Book Review: The Structure Of Scientific Revolutions

When I hear sci­en­tists talk about Thomas Kuhn, he sounds very rea­son­able. Scien­tists have the­o­ries that guide their work. Some­times they run into things their the­o­ries can’t ex­plain. Then some ge­nius de­vel­ops a new the­ory, and sci­en­tists are guided by that one. So the cy­cle re­peats, knowl­edge gained with ev­ery step.

When I hear philoso­phers talk about Thomas Kuhn, he sounds like a mad­man. There is no such thing as ground-level truth! Only the­ory! No ob­jec­tive sense-data! Only the­ory! No ba­sis for ac­cept­ing or re­ject­ing any the­ory over any other! Only the­ory! No sci­en­tists! Only the­o­ries, wear­ing lab coats and fake beards, hop­ing no­body will no­tice the cha­rade!

I de­cided to read Kuhn’s The Struc­ture Of Scien­tific Revolu­tions in or­der to un­der­stand this bet­ter. Hav­ing finished, I have come to a con­clu­sion: yup, I can see why this book causes so much con­fu­sion.

At first Kuhn’s the­sis ap­pears sim­ple, maybe even ob­vi­ous. I found my­self wor­ry­ing at times that he was knock­ing down a straw man, al­though of course we have to read the his­tory of philos­o­phy back­wards and re­mem­ber that Kuhn may already be in the wa­ter sup­ply, so to speak. He ar­gues against a sim­plis­tic view of sci­ence in which it is merely the grad­ual ac­cu­mu­la­tion of facts. So Aris­to­tle dis­cov­ered a few true facts, Gal­ileo added a few more on, then New­ton dis­cov­ered a few more, and now we have very many facts in­deed.

In this model, good sci­ence can­not dis­agree with other good sci­ence. You’re ei­ther wrong – as var­i­ous pseu­do­scien­tists and failed sci­en­tists have been through­out his­tory, posit­ing false ideas like “the brain is only there to cool the blood” or “the sun or­bits the earth”. Or you’re right, your ideas are en­shrined in the Sacristry Of Set­tled Science, and your facts join the ac­cu­mu­lated store that passes through the ages.

Sim­ple-ver­sion-of-Kuhn says this isn’t true. Science isn’t just facts. It’s paradigms – whole ways of look­ing at the world. Without a paradigm, sci­en­tists wouldn’t know what facts to gather, how to col­lect them, or what to do with them once they had them. With a paradigm, sci­en­tists gather and pro­cess facts in the ways the paradigm sug­gests (“nor­mal sci­ence”). Even­tu­ally, this pro­cess runs into a hitch – ap­par­ent con­tra­dic­tions, or things that don’t quite fit pre­dic­tions, or just a gi­ant ugly mess of epicy­cles. Some ge­nius de­vel­ops a new paradigm (“paradigm shift” or “sci­en­tific rev­olu­tion”). Then the pro­cess be­gins again. Facts can be ac­cu­mu­lated within a paradigm. And many of the facts ac­cu­mu­lated in one paradigm can sur­vive, with only slight trans­la­tion effort, into a new paradigm. But sci­en­tific progress is the story of one rel­a­tively-suc­cess­ful and gen­uinely-sci­en­tific effort giv­ing way to a differ­ent and con­tra­dic­tory rel­a­tively-suc­cess­ful and gen­uinely-sci­en­tific effort. It’s the story of sci­en­tists con­stantly toss­ing out one an­other’s work and be­gin­ning anew.

This gets awk­ward be­cause paradigms look a lot like facts. The atomic the­ory – the cur­rent paradigm in a lot of chem­istry – looks a lot like the fact “ev­ery­thing is made of atoms and molecules”. But this is only the ice­berg’s tip. Once you have atomic the­ory, chem­istry starts look­ing a lot differ­ent. Your first ques­tion when con­fronted with an un­known chem­i­cal is “what is the molec­u­lar struc­ture?” and you have pretty good ideas for how to figure this out. You are not par­tic­u­larly in­ter­ested in the sur­face ap­pear­ance of chem­i­cals, since you know that iron and silver can look al­ike but are to­tally differ­ent el­e­ments; you may be much more in­ter­ested in the weight ra­tio at which two chem­i­cals re­act (which might seem to the un­ini­ti­ated like a pretty ran­dom and silly thing to care about). If con­fronted with a gas, you might ask things like “which gas is it?” as op­posed to think­ing all gases are the same thing, or won­der­ing what it would even mean for two gases to be differ­ent. You can even think things like “this is a mix­ture of two differ­ent types of gas” with­out ag­o­niz­ing about how a perfectly uniform sub­stance can be a mix­ture of any­thing. If some­one asks you “How no­ble and close to God would say this chem­i­cal sam­ple is?” you can tell them that this is not re­ally a le­gi­t­i­mate chem­i­cal ques­tion, un­less you mean “no­ble” in the sense of the no­ble gases. If some­one tells you a cer­tain chem­i­cal is toxic be­cause tox­i­c­ity is a fun­da­men­tal prop­erty of its essence, you can tell them that no, it prob­a­bly has to do with some re­ac­tion it causes or fails to cause with chem­i­cals in the body. And if some­one tells you that a cer­tain chem­i­cal has changed into a differ­ent chem­i­cal be­cause it got colder, you can tell them that cold might have done some­thing to it, it might even have caused it to re­act with the air or some­thing, but chem­i­cals don’t change into other chem­i­cals in a fun­da­men­tal way just be­cause of the tem­per­a­ture. None of these things are ob­vi­ous. All of them are hard-won dis­cov­er­ies.

