A Curious Miscellany of Items Philosophical, Historical, and Literary

Manus haec inimica tyrannis.

Friday, April 23, 2010

Learning from Austrian Economists

Since the financial meltdown of 2008 and the ensuing “Great Recession”, economists have begun to re-examine the roots of their discipline. This re-examination seems to have crystallized into a revival of hitherto discredited Keynesian doctrines that had been largely abandoned after the 1980s. The attraction of Keynesianism is that it seems to give fairly clear prescriptions which rulers and policymakers can clutch at in order that they might be seen to be doing something – anything – to mitigate a perceived economic crisis.

To the powers that be, it matters little that the Keynesian prescriptions are wrong and will do more harm than good in the long run. After all, in Keynes’ famous words, “in the long run we’re all dead”. And if you’re a politician, the short run – generally around four years – matters much more.

Other economists, more sane though fewer in number, have instead turned to a rediscovery of the ideas of another historical strand of economic thought, the so-called “Austrian School”, the most well-known exponents of which were Friedrich Hayek (1899-1992) and Ludwig von Mises (pictured, 1881-1973). Their thinking has not caught on quite as much in the current “crisis”, mainly because their advice consists largely of avoiding action, which is not a prescription likely to endear them to regulators, nor to politicians and the mob that elects them.

Nevertheless, I have been immersing myself in the literature of the Austrian economists for the past year and a half or so, and in this post I’d like to impart some of the things that I’ve learned from them. In the next post I’ll play devil’s advocate and share what I take to be some of their shortcomings.

* * * *

1. Knowledge is Distributed

Human knowledge is not contained in the mind of a single person. The knowledge we require to run our society is dispersed across the minds of some six and a half billion people. This has pretty sweeping political implications. There is no single omnipotent person or planning bureau possessing all the knowledge necessary to coordinate the economic activities of millions of people. Since that knowledge is instead distributed among those millions of people, for the most part it’s best to let them coordinate themselves. Planners and policymakers should contain their hubris within the bounds of their capabilities.

2. Prices are Subjective

In medieval times theologians debated what the “just” price for widgets was, or what the “just” rate of interest to charge for a loan was. Now, what is a widget worth? It’s worth whatever people are willing to pay for it. Now, this sounds simple, but it surpassed the knowledge of the greatest thinkers of the Middle Ages. And we must ourselves avoid the fallacy of thinking that what people are willing to pay for a widget is somehow related to the intrinsic value of widgets in general, for if that were the case the price of widgets would be unchanging. Instead, the value of a widget is conceptually related to the opportunity cost of its purchase. In other words, what people are willing to pay for a widget depends on what other things they are willing to forego in order to purchase it. If I only have $X dollars to spend, and a seller is charging $Y for a widget, then buying a widget means I would only have $X-Y to spend on other things. If there is something else I need more than I need a widget and its price is greater than $X-Y, then I will have to forego purchasing the widget, because in purchasing the widget I would thereby lose the opportunity to spend my money on something more valuable to me.

Price is a function of the subjective valuations of many individuals. It is not the expression of the intrinsic value of a good. Indeed, there is no such thing as a good’s “intrinsic value”.

3. Markets are Information-Processing Mechanisms

The only way of finding out the value of a good is by having it exchanged in a market. A socialist planning bureau would have no way of efficiently pricing the widgets it ordered to be produced because it would have no way of finding out what the effective demand is for widgets, nor of finding out the cost of those factors going into its production. Those socialist economies that have hitherto existed have achieved whatever paltry production success they have by piggybacking, in parasitic fashion, on the pricing mechanisms (i.e. markets) of capitalist nations. It was market exchanges in free countries that allowed the USSR to have some rough idea of how many pairs of eyeglasses to produce at what cost (and even then, I can remember the semi-apocryphal Soviet-era stories about, say, millions of eyeglass frames being produced without lenses to go with them!).

For Austrian economists, the market is more than a place where things are bought and sold. The very act of buying and selling involves more than an exchange of money. It is at the same time an exchange of information, namely information about the preferences of people. And it is this information, formerly dispersed, which is aggregated in markets, allowing scarce resources to be allocated to their best (i.e. most preferred) uses. As von Mises put it, “to the entrepreneur of capitalist society a factor of production through its price sends out a warning: Don’t touch me, I am earmarked for the satisfaction of another, more urgent need. But under socialism these factors of production are mute”. Too much interference by planning and regulation disrupts this complex signaling process and distorts market information.

(Incidentally, all resources with which economists are concerned are scarce. If they weren’t in limited supply, economists would have nothing to say about them. Because they would have no opportunity costs attached to them, they could neither be bought nor sold, which would put them outside the realm of properly economic phenomena. It’s hard to name such an unlimited good. Perhaps air might be such a good, but only in certain contexts. If people were fish, water might be another example. Beyond the earth’s surface, most of the space in the universe is effectively unlimited, which in part is why there isn’t much of a market for ownership of distant stars.)

4. Social Order is a Spontaneous Growth

Because knowledge is dispersed across millions of people, the best way to organize social affairs is to not organize them, at least not in the usual sense. Social order comes about by people pursuing their own goals and objectives based on whatever knowledge they have. In order for this to happen, people must be allowed to communicate and exchange freely. As social beings we subordinate ourselves to rules and institutions which are largely the product of spontaneous growth, not conscious planning; of our making, but not of our design. The freedom of the market order often seems chaotic, but there’s a complex order underlying it, an order which we are in danger of destroying by trying to “rationalize” it through centralized planning.

5. The Rule of Law and its Decay

An example of such spontaneous order, particularly emphasized by Hayek, is the rule of law, as exemplified by the common law. The common law was a gradual and spontaneous growth, built up incrementally, not consciously planned or designed by anyone in particular. Those who were subject to the common law truly lived under “the rule of laws, not of men”. However, as time went on, a penchant for government activism has led to a situation where much of the common law has been replaced by a mass of statutes, designed and enacted, but which, taken collectively, possess little or no underlying rationale.

Symptomatic of this state of affairs is the cancerous growth of that monstrosity which goes under the oxymoronic name of “administrative law”. Administrative law is really the “law” pertaining to the arbitrary exercise of power by government bureaus. Administrative law has been replacing real law for some time now. In truth, attempting to legislate the exercise of arbitrary power is akin to attempting to go mad by rules.

Another symptom of the decay of the rule of law is the growing prominence of public law over private law. Law was once mostly private, governing the relations between private citizens. The bulk of law now takes the form of an ever-growing mass of statutes governing the relations between government and private citizens; or, more properly, it takes the form of statutes promulgated by government commanding citizens to do or not do (or both) various things, so that the latter’s sphere of free activity shrinks to the point that it no longer makes sense to speak of them as private citizens as such.

The worst part about all this is that it has been going on for so long now that we barely notice it (unless, say, you’re a smoker, which I am not). We assume that it is necessary for the government to perform the countless functions it now does, which necessitates all this rule-fetishism. Critics sneeringly accuse the Austrians of advocating a “night watchman” state, whose only role is to protect property, enforce contracts, punish criminals, and keep up the military. On the other hand, said von Mises, “it is difficult to see why the night-watchman state should be any more ridiculous or worse than the state that concerns itself with the preparation of sauerkraut, with the manufacture of trouser buttons, or with the publication of newspapers”.

My personal views about the duties of government fall somewhere between these two extremes. However, with my government trying to ensure that I eat enough vegetables, I know that the pendulum has swung too far in the direction of nanny-ism. Even fascist dictators have better sense than to concern themselves with such matters.

