William Labov: A Popperian or not a Popperian?

 

Proceedings of the Tenth International Symposium on Theoretical & Applied Linguistics, April 24-6 1996. Department of Theoretical and Applied Linguistics, School of English, Aristotle University of Thessaloniki.

  

William Labov:  A Popperian or not a Popperian? 

PERICLES A. DALTAS

Department of Linguistics,  University of Ioannina.

  

ABSTRACT 

The epistemological criticism of W. Labov's quantitative paradigm of (socio)linguistic variability has been focused on two issues, namely, whether or not  (a)  variable rules are constructed on a positivist basis, and  (b)  the inclusion of probabilities in them makes theoretical sense.

Adopting a popperian outlook, S. Romaine dismisses the labovian paradigm as positivist on the grounds that Labov has expressed the view that he feels entitled to consider his model right since he has diligently tried time and again to prove it wrong but without success.  However, this kind of rhetorical outburst on the part of theoreticians in favour of their beloved models is quite common (Popper himself does it) and does not in itself jeopardize the theoretical value of such models;  for, their value crucially depends on their content, that is, on whether they are constructed in such a way as to facilitate the most detailed criticism possible.  Judging by the severe and meticulous criticism that has been levelled against the quantitative paradigm over the years, we can safely conclude that this model has been endowed with a high degree of refutability.

Romaine also considers it unpopperian that probabilities are included in labovian rules of variability on the grounds that no finite number of observations would suffice to prove a probabilistic statement wrong, for which purpose an infinite number of observations would be required.  However, Labov's approach to sociolinguistics is commonsense in the popperian sense of the word:  it aims at gradually improving, through criticism, both our understanding of the relevant phenomena and the quality of our theoretical tools rather than striving to secure an unshakeable a priori theoretical basis, i.e., before the actual work of sociolinguistic analysis ever begins.  In this context, probabilities deriving from observed frequencies have to be taken with a sizeable pinch of salt:  they cannot be said to reflect any inherent qualities of the sociolinguistic data since the nature of such qualities is normally beyond the researcher's reach;  yet, they serve as statements open to criticism, i.e. they are reproducible and refutable, and, therefore, may lead to better theoretical models.

 

1.  INTRODUCTION 

The epistemological criticism of W. Labov's quantitative paradigm of (socio)linguistic variability focuses on two issues, namely,  (a)  whether variable rules are constructed on a positivist rather than on a Popperian basis, and  (b)  whether the inclusion of probabilities in them makes theoretical sense.

 

2.  LABOV, POPPER AND POSITIVISM 

Quantitative sociolinguists of the classical Labovian school of thought have been described as positivist (Romaine 1981) in that they seem to proceed from inductive observation to the construction of general models of analysis.  To put it in DeCamp's (1970) apposite terms, such researchers still appear to be in their ΅butterfly collectionΆ stage.

Sir Karl Popper pointed out the inadequacy of the positivist paradigm in the 1930's, when he insisted that ¤there is no induction¥ (1976:86) since all observation takes place in the context of expressly stated or tacitly accepted theories, or, as he characteristically put it (1972:71), ¤all knowledge is theory-impregnated, including our observations¥.  Thus, scientific method does not consist in attempting to corroborate our theories, for no single observation or test can conclusively prove a general theory right:  however many white swans we may spot we can never safely conclude that All swans are white, since the possibility of coming across a swan of a different colour cannot logically be excluded (Popper 1976:43).  On the contrary, scientific progress results from the refutation of our best theories:  a single counter-example is enough to demolish even the best theory, or at least a problematic phenomenon may show a theory to be of limited applicability and mark the beginning of an attempt to improve it.  As a result, the corroboration of a theory takes in effect the form of temporarily surviving the severest possible tests:  the best among a number of alternative theories is the one which up to now has fared better than the competition in the face of criticism by the scientific community;  this, however, does not mean that the theory is certain to do just as well in the next test, or that there may not arise a new theory that does even better.