A field with­out paradigms looks like the STEM supremacist’s stereo­type of philos­o­phy. There are all kinds of differ­ent schools – Kan­ti­ans, Aris­totelians, Lock­eans – who all dis­agree with each other. There may be progress within a school – some Aris­totelian may come up with a re­ally cool new Aris­totelian way to look at bioethics, and all the other Aris­totelians may agree that it’s great – but the field as a whole does not progress. Peo­ple will talk past one an­other; the Aris­totelian can go on all day about the telos of the em­bryo, but the util­i­tar­ian is just go­ing to ask what the hell a telos is, why any­one would think em­bryos have one, and how many utils the em­bryo is bring­ing peo­ple. “De­bates” be­tween the Aris­totelian and the util­i­tar­ian may not be liter­ally im­pos­si­ble, but they are go­ing to have to go all the way to first prin­ci­ples, in a way that never works. Kuhn in­ter­est­ingly dis­misses these ar­eas as “the fields where peo­ple write books” – if you want to say any­thing, you might as well ad­dress it to a pop­u­lar au­di­ence for all the good other peo­ple’s pre-ex­ist­ing knowl­edge will do you, and you may have to spend hun­dreds of pages ex­plain­ing your en­tire sys­tem from the ground up. He throws all the so­cial sci­ences in this bin – you may read Freud, Sk­in­ner, and Beck in­stead of Aris­to­tle, Locke, and Kant, but it’s the same situ­a­tion.

A real sci­ence is one where ev­ery­one agrees on a sin­gle paradigm. New­to­ni­anism and Ein­stei­ni­anism are the same kind of things as Aris­totelianism and util­i­tar­i­anism; but in 1850, ev­ery­body be­lieved the former, and in 1950, the lat­ter.

I got con­fused by this – is Aris­totelian philos­o­phy a sci­ence? Would it be one if the Aris­totelians forced ev­ery non-Aris­totelian philoso­pher out of the academy, so that 100% of philoso­phers fell in line be­hind Aris­to­tle? I think Kuhn’s an­swer to this is that it’s tel­ling that Aris­totelians haven’t been able to do this (at least not lately); ei­ther Aris­to­tle’s the­o­ries are too weak, or philos­o­phy too in­tractable. But all physi­cists unite be­hind Ein­stein in a way that all philoso­phers can­not be­hind Aris­to­tle. Be­cause of this, all physi­cists mean more or less the same thing when they talk about “space” and “time”, and they can work to­gether on ex­plain­ing these con­cepts with­out con­stantly ar­gu­ing to each other about what they mean or whether they’re the right way to think about things at all (and a New­to­nian and Ein­stei­nian would not be able to do this with each other, any more than an Aris­totelian and util­i­tar­ian).

So how does sci­ence set­tle on a sin­gle paradigm when other fields can’t? Is this the part where we ad­mit it’s be­cause sci­ence has ob­jec­tive truth so you can just set­tle ques­tions with ex­per­i­ments?

This is very much not that part. Kuhn doesn’t think it’s any­where near that sim­ple, for a few rea­sons.

First, there is rarely a sin­gle ex­per­i­ment that one paradigm fails and an­other passes. Rather, there are dozens of ex­per­i­ments. One paradigm does bet­ter on some, the other paradigm does bet­ter on oth­ers, and ev­ery­one ar­gues over which ones should or shouldn’t count.

For ex­am­ple, one might try to test the Coper­ni­can vs. Ptole­maic wor­ld­views by ob­serv­ing the par­al­lax of the fixed stars over the course of a year. Coper­ni­cus pre­dicts it should be visi­ble; Ptolemy pre­dicts it shouldn’t be. It isn’t, which means ei­ther the Earth is fixed and un­mov­ing, or the stars are un­ut­ter­ably uni­mag­in­ably im­mensely im­pos­si­bly far away. No­body ex­pected the stars to be that far away, so ad­van­tage Ptolemy. Mean­while, the Coper­ni­cans posit far-off stars in or­der to save their paradigm. What looked like a test to se­lect one paradigm or the other has turned into a wedge push­ing the two paradigms even fur­ther apart.

