Observation
Some of us are more easily satisfied than others. Each of usis more easily satisfied on some issues than on others: moreeasily satisfied on the issues that matter less. But there isa limit: when we get down to our own direct observation,there is nowhere deeper to look. Someone else's report, evenof direct observation, has not quite this finality for us. Wemay have good reason to trust such a report, but in trustingit we are making an inference from other evidence, otherobservations of our own. What we are directly observing israther the report itself, the spoken or written words. Wehave then to draw on past linguistic experience of theveracity of this or other speakers. A memory or evena written record of our own direct observation is still atsome remove from the original observation itself, thought wecan seldom ask better. What we are directly observing in thecase of our own written record is, like our friend's report,only inferentially related to our original observation;thought in our own case we have the best of reasons to abideby the inference.
Thus the ultimate evidence that our whole system ofbeliefs has to answer up consists strictly of our own directobservations - including our observations of our notes andother people reports. Naturally we leave many pointsunchecked. Lore is handed down from our forebears. Suchactual evidence as any one of us does have, however, is inthe end the direct evidence of the senses. Likewise suchevidence as there is and ever was, collectively, for thewhole overwhelming edifice of science, has consisted only inthe direct evidence of many peoples' senses.
The world with its quarks and chromosomes, its distantlands and spiral nebulae, is like a vast computer in a blackbox, forever sealed except for its input and outputregisters. These we directly observe, and in the light ofthem we speculate on the structure of the machine, theuniverse. Thus it is that we think up the quarks andchromosomes, the distant lands and the nebulae; they wouldaccount for the observable data. When an observation turnsout expectedly, we may try modifying our theory of thatstructure at one or another point.
When an observation shows that a system of beliefs mustbe overhauled, it leaves us to choose which of thoseinterlocking beliefs to revise; this important fact has comeup repeatedly. The beliefs face the tribunal of observationnot singly but in a body. But note now that the observationsentence itself, the sentence that reports or predictsa present or imminent observation, is peculiar on this score.It does face the tribunal singly, in the usual case, andsimply stands or falls with the observation that it reportsor predicts. And, standing or falling, it sustains or letsdown the system of beliefs that implied it.
What are observations? Some philosophers have taken themto be sensory events: the occurrence of smells, feels, noises,color patches. This way lies frustration. What we ordinarilynotice and testify to are rather the objects and events outin the world. It is to these that our very language isgeared, because language is a social institution, learnedfrom other people who share the scene to which the wordsrefer. Observation sentences, like theoretical sentences, arefor the most part sentences about external objects. This iswhy they can enter into logical relations with scientifictheory, confirming or refuting it.
In an early page we asked what sorts of things wereobjects of belief. Then we gratefully dropped that question,noticing that we could instead talk of sentences and ofbelieving them true. Now a similar maneuver conduces toclarity in dealing with the notion of observation: let us askno longer what counts as an observation, but turn rather tolanguage and ask what counts as an observation sentence.
What makes a sentence an observation sentence is notwhat sort of event or situation it describes, but how itdescribes it. Thus, I may see the dean of the law school maila birthday check to his daughter in Belgium. Saying so inthese terms does not qualify as an observation sentence. Ifon the other hand I describe that same event by saying thatI saw a stout man with a broad face, a gray moustache,rimless spectacles, a Homburg hat, ant a walking stick,putting a small white flat flimsy object into the slot of amailbox, this is an observation sentence. What makes it anobservation sentence is that any second witness would bebound to agree with me on all points then and there, grantedmerely an understanding of my language. The witness would notbe bound to agree that it was the dean, whom he or she mightnot know, nor expected to know anything about the check ora daughter in Belgium.