6. Prices are Interconnected

Eyeglass frames are not worth much without an adequate number of lenses to go with them. If too few lenses are produced, the price of lenses will rise (because there aren’t enough to go around), while the price of frames will drop (because they’re useless without lenses). The prices of lenses and frames bear an inverse relation to each other. There are other goods which are similarly related, but not inversely. For example, if the price of a certain good rises, so too will the price of another good which can be used as a substitute for it.

Socialist and interventionist economists, or at least the more stupid ones, have tended to treat the prices of various goods as if they were independent of each other. Thus, when proposing a policy of price controls, they assume that fixing the price of one good will have little or no effect on the prices of other goods, and on the economy more generally. But because of what von Mises calls the “connexity of prices”, such interference can have widespread effects.

7. The Pervasiveness of Unintended Consequences

Because of phenomena like distributed knowledge, the connexity of prices, and spontaneous order, we must be wary of attempts to control phenomena through rational planning, for the consequences of our actions can be unforeseen and widespread. Thus, the Austrian analysis seems to contain an inherent tendency towards conservatism. Despite this, Hayek himself explicitly denied any tendency towards conservatism in his thought (a denial which, incidentally, I find unconvincing). Karl Popper, a friend of Hayek’s, admitted the necessity of planned change, but suggested that it ought to be done by what he called “piecemeal social engineering”, that is, the tentative implementation of small incremental changes that are reversible.

8. Money is a Commodity

According to Carl Menger (1840-1921), the founder of the Austrian tradition in economics, the use of money originated as a form of indirect barter. You might offer me a number of chickens for my pig, but I might have no use for chickens at the moment, even though I have too many pigs for my own use. But if I knew that you had gold and that I could easily exchange that gold with someone else for something that I could use, then I might accept gold as a substitute (or you might exchange your chickens for gold with someone else and then come to me with the gold). Essentially this latter transaction is still a matter of barter, only more indirect. The only thing that makes gold different from pigs or chickens is that, for various historically contingent reasons, the demand for gold is (almost) always stable and universal, enabling it to become a standard medium of exchange.

Of course, eventually, instead of accepting gold, which is attended with certain inconveniencies, people came to accept paper issued from a trustworthy source which represented a given quantity of gold and which was technically exchangeable for the latter.

Short of coin clipping and other crude traditional ways of debasing the currency, when money is tied to a commodity it is difficult for a government to interfere with the money supply for its own nefarious purposes ― sooner or later the market catches on and price levels adjust accordingly. Gresham’s Law (“bad money drives out good”) will take over: people will pay their debts in devalued money, hoard gold and avoid lending, and prefer to consume their capital rather than invest it. Manipulation of a commodity-based money supply will be thwarted in the long run, one way or another.

However, when currency is decoupled from its commodity basis and the money supply is treated as if it were something that can be expanded indefinitely, the government is free to manipulate it for its short-term interests. These interests will usually involve inflating the currency through credit expansion, which – thanks in part to the false ideas of Keynes and his school – it believes it can do indefinitely. Keynesian economics seems to imply the existence of free lunches. Austrian economics prefers to “go Dutch”.

The Austrians adhered to a “sound money” policy. Such a policy limits the ability of governments to interfere with the normal operations of the market. For adherents of Austrian economics, the abandonment of the gold standard by the US in 1971 represented something close to the beginning of the end of capitalism. Who knows? They might turn out to be right. Such alarmism certainly seemed warranted during the great stagflation of the 1970s.

9. Beware the Boom, not the Bust

Keynesianism is all about shortening and mitigating economic downturns. For them, busts are caused by a lack of aggregate demand. They are unnatural. The cure for them is to increase consumption, even if this requires deficit spending by the public sector. The obvious danger to such a policy is inflation. But until the 1970s it was believed that there was an inverse relation between inflation and unemployment (the so-called “Phillips curve”), so a little inflation seemed a worthwhile tradeoff. As it turned out, this belief was mistaken: it was possible to have both inflation and high unemployment.

For the Austrian School, downturns are natural and inevitable, and the more you do to avoid them, the worse and more protracted you make them. Again, there’s no such thing as a free lunch.

It’s actually booms that are unnatural, especially when they’re fuelled by the expansion of credit and the money supply. When markets are awash in capital (or seemingly so, due to the short-term cognitive illusion caused by inflation), capital and resources get allocated inefficiently. Rather than investing in viable business enterprises, money is invested in imprudent schemes (like housing bubbles?). On the aggregate level, all of this represents a huge misallocation (or in von Mises’ jargon, a “malinvestment”) of resources.

Such misallocation will be felt sorely in the inevitable bust. Money stupidly invested in certain enterprises during flush times cannot be instantly taken out and reinvested more profitably elsewhere. Investment is not a perfectly reversible process. Capital gets tied up precisely when it needs to be freed up for use elsewhere, thereby exacerbating the bust. Eventually these effects will work themselves out, but this takes time, and the longer the boom lasts, the longer it will take to recover from the hangover.

Further to be considered are the effects of inflation during the boom. Most of these, including liquidity preference (i.e. cash hoarding), reduced lending, and preference of capital consumption over investment, are well-known, but the Austrians added a new twist. Keynesians often speak as if inflation is a phenomenon that works the same effects on the prices of all goods, across the board. They talk in terms of the rising price level. In contrast, the Austrians noted that, thanks in part to the connexity of prices, the effects of inflation will not spread themselves evenly throughout the economy. Some prices will go up (at different rates) while others will actually go down. This state of affairs amounts to a serious distortion of market signals, leading investors and consumers to make bad choices. Again, inflation leads to the misallocation of resources and market inefficiency.

All of this may sound too much as if I’m offering an unqualified defense of Austrian economics. Perhaps I am being too liberal in my praise, if only to offer a counterbalance to all the Keynesian “conventional wisdom” floating around right now. As promised, my next post will be devoted to an exposition of some of the shortcomings I’ve found in reading the Austrian literature.

Saturday, April 10, 2010

Le Jeu de la Mort

I recently heard about a disturbing new "reality" television show in France which was the basis of a documentary. The show was called Le Jeu de la Mort("The Game of Death"). It was based on Stanley Milgram's famous obedience experiments, in which he had unsuspecting subjects administer putatively fatal or near-fatal electric shocks to a hapless victim. Milgram managed to elicit a 64% compliance rate. Unfortunately, Le Jeu elicited a rate closer to 80%, which might mean that either (i) French people are more amoral than Americans, which I think unlikely, (ii) people have become even more amoral than Milgram's subjects were in the early 1960s, or else (iii) people are simply willing to sink to previously unplumbed depths for the glory of television, depths to which the glory of mere science could not compel them.

Perhaps the show is not the end of civilization as we know it, but one can pretty much see it from there. In any case, it put me in mind of a paper I wrote but never published. I put it aside sometime during the peer review process and didn't touch it again after I ceased being an academic. I began to think that I should really do something with it, so I'm posting it here on my blog.

Please read the endnotes. They're not just citations of literature. I put some consideration into them.

Setting People Up for Failure: The
Situationist Critique of Virtue Ethics

Though virtue ethicists take diverse standpoints, they generally have the following view in common: they believe that the appropriate object of moral evaluation ought to be agents — or more specifically, the characters of agents — rather than the actions of such agents. Thus, instead of evaluating courses of conduct, we should instead be evaluating whatever are the stable traits of character from which they spring. Actions are the signatures of character, and it is character that the moral philosopher should be interested in. In short, the purpose of ethics is to answer not the question, “What should I do, or according to what principles should I act?” but rather the question, “What sort of person should I be?”