What looks like a de­ci­sive vic­tory to one side may look like ran­dom noise to an­other. Did you know weird tech­nolog­i­cally ad­vanced ar­ti­facts are some­times found en­cased in rocks that our cur­rent un­der­stand­ing of ge­ol­ogy says are mil­lions of years old? Creation­ists have no trou­ble ex­plain­ing those – the rocks are much younger, and the ar­ti­facts were prob­a­bly planted by nephilim. Evolu­tion­ists have no idea how to ex­plain those, and de­fault to things like “the ar­ti­facts are hoaxes” or “the min­ers were re­ally care­less and a screw slipped from their pocket into the rock vein while they were min­ing”. I’m an evolu­tion­ist and I agree the ar­ti­facts are prob­a­bly hoaxes or mis­takes, even when there is no par­tic­u­lar ev­i­dence that they are. Mean­while, prob­a­bly cre­ation­ists say that some fos­sil or other in­com­pat­i­ble with cre­ation­ism is a hoax or a mis­take. But that means the “find some­thing pre­dicted by one paradigm but not the other, and then the failed the­ory comes crash­ing down” over­sim­plifi­ca­tion doesn’t work. Find some­thing pre­dicted by one paradigm but not the other, and of­ten the pro­po­nents of the dis­ad­van­taged paradigm can – and should – just shrug and say “what­ever”.

In 1870, flat-earther Sa­muel Row­botham performed a se­ries of ex­per­i­ments to show the Earth could not be a globe. In the most fa­mous, he placed sev­eral flags miles apart along a perfectly straight canal. Then he looked through a telescope and was able to see all of them in a row, even though the fur­thest should have been hid­den by the Earth’s cur­va­ture. Hav­ing done so, he con­cluded the Earth was flat, and the spher­i­cal-earth paradigm de­bunked. Alfred Wal­lace (more fa­mous for pre-empt­ing Dar­win on evolu­tion) took up the challenge, and showed that the bend­ing of light rays by at­mo­spheric re­frac­tion ex­plained Row­botham’s re­sult. It turns out that light rays curve down­ward at a rate equal to the cur­va­ture of the Earth’s sur­face! Luck­ily for Wal­lace, re­frac­tion was already a known phe­nomenon; if not, it would have been the same kind of wedge-be­tween-paradigms as the Coper­ni­cans hav­ing to change the dis­tance to the fixed stars.

It is all nice and well to say “Sure, it looks like your paradigm is right, but once we ad­just for this new idea about the dis­tance to the stars /​ the re­frac­tion of light, the ev­i­dence ac­tu­ally sup­ports my paradigm”. But the sup­port­ers of old paradigms can do that too! The Ptole­maics are rightly mocked for adding epicy­cle af­ter epicy­cle un­til their sys­tem gave the right re­sult. But to a hos­tile ob­server, posit­ing re­frac­tion effects that ex­actly coun­ter­bal­ance the cur­va­ture of the Earth sure looks like adding epicy­cles. At some point a new paradigm will win out, and its “epicy­cles” will look like perfectly rea­son­able ad­just­ments for re­al­ity’s sur­pris­ing amount of de­tail. And the old paradigm will lose, and its “epicy­cles” will look like ob­vi­ous kludges to cover up that it never re­ally worked. Be­fore that hap­pens…well, good luck.

Se­cond, two paradigms may not even ad­dress or care about the same ques­tions.

Let’s go back to util­i­tar­i­anism vs. Aris­totelianism. Many peo­ple as­so­ci­ate util­i­tar­i­anism with the trol­ley prob­lem, which is in­deed a good way to think about some of the is­sues in­volved. It might be tempt­ing for a util­i­tar­ian to think of Aris­totelian ethics as hav­ing some differ­ent an­swer to the trol­ley prob­lem. Maybe it does, I don’t know. But Aris­to­tle doesn’t talk about how he would solve what­ever the 4th-cen­tury BC equiv­a­lent of the trol­ley prob­lem was. He talks more about “what is the true mean­ing of jus­tice?” and stuff like that. While you can twist Aris­to­tle into hav­ing an opinion on trol­leys, he’s not re­ally op­ti­miz­ing for that. And while you can make util­i­tar­i­anism have some idea what the true mean­ing of jus­tice is, it’s not re­ally op­ti­mized for that ei­ther.

An Aris­totelian can say their paradigm is best, be­cause it does a great job ex­pli­cat­ing all the lit­tle types and sub­types of jus­tice. A util­i­tar­ian can say their paradigm is best, be­cause it does a great job tel­ling you how to act in var­i­ous con­trived moral dilem­mas.

It’s ac­tu­ally even worse than this. The clos­est thing I can think of to an an­cient Greek moral dilemma is the story of Anti­gone. Anti­gone’s un­cle de­clares that her traitorous dead brother may not be buried with the proper rites. Anti­gone is torn be­tween her duty to obey her un­cle, and her de­sire to honor her dead brother. Utili­tar­i­anism is…not re­ally de­signed for this sort of moral dilemma. Is ig­nor­ing her fam­ily squab­bles and try­ing to cure ty­phus an op­tion? No?