Popper (1976:132) proposes the following tetradic schema of theory development:

 

P₁    >    TT    >    EE     >    P₂

 

According to it, an attempt at theory formation always starts with a problem (P₁) for whose solution a tentative theory (TT) is formulated.  The next step is to subject the theory to critical examination in order to eliminate possible errors (EE).  Finally, the tentative solution to the initial problem may cause a series of new problems to emerge or lead to the discovery of previously unidentified ones (P₂).  This triggers the reapplication of the schema, not cyclically, for P₂ is different from P₁, but spirally:  problem-solving has now moved on to a new level.

The tetradic schema underlies not only scientific problem-solving but also dealing with problems in our daily lives.  For instance, faced with the need for improved transportation, we built and have since constantly improved the motor car, but our tentative solution to the original problem creates new problems, such as congested roads, air pollution or traffic accidents, which require, at least partly, their own solutions.

The application of the tetradic schema does not require a conscious or even human subject:  all living organisms go through the same four stages in dealing with problems posed by the specific environment in which they live.  From this point of view, "from the amoeba to Einstein is just one step" (Popper 1972:246), as the only difference between the two is that only Einstein is in a position to eliminate his errors exosomatically, i.e. by criticising one theory after another on paper rather than having to either evolve biologically or personally suffer the consequences of his mistakes.

It follows from the Popperian theory of problem-solving that both the growth of our scientific theories and our everyday problem-solving necessarily conform to the above tetradic schema, irrespective of what theoreticians (or ordinary people) think that they are doing when they identify problems and devise solutions.  This means that Labovian sociolinguists may be as positivist as they like in striving to corroborate rather than refute their theoretical approach, i.e., by looking not for counter-examples but for additional examples in support of their theories.  This is so because, to begin with, such theoreticians will readily acknowledge problematic cases that may be pointed out to them or that they themselves may come across.  More importantly though, in the context of the scientific community their theorising necessarily conforms to the Popperian tetradic schema since sooner or later it will be subjected to criticism by other (socio)linguists.  Of course, since it too is necessarily imbued in theory, criticism by peers may take ages to become available, may concentrate on only certain aspects of the original theory, or may in other ways prove equally tentative.

The distinction between what a theoretician thinks he is doing and what he is actually doing does not seem to be fully grasped by Romaine (1981:108) in her otherwise correct characterisation as positivist of a statement by Labov (1972:99) that scientific methodology consists in “trying to prove to yourself that you are wrong.  To be right means that you have finally, abjectly, hopelessly failed to prove yourself wrong.”

Indeed, from a Popperian point of view Labov should know that he or anybody else can only temporarily fail to prove his models wrong:  at any particular moment we may safely assume that our models are already wrong, though we may not be in a position as yet to demonstrate their falsity (still, that does not mean that we should make it too easy for critics to demolish our models lest precious features are lost for ever).  However, despite Labov's (all too human) emotional outburst in favour of his model, what really matters is  (a)  the extent to which this model is in a position to describe aspects of sociolinguistic reality that as yet other models cannot describe or cannot as well, and  (b)  its content.

With respect to (a) above, it seems to me that the Labovian quantitative paradigm, though less fashionable of late than it used to be, is still in a unique position to allow us to think of a sociolinguistic setup in terms of a system of rules.  Furthermore, as I argue elsewhere (Daltas 1996), the paradigm could be profitably developed to account for the fuzzy edges that such rules exhibit in actual use, i.e., when handled by members of the community at various degrees of proficiency or conformity to the norm.

It should also be stressed that, unlike more qualitative approaches, the Labovian quantitative paradigm allows us to make objective observations and reach testable conclusions that linguistics cannot do without.  The reason for this is that, as Labov revealed to us, not only the layman but also the linguist is an unreliable source of data of even his own mother tongue, let alone the 5,000 or so languages of the world, as he has no direct or conscious access to large chunks of linguistic structure.  As a result, methodological expediency obliges him to go through the inductive-looking stage of ΅butterfly collectionΆ, i.e., to act as if he were inductively building new general theories or improving existing ones.  Yet, as we saw above, however objective his observations or testable his conclusions, they are both necessarily steeped in theory.

We said above that Romaine (1981:108) is justified in pointing out Labov's un-Popperian belief that he might be right.  Indeed, the essence of Popper's teaching is that we can consider a theory successful only with respect to the past, for it is always possible to fail the next test that a member of the scientific community may devise on the basis on new data.  However, Romaine's criticism is of limited value since it is not addressed to all theoreticians that do not always adhere to the Popperian principle of refutation, rather than corroboration, as the organon of theoretical progress. 