However, a controversial and disconcerting argument has been offered to the effect that there is no such thing as character.[1] Obviously, if character were to prove to be a fictitious entity, the characterological programs like virtue ethics would be fundamentally undermined. Thus, the argument has been described as “deflationary” (Merritt 2000:365; Swanton 2003:19). The challenge is known as situationism, which, simply stated, is the view that behavior springs less from stable traits of character, and more from properties of the situations that elicit agent behavior. Its philosophical proponents mostly find their inspiration in research done over the last half century in the field of social psychology. Upon examination, it will be found that the argument relies on inferences from empirical data that are largely unwarranted.


The opening salvo of the situationist challenge to virtue ethics began with a paper by Gilbert Harman (Harman 1999). There he argued that if virtue ethics depends on the existence of stable traits of character, then social-psychological research done in the past forty or so years indicates that such traits of character do not exist. Therefore, virtue ethics is a dead end. Despite later softening, his language in this paper is strong. For example, he castigates Owen Flanagan (ibid. 326) for compromising with the enemy by suggesting that there are traits of character, though they are somewhat context-dependent and not as robust as virtue ethicists often imagine them to be (see Flanagan 1991:280-281). And he dismisses the notion of character education in one sentence: “If there is no such thing as character, then there is no such thing as character building” (Harman 1999:328).

Doris (2002) offers a considerably more nuanced version of Harman’s somewhat brusque argument, having the relative luxury of developing his position over the course of a book-length study. He first lays out the position that he is attacking, namely characterological “globalism”. According to Doris, trait globalism involves three theses (though he believes it is possible that the third can be jettisoned without affecting the other two). These theses (following Doris 2002:22) are:

1. Consistency. Traits are reliably displayed across a diversity of trait-relevant situations. Thus, if it is appropriate to φ in S1 and S2, I will φ in both.

2. Stability. Traits are reliably displayed over iterations of the same or similar trait-relevant situations. Thus, if I φ in S1, I will also φ in S2, where S1 ≈ S2.

3. Evaluative integration. The occurrence of a trait with a particular “evaluative valence” (i.e. a virtue or a vice) is probabilistically related to the occurrence of other traits with a similar evaluative valence.

What consistency means is that if I refrain from cheating on tests, thus displaying the trait of honesty, I will display this honesty by returning excess change at the cafeteria too. Stability means that if I refrain from cheating on my economics test, I will not subsequently cheat on my next economics test, or even on my physics test.[2] Evaluative integration is basically an expression of the traditional “unity of the virtues” thesis, the idea that the possession of one virtue (or vice) implies the possession of other virtues (or vices). Thus, one is not truly honest unless one is also courageous, loyal, and so on.[3]

Supposedly, globalism captures the intuitions of our folk psychology regarding character, and is presupposed by most, if not all versions of virtue ethics. For reasons that will become clear, if not entirely convincing, Doris rejects consistency and evaluative integration on empirical grounds. He does allow the possibility that “local” traits are consistent with the empirical data. However, the implication here is that although the thesis of stability may hold, it will not be of much use for virtue ethicists. This is because local traits are too local to be of much theoretical use: instead of a broad trait like honesty, we would only be able to speak of traits like honest-in-economics-tests, or honest-in-returning-excess-change. If character is the sum of an agent’s traits of character, then the existence of local traits would mean that character is highly fragmented. Such fragmented character would allow for little power to predict or explain behavior. There is, however, empirical research, mostly done through longitudinal studies of children, — and mostly ignored by Doris — indicating that there actually is an impressive degree of cross-situational consistency in traits of character (for an overview, see Eisenberg and Mussen 1989:18-22). Though consensus is lacking, we may at least say that Doris’ empirical premises are not as sound as they seem.[4]


According to situationists, globalism must be rejected or revised in light of data produced by social psychologists. There is not enough space in this study to examine all of the relevant research. Brief mention of a few studies must suffice, with more detailed discussion of a further one. For the most part, the studies cited are the ones most commonly appealed to by philosophical situationists.

Group effect was vividly demonstrated in a simulated prison at Stanford University (Haney et al. 1973). This disturbing experiment was supposed to have lasted two weeks but was stopped after just six days. A group of seemingly normal subjects was divided into “prisoners” and “guards” and put into a simulated prison environment. Typically, guards enjoyed their roles. The prisoners were not so lucky. Five had to be released prematurely due to the mental and emotional strain. Without delving into details, guards engaged in every ingenious form of cruelty, being prohibited from inflicting physical punishment. Inhibitions were shed as individual guards were subsumed in the flow of the group.

On a more banal note, mood effect was demonstrated in an experiment which found that people were significantly more likely to help a person pick up a dropped armload of papers immediately after finding a dime in a phone booth than were people who did not find a dime (Isen and Levin 1972). It was posited that the slight effect on subjects’ moods after finding something as inconsequential as a dime had a disproportionate effect on their subsequent behavior. Of those subjects who found a dime, fourteen helped and two did not; of those who did not find a dime, only one helped, while twenty-four did not.

In another experiment (Darley and Batson 1973), a group of theology students was informed that they were required to give a talk in another building. Some were told that they were running late for the appointment, while others were told they had enough time. Those who thought they were in a hurry were significantly less likely to help a distressed person lying in a doorway along their path than those who were not hurried; helping behavior seemed to vary with degree of hurry. Some of the subjects (in a twist of irony) were required to give their talks on the parable of the Good Samaritan, but this ultimately had had little effect on subsequent behavior.

What these experiments are held to show is that behavior is less a function of an agent’s character or endogenous motivational dispositions, and more a function of variables in the situation in which the agent acts. And furthermore, these situational variables may be 1) relatively insignificant, and 2) seemingly unrelated to the behavior they supposedly elicit. These two points are most clearly shown in the dime experiment, above.

In what follows we will focus mainly on a fourth experiment, or rather set of experiments, the Milgram obedience experiments (Milgram 1974). There are several reasons for devoting more attention to this particular example. First, focusing on a single experiment will help us make the best use of a limited amount of space, and the Milgram experiments can tell us most of what we need to know about the situationist mode of inference. Indeed, one advocate has claimed that Doris could have largely rested his case on the Milgram experiments alone (Webber 2006:656). Second, the generality of people have at least some familiarity with it, as it was well reported in the popular media. Third, it is the experiment most commonly cited in the situationist literature, and thus, animadversions on it generalize well. And finally, Milgram’s book (still in print) is a pleasure to read and can still provide much profit to the reader who, it is hoped, will be encouraged to pick it up.

The standard, baseline setup of the experiment was as follows. Volunteer subjects, called “teachers”, posed questions to a “learner” who, unbeknownst to the teacher, was complicit with the experimenter. The putative purpose of the experiment was to study the effect of punishment on learning. The learner was hooked up to a convincing-looking dummy shock generator and asked questions by the teacher. He would answer incorrectly according to a set pattern. The teacher was to administer a shock to him every time he got an answer wrong. With each wrong answer the shock voltage was to be increased. In the basic set-up the learner is in a separate room from the teacher. At 150 volts the learner protests audibly and demands to be let out of the experiment. In his agitation he yells and pounds his fists on the wall. At 330 volts the learner becomes silent and no longer answers questions. At 450 volts (the maximum level) the experiment is terminated. The dial the teacher uses to set the shock level is labeled as follows: “Slight Shock”, “Moderate Shock”, “Strong Shock”, “Very Strong Shock”, “Intense Shock”, “Extreme Intensity Shock”, “Danger: Severe Shock”, and finally “XXX”. If the teacher hesitates to go on, the experimenter goes through a series of four verbal “prods” to get him to continue. After the fourth prod the experiment is terminated.