But then util­i­tar­i­anism’s prob­lems are deeper than just “comes to a differ­ent con­clu­sion than an­cient Greek morals would have”. The util­i­tar­ian’s job isn’t to change the an­cient Greek’s mind about the an­swer to a cer­tain prob­lem. It’s to con­vince him to stop car­ing about ba­si­cally all the prob­lems he cares about, and care about differ­ent prob­lems in­stead.

Third, two paradigms may dis­agree on what kind of an­swers are al­lowed, or what counts as solv­ing a prob­lem.

Kuhn talks about the 17th cen­tury “dor­mi­tive po­tency” dis­course. Aris­to­tle tended to ex­plain phe­nom­ena by ap­peal­ing to essences; trees grew be­cause it was “in their na­ture” to grow. Descartes gets a bad rap for in­vent­ing du­al­ism, but this is un­de­served – what he was re­ally do­ing was in­vent­ing the con­cept of “mat­ter” as we un­der­stand it, a what-you-see-is-what-you-get kind of stuff with no hid­den essences that re­sponds me­chan­i­cally to forces (and once you have this idea, you nat­u­rally need some other kind of stuff to be the mind). With Carte­sian mat­ter firmly in place, ev­ery­one made fun of Aris­to­tle for think­ing he had “solved” the “why do trees grow?” ques­tion by an­swer­ing “be­cause it is in their na­ture”, and this cli­maxed with the play­wright Moliere por­tray­ing a buf­foon­ish doc­tor who claimed to have dis­cov­ered how opium put peo­ple to sleep – it was be­cause it had a dor­mi­tive po­tency!

In Aris­to­tle’s view of mat­ter, say­ing “be­cause it’s their essence” suc­cess­fully an­swers ques­tions like “why do trees grow?”. The Carte­sian paradigm for­bade this kind of an­swer, and so many pre­vi­ously “solved” prob­lems like why trees grow be­came mys­te­ri­ous again – a step back­wards, sort of. For Descartes, you were only al­lowed to an­swer ques­tions if you could ex­plain how purely-me­chan­i­cal mat­ter smash­ing against other purely-me­chan­i­cal mat­ter in a billiard-ball-like way could pro­duce an effect; a more vir­tu­ous and Descartes-aware doc­tor ex­plained opium’s prop­er­ties by say­ing opium cor­pus­cles must have a sand­pa­per-like shape that smooths the neu­rons!

Then New­ton dis­cov­ered grav­ity and caused an up­roar. Grav­ity posits no cor­pus­cles jostling other cor­pus­cles. It sounds al­most Aris­totelian: “It is the na­ture of mat­ter to at­tract other mat­ter”. New­ton was de­nounced as try­ing to smug­gle oc­cultism into sci­ence. How much do you dis­count a the­ory for hav­ing oc­cult el­e­ments? If some con­cep­tion of quan­tum the­ory pre­dicts the data beau­tifully, but says mat­ter be­haves differ­ently de­pend­ing on whether some­one’s watch­ing it or not, is that okay? What if it says that a cer­tain elec­tron has a 50% chance of be­ing in a cer­tain place, full stop, and there is no con­ceiv­able ex­pla­na­tion for which of the two pos­si­bil­ities is re­al­ized, and you’re not even al­lowed to ask the ques­tion? What if my ex­pla­na­tion for dark mat­ter is “in­visi­ble grem­lins”? How do you figure out when you need to re­lax your as­sump­tions about what counts as sci­ence, ver­sus when some­body is just cheat­ing?

A less dra­matic ex­am­ple: Lavoisier’s the­ory of com­bus­tion boasts an abil­ity to ex­plain why some sub­stances gain weight when burned; they are ab­sorb­ing oxy­gen from the air. A brilli­ant ex­am­ple of an anomaly ex­plained, which proves the su­pe­ri­or­ity of com­bus­tion the­ory to other paradigms that can­not ac­count for the phe­nomenon? No – “things shouldn’t ran­domly gain weight” comes to us as a prin­ci­ple of the chem­i­cal rev­olu­tion of which Lavoisier was a part:

In the sev­en­teenth cen­tury, [an ex­pla­na­tion of weight gain] seemed un­nec­es­sary to most chemists. If chem­i­cal re­ac­tions could al­ter the vol­ume, color, and tex­ture of the in­gre­di­ents, why should they not al­ter weight as well? Weight was not always taken to be the mea­sure of quan­tity of mat­ter. Be­sides, weight-gain on roast­ing re­mained an iso­lated phe­nomenon. Most nat­u­ral bod­ies (eg wood) lose weight on roast­ing as the phlo­gis­ton the­ory was later to say they should.

