsfeld.
This group has investigated how the correlation between and among three or more variables can help the analysis to sort out which of these variables can be said to cause which. This is an implication and formalization of the analysis seeking to investigate whether a hidden third factor is responsible for correlation between two other variables.
This kind of a study of causal ordering is quite useful and important but does not achieve the purported results. For instance, if the investigator starts with three variables none of which should really be said to be cause of another, the analysis can tell us nothing about whether or not the relationship between two given variables should be called causal.
Schemes of this kind aimed at labelling relationship as causal or non-causal depend heavily upon the use of extraneous knowledge to help us sort out the relationship. For instance, the knowledge that a certain event precedes all the others in time and hence cannot be the effect of these events.
Thus, the whole thing boils down to the assertion that a relationship is causal unless proved otherwise by tests for spuriousness. Such a scheme obviously does not afford an operational definition indicating whether a given relationship should be called causal. At best it can only suggest that within a set of variables one relationship is more causal than another.
An overview of these various attempts leads one to the conclusion, that no definition has been created that fits customary scientific usage, though this is the stated aim of all of them. It is not surprising also that no perfect or near perfect definition has yet been generated. Even the best operational definition does not lead every one to classify all examples of such concepts in exactly the same way.
There always are exceptions at the border line. It is, therefore, quite understandable that such terms as cause and effect which are so very complex and abstract would be much harder to define satisfactorily and would have many more border line cases on which people disagree when classifying situations as causal and non-causal.
Whether or not a situation is closely analogous to a controlled experiment does not provide a complete definition of causality. Furthermore, even in controlled experiment there is often no help for specification error except subject-matter knowledge.
In the light of the above discussion a working definition of causal relationship may be offered as under:
A causal relationship is expressed in a statement that has the following important characteristics: Firstly, it is an association that is strong enough for the observer to believe that it has a predictive (explanatory) power that is great enough to be scientifically useful or interesting.
For example, if the observed correlation is 0.6 even if the sample is large enough to justify the correlation as statistically significant, i.e., unimportant relationships are not likely to be labelled causal. Secondly, the more tightly a relationship is bound, that is, compatible with a general theoretical framework, the stronger its claim to be designated as causal.
Connections with a theoretical framework afford a support to the belief that the side conditions necessary for the statement to hold true are not restricted and that the changes of spurious correction are not substantial; because a statement tends to stand or fall as the rest of the system stands or falls.
It may be noted that the term causal is more likely to have different meaning, to the decision-maker and to the scientist. The decision-maker will call a relationship causal if he expects to be able to manipulate it successfully. For example, smoking may be considered causal by a decision-maker wanting to reduce deaths from diseases statistically related to smoking.
But to the scientists the word cause is likely to mean that the situation does not require further exploration. In the case of cigarettes perhaps only one ingredient in the cigarette does the damage and the scientists searching for this ingredient may choose to withhold the word cause from smoking itself.
The difference in meaning and use of causal concept between decision-making and pure investigation situation is one example of the general proposition that attribution of causality depends upon one’s purpose.
The causal concept is perhaps most necessary for a policy-maker especially when he is considering changing one variable in the hope of achieving change in another variable of, e.g., fertility, in reproduction parlance.
The classification causal and non-causal is an attempt to discriminate between situations that he believes allow such control and those that do not. On the other hand, the causal concept is not at all necessary for a person who is expected to forecast for he has no interest in trying to manipulate the independent variables. The causal concept may or may not be necessary for the pure investigator.
Berttrand Russell and most contemporary physicists seem to believe that it was neither necessary nor useful in the physical/natural sciences. Many non-policy scholars in the social science, however seem to find the concept of causality useful in classifying situations for future research.
The difference among disciplines in respect of the variable to be called causal also illustrates how causal labelling depends upon purpose. In those cases in which variables are complementary, like achievement motivation and investment, it is perhaps unnecessary for the psychologist or economist to deny the causal label to one variable in order to apply it to another variable.
But when the variables are hierarchical then they may be causally incompatible and particular investigators, depending upon their disciplines, must choose which label to give to study and call it causal on the basis of the label they consider most fruitful.
In regard to the meaning of causality as evidenced in social scientific usage of the term, there appears to be considerable consensus among scientists on which relationships are causal and which are not. J.L. Simon proposes an operational definition of causality.
“A statement”, he says, “shall be called causal if the relationship is close enough to be useful or interesting, if it does not require so many statements of side conditions as to get its generality and importance; if enough … third factor variables have been tried to provide some assurance that the relationship is not spurious; and if the relationship can be deductively connected to a larger body of theory or…be supported by a set of auxiliary propositions that explains the mechanism by which the relationship works.”