It was expected that very few subjects would proceed to the 450-volt level, and few subjects expected themselves to do so. Yet on this standard set-up 65% of subjects went all the way to the maximum shock level (i.e. were “obedient”). Of “disobedient” subjects, none disobeyed before reaching 300 volts (“Intense Shock”). The mean shock level was 405 volts (“Danger: Severe Shock”).


Situationists take the Milgram experiments to demonstrate that subjects can be induced to engage in behavior that they do not identify with — indeed, behavior that in most cases they are extremely averse to — at the calm prodding of a technician. From this finding, situationists like Doris adduce that the prosocial trait of compassion is either 1) non-existent, 2) not very strong, in that it may easily be overridden, or relatedly, 3) local (in the sense explained in §1, above).

Critics of situationists tend to point out that what the findings really show is that compassion is rare, not very widely distributed across the population. Furthermore, they argue, it should not be surprising to find that virtue is rare (Athanassoulis 2000:217-219; Kupperman 2001:242-244); the rarity of virtue is exactly what we ought to expect. Situationists in turn accuse such critics of elitism, of making virtue the possession of the few.[5] Besides not being a very useful conception of morality, putting as it does ethical life beyond the reach of most people, this sort of elitism also runs afoul of Flanagan’s useful Principle of Minimal Psychological Realism (PMPR): “Make sure when constructing a moral theory or projecting a moral ideal that the character, decision processing, and behavior prescribed are possible, or are perceived to be possible, for creatures like us” (Flanagan 1991:32). It may be an (unattainable) ideal, but it cannot be said to be descriptive of most — or even many — agents.

I must agree with the situationists on this point: if it is indeed the case that obedient subjects lack compassion, then it is a weak response to say that we ought to be satisfied that compassion is rare. There is no way around the fact that even for the most jaded and cynical among us, the high obedience rate in the Milgram experiment is — pardon the pun — shocking. It may be understandable that most people would be too thoughtless to help someone pick up a pile of papers, as in the Isen and Levin experiment.[6] This sort of narrowed conception of compassion we may be able to live with. But we still expect that the majority of our fellow citizens will not apply lethal doses of electricity to the innocent on the mere word of a labcoated stranger. We expect more from people, and we are not mollified by being instructed in the precious rarity of virtue. Luckily, we need not grant that the Milgram experiment shows that people lack compassion.


We could be pardoned if, upon reading the situationist literature surrounding Milgram’s experiments, we came away with the impression that subjects went about administering lethal shocks to their victims with no more compunction than would be shown in swatting mosquitoes. Indeed, Jonathan Webber claims that “almost all volunteers” were obedient (Webber 2006:652). With such descriptions, it is difficult to remember that close to a third of subjects were ultimately disobedient. But upon reading Milgram’s own account, we come away with a different impression. Indeed, part of what makes the account of these experiments so disturbing is that the vast majority of the subjects were what most of us would be willing to call compassionate, to some intelligible degree. This compassion, though it turned out not to issue in disobedience to authority, was certainly displayed by virtually all the subjects, as seen in the various manifestations of the distress they were experiencing.­ As Milgram noted,

the second unanticipated effect [besides the high voltage subjects generally administered] was the tension generated by the procedures. One might suppose that a subject would simply break off or continue as his conscience dictated. This is very far from what happened. There were in some subjects striking reactions of emotional strain. (Milgram 1974:41)

Furthermore, Milgram found that “obedient subjects reported themselves as having been slightly more tense and nervous than the defiant subjects at the point of maximum tension.” (ibid. 1974:42).

How did this tension play out? It varied from subject to subject. Some sweated profusely, some pulled at their hair or banged their fists on the table as they administered shocks. They also devised various coping strategies to make what they were doing easier on themselves. Commonly, in the variation of the experiment in which subject and victim were in the same room, subjects would try to position themselves in such a way as to avoid making eye contact with their victim; others, in reading out the questions, gave verbal cues to signal the correct responses to the learners.[7]

This latter strategy is an interesting one, insofar as it indicates the possibility that situationists are too crude in their definition of “disobedience”. In the context of the experiment, giving verbal cues is a (quiet) subversion of authority if ever there was one, as this would certainly undermine the heuristic value for the experimenter of the experiment the subjects thought they were participating in. In this sense, avoidance of inflicting harm on an innocent person became, for these subjects, a more important objective than was helping the experimenters achieve their putative end.[8] Seen in this way, we seem to have indications of a shading from obedience to disobedience, rather than the simple “on/off switch” of either continuing or quitting participation in the experiment.

Let us imagine that two people find themselves pressed into a firing squad, working for what they believe to be a wicked regime. Furthermore, they are about to execute what they are convinced is an innocent person. One of the two throws down his rifle, refusing to take part in the execution. He is shot, along with the victim. The other conscript simply fills her rifle with blanks (or purposely takes poor aim of her target). Are we to say the latter lacks compassion? And which of the two strategies was the wiser one, given that in neither was the victim’s life saved, while one resulted in two dead innocents instead of one? During the Holocaust, those righteous who resisted Nazi authority did so in their own particular ways. Is the citizen who hid a Jewish child in his attic to be regarded as lacking compassion because someone else (someone perhaps more foolhardy) put himself between a bullet and a Jew to save the latter?

If, as Doris contends, disobedience is a “prosocial” behavior in the context of the Milgram experiments, and if there are different ways of displaying it (as I contend), then might it not be the case that an agent’s adoption of one style of disobedience over another is an expression of something akin to individual character or personality? To take an example, research indicates that styles of prosocial behavior might be correlated with introversion or extroversion. The prosocial behavior of extroverted children tends to be more overt, more active, more extroverted than that of introverts. For example, in a study of children’s reactions to a simulated emergency situation in the presence of other children, it was found that “in general, extroverted children engaged in more active helping, whereas introverts tended to prefer passive modes of helping that did not require social interaction” (Eisenberg and Mussen 1989:61, citing Suda and Fouts 1980). As the saying goes, there is more than one way to skin a cat, and each such way may be expressive of individual differences in personality or character.

Returning to the point that most subjects in Milgram’s experiments displayed some manifestation of what most would call compassion, we note two further interesting facts. First, in the experimental variant in which they were permitted to choose their own shock level, none of the subjects went above seventy-five volts. Indeed, most of them never even reached the level at which the victim begins to express discomfort (Milgram 1974:71-72). This seems to indicate that there was no positive aggression or wanton cruelty implicated in the experimental results, and we might go so far as to claim that positive compassion was displayed, in at least the minimal sense that subjects showed some implicit regard for the principle that one ought not inflict harm on another who has not harmed oneself.

The second interesting fact Milgram noticed was that

when the experimenter was absent, subjects displayed an interesting form of behavior that had not occurred under his surveillance. Though continuing with the experiment, several subjects administered lower shocks than were required and never informed the experimenter of their deviation from the correct procedure. Indeed… some subjects specifically assured the experimenter that they were raising the shock level according to instruction, while, in reality, they repeatedly used the lowest shock on the board. (Milgram 1974:62)

Again, this indicates that there is more than one way to disobey authority. By limiting “disobedience” to an open break with authority, Doris and Harman (and to some extent even Milgram himself) see results that are more shocking than they need be. Once we widen our conception of what it means to disobey authority, we see that people are not as mindlessly obedient as they may appear. And certainly they come off seeming more compassionate than most situationists would seem willing to grant.[9] Some subjects were outright obedient or disobedient. Others, mostly in non-standard permutations of the experiments, were what we might call subversive. However, most displayed what we might call “compassion-consistent affect”.