In pre­vi­ous paradigms, weight gain wasn’t even an anomaly to be ex­plained. It was just a perfectly okay thing that might hap­pen. It’s only within the con­stel­la­tion of new meth­ods and rules we learned around Lavoisier’s time, that Lavoisier’s the­o­ries solved any­thing at all.

So how do sci­en­tists ever switch paradigms?

Kuhn thinks it’s kind of an ugly pro­cess. It starts with ex­as­per­a­tion; the old paradigm is clearly in­ad­e­quate. Progress is stag­nat­ing.

Aware­ness [of the in­ad­e­quacy of geo­cen­tric as­tron­omy] did come. By the thir­teenth cen­tury Alfonso X could pro­claim that if God had con­sulted him when cre­at­ing the uni­verse, he would have re­ceived good ad­vice. In the six­teenth cen­tury, Coper­ni­cus’ coworker, Domenico da No­vara, held that no sys­tem so cum­ber­some and in­ac­cu­rate as the Ptole­maic had be­come could pos­si­bly be true of na­ture. And Coper­ni­cus him­self wrote in the Pre­face to the De Revolu­tionibus that the as­tro­nom­i­cal tra­di­tion he in­her­ited had fi­nally cre­ated only a mon­ster.

Then some­one pro­poses a new paradigm. In its origi­nal form, it is woe­fully un­der­speci­fied, bad at match­ing re­al­ity, and only beats the old paradigm in a few test cases. For what­ever rea­son, a few peo­ple jump on board. Some­times the new paradigm is sim­ply more math­e­mat­i­cally el­e­gant, more beau­tiful. Other times it’s petty things, like a French­man in­vented the old paradigm and a Ger­man the new one, and you’re Ger­man. Some­times it’s just that there’s noth­ing bet­ter. Th­ese peo­ple grad­u­ally ex­pand the new paradigm to cover more and more cases. At some point, the new paradigm ex­plains things a lit­tle bet­ter than the old paradigm. Some of its pre­dic­tions are spook­ily good. The old paradigm is never con­clu­sively de­bunked. But the new paradigm now has enough ad­van­tages that more and more peo­ple hop on the band­wagon. Grad­u­ally the old paradigm be­comes a laugh­ing­stock, peo­ple for­get the con­text in which it ever made sense, and it is re­mem­bered only as a bunch of jokes about dor­mi­tive po­tency.

But now that it’s been adopted and ex­panded and reached the zenith of its power, this is the point at which we can ad­mit it’s ob­jec­tively bet­ter, right?

For a bet­ter treat­ment of this ques­tion than I can give, see Samz­dat’s Science Can­not Count To Red. But my im­pres­sion is that Kuhn is not re­ally will­ing to say this. I think he is of the “all mod­els are wrong, some are use­ful” camp, thinks of paradigms as mod­els, and would be will­ing to ad­mit a new paradigm may be more use­ful than an old one.

Can we sep­a­rate the fact around which a paradigm is based (like “the Earth or­bits the sun”) from the paradigm it­self (be­ing a col­lec­tion of defi­ni­tions of eg “planet” and “or­bit”, ways of think­ing, math­e­mat­i­cal meth­ods, and rules for what kind of sci­ence will and won’t be ac­cepted)? And then say the earth fac­tu­ally or­bits the sun, and the paradigm is just a use­ful tool that shouldn’t be judged ob­jec­tively? I think Kuhn’s an­swer is that facts can­not be paradigm-in­de­pen­dent. A me­dieval would not hear “the Earth or­bits the sun” and hear the same claim we hear (albeit, in his view wrong). He would, for ex­am­ple, in­ter­pret it to mean the Earth was set in a slowly-turn­ing crys­tal sphere with the sun at its cen­ter. Then he might ask – where does the sphere in­ter­sect the Earth? How come we can’t see it? Is Marco Polo go­ing to try to travel to China and then hit a huge in­visi­ble wall halfway across the Hi­malayas? And what about grav­ity? My un­der­stand­ing is the Ptole­maics didn’t be­lieve in grav­ity as we un­der­stand it at all. They be­lieved ob­jects had a nat­u­ral ten­dency to seek the cen­ter of the uni­verse. So if the sun is more cen­tral, why isn’t ev­ery­thing fal­ling into the sun? To a me­dieval the state­ment “the Earth or­bits the sun” has a bunch of com­mon-sense dis­proofs ev­ery­where you look. It’s only when at­tached to the rest of the Coper­ni­can paradigm that it starts to make sense.

This im­presses me less than it im­presses Kuhn. I would say “if you have many false be­liefs, then true state­ments may be con­fus­ing in that they seem to im­ply false state­ments – but true state­ments are still ob­jec­tively true”. Per­haps I am mi­s­un­der­stand­ing Kuhn’s ar­gu­ment here; the above is an amalgam of var­i­ous things and not some­thing Kuhn says out­right in the book. But what­ever his ar­gu­ment, Kuhn is not re­ally will­ing to say that there are definite paradigm-in­de­pen­dent ob­jec­tive facts, at least not with­out a lot of caveats.