One important example of this is Popper himself, who in his writings, apart from acknowledging his shortcomings time and again, also expresses his delight at being right on a great number of occasions.  Perhaps he is at his most jubilant, and quite rightly so, when he states (Popper 1972:1): 

“I think that I have solved a major philosophical problem:  the problem of induction.” 

at the beginning of his seminal article with the telling (for the purposes of the present argument) title "Conjectural Knowledge:  My Solution of the Problem of Induction".  The next two periods, which complete the first paragraph of his article, are phrased in the same triumphant, and therefore un-Popperian, certainty: 

“(I must have reached the solution in 1927 or thereabouts.)  This solution has been extremely fruitful, and it has enabled me to solve a good number of other philosophical problems.” 

This brings us to point (b) above, i.e. a theory's content (rather than the love its creator feels for it) as a criterion of evaluation.  According to Popper (1976:79, 86), content is directly proportional to a theory's refutability, i.e., its openness to the severest possible criticism: 

“Thus the whole problem of scientific method cleared itself up, and with it the problem of scientific progress.  Progress consisted in moving towards theories which tell us more and more - theories of ever greater content.  But the more a theory says the more it excludes or forbids, and the greater are the opportunities for falsifying it.  So a theory with greater content is one which can be more severely tested. [...] scientific progress turned out not to consist in the accumulation of observations but in the overthrow of less good theories and their replacement by better ones, in particular by theories of greater content.” (op.cit.:79) 

It would appear therefore that Romaine's criticism against Labov (which in the interests of fairness should have been addressed to Popper too) for his positivist belief that he is right is more rhetorical than anything else.  We would have had a point in considering Labov not to be a Popperian if it could be demonstrated that the Labovian paradigm is so constructed as to be immune to criticism.  Yet, the severe and multi-faceted criticism levelled against this model over the years by Romaine herself and by numerous other critics is in fact evidence of its high degree of refutability, and therefore of its fundamentally Popperian quality (for a review see Ντάλτας 1996).

  

3.  VARIABLE RULES AND PROBABILITIES 

Labov has also been criticised for allowing probabilities to be included in the structural description of his variable rules.  Following Popper's (1972) critique against probability theory, Romaine (1981) correctly points out that formulating in probabilistic terms a theoretical statement concerning the description of a phenomenon severely reduces the degree of refutability of this statement.  This is necessarily so, for no finite number of observations would suffice to demonstrate the falsity of a probabilistic statement:  for this purpose an infinite number of observations would be required.

It has also been considered unsound to derive probabilities from the observed frequencies of a finite language corpus, as these probabilities are often tacitly taken to be independent from the particular corpus and as underlying an unlimited number of language corpora with the same sociolinguistic characteristics.  As Romaine (1981:114) points out, our sociolinguistic behaviour, beyond the physiology of its phonetic side, does not obey laws, as do natural phenomena, but rules.  "Rules may be broken, but laws are never violated." (op.cit.)  It is therefore epistemologically unwarranted to generalize from probabilities deriving from observed frequencies of a particular corpus, however extensive or varied, in order to make predictions concerning other corpora.

However, there are two kinds of law.  There are the laws of nature and there are the laws of science, the latter being theoretical statements concerning the former.  Scientific laws are formulated in the context of specific theoretical schools or traditions, and we are simply hoping that they reflect fairly faithfully the laws of objective reality.  This means three things.

To begin with, we simply suppose that natural laws may not be broken - it is theoretically conceivable that they may or that they may under certain conditions.  Indeed, according to one interpretation of the facts, the probabilistic character of the activity of subatomic particles is inherent and does not depend on the influence of independent variables.  If this interpretation of the behaviour of the natural world is correct (Popper (1972:285f.f.) for one does not think it is), then the distinction between natural laws and sociolinguistic rules is not fundamental.  This of course does not mean that the distinction is unimportant, as the intentionality and symbolic character of human actions interferes with the application of rules but not of laws - although in the case of organic nature things are not so simple and at least intentionality cannot be excluded.  Moreover, there is no reason to suppose that the universe, along with its laws, has stopped evolving, i.e. changing.