What we can infer from the discussion of the Milgram experiments so far is that situationists have a narrow allowance of what counts as “behavior” in the experimental context. There are only two possibilities of action: either continue with the experiment (obedience), or walk away from it (disobedience). A whole range of possibilities in between are eliminated. Indeed, anyone who has spent any amount of time with a young child knows the creativity of children in testing the limits of disobedience. The baseline variant of the experiment seems perversely engineered to eliminate those middling possibilities of action that open up when some of the situational restrictions are relaxed. It is a forced choice experiment, and forced choice experiments overlook nuance in the quest for clear data.

Situationists have an artificially narrow definition of compassion in the context of the Milgram experiment. They assume that compassion is a trait that can only be expressed in one way: overt disobedience. Again, we have a fixed choice, which, when relaxed, allows for a variety of expressions of compassion. What the Milgram experiments say about compassion, if anything, is that it is a trait that best finds expression in a situation whose constraints allow for its expression. The more constraints that were removed from the standard Milgram experiment, the more ways subjects found to subvert the will of the authority figure. It was not shown that 65% of people lack compassion. Rather, it was shown that most people have at least a middling sort of compassion that will show itself in a variety of unspectacular ways.

Christian Miller makes much the same point: “[I]t has rarely been part of the view that possession of a virtue is an all or nothing phenomenon; rather, it comes in degrees” (Miller 2003:378). Now, such a weak virtue is not enough to underwrite strong evaluative trait attributions; we would not approvingly single its possessor out as being a compassionate person, in any full-blooded sense. Certainly such an attribution could serve no individuating function, as it is a trait he shares with most others. But neither does it accord with situationist claims for the non-existence of traits. Even with such a weak trait, we could imagine some situation in which it would make itself known. For example, might a person with weak compassion be disobedient in the Milgram situation if the authority figure left him to carry out the experiment on his own?

Much of this could have been surmised with little effort by comparing the situationist thesis with the title of Milgram’s book. Doris claims that Milgram’s work shows that the generality of people lack compassion. But it was not Milgram’s intent to study compassion; the title of his book is Obedience to Authority, and it was obedience that he was putatively studying. In order for Milgram’s work to be understood as saying anything about compassion, it must be demonstrated that there is some necessary relationship between compassion and obedience (or disobedience, as the case may be). To my knowledge, this has not been explicitly demonstrated in any of the relevant situationist literature, including Doris and Harman. Now, intuitively it is likely that there is some relationship between the two, but it is probably a lot more complex than these authors would have us believe. A couple of examples should make clear my point.

The first example is basically that of the firing squad, above. Imagine you are forced by an authoritarian military government to be part of a firing squad, and you are about to take part in the execution of someone you know to be innocent of any crime. You take part in the execution but purposely aim poorly in order not to kill the prisoner. I too am with you in the firing squad. However, I throw down my rifle and refuse to take part in the execution. I am shot along with the prisoner. Which of us is compassionate and which of us is not? Probably both of us are reasonably compassionate, but only I disobeyed. This indicates that compassion alone is not sufficient for disobedience, though compassion and courage might do the trick.

Here is a second example, this time a more mundane one. My boss asks me yet again to stay late on a Friday evening and work overtime. I angrily refuse. I am fired. My action is certainly an example of disobedience. And yet compassion seems to play no role here, though if there is a virtue at work we might characterize it as self-respect, or loyalty (to labor laws, or to my family who I am not seeing much of these days), or both, or neither. The point of these two examples is to illustrate that there is no necessary connection between compassion and (dis)obedience to authority. And where compassion is implicated, it may be implicated as just one component of a more complex motivational structure.


It has been objected to the situationist interpretation of the Milgram experiments that there are other ways in which it was contrived to force obedience. Sabini and Silver believe “that the key to the subjects’ obedience is the experimenter” (Sabini and Silver 2005:549). They are “strangers to the world of the laboratory or, better, visitors to this world. They are looking for clues about how to act, and the experimenter provides them” (ibid. 550). Presumably, the situation — let us call it the Milgram-situation — is not like anything most of the subjects are used to finding themselves in. They are disoriented, and nothing in their experience tells them how they are to behave. They desperately want some advice as to protocol. It is natural under such circumstances to advert to some established authority, and one is provided in the form of a labcoated technician. Furthermore, they are operating in an institutional framework. By “institutional framework” Sabini and Silver are not talking about Yale University in particular (where the experiments were conducted), but rather “the idea that this experiment is a going concern, that lots of people have been through this experience; the experimenter is a guide to this little bit of the social/moral world. He is taken to be someone who sees clearly what is objectively needed” (ibid. 550-551).

Erik Wielenberg holds a similar position to Sabini and Silver. Situationists take phenomena like the Milgram experiments to demonstrate that supposedly insignificant situational factors can have a disproportionate effect on subjects’ behaviour. On the contrary, Wielenberg turns things upside-down and claims instead that the results show that factors thought to be insignificant were in fact quite significant (Wielenberg 2006:480). The technician’s calm but insistent voice is not background, it is foreground.

Thus, redescribed according to Sabini and Silver, the Milgram-situation seems more like a setup for moral lapse than the sort of neutral and uncoercive environment described by situationists. As such, our natural reaction ought to be a knowing nod of the head rather than hand-wringing. However, situationists can here object that, whatever the exact characterization of the situation one prefers, their point still stands: aspects of the situation induced the behavior from the agents, not the agents’ own endogenous motivational structures. To answer this objection, I must discuss what I take to be the most powerful single criticism of situationist social psychology, a criticism that in my opinion has not been sufficiently addressed by situationists, nor sufficiently stressed by their opponents.


Due to inherent logistical limitations, experiments cited in the situationist literature have been conducted on subjects in single trials. This is a serious problem if one hopes to draw from them inferences about the presence or robustness of traits of character. If we accept a dispositional account of traits of character, we believe, very roughly, that a trait of character is a disposition to behave in certain ways in certain kinds of situations.[10] But single trial experiments cannot tell us anything about the dispositions of subjects. As Athanassoulis notes, “the experiments were not designed to uncover differences in character traits. They examined a specific reaction to a specific situation and did not reveal much about the subjects’ long-term dispositions.” (Athanassoulis 2000:217).

Thus, rather than having something to say about the existence or not of a given character trait, all that such experiments can tell us is that a certain proportion of the general population will likely behave a certain way when faced with a certain kind of novel (and artificial) situation. But if this is to speak to the issue of a trait’s stability (as defined in §1, above), the experiment would have to be repeated on the same subjects. Longitudinal study is required to warrant trait attribution. This is not possible in setups like the Milgram experiments, for the obvious reason that subjects would know what to expect on subsequent trials and thus would be less likely to behave in the same way, or in other words, they would be more likely to be disobedient.

Thinking closely about the last sentence, we may notice something strange about it, something that to my knowledge situationists have not been called on to answer for. If the situationist thesis is correct, and behavior is more or less determined by factors in the situation rather than by an agent’s endogenous motivational structures, we should expect them to behave in the same way on subsequent trials. After all, it is supposedly not the situation that changes on subsequent trials.

A situationist could conceivably reply here that the first trial induces in the agent a particular emotional response or mood effect, causing them to construe the situation differently on subsequent trials, the situation itself remaining objectively the same. This assumes that the later construal is wrong, because supposedly based on affect, and because supposedly objective situational variables are unchanged.