So where is the point at which we ad­mit some things are ob­jec­tively true and that’s what this whole en­ter­prise rests on?

Kuhn only barely touches on this, in the last page of the book:

Any­one who has fol­lowed the ar­gu­ment this far will nev­er­the­less feel the need to ask why the evolu­tion­ary pro­cess should work. What must na­ture, in­clud­ing man, be like in or­der that sci­ence be pos­si­ble at all? Why should sci­en­tific com­mu­ni­ties be able to reach a firm con­sen­sus unattain­able in other fields? Why should con­sen­sus en­dure across one paradigm change af­ter an­other? And why should paradigm change in­vari­ably pro­duce an in­stru­ment more perfect in any sense than those known be­fore? From one point of view those ques­tions, ex­cept­ing the first, have already been an­swered. But from an­other they are as open as they were when this es­say be­gan. It is not only the sci­en­tific com­mu­nity that must be spe­cial. The world of which that com­mu­nity is a part must also pos­sess quite spe­cial char­ac­ter­is­tics, and we are no closer than we were at the start to know­ing what these must be. That prob­lem— What must the world be like in or­der that man may know it?— was not, how­ever, cre­ated by this es­say. On the con­trary, it is as old as sci­ence it­self, and it re­mains unan­swered. But it need not be an­swered in this place.


Hon­estly this de­creases my trust in some of what’s come be­fore. Maybe he wrote all those sec­tions about in­com­men­su­rable paradigms be­cause paradigms re­ally are that in­com­men­su­rable. Or maybe it’s be­cause he thinks he’s play­ing some kind of ridicu­lous game where the first per­son to ad­mit the ex­is­tence of ob­jec­tive re­al­ity loses.


A lot of the ex­am­ples above are mine, not Kuhn’s. Some of them even come from philos­o­phy or other non­scien­tific fields. Shouldn’t I have used the book’s own ex­am­ples?

Yes. But one of my big com­plaints about this book is that, for a pur­ported de­scrip­tion of How Science Every­where Is Always Prac­ticed, it re­ally just gives five ex­am­ples. Ptolemy/​Coper­ni­cus on as­tron­omy. Alchemy/​Dal­ton on chem­istry. Phlo­gis­ton/​Lavoisier on com­bus­tion. Aris­to­tle/​Gal­ileo/​New­ton/​Ein­stein on mo­tion. And ???/​Fran­klin/​Coulomb on elec­tric­ity.

It doesn’t ex­plain any of the ex­am­ples. If you don’t already know what Coulomb’s con­tri­bu­tion to elec­tric­ity is and what pre­vi­ous ideas he over­turned, you’re out of luck. And don’t try look­ing it up in a book ei­ther. Kuhn says that all the books have been writ­ten by peo­ple so en­grossed in the cur­rent paradigm that they un­con­sciously jam past sci­en­tists into it, re­mov­ing all ev­i­dence of paradigm shift. This made parts of the book a lit­tle be­yond my level, since my knowl­edge of Coulomb be­gins and ends with “one amp per sec­ond”.

Even say­ing Kuhn has five ex­am­ples is giv­ing him too much credit. He usu­ally brings in one of his five per point he’s try­ing to make, mean­ing that you never get a re­ally full view of how any of the five ex­am­ples ex­actly fit into his sys­tem.

And all five ex­am­ples are from physics. Kuhn says at the be­gin­ning that he wished he had time to talk about how his sys­tem fits biol­ogy, but he doesn’t. He’s un­sure whether any of the so­cial sci­ences are sci­ences at all, and noth­ing else even gets men­tioned. This means we have to figure out how Kuhn’s the­ory fits ev­ery­thing from scat­ter­shot looks at the his­tory of elec­tric­ity and as­tron­omy and a few other things. This is pretty hard. For ex­am­ple, con­sider three sci­en­tific pa­pers I’ve looked at on this blog re­cently:

Cipri­ani, Ioan­ni­dis, et al perform a meta-anal­y­sis of an­tide­pres­sant effect sizes and find that al­though al­most all of them seem to work, amitriptyline works best.

Ce­bal­los, Ehrlich, et al calcu­late whether more species have be­come ex­tinct re­cently than would be ex­pected based on his­tor­i­cal back­ground rates; af­ter find­ing al­most 500 ex­tinc­tions since 1900, they con­clude they definitely have.

Ter­rell et al ex­am­ine con­tri­bu­tions to open source pro­jects and find that men are more likely to be ac­cepted than women when ad­justed for some mea­sure of com­pe­tence they be­lieve is ap­pro­pri­ate, sug­gest­ing a gen­der bias.