In short, an observation sentence is something that wecan depend on other witnesses to agree to at the time of theevent or situation described. A witness might of courseforget and give divergent testimony later, or might fail tonotice a feature at the time until it was pointed out. Butthe witness can check and assent if asked at the time. Thereason for such agreement is that the terms used in anobservation sentence are terms that we can all apply to theirobjects on sight: terms like "mailbox", "stout man", "graymoustache", "rimless spectacles", "Homburg hat", Walkingstick". They are terms unlike "dean of the law school","birthday", "daughter in Belgium"; for in applying theseterms to the present situation we depend on past experiencesthat few have been privileged to share.
"The cat is on the mat" qualifies as an observationsentence. "My cat is on the mat" does not, on our definition,since another witness might not know whose cat it was. Evenour observation sentence may sometimes be truthfully utteredwithout reporting a present observation; thus "The cat is onthe mat" may sometimes express a belief based on earlierobservation or mere hearsay. In calling it an observationsentence we mean that it is a form of words that can be usedto report a present event or situation, and that otherwitnesses can then be counted on to concur if queried at thetime.
It is easy to see why some of our sentences are bound tobe of this kind, if we reflect on how we learn language. Someterms, and short sentences containing them, are learned inthe sensible presence of something that the term describes,or in the circumstances that the sentence reports. This wayof learning expressions is what philosophers call ostensive.It is a simple matter of learning to associate the heardwords with things simultaneously observed - a matter, asmodern psychologists put it, of conditioning. Thus, we mayventure to volunteer or assent to the word "yellow" in thepresence of something yellow, on hearing others do so. Thisway of responding will be reinforced, as psychologists say bysocial approval or successful communication, and so becomehabitual. The part of language that we learn first must belearned ostensively, thus not depending on otherlanguage-learning.
Further vocabulary is acquired afterward by processesthat depend on prior acquisitions. Learning by ostensiondepends on prior acquisitions. By ostension we learn to useand react to observation sentences.
Typical observation sentences are about bodies: "This istable", "This table is square," "The cat is on the mat".Always the situation that makes an observation sentence truewill be a situation that is intersubjectively observable;that is, it will be sort of situation to which multiplewitness could, if present, attest. Further, it will bea situation that the witnesses can witness one another'switnessing of. These crucial traits are assured by thedistinctive nature of ostension. The learner of the languagehas to be able to observe the relevant situationsimultaneously with hearing the veteran speaker affirm thesentence, and must also be able to observe that thespeaker's affirming of the sentence is accompanied byobservations of that same situation. Correspondingly, theveteran speaker who ventures to judge the learner'sperformance has to be able to observe that the learner, whenaffirming the sentence, is observing the appropriatesituation.
There are two traits of observation sentences which,when considered side by side, invite a philosophicalquestion. The distinguishing feature of observation sentencesis that they can be checked on the spot. Yet these sentencesare commonly about enduring bodies - cats, mats, tables. Howis this possible? That there are enduring bodies at all,behind the passing show of sensory appearance, is a point ofphysical theory - a rudimentary point, but still somethingbeyond the observable present occasion. How then cana sentence about bodies be at the same time an observationsentence, for which the whole occasion for affirmation is theobservable present?
This puzzle comes of viewing the matter from wrong end.The special virtue of observation sentences is that we can,in principle, learn them by ostension as wholes, keyed aswholes to the appropriate observable occasions, before everlearning to link the component words to enduring bodies. "Thecat is on the mat" can be learned ostensively as a unitarystring of syllables in association with a certain range ofpossible scenes. All of us necessarily learned someobservation sentences thus. Then, as we gradually caught onto the theory of enduring bodies, we came to treat some ofthe component words as referring to bodies. Learning byostension, as a trained animal might, to associate wholeobservation sentences with appropriate patterns ofstimulation, is a first indispensable step toward learningphysical theory. we get on into the theory afterward, bit bybit, as we learn to dismember the observation sentences andmake further use of their component words. It is to thisprimary, ostensive learning of observation sentences aswholes that physical theory itself owes its vital continuingconnection with sensory evidence.