However, to this the agent may very well retort that it was the first construal that was wrong, based as it was on incomplete information; whatever mood or emotion is induced is a perfectly appropriate response in light of fuller knowledge and experience of the situation. Furthermore, this is perfectly consistent with the empiricist axiom (which situationists, being largely empiricists, would presumably embrace) that if there is such a thing as objective knowledge, experience is a way of acquiring it. Therefore, even if we are to assume that the situation does remain the same across iterations of it, we ought to consider the possibility that the agent’s specification of it on subsequent trials is more apt to be the “correct” one. Thus, either a) supposed repetitions of the experimental situation constitute different situations (nominalism), or b) the same situation is iterated (realism), but it is later specifications of the situation that are presumed to be veridical. In either case, the situationist position is undercut.

If the subjects truly lacked the trait of compassion, it needs to be explained why it is that they would suddenly and reliably gain it in time for the next trial of the experiment. Let S1 be a subject’s first exposure to the Milgram-situation, and S2 to be the second, and so on. Let φ be “obedience” and let ~ φ be “disobedience”. Now, situationists are fascinated by the fact that S1(φ) holds for most subjects. So much so in fact, that they do not seem to notice what should be a very interesting fact, that the series S1(φ), S2(~ φ) holds (or is presumed to hold) for most subjects. And, presumably, the series S1(φ), S2(~ φ), S3­(~ φ)… S­n(~ φ).­ How is this phenomenon to be explained by appealing to situational factors only, if S is supposed to be constant throughout? If behavior were strongly situationally-dependent, we should rather expect the series S1(φ), S2(φ), S3(φ)… Sn(φ).

It makes much more sense to say that compassion (or some other endogenous motivational structure) was there all along, but that for reasons outlined in sections 5 and 6, above, subjects were induced to act out of character.[11] Acting out of character is something we are all familiar with. What the Milgram experiments demonstrate vividly are the possibly disastrous consequences of acting out of character, valuable knowledge that we perhaps need to be reminded of. This is basically Joel Kupperman’s interpretation of the Milgram experiments:

Various psychological experiments, the most famous of which are the ones initiated by Stanley Milgram, have shown that a majority of people (most of whom must be presumed to have been moderately decent in ordinary life) will do appalling things in circumstances so unusual that ordinary standards might seem not to apply, especially if someone who seems reliable suggests to them that what they are about to do is really quite normal. (Kupperman 2004:105)

Unfortunately, subjects had no warning that within a matter of moments they were going to pass into the nightmarish world of the Milgram-situation. For most of them, probably very little in their experience prepared them for this eventuality, and certainly they were not warned by Milgram or his associates.

As an obiter dictum, one could also point out to situationists that agents do not only act out of character when they do something bad. The stock example here might be Oskar Schindler, who felt compelled to protect Jewish employees, while at the same time feeling guilty for failing to maximize his profits from them. This phenomenon has been called “inverse akrasia,” a weakness of will that results in right conduct (Arpaly and Schroeder 1999:162; see also Hursthouse 1999:150ff.). This means that we cannot automatically assume that subjects who disobeyed the experimenter in the Milgram situations did so because of their upright characters. On the face of it, the situationist could agree wholeheartedly, saying “Indeed, that is exactly our point. There is no such thing as character, so both the disobedient and the obedient cannot have character attributed to them.” The problem is, if this is the case, then no inferences can be made about character from an agent’s conduct on a single occasion. Neither obedience nor disobedience on such an occasion can tell us anything about that agent’s character, exactly because both types of agents could be acting out of character. As Arpaly and Schroeder advise, to make such inferences, we must look at the agent’s whole character, over stretches of time, in a diversity of situations.

It should also be added that to make such an illicit inference from isolated action to the non-existence of character, as the situationists do, is to ignore the dimension of character’s development over time. Thus, a subject’s obedience, rather than demonstrating his lack of character, might prove to be better described as the beginning of a new stage in that character’s development. His participation in the Milgram experiment might teach him something new about himself, having an effect on his future conduct. In short, there is a reflexivity here: situations and actions are ingredients in a character’s ongoing development.


Perhaps we were somewhat hasty in the previous section in writing that it is “not the situation that changes on subsequent trials” of the Milgram experiment with the same subject, or that “S is constant throughout”. When viewed from the perspective of the subject, an iteration of the same situation makes for quite a different situation.[12] On the first trial, the situation is a bewildering one in which the agent knows not how to proceed, and so looks to the experimenter for cues. On the next trial, the situation is characterizable from the agent’s perspective as a somewhat sadistic exercise to test her moral capacities; she knows the setup going into it and she can make an informed choice on whether to even take part, as well as what voltage level she ought to stop at. She may also know not to trust the experimenter as an authority anymore.[13]

Unfortunately, many situationists seem to assume that there is only one characterization of a situation, and that is the third-personal, “objective” one (Ross and Nisbett are possible exceptions here, as they do emphasize the role of the agent’s construal of the situation). But a researcher’s construal of the experimental situation need not be congruent with that of the subject’s own construal. Sreenivasan makes a similar point:

It is one of the hallmarks of situationism to privilege objective behavioral measures in the assessment of character traits, at the expense of various forms of subjective assessment favored by traditional theorists of personality — for example, self-reports, peer evaluations, and personality assessment scales. (Sreenivasan 2002:50)

In order, says Sreenivasan, to articulate our understanding of a trait such as compassion, we have to specify what responses are to count as compassion-relevant behavior. But whose specification is to count (ibid. 57-58)? By default, it is the experimenter’s, and as we noted above, this specification can be unduly narrow. Not only is this the case for behavior, but the very situation itself must be specified, and the subject’s specification may differ from the researcher’s. To get over this problem, and to be able to make warranted inferences about the possession of traits of character, or their non-possession, at least three conditions should be met:

1. The experimental situation should be a paradigm one for the expression of the trait (and we must add, for reasons to be made clear below, only that trait).[14]

2. The experimental situation should present the subject with no overriding circumstances or reasons for action that would negate the appropriate response.

3. The characterization of the situation (and the appropriate response) must be the same for both the observer and the subject.

I suggest that the Milgram experiments do not satisfy 3 and, though it can be debated, it is doubtful that they satisfy 1 or 2 either.[15] In any case, they do not warrant inferences to the existence or nonexistence of character traits. Also, note that 3 implies a methodology different from that typically undertaken by researchers. In a case like that of the Milgram experiment, if a subject’s behavior seems counterintuitive, then the first line of explanation must be to see if and how the subject’s understanding of the situation differs from the researcher’s. Instead, situationists tend to take the subject’s understanding as irrelevant, inferring that behavior was simply determined in some way by situational factors, without consideration of the subject’s own agency. The subjective point of view counts for little; what is pertinent is the observer’s perspective and understanding of the proceedings. Furthermore, subjects’ after-the-event explanations of their actions tend to be dismissed as post hoc rationalizations or justifications, instead of being taken for what they indeed could be: straightforward explanations, to the best of their knowledge, of why they did what they did. Of course, this is not to suggest that their accounts should always be taken at face value; after all, people are subject to confabulation and self-deception. But such accounts should not be dismissed tout court either.

The fact is, when we talk about “situations”, we necessarily speak loosely (interestingly, the same might be said of “actions”). There is not simply one intersubjective situation, the same for all. Any specification of a situation must include a description of how that situation is represented by the agent herself. This implies that the agent’s specification of the situation is a function of her own cognitive and endogenous motivational structures. Rather than viewing the situation as external and character as internal, we should think of them as different sides of the same coin. Situationists are only looking at one side, and perhaps virtue theorists commit the similar error of looking only at the other side. Ross and Nisbett make a similar point in their explanation of the fundamental attribution error, that is, the bias we laypeople have towards explaining behavior through reference to personality and dispositions rather than referring to aspects of the situation. They characterize the error as a perceptual Gestalt problem of foreground versus background (or, as they term it, “figure” versus “ground”): the agent is the foreground, and situation recedes to the background (Ross and Nisbett 1991:139-140).