The above definition is more in the nature of a check list of criteria. Whether or not a given relationship meets the criteria sufficiently to be called causal is neither automatic nor objective. The determination requires judgement and substantive knowledge of the entire context.
It should be clear, therefore, that science goes about its ordained business of events by disclosing their ‘efficient causes.’ This simply means that the event in question is shown to be determined by the preceding events.
The remarks of the philosopher of science, A.E. Taylor, can barely be excelled. He says, “The notion of causation as a transaction between two things is replaced in the experimental science by the conception of it as merely the determination of an event by antecedent events.
As it becomes more apparent that the antecedent events which condition an occurrence are a complex plurality, and include states of what is popularly called the things acted upon as well as the processes in the so-called agent, sciences substitutes for the distinction between ‘agent’ and ‘patient’, the concept of a system of reciprocally dependent interacting factors … the current scientific conception of cause (is thus) the ‘totality of conditions’ in the presence of which an event occurs and in the absence of any member of which it does not occur.
More briefly, causation in the current scientific sense means sequence under definitely known conditions.”
In modern science the emphasis is on a multiplicity of ‘determining conditions’ which together make the occurrence of a given event or effect probable. Scientific thinking is concerned with discovering ‘necessary’ and ‘sufficient’ condition for an effect.
While ‘common sense’ leads one to expect that one factor may provide a complete explanation, the scientist rarely expects to find a single factor or condition that is both necessary and sufficient to bring about an effect.
Rather, he is interested in ‘contributory conditions,’ ‘alternative conditions’, all of which he will expect to find operating to make the occurrence of a given event or effect probable (but not certain). We shall now briefly explain and illustrate the above ‘conditions.’
(a) A necessary condition is said to be one that must occur if the phenomenon of which it is a ’cause’ is to occur, e.g., if X is a necessary condition of Y, then Y will never occur unless X occurs. Such a relationship between X and Y may be designated as ‘producer-product’ relationship. Such ‘producer-product’ relationships are the especial concerns of social and behavioural sciences.
By way of illustration, we may say that differentiation is a necessary condition of social stratification, that is, social stratification would never occur if persons in the course of interaction did not get differentiated.
(b) A sufficient condition is one that is always followed by the phenomenon of which it is a ’cause.’ If X is a sufficient condition of Y, then wherever X occurs Y will always occur. It must be borne in mind that in this strict sense of ’cause-effect’, no object or event can by itself be said to be the cause of another object or event.
The effect that an object or event has on another, always depends on its environment, e.g., mere striking the bell will not cause the subsequent sound if the bell is struck in a vacuum. Such a relationship between X and Y is studied primarily in ‘mechanistic system.’
(c) A contributory condition is one that enhances the likelihood that a given phenomenon will occur but does not make its occurrence certain since it is only one of a number of factors that together determine the occurrence of the given phenomenon.
Some sociological studies have suggested that the absence of a father-figure from the home during childhood is a contributory condition in the generation of drug-addiction among adolescents in the family.
(d) A contingent condition is one under which a given factor is a contributory factor in producing a given phenomenon (effect). In the above example, the contributory condition, i.e., absence of the father figure, may contribute to the incidence of drug-addiction among adolescents only in neighborhoods were the use of drugs is quite pervasive.
In this case, such a neighbourhood is a contingent under which the contributory condition, viz., absence of father-figure, contributes to the probability of occurrence of the ‘effect.’
(e) Alternative conditions are conditions which may all contribute toward the occurrence of a given phenomenon or effect.
In the example cited above, it may be seen that the absence of the father-figure (contributory condition No.1) or the father-figure expressing variously antipathy toward children (contributory condition No.1) both contribute toward producing the effect, i.e., drug-addiction. These conditions are known as the alternative conditions.
It is impossible, to demonstrate directly that a given characteristic or event X determines another characteristics or event Y, either by itself or in conjunction with other characteristics or events.
We are rather in a position of inferring from the observed data that the hypothesis that X is a condition for the occurrence of Y is (or is not) tenable with some particular measure of confidence. Let us now consider what evidence is necessary to justify any inference of causal relationship.
(a) One type of relevant evidence concerns concomitant variation, i.e., the extent to which X and Y occur together or vary together.
Suppose we wish to test the hypothesis that X is contributory condition of Y, we shall have to find out whether the proportion of cause having the characteristic Y is significantly greater among cases having characteristic X than among the cases not having the characteristic X. Unless we can get at such an evidence, we will ordinarily conclude that the hypothesis is not tenable.