Secondly, a scientific law, being a hypothetical statement concerning a natural phenomenon, may be wrong.  In this case, there will be a difference between, on the one hand, probabilities calculated on the basis of features of the phenomenon that are taken to be inherent, and on the other, observed frequencies.  Thus, light was supposed to move in a straight line until Einstein showed that a ray passing by a star bends as a result of the pull that the star exerts on it.

Third, as a consequence of the above two points, probabilities of natural phenomena are calculated, as Sterelny (1983:57f.f.) points out, on the basis of their assumed characteristics and not from observations of their behaviour:  we do not need to observe the behaviour of a dice to calculate the probability of any number between one and six coming up in the next x number of throws.  The mathematical calculation will suffice, though an infinite number of observations will be subsequently required to decide whether the dice is fair or not.  Yet, in sociolinguistics, where we are normally ignorant of the inherent characteristics of phenomena, deriving probabilities from observed frequencies, though epistemologically unsound, is an expedient way of formulating refutable hypotheses.

From the above, it becomes clear that Labov is a commonsense theoretician in the Popper's (1972:37) sense of the word:  according to it the commonsense approach to science aims at a gradual improvement of our theories through criticism and not at a priori securing 'certainty'.  Labov (1972b:259) believes that there is no guarantee against error nor need there be any.  In this he is in complete agreement with Popper (op.cit.:41).  For this reason, his programmatic aim is not to secure an unshakeable theoretical basis before he even starts his research, nor to invent a number of grand theories (according to Popper (1976:62, 71), trying to be original interferes with our doing our job properly), but to study language in a more honest and reliable way on a trial and error basis.

To this purpose, that is to achieve a more responsible kind of research, Labov devises World 3 creatures, i.e. explicitly formulated hypotheses inhabiting the third Popperian world (World 1 is inhabited by objects outside us, and World 2 by our psychological states or ideas that have not been made explicit yet).  The sociolinguistic variables and quantitative rules proposed by Labov and the other quantitative sociolinguists are creatures of this type:  they have been formulated in such a way as to facilitate rather than obstruct their critical analysis, as is obvious from the extremely severe criticism that they have actually been subjected to.  In the light of criticism, these hypothetical schemata reveal considerable theoretical weaknesses.  Some of them have been eliminated over the years, others still persist.  For instance, it is far easier to identify phonological or morphological sociolinguistic variables than syntactic or semantic ones.  Also, not all variables are equally amenable to quantitative analysis as the frequencies of some are very unequally distributed in the various environments of the relevant rule (Ντάλτας 1996).  On the other hand, we are now much richer in knowledge than we were thirty years ago with respect to the quantitative structure of sociolinguistic phenomena.  Also, we are much better equipped to deal with linguistic theories that are the product of navel-watching rather than of interaction with World 3 sociolinguistic inhabitants.  The fact that criticism has already drawn the limits of quantitative variationism is evidence of the high degree of refutability of the theory, and therefore of its Popperian underpinnings.

REFERENCES

 

DeCAMP, D. 1970. Is a sociolinguistic theory possible?  In J. Alatis (ed.) 1970. Linguistics and the Teaching of Standard English to Speakers of Other Languages and Dialects. Georgetown University Press. Washington D.C.

LABOV, W. 1972. Some principles of linguistic methodology. Language in Society 1.

LABOV, W. 1972b. Sociolinguistic Patterns. Blackwell. Oxford.

DALTAS, P. 1997. Κοινωνιογλωσσική Μεταβλητότητα: Θεωρητικά Υποδείγματα και Μεθοδολογία της Έρευνας. Επικαιρότητα. Αθήνα. [Sociolinguistic Variability: Theoretical Models and Research Methodology]

POPPER, K.R. 1972. Objective Knowledge: An Evolutionary Approach. Oxford University Press. London.

POPPER, K.R. 1976. Unended Quest: An Intellectual Autobiography. Fontana/Collins. Glasgow.

ROMAINE, S. 1981. The status of variable rules in sociolinguistic theory. Journal of Linguistics 17.

STERELNY, K. 1983. Linguistic theory and variable rules. Language & Communication 3.1.