Probably none of us in fact learned "The cat is on themat" outright by ostension, but we could have. A likelierexample is "(This is a ) ball", or "Yellow". An importanttrait of language is that people learn it by differentroutes, and do record of the route is preserved in the wordslearned. What makes a sentence an observation sentence is notthat it was learned ostensively but that it is of a sort thatcould have been. And what sort is that? We already said: itis a sentence whose whole occasion of affirmation, nearlyenough, is the intersubjectively observable present occasion.This is a straightforward trait attaching to some sentencesand not others. And it is a trait that is socially traceable,for what it comes to is just that all speakers of thelanguage, nearly enough, will assert to the sentence underthe same concurrent stimulations. "The table is square" and"The cat is on the mat" will pas this test and so qualify asobservation sentences. "This is a bachelor" will not qualifyas an observation sentence, since one of two tested speakersmay happen to know that the man pointed to is a bachelorwhile the other does not.
Ostension accounts for our acquisition of only modestpart of our language. A major source is an elaborate andlargely unconscious process of abstraction andgeneralization, working partly from what we have previouslylearned by ostension and depending heavily on imitation ofobserved use. We guess the force of one sentence by noting itsuse in relation to other sentences; we grasp the use ofa word by abstraction from sentences in which it turns up;and we learn how to build new sentences by copying thestructure observed in old ones. There is much that could besaid, and much more still to be learned, about these methods.
A less mysterious form that such derivative acquisitioncan take, though not the most frequent, is definition. Thesimplest form of definition, in turn, is that in which the newexpression is equated outright to some expression that ispresumed to have been already intelligible. Thus, if wesuppose the words "parent", "brother", "married", and "man"already to have been somehow acquired, we might explain"uncle" and "bachelor" by equating them to "parent's brother"and "unmarried man". Other definition are contextual; in thesethe new expression is not equated to anything outright, butsystematic instructions are given for translating all desiredsentences containing the expression. For instance, we mightdefine "brother", not by formulating any direct substitutefor the word by itself, but by systematically explaining allsentences in which the word occurs followed by "of".This wecould do by translating "brother of x" as "male other thanx whose parents are parents of x". Or again, we might definethe connective "if and only if", not outright, but bysystematically explaining all the compound sentences that areobtained by putting "if and only if" between sentences. Wesimply explain "p if and only if q" as "if p then q and ifq then p".
Observation sentences are the bottom edge of language,where it touches experience: where speech is conditioned tostimulation. It is ultimately through them that language ingeneral gains its meaning, its bearing of reality. This iswhy it is they that convey the basic evidence for all belief,all scientific theory. They play this fundamental role notonly when someone is checking over his or her beliefs after aprediction has gone wrong, but equally when someone ismarshaling evidence for a belief that has been challenged bycolleague. And it is here that the social trait just nowattributed to observation sentences is crucial - that allspeakers assent to such a sentence under the samestimulations. As dissident theorists converge towardobservation sentences, they converge to agreement.
An observation may be made by an individual; but, as wehave emphasized, the truth of the observation sentence is anintersubjective matter. Here a favorite old irrationalistdoctrine finds both its seductiveness and its rebuttal. Thehoary view contends that truth is relative to believer;there's truth for me and truth for you, and theirreconciliation is generally neither possible nor desirable.Now the variable ownership of acts of observation might becited in support of this doctrine. For haven't we said thatobservations are the ultimate basis for belief systems? Andcan't e be expected, you and I, to make differentobservations? Maybe so; but which observation sentences aretrue will not thus depend on either of us, nor on any otherobserver. The firmness of our respective grounds foraccepting a given observation sentence may vary, and so mayour appraisal of that sentence; but its truth cannot.Happily, we need not acquiesce in the ultimacy ofdisagreement in order to appreciate its sources.Intersubjective conflict, to be sure, is unlikeintersubjective conflict in one important way: the former,but not like latter, may be recognized as such withoutthereby gaining impetus for giving way. But where yourbeliefs and mine are mutually inconsistent we cannot both beright, any more than I alone can be right in each of severalincompatible beliefs.