Further, if the hypothesis also specifies that the amount of Y is determined by the amount of X, we will have also to find evidence to the effect that, on the whole, those cases that show a higher amount of X also exhibit a higher amount of Y.
Other type of causal hypotheses, e.g., that X is necessary or sufficient ’cause’ of Y or that X as a contingent cause in association with M and an alternate cause with N, would require identifying particular patterns of association between X and Y.
Let us try to understand this with the help of an example. Suppose in a small town a doctor on the basis of his observations, advances the hypothesis that eating of a particular seasonal fruit (X) may lead to severe cold (Y).
An inquiry is then conducted with a view to testing the hypothesis. If consequent upon the inquiry it is found that among those who have it the proportion of those who ate the seasonal fruit (X) was almost equal, we would reject the hypothesis that X leads to Y.
Of course, before rejecting the hypothesis, a careful investigation would need to be carried out with a view to finding out whether eating the seasonal fruit (X) is a contributory condition of cold (Y) under some contingent condition, e.g., say, general debility.
Suppose, the inquiry revealed that the persons who had eaten the fruit and suffered from general weakness were in an overwhelming proportion in the ones suffering from cold, then we can say the seasonal fruit (X) is a contributory condition of severe cold (Y) under the contingent condition of general debility (M).
If, on the other hand, the inquiry indicated that 92% of people suffering from cold had eaten the seasonal fruit and only 25% of the people not suffering from cold had eaten the fruit, we would conclude that the hypothesis that X is the contributory ‘ cause’ of Y is tenable.
It must be remembered that the hypothesis would be simply tenable, not proved, since other possible explanations of the observed relation between X and Y may be invoked and this would be equally tenable, viz.:
(1) Affliction of cold in some way created a craving for the fruit, which means that eating fruits did not lead to cold; it is rather the other way round, i.e., cold (Y) created an urge for eating the fruit (X).
(2) Some other condition (Z) led to both eating the seasonal fruit and having cold.
(3) Yet another condition (W) like impurity which merely happened to be associated with eating the seasonal fruit was responsible for cold, i.e., tap water.
(b) Second type of evidence relevant to inference about causality is the time order of two events X and Y. One event reasonably be considered the cause of the other if it occurs after the other events.
By definition, an effect cannot be produced by an event which occurs only after the effect has taken place. In our example, X cannot be considered the ’cause’ of Y, if as proposed in the alternative hypothesis No. 1, the condition of severe cold (Y) led to a craving for the seasonal fruit (X).
It would be well to remember that time order may not be accepted by some as an automatic test of causality. This argument may be replied by pointing out that just because there is no logical connection, it would not follow that the time lags are no help in establishing causality.
We must recognize after all that to use time lag or time order to infer the direction of causality in a particular relationship is to make use of one of the most general inferences based on all the experimentation that has been undertaken, namely, that actions of the present do not appear to modify the past.
But this is a statistical empirical hypothesis, not without known exceptions. Therefore, to put this inference to sensible use, one needs to adduce other additional reasons to justify that the hypotheses may be believed to apply in a particular case.
It should, also be noted that the occurrence of a causal event may precede or may be contemporaneous with the occurrence of an effect. It is also possible for each factor in the relationship to be both a ’cause’ and an ‘effect’ of the other factor.
This is an instance of the symmetrical causal relationship. George Homan’s hypothesis:
“The higher the rank of a person within a group, the more nearly his activities conform to the norms of the group” typifies the symmetrical causal relationship in as much as the reverse of the hypothesis is also true, i.e., the closer the activities of a person come to the norm, the higher his rank will tend to be.
Although symmetrical causal relationships are frequently found in the realm of social phenomena, it is useful to focus upon the influence of any one factor on the other.
In distinguishing between ’cause’ and ‘effect’ it is useful to establish which of the two events came first, assuming they did not occur simultaneously. Knowing that an increase in rank in a specific instance, preceded an increase in conformity to group norms, we understand that the increase in conformity was not the causal factor.
However, knowledge of temporal priority is not in itself sufficient for inferring causality. In our example, even if we had established for certain that X preceded Y, this was not enough to say that the eating of seasonal fruit (X) caused severe cold (Y).
Two other alternative hypotheses (No. 2 and No. 3) need to be considered, i.e., that some other condition led to both (X) and (Y) or some other condition associated with X was responsible for Y.
(c) We must, therefore, get at the evidence which would establish that no other factor save the hypothesized one (X) was the ’cause’ of the hypothesized effect (Y). Till such time as the evidence ruling out other factor as possible determining condition of the hypothesized effect is secured, we shall not be able to say that X is the ’cause’ of Y.