What paradigm is each of these work­ing from?

You could ar­gue that the an­tide­pres­sant study is work­ing off of the “biolog­i­cal psy­chi­a­try” paradigm, a ven­er­a­ble col­lec­tion of as­sump­tions that can be prof­itably con­trasted with other paradigms like psy­cho­anal­y­sis. But couldn’t a Hip­po­cratic four-hu­mors physi­cian of a thou­sand years ago done the same thing? A meta-anal­y­sis of the effect sizes of var­i­ous kinds of leeches for de­pres­sion? Sure, leeches are differ­ent from an­tide­pres­sants, but it doesn’t look like the be­lief in biolog­i­cal psy­chi­a­try is af­fect­ing any­thing about the re­search other than the topic. And al­though the topic is cer­tainly im­por­tant, Kuhn led me to ex­pect some­thing more profound than that. Maybe the paradigm is ev­i­dence-based-medicine it­self, the prac­tice of do­ing RCTs and meta-analy­ses on things? I think this is a stronger case, but a paradigm com­pletely di­vorced from the con­tent of what it’s study­ing is ex­actly the sort of weird thing that makes me wish Kuhn had in­cluded more than five ex­am­ples.

As for the ex­tinc­tion pa­per, surely it can be at­tributed to some chain of thought start­ing with Cu­vier’s catas­trophism, pass­ing through Lyell, and con­tin­u­ing on to the cur­rent day, based on the idea that the world has changed dra­mat­i­cally over its his­tory and new species can arise and old ones dis­ap­pear. But is that “the” paradigm of biol­ogy, or ecol­ogy, or what­ever field Ce­bal­los and Lyell are work­ing in? Doesn’t it also de­pend on the idea of species, a differ­ent paradigm start­ing with Lin­naeus and de­vel­oped by zo­ol­o­gists over the en­su­ing cen­turies? It look like it dips into a bunch of differ­ent paradigms, but is not wholly within any.

And the open source pa­per? Is “fem­i­nism” a paradigm? But surely this is no differ­ent than what would be done to in­ves­ti­gate racist bi­ases in open source. Or some right-winger look­ing for anti-Chris­tian bi­ases in open source. Is the paradigm just “look­ing for bi­ases in things?”

What about my fa­vorite triv­ial ex­am­ple, look­ing both ways when you cross the street so you don’t get hit by a bus? Is it based on a paradigm of mo­tor­ized trans­porta­tion? Does it use as­sump­tions like “buses ex­ist” and “roads are there to be crossed”? Was there a paradigm shift be­tween the bad old days of look­ing one way be­fore cross­ing, and the ex­cit­ing new de­vel­op­ment of look­ing both ways be­fore cross­ing? Is this re­ally that much more of a stretch than call­ing look­ing for bi­ases in things a paradigm?

Out­side the five ex­am­ples Kuhn gives from the phys­i­cal sci­ences, iden­ti­fy­ing paradigms seems pretty hard – or maybe too easy. Is it all frac­tal? Are there over­ar­ch­ing paradigms like atomic the­ory, and then lower-level paradigms like or­ganic chem­istry, and then tiny sub­sub­paradigms like “how we deal with this one or­ganic com­pound”? Does ev­ery sci­en­tific ex­per­i­ment use lots of differ­ent paradigms from differ­ent tra­di­tions and differ­ent lev­els? This is the kind of thing I wish Kuhn’s book an­swered in­stead of just talk­ing about Coulumb and Coper­ni­cus over and over again.


In con­clu­sion, all of this is about pre­dic­tive cod­ing.

It’s the same thing. Per­cep­tion get­ting guided equally by top-down ex­pec­ta­tions and bot­tom-up ev­i­dence. Oh, I know what you’re think­ing. “There goes Scott again, see­ing pre­dic­tive cod­ing in ev­ery­thing”. And yes. But also, Kuhn does ev­ery­thing short of come out and say “When you guys get around to in­vent­ing pre­dic­tive cod­ing, make sure to no­tice that’s what I was get­ting at this whole time.”

Don’t be­lieve me? From the chap­ter Ano­maly And The Emer­gence Of Scien­tific Dis­cov­ery (my em­pha­sis, and for “anomaly”, read “sur­prisal”):

The char­ac­ter­is­tics com­mon to the three ex­am­ples above are char­ac­ter­is­tic of all dis­cov­er­ies from which new sorts of phe­nom­ena emerge. Those char­ac­ter­is­tics in­clude: the pre­vi­ous aware­ness of anomaly, the grad­ual and si­mul­ta­neous emer­gence of both ob­ser­va­tional and con­cep­tual recog­ni­tion, and the con­se­quent change of paradigm cat­e­gories and pro­ce­dures of­ten ac­com­panied by re­sis­tance.