So we all contribute, with our respective observations,to the knowledge that we all share. We find here an elementthat at once makes science hard and makes it possible. It ishard because it must build a coherent system from the diverseevidence gleaned and reported by people of different times,places, cultures, and interests; it is possible because thereis thus so much to draw on.
Are observation sentences infallible? Nearly, if we setaside those offered disingenuously and those uttered byspeakers who have not quite learned the language. It wouldstrain the very meaning of the words, in such sentences, tosuppose any appreciable fallibility; for the words arethemselves acquired through the association of observationsentences with the observable circumstances of theirutterance.
A trace of fallibility, indeed, there is. Normally,observation is the tug that tows the ship of theory; but inan extreme case the theory pulls so hard that observationyields. It can happen that a theory has long goneunchallenged, neatly conforming to countless relevantobservations on every hand, and that now one observationconflicts with it. Chances are that we will waive the onewayward observation. This still does not mean going back onour definition of an observation sentence. We defined it asa sentence to which all witnesses are bound to accede at thetime of the observed event; we left them free to change theirminds afterward. In the cases where we waive an observation- and they had better be pretty special - we are changing ourminds after the occasion, or, more usually, doubting someoneelse's report.
It is never a matter of rejecting an observation sentenceon the occasion of the observation. And an observationsentence ceases to be an observation sentence, after all,when we change the tense of its verb. Reports of pastobservations involve inference, as lately remarked. It isonly these, and not strictly observations sentences, that weare second-guessing when we waive the wayward observation.
despite any such legalisms, however, our memories arenot to be lightly dismissed, much less our records. Nor arethe reports of observation by trusted colleagues, though thetrust in this quarter admits of degrees. It is only a strongand long unchallenged theory that will occasionally resistthe adverse testimony of a remembered a recordered orreported observation. In such an extremity we may attributethe wayward evidence to unexplained interference, even tohallucination. If such alleged cases of hallucination tend tocluster in a few persons, who may then be seen as prone tohallucination, so much the better for our scientificconscience. There is then hope of accommodating the verywaywardness of those wayward observations in a theory too,a theory of psychopathology. Law may thus be sought in theapparent breaches of law.
Even when observations persist in conflicting with atheory, the theory will not necessarily be abandonedforthwith. It will linger until a plausible substitute isfound; the conflicting observations will stand unexplainedand the sense of crisis will mount.
"Galileo's contribution to the study of motion dependedclosely upon difficulties discovered in Aristotle's theory byscholastic critics. Newton's new theory of light and colororiginated in the discovery that none of the existing ...theories would account for the length of the spectrum, and thewave theory that replaced Newton's was announced in the midstof growing concern about anomalies in the relation ofdiffraction and polarization effects to Newton's theory.Thermodynamics was born from the collision of two existingnineteenth-century physical theories, and quantum mechanicsfrom variety of difficulties surrounding black-bodyradiation, specific heats, and the photoelectric effect.Furthermore, in all these cases except that of Newton theawareness of anomaly had lasted so long and penetrated sodeep that one can appropriately describe the fields affectedby it as in a state of growing crisis. Because it demands ...major shifts in the problems and techniques of normalscience, the emergence of new theories is generally precededby a period of pronounced professional insecurity. As onemight expect, that insecurity is generated by the persistentfailure of the puzzles of normal science to come out as theyshould. Failure of existing rules is the prelude to a searchfor new ones." (T.S.Kuhn, The structure of ScientificRevolutions, The University of Chicago, 1962 pp. 67.)