In our example, it may be that some third factor, e.g., glandular secretions, led both to desire eating the seasonal fruit as also to severe cold. If we can disprove this, the other alternative possibility still remains to be reckoned, i.e., some other factor which merely happened to be associated with eating of seasonal fruit led to cold.
Suppose it was found that people who had bought the fruit from a particular shop where the fruit was kept in the open for a long time were the ones who mostly suffered from cold, whereas the few who had bought from other shop where the fruits were kept in a cold storage mostly did not suffer from cold; then the hypothesis that the seasonal fruit (X) itself was the cause of severe cold (Y) would have to be discarded and attention would be turned on to the effects of the storing system which might have brought about a chemical reaction on fruits in one shop but not in the other.
Under these circumstances, the effect Y would properly be attributed to the chemical factor. It must be stressed that the three kinds of evidence, i.e., concomitant variation, time sequence of variables and evidence ruling out other factor as ’cause’ is or in not cause of the effect. It does not, however, provide any absolute certainty.
That is, we may, on the basis of our evidence, conclude that it is reasonable to believe that X is the ’cause’ of Y but we can never be certain that the relationship has been conclusively demonstrated.
In our above example, the procedures suggested for testing the hypothesis that X is a cause of Y, called for a number of different studies. None of these separate studies could provide a very secure ground for testing the hypothesis because it left the alternative hypotheses unscathed and untested.
An experimental design provides for the gathering of various kinds of evidence simultaneously so that all the alternative hypotheses can be tested. In an experimental test of the hypothesis in our example, the researcher would arrange for a number of subjects to eat the seasonal fruit (‘x) and for a number of comparable subjects not to eat the fruit.
The groups are to be chosen such that they do not differ from each other except by chance, before eating the seasonal fruit. Now comparison of the incidence of cold (Y) in two groups after one group which has not eaten it, would provide evidence of whether eating of the fruit (X) and cold (Y) vary together.
By keeping a careful record of the time of eating the fruit (X) and the time of the ons2 et of cold (Y) the researcher would get the proof as to which of the variables came first.
By introducing ‘controls’ to protect against the possibility that different exposures or experiences during the experiment (other than eating of seasonal fruit or not eating it) which might affect the occurrence of cold, he would ensure that the two groups differ from each other only with respect to (X).
The researcher, in addition, could build into his experiment, the provision for testing hypotheses about particular alternative causal factors. For example, the researcher would test the hypotheses about the effects of storing system by having some of the subjects eat seasonal fruit that had been stored in cold-storage and some eat other fruit (not the seasonal fruit in question) stored in the open.
This would help him “to ascertain whether the ‘open’ storing system alone was productive of (Y) or whether the ‘open storage’ interacted with the seasonal fruit (X) and the product of interaction (V) produced (Y), i.e., cold.
Thus, we see that experimental design wherever it is feasible is the most effective device for testing a causal hypothesis. But then, experiments are not possible to be set up in certain situations.
Suppose a researcher is interested in studying the effects of different methods of child-rearing on the personality structure of a person. He cannot conceivably assign certain children to be brought up in one way, others in another.
In such a case he would have no other alternative but to proceed by locating children who have been brought up in different ways and then assessing their personalities.
Hypotheses about the effect of attributes of the individuals are not often amenable to experimental investigation since the manipulation of the ‘independent’ variable (experimental variable or the factor which has been hypothesized as the ’cause’) is either extremely difficult or impossible. Let us say, we want to see the effect of feeblemindedness (X) on perception (Y).
It would not be possible in this case to manipulate (increase or decrease) feeblemindedness. The only alternative open to us will be to achieve this variation by selecting individuals in whom this variable is present or absent; more or less.
Occasionally, natural situations may provide the desired contrasting conditions (e.g., very high I.Q.) and thus the opportunity for sufficiently rigorous procedures to make possible a reasonably sound basis for inference.
Ordinarily, however, the natural situations are complicated and do not admit of an assumption on the part of the researcher that two or more groups that he has chosen for the purpose of experimentation variable. It is understandable that without a sound basis for such an assumption which a created artificial situation affords. The results of the experiment can have only a doubtful reliability.
Of course, there is no absolute certainty about the validity of inference. No matter how carefully controlled the experiment, there always lurks a possibility that the influence of some factor was not taken into account.
Especially, in social sciences, where there is little knowledge about what factors to control and where many of the relevant factors (e.g., attributes of the individual) are not quite amenable to control, this possibility has to be contended with.