|
Article Excerpt
ABSTRACT. In the present paper, I contrast empirical and non-empirical approaches to linguistics by examining the extent to which they practice the scientific method. I provide examples of linguistic analyses that follow and depart from the scientific method and argue that valid explanations about actual language processing rely on adherence to scientific methodology. However, this is not a requirement for philosophical arguments about abstract language structure. Charges of pseudoscience arise when empirical significance is attached to analyses that fail to follow the scientific method. Progress in linguistics is only made to the extent that researchers adopt the method that is standard in scientific endeavors.
1. INTRODUCTION. If you were to randomly select an introductory science textbook the chances are very high that you would find a section dedicated to the scientific method. The steps of the scientific method are usually enumerated in this way:
(1) Observe a phenomenon.
(2) Formulate a hypothesis to explain it.
(3) Carry out an experiment or collect other observations to test the hypothesis.
(4) Analyze the results to determine whether they confirm or refute the hypothesis.
Studies which follow these steps are considered empirical or scientific. Of course, not every field of study is scientific. Literature, philosophy, pure mathematics, formal logic, and art are examples of these. All of them are worthy areas of inquiry even though they are not scientific, and in general they do not purport to be.
In contrast, some fields have all the trappings of a scientific enterprise, but unlike philosophy or literature, for instance, they appear to make scientific assertions. Such enterprises are considered pseudosciences because they make scientific sounding claims but arrive at their conclusions without follow the scientific method. Astrology, extra-sensory perception, UFOlogy, and studies that tout magic diet pills fall into this category. Besides their failure to follow scientific method, pseudosciences demonstrate a number of other characteristics as well (Ruscio 2002): (1)
(1) They ignore contradictory evidence.
(2) Their proponents often react in a hostile manner when their orthodoxy ms challenged.
(3) They use an inordinate amount of technical jargon.
(4) The correctness of their ideas is supported by argumentation, reasoning, intuition, introspection, and reference to authority figures rather than tangible evidence.
(5) Very little new real-world knowledge is produced.
(6) It is impossible to subject their theories to scrutiny.
(6) Explanations are vague and often involve scientific terms used out of context.
My contention in this paper is not to argue that the whole of linguistics is either scientific or pseudoscientific. It is patently unfair to subsume entire fields under one rubric or another since there is a great deal of diversity regarding how individual studies are carried out. I will, in fact, review some instances of pseudoscientific physics. However, the principal focus of this paper is to evaluate specific linguistic analyses and hypotheses in terms of their adherence to the scientific method. It is my hope that a discussion of scientific methodology will highlight the importance of well-designed experimental and quantitative analyses, and how these lead to progress in understanding linguistic processes.
2. OBSERVATION. The first step in the scientific method is to observe some phenomenon. Linguists are masters of observation and hardly an observable generalization is left untouched in some framework. Slips-of-the-tongue, bizarre syntax, and unusual pronunciations rarely go unnoticed either, and partners of linguists often lament that what they say is not heard, but rather, how they say it.
3. HYPOTHESIS FORMATION. The second step in the scientific method is to turn the observation into a testable statement. Observation of a linguistic phenomenon should naturally lead to speculations about why it occurs, how it is formed, and how it may be explained. Such speculation constitutes a scientific hypothesis if it is formulated in such a way that it makes some sort of prediction. Statements in many areas of inquiry such as formal logic and philosophy do not purport to make any kind of prediction about the real world. Pedagogical grammars are written as precise descriptions of a language, but are not designed to predict anything. As a result, such studies exclude themselves from empirical test even though they are valid realms of inquiry. Pseudoscience never predicts what may happen under certain circumstances, but explains what happened after the fact. For example, Freudian psychoanalysis and the 'predictions' of Nostradamus always involve post hoc analyses of events and are not helpful in predicting future behavior or events.
3.1. FALSIFIABILITY. A crucial aspect of a scientific hypothesis is that it must make clear what predicted outcomes would prove the hypothesis wrong. Allow me to borrow an example from Stanovich (1996:21-22). In 1793, there was an outbreak of yellow fever in Philadelphia. A local physician, Benjamin Rush, had the hypothesis that blood letting would cure the fever as long as the disease had not progressed too far. When a patient lived, he counted it as evidence for the efficacy of his practice. However, when a patient died in spite of a bloodletting, he assumed that it was because the malady was too far advanced to yield to bloodletting. It should be clear that Rush's hypothesis was unfalsifiable. No evidence whatsoever could prove it wrong. Both his patients that lived and his patients that died confirmed his hypothesis.
This anecdote exemplifies the requirement that a scientific hypothesis must be subject to possible falsification (Popper 1968). A hypothesis that is not formulated in this way does not lead to scientific progress. While it is easy to dismiss Rush's flawed theory in hindsight, the same sort of irrefutable hypotheses are not hard to find in linguistics. For example, Kahn (1976) formulated a rule of /t/ flapping in American English that, among other criteria such as stress, states that flapping can only occur if the preceding segment is [-consonantal]. This accounts for the fact that the flap may follow either a vowel (e.g. city [sIri]) or [[??]] (e.g. sorted [[TEXT NOT REPRODUCIBLE IN ASCII.] ]). However, flapping is variable after [l] (e.g. altar [[TEXT NOT REPRODUCIBLE IN ASCII.] ] or [alr[??]). To account for this he suggests that [1] must be [+cons] when flapping is not found and [-cons] when it does occur. This sleight-of-hand parallels Rush's in that it effectively shields the hypothesis from possible refutation. It is scientifically useless because it rules nothing out.
Lest one think that such...
|