There are some points at which, without deliberateconsideration of theories, all of us find it second nature toedit observation. We learn to take it that sticks appearingbent while partially immersed in water should in fact bejudged straight. We learn not to suppose that the moon islarger when near the horizon than when higher in the sky.When the colors before us begin to vibrate, we do not imaginethat the proprieties of light have changed. But in all theseexamples, again, we are at pains in the end to accommodatethe waywardness of observation in a theory too. The illusionof the immersed sticks is covered by physical theory ofrefraction; the illusion of the low moon is coped with bysome psychological hypotheses; and a general visualdisruption is apt to set us speculating about something weate or drank. Observations thus stubbornly retain theirprimacy. They remain the boundary conditions of our body ofbeliefs.
It must be confessed however that not all observations,or reports of observations, are so conscientiouslyaccommodated. Some of them, uncongenial to existing theory,get passed over with even less acknowledgment than it wouldtake to rate them as hallucinations. Persistent reports ofoccult experiences receive this short treatment, as also, oflate, many of the reports of unidentified flying objects.Note, however, that a good scientist does not treat anuncongenial observation in this high-handed way when theobservation is induced by an experiment of his own. For hisexperiment will have been designed for the very purpose ofdeciding between two alternative moves in the development ofhis theory, two preconceived alternative beliefs. But he willperhaps dismiss a puzzling observation, reported to him withpalpable sincerity or even made by himself, if he has in mindno specific change of theory that might accommodate theobservation and still jibe with previous data. Up to a pointthis highhandedness is justifiable. If a scientist were tointerrupt existing projects in order to find plausiblehypothesis for every puzzling experience outside thelaboratory, and if he were to lend a patient and judiciousear to every crank and gossip, he would learn less.
Scientists are so good nowadays at discovering truth thatit is trivial to condone their methods and absurd tocriticize them. At the same time it is evident that waivingobservations is always a delicate business. A theory that issustained only at the cost of systematic waiving is anundependable instrument of prediction and not a good exampleof scientific method.
Just because it is not feasible to accommodate allobservations all the time, some philosophers have wanted toscout the whole idea of observation. Their doubts have beenaggravated by a further consideration: the air ofsubjectivity that seems to them to render the very idea ofobservation hopelessly vague. Where the untrained eyeobserves a wired metal box, the trained eye observesa condenser. Where the untrained eye observes nothing, thetrained eye observes the recent trace of a deer. But againthese discrepancies are no ground for misgivings whenproperly viewed; they are only a play on the careless use of aword.
For philosophical purposes the notion of observation, andof observation sentence, needs to be taken with anunimaginative literalness. A straightforward criterion to thepurpose is already before us: that all reasonably competentspeakers of the language be disposed, if asked, to assent tothe sentence under the same simulations of their sensorysurfaces. On this criterion "That's a condenser" simply doesnot count as an observation sentence, trained eyenotwithstanding. Naturally the experts, being reasonable,will stop pressing for further evidence anyway as soon asthey can agree. The can agree that it is a condenser, so theystop there, rather than press on compulsively to genuineobservation sentences, in our sense of the term; but theyalways could press on. If they care to use the termobservation for their intermediate stopping point, let us notdispute about the term. They might be said to be simplynarrowing the category of "competent speakers of thelanguage" to their specialized group.
We remarked that some philosophers have identifiedobservations with events of sensation. It is thus not to bewondered at that, in some philosophical writings, the titleof observation sentence is reserved for sentences verydifferent from observation sentences as we have defined them.It is reserved for introspective reports such as "I am inpain" and "I seem to see blue now". Such reports also havebeen rated as infallible. It must be conceded that they tendto be incontestable, because of the speaker's privilegedaccess to his or her private experience. But on this verypoint they differ diametrically from observation sentences inour sense. The situations that make them true are not ones towhich multiple witnesses could attest. What is open to publicobservation in such a case is rather the introspective reportitself. What is comparable to the cat's being on the mat isnot the person's feeling pain or seeing blue, but thereporting pain or blue - the verbal behavior. This verbalbehavior is indeed available as a datum for furthertheorizing; it is a datum to which multiple witnesses mightattest.