There is even ev­i­dence that these same char­ac­ter­is­tics are built into the na­ture of the per­cep­tual pro­cess it­self. In a psy­cholog­i­cal ex­per­i­ment that de­serves to be far bet­ter known out­side the trade, Bruner and Post­man asked ex­per­i­men­tal sub­jects to iden­tify on short and con­trol­led ex­po­sure a se­ries of play­ing cards. Many of the cards were nor­mal, but some were made anoma­lous, e.g., a red six of spades and a black four of hearts. Each ex­per­i­men­tal run was con­sti­tuted by the dis­play of a sin­gle card to a sin­gle sub­ject in a se­ries of grad­u­ally in­creased ex­po­sures. After each ex­po­sure the sub­ject was asked what he had seen, and the run was ter­mi­nated by two suc­ces­sive cor­rect iden­ti­fi­ca­tions.

Even on the short­est ex­po­sures many sub­jects iden­ti­fied most of the cards, and af­ter a small in­crease all the sub­jects iden­ti­fied them all. For the nor­mal cards these iden­ti­fi­ca­tions were usu­ally cor­rect, but the anoma­lous cards were al­most always iden­ti­fied, with­out ap­par­ent hes­i­ta­tion or puz­zle­ment, as nor­mal. The black four of hearts might, for ex­am­ple, be iden­ti­fied as the four of ei­ther spades or hearts. Without any aware­ness of trou­ble, it was im­me­di­ately fit­ted to one of the con­cep­tual cat­e­gories pre­pared by prior ex­pe­rience. One would not even like to say that the sub­jects had seen some­thing differ­ent from what they iden­ti­fied. With a fur­ther in­crease of ex­po­sure to the anoma­lous cards, sub­jects did be­gin to hes­i­tate and to dis­play aware­ness of anomaly. Ex­posed, for ex­am­ple, to the red six of spades, some would say: That’s the six of spades, but there’s some­thing wrong with it— the black has a red bor­der. Fur­ther in­crease of ex­po­sure re­sulted in still more hes­i­ta­tion and con­fu­sion un­til fi­nally, and some­times quite sud­denly, most sub­jects would pro­duce the cor­rect iden­ti­fi­ca­tion with­out hes­i­ta­tion. More­over, af­ter do­ing this with two or three of the anoma­lous cards, they would have lit­tle fur­ther difficulty with the oth­ers. A few sub­jects, how­ever, were never able to make the req­ui­site ad­just­ment of their cat­e­gories. Even at forty times the av­er­age ex­po­sure re­quired to rec­og­nize nor­mal cards for what they were, more than 10 per cent of the anoma­lous cards were not cor­rectly iden­ti­fied. And the sub­jects who then failed of­ten ex­pe­rienced acute per­sonal dis­tress. One of them ex­claimed: “I can’t make the suit out, what­ever it is. It didn’t even look like a card that time. I don’t know what color it is now or whether it’s a spade or a heart. I’m not even sure now what a spade looks like. My God!” In the next sec­tion we shall oc­ca­sion­ally see sci­en­tists be­hav­ing this way too.

Either as a metaphor or be­cause it re­flects the na­ture of the mind, that psy­cholog­i­cal ex­per­i­ment pro­vides a won­der­fully sim­ple and co­gent schema for the pro­cess of sci­en­tific dis­cov­ery.

And from Revolu­tions As Changes Of World-View:

Sur­vey­ing the rich ex­per­i­men­tal liter­a­ture from which these ex­am­ples are drawn makes one sus­pect that some­thing like a paradigm is pre­req­ui­site to per­cep­tion it­self. What a man sees de­pends both upon what he looks at and also upon what his pre­vi­ous vi­sual-con­cep­tual ex­pe­rience has taught him to see. In the ab­sence of such train­ing there can only be, in William James’s phrase, “a bloomin’ buzzin’ con­fu­sion.” In re­cent years sev­eral of those con­cerned with the his­tory of sci­ence have found the sorts of ex­per­i­ments de­scribed above im­mensely sug­ges­tive.

If you can read those para­graphs and hon­estly still think I’m just just ir­ra­tionally read­ing pre­dic­tive cod­ing into a perfectly in­no­cent book, I have noth­ing to say to you.

I think this is my best an­swer to the whole “is Kuhn deny­ing an ob­jec­tive re­al­ity” is­sue. If Kuhn and the pre­dic­tive cod­ing peo­ple are grasp­ing at the same thing from differ­ent an­gles, then both shed some light on each other. I think I un­der­stand the way that pre­dic­tive cod­ing bal­ances the im­por­tance of pre-ex­ist­ing struc­tures and cat­e­gories with a pre­served be­lief in ob­jec­tivity. If Kuhn is try­ing to use some­thing like the pre­dic­tive cod­ing model of the brain pro­cess­ing in­for­ma­tion to un­der­stand the way the sci­en­tific com­mu­nity as a whole pro­cesses it, then maybe we can im­port the same bal­ance and not worry about it as much.