I would add that if this characterization of the fundamental attribution error is more than a metaphor, the opposite error can also be made, and is made, by situationists. They take the situation as foreground, while the agent is background. Both are biases, and neither is a priori the correct view. Think of Wittgenstein’s duck-rabbit figure (Philosophical Investigations II.xi): the situationist says it is a rabbit and claims that those who say it is a duck are mistakenly “duck-biased”; duck-theorists can just as rightfully claim that situationists are subject to “rabbit bias”. Again, they are two sides of the same coin. A complete description of the figure must account for both the duck and the rabbit, and their relation to each other. We might say that situation is a function of external factors as perceived through the lens of the subject, and the subject’s agency is itself only manifested through situations. This was Henry James’ point when he wrote, “What is character but the determination of incident? What is incident but the illustration of character?”[16] We also note, before moving on, that insofar as situationists are correct in claiming we are subject to attribution bias, there is research indicating that such biases are not ironclad laws of human nature, but rather they are modifiable by those subjects who are made aware of them (Beaman et al. 1978). The same goes for knowledge of the situational factors that seem to affect behavior: “There is evidence that as people come to learn about these biases and situational effects that influence behavior, the biases begin to lose their power” (Andrews 2004:248). Certainly this accords both with our intuitions and much personal experience.


What we can take away from the criticisms of situationism in this paper are the following ideas. First, situations have an agent-relative aspect, in that any characterization of a situation must be specified through the lens of some particular agent, whether it be from the first-person perspective of the agent in the situation, or from the third-personal perspective of an observer.

Second, from the first-person perspective, situations are nominal, in that they are non-iterable. This is because the agent is necessarily changed by his initial experience in the situation, however imperceptibly. In this sense, it is a different agent who meets the iterated situation, and since situations are agent-relative, it follows that we must describe what is supposedly an iteration of the same situation as a different situation. Every situation is particular, and thus we must be nominalists with regard to situations.[17] Thus, what the third person observer would call an iteration of the Milgram-situation, the agent who finds himself in it might likely see as a different situation from what came before, even if similar in several respects: “Taking subjective construal into account requires us to note that subjects in the same situation, objectively speaking, are not in the same situation on the situationist understanding, because subjective construals of the situation may vary” (Kamtekar 2004:470).

Third, what all of this implies is that we are wrong to make a hard demarcation between agent and situation: any adequate description of the latter would necessarily include the former as part of it. Strict demarcation between agent and situation leads us to crudely view agents as semi-inanimate objects moved around by external forces. Or it may instead lead us to the opposite error of viewing agents as self-moving objects without reference to situational facts that could make their movements intelligible. This latter error is the root of what is called the Thesis of Motivational Self-Sufficiency of Character, the idea that motivation has its sources with the agent, and that behavioral dispositions are largely independent of external factors, not sustained by, for example, social conditions and relationships (Merritt 2000:374; Swanton 2003:31-32). Epictetus, for example, seems to have had such a thesis in mind when he wrote that “we ought not to look for the motive [for our actions] anywhere outside of ourselves” (Discourses 1.11.28).

An earlier Stoic philosopher, Chrysippus, was implicitly recognized this dichotomy between motivational self-sufficiency and situational causality. In discussing moral responsibility, Chrysippus distinguished between external causes of action (roughly, situationism), and internal causes of action (roughly, character). He used the analogy of a cylinder pushed down an incline: it is the push (external cause) that gets the cylinder moving, but this force would be inefficacious without assistance from the constitution — the shape, weight, etc. — of the cylinder itself (internal cause), which ensures that the cylinder actually rolls (Long and Sedley 1987:387-388; Salles 2005:42-47). What differentiates human agents from cylinders is the greater degree to which internal causes motivate them. But neither kind of causality absolutely excludes the other. Indeed, Chrysippean internal causes are to a significant degree shaped by external causes, meaning that the demarcation between internal and external causes is not as sharp as we might like; a cylinder, after all, becomes a cylinder through external causes, and humans do not bring themselves into existence. But once a cylinder or human being comes to be, any explanation of its changes and activities must be done at least in part through reference to internal causes.

Thus, perhaps neither motivational self-sufficiency nor situationism is an error per se. Perhaps they are simply two sides of the same coin: for some purposes we look to an explanation in terms of external causality (situationism), while for other purposes we find an internal motivational story appropriate (motivational self-sufficiency). In other words, perhaps, Spinoza-like, we view agents under the light either of activity or passivity, of natura naturans or natural naturata, without either one being a priori incorrect.[18] I leave this door open, while recognizing that to pass through it would require a broader metaphysical scheme than I intend to offer herein. It would also require an account of what makes explanations in terms of agential activity or passivity appropriate in given contexts.


[1] Or put in its weaker form: there is no such thing as character as we know it in its folk-psychological form. Situationists have a tendency to state the strong version of the argument and then switch to the weaker one when confronted by critics, claiming that it was the weaker one they actually meant — at least until their next writing. See for example (Harman 1999), where the strong thesis is stated, followed by (Harman 2000), where, in response to (Athanassoulis 2000) he reverts to the weak version, claiming it was what he intended all along. Similarly, (Doris 2002) switches back and forth, depending on where his argument takes him, eventually settling on the weak version (Doris and Stich 2005). As this paper will show, the strong version is implausible. The weak version is less so, but insofar as it tells us behavior is a function of situations and internal motivational structures, it says little we did not already suspect.

[2] We leave it as an open question what degree of similarity globalism requires between situations for an attribution of stability. Is the context of an economics test sufficiently like the context of a physics test? Much depends on the details. Indeed, as will be seen, situationists hold that any seemingly minute detail, including ones not seemingly related to the expression of the relevant trait, may have an effect on behavior that will nullify the attribution of stability. If they are right, then it may be the case that there is, practically speaking, no such thing as the iteration of the same situation. But we are getting ahead of ourselves.

[3] For a classical statement of this thesis, see Plato’s Protagoras. For a deeper examination of the “unity of the virtues” doctrine in classical times, see (Cooper 1998).

[4] Consensus is lacking, in part, because it is not clear what studies of children can tell us about formed adult human character (Doris 2002:63; Webber 2006:652).

[5] Certainly this elitism is to be found in the classics of virtue ethical literature from Plato and Aristotle onwards. And insofar as modern virtue ethicists draw inspiration from these sources, they too often leave themselves open to this objection (if objection it be).

[6] Sabini and Silver express doubt that this experiment is of much value in inferring the existence or non-existence of traits of character: “we just do not believe that picking up or not picking up your papers is a very important manifestation of a moral trait” (Sabini and Silver 2005:539-540).

[7] Few authors pay enough attention to the significance of the subjects’ distress. Exceptions — from both sides of the debate — include Erik Wielenberg and Jonathan Webber (Wielenberg 2006:473; Webber 2006:656).

[8] Remember that the subjects thought the purpose of these experiments was, ostensibly, to study the effects of punishment on learning. Thus, punishment (in the form of electric shocks) was an integral part of the success of the experiments, from the learner’s point of view. Actions tending to undermine the efficacy of punishment can therefore be read as a subversion of the experiment’s (and the authority’s) stated goal.

[9] It is notable that these more complex reactions of subjects are completely missing from Harman’s account of the experiments. He simply reports the final results and draws his inferences. Doris does mention the discomfort of the subjects, but only in order to counter the objection that the latter were aware of the fictionality of the experimental set-up; he does not seem aware that such discomfort casts doubt on the legitimacy of the claim that subjects lacked compassion (Doris 2002:43-44).

[10] On the issues surrounding dispositional accounts of traits of character, see (Brandt 1970).

[11] This is the conclusion that Peter Goldie also comes to, though for different reasons than those given here. See (Goldie 2000:151-175).

[12] As Kristin Andrews has pointed out to me, it may go deeper than this. The agent may view all relevant objective aspects of the situation in the same way, and still view it through the lens of a varying mood.

[13] Situationists tend to dismiss any fuss over the questionable ethics of the Milgram experiment, but I think it is important that the reader be aware of this controversial aspect of the research. Obviously, telling the agents the real purpose of the experiment (to study subjects’ willingness to administer electric shocks to fellow citizens) would have hopelessly polluted whatever data was obtained. On the other hand, armed with this knowledge, how many subjects would have refused to participate? And of the ones who would have participated, how many of them would have disobeyed earlier than they did? If the answer to either of these questions is “a significant number”, then situationists must revise their assertion of a lack of compassion among the general populace. The body of literature on the ethical implications of the Milgram experiments is quite large. For an overview, see (Blass 2004:116-118 et passim).

[14] “How, then, can virtuous character be detected? If an important part of virtuous character is the possession of certain capacities and dispositions, then the best way to test for the presence of character is to place subjects in situations in which these capacities are called for” (Wielenberg 2006:479). Wielenberg makes this point in connection with the preferability of practical over “pencil-and-paper” studies of character; a subject may say she is compassionate, but, as the Milgram experiments show, there is often a gap between belief and practice.

[15] Sceptics may wonder if 3 can ever be satisfied. This is not a question I propose to explore here.

[16] Ironically, this quote appears as the epigraph of the preface to Doris’ book (Doris 2000:ix). It would seem that he used it without comprehending its full import.

[17] This does not mean that two situations cannot overlap in many key respects — or rather, that an agent’s construal of them cannot bear key similarities. However, we must not be surprised if an agent’s behavior does not accord with the predictions we make of it on the basis of situational similarity. Differences between the situations may lie in the agent (and in her representation of the situations) rather than in our characterization of the situation.

[18] “Nature naturing” or “nature natured” (Spinoza, Ethics, Schol. Pr. 29, Pt. I).


Andrews, Kristin. “How to Learn from Our Mistakes: Explanation and Moral Justification,” Philosophical Explorations 7 (2004), 247-263.

Arpaly, Nomy and Timothy Schroeder. “Praise, Blame, and the Whole Self,” Philosophical Studies 93 (1999), 161-188.

Athanassoulis, Nafsika. “A Response to Harman: Virtue Ethics and Character Traits,” Proceedings of the Aristotelian Society 100 (2000), 215-221.

Beaman, A. L., P. J. Barnes, B. Klentz and B. McQuirk. “Increasing Helping Rates through Information Dissemination: Teaching Pays,” Personality and Social Psychology Bulletin 4 (1978), 406-411.

Blass, Thomas. The Man Who Shocked the World: The Life and Legacy of Stanley Milgram. New York: Basic Books, 2004.

Brandt, Richard B. “Traits of Character: A Conceptual Analysis” (1970), reprinted in Morality, Utilitarianism, and Rights. Cambridge: Cambridge University Press, 1992.

Cooper, John M. “The Unity of Virtue,” in Ellen Frankel Paul, Fred D. Miller, Jr. and Jeffrey Paul (eds.), Virtue and Vice. Cambridge: Cambridge University Press, 1998.

Darley, J. M. and C. D. BATSON. “‘From Jerusalem to Jericho’: A Study of Situational and Dispositional Variables in Helping Behavior,” Journal of Personality and Social Psychology 27 (1973), 100-108.

Doris, John M. Lack of Character: Personality and Moral Behavior. Cambridge: Cambridge University Press, 2002.

Doris, John M. and Stephen P. STICH. “As a Matter of Fact: Empirical Perspectives on Ethics,” in Frank Jackson and Michael Smith (eds.), The Oxford Handbook of Contemporary Philosophy. Oxford: Oxford University Press, 2005.

Eisenberg, Nancy and Paul H. Mussen. The Roots of Prosocial Behavior in Children. Cambridge: Cambridge University Press, 1989.

Flanagan, Owen. Varieties of Moral Personality: Ethics and Psychological Realism. Cambridge, MA: Harvard University Press, 1991.

Goldie, Peter. The Emotions: A Philosophical Exploration. Oxford: Clarendon Press, 2000.

Harman, Gilbert. “Moral Philosophy Meets Social Psychology: Virtue Ethics and the Fundamental Attribution Error,” Proceedings of the Aristotelian Society 99 (1999), 315-331.

—— “The Nonexistence of Character Traits,” Proceedings of the Aristotelian Society 100 (2000), 223-225.

Isen, A. M. and P. F. Levin. “Effect of Feeling Good on Helping: Cookies and Kindness,” Journal of Personality and Social Psychology 21 (1972), 384-388.

Kamtekar, Rachana. “Situationism and Virtue Ethics on the Content of Our Character,” Ethics 114 (2004), 458-492.

Kupperman, Joel J. “The Indispensability of Character,” Philosophy 76 (2001), 239-250.

—— “Tradition and Community in the Formation of Character and Self,” in Kwong-Loi Shun and David B. Wong (eds.), Confucian Ethics: A Comparative Study of Self, Autonomy, and Community. Cambridge: Cambridge University Press, 2004.

Long, A. A. and D. N. Sedley (eds.). The Hellenistic Philosophers, Volume One: Translations of the Principal Sources, with Philosophical Commentary. Cambridge: Cambridge University Press, 1987.

Merritt, Maria. “Virtue Ethics and Situationist Personality Psychology,” Ethical Theory and Moral Practice 3 (2000), 365-383.

Milgram, Stanley. Obedience to Authority: An Experimental View. New York: Harper and Row, 1974.

Miller, Christian. “Social Psychology and Virtue Ethics,” Journal of Ethics 7 (2003), 365-392.

Ross, Lee and Richard E. Nisbett. The Person and the Situation: Perspectives of Social Psychology. New York: McGraw-Hill, 1991.

Sabini, John and Maury Silver. “Lack of Character? Situationism Critiqued,” Ethics 115 (2005), 535-562.

Salles, Ricardo. The Stoics on Determinism and Compatibilism. Aldershot, UK: Ashgate, 2005.

Spinoza, Benedictus de. The Collected Works of Spinoza (Vol. I). Edwin Curley (trans.). Princeton: Princeton University Press, 1985.

Sreenivasan, Gopal. “Errors about Errors: Virtue Theory and Trait Attribution,” Mind 111 (2002), 47-68.

Suda, W. and G. Fouts. “Effects of Peer Presence on Helping in Introverted and Extroverted Children,” Child Development 51 (1980), 1272-1275.

Swanton, Christine. Virtue Ethics: A Pluralistic View. Oxford: Oxford University Press, 2003.

Webber, Jonathan. “Character, Consistency, and Classification,” Mind 115 (2006), 651-658.

Wielenberg, Erik J. “Saving Character,” Ethical Theory and Moral Practice 9 (2006), 461-491.

Wittgenstein, Ludwig. Philosophical Investigations. Oxford: Basil Blackwell, 1968.